Professional Documents
Culture Documents
WATSON
PRESS
A TREATISE ON
THE THEORY OF
BESSEL FUNCTIONS
G. N. Watson
The late Professor G. N. Watson wrote his
monumental treatise on the theory of Bessel
functions with two objects in view. The first
was the development of applications of the
fundamental processes of the theory of com-
plex variables, and the second the compila-
tion of a collection of results of value to
mathematicians and physicists who en-
counter Bessel functions in the course of
their researches.
The completeness of the theoretical ac-
count, combined with the wide scope of the
collection of practical examples and the
extensive numerical tables, have resulted in
a book which is indispensable to pure mathe-
maticians, to applied mathematicians, and
to physicists alike.
W. B. F.
A TREATISE ON THE
THEORY OF
BESSEL FUNCTIONS
BY
G. N. WATSON
SECOND EDITION
CAMBRIDGE
AT THE UNIVERSITY PRESS
1966
PUBLISHED BY
THE SYNDICS OF THE CAMBRIDGE UNIVERSITY PRESS
Bentley House, 200 Euston Road, London, N.W. 1
American Branch: 32 East 57th Street, New York, N.Y. 10022
West African Office: P.M.B. 5181, Ibadan, Nigeria
/ ~--fcU^6\\*s
Since the book is, for the most part, a development of the theory of func-
G. N. W.
G. N. W.
March 31, 1941.
CONTENTS
CHAP. PAGE
I. BESSEL FUNCTIONS BEFORE 1826 1
BIBLIOGRAPHY 753
particulars, belongeth to the first author of the story : but to use brevity,
and avoid much labouring of the work, is to be granted to him that will
make an abridgement.
2 Maccabees ii. 30, 31.
CHAPTER I
dy = yydx + xxdx,
and states, more than Thus he writes (Jan. 27,
once, his inability to solve it.
equation of the second order and wrote to Leibniz (Nov. 15, 1702) " Qua :
il The connexion between this equation and a special form of Bessel's equation will be seen
in 4-3.
,
X 3 X 7 X 11
XVa X19
3 3.4.7 3.4.7.8.11 3. 4.7.8. 11 . 12. 15^3. 4. 7.8. 11. 12. 15. 16. 19
V , X* X* X 12 x lti
3.4
T 3.4.7.8.11.12
T
3.4.7.8 3.4.7.8.11.12.15.16
quae series quidem actuali divisione in unam conflari possunt, sed in qua
ratio progressionis non tarn facile patescat, scil.
_a? x1 2x11 13-r15
y ~ + + + +
3 3.3. 7 373. 3. 7. 11 3.3.3.3.5. 7777II
Of course, at that time, mathematicians concentrated their energy, so far
as differential equations were concerned, on obtaining solutions in finite terms,
and consequently James Bernoulli seems to have received hardly the full credit
to which his discovery entitled him. Thus, twenty-two years later, the paperf
in which Count Riccati first referred to an equation of the type which now
bears his name, was followed by a note:,: by Daniel Bernoulli in which it was
stated that the solution of the equation
ax dx + uudx = bdu
11
characteribus occultis involuta 24a, 6b, 6c, Sd, 33e, hf, 2g, 4>h, 33i, 61, 21m,
26n, 16o, Sp, oq, I7r, 16s, 25t, 32m, 5#, By, +, -, , , =, 4, 2, 1."
The anagram appears never to have been solved ; but Bernoulli published
his solution ||
of the problem about a year after the publication of the anagram.
The solution consists of the determination of a set of values of n, namely
4m/(27/i 1), where m is any integer, for any one of which the equation is
soluble in finite terms; the details of this solution will be given in 4*1, 4*11.
* See Leibnizens gesamellte Werke, Dritte Folge (Mathematik), m. (Halle, 1855), p. 75.
t Acta Eruditorum, Suppl. vm. (1724), pp. 6673. The form in which Riccati took the
equation was
xmdq = du + uu dx:q,
where q = xn .
% Ibid. pp. 73
75. Daniel Bernoulli mentioned that solutions had been obtained by three
other
members of his family John, Nicholas and the younger Nicholas.
The reader should observe that the substitution
u= bdzz dx
gives rise toan equation which is easily soluble in series.
Exercitationes quaedam matheviaticae (Venice, 1724), pp. 77
|j
80; Acta Eruditorum, 1725,
pp. 465473.
1*2] BESSEL FUNCTIONS BEFORE 1826 3
John Bernoulli's equation* has resulted in the name of Riccati being associated
not only with the equation which he discussed without solving, but also with
a still more general type of equation.
It is now customary to give the namef Riccati's generalised equation to
any equation of the form
n
+ 4ww +
4 9 16n4 4.9.16. 25ns
'
4 9n
. . .
See James Bernoulli, Opera Omnia, n. (Geneva, 1744), pp. 1054 1057 it is stated that the
* ;
point of Riccati's problem is the determination of a solution in finite terms, and a solution which
X Iiutitutiones Calculi Integralis, n. (Petersburg, 1769), 831, pp. 8889. In connexion with
the reduction, see James Bernoulli's letter to Leibniz already quoted.
Novi Coram. Acad. Petrop. vni. (17601761), [published 1763], p. 32.
|| 163164.
Ibid. ix. (17621763), [published 1764], pp.
If " Theoremata de osoillationibus corporum filo flexili connexorum et catenae verticaliter
suspensae," Comm. Acad. Set. Imp. Petrop. vi. (1732 3), [published 1738], pp. 108 122.
** Loc. cit: p. 116.
ft The length of the simple equivalent pendulum is n.
4 THEORY OF BESSEL FUNCTIONS [CHAP. I
The last series is now described as a Bessel function * of order zero and
argument 2 V(#/n); and the last quotation states that this function has an
infinite number of zeros.
Let p be the line density of the chain (supposed uniform) and let T be the tension at
height x above the lowest point of the chain in its undisturbed position. The motion being
transversal, we obtain the equation bT=gpbx by resolving vertically for an element of
chain of length bx. The integral of the equation is Tgpx.
The horizontal component of the tension is, effectively, T{dyjdx) where y is the (hori-
zontal) displacement of the element ; and so the equation of motion is
'*-( r 2)-
If we substitute for T and proceed to the limit, we find that
dt*
y dx\ dxj '
If / is the length of the simple equivalent pendulum for any one normal vibration, we
write
where A and f are constants ; and then n (x/f) is a solution of the equation
A d
( -\ +- =
dx\ dx) f
If x/f= ii, we obtain the solution in the form of Bernoulli's series, namely
u
v= l H
ul
1 1.4
+ u*
v?
1.4.9 1.4.9.16
* On the Continent, the functions are usually called cylinder functions, or, occasionally, func-
tions of Fourier -Bessel, after Heine, Journal fur Math. lxix. (1868), p. 128; see also Math. Ann.
m. (1871), pp. 609610.
t Comm. Acad. Petrop. vn. (17345), [published 1740], pp. 162179.
X Acta Acad. Petrop. v. pars 1 (Mathematics), (1781), [published 1784], pp. 157177. Euler
took the weight of length e of the chain to be E, and he denned g to be the measure of the
distance (not twice the distance) fallen by a particle from rest under gravity in a second. Euler's
notation has been followed in the text apart from the significance of g and the introduction of
p and 5 (for d).
~
. , where C and
D are constants. Since y is finite when #=0, C must be zero.
If a is the whole length of the chain, y =0 when . = a, and so the equation to determine/* is
a% a3
j. _
1
" + TT4/2 ~ 4. 9/3
""' a
1 ./ 1 .
By an extremely ingenious analysis, which will be given fully in Chapter xv, Euler
proceeded to shew that the three smallest roots of the equation in a/f are 1-445795, 7*6658
and 18-63. [More accurate values are 1*4457965, 7*6178156 and 18*7217517.]
In 'the memoir'"' immediately following this investigation Euler obtained the general
solution (in the form of series) of the equation -j- (u-r-\+v0, but his statement of the
law of formation of successive coefficients is rather incomplete. The law of formation had,
however, been stated in his Institutiones Calculi Integrality, n. (Petersburg, 1769), 977,
pp. 233-235.
The solution of this equation which is finite at the origin is given on p. 256
of Euler's memoir; it is
w = ?-P \l + 2.4(w+l)(n
{ 2(w+l)e 2
+ 3)e4 ~) ,
Save for an omitted constant factor the series is now called a Bessel
coefficient of order /S and argument arje. The periods of vibration, 2irja, of a
* Acta Acad. Petrqp. v. pars 1 (Mathematics), (1781), [published 1784], pp. 178190.
t See also 935, 936 (p. 187 et seq.) for the solution of an associated equation which will be
discussed in 3*52.
X Novi Comm. Acad. Petrop. x. (1764), [published 1766], pp. 243260.
The reason why Euler made this change of notation is not obvious.
|| If /3 were not an integer, the displacement would not be a one-valued function of position,
in view of the factor sin {p</> + B).
6 THEORY OF BESSEL FUNCTIONS [CHAP. I
E=M+ n=l
Z A n sinnM, -=l + he + 2 B H cosnM,
&
2
n=\
in which a and e are the semi-major axis and the eccentricity of the orbit, and
_9 (-)" M -n-i 6 n+m
ti (_)m H + m n-t-m-i n+2m
.
n ~ "m=o n ~~
^ ^
-
(
l+2 t
) , ro e
2+"*"ro (n !
+ w)'! '
mt 2' w!(w + m)!
Lagrange gave these expressions for n = 1, 2, 3. The object of the expansions
is to obtain expressions for the eccentric anomaly and the radius vector in
terms of the time.
In modern notation these formulae are written
j?
f dA n
n at
a memoir by Lefort, Journal de Math. xi. (1846), pp. 142 152, in which an error made by
Poisson is corrected, should also be consulted.
* Cf. Bourget, Ann. Sci. de VEcole norm. sup. in. (1866), pp. 5595, and Chree, Quarterly
Journal, xxi. (1886), p. 298.
t Hist, de VAcad. R. des Sci. de Berlin, xxv. (1769), [published 17711, pp. 204233. [Oeuvres,
in. (1869), pp. 113138.]
X Ricerche sulla convergenza delta serie che serva alia soluzione del problema di Keplero
(Milan, 1817). This work was translated into German by Jacobi, Astr. Nach. xxx. (1850),
col.197254 [Werke, yii. (1891), pp. 189245]. See also two papers by Scheibner dated 1856,
reprinted in Math. Ann. xvn. (1880), pp. 531544, 545560.
14] BESSEL FUNCTIONS BEFORE 1826 7
c
2
^ + <^-"- 2 (l-e 2 )J B ==0.
Define n by the formula A n =2nn - i
e
/udt
/n ! and then
2
(^+2 ) + tt- 2
(l-2 ) = 0.
If u (n a ) we should expect
and du/dt to l>e (n 2a ) and (n a ) respectively and
u- ;
on considering the highest powers of n in the various terms of the last differential equation,
we find that a= 1. It is consequently assumed that u admits of an expansion in descending
powers of n in the form
On substituting this series in the differential equation of the first order and equating to
/-ero the coefficients of the various powers of n, we find that
o
2 = (l- f 2 )/*2 , + 2oi)+Mo=0, ...
judt-n^ogj^f2)vA(1-e 2
) + l}-ilog(l- f 2 ) + ...,
f
n exp {nj{\- t2)J
B
v/(^ M *(i-<2 )*{i+N/(i-2 )} n
'
and this is the result obtained by Carlini. This method of approximation has been carried
much further by Meissel (see 8*11), while Cauchy* has also discussed approximate
formulae for An in the case of comets moving in nearly parabolic orbits (see 8"42), for
The investigation of which an account has just been given is much more
plausible than the arguments employed by Laplace + to establish the corre-
sponding approximation for Bn .
The investigation given by Laplace is quite rigorous and the method which
he uses is of considerable importance when the value of Bn ismodified by
taking all the coefficients in the series to be positive or, alternatively, by
supposing that e is a pure imaginary. But Laplace goes on to argue that an
approximation established in the case of purely imaginary variables may be
used sans crainte in the case of real variables. To anyone who is acquainted
'
'
with the modem theory of asymptotic series, the fallacious character of such
reasoning will be evident.
For other and more recent applications of this principle, see Stokes, Proc. Camb. Phil.
362366 [Math, and Phys. Papers, v. (1905), pp. 221225], and Hardy,
Soc. vi. (1889), pp.
Proc. London Math. Soc. (2) n. (1905), pp. 332339 Messenger, xxxiv. (1905)., pp. 97101. ;
A statement of the principle was given by Borel, Acta Mathematica, xx. (1897), pp. 393
394.
,=o
in which n is large and t. has a fixed positive value. The greatest term is that for which
mfi, where /u is the greatest integer such that
Now, if u m denotes the general term in BJ~ l \ it is easy to verify by Stirling's theorem
~2ttMX/{7r/(l-?)},
since* q is nearly equal to 1.
and so BJ l)r f
2 v'(l + c2 ))^*
) e exp { WU
+ 2 )}
\ nn3 J {1 + ^(1 + ,')} *
This approximate formula happens to be valid when e < 1 (though the reason
for this restriction is not apparent, apart from the fact that it is obviously
necessary), but it is difficult to prove it without using the methods of contour
2
The formula 1 + 2 2<f' ~ N/W(l-g)} maT be inferred from general theorems on series;
cf. Bromwieh, Theory of Infinite Series, 51. It is also a consequence of Jacobi's transformation
formula in the theory of elliptic functions,
*3 (0 \r) = (-*)-* M0 -T- I
1
);
see Modern Analysis, 21*51.
;
integration (cf. 8*31). Laplace seems to have been dubious as to the validity
of his inference because, immediately after his statement about real and
imaginary variables, he mentioned, by way of confirmation, that he had
another proof; but the latter proof does not appear to be extant.
~ + = 1 |
f*
cos (a sin x) dx,
22 2 2 42
. 2 2 42 6 2
. .
had been proved some years earlier by Parseval; it is a special case of what
are now known as Bessel's and Poisson's integrals 22, 23). (
* The memoir deposited in the archives
greater part of Fourier's researches was contained in a
of the French Institute on Sept. 28, 1811, and crowned on Jan. 6, 1812. This memoir is to be
found in the Mim. de VAcad. des Sci., iv. (1819), [published 1824], pp. 185555; v. (1820),
[published 1826], pp. 153246.
t This is a generalisation of Bernoulli's statement quoted in 1*2.
t Math. Ann. xxxin. (1889), pp. 246266.
M4m. des savant itrangers, i. (1805), pp. 639
648. This paper also contains the formal
statement of the theorem on Fourier constants which is sometimes called Parseval's theorem
another paper by this little known writer, Mim. des savans Strangers, i. (1805), pp. 379 398, con-
tains a general solution of Laplace's equation in a form involving arbitrary functions.
10 THEORY OF BESSEL FUNCTIONS [CHAP. I
(1909), pp. 359388; and Young, Proc. London Math. Soc. (2) xvni. (1920), pp. 163200.
This expansion will be dealt with in Chapter xvni.
V .'o
cos (h\ cos co) sin2n G>cifo>,
where w= 0, 1, 2, ... and \ is the distance from the axis of the cylinder. The
integral is now known as Poisson's integral ( 2'3).
*&ffl + 1+ i),_a
(
* Journal de I'ficole JR. Polytechnique, xn. (cahier 19), (1823), pp.
249403.
+ Ibid. 300 etseq. The equation was also studied by Plana, Mem. della R. Accad. delle Sei.
p.
di Torino, xxv. (1821), pp. 532 534, and has since been studied by numerous writers, some of
whom are mentioned in 4*3. See also Poisson, La TMorie Mathimatique de la Chaleur (Paris,
1835), pp. 366, 369.
t See also Rohrs, Proc. London Math. Soc. v. (1874), pp. 136 187. The notation J (k) was
not used by Poisson.
;
When k is large, l/(4 2 ) may be neglected in comparison with unity and so we may expect
that Jo{k)sJk is approximately of the form A cos k + Bsm k where A and B are constants.
To determine A and B observe that
cos k . J (k) - sin k . /' (k) - \ cos 2 Ao cos (2k sin 2 i) o?<u
= "^'
^j I
1
-2-)
S
V(2 * }
and similarly sin ../(*) + cos . J '
() = ^^ /"
fl -^Y" nrnx^dx.
VW (l-^V.^
But lim
*-^oo yf o \ 2kJ sin f^V.^K/^),
J sin
by a well known formula*.
[Note. It is not easy to prove rigorously that the passage to the limit is permissible
the simplest procedure is to appeal to Bromwich's integral form of Tannery's theorem,
Bromwich, Theory of Infinite Series, 174.]
It follows that
( 1
cos k . J (k) -sink. J '
(k) = j^ ( 1 4- t ),
J ^ = K1 + ffc C S
*+^+ ^ sin ^'
J(*k) ^
where A =.5=1. The series are, however, not convergent but asymptotic, and the validity
of this expansion was not established, until nearly forty years later, when it was investi-
gated by Lipschitz, Journal fur Math. lvi. (1859), pp. 189 196.
The result of formally operating on the expansion assumed by Poisson for the function
eft i
J (k) *J(irk) with the operator -jt2 + 1 + rn is
+ 8m *[_ p + Ji
+
+- J'
* Cf. Watson, Complex Integration and Cauchy's Theorem (Camb. Math. Tracts, no. 15, 1914),
p. 71, for a proof of these results by using contour integrals.
:
A---H,
A'- 1
AA" -"
2 g2
A
*, A -
A'"*=
9 25
'
2.3.83
R'
'
/""
/V\*IY 1 - 1 9 9-25 \ ,
OM(*ooe.)A,-^J 8I--2T8^
+ 2V3."83l3+-" C08 *
j o LV >
)
/. 1 9 9.25 , \ . ,1
1 + 8l"2T8^~2.3.88ife + -; 8m
,
V *J-
But, since the series on the right are not convergent, the researches of Lipschitz and
subsequent writers are a necessary preliminary to the investigation of the significance of
the latter portion of Poisson's investigation.
It should be mentioned that an explicit formula for the general term in the expansion
was first given by W. B. Hamilton, Trans. B. Irish Acad. xix. (1843), p. 313; his result
was expressed thus
- "["
cos (20 sin a) da = -^ 1 [0] ([ - *]) 2 (43) cos (20 - $n,r - }n),
and he described the expansion as semi-convergent; the expressions -B and [-$]" are
[0]
to be interpreted as 1/n and (-|)( f) ... (-+)
!
~
y = Ax% + Bx$ e hxcoaa log (x sin
2
is
f e- hxcoSu> da> \
m) da>,
Jo Jo
dxr x dx
-
is y = A\ e~ hxcoau da) +B I e hxcoa '\og (a? sin2 w) cZo>.
.'o Jo
general integral of an associated partial differential equation was given in an earlier memoir,
ibid, p. 227.
t Camb. Phil. Trans, ix. (1856), p. [38], [Math, and Phys. Papers, hi. (1901), p. 42].
- = 1 + ^
2
+ 2 Bn cosnM,
where Bn = / sin u sin (nu ne sin u) du ;
this expression for Bn should be compared with the series given in 1'4.
In the memoir of 1824 Bessel investigated systematically the function I h
k
defined by the integral J
Ikh
If
= k I
2"
cos (hu k sin u)' du.
zvrJo
He took h to be an integer and obtained many of the results which will be
given in detail in Chapter n. not adapted for defining the
Bessel's integral is
function which most worth study when h is not an integer (see 10-1) the
is ;
function which is of most interest for non-integral values of h is not Ikh but
the function defined by Lommel which will be studied in Chapter in.
After the time of Bessel investigations on the functions became so numerous
that it seems convenient at this stage to abandon the chronological account
An account of researches from the time of Fourier to 1858 has been compiled
historical
by Wagner, Bern MittheUungen, 1894, pp. 204266 a briefer account of the early history
;
was given by Maggi, Atti delta It. Accad. dei Lincei, (Tra/isunti), (3) iv. (1880), pp. 259263.
a formula given by Poisson, Connaissance des Terns, 1825 [published 1822], p. 383.
;
CHAPTEE II
nition,and proved incidentally that the functions thus defined are equal to the
definite integrals by which they had previously been defined by Bessel f. It
should, however, be mentioned that the converse theorem that Bessel's inte-
grals are equal to the coefficients in the expansion, was discovered by Hansen:}:
fourteen years before the publication of Schlomilch's memoir. Some similar
results had been published in 1836 by Jacobi ( 222).
It will be shewn that this function can be developed into a Laurent series,
n
qua function of t; the coefficient of t in the expansion is called the Bessel
is denoted by the symbol Jn (z),
coefficient of argument z and order n, and it
so that
*('"?)- I t"Jn (z).
(1)
=-oo
zt
To establish this development, observe that e^ can be expanded into an
absolutely convergent series of and for all values of t,
ascending powers of t ;
and t, t = excepted.
form (1), which is valid for all values of z
* ZeiUchriftfiir
Math, und Phys. n. (1857), pp. 137 165. For a somewhat similar expansion,
SCO80 , see Frullani,
Mem. Soc. Ital. (Modena), xvm. (1820), p. 503. It must be
namely that of e
pointed out that Schlomilch, following Hansen, denoted by Jftn what we now write as Jn {2z)
but the definition given in the text is now universally adopted. Traces of Hansen's notation
are to be found elsewhere, e.g. Schlafli, Math. Ann. m. (1871), p. 148.
22.
t Berliner Abh. 1824 [published 1826], p.
+ Ermittelung der Absoluten Storungen in Ellipsen von beliebiger Excentricitdt und Neigung,
i. der Sternwarte Seeburg : Gotha, 1843], p. 106. See also the French transla-
theil, [Schriften
n -x
= 5 i-tyj_ n { Z ),
n x
eiz(t-llt) =J + J n
+{_ )n t - n]
(3) q {z) {t Jn {2)
w-1
The function of order unity has been encountered by Turriere, Nouv. Ann. de
Math. (4)
ix. (1909), pp. 433441, in connexion with the steepest curves on the surface z=y (5.r2 -y).
m=o(n-rm)l ml
We therefore have the result
(1) Jn(z)= S l ^ K * Z)
where n is a positive integer or zero. The first few terms of the series are
given by the formula
zn ( z* * 1
Jn () = +
(2) 2^1 { 1 ~ 2 2
. 1 . (n + 1 2- . 1 . 2 . ( + 1) (n + 2)
'"]'
In particular
(3) Jo (?) 1 22 ^ 22 42 . 22 . 42 6 2.
"
To obtain the Bessel coefficients of negative order, we select the terms in-
volving irn in the product of the series representing exp tyzt) and exp(- \z\t\
where n is still a positive integer. The term of the second series which, when
n
associated with the general term of the first series gives rise to a term in t~
is the term for which ra = n + r ; and so we have
J -n{Z) ~ '
be observed that, in the series (1), the ratio of the (w + l)th term
It is to
to the rath term is - s /{ra (n + ra)}, and this tends to zero as ra
2 oo for all ,
the series representing Jn (z) is convergent for all values of z and n, and so it
From (1) we can obtain two useful inequalities, which are of some import-
ance (cf. Chapter xvi) in the discussion of series whose general term is a
multiple of a Bessel coefficient.
Whether z be real or complex, we have
ao i l z i2W
(4) !
J. < & exp (|!^) <
I
I
^ exp (J | . |>).
xm.
This result was given in substance by Cauchy, Comptes Rendus, (1841), pp. 687,
.l< -^P(l*l ,
r
l
lJ l.
By considering all the terms of the series for Jn (z) except the first, it is
found that
() j.()-^(i+),
.
n + 1
It should be observed that the series on the right in 2*1 (1) converges uniformly in
any, bounded domain of the variables z and t which does not contain the origin in the
*-plane. For if 8, A and R are positive constants and if
d<|t|<A, |*| < | A.I,
the terms in the expansion of exp (\zt) exp ($z/t) do not exceed in absolute value the corre-
sponding terms of the product exp ($RA) exp QR/8), and the uniformity of the convergence
follows from the test of Weierstrass. Similar considerations apply to the series obtained
by term-by -term differentiations of the expansion 2^*t/n (z), whether the differentiations be
performed with respect to z or t or both z and t.
$z(l + l/t*)e
iz(t ~ m= 2 nt^Jniz),
so that
n Jn (z)= 2
^(1 + 1/e ) 5 2
t nt^Jniz).
= =
If the expression on the left is arranged in powers of t and coefficients of n_1
tf
are equated in the two Laurent series, which are identically equal, it is evident
that
\z {Jn-i (z) + Jr* i
(z)) =nJn (z),
which is the first of the formulae f.
* Throughout the work primes are used to denote the derivate of a function with respect to
its argument.
f Differentiations are permissible because ( 2*11) the resulting series are uniformly convergent.
The equating of coefficients is permissible because Laurent expansions are unique.
18 THEORY OF BESSEL FUNCTIONS [CHAP. II
^(t-l/t)e
iz<t - m= t
n Jn '(z),
= -oo
(g)
2r
^T-)= 2 {-TrCm .Jn -r+n {z\
where r Cm is a binomial coefficient.
=i
= 4 1 (-y-*nJm (z),
since zJ^+i (z) -* as iV-*- 00 , by 211 (4).
2-13, 2'2] THE BESSEL COEFFICIENTS 19
=4 I - nJ2n (z),
(11) *./,(*) {-)n l
n-l
is useful in the developments of Neumann's theory of Bessel functions ( 3'o7).
d
dz
f-^Wni*)} = -z*- n Jn (z),
that is to say
(1) zi
W) +z
dJAz)
+ (z2 _ wl) Jh {z) ^
The analysis is simplified by using the operator ^ defined as z (dldz).
"'
(1) Jn (z) = J- I cos {n$ - z sin B) d6.
Lit 'o
* Berliner Abh. 1824 [published 1826], p. 34 ; see also Frullani, Mem. Soc. ItaL.(Modena), xvm.
(1820), p. 504.
t Ibid. pp. 22 and 35.
;
(3) J n (z) = 2~ I
cos(n0-zsm0)d0,
2771 J =- Z7rt J
The integrals on the right all vanish except the one for which m = n; and
so we obtain the formula
Jn (z) = -L f {
e*(*-z n ) + e - Une - z sin 9) Q J
and equation (2), from which (1) may be deduced, is now obvious.
Jn (z) =
1
IT
f
I
Jo
cos nd cos (z sin 0)d0-i If...
n0
TT J
I sin sin (z sin 0) d0.
* Term -by- term integration is permitted because the expansion uniformly convergent on the
is
convenient to use the symbol (a+) to denote integration round a contour encircling
contour. It is J
the point a once counterclockwise.
t Ermiitelung der absoluten Storungen (Gotha, 1843), p. 105.
" .
'**
1
Jn {z)- I cos n$ cos (z sin 6) dd
K) ("even).
2 ,^ f
= - I cos w cos (z sin &) d0
ir Jo
If be replaced by tt - 77 in the latter parts of (6) and (7), it is found that
( 8) /n (*) -- (-)*< 1}
I cos n?7 sin (s cos 17) <fr; (w odd),
(9) /ii (z) =-(-)** cos nr) cos (z cos 17) cfy (n even).
The reader wijl find it interesting to obtain (after Bessel) the formulae 2-12 (1) and
2-12 (4) from Bessel's integral.
Twoformulae involving definite integrals which are closely connected with PaTseval's
integral formula are worth notice. The first, namely
1 />
J
e v cos 0+vs sin dfy
J ^
(y cos 6 + iz sin 6f +
dd = 0,
f" (y cos 6 + iz sin 6f dd =
2r ( w + *)r(i)
(y2 _ ^
the formula then follows without difficulty.
* Journal fUr Math. xv. (1836), pp. 1213. [G<?s. Math. Werke, vi. (1891), pp. 100102] ; the
integrals actually given by Jacobi had limits and w with factors I/*- replacing the factors 2/*-.
See also Anger, Neueste Schriften der Naturf. Ges. in Danzig, v. (1855), p. 1, and Cauchy,
Comptes Rendus, xxxvm. (1854), pp. 910 913.
t Berliner Abh., 1824 [published 1826], p. 37. See also Anger, Neueste Schriften der Naturf.
Ges. in Danzig, v. (1855), p. 10, and Lommel, ZeiUchrift fUr Math, und Phys. xv_(1870), p. 151.
t Bulletin de Vdcad. R. de Belgique, (2) xli. (1876), p. 938.
)
no that
p^ + ! )^ = l
(0 + N
= J_. /"
)
ex ;
[* exp {*+*-} <W,
by taking the contour to be a unit circle ; the result then follows by bisecting the range of
integration.
Two 'series, which are closely connected with Bessel's integral, were dis-
covered by Jacobi*. The simplest method of obtaining them is to write
t = e
ie
in the fundamental expansion 21 (3). We thus get
M=l
On adding and subtracting the two results which are combined in this formula,
we find
=o
The results (3) and (4) were given by Jacobi, while the others were obtained later by
Anger t. Jacobi's procedure was to expand cos (s cos 17) and sin (z cos q) into a series of
cosines of multiples of t}, and use Fourier's rule to obtain the coefficients in the form of
integrals which are seen to be associated with Bessel's integrals.
In view of the fact that the first terms in (1) and (3) are not formed
according to the same law as the other terms, it is convenient to introduce
Neumann's factor* en> which is defined to be equal to 2 when n is not zero,
and to be equal to 1 when n is zero. The employment of this factor, which
* Journal fUr Math. xv. (1836), p. 12. [Ges. Math. U'erke, vi. (1891), p. 101.]
will be of frequent occurrence in the sequel, enables us to write (1) and (2) in
the compact forms:
oo
=o
ao
0) 1=2 em Jm (z).
If we differentiate (5) and (6) any number of times before putting 6 = 0, we
obtain expressions for various polynomials as series of Bessel coefficients. We
shall, however, use a slightly different method subsequently
( 27) to prove
that zm is of. Bessel coefficients when m is any positive
expansible into a series
integer. any polynomial is thus expansible. This is a
It is then obvious that
special case of an expansion theorem, due to Neumann, which will be investi-
gated in Chapter xvi.
For the present, we will merely notice that, if (6) be differentiated once
before 6 is put equal to 0, there results
Note. The expression exp {$z{t- 1/0} introduced in $ 2 1 is not a generating function
in the strict sense. The generating function t associated with en Jn (?) is 2 **"/ (z).
If this expression be called S, by using the recurrence formula 2-12 (2), we have
A result equivalent to this was given by Brenke, Bull. American Math. Soc. xvi. (1910),
pp. 225230.
I fir I oo / \m sm fir
-
7tJ
cos (* cos 0)8^0(20 = - t
7T m =o
V(2W)! Jo
cos2'0sin=0<W
= '
"
R=rn+1 I cos (rp cos ) sin 2
"
1 1
udu,
Jo
then R satisfies the differential equation
dr1 r-
Nielsen, Handbuch der Tfieorie der Cylinderfunktionen (Leipzig, 1904), p. 51, calls them
Bessel's second integral, but the above nomenclature seems preferable.
|| The series to be integrated is obviously uniformly convergent ; the procedure adopted is due
to Poisson, ibid. pp. 314, 340.
2-3, 2*31] THE BESSEL COEFFICIENTS 25
*
(' e^* sin8'* Odd = f cos (z cos 0) sin 0d0\
Jo Jo
this is evident when we consider the arithmetic mean of the integral on the
left and the integral derived from it by replacing by it 0.
We thus get
(3) /(*) =
r(n
4|^V, (1 _ PY - iM>
has suggested important developments (cf. 6*1) in the theory of Bessel
functions.
! '
(4) ' cos (z cos 0) sin2" 0d0 = 2 \ cos {z cos 0) sin1" 0d0
Jo Jo
f*
=2
Jo
cosCssin^cos*1 ^,
and each of these expressions gives rise to a modified form of Poisson's integral.
An interesting application of Bessel's and Poisson's integrals was obtained
by Lommelf who multiplied the formula
, / ,m
* 4w8 {4n8 -28 }...{4n2 -(2ro-2)2} . .
cos2n0= 2 (-) s
/a v.
J
-sin*"^
-o (zm)\
by cos (z cos 0) and integrated. It thus follows that
The proof, that Jn (z) is equal to Poisson's integral, which was given by
Bessel J, is somewhat elaborate; it is substantially as follows
-^ cos
I sin2"
-1
cos (z cos 0) 9 ,- sin2*1
" "
1 1
sin (z cos 0)
**
= |
(2w - 1) sin8"-8 - 2n sin " + 8
sin2"*8 cos (* cos 0),
* Poisson actually made the statement (p. 293) concerning the integral which contains
sinSH+1 6 ; but, as he points out on p. 340, odd powers may be replaced by even powers throughout
his aualysis.
^2 r*
+ 2n + i J cos (* cos 0) sin2"-*" 8 0d0 = 0.
If now we write
rin + ^ra) /^
008 <* cos 0) sin2n ^ s * <>
the last formula shews that
<f>(0)
=J (z),
when w = 0, 1, 2, 3, ....
si^ = ^
.
} sinn *>
where fi = cos 0. We shall assume this formula for the moment, and, since no
simple direct proof of it seems to have been previously published, we shall
give an account of various proofs in 2*321 2323.
we observe that the
If first n /i2 ) M-i with respect to
1 derivates of (1 ,
= (-) n
J_
COS foa - \ mr) -^ d. }
i di
* Journal fiir Math. xv. (1830), pp. 1213. [Ges. Math. Werke, vi. (1801), pp. 101102.] See
also Journal de Math. i. (1836), pp. 195196.
2*32, 2*321] THE BESSEL COEFFICIENTS 27
{Zfi
,
\n-jr) dfi
=1 . 3 5
. . . . (2n - 1) cos (z cos - ^nir) cos n0d0
Jo
= 1.3.5...(2n-l)wJ n (5),
by Jacobi's modification of 2'2 (8) and (9), since cos^cos 6 \nir) is equal
to ( )*** cos (z cos 6) or (-)^ n-1) sin(^cos 9) according as n is even or odd; and
this establishes the transformation.
as a special case of a formula due to Lacroix*; but the proof which Lacroix gave of
his formula is open to objection in that it involves the use of infinite series to obtain
a result of an elementary character. A proof, based on the theory of linear differential
equations, was discovered by Liouville, Journal de Math. vi. (1841), pp. 69 73; this
proof will be given in 2*322. Two years after Liouville, an interesting symbolic proof
was published by Boole, Camb. Math. Journal, in. (1843), pp. 216224. An elementary
proof by induction was given by Grunert, Archio der Math, und Phys. iv. (1844), pp. 104
109. This proof consists in shewing that, if
<p- i(W) -
dQa /"(*
then n+1 = (l- M )-
2 --2/i9B -M(H-l) B rfM ,
5 J
and that ( - n~ 1.3.5.;. (2 the same recurrence formula.
) ' 1) (sin nff)jn satisfies
Other proofs of this character have been given by Todhunter, Differential Calculus
(London 1871), Ch. xxviil, and Crawford t, Proc. Edinburgh Math. Soc. xx. (1902),
pp. 11
15, but all these proofs involve complicated algebra.
A proofdepending on the use of contour integration is due to Schlafli, Ann. di Mat. (2)
v. (1873), pp. 201 202.
The contour integrals are of the type used in establishing
Lagrange's expansion; and in 2*323 we shall give the modification of Schlafli's proof,
in which the use of contour integrals is replaced by a use of Lagrange's expansion.
"
=0 a
a"
\ m ~t"2J
M
= 2 (-)".C'!m + i(ainitf)* + (oo8itf)"-*- 1
m=0
=sin(2wx0),
and this is the transformation required J.
(l-V)Z)y+(2n-l) My=0.
Differentiate this equation n times ; and then
(1 - fi2 ) Dtt+l i/
- (iD^+^D^-^^O :
but (l"^-^-^^^-"^.
so that (^ + J
Z)- y = 0.
de - * (l^-*)*-(l^*)
and so Jfl J)
1.3.5...(2n-l) Sl
f(n) (C S *} 8inm * dx >
1
f
w
XM , J. 2n 2n(n-l) A
1 ,
according as n is even or odd, and we deduce from 2*2 (8) and (9) that
It has been stated that the expansion is valid when /(cos x) is a poly-
nomial in cos a;; it can, however, be established when /(cos x) is merely re-
stricted to be an integral function of cos x, say
* 6w cos n a?
The simplest way of proving this result is from the formula 2*2 (4), which
gives
1 0+)
f<
M_1 * {y+* ~ vt
Jn {y + *) = 5^-. '~
{t
' dt
= ~I po+) oo
X t
m -n- Jm (y)e*z -
1 <t 1/t
>dt
i oo no+)
= _i-. X Jm (y)\ p--i eft-Milt
00
- 2 Jm (y)Jn-m{z),
wi -oo
on changing the order of summation and integration in the third line of the
analysis and this is the result to be established.
;
.-!/,
.j I- rJr (z))\ I-oo tm Jm (z)\.
[ r= oo ) Km- J
J(2s) = Jr {z)Jn-*(z).
In particular, taking ?i = 0, we have
(1) Jo (2s) = Jo* (:) + 2S (- )' Jr2 (*) - (-)' e r Jr* (z).
r=l r=0
f Studienilber die BesseVschen Functionen (Leipzig, 1868), pp. 26 27 ; see also Schlafli, Math.
Ann. in. (1871), pp. 135137.
+ Ermittelung der absoluten Stiirungen (Gotha, 1843), p. 107 et seq. Hansen did not give (4),
and be gave only the special case of (2) in which n = l. The more general formulae are dne to
Loramel, Studien Uber die BesseVschen Functionen (Leipzig, 1868), p. 33.
For brevity, JH2 (z) is written in place of {</ (z)) 2 .
2'4-2'6] THE BESSEL COEFFICIENTS 31
(2) Jn (2z) -S/f (#) Jn-r (*) + 2 (-Y Jt (*) Jn+r (z),
r=0 r=l
when the Bessel coefficients of negative order are removed by using 2*1 (2).
Similarly, since
it follows that
and so
1 f* f*
Jn2 (z) = |-5 j
I *>+*> e -(nfl+sin +, dOdj).
To reduce this double integral to a single integral take new variables defined
by the equations
8-<f> = 2X ,
0+<f> = 2yJr,
so that
d(0,<f>)
= 2.
9 (x Vr )
It follows that
1 C"
=- \
J** (2* cos x) <%
if Jo
=
Jr? (s)
^f Jo (2* sin yjr) j***
<ty,
so that
i r*
/o(2*sin^)cos2n^(fy
(3) ^""^"Srl
1 f*
a - J (2z sin i/r) cos 2nyjr dijr,
if h
a result which SchlafliJ attributed to Neumann.
a
2'61. Neumann's series for J (z).
~ TO = m\{tn+m)\{(n + m)l\ %
'
t Einleitung in die Theorie der BesteVtchen Funktionen, n. (Bern, 1900), pp. 81 85.
of Neumann's treatise.
J The formula is an immediate consequence of equation 16 on p. 69
Math. Ann. in. (1871), p. 603. The memoir, in which 4his result
was given, was first pub-
lished in the Leipziger Berichte, mi. (1869), pp. 221256.
,
2m-
(2n+l)(2n + 3)
(2)
4
~(2n + 2)(2w + 4)'
We shall now obtain the result which was foreshadowed in 2*22 con-
cerning the expansibility of zm in a series of Bessel coefficients, where is any m
positive integer. The result for m
= has already been given in 2*22 (7).
In the results 2*22(1) and (2) substitute for cos 2nd and sin(2n+ 1)0
their expansions in powers of sins 6. These expansions are*
n.(n + s -1)!
cos2n#= 2 (-)' , wo\t (2 sin ey,
,=o (n-s)!(2s)!
|
sin (x sin 9) = 1 Jm+ , (X)
{
I (-)
gj$| (* - )""(
(~)*(2sin0) +1
1 , n\
sin (z sin 6)=
5
X L -L ro TT\i \
f
" (2n
2 +
r^-^r
l).(n + s)!
r ,
J+i(*)
J
l=Jo(s) + 2 5 J2n (4
n=l
Jm+1 (z >- *
I
(
* ' "-. Jn^7)\ (
* *' 2>
>
The first of these is the result already obtained ; the others may be com-
bined into the single formula
I C + *.)( +
. -Dl
(1) (i . r = w=0 Jm> w (m _ 1( 2| 3...)
The rearrangement of the double series now needs justification the rearrangement ; is
permissible if we can establish the absolute convergence of the double series.
and so the series of moduli is convergent. The series for cos (z sin $) may be treated in
a similar manner.
The somewhat elaborate analysis which has just been given is avoided in
Lommel's proof by induction, but this proof suffers from the fact that it is
supposed that the form of the expansion is known and merely needs verifica-
tion. If, following Lommel, we assume that
i
tor- n=0
< + * t
>-iT
+ - 1 )'
jM ,(A
* ZeitschriftfUr Math, und Phys. n. (1857), pp. 140141.
f Theorie der BesseVschen Functionen (Leipzig, 1867), p. 38.
t Studien liber die BesseF?chen Functionen (Leipzig, 1868), pp. 35 36. Lommel's investigation
in given later in this section.
-
[which has been proved in 2'22 (8) in the special case m = 1], we have
/i \,+, S (m + 2n) (m + n 1) .
! , r , v
n=0 n -
* J~+ (*)
, x
n y I
2 (wi -Mi)! f , r , , r / ,,
=0 n\
=0 w!
Let t be a complex variable and let u be denned by the equation u= - 2 , so that when
wi! /"
expi-ia^-l/O}---^*
^/ lw
(m +2n).(m + n-l)! .
-p<-^'-'/',
*
r +2n)
-ir'
, -'
i! '--"
*
!
'
/ m+2n
\
zh
n=0
when we calculate the sum of the residues at the origin for the last integral ; the inter-
change of the order of summation and integration is permitted because the series converges
uniformly on the contour and the required result is obtained.
;
The formulae
n=0 m=0
in which p is any positive integer [zero included in (2) but not in (1)] and 1* is a numeri-
cal coefficient, are evidently very closely connected with the results of 27. The formulae
* Messenger, xxxii. (1903), p. 8; a proof on the same lines for the case i =l had been pre-
viously given by Kapteyn, Nieuw Archief voor Wiskunde, xx. (1893), p. 120.
:
were obtained by Schlomilch, Zeitschrift fur Math, und Phys. II. (1857), p. 141, and he
gave, as the value of P^ p \
m
w>
<* i-r-c*(-**r
{) m ~A=o 2m .m!
where m Ck is a binomial coefficient and the last term of the summation is that for which k
ishm - 1 or \{rti 1). To prove the first formula, take the equation 2*22 (.1), differentiate
2p times with respect to 0, and then make 6 equal to zero. It is thus found that
/<
2
xn 2 /o
- ).<*.). *.
x*. r /x
w R**"
_[__ ^
cos (* sin *)"!
;
j,. o =
V
[^
d *p
Z
Mi;
(-yfaiiPT]
(2m) ,
j,,,,
The terms of the series for which m>p, when expanded in ascending powers of 6,
contain no term in 0*, and so it is sufficient to evaluate
m=0 v * wl7 !
Jfc-0
p
= 2(-)lS jP-Zff,
=o
since terms equidistant from the beginning and the end of the summation with respect to
I- are equal. The truth of equation (1) is now evident, and equation (2) is proved in a
similar manner from 2*22 (2).
The reader will easily establish the following special cases, which were stated by
Schlomilch
P J (*) + 33 Ja (i + 53 Jb (*) + = (+**),
x ) ...
(
4) '2V2 ?)+4V4 W+eV6 ()+...=^,
(
'-A- +
v (2m + 2n).(2m n-l)!
am J,Wi+m (2zsm0),
-.
,
(2sin0)= 2 v /a .
/-*<*>
"(2m)!*, .*!
2-72] THE BESSEL COEFFICIENTS 37
(2)
<^-S.L-ri^<*
and this is true when m = 0, for it then reduces to Hansen's formula of 25
(3)
**= ~
tJ. 4=2
* 4n* (4 - 22) J* (*),
If we differentiate (1), use 2*12 (2) and then rearrange, it is readily found that
* Leipziger Berichte, xxi. (1869), p. 226. [Math. Ann. in. (1871), p. 585.]
CHAPTER III
BESSEL FUNCTIONS
The Bessel coefficients, which were discussed in Chapter II, are functions
of two variables, z and n, of which z is unrestricted but n has hitherto been
required to be an integer. We shall now generalise these functions so as to
have functions of two unrestricted (complex) variables.
This generalisation was effected by Lommel*, whose definition of a Bessel
function was effected by a generalisation of Poisson's integral ; in the course
of his analysis he shewed that the function, so defined, is a solution of the
linear differential equation which is to be discussed in this section. Lommel's
definition of the Bessel function Jv {z) of argument z and order v wasf
and the integral on the right is convergent for general complex values .of v
for which R(v) exceeds |. Lommel apparently contemplated only real
values of v, the extension to complex values being effected by HankelJ;
functions of order less than - \ were defined by Lommel by means of an ex-
tension of the recurrence formulae of 212.
Let us now construct a solution of (1) which is valid near the oiigin; the
form assumed for such a solution is a series of ascending powers of z, say
y= 2 cm z*
+m
,
m-0
where the index o and the coefficients c m are to be determined, with the pro-
viso that c is not zero.
For brevity the differential operator which occurs in (1) will be called V,
so that
V, 2 cm z* +m = 2 c w {(a + m) 2
-i/2 }s" +m + 2 cm z a+m+a .
The expression on the right reduces to the first term of the first series,
namely c (a8 Vs) za if we choose the coefficients cm so that the coefficients of
,
/(^{(a + l)*-* 2
}
=0
ca {(a + 2) -*, j+c
2 2
=0
c {(a + 3) -i/ + Cl
8
2 2
}
=0
(3)
m-0
From this result, it is evident that the postulated series can be a solution
of (1) only if a =+ v; for c is not zero, and za vanishes only for exceptional
values of z.
Now
consider the mth. equation in the system (3) when w > 1. It can be
written in the form
cm (o - v + m) (a + v + m) + Cm-* = 0,
and so it determines cm in terms of Cm^ for all values of m greater than 1
unless a-i>ora + i/isa negative integer, that is, is a negative
unless 2v
integer (when a= v) or unless 2v is a negative integer (when a v).
We disregard these exceptional values of v for the moment (see 3" 11,
3*5), and then (o + m) p*
2
does not vanish when m 1, 2, 3, .... It now
* When the constants a and c m have been determined by the following analysis, the series
obtained by formal processes is easily seen to be convergent and differentiable, so that the formal
follows from the equations (3) that d = c3 = c5 = ... = 0, and that c 2Ml is ex-
c"* = (-r* .
(5) c zv 1+ 2
-i m
! (v
(-)" (**)""
+ 1) {v + 2) + m)
. . . (v
is a formal solution
soli of equation (1). If we take a= v, we obtain a second
formal solution
(-)m (^y
(6) c 'z-"\i+ s
,-i m\(- v + 1) (- v + 2) ... (-v + m)]'
In the latter, c has been written in place of c because the procedure of
'
,
obtaining (6) can evidently be carried out without reference to the existence
of (5), so that the constants c and c ' are independent.
(') co = oirv..
i \
c =
2T(v+l)' i
2-T{-p + l)'
The series (5) and (6) may now be written
An interesting symbolic solution of Bessel's equation has been given by Cotter* in the
form
"- 1
[l+zv D- 1 z- 2 D- 1 zv+1 ]- 1 (Azv +Bz- v ),
where Dsd/dz while A and B are constants. This may be derived by writing successively
[D(zD~2v)+z]zv y=0,
[zD-2p+D~ 1 z]z"y= -2vB,
zD (z~ vy) + z~ 2v D~ V+V= - 1vBz- % \
z- vy+D- z-^- x D- x z v+l y=A +Bz~
x tv
,
and so
( y q
(2) Ctm =
2.4...(2m).(2r + 3)(2r + 5)...(2r + 27/i + l")
which is the value of c^ given by 3'1 when a and v are replaced by r + \.
If we take
1
c
-2~r"+4r(r + f)'
we obtain the solution
,= m!r(r + m + f)'
which is naturally denoted by the symbol Jr+ ^(z), so that the definition of
Cgr+i + Czr l 0,
and this equation is satisfied by an arbitrary value of Car+r> when ra > r, C2m+i
is defined by the equation
-r c
. = (-)w 2r +i
27,1+1
(2r + 3) (2r + 5) ... (2m + 1) . 2 . 4 ... (2m- 2r)
* Proc. R. Irish.
Acad. xxvn. (A), (1909), pp. 157161.
f The cases combine to form the case in which 2v is an integer.
42 THEORY OF BESSEL FUNCTIONS [CHAP. Ill
Co
2-"-* T(\-r) /_,_! (z) + Czr+1 2*+* r (r + f) Jr+i (z).
It follows that no modification in the definition of Jv (z) is necessary when
v=(r + ); the real peculiarity of the solution in this case is that the
negative root of the indicial equation gives rise to a series containing two
arbitrary constants, c and c^+i, i.o. to the general solution of the differential
equation.
respect to the independent variable, then the solutions are linearly inde-
pendent if the Wronskian determinant^
y* y*
yl y*
does not vanish identically; and if the Wronskian does vanish identically,
then, either one of the two solutions vanishes identically, or else the ratio of
the two solutions is a constant.
If the Wronskian does not vanish identically, then any solution of the
differential equation is expressible in the form cx yx + d y2 where cx and ca are
constants depending on the particular solution under consideration; the
solutions y and y2 are then
x said to form a fundamental system.
W,{y,,y}, 8K{yi,y.},
the former being used when it is necessary to specify the independent variable.
V v J_ v {z) = 0, V v J,(z) = Q
by J (z), J_ v (z) respectively and subtract the results, we obtain an equation
which may be written in the form
( 23), p. 109. Proofs of the theorems quoted in the text are given by Forsyth, Treatise on
Differential Equations (1914), 72 74.
3*12] BESSEL FUNCTIONS 43
with similar expressions for J-V (z) and /'_ (z); and hence
V '
TTZ
If we compare this result with (1), it is evident that the expression on the
right which is 0(z) must vanish, and so*
Since sini^r is not zero (because v is not an integer), the functions Jv (z),
/_ {z) form a fundamental system of solutions of equation 31 (1).
When v is an integer, n, we have seen that, with the definition of 21 (2),
nK) m\r(-n + m + Ty
mZ
Since the first n terms of the last series vanish, the series is easily reduced to
( )n Jn(z)y so that the two definitions of J-n(z) are equivalent, and the
functions Jn (z), J_n (z) do not form a fundamental system of solutions of
Bessel's equation for functions of order n. The determination of a fundamental
system jn this case will be investigated in 363.
To sum Jv (z) is defined, for all values of v, by the
up, the function
expansion of and J (z), so defined, is always a solution of the equation
31 (8);
Vt/ = 0. When v is not an integer, a fundamental system of solutions of this
equation is formed by the functions Jy (z) and /_ (z).
the base, and the basic Gamma function rp (v) is defined to satisfy the recurrence formula
Tp (u + l) = [u].rp (v).
The basic Bessel function is then defined by replacing the numbers which occur in the
series for the Bessel function by basic numbers. It has been shewn that very many theorems
* This result is due to Lommel, Math. Ann. iv. (1871), p. 104. He derived the value of C by
making z -* <x> and using the approximate formulae which will be investigated in Chapter vn.
44 THEORY OF BESSEL FUNCTIONS [CHAP. Ill
concerning Bessel functions have their analogues in the theory of basic Bessel functions,
but the discussion of these analogues is outside the scope of this work. Jackson's main
results are to be found in a series of papers, Proc. Edinburgh Math. Soc. xxi. (1903), pp.
6572 xxn. (1904), pp. 8085; Proc. Royal Soc. Edinburgh, xxv. (1904), pp. 273276;
;
Trans. Royal Soc. Edinburgh, xli. (1905), pp. 128, 105118, 399-408; Proc. London
Math. Soc. (2) I. (1904), pp. 361366; (2) II. (1905), pp. 192220; (2) ill. (1905), pp. 123.
The more obvious generalisation of the Bessel function, obtained by increasing the
number of sets of factors in the denominators of the terms of the series, will be dealt with
in 4 4. In connexion with this generalisation see Cailler, Mem. de la Soc. de Phys. de
#
Geneve, xxxiv. (1905), p. 354; another generalisation, in the shape of Bessel functions of two
variables, has been dealt with by Whittaker, Math. Ann. lvil (1903), p. 351, and Peres,
Comptes Rendus, clxi. (1915), pp. 168 170.
The series which defines Jv (z) converges absolutely and uniformly* in any
closed domain of values of z [the origin not being a point of the domain when
R (y) < 0], and in any bounded domain of values of v.
For, when \v\ % N and \z\ % A, the test ratio for this series is
-I*
m(v + m) m (m A ) r
This choice of to being independent of v and z, the result stated follows from
the test of Weierstrass.
Hence f an analytic function of z for all values ofz(z = possibly
Jv (z) is
(i)
W=tvt)( 1+ >
where 6\ <exp 1'
|
{jJi]}-
and j
v +l \
is the smallest of the numbers v + 1 v+2
|
1, | |, 1 1+3|, ....
This result may be proved in exactly the same way as 2'11 (5) ; it should be com-
pared with the inequalities which will be given in 3*3.
Finally, the function z", which is a factor of Jv (z), needs precise specifica-
(1902), p. 494.
l
/ _ \m ^2 v i +ai
~ ** 7\.
1+im
Zo2"- .m\r{v + m)
- z" Jv^ (z).
When we differentiate out the product on the left, we at once obtain (3).
In like manner,
/_\r^2wi
=1 2 K+2m-i .
( W _ 1)! T (l/ +W+ 1)
oc / \m+i n+i
= +m+1
w = 2' . ml r (v + w + 2)
whence (4) is obvious; and (2) and (1) may be obtained by adding and sub-
tracting (3) and (4).
* StUdien Uber die BesseVschen Functionen (Leipzig, 1868), pp. 2, 6, 7. Formula (3) was given
when v is half of an odd integer by Plana, Mem. delta R. Accad. delie Sci. di Torino, xxvi. (1821),
p. 533.
1
46 THEORY OF BESS EL FUNCTIONS [CHAP. Ill
= *-~ j- <*>'
(5)
(AJ [zvj > {z)]
W
(6) ir-J.{*)} =(-)' z>"Jv+m (z)
(^)
by repeated differentiations, when m is any positive integer.
Lommel obtained all these results from his generalisation of Poisson's
integral which has been described in 3*1.
The formula (1) has been extensively used* in the construction of Tables
of Bessel functions.
(7)
T /vtv + JL
J. (z) J^ v
T-/\r/\2
J
(z) = - (z) v ., (z)
sin vtr
7TZ
from formula (2) of 312.
(1) ~ {K
dx*
Vi
*(a) iS(x)} + {K Vth (x) iS(w)}
+ l*
+ 2(fr) dx
[Kv ^x) iS{*)) - 0,
x
(2) K w+ltM (x)
.
= K {x) + K" (x),
Vtll VttL
t Math. Ann. iv. (1871), p. 105. Some associated formulae are given in 3*63.
solutions of this equation in the form of series were given by Boole* many
years ago.
this expression was adopted by Lommelj as the definition of Jv (z) for positive
values of v + \.
Lommel subsequently proved that the function, so denned, is a solution of Bessel's
generalised equation and that it satisfies the recurrence formulae of 32 and he then
;
defined Jv (z) for values of v in the intervals (-, -#),(_#, _), (J| -J), t ... by suc-
cessive applications of 3 - 2 (1).
To deduce (1) from the, definition of Jv {z) adopted in this work, we trans-
form the general term of the series for Jv (z) in the following manner:
(-)rw (i*),+swt
= (-y*Q*Y ** r(y+j)r(m + i)
m!I> + m + l) I> + )r(*)-(2ro)f I> + m + l)
R (v) > |.
provided that
Now when R (v) > , the series
l= i (2wz)!
* Phil. Tram, of the Royal Soc. 1844, p. 239. See also a question set in the Mathematical
Tripos, 1894.
t Journal de VEcole R. Poly technique, xn. (cahier 19), (1823), pp. 300 et teq., 340 et $eq.
Strictly speaking, Poiason shewed that, when 2v is an odd integer, the expression on the
right
multiplied by Jz is a solution of the equation derived from Bessel's equation by the appropriate
change of dependent variable.
t Studien liber die BetseVschen Functionen (Leipzig, 1868), pp. 1 et seq.
48 THEORY OF BESSEL FUNCTIONS [CHAP. Iir
whence the result stated follows by making the substitution t = sin2 6 and
using the fact that the integrand is unaffected by writing ir 6 in place of 6-
When -&< R(v)<$, the analysis necessary to establish the last equation is a little
more The simplest procedure seems
elaborate. to be to take the series with the first two
terms omitted and integrate by parts, thus
_ (1
Jov+iU 2 (2mj! '>
J*
(1
/ +**ui (2m)i j*
on integrating by parts a second time. The interchange of the order of summation and
integration in the second line of analysis is permissible on account of the uniformity of
convergence of the series. On adding the integrals corresponding to the terms m0, m = 1
(which are convergent), we obtain the desired result.
J- <" "
r(5(i) />-* V ~ ** cos I* G - ' dt
Obvious transformations of this result, in addition to (1), are the following:
<2 >
J - <*> -
nffir-w ,C -
(1 <'>"""' cos <*> *
<3>
* <*> - nMm fj 1 - ,r * cos (zt) *
2
<5) ^ (z) = cos (z cos 9) sin" " d "'
r(- +i)r( C'
f ^ ( cos ^) sin2
""^ cos ^'
is sometimes usefuiy^t ^s valid, only when R{v)> \.
3*31] BESSEL FUNCTIONS 49
An expansion involving Bernoullian polynomials has been obtained from (4) by Nielsen*
with the help of the expansion
'1
f-*-(i-.-)i
rfl
*" +,tf+1)
'= (n + 1)!
in which <f> n ()
denotes the nth Bernoullian polynomial and aizt.
[Note. Integrals of the type (3) were studied before Poissou by Plana, Mem. delta R.
Accad. delle Sci. di Torino, xxvi. (1821), pp. 519
538, and subsequently by Kummer,
Journal
Math, xn. (1834), pp. 14-4 147; Lobatto, Journal fiir Math. xvn. (1837), pp.
filr
363371; and Duhamel, Cours d' Analyse, II. (Paris, 1840), pp. 118 121.
A function, substantially equivalent to Jv (2), denned by the equation
zv
e
1
{v-af- (v-P)"- 1 dv
f*
was also discussed by Lommel, Archiv der Math, und Phys. XL. (1863), pp. 101126. In
connexion with this integral see also Euler, Inst. Cole. Int. II. (Petersburg, 1769), 1036,
and Petzval, Integration der linearen Differentialgleichungen (Vienna, 1851), p. 48.]
(i) !-/.<*)!
rj+ffrft-j
/' exp l ' w '
sin " ede
By using the recurrence formulae 3-*2 (1) and (4), we deduce in a similar
manner that
(2) |JpW |
<TT
i^jl + jFi lggL J]
}exp|/W| <-|<-i>,
By using the expressionf {2/(7rz)}* cos z for J^ (z) it may be shewn that
(1) is valid when v = \.
These inequalities should be compared with the less stringent inequalities
obtained in 313. When v is complex, inequalities of a more complicated
character can be obtained in the same manner, but they are of no great im-
portance.
* Math. Ann. lix. (1904), p. 108. The notation used in the text is that given in Modern
Analysis, 7*2; Nielsen uses a different notation.
t The reader should have no difficulty in verifying this result. A formal proof of a more
general theorem will be given in 3*4.
*
-"
in which Gn v (t) is the coefficient of a" in the expansion of (1 2at + o 2
)
in
ascending powers of a, is due to Gegenbauer* the formula is valid when ;
'
'* W - r(, + i'5*)r( )i>
(
t
(1 - (=) ' + ""'
^ ,,()
(-2i)nr(, + ,,4-^ra)J_ 1
e
1 <** p-
Now it is known thatj
dn (l- t*)*+-l (- 2) n T ( v + n +
!
$) T (2i/)
(1 -)-* Cn "(0,
whence we have
n
w J ^ - (-O -r(2,).7 ,!(^ p ?
c w<.
(3) t/ (2)
(
^-r( v + i)r^)r(2 + n)J_ I; 1
e (1 -
*** *"'
(7(^)) - r(^i^+V^ (P )
(5)
1
in which i*"* denotes a generalised Legendre function, is due to Filon, Phil. Mag. (6) vi.
Jv (Z) = %-r-r
TT 1 (V) Jo
(* eiz sin8" .
f sin *"
.'
8 1
yjr dyjr,
J* () = e iv * *9
sin2 ' sin8-- 1 fd0dyfr
^FJ-\ I"!*
_ (W [(
* It is supposed tbat
9*y _ Y{y + l)z v ->x
dz*
~ r("-M+i)
t Wiener Sitzungsberichte, ucxiv. (2), (1877), pp. 128 129.
t This method is effective in proving numerous formulae of which analytical proofs were
given by Gegenbauer ; and it seems not unlikely that he discovered these formulae by the method
in question ; cf. 12*12, 12-14. The device is used by Beltrami, Lombardo Reiidiconti, (2) xiii.
(1880), p. 328, for a father different purpose.
The symbol jj m ^ means that the integration extends over the surface of the hemisphere on
which i is positive.
52 THEORY OF BESSEL FUNCTIONS [CHAP. Ill
Now the integrandan integral periodic function of yjr. and so the limits of
is
We thus get
(ktffY /'i"'/" a +2w
J v (w) = -, /. e iv sin
e cos * cos2 " -1 sin Bdslr d$
TTl (v)Jo J*
W
= -te K iw8i,l * C08 (* +
cos2 ""1
'
so that
(Aw)" /*" f 2 "
J v (a) = - L, 1 I exp [i (Z z cos <f>)
sin cos yfr
iz sin <f>
sin t/t sin 0]
2 '- 1
con 6 sin Bdyfrde.
The only difference between this formula and the formula
"- 1
Jv {m) = v
l, / exp [iw sin cos -f] cos 2 sin 0ety> d0
7Tl (v) Jo .'
is in the form of the exponential factor ; and we now retrace the steps of the
analysis with the modified form of the exponential factor. When the steps are
retraced the successive exponents are
(IwY
y,,
ff ex P [*^ cos
I
** ( C(>s <}> cos + sin <f>
sin cos yfr)]
"" 1
sin 2
^ sin 2"
0d\frd0,
and consequently Gegenbauer's formula is established.
[Note. The device of using transformations of polar coordinates, after the manner of
this section, to evaluate definite integrals seems to be due to Legendre, M4m. de I' Acad, des
Sci., 1789, p. 372, and Poisson, Mem. de FA cad. des Sci. in. (1818), p. 126.]
We shall now deduce from Poisson's integral the important theorem that,
when v is half of an odd integer, the function Jv (z) is expressible in finite teivns
by means of algebraic and trigonometrical functions of z.
It will appear later ( 474) that, when
such a value, then ,TV (z) v has not
is not so expressible ; but of course this converse theorem is of a much more
recondite character than the theorem which is now about to be proved.
.
[Note. Solutions in finite terms of differential equations associated with Jn+ i (z) were ob-
tained by various early writers ; it was observed by Euler, Misc. Taurinensia, in. (1762
1765), p. 76 that a solution of the equation for e i2 J)t + , (z) is expressible in finite terms ; while
the equation satisfied by 2* Jn+i (.s) was solved in finite terms by Laplace, Conn, des Terns.
1823 [1820], pp. 245257 and Mecanique Celeste, v. (Paris, 1825), pp. 8284 ; by Plana, Mem.
della R. Accad. delle Sci. di Torino, xxvi.
(1821), pp. 533534; by Paoli, Mem. di Mat. e
di Fis. (Modena), xx. (1828), pp. 183188; and also by Stokes in
1850, Trans. Camh. Phil.
Soc. ix. (1856), p. 187 [Math, and Phys. Papers, n.
(1883), p. 356]. The investigation
which will now be given is based on the work of Lommel, Studien liber die BesseVsehen
Fimctionen (Leipzig, 1868), pp. 51 56.]
It is convenient to restrict n to be a positive integer (zero included), and
then, by 3-3 (4),
e r+x
nly/ir \_ r=0 z dtr J_i'
when we + times
integrate by parts 2n 1 ; since (1 - t*)
n is a
polynomial of
degree 2n, the process then terminates.
To simplify the last expression we observe that if dr (l -t2 n /dtr be cal-
)
culated from Leibniz' theorem by writing (1 -t2 n = (1 - t)n (l + t) n the only
) ,
term which does not vanish at the upper limit arises from differentiating n
times the factor (1 - t)n and therefore from differentiating the other factor
,
r n times; so that wo need consider only the terms for which r^n.
It follows that
Um * KZ) Jir L
( }
r rn ^+1 .(r-n)!(2n-r)!
2 /"_ V+l Oan-r ~\
+ (_yi+i e-i2 v
v V L -
-,
1 t
V '
,.-s' +1 .(r-w)!(2n-r)! '
and hence
= -J [> *~"<n + r)l (- t-r-n-x (n + r) ;-|
(1) </ +i (s) i
This result may be written in the form*
(2) J^(.)-fiy[n(,-
+1W WL
i )'S' -,<-<*2L_
5
(2r)I(n-2r)i(2*) r-c,
cos^-i^r' 2
(-M* ++!)! 1
_o P (2?- + l)!(n-2r-l)!(2^'+' '
In particular we have
S
(3) Jj (*) = () sin m, J %
(z) =
(-) (
-^ - cos z) ;
the former of these results is also obvious from the power series for J (z).
\zdz) z
iz and
as a polynomial in \jz multiplied by e , so we must have
oiz
(y-(~ + r) l
y.W<* V****.
'
,_"o r (n
!
- r) S (2*)
r ~ V ;
WW * '
for, if not, the preceding identity would lead to a result of the form
where fa (z) and fa (z) are polynomials in 1/z; and such an identity is obviously
impossible*.
"
1 e
r -o r!(n-r)!(2^" ,= r! (w-r)! (2-) r
n
= (-r (2tt> *+> {*-* j_j (*>i
( i)
= (-)(27r^./_n_ J (^).
Consequently
(4) J
</_, M;a)= -J [>
W _ v(te) ^ |
2 _____
**.( +
+ ^I (-Y+.(n
_
+ r)n
^
r)i
e
,
rS(n r)!(2 j
j-* um-_- *
I (-r <**)
r(j) ,to4-# ...(m-i)'
and hence
_
In particular, we have
The explicit expression of a number of these functions can be written down from
numerical results contained in a letter from Hermite to Gordan, Journal fur Math. lxvi.
(1873), pp. 303311.
In his researches on vibrating spheres surrounded by a gas, Stokes, Phil. Trans, of the
Royal Soc. clviii. (1868), p. 451 {Math, and Phys. Papers, iv. (1904), p. 306], made use of
the series
n( + l ) (-!)( + !) ( + 2 )
"
r
2.imr 2. 4. fair) 2
"
K "'
which is annihilated by the operator
d* . d n(n + l)
dr2 dr r2
2
'' l
2
2 d
+ -*-+.,
, ,
- r- sin 6.
1
.
6
55
f
\
.
sin 6 52}
. v)
+ m2 .
cr* " or
r c6 { dff)
In this notation Stokes was followed by Rayleigh, Proc. London Math. Soc. iv. (,1873),
pp. 93 103, 253283, and again Proc. Royal Soc. i.xxii. (1903), pp. 4041 [Scientific
Papers, v. (1912), pp. 112114], apart from the comparatively trivial change that Rayleigh
In order to obtain a solution finite at the origin, Rayleigh found it necessary to take
Sn = -
'
(
n+ 1
Sn in the course of his analysis, and then
)
t>
ir i (ir\ / r \ n e~ ir
rn+1 \ rcrj r
In order to have a simple notation for the combinations of the types e* /(tr) which
ir
are required for solutions finite at the origin, Lamb found it convenient to write
in his earlier papers, Proc. London Math. Soc. xm. (1882), pp. 5166; 189 212; xv.
(1884), pp. 139149 xvi. (1885), pp. 2743 Phil. Trans, of the Royal Soc. clxxiv. (1883),
; ;
pp. 519549; and he was followed by Rayleigh, Proc. Royal Soc. lxxvii. A, (1906),
pp. 486499 [Scientific Papers, v. (1912), pp. 300312], and by Love* Proc. London
Math. Soc. xxx. (1899), pp. 308321.
With this notation it is evident that
.=p^,=(-)-...(
! !. + .).( i)'^. r
Subsequently, however, Lamb found it convenient to modify this notation, and accord-
ingly in his treatise on Hydrodynamics and also Proc. London Math. Soc. xxxn. (1901),
-j-\
n e~"
=*n(*) -l'V'i(*)
so that * n () = pr+i
>t+ ^
" ,
'
vnw-
<A
= () to
Y~
while Rayleigh, Phil. Trans, of the Royal Soc. cciii. A, (1904), pp. 87110 [Scientific Papers,
v. (t912) pp. 149161] found it convenient to replace the symbol /(*) by x*(*)- Love,
Phil. Trans, of the Royal Soc. ccxv. A, (1915), p. 112 omitted the factor (-) and wrote
, .
( dye-** . , x
/rf\"sin2
while yet another notation has been used by Sommerfeld, Ann. der Physik und Chemie, (4)
xxvm. (1909), pp. 665736, and two of his pupils, namely March, Ann. der Physik und
Chemie, (4) xxxvu. (1912), p. 29 and Rybczyriski, Ann. der Physik und Chemie, (4) xli.
(1913), p. 191 ; this notation is
^(.HW^-^ l
(-ai)"T'
C(*)(MH'+i (*) + (-)* tf--*(*>l
and it is certainly the best adapted for the investigation on electric waves which was the
subject of their researches.
^ J ,
but, as
t This is nearer the notation used by Heine, Handbuch der Kugelfiinctionen, i. (Berlin, 1878),
p. 82 ; except that Heine defined ^ n (z) to be twice the expression on the right in his treatise, but
not in his memoir, Journal filr Math. lxix. (1869), pp. 128 141.
3*5] BESSEL FUNCTIONS 57
Sommerfeld's notation is a slightly modified form of the notation used by L. Lorenz, who
used vn and vn +( ) n iwn in place of >frH and fn see his memoir on reflexion and refraction
;
of light, K. Danske Videnskabernes Selskabs Skrifter, (6) vi. (1890), [Oeuvres scientifiqucs, I.
linearly independent of Jn (z); and the combination of this solution with JH (z)
will give a fundamental system of solutions.
The solution which will now be constructed was obtained by Hankel*;
the full details of the analysis involved in the construction were first published
by BOcherf.
An method of constructing HankeFs solution was discovered by Forsyth
alternative
his procedure is based on the generalmethod of Frobenius, Journal fiir Math, lxxvi. (1874),
pp. 214235, for dealing with any linear differential equation. Forsyth's solution was
contained in his lectures on differential equations delivered in Cambridge in 1894, and it
has since been published in his Theory of Differential Equations, iv. (Cambridge, 1902),
pp. 101
102, and in his Treatise on Differential Equations (London, 1903 and 1914),
Chapter vi. note 1.
t Annals of Math. vi. (1892), pp. 85 90. See also Niemoller, Zeitsehri/t fiir Math, und Phys.
xxv. (1880), pp. 65-71
X The essence of Hankel's investigation is the construction of an expression which satisfies
the equation when v is not an integer, which assumes an undetermined form when v is equal to
the integer n and which has a limit when v-*~n.
w. B. F. 3
58 THEORY OF BESSEL FUNCTIONS [CHAP. Ill
It is evident that
l ;
dv du
as v - n, since both of the differential coefficients existf .
Hence
^ w v n
exists; it is called a Bessel function of the second kind of order n.
v n
v-*- n
and also
dJA z )
dJA Vt dJ"-,(fY
(2) *.(*)- =-
(
K >
dv dv _
* The reader will realise that, given a solution of a differential equation, it is not obvious that
a limiting form of this solution is a solution of the corresponding limiting form of the equation.
+ See 3-1. It is conventional to write differentiations with respect to z as total differential
coefficients while differentiations with respect to v are written as partial differential coefficient-.
Of course, in many parts of the theory, variations in v are not contemplated.
% The symbol Yn (z), which was actually used by Hankel, is used in this work to denote a
function equal to ljir times Hankel's function ( 3*54).
so that
= l im
J-A*)-(-YJAz)
M ^ fi + n '
* SUidien iiber die BesseVschen Functionen (Leipzig, 1868), p. 87. Lommel actually proved
this result for what is sometimes called Neumann's function of the second kind. See 3-58 (8).
t Court d Analyse,
1
n. (Paris, 1840), pp. 122 124.
:
=2 2
rra^ (log^-^(m + l)},
where yfr denotes, as is customary, the logarithmic derivate of the Gamma-
function*.
Since 0< ^ (i + 1) < m when w=l, 2, 3, ... the convergence of the series for YoG?)
may be established by using D'Alembert's ratio-test for the series in which ^(m+1) is
J - X
log (*) . (s)
^yy * (m +
1)J
,
Y. () -2 -2 JS^ 1
(3)
The reader
fry
+ log ft*)} ./.(*)
will observe that
yy | I+ g + + raj
!
This function was adopted as the canonical function of the second kind of order zero by
Neumann, Theorie der BesseVschen Functional (Leipzig, 1867), pp. 42 44; see 3 57. -
But the series was obtained as a solution of Bessel's equation, long before, by Euler \.
Eider's result in his own notatiou is that the general solution of the equation
V~ n*
X
\JW X + 1.8.27'^ 1.8.27.64^
+et -
+A 1 - 3- r + JL- X2 _ _L_ tT 3
g
+ 1.4.9. .* - etc Ix
(
\ nn
.
+a nn
X+ 1.4n* X 1 . 4 . 9 f
-
C +1
. 4 . 9 . 16
*'
* Modern Atialy-iis, Ch. xn. It is to be remembered that, when m is a positive integer, then
*U)=-7, V'(' + 1 )=
J
+ 2+ - +-->
where 7 denotes Euler 's constant, 0-5772157 ....
t Inst. Calc. Int. 11. (Petersburg, 1709), 977, pp. 233235. See also Acta Acad. Petrop. v.
where A and a are arbitrary constants. He gave the following law to determine successive
numerators in the first line
w 2(1 + *1+...+!) =
mj ^, m
\1 !
<r m+ i = (2m+l)<r-m ii
m.1
(r .
It is clear that
?^.<*>- 3 (- )m (i*) H * w |
| (
dv m =()dv \m\ T(i> + l)jm+
" V
' l m!
TO r(-i/ + m + l) *! r(-i/ +m+ l)'
1 , r (v in) sin (v m) it
by
V(-v + m+l) *
'
[~ (
(^)~'" m r (y - wt) sin (v - m) tt
"
t
) 1
Hence
pM*)1 . v 1
(-) n r(-TO)(|^)-" 4 * ,t
+
,;. J(-,+.) it- h <*') + <- + ' +i '
that is to say
() ( **> +(
L dp _U- ( }
w =o m! > .t m!(n + m)!
x {
lo g(i*)-"f (m+1)},
when we replace m by n + m in the second series.
On combining (1) and (2) we have Hankel's formula, namely
w w m=o *!
a
m=0 w!(n + m)!
x {2 log ($z) - yfr (m + 1) - i/r (n + m + 1)}
w 1
(n 1} l
11, 1
12 n
It is frequently convenient (following Lommel*) to write
d Z)
(4) %(*) = -^ -Mz)logz,
so that
<5
>
^^-S J<.nXLi) ll082+n ^ m+1)];
when v is a negative integer, ^ (s) is defined by the limit of the expression
on the right.
We thus have
The complete solution of .v -v^ + ay=0 was given in the form of a series (part of which
contained a logarithmic factor) by Euler, Inst. Calc. Int. n. (Petersburg, 1769), 935,
936 solutions of this equation are
;
** J (2aM),
x #* Yi (2aM).
Euler also gave (ibid. 937, 938) the complete solution of x% -r^ + ay =0 ; solutions of
Hitherto the function of the second kind has been defined only when its
order is an integer. The definition which was adopted by Hankel* for un-
restricted values of v (integral values of 2v excepted) is
sin2i/7r
This definition fails both when v is an integer and when v is half of an
odd integer, because of the vanishing of sin 2vtt. The failure is complete in
the latter case ; but, in the former case, the function is defined by the limit
of the expression on the right and it is easy to reconcile this definition with
the definition of 3'5.
v n
= J z COSv7r -
(-)* lim \ "( )
J-AzY \
v+n L v-n J
= YM (z) + lim > ~n Jv (z)
r+n L v J
= Y n (s),
and so we have proved that
(2) limY(*) = Yn (*).
It is now evident thatY (z), defined either by (1) or by the limiting form
of that equation, a solution of Bessel's equation for functions of order v both
is
when (i) v has any value for which 2v is not an integer, and when (ii) v is an
integer the latter result follows from equation (2) combined with 3*5 (3).
:
The
function Y (z), defined in this way, is called a Bessel function of the
second kind (of Hankel's type) of order v ; and the definition fails only when
v +1 is an integer.
Note. The reader should be
careful to observe that, in spite of the change of form, the
function Y (z),
qua function of v, is continuous at v = n, except when z is zero; and, in
fact, Jv {z) and Y(z) approach their limits J (z) and Y (z), as v-*~n, uniformly with
n
respect to z, except in the neighbourhood of z = 0, where n is any integer, positive or negative.
The function which was adopted by Weber as the canonical function of the
second kind is expressible in terms of functions of the first kind by the formula*
Subsequent writers, however, have usually omitted this factor \tr, e.g. Graf
and Gubler in their treatise f, and also Nielsen, so that these writers work with
Weber's function.
The symbol K is, however, used largely in this country, especially by
Physicists, to denote a completely different type of Bessel function ( 3'7),
and so it is advisable to use a different notation. The procedure which seems
to produce least confusion is to use the symbol Yv {z) to denote Weber s function,
after the manner of Nielsen J, and to adopt this as the canonical function of
the second kind, save in rare instances when the use of Hankel's function of
integral order saves the insertion of the number tr in certain formulae.
We thus have
J (jOcgg '" --*--(*>
= cos^;
(1) Yv (z) = "
JM^^JLrl-A^ = I
(2) Yn (z) = lim
Sin VTT 7T
Yn (0 .
,-*. n
[Note.Schlafli's fuuction has been used by B6cher, Annals of Math. vi. (1892),
pp. 8590, and by McMahon, Annals of Math. vm. (1894), pp. 5761; ix. (1895),
pp. 23 30. Schafheitlin and Heaviside use Weber's function with the sign changed, so
that the function which we (with Nielsen) denote by Yv (z) is written as Yv (z) by
Schafheitlin and (when v = n) as GM (z) by Heaviside||.
Gray and Mathews sometiinesIT use Weber's function, and they denote it by the
symbol Yn .
* Weber's definition was by an integral (see 6*1) which is equal to this expression ; the
expression (with the factor \ir inserted) was actually given by Schlafii.
+ Einleitung in die Theorie der Bexsel'xchen Funktioiien, i. (Bern, 1898), p. 34 et seq.
% Nielsen, as in the case of other functions, writes the number indicating the order as an
index, thus Y v (z), Handbuch der Theorie der Cylinderfuiilctionen (Leipzig, 190-1), p. 11. There
are obvious objections to such a notation, and we reserve it for the obsolete function used by
Neumann ( 3*58).
See, e.g. Journal fUr Math. cxiv. (181)5), pp. 31 44, and other papers; also Die Theorie der
BesseVschen Funktioiien (Leipzig, HK)8).
I!
Proc. lloyul Sue. liv. (1893), p. 138, and Electromagnetic Theory, n. (Loudon, 1899), p. 255;
a change in sign has been made from his Electrical Papers, n. (London, 1892), p. 445.
11 A TreatUe on Bessel Functions (London, 1895), pp. G5 66.
;
Lommel, in his later work, used Neumann's function of the second kind (see 3'57), but
in his Studien iiber die BesseVschen Functionen (Leipzig, 1868), pp. 85 86, he used the
function
7rFM (*) + {,/,(+4) + log2} Jn (z),
where Yn (z) is the function of Weber. One disadvantage of this function is that the
presence of the term yjr (+) makes the recurrence formulae
for the function much more
complicated; see Julius, Archives Nterlandaises, xxviii. (1895), pp. 221 225, in this
connexion.]
The definition given by Heine* of the function of the second kind possesses
some advantages from the aspect of the theory of Legendre functions it ;
The use of Heine's function seems to have died out on the Continent many years ago
the function was occasionally used by Gray and Mathews in their treatise t, and they term
it On (z). In this form the function has been extensively tabulated first by AldisJ: and
Airey, and subsequently in British Association Reports, 1913, 1914 and 1916.
This revival of the use of Heine's function seems distinctly unfortunate, both on account
of the existing multiplicity of functions of the second kind and also on account of the fact
(which will become more apparent in Chapters vi and vn) that the relations between the
functions JH (z) and YH (z) present many points of resemblance to the relations between the
cosine and sine ; so that the adoption ||
of Jn (z) and On (z) as canonical functions is com-
parable to the use of cos and %ir sins as canonical functions. It must also be pointed
out that the symbol On (z) has been used in senses other than that just explained by at least
two writers, namely Heaviside, Proc. Royal Soc. liv. (1893), p. 138 (as was stated in 3*54),
and Dougall, Proc. Edinburgh Math. Soc. xvm. (1900), p. 36.
Note. An
error in sign on p. 245 of Heine's treatise has been pointed out by Morton,
Nature, lxiii. (1901), p. 29 ; the error is equivalent to a change in the sign of y in formula
3*51 (3) supra. It was also stated by Morton that this error had apparently been copied
by various other writers, including (as had been previously noticed by GraylT) J. J. Thomson,
Recent Researches in Electricity and Magnetism (Oxford, 1893), p. 263. A further error
noticed by Morton in Thomson's work seems to be due to a most confusing notation employed
by Heine for on p. 245 of his treatise Heine uses the symbol
; to denote the function K
called - \n J'q in this work, while on p. 248 the same symbol denotes - \n (Y -iJ ). K
if we multiply these by cot vir and cosec vrr, and then subtract, we have
whence (3) follows at once. Equation (4) is derived in a similar manner from
the formulae
By addition and subtraction of (3) and (4) we obtain (2) and (1).
The formulae are, so far, proved on the hypothesis that v is not an integer
but since Yv (z) and its derivatives are continuous functions of v, the result of
proceeding to the limit when v tends to an integral value w, is simply to
replace v by n.
Again, the effect of multiplying the four equations by ire rri sec vir, which
is equal to 7re " 1,,ri sec {y l)7r, is to replace the functions
(
Y by the functions
Y throughout.
In the case of functions of integral order, these formulae were given by Lommel,
Studien iiber die Bessel'schen Functionen (Leipzig, 1868), p. 87. The reader will find it
instructive to establish them for such functions directly from the series of 3-52.
Neumann's investigation connected with the formula (4) will be discussed iu 3*58.
;
The second solution of Bessel's equation for functions of order zero being
known to contain logarithms, Neumann assumed as a solution the expression
J (z) log z + w,
where w is a function of z to be determined.
If this expression is to be annihilated by V we must , have
V w = -V {J (*)log*}
= -2zJ '(z).
But, by 212 (11),
- 2z ^ (z) = 2z J, (s) = 8 2 {-T~'nJ2n (*)
and so, since V Jm (z) 4>n2 JM (z), we have
= 2V 2(-y-iJm (z)/n;
=i
the change of the order of the operations 2 and V is easily justified.
2 2(-y->Jm (z)/n,
n-l
and therefore Neumann's function F (0)
(z), defined by the equation
eixcoa
a
<
log (x sin 2 <o) dm
h
isa solution of Bessel's equation for functions of order zero and argument x ;
2 f*
- cos (z sin 0) log . (4iz cos2 0) dd
was identified by Neumann f with the function F (z) and the analysis by ,0 >
;
and so, if we assume that the order of summation and integration can be
changed, we deduce that
from this result combined with Parseval's integral ( 2 2) and the definition of
F (z), we at once obtain the formula
,0
The change of the order of summation and integration has now to be examined,
because 1n~ l cos 2n0 is non-uniformly convergent near 6=0. To overcome this difficulty
we observe that, since 2 ( - ) n J.in {z)jn is convergent, it follows from Abel's theorem that
%
Now, since a is less than 1, 2 (on cos 2n6)/n does converge uniformly throughout the
range of integration (by comparison with 2a"), and so the interchange is permissible that ;
is to say
n
2 /"*""
a .a cos2n8 . A 2 /"i* . a" cos 2nd
dOn
/ , ,
- 2 / eos(cos0) d6**>- I cos(cos#) 2
=1 Jo n n J n-1 n
1 lh"
= -- / cos ( cos d) log (1 - 2a cos 26 + a2 ) flW.
n Jo
Hence we have
(")* (*' *""
=_ lim i | cos (a cos d) log (1 - 2a cos 20 a2 ) <J9.
n-l M o-*.l-0 " y o
We now proceed to shew that*
lim / cos(*cos0){log(l-2acos20-M2 )-log(4asin 2 0)}c?0 = O.
a-*-l-0 J
I/,o
**
^ {log (1 - 2a cos 26+ a8 ) - log (4a sin2 6)} d6
f
Jo
-A i
/**"
\
( ~
|-22
a"cos20 .
+log (l/a>- 2
. . . ,
log (2 sin
. J
6)\d6
io I *i " J
= J*iUog(l/a),
term-by-term integration being permissible since a< 1. Hence, when a< 1,
/"*
- 2a cos 26 + a2 ) - log (4a sin2 0)} dd < Jtt4
I I
= -1 /"*"
cos (z cos 0) log (4 sin 2 8) dd,
.
The differential equation considered by Stokes \ in 1850 was -rf +- -r- - m2 y =0, where
m is a constant. This is Bessel's equation for functions of order zero and argument imz.
Stokes stated (presumably with reference to Poisson) that it was known that the general
solution was
/itr
{C+ D log (z sin 2 0)} cosh (mz cos 6) d&.
o
* The value of this limit was assumed by Neumann,
t If 2 is real, A l\ if not, A < exp {
1 (z) j
}
J Trans. Camb. Phil. Soc. ix. (1856), p. [38]. [Mathematical and Physical Papers, in. (1901),
p. 42.]
' +
It is easy to see that, with Neumann's notation, the value of the expression on the right
is
\n {C- D log (4m)} J (imz)+\nD F() (imz).
The expression was expanded into a series by Stokes ; it is equal to
* m 2" z2n f in
bn (C + D log z) J (imz) + 2D 2 ,_ ., I cos2" 6 log sin 6 dd,
=o v* n )
J o
and, by integrating by parts, Stokes obtained a recurrence formula from which it may be
deduced that
The Bessel function of the second kind, of integral order n, was denned by
Neumann* in terms of F <0)
(z) by induction from the formula
(1) z
dYi
^
which is a recurrence formula of the same type as 2*12(4). It
- nY^ (z) = - zYW (*),
is evident
from this equation that
(2) FW(jr)-(^)(^)*FW().
and, if we apply the operator f -j- to this equation n times, and use Leibniz'
ZCLZ
theorem, we get
and so
<3> ' &T F '' {l) + (2M + 2) (a)"" 7 " w (a)"
7 " (J) " *
*"
{zlzf ^^ {Z)] + (2n + 2)
{zlz)
{z
~ nYW
{Z)] + ^^ {Z) = -
This equation is at once reducible to
(4) V,Fw(5)-0,
and so F (w)
(s) is a solution of Bessel's equation for functions of order n.
so that
dF(n+I >(*) n+ 1
+ -^-
t
F<+-> (z) - F (*) = 0,
Tz z
whence we obtain another recurrence formula
d7 (Z)
(5) z
2 + nYm (*) = zY- (z).
When we combine (1) with (5) we at once deduce the other recurrence
formulae
dF< "'^
(7) F<"--> (*) - F<+ -> (*) = 2 .
as
Consequently F (w
(*) satisfies the same recurrence formulae as Jn (z), F (s)
and Yn (s). It follows from 357 (2) that
<
The generalisation of the formula 357 (1) has been given by Neumann*;
it is
i
|_ -^
On ro-i
i
, r /\
m (n + m) ^
v.
where sn
111
= T + B + 5+ ...+-,
1
s = 0.
,=i
1 Z o w
To establish this result, we first define the functions Ln (z) and Un (z) by
the equations
,__ -ig.
n-l 9n-m-i *> j r /_\
(2) Z. W.J .(.) log
5
We shall prove that Ln (z) and Un (z) satisfy the recurrence formulae
* Theorie der Bessel'schen Funetianen (Leipzig, 1867), p. 52. See also Lommel. Studien liber
die Bessel'schen Functionen (Leipzig, 1868), pp. 82 84 ; Otti, Bern Mittheilungen, 1898, pp. 34 35
and Haentzschel, Zeitschrift fiir Math, und Phys. xxxi. (1886), pp. 25 33.
72 THEORY OF BESSEL FUNCTIONS [CHAP. Ill
It is evident that
d_ [
L n (z) \ . d (JJs)) Mz) _ 2 -->.nl _d ( Jm (z) )
dz\
dz zn )
S dz\ zn j z n+1 m. (n-m).w!^Vsn - M j
Jn (z) n - r"- 1
1 _ ... n z} 2 .! (.. .Jm (z) J +i(z)\l vl
z1
zn
'
and the first part of (4) is proved. To prove the second part, we have
d [Un(z)]_
n
d [Jn (z)) | (-y(n + 2m) d J^fr) f )
J ( )m
= - * -=- + 7n *
(z) 1
~ -^ [mJr*^ (z) - (n + m) J1l+M (z)}
=i (n +
*7n + l(*)
n
and the second part of (4) is proved. It follows from 358 (2) that
zn dz\ zn j'
g-<'>-.
v.i(;v)- -
s
Note. It will be found interesting to establish this result by evaluating the coefficient
The Poisson-Neumann formula of 3-571 for F() (z) was generalised by Lommel,
Sttidien ilber die BesseVschen Functionen (Leipzig, 1868), p. 86, with a notation rather
different from Neumann's; to obtain Lommel's result in Neumann's notation, we first
observe that, by differentiation of Poisson's integral for Jv (*), we have
and
^^-Jy(z)logz=
so, from 3*582 (3),
F ^^^ f^ coa(z Sme)coH^B{log(ico^0)-ylr(u+i)}dB,
Since Jp {z) and F (z) satisfy the same recurrence formulae ( 3*2, 3'56),
in which the functions enter linearly, and since the functions of the third kind
* Ofversigt over det K. Danske Videnskabemes Sehkabs Forhandlinger, 1902,. p. 125. Hand-
bitch der Theorie der Cylinderfunktionen (Leipzig, 1904), p. 16.
t Nielsen uses the symbols Hf(z) t
Hv
2 (z).
74 THEORY OF BESSEL FUNCTIONS [CHAP. Ill
are linear functions (with constant coefficients) of Jv (z) and Y v (z), it follows that
these same recurrence formulae are satisfied by functions of the third kind.
Note. Bayleigh on several occasions, e.#. PAt7. J/ogr. (5) xliii. (1897), p. 266 (6) xiv. ;
(1907), pp.350359 [Scientific Papers, iv. (1904), p. 290; v. (1912), pp. 410418], has used the
symbol Dn (z) to denote the function which Nielsen calls \ triH
(2)
(s).
that some of the formulae are simply the definitions of the functions on the
left.
sm w 2t
H {x)
(z) = J-M-c-iJriz) = Y. v (z)-e-"- i Y v (z)
(5)
" t sin vtt sin vir '
The functions of the second and third kinds will now be defined for all
values of the argument by means of the equations 3"54 (1), 36 (1); and
then the construction of the following set of formulae is an easy matter:
+ 2e J^T
8
(6)
v ' Hf (zem) - a* H
v v 7
(2)
"
(z)
v ' sini/7r
7r
JyvV(z)'
Of these results, (3) was given by Hankel, Math. Ann. vin. (1875), p. 454, in the special
case when m=\ and v is an integer. Formulae equivalent to (5) and (6) were obtained by
Weber, Math. Ann. xxxvu. (1890), pp. 411, 412, when m=l see 6-11. And a memoir ;
by Graf, Zeitschrift fur Math, und Phys. xxxvm. (1893), pp. 115120, contains the general
formulae.
_2
TTz'
This result is established on the hypothesis that v is not an integer; but con-
siderations of continuity shew that
(1) M {J, (*), *,(#)} -2/(,
whether v be an integer or not. Hence Jv (z) and Y
v (z) always form a funda-
mental system of solutions.
It is easy to deduce that
Various formulae connected with (1) and (3) have been given by Basset, Proc. London
Math. Soc. xxi. (1889), p. 55 they are readily obtainable by expressing successive differ-
;
An associated formula, due to Lommel* Math. Ann. iv. (1871), p. 106, and Hankel,
Math. Ann. vm. (1875), p. 458, is
(
12 ) JAz)Yv ^{z)-Jv ^{z)Yv {z)=-~.
irz
This is proved in the same way as 32 (7).
It follows that, when v is not an integer, the functions Iv (z) and /_(*) form
a fundamental system of solutions of equation (1).
It may be verified, by the methods of 3*5, that Kn (z) is a solution of (1) when
the order v is equal to n.
The definition of n (z) is due to Basset, Proc. Camb. Phil. Soc. vi. (1889), p. 11, and
K
the infinite integrals by
his definition is equivalent to that given by equations (4) and (5) ;
which he actually defined the function will be discussed in $ 614, 61 5. Basset subse-
(Cambridge, 1888),
quently modified his definition of the function in his Hydrodynamics, n.
, Ai 1 rdl-y(z)
. 37, (*)1
pp. 1819, and his final definition is
equivalent to c ^Ti^ '
cV~J v -n
In order to obtain a function which satisfies the same recurrence formulae as 7 (z),
Gray and Mathews in their work, A Treatise on Bessel Functions (London, 1895), p. 67,
'
2|_ ov Cv Jv= n
of this definition to functions of unrestricted order is by the
The only simple extension
formula
K* (2) = \ir cot vrr{I- v (z) - 7 ()},
Proc. London Math. Soc. xxx. (1899), p. 167.
3*71] BESSEL FUNCTIONS 79
(cf.Modern Analysis, 1771) but this function suffers from the serious disadvantage that
it vanishes whenever 2 is an odd integer. Consequently in this work, Macdonald's
function will be used although it has the disadvantage of not satisfying the same
recur-
rence formulae as Iv (z).
An inspection of formula (8) shews that it would have been advantageous if a factor %n had
K
been omitted from the definition of v (*) but in view of the existence of extensive
;
tables
of Macdonald's function it is now inadvisable to make the change, and the
presence of the
factor is not so undesirable as the presence of the corresponding factor in Schlafli's
function
( 354) because linear combinations of /() and K
v (z) are not of common occurrence.
We shall now give various formulae for Iv (z) and v (z) analogous to K
those constructed in 3'23*6 for the ordinary Bessel functions. The proofs
of the formulae are left to the reader.
(1) /_, (,) - J+1 (,) = ^ I, (z), K. v x (z) - Kv+l (z) = - ^ Kv (z),
(2) /_, (z) + Iv+l (z) = 2/; (*), K^ (z) + Kv+1 (z) = - 2KJ (z),
(3) zi; (z) + vlv (z) = zlv_ x (z), zKJ (z) + vKv (z) = - z Kv . (z), x
(4) zlj (z) - vlv (z) = zlv+1 (z), zK: (z) - vK v (z) = - zKv+1 (z),
^L) {z"Iv (.)} = z^Iv_ m (z), (-*-) [z *Kv (z)} = {-r*K (z),
K }
\zdzj \ z" j
*'+ '
The following integral formulae are valid only when R(v + \) > :
(9) Iv (z) =
r y ^
I
cosh (z cos 0) sm^BdO
'
- i>?|rr (i) L (1 " t2y h cosh {zt) dt
<**)'
- ^os s i n2 , dQ
=
r^r|^)/rcosh {z cos d) sin2v0d0
+^ >
r f r!( W -r)!(2^J'
(12) /f + , (.) =
y r- 2
o;T ^-^ r ,
(13) *,-)'.-.
- ^ -
x (-)"(- m- 1)1
n \ ^.W-jS, s'
(15)
m!(w-m
eo /1 2 \n+am
+ <")n+1 2 ,
,
'
v. {log (**) - W (m + - & 1) (n + m +1 )],
^o (*) = --
*
(16) f
e*
008 '
{log (2* sin2 0) + 7} dO,
The integral involved in (16) has been discussed by Stokes (cf. 3"572).
The integrals involved in (9) and the series in (14) were discussed by Riemann in his
memoir "Zur Theorie der Nobili'schen Farbenringe," Ann. der Physik und Chemie, (2) xcv.
(1855), pp. 130 139, in the special case in which v=0; he also discussed the ascending
power series for 7 (2).
The recurrence formulae have been given by Basset, Proc. Camb. Phil. Soc. vi. (1889),
pp. 2
19; by Macdonald, Proc. London Math. Soc. xxix. (1899), pp. 110115; and by
Aichi, Proc. Phys. Math. Soc. of Japan, (3) 11. (1920), pp. 819.
Functions of this type whose order is half an odd integer, as in equations (10) and (12),
were used by Hertz in his Berlin Dissertation, 1880 [Oes. Werke, 1. (1895), pp. 7791]
and he added ^et another notation to those described in 3*41.
:
3'8. Thomson's functions ber (z) and bei (z) and their generalisations.
A class of functions
which occurs in certain electrical problems consists of
Bessel functions whose arguments have their phases equal to \tr or f ir.
The functions of order zero were first examined by W. Thomson * ; they
may be defined by the equation f
(1) ber (x) + i bei (x) = J (xi Ji) =I (x ^/i),
where x is real, and ber and bei denote real functions. For complex argu-
ments we adopt the definitions expressed by the formulae
(2) ber (z) i bei (z) -J (zi Vi *) = h (* V *')
Hence we have
Extensions of these definitions to functions of any order of the first, second and
third kinds have been effected by Russell:}: and Whitehead.
The functions of the second kind of order zero were defined by Russell by
a pair of equations resembling (2), the function IQ being replaced by the
function thus K ,
(8) ker (z) = \tr hei (z), kei (z) \ir her (z),
in consequence of 3'7 (8).
* Presidential Address to the Institute of Electrical Engineers, 1889. [Math, and Phys.
Papers, in. (1890), p. 492.]
f In the case of functions of zero order, it is customary to omit the suffix which indicates
the order.
t Phil. Mag. (6) xvii. (1909), pp. 524552.
Quarterly Journal, xui. (1911), pp. 316 342.
IIIntegrals equal to ker (z) and kei (z) occur in a memoir by Hertz, Ann. der Physik und Chemie,
(3) xxn. (1884), p. 450 \_Ges. Werhe, i. (1895), p. 289].
:
(10) kei (z) =- log (\z) . bei (*) - \ir ber (*)
It has also been observed by Russell that the first few terms of the expansion of
ber2 ()+bei 2 () have simple coefficients, thus
but this result had previously been obtained, with a different notation, by Nielsen (cf.
;
m
5-41) the coefficient of ($z)* in the expansion on the right is l/[(m !) 2 (2i) !]. .
(12) ber_ (z) cos vtr . ber (z) sin vrr . [hei (z) bei (z)],
(13) bei_ {z) = cos vir . bei (z) + sin inr . [her (z) ber (z)],
(14) her_(,z) = cos vir . herv (z) sin vtr . hei (z),
(15) hei_ v (z) = sin vir . her (z) + cos vir . hei v (z).
The reader will be able to construct the recurrence formulae which have
been worked out at length by Whitehead.
The functions of order unity have recently been examined in some detail
by B. A. Smith f.
Functions which satisfy only one of the two formulae are also discussed by
Sonine in his elaborate memoir a brief account of his researches will be given
;
in Chapter x.
Following Sonine we shall call any function ^(z), which satisfies both of
the formulae, a cylinder function. It will now be shewn that cylinder functions
are expressible in terms of Bessel functions.
u .
= -*#,(*).
that is to say
e**>Jm {kpf *
r m<t>
sin ^
(cf. 4-8 and Modern Analysis, 18*5).
84 THEORY OF BESSEL FUNCTIONS [CHAP. Ill
Some writers* following Heine t who called Jn (z) a Fourier- Dessel function, call Jn {z)
a Fourier function.
Although Bessel coefficients of any order were used long before the time of Bessel
(cf. 1-3, 1*4), it seems desirable to associate Bessel's name with them, not only because
it has become generally customary to do so, but also because of the great advance made by
Bessel on the work of his predecessors in the invention of a simple and compact notation
for the functions.
Bessel's name was associated with the functions by Jacobi, Journal fiir Math. xv.
(1836), p. 13 [Ges. Math. Werke, vi. (1891), p. 101]. "Transcendentium 7 naturam varios-
fc
celeberrima."
* E.g. Nicolas, Ann. Sci. de VEcole norm. sup. (2) xi. (1882), supplement,
f Journal fiir Math. lxix. (1868), p. 128. Heine also seems to be responsible for the term
cylinder function.
CHAPTER IV
DIFFERENTIAL EQUATIONS
4*1. Daniel Bernoulli 's solution of Riccati's equation.
The solution given by Bernoulli* of the equation
(!) -^ = azn + by 2
consisted in shewing that when the index n has any of the values
/9\
\*) n = 4m
2m 1
(4) v =- n - 4.
* Exercitationes quaedam muthematicae (Venice, 1724), pp. 7780 Acta Eruditorum, 1725, ;
pp. 473 475. The notation used by Bernoulli has been slightly modified and in this analysis ;
= Z, y = -
+ !-"*- Y-
[Note. The substitutions are possible because 1 is not included among the values of
n. The factor n+ 1 was not inserted by Bernoulli the effect of
in the denominator ; its
presence is that the transformed equation is more simple than if it were omitted.]
that is
z -.- = a + v + bv2 ;
dz
unci this is an equation with the variables separable.
Hence, in this limiting case, Riccati's equation is still soluble by the use
of elementary functions.
4-114'13] DIFFERENTIAL EQUATIONS 87
This solution was implicitly given by Euler, Inst. Cole. Int. n. (Petersburg, 1769), 933,
dz^
+ z*
-U >
(2)
cT
+r,2 ~ &zn~^ = '
and the soluble cases are those in which 1/q is an odd integer.
Define a new variable w by the equation
/A \ d 2
w _ dw
(4) -^ + 2cz*-> -^ + (q - 1 ) cz*--w = 0.
A solution in series of the last equation is
00
w = z-m-v 2 A r z~Qr ,
provided that'
Ar_+l _ (2qr + q + 1) (2qr +q- 1)
Ar 8qc(r + l)
* Nov. Comm. Acad. Petrop. vm. (17601761) 363; and
[1763], pp. ix. (17621763)
[1764], pp. 154169.
and so the series terminates with the term A m z~ qm if q has either of the values
l/(2m + 1); and this procedure gives the solution* examined by Bernoulli.
The general solution of Riccati's equation, which is not obvious by this method, was
given explicitly by Hargreave, Quarterly Journal, vn. (1866), pp. 256258, but Hargreave's
form of the solution was unnecessarily complicated; two years later Cayley, Phil. Mag. (4)
xxxvi. (1868), pp. 348351 [Collected Papers, vn. (1894), pp. 912], gave the general solu-
tion in a form which closely resembles Euler's particular solution, the chief difference between
the two solutions being the reversal of the order of the terms of the series involved.
Cayley used a slightly simpler form of the equation than (2), because he took constant
multiples of both variables in Riccati's equation in such a way as to reduce it to
fc W-^~ = 0.
2
(5)
dz
given in 413 and we shall now explain Cayley 'sf method of solving
(2);
this equation, which is to be regarded as a canonical form of Riccati's
equation.
(1)
*?_<-,*-.,,_<);
differentiations with
where d and C 2 are arbitrary constants and primes denote
respect to z.
i
?-l
q(q-l)
^^q(q-l)2q(2q-l)
czq+ -^*_
(g-l) (3g-l)
Vx ^
,
*
/,> c iq
_ (y-l)(3g-l )(og-l)
q{q-l)2q(2q-l)Sq(Sq-l)
*
and we take Ux to be exp (czq/q) multiplied by this series.
L ?(9-l) ?(0-l)2$(2gr-l)
- l)(5i- 1)_
+ _il_- 1JJ3 ?
"I
Z +
q(q- l)2q(2q- l)Sq(Sq-lf -J'
and both of these series terminate when q is the reciprocal of an odd positive
integer. Since the ratio Ci : J72 is the exponential function exp (2oz9 /<7)
multiplied by an algebraic function of zv, it cannot be a constant ; and so
U U
lt % form a fundamental system of solutions of (1).
If q were the reciprocal of an odd negative integer, we should write
equation (1) in the form
*-"W.)^(./.)-a
whence it follows that
d
V F = z exp( + + l)(Sq + l)
(q
lt 2 afljq) 1 -?-;. czfl + CV2 ,..] .-.
q(q + l)2q(2q + l)
The series which have now been obtained will be examined in much greater
detail in 4'4 4'42.
The reader should have no difficulty in constructing the following solutions of Riccati's
equation,when it is soluble in finite terms.
It is to be noticed that the series Di, U^ (or Vu V2 as the case may be) are supposed ,
to terminate with the term before the first term which has a zero factor in the numerator
see 4-42 and Glaisher, Phil. Trans, of the Royal Soc. clxxii. (1881), p. 773.
w. B. F. 4
90 THEOBY OF BESSEL FUNCTIONS [CHAP. IV
Among who have studied equation (1) are Kummer, Journal fur Math. xn.
the writers
(1834), pp.144147, Lobatto, Journal fur Math. xvn. (1837), pp. 363371, Glaisher (in
the memoir to which reference has just been made), and Suchar, Bull, de la Soc. Math, de
France, xxxn. (1904), pp. 103116; for other references see 43.
The reader when q0, the equation (1) is homogeneous and imme-
will observe that
diately soluble;and that the second order equation solved by James Bernoulli ( 1*1) is
obtainable by taking q2 in (1), and so it is not included among the soluble cases.
Zt
This is easily reduced to the form of 4*13(2) by taking t~a/a as a new
independent variable.
To solve the equation, Schlafli wrote
at
and arrived at the equation
t
m
Iff F(a,t)=X
1=9 w!r(o + m + l)'
the general solution of the equation in y is
Ct^Fia + 1, t) + c2 F{- a - 1, t)
u
^F^^ + cJ-HFi-a,*)
The connexion between Riccati's equation and Bessel's equation is thus
rendered evident ; but a somewhat tedious investigation is necessary ( 4*43)
to exhibit the connexion between Cayley's solution and Schlafli's solution.
, a 1 a2 1 a3
1+- + H-
.
Elements de Oeometrie (Paris, 1802), note 4, in the course of his proof that is irrational.
Later the function was studied (with a different notation) by Clifford; see a posthumous
fragment in his Math. Papers (London, 1882), pp. 346 349.
* Ann. di Mat. (2) i. (1868), p. 232. The reader will see that James Bernoulli's solution in
be associated with Schlafli's solution rather than with Cayley's solution,
series ( 1*1) is to
A solution of Riccati's equation, which involves definite integrals, was given by Murphy,
Trans. Camb. Phil. Soc. in. (1830), pp; 440 443. The equation which he considered is
y =4* h~^[(f> (A) exp (*i//A) + < (1/A) exp (A*V)] dh,
f
where w
<6(A)=e*A-
^ /
Jo
e~ h ha
- 1 dh= I ,
n =oa(a
^-
+ l)(a +\,
t
,.
2)..,(a + n)
-,
If 1/A be written for A in the second part of the integral, then the last expression given
for y reduces to wit multiplied by the residue at the origin of h~* < (A) exp (* 1/0/A), and the
connexion between Murphy's solution and Schlafli's solution ( 4*15) is evident.
An was published by Challis, Quarterly Journal, vn. (1866), pp. 51
investigation 53,
which shewed how to connect two equations of the type of 4*13 (2), namely
in one of which l/q is an odd positive integer, and in the other it is an odd negative
integer. This investigation is to be associated with the discovery of the two types of
solution given in 4*14.
which is easily transformed into an equation of Riccati's type by taking - + and z*u as i
new variables, was investigated by Rawson, Messenger, vn. (1878), pp. 69 72. He trans-
formed it into the equation
dv
-jt y+
a+a
bz m
-a 2 +a
y -cz" =Q,
by taking bu=czaly; two such equations are called cognate Riccati equations. A somewhat
similar equation was reduced to Riccati's type by Brassinne, Journal de Math. xvi. (1851),
pp. 255256.
The connexions between the various types of equations which different writers have
adopted as canonical forms of Riccati's equation have been set out in a paper by Greenhill,
Quarterly Journal, xvi. (1879), pp. 294 298.
* Greenhill, Engineering, cvh. (1919), p. 334 Phil. Mag. xxxvm. 501528
; (6) (1919), pp.
see also Engineering, cix. (1920), p. 851.
92 THEORY OF BESSEL FUNCTIONS [CHAP. IV
(1) ^ = P + Qy + Rf,
where P, Q, R are any given functions of z. This equation was investigated
by Euler*. It is supposed that neither P nor R is identically zero for, if ;
It was pointed out by Enestrom, Encyclop4die des Sci. Hath. n. 16, 10, p. 75, that a
special equation of this type namely
nxxdx nyydx-k-xxdyxydx
was studied by Manfredius, De constructions aequationum differentialum primi gradus
(Bologna, 1707), p. 167. "Sed tamen haec eadem aequatio non apparet quomodo construi-
bilis sit, neque enlni videmus quoinodd illam integremus, nee quomodo indeterminatas ab
invicem separemus."
The equation (1) is easily reduced to the linear equation of the second
order, by taking a new dependent variable u defined by the equation f
(2) --*--"
The equation then becomes
/o\ d2 u \r\ 1 dR) du _ _
(5) u = e^' dz ,
W
(6) ^^.fly-y.,
y '
dz p p**
which is of the same type as (1). The complete equivalence of the generalised
Riccati equation with the linear equation of the second order is consequently
established.
The equations of this section have been examined by Anisimov, Warschau Univ. Nach.
1896, pp. 1 33. [Jahrbuch iiber die Fortschritte der Math. 1896, p. 256.]
Nov. Comvi. Acad, retrop. vin. (1760 1761) [1763], p. 32 ; see also a short paper by W. W.
*
Johnson, Ami. of Math. in. (1887), pp. 112 115.
t This is the generalisation of James Bernoulli's substitution ( 1-1). See also Euler, Inst.
Calc. Int. n. (Petersburg, 1769), 831, 852, pp. 88, 104.
y
g-P+fc + JV.
and write y = y + 1/v. The equation in v is
dv
^ + (Q+2Ry )v + R = 0,
y-yi
The result of substituting (y x w y )/(w 1) for y in the equation is
y<>-yi *? ,
w d}h !_ dy*_ P >Q yw- .vo + p (yxw - y ,
and, when we substitute for (dyjdz) and (dy /dz) the values P + Qy + Ry*
1
wTz
=Ry- Ry^
so that w = c exp {f(Ry Ry ) dz],
x
yjrJf.o =l y*-y<>
y-yi C '
y.2 -y x
'
(yi-y*) (y*-y*)
(yi-y*)(y*-y*)
is independent ofz; for it is equal to
(0 -(7# )(Ci
1
-(Q"
In spite of the obvious character of this theorem, it does not seem to have
been noticed until some forty years ago*.
primitivo of the Riccati equation), has been studied by Vessiot, Ann. de la Fac. des Sci. de
Toulouse, ix. (1895), no. 6 and by Wallenburg, Journal fiir Math. cxxt. (1900), pp. 210 217-
and Comptes Rendus, cxxxvii. (1903), pp. 1033 1035.
* Weyr, Abh. biihm. Ges. Wiss. (G) vm. (18751876), Math. Mem. i.
p. 30 ; Picard, Ann. Sci.
de VEcole norm. sup. (2) vi. (1877), pp. 342 343. Picard's thesis, in which the result -is con-
tained, is devoted to the theory of surfaces and twisted curves a theory in which Riccati's
equation has various applications.
t Nouv. Ann. de Math. (4) n. (1902), pp. 529545.
4*3] DIFFERENTIAL EQUATIONS 95
The equations which we are now about to investigate are derived from
Bessel's equation by elementary transformations of the dependent and inde-
pendent variables.
(i)
p -*,.BiE
dz*t z
where c is an unrestricted constant. The equation is of frequent occurrence
in physical investigations, and, in such problems,
p is usually an integer.
The equation has been encountered in the Theory of Conduction of Heat
and the
Theory of Sound by Poisson, Journal de VlZcole Polytechnique,- xn. (cahier
19), (1823),
pp. 249403; Stokes, Phil. Tram, of the Royal Soc. 1868, pp. 447464 [Phil. Mag.
(4)
xxxvi. (1868), pp. 401421, Math, and Phys. Papers, iv.
(1904), pp. 299324]; Rayleigh,
Proc. London Math. Soc. iv. (1873), pp. 93103, 253283
[Scientific Papers, i. (1899)',
pp. 138, 139]. The special equation in which p=2 occurs in the Theory of the Figure of
the Earth; see Ellis, Camb. Math. Journal, n. (1841), pp. 169177, 193201.
, d? (uz-i) d (uz-i) ,
dz*
CV U '
z dz
of which the general solution is
W v = zP + i<@pH (ciz).
* See Plana, Mem. della R. Accad. delle Sci. di Torino, xxvi. (1821),
pp. 519538, andPaoli
Mem. di Mat. e di Fis. della Soc. Italiana delle Sci. xx.
(1828), pp. 183188.
t This was known to Plana, who studied equations and in the paper to
(1) (5) which reference
has just been made.
Equation (3), which has been studied in detail by Bach, Ann. Sci. de Vjtcole norm. sup.
(2) in. (1874), pp.
47 68, occurs in certain physical investigations; see L. Lorenz, Ann.
der Physik und Chemie, (2) xx. (1883), pp. 1 21 [Oeuvres Scientifiques, I. (1898), pp. 371
396]; and Lamb, Hydrodynamics (Cambridge, 1906), 287 291. Solutions of equation (3)
in the form of continued fractions (cf. 5*6, 9 '65) have been examined by Catalan, Bulletin
(5) Jr-c^-^o,
and its solution is
d*R
__ _ p2 /2= n(n + l) n
D _J__i
.,
Rf (Poisson),
S-*-*^* (Wisher).
* Phil. Tram, of the Royal Soc. clxxii. (1881), pp. 759 828 ; see also a paper by Curtis, Cam-
bridge and Dublin Math. Journal, ix. (1854), pp. 272 290.
t Zeitschriftfiir Math, und Phy*. xxxi. (1886), pp. 25 33.
% Proc. Camb. Phil. Soc. iv. (1883), pp. 6573.
:
&y 2i/-l dy n
.
/i\
(1)
i?- s + *-
,
"(2) y- *"#,(*).
Lommel proceeded by direct transformations to construct the equation whose
at
general solution is gfi"~ i^v (y^), where a, #, 7 are constants. His result,
which it will be sufficient to quote, is that the general solution of
is
The solution of (3) was given explicitly by Lommel in numerous special cases. It will
be sufficient to quote the following for reference
(7) z
S+ (1 -" )
S + 5"*=0; *- + V,W')-
(8) 2
^+ (1 _ v) g_i tt=0; -**.{iV*).
* Studien liber die Bessel'schen Functionen (Leipzig, 1868), pp. 98 120 ; Math. Ann. in.
(1871), pp. 475487.
f Math. Ann. xiv. (1879), pp. 510586.
98 THEORY OF BESSEL FUNCTIONS [CHAP. IV
+ (2 ,-i)^ + 2^x>)_aGf) Tf ^
(*"<*)
+
l*'(*)
W i.l y= 0.
(H) ^W-^WfeWI'^W^ 1
.
v ;
ck <(*) dz |_4 (</>(^j 2<(*)
y =o
is
is
(22)
H + ^^ = 0; y-*^^>
The independent researches of Pearson proceeded on very similar lines
except that he started from Bessel's equation instead of from the modified
form of it. The reader will find many special cases of equation (17) worked
out in his paper.
A partial differentia] equation closely connected with (7) and (8), namely
o2 du du
0Z l
v ' VZ
n Ct
has been investigated by Kepinski, Math. Ann. lxi. (1906), pp. 397 405, and Myller-
Lebedeff, Math. Ann. lxvi. (1909), pp. 325330. The reader may verify that Kepinski's
formula
The special case of the equation when i/= 1 was also investigated by Kepinski, Bull,
int. de I' Acad, des Sci. de Cracoirie, 1905, pp. 198 205.
4'32. Malmsten's differential equation.
3+;-(^>
which is obviously a generalisation of Bessel's equation ; and it is a special case of ^ 4*31
(15).
To reduce the equation, Malmsten chose new variables defined by the formulae
We choose p and q so that this may reduce to the equation of 4*3 (1) considered by
Plana, and therefore we take
2pq-q + l+qr=0, (m + 2)q = 2,
so that p = - ir fan.
The equation then reduces to
d? L( + 2) 2+ 4* _P
*
Camb. and Dublin Math. Journal, v. (1850), pp. 180 182. The case in which =0 had been
previously considered by Malmsten, Journal fUr Math, xxxix. (1850), pp. 108 115.
100 THEORY OF BESSEL FUNCTIONS [CHAP. IV
By 4*3 this is intcgrable in finite terms if
(2) m + 2= + ^+(l-r)>}
+
The equation is also obviously integrable in the trivial cases A =0 and m= -2.
rt=0 n J
KPi)n \Pi)n \Pq)n
In particular,
I < g >" -
n=0W!(p)n
CO .71
=0!(p)n'
The functions defined by the first three series are called generalised hyper-
geometric functions.
It is evident that
* Math. Ann. xxxvi. (1890), p. 84 ; xxxvm. (1891), pp. 227, 586, 587. Cf. 4-15.
t Proc. London Math. Soc. (2) v. (1907), p. 60. The modification due to Barnes is the insertion
of the suffixes p and q before and after the F to render evident the number of sets of factors.
. :
d2 u
-j
dz*
c*u = p(pr + \)- u,
z*
and obtain relations between them, which will for the most part be expressed
in Pochhammer's notation.
z*+* . F, (p + | ;
\c*z-) ; z~p -oF^-p; \c*z*).
v
dz* dz z*
which are suggested by the fact that the functions e e* are solutions of the
original equation with the right-hand side suppressed.
Now, by direct multiplication of series, the two expressions on the left are
expansible in ascending series involving z?+1 , z^ zp+3 , .... And the expressions
on the right similarly involve tr*, z1 ~p, z-'P, .... Since none of the two sets of
powers are the same when 2p is not an integer, we must have
(1) *. ,F, (p + 1 ; 2p + 2 ;
- 2cz) = e~ n F .
x x (p + 1 ; 2p + 2 ; 2cz)
These formulae are due to Kummerf. When (1) has been proved for general
values of p, the truth of (2) is obvious on replacing p by p 1 in (1).
* It follows from 3*1 that a special investigation is also necessary when p is half of an odd
integer.
t Journal fllr Math. xv. (1836), pp. 138141.
102 THEORY OF BESSEL FUNCTIONS [CHAP. IV
When p has any of the values , f f ..., the solutions which contain z~p
, ,
When p has any of the values 0, 1, 2, ... a comparison of the lowest powers
of z involved in the solutions shews that (1) still holds; but it is not obvious
that there are no relations of the form
z-p J\(\-p\ \<?z*) = z-ve*\F (-p\ -2p; -2cz) + k zP+\F
x 1 + ; \&z*)
1 (p
^z-Pe-^F.i-p; -2p; 2cz) + k2 zr+* <t
F {p + %; \c*z%
1
which forms part of equation (1) of 441, is a particular case of the more
general formula due to Kummer*
(i) i^(; p; 0-Wp-; />; -),,
which holds for all values of o and p subject to certain conventions (which will
be stated presently) which have to be made when a and p are negative integers.
We first suppose that p is not a negative integer and then the coefficient of
n
in the expansion of the product of the series for e* and (p a p ) is & ; ;
mto(n - m)\ ml(p)m " Wp)n .V^*' {p " ")m (1 " P ~ n) "
nl(p) n
= ()
'
n\(p)n
if we first use Vandermonde's theorem f and then reverse the order of the factors
in the numerator; and the last expression is the coefficient of fn in X X (a; p; *). F
The result required is therefore established when a and p have general complex
values J.
* Journal fur Math. xv. (1836), pp. 138 141; see alsoJBach, Ann. Sci. de I'ticole norm. sup. (2)
in. (1874), p. 55.
t See, e.g. Chrystal, Algebra, n. (1900), p. 9.
$ Another proof depending on the theory of contour integration has been given by Barnes,
Tram. Camb. Phil. Soc. xx. (1908), pp. 254257.
.
When these factors have been cancelled, the series for X F X ( N ; p ; ) and
iFt (p + iV p ; ; )
are both continuous functions of p near p = M, where
M is any of the integers N, + 1, + 2, N N . . .
Hence we may proceed to the limit when p -*. if, and the limiting form
of (1) may then be written*
the last term in which the numerator does not contain a zero factor, while
the symbol 1 means that the series is to proceed normally as far as the term
~N
in % M and then it is to continue with terms in f Jf+1 f->/+2 ... the vanishing
, , , ,
+ ( - )M ~ N
N
MKM~Ay.
** +I ^ (^ " ^ + 1 ;
3/ + 2 ;
?).
and from this result combined with (2), (3) and (4) we deduce that
(5) ^ (- iV; - Jf 01 - rf
; ^ (JV- if; - if; - ) 1.
This could have been derived directly from (1) by giving p a (instead of o)
an integralvalue, and then making p tend to its limit.
* Cf. Cayley, Messenger (old series), v. (1871), pp.77 82 [Collected Papers, vm. (1895), pp. 458
462], aDd Glaisher, Messenger, vm. (1879), pp. 2023.
.
(6) 0*J X (p +1 ; 2p + 2 ;
- 2cz) = & (p + f; \&z%
which forms the remainder of equation (1) in 4*41, and which is also due to
Kummer*.
If we suppose that 2p is not a negative integer, the coefficient of (cz) n in
the product of the series on the left in (6) is
^ (-2r (p + ix
= (-)* 1 2m Q> + u(---2;>-i)n-w
m= o(n-m)\ml(2p + 2) m (2p + 2) n m=0 ml(n-m)l
n
Now 1 2 2W . Cm (^ + l)m (- n - 2p - l)n-, is the coefficient of tn in the
expansion of (1 2t)~p~ l
(1 t) n+2p+1 and , so it is equal to
1 f(o+) 1 />+)
even.
Hence it follows that
m
(cz)
%
2n
.n!(p + f) n
'
=o2
and this is the result to be proved.
When we make p tend to the value of a negative integer, N, we find by
the same limiting process as before that
It follows that
Ifwe change the signs of c and z throughout and add the results so obtained,
we find that
(7) 2 tfx (f - N icV) = e. ; ^ (1 - N; 2 - 2tf; - 2cz) ~1
the other terms on the right cancelling by a use of equation This is, of
(1)
J_y+ j (icz) in finite terms with a different notation.
course, the expression for
The equation
(1 >
'g+g + C + ^-o.
which is a generalisation of Bessel's equation for functions of order zero,
occurs in the theory of the reflexion of sound by a paraboloid. It has been
investigated by Sharpe* who has shewn that the integral which reduces to
unity at the origin is
where
(3) 1 = f
"cos (4 log cot \0) dd.
Jo
It is easy to see from this form of the integral that it converges for (complex)
values of A for which \I(A)\< 1, andf
2
C=- cosh hvA.
ir
The integral has been investigated in great detail by Sharpe and he has
given elaborate rules for calculating successive coefficients in the expansion of
y in powers of z.
A simple form of the solution (which was not given by Sharpe) is
y- to ,^(jTi^; 1; T 2iz).
* Messenger, x. (1881), pp. 174185 ; xn. (1884), pp. 6679 ; Proc. Camb. Phil. Soc. x. (1900),
pp. 101136.
t See, e.g. Watson, Complex Integration and Cauchy's Theorem (1914), pp. 64 65.
106 THEORY OF BESSEL FUNCTIONS [CHAP. IV
depends on the fact that cylinder functions exist for which the quotient
<$ (z)/^ v (z) is independent of z.
Each of the functions Jn (z) and Yn (z), of integral order, possesses this
property [ 2*31, 3*5]; and the functions of the third kind p
(z), H H {2)
(z)
v = n+\ and m = 2n + 1.
Hence if ^n denotes either Jn or Yn , we have
From this equation we obtain Lommel's result that the functions z* n Jn (y \Jz) t
7 = ic exp (riri/n). (r = 0, 1, 2, . .
.
, n- 1)
where 7 has any value such that 72n+1 = cm+1 e- (n+ * ),ri
, so that
has been discussed by Molins, MSm. de VAcad. des Sci. de Toulouse, (7) vin. (1876), pp. 167 189.
4-5] DIFFERENTIAL EQUATIONS 107
For some applications of these results, see Forsyth, Quarterly Journal, xix. (1883),
pp. 317320.
In view of (1), which holds when m is an integer, Lommel, Math. Ann. n.
(1870), p. 635,
has suggested an interpretation of a "fractional differential coefficient." Thus
he would
exp(yV) to mean 90o (yijz). The idea has been developed at some
JgJ
length by Heaviside in various papers.
a P M V
1 JL
2 1
2.
1 2
* % * 1
-1 ~\ 3 10
These four cases give the following equations and their solutions
(7) ~ S} = 4
{?
22tt; =^ 1
{^2 (2^)+^2(2tV^},
(9)
t S} =2 ,2 tt;
-^{^(~*) +*(-*)}.
These seem to be the only equations of Nicholson's type which are soluble with the aid
of Bessel functionsin the case fi
; 2, the equation (5) is homogeneous. Nicholson's general
equation is associated with the function
It has been seen ( 4*3) that the general solution of the equation is
(ae z
+ Pe-^ls/z,
where a and /3 are constants, it follows that the general solution of (1) may
be written
, / d y .
ae + &e-
6* c*
r
.
where a' = a/c, #'= -fife. This may be seen by differentiating CLeP+ffe-*2 once.
* Phil. Tram, of Royal Soc. clxzii. (1881), p. 813. It was remarked by Glaisher that
the
equation (3) is Earnshaw, Partial Differential Equations (London, 1871),
substantially given by
p. 92. See also Glaisher, Quarterly Journal, xi. (1871), p. 269, formula (9), and p. 270.
4 * 6] DIFFERENTIAL EQUATIONS 109
Note. A result equivalent to (2) was set by Gaskin as a problem* in the
Senate House
Examination, 1839; and a proof was published by Leslie Ellis, Camb.
Math. Journal, n.
(1841), pp. 193195, and also by Donkin, PhU. Trans, of the Royal Soc.
cxlvii. (1857),
pp. 4357. In the question as set by Gaskin, the sign of c 2 was changed, so that the
solu-
tion involved circular functions instead of exponential functions.
It is convenient to write
= <*- [<* + 2p - 2) (* + 2p - 4) (* + . . .
2) * -<*-*>]
.
= J-[( 23 lYJL~\
z^ 1
\_\ dz) z*-*]
'
L ~243 "350
JJ:;iZ S pp 240
(1876)> ' 349 ; and PhiL Tram of ' t,ie Rmjal Soc -
'
When we transform (2) and (3) with the aid of (4) and (6), we see that
the general solution of (] ) is expressible in the following forms
{1) u z
*-
~zp+A dz) >
l / -dy+WtP + fftr *
^
1
v
/Q
8 uss '
< >
**\*di)
The solutions of the equation
2
2pdv
-j- cH = 0,
d
--v .
- -E- ,
[(3) of 4-3], which correspond to (2), (3), (7) and (8) are
<9 >
-^Ub) i
(10) v = ^ +1 (aV* + 0'e-a ),
(-^J
/nx 1^, dya f + fier"
<
12 > = ^s) ?!
A different and more direct (7) is due to Boole, Phil. Trans, of the
method of obtaining
Royal Soc. 1844, pp. 251, 252 on ;
Equations (London, 1872), ch. xvn.
Treatise
Differential
IX. (1854), p. 281.
pp. 423425; see also Curtis, Cambridge and Dublin Math. Journal,
The solution (9) was first given by Leslie Ellis, Camb. Math. Journal, II. (1841), pp. 169,
193, and Lebesgue, Journal de Math. xi. (1846), p. 338; developments in series were
obtained from it by Bach, Ann. Sci. de I'jtcole norm, sup. (2) in. (1874), p. 61.
A transformation of the solution (9), due to Williamson, Phil. Mag. (4) xi. (1856),
pp. 364371, is
/3 IV
(13) v==c
^\dc'c) (
ae" + 3e ~
'
C
*)-
1 r) if)
This is derived from the equivalence of the operators - ~- , - g- ,
when they operate on
functions of cz.
is then manifest.
,
(II) If fr (z) denotes any function of order r, then any algebraic function
of functions of the types
/m+i ( z ) an algebraic function of one or more functions of the types lfm (z)
is
t
sfm ( z )> efm (z) as well as (possibly) of functions whose order does not exceed
m. Let us concentrate our attention on a particular function of one of the
three types, and let it be called 6, ^ or according to its type.
(I) We shall first shew how to prove that, if (1) has a solution of order
m + 1, then a solution can be constructed which does not involve functions of
the types 6 and S-.
For, if possible, let /, )l+ i (z) = F(z, 0), where F is an algebraic function of ;
and any function of z (other than itself) of order m+1 which occurs in F
is algebraically independent of 0.
/qx
{6)
*F / n -
F X K)
"
W+ 2 df (z) d* F
Tz*~ dz> fm (z) dz d6dz
% Null solutions are disregarded ; if u were of order less than , then - --^ would be of order
^
d-F(z,0 + c)
'-
.
-F(z,0 + c). x (z),
,
, . .
dF(z, + c) o-F(z, + c)
oc
'
dc> ' '"
where 4> involves transcendants (of order not exceeding + 1) which are m
algebraically independent of 0. But this is impossible because e Ae is not an
algebraic function of ; and therefore and e form a fundamental system F F
of solutions o/'(l).
Hence Fee is expressible in terms of F and F e by an equation of the form
FM = AF, + BF,
where A and B are constants. Now this may be regarded as a linear equation
in (with constant coefficients) and its solution is
It follows that, in so far as fm+x (z) involves functions of the types and ^,
it involves them linearly, so that we may write
Take any one of the terms in fm+l (z) which is of the highest degree, qua
function of lt 0.,, ...%,%, ..., and let it be
Ox {z)0t {z) ... P (z).* 1 (z) ... * Q (t).irPiQ (z).
Then, by arguments resembling those previously used, it follows that
"
d_d_ d_ d_ d_ _a_
/**.(*)
d0! d02
'"
d0p'd& 1 d*S 2
"*
S^J
is a solution of (1) ; i.e. yffp^iz) is a solution of (1).
But ^p t Q (z) is either a function of order not exceeding m, or else it is a
function of order m+1 which involves functions of the type and not of
the types and ^.
In the former case, we repeat the process of reduction to functions of lower
order, and in the latter case we see that some solution of the equation is an
algebraic function of functions of the type .
(II) We shall next prove that, whenever (1) has a solution which is a
transcendant of order greater than n, then it has a solution which involves
^ '-G(*,ce).x(),
4*71] DIFFERENTIAL EQUATIONS 115
so that
/a\
(6)
d*G(z,c%);
^ , ~.
-G(z,c&).
,
X (z) =
.
0.
ft
dG(z,S) ^ d*G(z,e) "
d@ *
a@ 2 '
In the former case G Q% A , and in the latter G has one of the values
where 4>, Q>lt <t>8 are functions of z of order jl+ 1 at most, any functions of
order fi+1 which are involved being algebraically independent of ; while
7 and $ are the roots of the equation
x (x - 1) - Ax - B = 0.
In any case, G either contains only by a factor which is a power of or else
G is the sum of two expressions which contain only in that manner. In the
latter case*,
G(z, c%)-c*G(z, 0)
(*)exp|2 7log*V>
* If $j is not identically zero ; if it is, then 4> 0* is a solution of the specified type.
2
116 THEORY OF BESSEL FUNCTIONS |CHAP. IV
(!) ;i-X<*)
[in which x ( z ) is f order n] has a solution which is an elementary tran-
scendant of order greater than n, then it must have a solution of the form
where /* ^ n. If the equation has more than one solution of this type, let a
solution for which /* has the smallest value be chosen, and let it be called ,.
Liouville's theorem, which we shall now prove, is that, for this solution, the
order of d (log uj/dz is equal to n.
Let
d log i
= C'
dz
and then t is of order p at most ; let the order of t be N, where iV < /*.
If N=n, the theorem required is proved. If N>n, then the equation
satisfied by t, namely
t = F(z,0).
By arguments resembling those used in 4*71, we find that, if N> n, then
F(z,0 + c)
is also a solution of (2). The corresponding solution of (1) is
expJJF(*, + c)dz,
and this is a solution for all values of c independent of z. Hence, by
differentiation with respect to c, we find that the function u^ defined as
where C is a constant.
If C= 0, Fis independent of 0, which is contrary to hypothesis ;soG^ 0, and
m, = J(C/F,).
Hence an algebraic function of
m, is and similarly it is an algebraic function
;
&), t = G (z,
and, by arguments resembling those used in
471 and those used earlier in
this section, we find that the function u., defined as
3 r i
r- exp }&(z, c<d)dz
so that
1
__ i(j _= ! ,,.eG 9 .
,-V{W(?.)l.
Consequently u x an algebraic function, not only of all the transcendants of
is
the types and ^, but also of those of type which occur in t and therefore ;
4 #
73. Liouville's theoi-emf that Bessel's equation has no algebraic integral.
We shall now shew that the equation
various modifications) to Bessel's equation was given in his later paper, Journal de Math. vi.
(1841), pp. 113, 36.
,
-"%<*>
dhi ,
x
where
(2) X (.).(l)-i.
If possible, let Bessel's equation have an algebraic integral then (1) also ;
has an algebraic integral. Let the equation which expresses this integral, u,
as an algebraic function of z be
(3) 64(u,z) = 0,
where 64 is a polynomial both in u and in z ; and it is supposed that 64 ik
irreducible*.
The equations (3) and (4) have a common root, and hence all the roots
of (3) satisfy (4).
For, if not, the left-hand sides of (3) and (4) (qua functions of u) would
have a highest common factor other than 64 itself, and this would be a
polynomial in u and in z. Hence 64 would be reducible, which is contrary to
hypothesis.
Let all the roots of (3) be ,, , ... uM . Then, if s is any positive integer,
tt/ + W *+2 + Ujg*
is a rational function of z ; and there is at least one value of s not exceeding
=i
where r = 1, 2, ... 5. Since ** 2 , ... m,, are all solutions} of (1), it is easy to
prove that
dWo
<*\ -w
(6) ^=W^ + r(s-r+l) X (z)W .
1 r l ,
(r = 1, 2, ... - 1)
in z or in both u and
-J^ has no factors which are polynomials in u or
* .
That is to say,
It follows by an easy induction from (5) and (6) that the highest power of
l Kz - 9 ) in Wr is lj(z - a g )
p+r where r =
1, 2,
, ... s.
Hence there is a higher power on the left of (7) than on the right. This
contradiction shews that there are no terms of the type 2?
n>9 (z - a q )~ in
n W
and so a
W = X A n z\
n= -it
where the suffix r +1 indicates that the first r +1 terms only of the hyper-
geometric series are to be taken.
If s
odd, the terms of highest degree on the left and right of
is
(7) are
of degreesX - 2 and X respectively, which is impossible. Hence vanishes W
whenever s is odd.
When s is even, the result of equating coefficients of zx ~ in (7) x
is
We have therefore proved that, when s is odd, W vanishes, and that, when
s is even, W
n is expressible in the form
2 A ny8 z- n ,
m=0
where n a positive integer and, in the case of one branch at least, c does
is
not vanish because the constant terms in the functions r are not all zero. ^
And the series which are of the form
2 cm z- m,n
wi=0
are convergent* for all sufficiently large values of z.
When we substitute the series into the left-hand side of (1), we find that
the coefficient of the constant term in the result is c and so, for every branch, ,
c must be zero, contrary to what has just been proved. The contradiction
thus obtained shews that Bessel's equation has no algebraic integral.
Since x iz) is f order zero, it follows from 4"72 that, if Bessel's equation
has an integral expressible in finite terms, then (2) must have a solution
which is of order zero, i.e. it must have an algebraic integral.
If (2) has an algebraic integral, let the equation which expresses this
integral, t, as an algebraic function of z, be
(3) a(t,z) = 0,
where S4 is an irreducible polynomial in t and z.
First suppose that there are more than two branches of t, and let three
of them be called tlt t2 t^, the corresponding values of u (defined as exp jtdz)
,
being uly v?, us These functions are all solutions of (1) and so the Wronskians
.
n dus duz .
rf
Ua
~d~
~ U* Ua ' ~" ^'
and t3 2 is not zero, because, if it were zero, the equation (3) would have a
pair of equal roots, and would therefore be reducible.
Hence G x 0, and so
UtU3 = dfcU - 2 ).
But Wl =A /^-^,
and therefore ux is an algebraic function of z. This, as we have seen in 4-73,
cannot be the case, and so t has not more than two branches.
K) {V' + 4>UV=0.
Let V be factorised so that
V=Az*Tl(z-a q y<i,
where A is constant, Kq and \ are integers, and k,{ and a q are not zero.
122 THEORY OF BESSEL FUNCTIONS [CHAP. IV
\z q
4 (z aq )
<>
^ + f4i^ + {s +2 4(r^)F + ^n(,-a,r'- X (,),o.
Now consider the principal part of the expression on the left near a q . It
is evident that none of the numbers x q can be less than 2, and, if any one
of them is greater than 2 it must satisfy the equation
Kq t T!/f q =s V/>
the numbers tc
q
are equal to 2.
Again, if we consider the principal part near oo , we see that the highest
power in V must cancel with the 1 in % (z), so that \ = 2 Kq .
9
It follows that xJV is rational, and consequently (t, z) is reducible, which
is contrary to hypothesis.
Hence t cannot have as many as two branches and so it must be rational.
If we consider the principal part of the left-hand side near aq we see that
l/(z aq ) cannot occur in t to a higher power than the first and that
Bljq -&hq = 0,
so that Bhq =l.
we consider the principal parts near and oo we find
Similarly, if , that
*=1, (A^y-JL-t-pip+l); X~0, 4,' = -l.
Since p = v ^, we may take A- = p without loss of generality.Y
When we substitute 2 cmzm for w in (7) we find that the relation connecting
successive coefficients is
m (m 2p 1) cm 2icm-! (m p - 1) = 0,
and so the series for w cannot terminate unless in p 1 can vanish, i.e. unless
Hence the hypothesis that Bessel's equation is soluble in finite terms leads
of necessity to the consequence that one of the numbers + v \ is zero or a
positive integer; and this is the case if, and only if, 2v is an odd integer.
Conversely we have seen ( 34) that, when 2v is an odd integer, Bessel's
equation actually possesses a fundamental system of solutions expressible in
finite terms. The investigation of the solubility of the equation is therefore
complete.
Some applications of this theorem to equations of the types discussed in 4*3 have
been recorded by Lebesgue, Journal de Math. xi. (1846), pp. 338 340.
fr az + brf
with a view to proving that it is, in general, not integrable in finite terms.
4w
(i-0, 1, 2, ...)
2w + l
This theorem was proved by Liouville, Journal de Math. vi. (1841), pp. 1 13. It
seems impossible to establish it by any method which is appreciably more brief than the
analysis used in the preceding sections.
124 THEORY OF BESSEL FUNCTIONS [CHAP. IV
The first appearance in analysis of the general Bessel coefficient has been
seen ( 1*3) to be in connexion with an equation, equivalent to Laplace's
equation, which occurs in the problem of the vibrations of a circular membrane.
We shall now shew how Bessel coefficients arise in a natural manner from
Whittaker's* solution of Laplace's equation
(1)
^ + ^ + ^= -
J
ek (z + ixcosu + iysinu) cog mu ^
in which k is any constant and m is any integer.
If we take cylindrical -polar coordinates, defined by the equations
0**
|
eiipcog(u-*) cosmM ^ M=::e*z I
eik* CMV cos m (v + <f>)
dv,
J n J it
J
I
IT
gk (z + ix cos u + iy sin u) sm mu ^
and this is equal to 2vim e kz sin m<p . Jm (kp). Both of these solutions are
analytic near the origin.
* Monthly Notices of the R. A. S. lxii. (1902), pp. 617620; Math. Ann. lvii. (1902),
pp. 333341.
t The simplest method of effecting the transformation is by using Green's theorem. See
W. Thomson, Camb. Math. Journal, iv. (1845), pp. 3342.
4-8, 4-81] DIFFERENTIAL EQUATIONS 125
and a normal solution of this equation of which ekz is a factor must be such that
lcTV
Vd<p*
is independent of <f>,
and, if the solution
is to be one- valued, it must be equal
to m where 2
m is an integer.
Consequently the function of p which is a
factor of V must be annihilated by
dp 3 p dp \
and therefore it must be a multiple of Jm (Jcp) if it is to be analytic along the
line p 0.
e**
sin
. m<b
r . Jm (kp).
r/
' \
These solutions have been derived by Hobson* from the solution ^*J (kp) by Clerk
Maxwell's method of differentiating harmonics with respect to axes.
Another solution of Laplace's equation involving Bessel functions has been obtained by
Hobson (ibid. p. 447) from the equation in cylindrical-polar coordinates by regarding d/dz
as a symbolic operator. The solution so obtained is
cos
sin
. m+ 'V(pz)M>
where/(s) is an arbitrary function but the interpretation of this solution when <@ involves
;
m
a function of the second kind is open to question. Other solutions involving a Bessel
function of an operator acting on an arbitrary function have been given by Hobson, Proc.
London Math. Soc. xxiv. (1893), pp. 5567 xxvi. (1895), pp. 492494. ;
K } + + ~&
dx2 dy3 dz 3 dt 2 '
in which t represents the time and c the velocity of propagation of the waves,
from the same aspect.
Whittaker'sf solution of this equation is
-r f,i
gik(xainucosn + yainuainv+zcoiiu + et)
J? / \
dU fJv
The physical importance of this particular solution lies in the fact that it
is the general solution in which the waves all have the same frequency kc.
angular coordinates of the direction which the (u, v) referred to new axes for
polar axis is the direction (6, <f>) and the plane y\r = passes through the
.s-axis. The well-known formulae of spherical trigonometry then shew that
Vn = e** 1 (*
J
e*rc08 *- Sn (0, <f>
; to,
ty) sin wdtodty.
= (2*)Ue^^^A n (0,4>)
by 332.
Now the equation of wave motions is unaffected if we multiply x, y, z and
t by the same constant factor, i.e. if we multiply r and t by the same constant
factor, leaving and unaltered so that A n (0, <f>) may be taken to be in-
<f>
;
* This follows from the fact that Laplace's operator is an invariant for changes of rectangular
axes.
t This is otherwise obvious, because Sn may be taken independent of k.
4-82] DIFFERENTIAL EQUATIONS 127
*("*) -<++*'.
so that if this factor is to be analytic at the origin it must be a multiple of Jn + {hr))^,:
e^t/o [k>s/(p* +a - 2
lap cos <f>)},
by making a change of origin to the point (a, 0, 0), we see that it is a solution
of Laplace's equation in cylindrical-polar coordinates. This solution has e** as
a factor and it is analytic at all points of space. It is therefore natural to
expect it to be expansible in the form
= Wl l J
Assuming the possibility of this expansion, we observe that the function under
consideration is an even function of 0, and so Bm = and, from the symmetry ;
Jo {kx'Xp* +a
2
lap cos <)} = 2 em cm Jm (kp) Jm (ka) cos m<f>.
m=0
If we expand both sides in powers of p, a and cos <f>,
and compare the
coefficients of (k-pa cos <)
m we
, get
and which represents a wave moving in the direction of the axis of z from
(^) e
iket
2 c n i* JnH (kr)P n (cos 0),
\K)J =o
(If)*
n on each we find that
If we compare the coefficients of (kr cos 6) side,
n \-^
1T)
2+*r(n + f)"2.(n!r
i
e
ikrc se = fiE) i (n + l)iJn+ i(kr)P n (cos0),
\krj =o v
The analysis just explained has been extended by Hobson to the case of
the equation
'
+ + ^
a?2 r 9r
ikct
and so such a solution, containing a time-factor e , must be of the form
iket <& ilp -*
e (kr)l{kr)*^K
* This is due to Neumann, Theorie der Bessel'schen Functionen (Leipzig, 1867), pp. 59
65.
t The harmonics do not occur because the function is symmetrical about the axis of
tesseral z.
X This expansion is due to Bauer, Journal filr Math. lvi. (1859), pp. 104, 106.
Proc. London Math. Soc. xxv. (1894), pp. 4975.
4-83] DIFFERENTIAL EQUATIONS 129
f -
Hobson describes the quotient ^\{ P-i)(kr)l(kr)^fh %) as a cylinder function
of rank p ; such a function may be written in the form
<$(kr\p).
J {k v^r + o - 2 2
2ar cos <j>) \ p),
no other coordinates.
Hence
** J {k V(r2 + a? - 2ar cos <f>) j p)
is annihilated by the operator
'
dp2 p dp dx
that is to say, by the operator
& ,
J>~1 9 . (ff-2)cos4> d 1 y
dr* r dr r*sin<j> d<f> r*d<p*
Now normal functions which are annihilated by this operator are of the form
"
(hr)ir->
pn (cos <p\p),
where Pn (cos <p j p) is the coefficient* of on in the expansion of
(1 - 2a cos <f>
+ cpy-*P.
By the reasoning used in 4*82, we infer that
J \k V^r + 2
a* 2ar cos <f>) \ p)
00
1
~ (/fca^-^/fcr)^- 1
^ AnJn+ip~ ( kr ) ^h-Jp-i (ha) Pn (cos
1
<f> \p).
Now expand all the Bessel functions and equate the coefficients of
n on each side
(k?ar cos <f>) we find that ;
so that A n = 2**- 1
( + %p - 1) T (|p - 1 ).
Pn (cos p_ * (cos
j p) S3 C* 0).
130 THEORY OF BESSEL FUNCTIONS [CHAP. IV
(r 2
+a 2
- 2ar cos <)il>_ *
' + + 2
+
c2 dt*
* dx1i dxf 9a;, dx4*
have been investigated by Bateman*, who also established a connexion between
the two systems.
* '
{dp 2
+ + 2 2
+
[do*
+
d* o*
'
1 &V + 1 92 F ld*V
'
2
r cos < 2
9x r2 sin2 4* W* c2 ^'
Now normal solutions of this equation which have e* (MX +"* + * c<) as a factor
are annihilated by the operator
3 3 d_
dr3 r dr
* Messenger, xzxiu. (1904), pp. 182188; Proc. London Math. Soc. (2) hi. (1905), pp. 111123.
,
and since such solutions are expressible as the product of a function of r and
a function of they must be annihilated by each of the operators
<f>
y 3 9 4X(X+1)
dr2 rdr r* '
d *
\2 + (cot<t>-t*n<l>)^r
^ + 4 \(\+l)- l
**
9< 3^> cos2 ^> sin 2 '
x 2 F (* - X,
X ^- + \ + 1 ;
+ l ;
sin 2 + \ *ox+*++**>.
2 a A (kr)~ l
J^ (kr) cos*< sin"< . 2JP, (~^ - \ ^^ + X + 1 ; v + I ; sin 2 <f>)
where the summation extends over various values of X, and the coefficients a*
depend on X and <J>, but not on r or <j>. By symmetry it is clear that
aA = 6 X cos*<I> sin" 2 F .
t
(^-J^ - X, ^^ + X + 1 ; v + 1; sin 2 <l> \ ,
We shall not give Bateman's proof, which is based on the theory of linear
differential equations,* but later ( 11*6) we shall establish the expansion of
J (kr cos <f>
cos 4>) Jv (kr sin <f>
sin <I>) by a direct transformation.
CHAPTER V
MISCELLANEOUS PROPERTIES OF BESSEL FUNCTIONS
The recurrence formulae 3.9 (5) and (6) at once lead to the results
(1)
J
V>% (*) dz = *H- #v+i (z)>
^z^f{z)%%(z)dz;
let this integral be equal to
z"+> {A (z) <$ v (z) + B (z) <&v+1 (z)},
In order that A (z) and B (z) may not depend on the cylinder function, we
take A (z) = B' (z), and then
(3)
J"
** +1 \b"(z) + ^~ B' (z) + B (*)| # (*) dz
= *+ {B' (z) <@ v (z) + B (z) <@ v+1 (*)}.
This result was obtained by Sonine, Math. Ann. xvi. (1880), p. 30, though an equivalent
formula (with a different notation) had been obtained previously by Lommel, Studien fiber
die BesseVschen Functionen (Leipzig, 1868), p. 70. Some developments of formula (3) are
due to Nielsen, Nyt Tidsshrift, ix. (1898), pp. 7383 and Ann. di Mat. (3) vi. (1901),
pp. 4346.
For some associated integrals which involve the functions ber and bei, see Whitehead,
Quarterly Journal, xlii. (1911), pp. 338 340.
5-1, 5-11] MISCELLANEOUS THEOREMS 133
should be noted:
(4) I V+ ir
^ 0) ds = - (ft - v*)f'* - &r (z) dz
2 L li
The simplest integrals which contain two Bessel functions are those
derived from the Wronskian formula of 312 (2), namely
r / \ T = 2 sin-
I/7T
Jr v
/
(z)
\ T
J'_ (z)
/ \
- J- v (z)Jv
' /
(z)
\
,
which gives
K
J zJ*(z) 2sini/7r Jv (z)
'
, v
/"* dt 7T . /_,, (.g)
9
zJv (z) J_ <>) ~
Kl) 10g
) 2 sin *tt J v (z)
'
,ox [
z
dz _ir Fy (>)
^
K }
['
J
dx _
zYv\z)~
it
2
Jv (z)
Y v (z)'
The reader should have no difficulty in evaluating the similar integrals which contain
any two cylinder functions of the same order in the denominator. The formulae actually
given are due to Lommel, Math. Ann. iv. (1871), pp. 103 116. The reader should compare
(3) with the result due to Euler which was quoted in 1*2.
Some more interesting results, also due to Lommel*, are obtained from
generalisations of Bessel's equation.
d*y n d?n ~ rt
then
j\P -Q)ynd^y^-v^.
* Math. Ann. xiv. (1879), pp. 520536.
134 THEORY OF BESSEL FUNCTIONS [CHAP. V
Now apply this result to any two equations of the type of 431 (17). If
*^V> ^i, denote any two cylinder functions of orders fi and v respectively, we have
where <f>
(z) and i/r (2) are arbitrary functions of z.
<f>
(z) and yjr (z) to be multiples of z, say kz and Iz. It is then found that
Z
{{^-P)z-^^\^{kz)^
^
(7) [ v (lz)dz
d
= ,fa(kz) d^-^(lz)
dz dz
= z {k^ +1 (kz) W. (Iz) - 1&* (kz) # r+1 (lz)} -(/*-) ^M (kz) <@ v (Iz).
The expression on the left simplifies still further in two special cases (i)fiv,
(ii)k=*l.
It becomes nugatory when k = I, for the denominator is then zero, while the
numerator is a constant.
If this constant is omitted, an application of l'Hospital's rule shews that,
when I -*- k,
(9)
J"
z<@ (kz) V (kz) dz^-^ {kz <^M+1 (kz) #/ (kz)
- ks&r (kz) W^(kz) - ^V (kz) ^ M+1 (kz)}.
The result of using recurrence formulae to remove the derivates on the
right of (9) is
(12) [*?<&,, (kz) LM (kz) dz = * {2^M (kz) iLM (fa) + ^ (fa) f _M _x (fa)
+ ^ +1 (fa)iU +1 (fa)},
as a cylinder
function of order /*.
is found that
V^(,)0, V^ d
^^ = 2f^ li
(z)
by -
\<f*(z)
Z
OI&
,
z
1
-^^(z) respectively, subtracting and integrating, and then re-
placing z by kz.
(is)
J
jj(kz)^ -g{/ w (fa)gKfa)-^(fa)gu +1 (fa))+~j |l-(fa)L
other formulae of this section seem to have been discovered before the publication of
Lommel's memoir.
Various special cases of the formulae have been worked out in detail by Marcolongo,
Napoli Rendiconti, (2) in. (1889), pp. 9199 and by Chessin, Trans. Acad. Sci. of St Louis,
xii. (1902), pp. 99108.
136 THEORY OF BESSEL FUNCTIONS [CHAP. V
5"12. Indefinite integrals containing two cylinder functions; Lommel's
second method.
^ {zo S?
M+1 (z) & v+ > CO} =* [V, (z) W v+X (z) + ^M+1 (z) (z)}
+ (p- H,-p-2)z'>-> <@+1 (z) W v+1 (z),
so that
(p + ft + v) j'z'- 1
^ (z) <@ v (z)dz + (p-p-v- 2)
J
*
z"- 1 <@r+l (z) W +1 (z) dz
Thus we have
e
(2)
J
^ l
^()?r Wdf- f ~7^ )
i^(f),(f) + ^(f)?rH (i)i,
As special cases of these
(4)
rz
/
^ +1 r
^ (*) <fc = ^_ Wv+
(*) + *V+i (*)}
In particular, if fi = v = 0,
(6) /'*.(#)*.(*)*
/<
%(z)<& (z)^,
For a special case of (1) see Rayleigh, Phil. Mag. (5) xi. (1881), p. 217. [Scientific Papers,
I. (1899), p. 516.]
The
analysis of 5-1 has been extended by Sonine, Math. Ann. xvi. (1880), pp. 30 33,
to the discussion of conditions that
but the results are too complicated and not sufficiently important to justify their insertion
here.
which is a natural extension of the formula 5*1 (4), has been discovered by
Schafheitlin* and applied by him to discuss the rate of change of the zeros of
<&9 {z) as v varies ( 156).
*
= [- ***<&. (z) <@: (*>] + 1 {*+#/ (*)+(,* + 1 >
**+ %% (z) 'w: (* dz.
+2
J
V <@: (z) {*&: (z) + (* - *> k (^)j cte,
(fi + 1)
J
V+ 2 <&,'* (z) dz = |>+ W (z)] + 2
J'**
1
(z* - i/
2
) # (z) <&,' (z) dz.
Hence, on substitution,
(/* + 1) [ V {? - 1/
2
) # (z) dz
= |>+3 ^V 2
(*) - (ji + 1) ^+ 2
^ (*) #/ (*)]
+ 2 [ V+ 3 <@v (z) <@: (z) dz + {(/a + l)2 - M j"V+# r (*) ^V(s) ek
= |>+ 3
^ v
'
2
(*) - (/*+ 1) ^+ 2
^(*) #/ (2) + z* +3 * (z)
+ {(/* + 1)2 -"'l* * 4 1
^ ^)]3
Z
-(p+3)j z+*'@vi (z)-(ti + 1) { (/* + 1) - 2
v*\ J'rW(t)dz.
By rearranging we find that
(fi + 2)
J*
** #, (5) (2c = (n+ 1) (i/
2
-\ (jjl + 1 )2 } [ V r
#2 (*) <2e
We shall now discuss some of the simplest expansions of the type ob-
tained for (\z)m in 2*7. The general theory of such expansions is reserved
for Chapter xvi.
n= ni
n J s r(u + w) T ,. r(At+n + i) r 1
Lm-o n - -o w -
J
* Wiener Sitzungsberichte, lxxiv. (2), (1877), pp. 124130.
5*2, 5*21] MISCELLANEOUS THEOREMS 139
and so the' sum is a constant. When we make z--0, we see that the constant
is unity; that is to say
The reader will find that it is not difficult to verify that when the expansion on the
right in (1) is rearranged in powers of z, all the coefficients except that of z* vanish ; but
this is a crude method of proving the result.
The expansion
(i) ($z)-'J (z) = az)-r( v + i-fi)
Qi+2n)r(g+n)
v 5
.to nir(F + l-/*-)r(ir + n + l)
T
jM+2n W
(
.
integers.
(*zY *' \z )
~ * i n/ TT\
".?.
by Vandermonde's theorem ; and the result is established.
(8) (**)-*(*)-*.?.,,#&
x 2F,(/x + n, -n; y + 1; h?) (fi + 2n) J^+^iz).
This formula will be required in establishing some more general expansions
in 11-6.
Again, (z + h) "Jv {<J(z+ h)} is analytic except when z+h = 0; and so,
provided that h | j
< I z \
, we have
(2)we deduce from 3*4, after making some slight changes in notation,
equation (4) being true only when J*|<i|*i- These formulae are due to
Glaisherf, who regarded the left-hand sides as the generating functions
associated with the functions whose order is half of an odd integer, just as
exp {\z (t ljt)\ is the generating function associated with the Bessel co-
efficients.
Proofs of (3) and (4) by direct expansion of the right-hand sides have
been given by Glaisher the algebra involved in investigations of this nature
;
is somewhat formidable.
* Studien iiber die Bessel' schen Functionen (Leipzig, 1868), pp. 11 16. Formula (1) was given
by Bessel, Berliner Abh. 1824 [1826], p. 35, for the Bessel coefficients.
t Quarterly Journal, xn. (1873), p. 136 ; British Association Report, 1878, pp. 469470. Phil.
Tram, of the Royal Soc. clxxii. (1881), pp. 774 781, 813.
5-22] MISCELLANEOUS THEOREMS 141
(-) w (i^) ?
(5) jv {W(i + *)} = a + wi m=0 m -
V y+w (^) >
and, in particular,
l (i/ + 1) TO=0 m!
In like manner, from (2),
provided that the series on the right is convergent. The convergence is obvious
= _( 4f^sin^
+1 l '
7TTO*
Hence the condition for convergence is R (v) > 0, and if the condition is
satisfied, the convergence is absolute. Consequently, when R (v) > 0, and also
when v is any integer,
{
- )m{h * )m
(9) I Jv- m (z) = Q.
m= ml
In like manner, if R(v) > 1, and also when v is any integer, we have
and k <
J
\ 1 ; so that
when v tends to an integral value, we see that they hold for functions of
integral order.
the first kind, we see that we may substitute the symbol ^ for the symbol
Y throughout.
These formulae were noted by Lommel, Studien, p. 87. Numerous generalisations
last
of them be given in Chapter xi. It has been observed by Airey, Phil. Mag. (6) xxxvi.
will
When we combine (5) and (13), and then replace V(l + k) by X, we find
when |\-1|<1,
that,
(15) 9. (\m) = X* 2
m-0
(
~ )m(V
",
m.
1)W(
^ ^h ,
and, in particular, when X is unrestricted,
d) /.(x.) = v I H-fr'-^fr^ w
These two results are frequently described* as multiplication theorems for
Bessel functions.
It may be observed that the result of treating (14) in the same way as (8) is that
(when v is taken equal to an integer n)
(17)
x
'
-(n-l)l (2/ 2 )=tt 2
ra=0 m Yn . m {z).
i-Jf-
An alternative proof of the multiplication formula has bee.n given by Bohmer, Berliner
Sitzungsberichte, xiii. (1913), p. 35, with the aid of the methods of complex integration
see also Nielsen, Math. Ann. lix. (1904), p. 108, and (for numerous extensions of the
formulae) Wagner, Bern MittheHungen, 1895, pp. 115119; 1896, pp. 5360.
[Note. special case of formula (1), namely that in which v = l, was discovered by
A
Lommel seven years before the publications of his treatise ; see Archiv der Math, xxxvn.
(1861), p. 356.
His method consisted in taking the integral
/ / co (r cos + ijr sin 6) ddt)
by two different methods.
over the area of the circle * + $= 1, and evaluating it
i fir . M -\JQ-P> d
j (r cos
sin vs + r)r sin 0) -
a
.
rsin.0
2ttJ_i|_ _W0-f)
\
* See, e.g. Schafheitlin, Die Theorie der Besselscken Funktionen (Leipzig, 1908), p. 88.
5*23, 5*3] MISCELLANEOUS THEOREMS 143
1
5~
4bW J
fn
/
v J
f
I
o
1
=
^ ff cos tfr) ^d^^-jjl- ?)* cos (r) <*=./, (r)/r.
If we compare these equations we obtain (1) in the case i/ =l with z and h replaced by
r2 cos2 and r2 sin 2 *.]
a) j, w = r^+1) | n.-^^
M r + m+
1 (v ft) = (v 1)
o^ m! * +TO v '
This formula was given by Lommel, Studien iiber die Bessel'schen Functionen (Leipzig,
1868), pp.2223, in the special case M =0; by differentiating with respect to v and then
putting *=0, it is found that
, , *{V' (l +i)+V''(-/*)-'<Kn-M)K
and, when /i=0, we have Lommel's formula
* MatA. Ann. hi. (1871), pp. 135137. ( JWd. xvi. (1880), pp. 78.
144 THEORY OF BESSEL FUNCTIONS [CHAP. V
It will first be shewn that the series on the right of (1) is a uniformly
convergent series of analytic functions of both z and t when
|*|<r, R^\t\^A,
where r, R, A are unequal positive numbers in ascending order of magnitude.
When m is large and positive, J ^m (t)Jm {z) is comparable
v with
and the convergence of the series is comparable with that of the binomial
series for (1 r/R)". When m is large and negative (= w), the general
term is comparable with
(-)"(*A)'(jAr)"
T(v + n+l).n\
and the uniformity of the convergence follows for both sets of values of m by
the test of Weierstrass.
Term-by- term differentiation is consequently permissible*, so that
i *
=H 2 { J,--! (0 - Jv- m+1 (t)} Jm (Z)
" m= ao
1 "
-^* 2 J -m(t){Jm- (z)-Jm+1 (z)}, v 1
m= -oo
and it is seen, on rearrangement, that all the terms on the right cancel, so
that
d-a)J./~< >'-<'>- a ,
30
Hence, when z | \
< \t \ , the series 2 Jv -m (t) Jm (z) is an analytic function
m= -oo
Math. Ann. xliii. (1893), pp. 141 142. Various generalisations of these formulae will be
given in Chapter xi.
The ascending series for the product J. (z) Jv (z) has been given by various
published it in 1877, but it had, in fact, been previously published (in 1870)
by SchlafliJ. More recently the product has been examined by Orr, while
Nicholson has given expansions
|| (cf. 5'42) for products of the forms
by the product of two Bessel functions, and solve it in series. We shall then
( 541) obtain the expansion anew by direct multiplication of series.
* See also Epstein, Die vier Reehnungsoperationen mit BeueVsehen Functionen (Bern, 1894),
[Jahrbuch tiber die ForUehritte der Math. 18931891, pp. 845846].
t Ueber die Auswerthung bettimmttr Integrate mit Hillfe von Ver&ndetungen det Integrationsweget
(Bern, 1877), p. 13. The authorities who attribute the expansion to SchSnholzer include Graf and
Oubler, Einleitung in die Theorie der BesseVschen Funktionen, u. (Bern, 1900), pp. 8687, and
Nielsen, Ann. Sci. de Vltcole norm. sup. (3) xvin. (1901), p. 50 ; Handbueh der Theorie der Cylin-
der/unktionen (Leipzig, 1904), p. 80. According to Nielsen, Nouv. Ann. de Math. (4) u. (1902),
p. 396, Meisselobtained some series for products in the Iserlohn Programm, 1862.
% Math. Ann. hi. (1871), pp. 141142. A trivial defect in Schlafli's proof is that he uses a
contour integral which (as he points out) converges only when B(/t+r + l)>0.
Proe. Camb. Phil. Soc. x. (1900), pp. 93100.
||
Quarterly Journal, xliii. (1912), pp. 78100.
'
d
It follows that ~ \y" + (J + J) y] = 2v'V + 2v'w"
dz
= - 2Ivw' - 2Jv'w
and hence y'" + 2 (/ + /) y + (/' + J') y = (/ - J) (v'w - vw').
y + 4/y' + 2/>=0;
/'/
(1)
but, if IJ, it is easy to shew by differentiation that
(2)
* qr+w+wv+J-), }
_ _ (/ _ ^
This is the form of the differential equation used by Orr ; in connexion with (1), see
Appell, Comptes Rendm, xci. (1880), pp. 211 214.
To apply these results to Bessel's equation, the equation has to be reduced
to a normal form ; both Orr and Nicholson effect the reduction by taking
z 9!> v (z) as a new dependent variable, but, for purposes of solution in series, it
d_
dd
0+2(2^-^-i, )^| + 4^y| + (^-^) 8 2
y = 0,
that is to say
* 2 {-)m cm z
where a /* v and
+ 2m !)( + 2m) ctn _i
4 (a
m
(a + ft + + 2m)(a + p -v + 2m)(a- p+ v + 2m)(a- fi-p + 2m)
1>
If we take a = /* + v and
^-2^+"r(/*+i)r(i/ + i)'
we obtain the series
o (_)w
Qgy,+p+m, Y( fl + y+ 2m + 1)
w?o m!r(/* + v + m + l)r( A* + w+l)r(i/+w+l)
and the other series which are solutions of (3) are obtained by changing the
signs of either /i or v or both /* and v.
"
is easy to infer that, if 2/*, 2v and 2 (/* + v) are not negative integers and if
fi* ^ , then
In like manner, by solving (4) in series, we find that, when 2i> is not a
negative integer, then
m! T(/* + m+ l) p (i> + w + l)
(n +v+m + l)m
m!r(/t + w+l)r(i/+m + l)'
when Vandermonde's theorem is used to sum the finite series.
----- ' *~
'
I \ . H--' S
148 THEORY OF BESSEL FUNCTIONS [CHAP. V
This obvious mode of procedure does not seem to have been noticed by any of the
was given by Nielsen, Math. Ann. lii. (1899), p. 228.
earlier writers; it
The series for cos z and J {z)e,mz were obtained by Bessel, Berliner Abh. 1824,
J {z)
m Qitn2n A2n
ml r<> + m + l)r(i/ + l)
and so
m=0 t! r(/i+TO+l)
One case of reduction is the case b = a, which has already been discussed
another is the case b = ia, provided that ft* = v*.
In this case we use the formula*
* Cf. Kummer, Journal fUr Math. xv. (1836), p. 78, formula (53).
5*42] MISCELLANEOUS THEOREMS 149
The formulae (3), (4), (5) were discovered by Nielsen, Attidella R. Accad. dei Lined, (5)
xv. (1906), pp. 490 497 and Monatshefte fiir Math, und Phys. xix. (1908), pp. 164170,
from a consideration of the differential equation satisfied by Jv (az)Jv (bz).
Some series have been given, Quarterly Journal, xli. (1910), p. 55, for products of the
types Jy3 (z) and J% (z) t/_ (z), but they are too cumbrous to be of any importance.
The special case of this formula in which 2v is an integer has been given
by Hobson*.
We have
{
JM (,) J, (*)} - log {\z). J (z) J (z)
l~
v I"
(~)r (^>t+,,+8r T(p + v + 2r + l)
and
r (fi - v + 2r + 1)
(-)" (^y4-"*"*
r=0 Lr: rO-i/ + r+l)r(/* + r+l)r(-i/ + r + l)
x fi/r (p - y + 2r -H 1) - ^ (/* - v + r + 1) - yfr (- + r + 1)} 1/
n-l oo
We divide the last series into two parts, 2 and X . In the former part we
have
* Proc. London Math. Soc. xxv. (1894), p. 06 8 je also Cailler, ilfm. d ia Soc. de Phys. de
;
r -o r!(n+r)ir(^ + r+l)
The series for Ym (z) Yn (z) can be calculated by constructing series for
~
[ d-J h (z)Jv (z)
|_
dfldv jn-m, v-n
to the result of 5
-
41 ; the integral has this value when m 0, 1, 2, provided . .
. ,
J(*)J(*) = - 2
m -o.'o
7r
(
,w .. + m ..n
m!r(/*+v + l) T
^s(fi-v)dd0,
* This formula is due to Cauchy ; for a proof by contour integration, see Modern Analysis,
p. 263.
5-43-5-51] MISCELLANEOUS THEOREMS 151
(2) JM (z) Jn (z) = *i~L P'j^ (2z cos 6) cos + n) 0d0, (fi
T Jo
and this formula is also true if /* and n are both integers, but are otherwise
unrestricted.
Formula (1) was given by Schlafli, Math. Ann. in. (1871), p. 142, when nv are both
integers; the general formula is due to Gegenbauer, Wiener Sitzungsberichte, cxi. (2a),
(1902), p. 567.
(1) (**r -
r{/M + v+1 y
m-0 m.
The formula is true if fi and v are not negative integers, but the following
proof applies only if R(fi + v + l)> 1.
From 5*2 we have
(.>-.- 1 0>+>+)ro. + > + m) J ^^
If we multiply by cos (fi v)0 and integrate, it is clear from 5*43 that
2^+
:p\*".co.-)M- 2 0>+ + )ro.+>+)
Jo to-0 ml
and the result follows by evaluating the integral on the left for other values ;
+h)<J,yW,
i8 j> H (.).i"'
n=0 Z dz
so that
2 J'xJK-i
(*) <fc = 2 (* + 2) JV2 (*)1
J
* The results of this section will be found in Math. Ann. ii. (1870), pp. 632633 ; xiv. (1878),
p. 532 ; MUnchener Abh. xv. (1886), pp. 548549.
- ;
(1) ** {./
2
_ x (z) - J^ (z) J (z)} = 2 (i; + 2n) JV (*), 2
n-0
on taking zero as the lower limit when R (v) > by adding on terms at the ;
beginning of the series, it may be seen that the restriction R,(v) > is super-
fluous.
(3)
*. (-r<n + *)/+(*).
2 n J\ + (s)
-o
for it is evident that
a
= ivJ?(z)\z,
and so. when iJ(v) >0, we obtain a modification of Hansen's formula (2 -
5),
namely
00
f* fit
(5) 2 en J\+n (z) = 1v\ Jf(i%.
=o Jo *
m=o tJo
f 2 sin2 *]* 2 f* . _ ,cft
7T t 7T.I *
|_ Jo
and so
(6) 2 /'
n+J (*) = -i(2*),
n=0 T
where, as usual, the symbol Si denotes the "sine integral." This result is given
by Lommel in the third of the memoirs to which reference has been made.
5-6] MISCELLANEOUS THEOREMS 153
^ */_,(*) 1 - 1 - 1
1 J v+m\ z )
This formula is easily transformed into
'
K ' J ^ (z)
v 2v/z -2(v + l)/z-...-2(v + m)lz - Jv+m (z)
These results are true for general values of v\ (1) was discovered by
Bessel* for integral values of v. An equivalent result, due to Schlomilchf, is
Jv+1 (z) z z*
J* z2 zJv+m+1 (z)
(4)
Jv (z) 2(i/ + l)-2(i/ + 2)-2(i/ + 3)-...-2(/ + m)- J, +m (z) '
'
The Bessel functions in all these formulae may obviously be replaced by any
cylinder functions.
It was assumed by Bessel that, when m -*- <x> , the last quotient may be
neglected, so that
* Berliner Abh. (1824), [1826], p; 31. Formula (2) seems not to have been given by the earlier
writers; see Encyclopedic des Sci. Math. n. 28, 58, p. 217. A slightly different form is used by
Graf, Ann. di Mat. (2) xxin. (1895), p. 47.
t Zeitschrift fiir Math, und Phys. n. (1857), p. 142; Schldmilch considered
integral values of
v only.
J Studien fiber die BesseVschen Functionen (Leipzig, 18G8), p. 5 see also Spitzer, Archiv der
;
Math, und Phys. xxx. (1858), p. 332, and Gunther, Archiv der Math, und Phy*. lvi. (1874),
pp. 292297.
W. B. F. 6
154 THEORY OF BESSEL FUNCTIONS [CHAP. V
It is not obvious that this assumption is justifiable,
though it happens to
be and a rigorous proof of the expansion of a quotient of Bessel functions
so,
into an infinite continued fraction will be given in 9'65 with the help of the
theory of " Lommel's polynomials."
[Note. The reason why the assumption is not obviously correct is that, even though
The reader will find an elaborate discussion on the representation of Jv (z)/Jv -i (z) as a
continued fraction in a memoir* by Perron, Munchener Sitzungsberichte, xxxvn. (1907),
pp. 483
504; solutions of Riccati's equation, depending on such a representation, have
been considered by Wilton, Quarterly Journal, xlvi. (1915), pp. 320323. The connexion
between continued fractions of the types considered in this section and the relations con-
necting contiguous hypergeometric functions has been noticed by Heine, Journal fiir
Math. lvii. (1860), pp. 231247 and Christoffel, Journal fur Math, lviii. (1861),
pp. 9092.
(i) J. W -iin r a^ J
..i'1
( > M;+i;-^).
We shall prove this result for general (complex) values of v and z when \ and
fi tend to infinity through complex values.
/,
K
* }
=-, n [(1 + rh) (1 + rv )\
This a continuous function of 8 and rj and, if 8 r) are arbitrary positive
is ; ,
numbers (less than 2 z j"1 ), the series of which it is the (wi + l)th term con-
|
For the term in question is numerically less than the modulus of the (m + 1 )th
term of the (absolutely convergent) expansion of
and the uniformity of the convergence follows from the test of Weierstrass.
function of both the variables ($, rj) at (0, 0), and so the limit of the series is
Now consider the nature of the structure of spheres, cones and planes
associated with polar coordinates in a region of space at a great distance from
the origin near the axis of harmonics. The spheres approximate to planes and
the cones approximate to cylinders, and the structure resembles the structure
associated with cylindrical-polar coordinates ; and normal solutions of Laplace's
equation referred to such coordinates are of the form ( 4*8)
e k*Jm (kp)^m<f>.
i. (1899), pp. 338341 vi. (1920), pp. 393397]; and Giuliani, Giorn. di Mat. xxn. (1884),
;
pp. 236 239. The result has been extended to generalised Legendre functions by Heine
and Rayleigh.
It has usually been assumed that n tends to infinity through integral
values in proving (1); but it is easier to prove it when n tends to infinity as
a continuous real variable.
* If ?i were not large, the approximate formula for Pnm (cos 0) would be (sin m 0)/m !.
and the reasoning of the preceding section is applicable with the slight
modification that we use the inequality
|rin(f/)|*f|(*/n)|,
when |
\z |
$ 2 n | |
, and then we can compare the two series
is made when \n\>lj8 The details of the proof may now be left to the reader.
.
the proof. This fact was first noticed by Giuliani the earlier writers took for ;
Pw- (cos -J = r
v
^_
'
'
& (- n, n+ 1 m + 1; ; sin" **/).
so that
z\ tanh (iz/n) m
Pn~m (cosh
r> (
= i
F (~ n,n + l; m + 1; ~ sinh,,\z\n), _, ,
, _ .
2 , s
-J r m ^i) (
* *
so that
rn m sinri7r ~ z\
,,,
(4)
,.
hm
w _.ooLsm
-^
m+ r Qnm ( cosh -
n)7r \ nj
,
Km (z).
(
This formula has been given (with a different notation) by Heine +; it is most
easily proved by substituting the integral of Laplace's type for the Legendre
function, proceeding to the limit and using formula (5) of 62 2.
* Cf. Bromwioh, Theory of Infinite Series, 49.
(5)
.
K
"/(il*)" /,(
* );
this is due to Laurent*, and it may be proved by using the second of Murphy's
formulae, namely
Pn (cos 6) = cos" %6 F .
2 X (- n, -n ; 1 ; - tan 2
\6).
[Note. The existence of the formulae of this section must be emphasized because it
used to be generally believed that there was no connexion between Legendre functions and
Bessel functions. Thus it was stated by Todhunter
in his Elementary Treatise on Laplace's
Functions, Lame's Functions and
BesseVs Functions (London, 1875), p. vi, that "these
[i.e. Bessel functions] are not connected with the main subject of this book."]
Since
n log {cos (z/n) + i sin (z/n) cos <} -*- iz cos <f>
Heine f and de Ball J have made similar passages to the limit with integrals
of Laplace's type for Legendre functions. In this way Heine has
defined
Bessel functions of the second and third kinds ; reference will be made to his
results in 6*22 when we deal with integral representations of F (z).
Mehler has also given a proof of his formula by using the Mehler- Dirichlet integral
Pn (cos6) = 2- [
*<&4*L .
P
i n (cos zln\ - I'
(cos*/*)*-J +
o
C0S
-^ W }
,
but the passage to the limit presents some little difficulty because the integral is an im-
proper integral.
Various formulae have been given recently which exhibit the way in which
the Legendre function approaches its limit as its degree tends to infinity.
(1) Pn -(cos0)
= (n + )"' (cos 0)- [Jm (x)
+ sin 2
\0 {^x J^ (x) - Jw (x) + ^x' Jm+l 1
(x)} + . .],
Other formulae, which exhibit an upper limit for the error due to replacing
a Legendre function of large degree by a Bessel function, aref
(2) Pn (cos rj) iir- 1 Qn (cos rj)
= ^(sec rj) e <* + i>,- t*n '>> [Jo {( + h) ^n v} iY {( + ) tan rj\]
4^i-\/(sec v)
+ 7r{R(n)
+ ^'
less than unity in absolute magnitude, and n may be complex provided that
its real part is positive. But the proof of these results is too lengthy to be
given here.
+ + y u
dz* z dz V z* )
'
and compare the equation with the equation defined by the scheme
fa, b, c,
1
Pio, , y, z\,
l', ff, 7. -I
* Proc.
London Math. Soc. (2) xm. (1914), pp. 220 221 eome associated results had been ;
obtained previously by the same writer, Proc. London Math. Soc. xxxi. (1899), p. 269.
t Watson, Trans. Camb. Phil. Soc. xxn. (1918), pp. 277308 Messenger, xlvii. (1918), ;
pp. 151160.
X Nova Acta Caes.-Leop.-Acad. (Halle), 1888, pp. 148.
5*73] MISCELLANEOUS THEOREMS 159
namely
d* y Q- a -a '
\-fi-fi' l-y-yl dy
dz* \ z-a z-b z-c ) dz
\ z-a )
y = 0,
(z -a)(z - b) {z c)
we see that the latter reduces to the former if
a = 0, a = v fi, a = v fi,
while b, c, fi, fi', 7, y tend to infinity in such a way that + fi' and 7 +
fi
7'
( 0, 2ifi, -2ifi, \
where 7, 7 - M + i V{0* + ) + tf 2 2
}-
( 0, ifi, oo,
J
limP< v-fi, fi, y, z\
"*"
i-v-fi, -fi, y, )
( 0, qo, -4a& \
Olbricht has given other schemes but they are of no great importance and
those which have now been constructed will be sufficient examples.
Note. It has been observed by Haentzschel, Zeitschrift fiir Math, und Pkyt. xxxi.
(1886), p. 31, that the equation
whose solution ( 4*3) is xS'Wv (hit), may be derived by confluence from Lamp's equation
when the invariants g2 and g3 of the Weierstrassian elliptic function are made to tend to
zero.
CHAPTER VI
INTEGRAL REPRESENTATIONS OF BESSEL FUNCTIONS
e izt Tdt
J a
of z.
b
V, L [ e
izt
Tdtl = z v+i I e
lzt
T{\- 2
) dt + (2i/ + +1
1) iz j e
izt
Ttdt
h b
= iz"+ l
e
izt
T (t* - 1)1 + iz+* e izt \{2v + 1) Tt - ^ {T (i -
2
1)}1 dt,
f
*
Math. Ann. i. (1869), pp. 473 485. The discussion of the corresponding integrals for Iv(z)
and K v (z)
is due to Schlafli, Ann. di Mat. (2) i. (1868), pp. 232242, though Schlafli's results
are expressed in the notation explained in 4-15. The integrals have also been examined in great
detail by Gubler, Zurich Vierteljahrsschrift, xxxm. (1888), pp. 147 172, and, from the aspect of
the theory of the linear differential equations which they satisfy, by Graf, Math. Ann. xlv. (1894),
pp. 235 2C2 lvi. (1903), pp. 432 444. See also de la Vallee Poussin, Ann. de la Soc. Sci. de
;
Fig. 1.
Fig. 2.
* It is supposed that v has not one of the values \,\,\\ ...; for then the integrands are analytic
at 1, and both integrals vanish, by Cauchy's theorem.
162 THEORY OF BESSEL FUNCTIONS [CHAP. VI
zu
/"(1+.-1-)
J*(P-l)'-*dt= 2
00
^-r-jm+, -1-) V*(P-l)*dt.
;m z Vrm,
Ja m=o ml Ja
Now tm (t I)" - * is an even
2
an odd function of t according as m is even
or
or odd; and so, taking the contour to be symmetrical with respect to the origin,
we see that the alternate terms of the series on the right vanish, and we are
then left with the equation
We thus get
fd+> n
I *-* (u - 1)"-* du = {-(- - e<"- '} / um -l (1 - )"- du
Jo Jo
. r(m + )I> + )
= \,/ -
Z
2t COS Z>7T '
.
r(m + /;+l)
Now both sides of the equation
"
J r (w + v + i)
are analytic functions of v for all values of v) and so, by the general theory of
analytic continuation*, this result, which has been proved when (v + ) > 0, R
persists for all values of v.
* Modern Analyst*, 5-5. The reader will also find it possible to obtain the result, when
Ii (" + )~-0, by repeatedly using the recurrence formula
Jo ++J Jo
which is obtained by integrating the formula
the integral is then expressed in terms of an integral of the same type in which the exponent of
u - 1 has a positive real part.
6-1] INTEGRAL REPRESENTATIONS 163
J a m-o (2m)!r(j/+m + l)
= 2" +i i r (i) r (i + $> cos i/tt . </ (z).
Next let us consider the second type of contour. Take the contour to lie
wholly outside the circle 1 = 1, and then (t 2 l)" -i is expansible in a series
1
|
m=o m ! I (| - i')
But
J oo i . J aa exp ia
* If v -J is a negative integer, the simplest way of evaluating the integral is to calculate the
residue of the integrand at u= 1.
tzt
t To justify the term-by-term integration, observe that /
'
| e dt l
is convergent ; let
J oot
-^
its value be A'. - l)"
Since the expansion of it follows that, wheu we
(t
2
converges uniformly,
are given a positive number
an integer e,independent of t, such that the remainder
we cau find M
after M
terms of the expansion does not exceed e/K in absolute value when M ^ We then M .
have at once
I
J cot jnir{-i>) J aci
=o
f(-l+.l-+ > ,izt
<eK--11 V '\e tzt dt\=e, '
J cot
aud the required result follows from the definition of the sum of an infinite series.
be deformed into the second of the two contours just considered. Hence the
analytic continuation of J- V (z) can be defined by the new integral over an
extended range of values of arg z; so that we have
r^-\,>" n7V/'(-i+,i+)
f
where arg z has any value between \ir + to and \ir + co.
When R (z) > and ^( + J)>0we may take the contour to be that shewn in Fig. 3,
\/
/\
Fig. 3.
in which it is supposed that the radii of the circles are ultimately made indefinitely small.
By taking each straight line in the contour separately, we get
J-> {z)=
T{ Y ei
" e
' sni(v ~
h) (1 -^" * i
K^n\) [/I
+ Je * e-3rC-
4 J) (1 - #)-\ dt
On bisecting the third path of integration and replacing t in the various integrals by it,
-t, t, t, it respectively, we obtain a formula for J- V (z), due to Gubler* which corre-
sponds to Poisson's integral for Jv (z) ; the formula is
2{ "
(4) J_ v (s)= ,
\fW 7Y-J S mvn f
e-'til + tty-hdt+f'cosizt + v^.il-fiy-idt],
r (v+#) "($) |_ jo Jo J
and, if this be combined with Poisson's integral, it is found that
Taking R (z) > 0, let us modify the two contours of 6*1 into the contours
Fig. 4. Fig. 5.
By making those portions of the contours which are parallel to the real
* Ziirich Vierteljahrsschrift, xxxm. (1888), p. 159. See also Graf, Zeitschrift fur Math, und
Phys. xxxvin. (1893), p. 115.
t Journal fUr Math, lxxvi. (1873), p. 9. Cf. Hayashi, Nyt Tidsskriftfor Math. xxni. b, (1912),
pp. 8690. The formula was examined in the case v=0 by Escherich, Monatshefie fiir Math,
und Phys. m. (1892), pp. 142, 234.
166 THEORY OF BESSEL FUNCTIONS [CHAP. VI
axis move off to infinity (so that the integrals along them tend to zero), we
obtain the two following formulae:
0) J.{z)
2ttiT(i)
d+)
e
izt
(t
2 - l)"-i dt + e izt (t2 - iy-i dt
_J l+i J -1+ooi
(2) J _{*)-
2 7nT()
r(i+) f(-i+) 1
e
rzt
(t
2
- ,
(3) J-.(*)-
2r(i)
/(i+)
d+) .
f(-i-) -i
1
e
izt
(t
2 - 1)"-* dt
,
+ e~ vin e
izt
(t
2
- 1)"-* dt
J 1+ooi J -1+ooi J
In the last of these integrals, the direction of the contour has been reversed
and the alteration in the convention determining the phase of t
2 1 has
necessitated the insertion of the factor e~ 2{y ~^ )H .
On comparing equations (1) and (3) with 3'61 equations (1) and (2), we
see that
(4)
T(\-v).(\z y[^ .
(5)
unless v is an integer, in which case equations (1) and (3) are not independent.
We can, however, obtain (4) and (5) in the case when v has an integral
value from a consideration of the fact that
(n), all the functions involved are
continuous functions of v near v n. Thus
dt
^(^-i)"-*^
7r * * <.) *exp(-t)
provided that, in both (6) and (7), the phase of z lies between \ir + <u and
Representations are thus obtained of H^ (z) when arg z has any value
between ir and 2tt, and of H v
{z>
(z) when arg s has any value between
2tt and jr.
If co be increased beyond the limits stated, it is necessary to make the contours coil
round the singular points of the integrand, and numerical errors are liable to occur
in the interpretation of the integrals unless great care is taken. Weber, however, has
adopted this procedure, Math. Ann. xxxvn. (1890), pp. 411 412, to determine the for-
mulae of 362 connecting HM V ( - z), HW V ( - z) with HJ (z), H V V) (*).
Note. The formula 2i*F (z) = ffvW(z) Hv V)(z) makes it possible to express Yv (z) in
terms of loop integrals, and in this manner Hankel obtained the series of 3'52 for
Y(); this investigation will not be reproduced in view of the greater simplicity of
Hankel's other method which has been described in 3-52.
In the formula 6'11 (6) suppose that the phase of z has any given value
between tr and 2tt, and define by the equation
arg z = a) + @,
so that \ir<fi< \tr.
where the phase of 1 + \iujz has its principal value. Again, if /3 be a given
acute angle (positive or negative), this formula affords a representation of
H w (z) valid over the sector of the s-plane in which
the contours into the line joining the origin to oo exp i/3, taken twice ; for the
integrals taken round a small circle (with centre at the origin) tend to zero
with the radius of the circle J.
On deforming the contour of (1) in the specified manner, we find that
(4) ff.(,).(l.)
T?
_r Jt
^^ (!_) *,,
f f .... The results may also be obtained for such values of v by expanding the integrands
,
,
The general formulae (3) and (4) ar_e of fundamental importance in the
discussion of asymptotic expansions of Jv (z) for large values of \z\. These
applications of the formulae will be dealt with in Chapter vn.
A useful modification of the formulae due to Schafheitlin. If we take
is
arg z = /3 (so that arg z is restricted to be an acute angle), and then write
u 1z cot 0, it follows that
W
( 6)
* To
W=
H<*(,)
"
, r(y+1)r(|)
^ ^
T(ir+J)r(i)Jo
obtain this formula, write
sin
!
eoe-lg.ar**-*
o
s^h,,
2 " +1
e dtf >
t + 1 = 6"^
z" 1
( - u), t - l = 2e w (1 - iujz).
^ F to
2 -*- - f**
cos-* . cos ( - vO + \6) ^ cntede
These formulae, which are of course valid only when R (v + 1) > 0, were
applied by Schafheitlin to obtain properties of the zeros of Bessel functions
( 15'32-*-15*35). They were obtained by him from the consideration that the
expressions on the right are solutions of Bessel's equation which behave in the
appropriate manner near the origin.
occurring in (3) and (4) when fi = v, has been studied in some detail by Nielsen, Math. Ann.
lix. (1904), pp. 89102.
The integrals of this section are also discussed from the aspect of the theory of asymp-
otic solutions of differential equations by Brajtzew, Warschau Polyt. Inst. Nach. 1902,
nos. 1, 2 [Jahrbuch iiber die Fortschritte der Math. 1903, pp. 575677].
^A-'J
<rS,\ r,^<
l'^
Fig. 6.
HP (w) = r (
^
-.{?7fr^- (1 " e-^-^) f>< (* - 2 1)"-* dt,
r
#<*> (as) =- ^~$;$^ (1 - e2 <"-i>) f
te < ( - l)r~* dt,
w
,o\ r
" w " r(i-.or(*).(iy
/ v 2 f"
J i
ain (xt).dt
(*
2
- i)- + *
When R (*) > and R (v +) > 0, it is evident from formula 6-11 (6) that
r(v+i)r(i)
(1+ ^ i
When v +\ is not a positive integer, the last expression may be regarded as a symbolic
representation of HJV> (z), on the understanding that/ (D) (e^'jz) is to be interpreted as
Afci
i I e M /(it)dt.
J i+i
Consequently
(1) ^ (1)
(?)=
r(v+i)r(i)
(l+^) * ; ,
and similarly
(2)
so that
#< , m
W - r(+J)r(j)^ (i +i^- 5Elti22L'
jr
'-f5lrIB (1+1 ^-..
< 3> '
The series obtained from (4) by expanding in ascending powers of D does not converge
unless it terminates; the series obtained in a similar manner from (3) converges only
when R(y)>\.
The expressions on the right of (3) and (4), with constant factors omitted, were given
by Hargreave, Phil. Trans, of the Royal Soc. 1848, p. 38 as solutions of Bessel's equation.
The exact formulae are due to Macdonald, Proc. London Math. Soc. xxix. (1898), p. 114.
An associated formula, valid for all values of v, is derivable from 6'11 (4). If n is any
positive integer, we see from the equation in question that
H W(z) = T ^~
" U+)
v) z)
(<2-l)"--l(-).(l + Z)2)e*<fc
7rtr ;^ f
v
C?/ J l + o
so that
= (-)"r(W)(^ p {ti _ iy - n . iiztdtf
This result, proved when R (z) > 0, is easily extended to all values of z by the theory of
analytic continuation ; it was discovered by Sonine, Math. Ann. xvi. (1880), p. 66, when
v = r and used by Steinthal, Quarterly Journal, xvm. (1882), p. 338 when v = n+%; in the
case when v = n+% the result was given slightly earlier (without the use of the notation
of Bessel functions) by Glaisher, Proc. Camb. Phil. Soc. ill. (1880), pp. 269271. proof A
based on arguments of a physical character has been given by Havelock, Proc. London
Math. Soc. (2) II. (1904), pp. 124125.
If we take <o = \tt in 6*1 (3) and then replace z by iz, we find that, when
i
arg z |
< \tt,
4
-1
Fig. 7.
*2^r<i)
2
[
(1 " <^> i?~* ( ' ~ *
r * dt
- "" - <2)"-* dt
+ i (e + e-3vni ) [ e~zt (1 ,
* Ann. dt Mat. (2) i. (1868), pp. 239241. Schlafli obtained the results (1) and (2) directly
by the method of 6-1. t Cf. Serret, Journal de Math. ix. (1844), p. 204.
172 THEORY OF BESSEL FUNCTIONS [CHAP. VI
(4) K v (,) = (*
2
- I)""* dt,
^j^ffi J"*-*
whence we obtain the formula
T
(5) K v (z) = ^) ^iT [V* c08h< sinn^A,
v
>
I
arg * < i i
77"' tne integral for H ^ (xze^) derived from
(
in the form
r( + *)-(i*>- (1+) *
H
n_v> ^xze _
(xze )-
v
^ ^T(^) f
J +00 (^-l)-+i-
Now, when R(v)^ \, the integral, taken round arcs of a circle from p to
p e
j"-t-*arg Z)
tends to zero as p <x>
by Jordan's lemma. Hence, by Cauchy's
,
theorem, the path of integration may be opened out until it becomes the line
_
=
-*" r (v + ) . (%xz)~ v r xi z'
e~ xzt dt
2TQ) )i,z (* -l)" +
2 *
r (V 4- 1) . (2z) v f*
00
f
_e
e-" du
(i) J->(^
and so we have Basset's formula
Kl) ^ v \xz)-
X V (\) (u2 +z 2 v+ i'
] o )
* The was examined in the case i=0 by Biemann, Ann. der Physik und
integral on the right
Chemie, 130139.
(2) xcv. (1855), pp.
The on the right in (1) was studied by numerous mathematicians before Basset.
integral
Among these investigators were Poisson (see 6'32), Journal de V&cole Polytechnique, IX.
(1813), pp. 239241; Catalan, Journal de Math. v. (1840), pp. 110114 (reprinted with
some Mem. de la Soc. R. des Set. de Liege, (2) xn. (1885), pp. 26 31); and
corrections,
Serret, Journal de Math. vm. (1843), pp. 20, 21; ix. (1844), pp. 193210; Sohlonrilch,
Analytischen Studien, n. (Leipzig, 1848), pp. 96 97. These writers evaluated the integral
in finite terms when v+$ is a positive integer.
Other writers who must be mentioned are Malmsten, K. Svenska V. Alcad. Handl. lxii.
(1841), pp. 65 -74 (see 7'23) Svanberg, Nova Acta Reg. Soc. Sci. Upsala, x. (1832), p. 232
;
Leslie Ellis, Trans. Camb. Phil. Soc. vm. (1849), pp. 213215 Enneper, Math.
Ann. xi. ;
Journal, xx. (1885), pp. 250260; and Oltramare, Comptes Rendus de V Assoc.
Francaise ' '
/ cos xu du .
_ (-) n-1 jr r d"' 1
/e~ xs ^P\~\
(u*+z*)* ~ 2**-. (n-1)! [dp^
o \~^fr~)] p=1
(~)w
~ iff ~
r d*~ e xtp i i
z* eia Tdt.
Jf a
T be
taken to be the Legendre function Q_j (t), the contour may start
If
and end at + ootexp( ia>), where w
is an acute angle (positive or negative)
First consider
J oo iexp (tui)
where the phase of t is zero at the point on the right of t =1 at which the
contour crosses the real axis. Take the contour to lie wholly outside the circle
|
|
= 1 and expand Q v -{t) in descending powers of t. It is thus found, as in
the similar analysis of 6*1, that
(1) J v (*) = K
* }
u #* Q.-i (0 dt,
7rlP Kt)
,
xiexp(-iw)
and therefore
*
(2) J_ r (,) = "*> S* Q__j (I) <fc
vr 1 ^f^ J o&iexp(-i<o)
If we combine these formulae and use the relation* connecting the two
kinds of Legendre functions, we find that
Again, consider
this is a solution of Bessel's equation, and, if the contour be taken to lie on the
right of the line R (t) = a, it is clear that the integral is [z* exp (a\z ))} as
J oo * exp ( i)
cf. Hobson, Phil. Tram, of the Royal Soc. clxxxvii. (1896), p. 461.
6-2] INTEGRAL REPRESENTATIONS 175
(4) ^ a,
<*)-
WJrm I
**Q.-Mdt,
H - W ra
'
(5) 9 (z) v e'* &- (*> dt >
(-M
4* P,_ x* (0
V <fc;
The integral which differs from (6) only by encircling the point +1 instead
of 1 is zero since the integrand is analytic inside such a contour.
In (5) and (6), arg (t+ 1) vanishes where the contour crosses the real axis
on the right of 1, and, in (5), arg (t 1) is ir at that point.
* Mathematical Collection, v. (Moscow, 1870) ; Math. Ann. xvi. (1880), pp. 929.
t Ann. di Mat. (2) v. (1873), p. 204. His memoir, Math. Ann. m. (1871),
pp. 184149, should
also be consulted. In addition, see Graf, Math. Ann. lvi. (1903),
pp. 423432, and Cbessin, John*
Hopkins University Circulars, xiv. (1895), pp. 20 21.
X Cf. Modern Analysis, 12-22. Cf. Modern Analysis, 532, 4-44.
.
*W-^n '-pI'-sI*
,
(i)
When |
arg z \ < \ir, we may swing round the contour about the origin until
it passes to infinity in a direction making an angle arg z with the negative
real axis.
1 f(+) 1\) f
ni \
-ni
Fig. 8.
4
If we write t + iri for w on the sides parallel to the real axis and id for w
on the lines joining to + iri, we get Schlafli's generalisation ofBessel's integral
(4)
l
Jv (z) = - J'cos {yd - z sin 0) d$ - ^^ J" e
-t-z**t dt>
If we make arg z ~- \tt, the first integral on the right is continuous and,
if R (y) > 0, so also is the second, and J v {z) is known to be continuous. So (4)
is still true when z is a pure imaginary if R{v) is positive.
the contour in (3) may be replaced by one which goes from oo (w yfr) i to
oo+ (V + yjr) i. By taking the contour to be three sides of a rectangle with
corners at oo {ir ^) i, (ir ^jr) i, (ir + ^r) i, and oo + (tt + yfr) i, we obtain,
as a modification of (4),
*
(6) J, (*) =1 f cos(i/0-*sin0)d0
""Jo
1 f g-zrinht
+ I *.*-,,(
gin (s^gjj C08 ^ - $ 1/W - 1^) (ft,
provided that |
arg z |
is less than both <ty
and tt ^.
When J? (i) > and z is positive (= x), we may take ^= in the last
formula, and get*
(7) J (#)=-! cos (i/0 - x sin 0) d0 + - I *->* sin (a? cosh * - Ai/tt) dt.
I? J TT J |)
in which c may have any positive value ; this integral is the basis of many of
Sonine's investigations.
Integrals which resemble those given in this section are of importance in the investiga-
tion of the diffraction of light
by a prism see Carslaw, Proc. London Math. Soc. xxx.
;
(1899), pp.121161 W. H. Jackson, Proc. London Math. Soc. (2) I. (1904), pp. 393414;
;
Whipple, Proc. London Math. Soc. (2) xvi. (1917), pp. 94111.
6*21. Integrals which represent functions of the second and third kinds.
we find that
ir Y (z) = cot vir I cos (vO zamd)dd- cosec vir I cos (v$ + z sin 8) dO
J o Jo
/"oo />
cosv7r/ e- t
-ZBinht
dt-l e**-""** dt.
Jo Jo
* Cf. Oubler, Math. Ann. xlii. (1897), pp. 583584.
178 THEORY OF BESSEL FUNCTIONS [CHAP. VI
ni
Fig. 9.
write t, i0, t + 7rt for w on the three parts of the contour ; we then see that
the expression is equal to
J
Too 1 fir g vni foo
_ . yt-zsmhtfa +_ /
eHzsiiiO-ye) 0 + _ e -v-sinh <ft,
7rl.'o '"Vo 7TI Jo
and this is equal to Jv (z) + 1 F (s) from formula (1) combined with 62 (4).
Hence, when arg z < \tr, we have | |
feo+iri
(2) # {1)
(*) = A 1
e*
Mnh W-' M dw,
'
(3) #<*>(*)=- .
f
"V *""""^.
811
Formulae equivalent to these were discovered by Sommerfeld, Math. Ann. xlvii. (1896),
pp. 327
357. The only difference between these formulae and Sommerfeld's is a rotation
of the contours through a right angle, with a corresponding change in the parametric
variable; see also Hopf and Sommerfeld, Archiv der Math, und Phys. (3) xvm. '{1911),
pp. 1-16.
By an obvious change of variable we may write (2) and (3) in the forms
the contours are those shewn in Fig. 10, emerging from the origin and then
bending round to the left and right respectively results equivalent to these ;
Fig. 10.
[Note. There is no difficulty in proving these results for integral values of v, in view
of the continuity of the functions involved ; cf. 6*11.]
We proceed to modify the contours involved in (4) and (5) to obtain the
analytic continuations of the functions on the left.
If o> is an angle between tr and rr such that |
to arg z < %ir, we have
|
and
1 /> exp (-)* ( / is")
(7)
^"W-sL,,. ^p{*.(-J]
du,
the contours being those shewn in Fig. 11 and Fig. 12; and these formulae
give the analytic continuations of the functions on the left over the range of
values of z for which to \tr< args< m + ^rr; and o> may have any value
between* v and ir.
the phase of u.
180 THEORY OF BESSEL FUNCTIONS [CHAP. VI
(8) H^ v (Z) =
g^vwi foo+Jirt
r-
TTl J _ao-Jiri
e i*cosh-vu>
fa
/"oo+Jiri
eizcosh cogh vw dW)
. v>
J
TTl Jo
p\vni j'ao ^iri
(9) H (z) = -^\
Wl J -ao+tni
e-izcoshu>-*w dw
Qghrxi ( Jtt*
= :- e
-izcosh w
^h vW # ^ Wj
7rt Jo
provided that arg z | \
< \tt.
Formulae of special interest arise by taking z positive (=#) in (6) and (7)
and 1 < R (*>) < 1. A double application of Jordan's lemma (to circles of
large and small radius respectively) shews that, in such circumstances, we may
take = %-rr in (6) and to = \tt in (7). It is thus clear, if u be replaced by
a>
+ ie l , that
p\vai foo OgJnrt Too
2 r
(13) Yv (x) = I cos (# cosh vtt) . cosh vt.dt;
TTJo
and, in particular (cf. 6*13),
/ta\ r / \ 2 fsinarf.<ft
(14) *<*>-; j, v^T)'
/, -. TT , . 2 f " cos # . <fo
(15)
^'"iJ.W^U'
when we replace cosh t by .
The two formulae are due to Mehler, Math. Ann. v. (1872), p. 142, and Sonine,
last
Math. Ann. xvi. (1880), p. 39, respectively ; and they have also been discussed by Basset,
Proc. Camb. Phil. Soc. "in. (1895), pp. 122128.
A slightly different form of (14) has been given by Hardy, Quarterly Journal, xxxn.
(1901), pp. 369384; if in (14) we write x=2 y/(ab), xt=au+b/u, we find that
(16)
f"
n (+) ^= "A {2 J{ab)}.
Note. The reader will find it instructive to obtain (14) from the formula
(1).
2 /"
j v{2(co
sin (
)} ^^ + )*
The modifications of the previous analysis which are involved in the dis-
cussion of Iv {) and K v (z) are of sufficient interest to be given fully; they are
due to Schlafli*, though he expressed his results mainly in terms of the
function F(a, t) of 415.
The analysis of 62 is easily modified so as to prove that
/-<>- (
&L'-*(' + s)*
and hence, when arg z < | |
\ir,
<2
> ''W- s?/.T u"" exp *(* +
1)}
du-
f
1 foo -Hri
/ (z)=er-. ^coshw-wr fa
(3)
ITTIJ aa-ni
The formulae (2) and (3) are valid when arg 2 = $-rr if R(v) > 0.
If in (3) the contour is taken to be three sides of a rectangle with corners
at 00 in, tri, in, 00 + iri, it is found that
and hence
(6)
2ttc 2i sin vtt
when 7r < &>< 7r and \ir + a> < arg z < \tt + a>.
Similarly
sinv7T /""expOr-ult - " -1 f
/ Jx")
(9) e"" /_ (*) - e~ vwi Iv (z) m exp -Us ( u + - )\ du
T J Oexp (-+) ( V tt/J
;
this is valid when < a>< 2ir and $tt + &> < arg z < ^7r + eo.
The contours for the formulae (8) and (9) are shewn in Figs. 13 and 14
respectively.
Further, when z is positive (= x) and 1 < R (v) < 1, the path of integra-
tion in (8) may be swung round until it becomes the positive half of the
imaginary axis ; it is thus found that
so that
j'*
(11) JSrr (*)=*J*'
<
<ri*sinht+,t (fc i
. .
(13) K v (x) =
COS|l/7Tj
^ I cos (a; sinh t) cosh j/ . dt,
and these formulae are all valid when x > and - 1 < R (v) < 1.
In particular
(14) jr.<.)-(*.<ho*-["!=|2fc*
.'o Jo v(f + -U
a result obtained by Mehler* in 1870.
It may be observed that if, in (7), we make the substitution ^ze t =T , we find that
(15) jr.W-K^^expj-r-^}^,
provided that ) R (z2 > 0.
The integral on the right has been studied by numerous mathe-
maticians, among whom may be mentioned Poisson, Journal de VEcole Polytechnique, ix.
(cahier 16), 1813, p. 237; Glaisher, British Association Report, 1872,
pp 1517; Proc.
Camb. Phil. Soc. in. (1880), pp. 512; and Kapteyn, Bvll. des Sci. Math. (2) xvi. (1892),
PP 4144. The integrals in which v has the special values \ and f were discussed by
Euler, Inst. Calc. Int. iv. (Petersburg, 1794), p. 415; and, when v is half of an odd
integer, the integral has been evaluated by Legendre, Exercices de Calcul Integral, I. (Paris,
1811), p. 366; Cauchy, Exercices des Math. (Paris, 1826), pp. 5456; and Schlomilch,
Journal fur Math. xxxm. (1846), pp. 268280. The integral in which the limits of inte-
gration are arbitrary has been examined by Binet, Comptes Rendus, xn. (1841),
pp. 958
962.
The integrals
f" a? f* a?
I sin t sin P" 1
dt, cost. cos f-.
1
dt,
Ja t Jq t
in which x > 0, 1 < R(v) < 1, have been examined by Hardy f as examples
of Du Bois Reymond's integrals
m t.t~*dt,
f(t)*
/,o cos
t
.
sin f' .
1
dt = x" f
I sin (a?e*) sin (xe~ l ) e vt dt .
JO t J _
__ in 1 fgitzcosht i g 2&ccosht girfzsinht g2tassinhtJ git
fa
./ 00
and similarly
"K2 "ITT"
(2)
f
lob cos ^ . r- 1
eft = , , (X (2a?) - Jv (2a?) + /_ (2a?) - / (2a?)].
When v has the special value zero, these formulae become
curious extension of Jacobi's formulae of 2*2 has been obtained in the case of J {x)
A
and Jx (x) by Theisinger, Monatshefte fur Math, und Phys. xxiv. (1913), pp. 337 341 we ;
shall now give a generalisation of Theisinger's formula which is valid for functions of order
v where -\<v<\.
If a is any positive number*, it is obvious from Poisson's integral that
2 " >-aii.
j9 < X) = flf)" ( C os (x cos 0) sin2 " 6dd
+ 2
(y f
**
(l - e-enn#) cos (# cos 0) sin 2 " 0tf0.
1-expJicm^-l/i)}
'
J
ft
-i~~5nh{i (*-V)}
sml1
^ "^(*jzl!*f* 2f J *'
where the contour passes above the origin. Take the contour to be the real axis with an
indentation at the origin, and write z= tan < on the two parts of the contour; we thus
find that the last expression is equal to
Jo
^
exp(-cmcot<6)
y
sin(.rcot<)
, ,.
.
sin (x
x
,
tan id>)
"^
, ,_..
.
ml cot2" <h
"*
., , d<b
-^-r
^sin<p
+ .[**
J
r**
o
1 - eXl (a
sin ;
(.r cot ^ <)
j\
</>)
sin <*tan
v
j
W^ .e-cot<j>\
2
^sin0
sin (.r tan i<A) . rf<
f vir)
. .
/i
=4,
I sm
Vi
j.
f/ cos (Max
(iax cot <6) vs cot
/i i.
<f> \.
2
r-^V cot " <f -r-1-^ ,
J
/ sin(#cot</>) ^sin#'
* In Theisinger's analysis, a is an even integer.
.
and therefore
r(y+
(1)
^ J v ( A.) [
**
e -.in C08 (-p cos 6) sin*" 0d0
+ 2
/**"
/
Jo
sin (iax
2
/i
cot
x j\
d>)
/i AJ
vi
.sin (a? tan Ad>)
cos (4<m; cot <b - vn) r-^-.
,,,
cot8 " d>
sm (j;cotd))
^~ ^sind)
dd>
^-^
The transformation fails when v >, because the integral round the indentation does
not tend to zero with the radius of the indentation. The form given by Theisinger iu the
case v = \ differs from (1) because he works with 3-3 (7) which gives
r(-j)r(j) ""2
(2)
2 Z-r - ^ {*) = ( ** e-"infl sin (x cos 6) sin
2
6 cos 6d6
+4a f**
/
,1
sm (lax x jx /i j
cot <6) cos (*a# cot <A - vir)
i.
.
sin 2 (Aar tan Ad>)
r-f ,
2
?Z cot "
,,_.,,
' d>
cW>
-r-~ ,
Three different types of integrals which represent K v (z) have now been
obtained in 6'15 (4), 622 (5) and 616 (1), namely
K iy ' h dt
'
(,,
'i|frr e~* {t *~
-
= f e zcoaht cosh vt.dt,
Jo
The equality of the first and second was directly demonstrated in 1871 by
Schlafli*; but Poisson proved the equivalence of the second and third as early
as 1813, while Malmsten gave a less direct proof of the equivalence of the
second and third in 1841. We proceed to describe the three transformations
in question.
We first give an abstract of the analysis used by Schlafli, Ann. di Mat. (2) v. (1873),
which arises from a comparison of two of the integral representations of K v (2), and which
may be established by analysis resembling that of 2*323.
We have, of course, to suppose that R(z)>0 to secure convergence, and it is convenient
at first to taket -<R(v)<l.
* An earlier proof is due to Kummer, Journal filr Math. xvn. (1837), pp. 228 242, but it is
on expanding the last factor of the integrand in powers of u and integrating term-by- term.
2"
-* ooBh^cM-?!^^ ("c-'^flsinhdsinh^dd
w"il^
r (l) f"
jo r(J) J
= 7^
r(l-i'), Jo
/ tf-'w'MtfsuihdcM
dtdx
-w=i)l'!'"^r*-
1}
r(i-r)Ji J/ (
' *(*-<)*
the inversion of the order of the integrations presents no great theoretical difficulty, and
the transformation is established.
I f c
-~^- ft_
f
I>+*)-(2*)" r cos(*M )rftt
isdue to Poisson* Journal de VEcole Polytechnique, ix. (1813), pp. 239241. The equation
is true when |argz|<r, x>0 and R(v)> -, but it is convenient to assume in the
course of the proof that R (v) > | and arg z < w, and to derive the result for other values
| |
. ,. f coa(xu)du 4r(A) f .
, , ,, , ,, , ,
* See also Paoli, Mem. di Mat. e di Fit. della Soe. Italiana delle Sci. xx. (1828), p. 172.
6-32, 6-33] INTEGRAL REPRESENTATIONS 187
Jo Jo
/OO /"OO
I
{exp ( - 2) cos #m dv) . .
*"- J exp ( ss2 ) ds,
o Jo
when we write ts (uP+z*) and change the order of the integrations*.
r(v + ^ )
/o"^Spi ==ir(* )
/o"'
, '"
IeXp{ -^- i ^ <i8
v(2z])"J
which establishes the result.
[Note. It is evident that 8=\xe~ tlz=\v1 lvlz. The only reason for modifying
i r
by taking v as a parametric variable is to obtain an integral which is ostensibly of the
same form as the integral actually investigated by Poisson ; with his notation the integral is
The method employed by Malmstent in proving that, when R(z)>0 and R {?)> -i,
then
m coB(xu)du r(i)(^)- m
r(+*) _ [ - v*, * - idt
[ c t
{l 1Y
X) ah
(**)T(J);. (*+^)" + i~ ro+*) Ji
isnot so direct as the analysis of 6*31, 6*32, inasmuch as it involves an appeal to the
theory of linear differential equations. It is first shewn by Malmsten that the three
expressions
and that as x-*~ + oo , the third is (**) while the first and second are bounded, provided
that R (v) > 0. It follows that the second and third expressions form a fundamental system
of solutions of the equation
S- <*->-*'-*
* Cf.Bromwich, Theory of Infinite Series, 177.
t Svenska V. Akad. Handl. lxii. (1841), pp. 6574.
A'.
(cos xu) du
/;
where V is independent of x. To determine C, make x-*-0 and then
and the required transformation follows, when R(v)>0, if we use the duplication formula
for the Gamma function.
c a gtt
An immediate consequence of Malmsten's transformation is that / ? .'
'
- is
7o (*+82 r
expressible in finite terms ; for it is equal to
n-i (2^)(2n-m- 1)
(2s)"-*(-l)! m=0 m.'(ra-ro-l)! *
This method of evaluating the integral is simpler than a method given by Catalan,
Journal de Math. v. (1840), pp. 110 114; and his investigation is not rigorous in all its
stages. The transformation is discussed by Serret, Journal de Math. vnr. (1843), pp. 20,
21; ix. (1844), pp. 193216; see also Cayley, Journal de Math. xn. (1847), p. 236
[Collected Papers, i. (1889), p. 313.]
The integral
* Tram. Camb. Phil. Soc. vi. (1838), pp. 379 402. Airy used the form
t The result was communicated to Airy on March 11, 1818 ; see Trans. Camb. Phil. Soc. vni.
(1849), pp. 595599.
X Quarterly Journal, xxxv. (1904), pp. 2266 ; Trans. Camb. Phil. Soe. xxi. (1912), pp. 1 48.
6'4] INTEGRAL REPRESENTATIONS 189
**-<>,
as!
and he also obtained the asymptotic expansions of the integral for large
values of x, both positive and negative.
As was observed by Stokes (foe. cit. p. 187), this differential equation can be reduced
to Bessel's equation ; cf. 4*3 (5) with 2g = 3. The expression of Airy's integral in terms of
Bessel functions of orderst was published first in a little-known paper by Wirtinger,
Berichte dea natur.-med. Vereins in dnruibruck, xxm. (1897), pp. 7 15, and later by
Nicholson, Phil. Mag. (6) xvn. (1909), pp. 617.
Subsequently Hardy, Quarterly Journal, xli. (1910), pp. 226 240, pointed out the con-
nexion between Airy's integral and the integrals discussed in 6*21, 6*22, and he then
examined various generalisations of Airy's integral (.10*2 10*22).
To evaluate Airy's integral J, we observe that it may be written in the form
The integrals along these arcs tend to zero as p -*- oo , by Jordan's lemma, and
hence, by Cauchy's theorem, we obtain Stokes' transformation
(<*>
1 / co expert
cos (t* xt) dt = o exp (? ixt) dt
Jd Zjooexpfiri
=\ I* {^
Ti
exp (- t **'T) + -** i exp(- t 3 e-^xr)} dr;
~ Jo
the contour of the second integral consists of two rays emerging from the
origin and the third integral is obtained by writing re*", re**1 for t on these
rays.
Now, since the resulting series are convergent, it may be shewn that
/,
exp(-T + ^a!T)dT=
w=0
2 ^ m ;
-
.'o
T*exp(-T)<*T,
* Trans. Camb. Phil. Soc. ix. (1856), pp. 166187. [Math, and Phys. Papers, h. (1883),
pp. 329349.] See also Stokes' letter of May 12, 1848, to Airy, Sir G. O. Stokes, Memoir and
Scientific Correspondence, n. (Cambridge, 1907), pp. 159 160.
+ For other occurrences of these functions, see Rayleigh, Phil. Mag. (6) xxvm. (1914),
pp. 609619; xxx. (1915), pp. 329338 [Scientific Papers, vi. (1920), pp. 266275; 841349]
on stability of motion of a viscous fluid; also Weyl, Math. Ann. Lxvm. (1910), p. 267, and, for
approximate formulae, 8*43 infra.
X The integral is convergent. Cf. Hardy, toe. cit. p. 228, or de la Vallee Poussin, Ann. de la
Soc. Sci. de Bruxelles, xvi. (1892), pp. 150180.
Bromwiob, Theory of Infinite Series, 176.
190 THEORY OF BESSEL FUNCTIONS [CHAP. VI
and so
| cos (<* +
~ art) eft = 2 '? rm exp
r\(- t*) dr
Jo m =o ml Jo /
"* + **
U-omir(m + |) mt.iti!r(m + |)J
This is the result obtained by De Morgan. When the series on the right
are expressed in terms of Bessel functions, we obtain the formulae (in which
a; is to be taken to be positive) due to Wirtinger and Nicholson
J ^ Z)e aS
- 27ri\ntoJoo 2.m\r(p + m + l)
'
Now suppose that R (v) > \, and choose the contour so that, on it,
5 r ( 2 m - g>
m =o2 .m!r(i/ + w + l) 2r'l
* Rend, del R. Istituto Lombardo, (2) xix. (1886), pp. 559562 ; Atti delta R. Accad. dei
Lincei, ser. 4, Rendieonti, iv. (1888), pp. 694700, 792799.
t Mellin has given a summary of his researches, Math. Ann. lxviii. (1910), pp. 305 337.
X Gamb. Phil. Tram. xx. (1908), pp. 270 279. For a bibliography of researches on inteRraU
of this type, see Barnes, Proc. London Math. Soe. (2) v. (1907), pp. 59 65.
'
The only poles of the integrand inside the contour are at 0, 1, 2, When
we calculate the sum of the residues at these poles, we find that
so that
(1) ^ ( ^ )e
- = _(ML- F i l (I
, + j ; 2 + i; -2iz),
M*) *= r *+ x yK ( +
{
(2) *-, 2" + l; 2iz).
These formulae, proved when B(v) > , are relations connecting functions
of v which are analytic for all values of v, and so, by the theory of analytic
continuation, they are universally true.
r(-*->r (-)(!)'+
qua function of s. It has poles at the points
5 = 0, 1, 2, ...; - v ,-v+l,-v + 2, ....
The residue at s m is
iri
v
(~)m (^Y +Sm
sin vnr ' ml r (v +m+ 1)
so that
(4) irew H v
w ( z) = -~lr(-v-s)r (- *) (- \izy*" ds,
where the contours start from and return to + oo after encircling the poles of
the integrand counter-clockwise. When |
arg iz \
< \ir in (3) or arg ( iz)
| \
< \tt
in (4) the contours may be opened out, so as to start from oo i and end at
oo i. If we reverse the directions of the contours we find that
c+ l
(5) 7re-*<" +1 "* H (z) = -U
I'm
f" r ( _ _ ,) r (- *) (ii>)' +M ds,
J -c-aoi
(6) vJW H p
o)
( z)
= -L r
frm]
e+x>%
-c-oot
r (_ _ S ) r (- s) (- \%zy*~ ds,
'.-K,$g&*>
and it is obtained in the same way as the preceding integrals ; the reader
will notice that, when j
s |
is large on the contour, the integrand is 0(\s |~* -1 )-
By using the duplication formula for the Gamma function we may write
the results just obtained in the form
m
UJ j (z) ** -
>rW 2ty*rJ.
^- r r^r(+ ^r^
-M+j).(2uy*
+ s+^sin**-*
y
in which the integrand differs from the integrand in (1) by a factor which is
periodic in s. It is to be supposed temporarily that 2v is not an integer and
that the path of integration is so drawn that the sequences of poles 0, 1, 2, ...
2v, 1 2v, 2 2i/, ... lie on the right of the contour while the sequence of
poles v \, v~\, v \, ... lies on the left of the contour. In the first
place, we shall shew that, if arg iz < f ir, the integral taken round a semi-
| \
circle of radius p on the right of the imaginary axis tends to zero as p * <x>
1
T(v + s+i).(2izy
g r(s + l)JT(2i/ + s+l)
~ pe i9
log (2iz) - {v + pe") (log p + id) + pe ie - ^ log (2tt) ;
and the real part of this tends to co when ^7r<6<\ir, because the dominant
term is p cos 6 log p. When is nearly equal to |7r, sin sir is comparable | \
with \ exp [pir sin 6 and the dominant term in the real part of the logarithm
\ j)
p cos 6 log 2z | |
p sin $ arg 2iz pcosd log p + pd sin 6 + p cos 6 - 2p7r|sin &\
.
Hence s times the integrand tends to zero all along the semicircle, and so
the integral round the semicircle tends to zero if the semicircle is drawn so as
to pass between (and not through) the poles of the integrand.
It follows from Cauchy's theorem that, when arg iz < \tr and 2v
J
j
is not an
integer, then
- ^r-
JLi^/tr j
Pr -oof
(- ) r (- 2* - ) r (* + + j) . ^y ds
may be calculated by evaluating the residues at the poles on the right of the
contour.
The residues of
r(-s)r(-2v-s)r(v+8 + i).(2izy
at s m and s = 2v + m are respectively
= ( 2,M r(, + i) +
sin 2vir I (2j/ + 1)
l
sin 2v7r '
i*->n)
e cog (,.) ^2^)"
(3) *,(*)--
7T*
*r r(-)r(-2v-*)r(i/+s + ^).(-2wyrf*.
There are really two aspects of the problem to be considered ; the investi-
gation when v is large is very different from the investigation when v is not
large. The former investigation is, in every respect, of a more recondite
character than the latter, and it is postponed until Chapter vm.
It must, however, be mentioned that the first step towards the solution of
the more recondite problem was made by Carlini* some years before Poisson 's -J"
investigation of the behaviour of J (x), for large positive values of x, was
published.
r z N
= / 2 \T , , , L 12 .3 2 l2 .3 8 .5 2 .7 a )
/o( *) COB g 1 + -"j
U) [
<
"^>-t -2l(8^ 4!(8*)'
l2 3 2 52
+ sin(*-i7r).j
(
IT^-3^ l2
+ ..
when x is large and positive. But, since the series on the right are not con-
vergent, and since Poisson gave no investigation of the remainders in the
series, his analysis (apart from his method of obtaining the dominant term) is
t Journal de VEcole Polytechnique, xn. (cahier 19), (1823), pp. 350 352 ; see l 6. An in- -
vestigation of Jv (x) similar to Poisson's investigation of J (x) has been constructed by Gray and
Mathews, A Treatise on Bessel Functions (London, 1895), pp. 34 38.
1
be seen in the course of this chapter that Poisson's series are asymptotic
It will ; this
has been proved by Lipschitz, Hankel, Schlafli, Weber, Stieltjes and Barnes.
It must be mentioned that Poisson merely indicated the law of formation of successive
terms of the series without giving an explicit expression for the general term such an ;
The analogous formal expansion for J x (x) is due to Hansenf and a few ;
years later, JacobiJ obtained the more general formula which is now usually
written in the form
gy^^
.
...JJ
These expansions for J (x) and Jx (x) were used by Hansen for purposes of numerical
computation, and a comparison of the results so obtained for isolated values of x with the
results obtained from the ascending series led Hansen to infer that the expansions, although
not convergent, could safely be used for purposes of computation .
Two years before the publication of Jacobi's expansion, Plana had dis- ||
covered a method of transforming Parse val's integral which placed the expansion
of / (x) on a much more satisfactory basis!". His work was followed by the
researches of Lipschitz**, who gave the first rigorous investigation of the
asymptotic expansion of Jo with the aid of the theory of contour integra-
(z)
tion ; Lipschitz also briefly indicated how his results could be applied to Jn (z).
The general formulae for Jv (z) and Y v (z), where v has any assigned (com-
and z is
plex) value large and complex, were obtained in the great memoir by
Hankelft, written in 1868.
* Some information concerning W. R. Hamilton's researches will be found in Sir George
Gabriel Stokes, Memoir and Scientific Correspondence, i. (Cambridge, 1907), pp. 130135.
+ Ermittelung der absoluten Storungen [Schriften der Stermoarte Seeburg], (Gotha, 1843),
pp. 119123.
Jacobi's result
t Astr. Nach. xxvm. (1849), col. 94. [Get. Math. Werke, vn. (1891), p. 174.]
is obtained by making the substitutions
n+1>cosx+(- l^-^sinx,
J2 cos(*- iir-iir) = - l)*<
. (
the addition of the corresponding formula for / (z), by Kirchhoff f and a little- ;
A close study of the remainders in the asymptotic expansions of J (x), T (x), I (x)
and Kq{x) has been made by Stieltjes, Ann. Sci. de VEcole norm. sup. (3) in. (1886),
pp. 233 252, and parts of his analysis have been extended by Callandreau, Bull, des Sci.
Math. (2) xiv. (1890), pp. 110 114, to include functions of any integral order; while
results concerning the remainders when the variables are complex have been obtained by
Weber, Math. Ann. xxxvu. (1890), pp. 404416.
The expansions have also been investigated by Adamoft, Petersburg Ann. Inst, polyt.
of integrals of Poisson's type. But Schlafli1[ has shewn that a large number of
results are obtainable by a peculiar treatment of integrals of Bessel's type,
while, more recently, Barnes** has discussed the asymptotic expansions by
means of the Pincherle-Mellin integrals, involving gamma-functions, which
were examined in 65, 6*51.
valid when -\ir<fi<\ir and |7r + j8< args < |7r + jS, provided that
(v-j)iu
+ (,,-fr)(-f).(^ + t
*"'
"*"
2z 270 s
but since this expansion is not convergent all along the path of integration,
we shall replace it by a finite number of terms plus a remainder.
For all positive integral values of p, we have*
1 + 2z) + (p-l)\{2iz) {1 tf 1 dt
\ -J -^TWz)
=0 J I 2iz)
ut
1 >sin8, |arg(l-^) <7T,
2iz
- (AY e**-**> Pi
1
(*-*>- r ( ,> + , +*>
H
*'w (z)
w-UJ 1L1
+ jKp
U-o ir(v + i).(Kf)- J'
where
[ 'J
when ^ is such that it + 28 ^ arg z ^ 2tt 28, 8 being any positive acute
angle ; and the symbol Bachmann-Landau symbol which denotes a
is the
function of the order of magnitude^ of z~* as z -* oo | \
The formula (1) is also valid when R (v - p |) > ; this may be seen by
* Cf. Modem Analysis, 5-41. The use of this form of the binomial expansion seems to be
due to Graf and Gabler, Einleitung in die Theorie der BesseFschen Funktionen, i. (Bern, 1898)
pp. 8687. Cf. Whittaker, Modem Analysis (Cambridge, 1902), 161; Gibson, Proe. Edinburgh
Math. Soc. xxxvm. (1920), pp. 6 9 ; and MaoBobert, ibid. pp. 10 19.
t Cf. Modem Analysis, 2-1.
:
provided that R (v + |) > and that the domain of values of z is now given
by the inequalities
- 2tt + 28 % arg z ^ tr - 25.
{4>v* - I s } {4* 2 -3 2
.. {tv* - (2m - l)2 }
= } .
2 2m .w!
these expansions become
(4) H, W = (A)*e-<-^g |^ + o
<)(*-*)] .
361 (7), which connect functions of the third kind of order v with the corre-
sponding functions of order v, that the restriction that the real part of v
exceeds - \ is unnecessary. So the formulae (1) (6) are valid for all values
of v, when z is confined to one or other of two sectors of angle just less than Sir.
y
of which (7) is valid when tt < arg z < 2-tt, and (8) when 2-ir < arg z < ir.
7*21] ASYMPTOTIC EXPANSIONS 199
/-n / i
2
Tt\ / VT[coe^-W-iir).^
i s ^
(-)w -(">2m)
(1) *(*)-[-) {2gym
/ i
i \ v (-)w -(^,2m + iy
-sin^-Ji/w-iir).^ J2zj^i
(-)w 2w)
(2) n W ~(iy sin(z \i vtt
/ i x v
$ir). 2.
m-0
.(f,
(2zfm
/i i x v (-)'*('. 2w+l)"
. .
( ^y w+1
( 4) ^.(ij^^w-wyt^-)
and (in the case of functions of integral order n only),
T.(.)~(
(5)
T) Ln(*-W-i).jS o (2 , )m
v
+ cos(^-^7r-|7r).^
o ( ^ )2m+1 J.
These formulae are all valid for large values of z provided that arg z
|
|
| |
< tt;
and the error due to stopping at any term is obviously of the order of magni-
tude of that term multiplied by \\z. Actually, however, this factor \\z may
be replaced by ljz2 this may be seen by taking the expansions of HJ$ (z) and
;
H v {z) to two terms further than the last term required in the particular
(2)
(1908), p. 60. The error appears to have originated from Todhunter, An Elementary Treatise on
Laplace's Functions, Lamfs Functions and Bessel's Functions (London, 1875), pp. 312 313.
1l 1 1 :
- l e-(f+)z M -i (2 - tu)-* du = 0,
o
and the analysis then proceeds on the lines already given; but in order to
obtain asymptotic expansions of a pair of solutions of Bessel's equation it
seems necessary to use a method which involves at some stage the loop
integrals discussed in Chapter vi.
Note. The method by which Lommel endeavoured to obtain the asymptotic expansion
of Yn (z) in his Studien, pp. 93 97, was by differentiating the expansions of J v (z) with
respect iX> v but of course it is now known that the term-by-term differentiation of an
;
It should be noticed that Lommel's later work, Math. Ann. iv. (1871), p. 103, is free from
the algebraical errors which occur in his earlier work. These errors have been enumerated
by Julius, Archives Nderlandaises, xxviit. (1895), pp. 221 225. The asymptotic expansions
of Jn (z) and Yn (z) have also been studied by McMahon, Ann. of Math. vni. (1894),
pp. 57 61, and Kapteyn, Monatsheftefiir Math, und Phys. xiv. (1903), pp. 275282.
t Ann. Sci. de Vticale norm. sup. (3) xxi. (1904), pp. 207'86b. 459534 ; Verh. des Int.
Math. Kongresses in Heidelberg, 1904, pp. 241 245.
% Acta Mathematica, xxxvn. (1914), pp. 113 140.
Quarterly Journal, xlvi. (1915), pp. 263283 ; Proc. London Math. Soc. (2) xv. (1916),
pp. 125.
w
7*22] ASYMPTOTIC EXPANSIONS 201
The formula 7-21 (1) for Jy (z) was established for values of z such that
|
arg z < 7r. If we took arg z to lie between
|
and 2ir (so that arg ze~ ni lies,
m-o \&ze )
and throughout the sector in which ir < arg z < 2rr, the function HP (z) has
the asymptotic expansion
W \irtj TO -o (2iz)m
( 2) iw>~)'.^^.sj$
VtTS/ TO _o (2t^) TO
The expansions (1) and (2) are both valid when < &rg z < tt ] now the
difference between them has the asymptotic expansion
where the constants c,, c2 have values which depend on the domain of values
assigned to arg z. And, if arg z is continually increased (or decreased) while
\z\'\% unaltered, the values of c x and c have to be changed abruptly at various
2
stages, the change in either constant being made when the function which
multiplies it is negligible compared with the function multiplying the other
constant. That is to say, changes in c, occur when I (z) is positive, while
changes in c2 occur when I(z) is negative.
It is not difficult to prove that the values to be assigned to the constants
Ci and g2 are as follows
<h = i^ c, = i**+iM
+1) +V"\ {v
[2P 7r < arg 2 < (2p + 2) tt],
where p is any integer, positive or negative.
The fact that the constants involved in the asymptotic expansion of the analytic function
Jv (z) are discontinuous
was discovered by Stokes in (March?) 1857, and the discovery was
apparently one of those which are made at three o'clock in the morning. See Sir George
Gabriel Stokes, Memoir and Scientific Correspondence, I. (Cambridge, 1907), p. 62. The
papers in which Stokes published his discovery are the folio wing t: Trans. Cavib. Phil.
Soc. x. (1864), pp. 106128; xi. (1871), pp. 412
425; Acta Math. xxvi. (1902), pp. 393
397. [Math, and Phys. Papers, iv. (1904), pp. 77 109 ; 283298 v. (1905), pp. 283-287.] ;
The third of these seems to have been the last paper written by Stokes.
The formula 7*2 (5) combined with equation 3*7 (8), which connects
K v (z) and if (l)
(iz), shews at once that
-(2z) e
L ~nr + 2i(8^
+
-J'
* Cf.Bromwich, Theory of Infinite Series, 133.
t Stokes illustrated the change witb the aid of Bessel functions whose orders are and ^
the latter being those associated with Airy's integral ( 6-4).
7*23, 7*24] ASYMPTOTIC EXPANSIONS 203
<o\ r , \
* v (-) w ( y
m) <r'+(,,+ *"'' 5 (", "0
* V (-) ) TO
e-'-<"+*>" m)
W
<*\ t t \
^~(2,r*)i m%> (2*) +
(>, ,
(2tt^ TO
| (y,
to(2^r'
provided that |7r < arg z < \ir.
The apparent discrepancy between (2) and (3) when z has a value for
which arg z lies between \tt and |7r is, of course, an example of Stokes'
phenomenon which has just been investigated.
The formulae are also given by Riemann, Ann. der Pkysik unci Chemie, (2) xcv. (1855),
p. 135, when v = 0. Proofs are to be found on pp. 496 498 of Hankel's memoir.
A number of extremely interesting symbolic investigations of the formulae are to be
found in Heaviside's* papers, but it is difficult to decide how valuable such researches are
to be considered when modern standards of rigour are adopted.
x[(2a)n + nC 1 (n + l).(2a)n- + n C. 1
i (n-rl)(n+2).(2ay-*-r...]
is obtained (cf. 6*3). This formula is written symbolically in the form
f
cos ax dx . ire~ a f _ 1
\2a + _1
Jo (1 +x 2
)
n+1 2 :W+1 .?i!( |>] j
It will be observed that this notation is different from the notation of 4*4.
* Proc. Royal Soc. lii. (1893), pp. 504529; Electromagnetic Theory, ii. (London, 1899). My
thanks are due to Dr Bromwich for bringing to my notice the results contained in the latter work.
t K. Svenska V. Akad. Handl. lxii. (1841), pp. 65-74.
X Quarterly Journal, xlii. (1911), pp. 329338.
:
be quoted here. The functions of zero order had been examined previously by
Russell*; he found it convenient to deal with the logarithms! of the functions
of the third kind which are involved, and his formulae may be written as follows
n ber <*) - exP a (*) cos
K
^
'
)
\bei(z)~ V(27r^)sin^
r (,\ W '
(9\
'
J
ker (*)
_ ex P(-^) c OS o
- , v
V
(kei(s) V(2*/tt) sin^ ;'
where
z \ 25 13
a (z)
. .
~ -r= + '
-v/2
'
We take first the case of / (x) when v > \. When we replace sin \B by
u, we have
a"' ,An'' em
I' ( * )=
uMml?2 (2a?)" e*
J
exp (- 2u*x) uiv (1 - u*y~i du .
i>+i)r<i).Jo i
2 (2a?)" e* S {\ - v)m f
1
,~ ,
Wa +2m eXp(
,
2Ma?)rfW '
-I> + 1)1^)^0 m! io
the last result being valid because the series of integrals is convergent.
We may write this equation in the form
v )m rzx
C\\
{) T (t\
K{a!) S (J [ fv+m-hf-tfif
l e dt
-r( V + l)V(i)^2x)J: ml(2xr]
e*
(l-v)m .y(v + m + ^,2x)
V(27r) TOto I> + i).ro!(2aOm '
t Gf. the similar procedure due to Meissel, which will be explained in 8-11.
t Bull, tit la Soc. Math, de France, xxxvi. (1908), pp. 7785. Of. Modern Analysis, 16-2.
7*25, 7*3] ASYMPTOTIC EXPANSIONS 205
y (v
+n + \, 2x)
ana tt +v >
r( + $)
is 0(x v + n + i.e~ Sx ) which is o(l) for each integral value of n.
In the case of the ordinary Bessel functions, we take the expression for
the function of the third kind
*' w H,
1+
<>-U)T>+tfJ.' "^( s)
d
i^K^ ^
and similarly
(3 ) ^w -( i) r
;
^- M )+o >
From these results it is easy to derive convergent series for the functions
of the first and second kinds.
Hadamard gave the formulae for functions of order zero only ; but the extension to
functions of any order exceeding \ is obvious.
w
2 +i).f;--{(^ir-(-r }*=^->'
* We may take v^Q without losing generality.
!
-
so that
\*
(2 [cos (x T \vtt - {-n) P (x, v) - sin (x *\vir-\ir)Q (x, v)\
J
(2) Yv (x) = (^)
yrrx/
[sin (x*\vTr-Tr)P (x, v) + cos (x + \mr tt) Q (x, v)\
Now I(v) and, in the analysis of 7*2, we may take B to be \ir since
the variables are real, and so A^, = 1.
(i1
I
x
2^
Y~* _ V
<*-"> f !?Y" + *'<*-"> f 2!\*
~ TOto m! V2tW "^ (2^)! ^2**/ '
(WW
Y~* + fi
WM*
1
-2 JLY" 4. 2* (-*> /
/i
+ j. Y~* " "v /
V
_,_
where 6 |
^1 ; and, since O is obviously real, - 1 ^ $ $ 1.
1
P( *'* )=
mto (2,n)!(2*)-
+
(2p)!(2^r(, + |)J ^ *
du >
and since
f
"
9 e~
u W +2p-* cZw U |
e~tt u+w-* dw = T (/ + 2p +
we see that the remainder after p terms in the expansion of (x, v) does not P
exceed the (p + l)th term in absolute value, provided that 2p>v \.
we find in a similar manner that the remainder after p terms in the expansion
ofQ (x, v) does not exceed the (p + 1 )th term in absolute value, provided that
2p>v-.
These results were given by Hankel, Math. Ann. I. (1869), pp. 491 494, and were
reproduced by Gray and Mathews in their Treatise on Bessel Functions (London, 1895),
p. 70, but small inaccuracies have been pointed out in both investigations by Orr, Trans.
Camb. Phil. Soc. xvn. (1899), pp. 172180.
* This result was obtained in a rather different manner by Lipschitz, Journal filr Math. lvi.
and
p
+ iX' = * (-lJLz2!k ( u y
h
(i
V 2x) mZ ml \2x)
(p-D!
where < 6 l <$ 1, and so, on integration,
v -\(
*.()- ()V (2ar) (2a )"pJ'
,t f
The results of 7 3 were put into a more precise form by Stieltjes*, who
-
proved not only that the remainders in the asymptotic expansions associated
with J (x),Y (x) and K
(x) are numerically less than the first terms neglected,
but also that the remainders have the same sign as those terms.
the terms all have the same sign whereas in the other three asymptotic expansions the
terms alternate in sign.
~ UX
P (*' 0) = e U~k K 1 + ** M) "* + - ** M >"*J du
2^r C (1 >
~ UX
Q (X '
0) =
2J^So
e
^ {(1 + url " (1 ~ hiu)
~'
]
du -
l+Ksin 2 9 Jo
K% *' ^
Jo </>
\ (a + \iu)-i + (i - \wr*} - 1 -
~f ({uY + ^jf^- ) -
M w_i and
If we multiply by the positive function e" * integrate, it is evident
that
ittsin9
i ((1
-
, ,
. v ,
4
+ iiu) 4 _ (i _ itU )-i - - j o
n ii i fi"
*^& dd>
,
we
m
find that
(2)
n/
c^,o)_-
n\ la
+
,
i'-
8! (8*)
3*- 5'
1 "
, r
K
.
)
f
l'.ff.5...(4p-8y
(2p-l)!(8)- 1
In the oscillatory parts of the series for P {x, v) and Q {x, v), the remainders
are of the same sign as, and numerically less than, the first terms neglected.
~ i]
p {x v) = " e~ ux " v-i {{1 + * iu)v
~h
+(1 " i
iu)v du '
>
2T^ry)/
Q{x, p)= e
~ ux
"""' {{l + ** M) ""* (1 ~ * iuy ~ i] du '
2iT^Y) \1
and, exactly as in 7 3 ;
we may shew that
trAtz^itr \&ni).
wi=0
( )P(i
(2p~2)?
ur~7
o
(1 - typ
"
2
{(1 + ^y^ - & - wr**) dt.
The reader will see that we can establish the theorem if we can prove that,
when 2p > v \. the last term on the right is of fixed sign and its sign is that of
- i M
1 f (1 - 1)**-* ii \w-*-i \e~ K {i+i - iut)
- e~ x <1 '
1 d\ dt
(2p-v-$)T(2p-v-$)] K
. l
' * Jo
|(1- t)
2?-*
sin ($\ut) dt = ( sin (l\ut) dt > 0.
Jo Jo
2F-T^Ti C ~ O^"*** K
(1 1 + W-* +
* - (1 " ImO'-*^} dt
into an infinite integral in which the integrand is positive, and so the expression
under consideration is positive.
That is to say,
i{(i + 4y-*+(i-iwy-}
"
mto (2wi)! (2p)!
where 0^0 when 2p> v \. And it has already been seen ( 7*3) that in
these circumstances |
6 ^
|
1. Consequently ^ ^1 ; and then, on multiplying
the last equation by e~ux u v ~* and integrating, we at once obtain the property
stated for P (x, v).
The corresponding property for Q (x, v) follows from the equation
Note. The analysis fails when \<v<\ if we take p=0, but then the phase of
(l$m) r -* lies between and \{v-\)ir, and so {{l+\iu) v -\ + (\ -iiw)"
- has the
\ *}
same sign as unity, and, in like manner, {(1 + %iu) v ~h (1 - \iu) v ~ i}/t has the same sign as
\{v-\)u, and hence P{x, v) and $(j-, v) have the same sign as the first terms in their
expansions, so the conclusions are still true ; and the conclusion is true for Q (x, v) when
ties owe their importance to the fact that they are true whether z and v are real
or complex. In the investigations which we shall give it will be supposed for
simplicity that v is real, though it will pe obvious that modifications of detail
only are adequate to make the mode of analysis applicable to complex values
j(iy y- -'-'.r e ,
-.,
M
>->f 1 +
y-'
d
2\i e'G-W-1")
If ^ v< ,
\TTz)
we use
;iy wr(- + i)
(l _^p
(l - ^*)} - A* -
// (z) = (2/*) ( +
and apply the inequality just obtained
1) ^ (*)
and similarly
(2) H (z)^0\ ($ir*)-*e-*fr-*"-W |
where
(3)
("<t)
The results may be called Weber's crude inequalities satisfied by Hv (z) {1)
and H
v (z). By an elegant piece of analysis, Weber succeeded in deducing
more refined inequalities from them in the following manner
Take the first p terms of the series involved in Hankel's two expansions
and denote them by the symbols 2 (z p), 2 2 (z p), so that (1)
;
(
'
'
;
2 u> (z
v) _ V
m=0
(-)m -(v>)
2 m (z
.
o) _ V (W)
It is easy to verify that
Yd
+ Z% + z " {z,p)
dz* dz z> -z*{-2izy-i
We regard this as an equation to be solved by the method of variation of para-
meters; we thus find that
It follows that
(z) + B (z) H. (z)
/^^ .
and so
il (#) - 4 - 1 Tr/^ {tT (5 + t)}-> 6**-*~-W P H (Z +
9 t) dt,
h 2 :^'f t)]p
where A is a constant.
We obtain a similar expression for B (z), and hence it follows that
2, m (* ; p) - {-4# (1
>
(*) + ##, ,2)
(*)} (**)* e
-'(*-*--i')
Hu y (z) = (
\ttz)
) J*-*"-**) {2 r (z; p) + Rp <
Sinee R (z) ^ 0, we have \z + t\^ ^/(r2 + t'), and so, by using the crude
inequalities, we see that. the modulus of the last integrand does not exceed
2*-p 7T"1 Q3 . (r3 + t*)-* lp+1) .
Hence
|
Rp w |
^ 2 1-* G*p |
(v, p) | f
(t* + )-*<p+ dt,
.'
W I-p
r(Jp + J).|(2*)r
K^w K ,p;i
These are the results obtained by Weber and it will be observed that in ;
the analysis no hypothesis has been made concerning the relative values of v
and p\ in this respect Weber's results differ from the results obtained by
other writers.
7'34] ASYMPTOTIC EXPANSIONS 213
When the argument of a Bessel function is not very large*, the asymptotic
expansion isnot well adapted for numerical computation because the smallest
term in it (with the remainder after the smallest term) is not particularly
small; at the same time the argument may be sufficiently large for the
ascending series to converge very slowly.
An ingenious method for meeting these numerical difficulties was devised
by Stieltjesf; we shall explain the method in detail as applied to the function
K (x) and state the results which were obtained by Stieltjes by applying the
method to J (x) and Y (x).
V
- 2 *-?-(-
V lu Bin* 0)' m d0dn
7T m = oJo JO M* *
That is to say,
n"^"(-i-^^ds*."
JO .'(
'o *(1 + |usin 2
0)
where
D
Rp=~i\
at f"
t'.'o
f*
.'o
e~ xu (hi sin8 0)*
w*(l
w, , ,
+ usm
r~-^rd0du.
2
0)
, a .
Now the value of m for which (0, m)j(2x) m is least is nearly equal to 2x
when x is large ; accordingly, in order to consider the remainder after the
smallest term of the series for K (x), we choose p so that
x= hp + <r,
tt*Jo Jo m*(1
.
..,
: -
2 m d0dn.
0)
Now, as u increases from to qo \ue~* H increases from , up to a maximum
_1
e (when u = 2) and then decreases to zero so we write ;
$-* = e~ l ~S\
The domain of integration becomes the whole of the (f, rj) plane ; and it is
found that
R 2f^r f-
e pM)
- {
l a r>s ? v>}dtdV ,
where
a o,o = i 2,o = 2o- 2 - ^j, ao,2 = i>
by some rather tedious arithmetic. It follows* that the dominant terms of
the asymptotic expansion of Rp for large values of p are
so that
(2xy K r '
W p
'
so that the error due to stopping at one of the smallest terms is roughly half of the first
term omitted.
In like manner Stieltjes proved that, if P (x, 0) and Q (x, 0) are defined as
in 7-3, then
(3) {> }
~Jlo (2*) K r p '
where
(5)
(6)
It would be of some interest to extend the results, which Stieltjes has established for
Bessel functions of zero order (as well as for the logarithmic integral and some other
functions), to Bessel functions of arbitrary order.
* Cf. Bromwich, Theory of Infinite Series, 133, 137, and 174, or the lemma which will be
proved in 8*3.
.
1 (v + i)
]
o cos 2 " +1 6
n, ,
(2x) v+ l [** sin"-* sin (i; - 4) . , ,
A (v + f). o cos 2 " +1
Q(x,v)/P(x,v)
is an increasing function of x when - < y < and that it is a decreasing
function of # when | < i>< f
For we have
Q'(x,v)P(x,v)-P'(x,p)Q(x,v)
( (2x) v+ l )
2
ft* ft*
where
F (e = - tan *> c s <* -
so that
>
*> l^^
(cos0cos<) :
< tan * *> e sin (" - *) <*>>
!(, *) + *<* = -
0)
g=|^|^ (tan tan </,) sin (i -,,)(- *).
* Ann. di Mat. (2) vi. (1875), pp. 120. The only standard work on Bessel functions in which
the importance of this memoir is recognised is the treatise by Graf and Gubler.
216 THEORY OF BESSEL FUNCTIONS [OHAP. VH
Jv (z) when z is large and v is fixed, yet Schlafli's method not only succeeds
in obtaining the expansion, but also it expresses the remainders in a neat and
compact form.
and selecting the contour of integration in such a way that, on it, the phase* of
Ire* (u - 2 + l/)
is constant. He took two contours, the constants for the respective contours
being and it ; and it is supposed that r is positive and a is real.
u = 1 + pe a ,
Hence we have
sin(a + 2fl) sinfl
(1)
{L) P U e
sm(a + 6)' sin(a + 0)
Next choose a new parametric variable <f>
such that
<f>
= 20 + a - 7T,
and then
/^ cos \ (a - <b) (u - l) a - r sin 2
v2)
( '
w= T7
,,
TT e *. re
.
~= T7(a JX cos i/ a .x -
<p) 9)
Now, as <f>
(ir a) to (ir a), u traces out a contour emerging
varies from
from the origin at an angle {ir a) with the positive real axis and passing to
infinity at an angle (ir a) with the positive real axis, provided that < a < 2v.
If this restriction is not laid on a the contour passes to infinity more
than once.
We shall now lay this restriction on a ; and then the contour is of the type
specified for formula 6*22 (9), provided that we give a> and arg z the same
value a, as is permissible.
It follows that
where u is
2i sin vrr
defined in terms of
^bC'-'M-DI-J* by equation (2).
<f>
* The reader will find it interesting to compare the general methods of this section with the
"method of steepest descents" which is applied to obtain various asymptotic expansions in
Chapter viii.
7*4] ASYMPTOTIC EXPANSIONS 217
the expression on the left by its value as a function of the third kind, we have
- re u (u - iy/u = t,
It follows that
( 4) n. +u-. = <*+j f
-- 1
,jr?(.t-}W
Hence
(f - 1)' + #/(re*)
" w = (C- l)""* (r+r + (- l)v (re-)P {(?- 1)' + #/(re*)}
j_ h+ w+ , , 1+ ,
a*
I> + m + i) = m\(v,m)
2-Tri]
* ^ ;
I>-m + ).(2m!) (2m)! *
where
(_)p //(+, l/w+.n-) <?-#>-* "+p-* ddt
"* 2mV(2V);
27rtV(27r)io J
i (- V)^ (re{(r- 1
1)' + ?*7(ia )}
W. B. F. *
218 THEORY OF BESSEL FUNCTIONS [CHAP.Vn
First consider
I r(+,i/+,i+) " +*>-*dZ
ton] (c-iy^-M^-iy+w^M
When p is so large that it exceeds both R (v ) and R{v ^), we take the
contour to be as shewn in Fig. 15; and when the radii of the large and small
circles tend to oo and respectively the integrals along them tend to zero.
If now we write
=e ,ri (l-a;)Joo
on the two rays (which are all that survives of the contour), we find that
Fig. 15.
Now the numerator of the integrand is positive (when v is real), and the
modulus of the denominator is never less than 1 when \nr< a< tr\ for other
values of a it is never less than |
sin a ]
Therefore
where | O |
is 1 or |
cosec a |
according as cos a is negative or positive. When v
is complex, it is easy to see that
(7) Vk cos
cosi/tt
R (i/tt)
I
o \(R(v),p)
(2r)*>
|
, ,
The modifications necessary for complex values of v are left to the reader.
ire<-(M-2 + l/tt)
to be zero. As before, we write
u= 1 + pe**,
and then reu />9 **/(! +pei9) is positive, and therefore equal to its conjugate
complex, so that we obtain anew equation (1). We then diverge from the
preceding analysis by writing
<p = -(20 + a)
so that
Now, as <f>
varies from a to a, u traces out a contour emerging from the
origin at an angle a with the positive real axis and passing to infinity at an
angle a with the positive real axis, provided that a lies between ir and it.
The contour is then of the type specified for formula 6'22 (8) if, as is per-
missible, we give a> and argz the same value a.
provided that v is real and p + 1 > u and .ft (02 ) ^ when I(z) ^ 0. If v is | |
;
complex the form of the remainder has to be modified, just as in the case of (8).
It should be observed that, since the integrands in (3) and (11) are even
functions of v, it is unnecessary in this investigation to suppose that R (v)
must exceed | , as was necessary in investigations based on integrals tof
Poisson's type.
r (- s) r (- 2v - s) r ( v + s + j) (- iizy ds
/. ao ivp
["'
= (- 2u)-*-* T(-s + p+p) r(-s-v +p) V(s-p + 1) (- 2iz)ds.
J -TCi
If arg
|
( iz) ^ f it
|
8, we have
r(-s+v+p)r(-s-v+p)r(s-p + i)(-2izyds\
I/:
1
ds\,
J -coi
and the last integral is convergent and so the first integral of all is
0{(-2i*)-'-p}.
But, by the arguments of 6'51, the first integral is 2iri times the sum
of the residues at the poles on the right of the contour, and so it is equal to
7T- Hv w {z)j\ei(z~ v " ) 27ri times the sum of the residues
cos pit (2z) v ] plus at
s = - p , p f, ..., p p + \. The residue at v m \ is
(-)"' r (p + m + ) T(- p + m + $)
m\(-2izy +m+ *
* Trans. Camb. Phil. Soc. xx. (1908), pp. 273279.
7'5, 7*51] ASYMPTOTIC EXPANSIONS 221
M (
- 1Fr(,
:yS +
' ti)
^-H
and this is equivalent to the result obtained in 7*2. The investigation of
HJ (z) may be constructed by replacing i by - i throughout.
The reader should notice that, although the determination of the order of
magnitude of the remainders by this method is transparently simple, it is not
possible to obtain concrete formulae, concerning the magnitude and the sign
of the remainders, which are ultimately supplied by the methods which have
been previously considered.
2 eg**ri0+>+i).ri t
2/^+2^-1
...}.
trz 1 4dz
the contouris to be chosen so that the poles of T (2s + 1) lie on the left of the
contour and the poles of the other four Gamma functions lie on the right of
the contour and it is temporarily supposed that /*, v and fiv are not
;
integers, so that the integrand has no double poles. The integral is convergent
provided that |
arg(t>) | < f 7r.
222 THBOBY OF BESSEL FUNCTIONS [CHAP. VH
First evaluate the integral by swinging round the contour to enclose the
sequences of poles which lie to the right of the original contour ; the expression
is equal to minus the sum of the residues at these poles, and the residue at
m + JO* + v) is
ggH+ww r (fi + v + 2m + 1) (-)m (^zy+* +am )
.
~~
sin fiir .sin vtr . sin (/a + v)ir' m\T + m + 1) T (v + ra + 1 ) T (ft + v + m + 1
(jj.
It follows that
xr (^- a r (-^-^) (
)
^ )s,<fo
x r (=*-) r(-^-*)cos*7r.(|*)<fc,
7*51] ASYMPTOTIC EXPANSIONS 223
and
(2) -
M . t +>))r [|J">WJ.W+F,(,)r.(.)
+ tan i 0 - ") [J* (') Y. (*) - F () /, (*))]
-BD^^er-'-) "^-) 1
We first take p to be an integer so large that the only poles of the in-
tegrands on the left of the line R(s) = -p-\ are poles of Y (2s + 1) ; and then
/oof raoi-pl
-x* J -xj p J
(when either integrand is inserted) is equal to 2wt times the sum of the
residues at the poles between the contours. Since
J
, x.-p-J
-xt-p-J
^ ^ ^ ^ ^_^
we deduce that the asymptotic expansions, when arg z <
j
i ir, are
(8) [ Jh (z) Jv (z) + F (z) F, (#)] - cot \{p - v) 7T . [J, (,) F Cr) - J v (z) F (*)]
. /* + ^
7T
"
.=o (2m + 1) !($*)*+
+1 +^ -- + 1 g"
ipi-rini^-Oir-^l 2 '
2 1, f
2'"W
and
(6) Jo (,)+I. *
<*)~~J& o (2m)l(&)-
Formula (5) seems to have been discovered by Lorenz, K. Danske Vidensk, Sehkabs
Skrifter, (6) vi. (1890). [Oeuvres scientifcpies, I. (1898), p. 435], while the more general
formulae (3)were stated by Orr, Proc. Camb. Phil. Soc. x. (1900), p. 99. A proof
and (4)
of (5) which depends on transformations of repeated integrals was given by Nielsen, Hand-
buch der Theorie der Cylinderfunktionen (Leipzig, 1904), pp. 245 247 the expansion (5) ;
is, however, attributed to Walter Gregory by A. Lodge, British Association Report, 1906,
pp. 494498.
It is not easy to estimate exactly the magnitude or the sign of the re-
mainder after any number of terms in these asymptotic expansions when this
method is used. An alternative method of obtaining the asymptotic expansion
of J,2 (z) + F2 (z) will be given in 13'75, and it will then be possible to form
such an estimate.
CHAPTER VIII
perties which
be examined are of primary importance when the orders of the
will
functions concerned are large, though niany of the results happen to be, true
for functions of all positive orders.
The trivial problem of determining the asymptotic expansion of J, (), when v is large
and z is fixed, may be noticed here.
It is evident, by applying Stirling's theorem to the expansion of 3*1, that
^(*)~exp{v+vlog(^)-(v+|)logv}.[co+^+^ + ...l,
where c 1/J(2n)', this result has been pointed out by Horn, Math. Ann. lii. (1899),
p. 369.
[Note. For physical applications of approximate formulae for functions of large order,
the following writers may be consulted: Macdonald, Proe. Royal Soc. lxxi. (1903), pp. 251
258; lxxii. (1904), pp. 5968; xc.A (1914), pp. 5061; Phil. Trans, of the Royal Soc. cxx. A
(1910), pp. 113
144; Debye, Ann. der Physik und Chemie, (4) xxx. (1909), pp. 57136;
March, Ann. der Physik und Chemie, (4) xxxvii. (1912),, pp. 2950; Rybczynski, Ann. der
* Phil. Mag. (5) xxiii. (1887), pp. 252255. [Math, and Phys. Papers, iv. (1910),
pp. 303306.]
In connexion with the principle, see Stokes, Trans. Comb. Phil. Soc. ix. (1856), p. 175 footnote.
[Math, and Phys. Papers, n. (1883), p. 341.]
t Math. Ann. txvu. (1909), pp. 535 558; MUnehener Sitzuvgsberichte, xl. [5], (1910).
5
Phyrii und Ckemie, (4) xli. (1913), pp. 191208 Nicholson, Phil. Mag. (6) xix. (1910), ;
pp. 516 537 Love, Phil. Trans, of the Royal Soc. ccxv. A (1915), pp. 105 131 Watson,
; ;
Proc. Royal Soc. xcv. A (1918), pp. 8399, 546563. The works quoted all deal with
the problem of the propagation of electric waves over the surface of the earth, and are
largely concerned with attempts to reconcile theoretical with experimental results.]
^
(1892), col. in the following
(1)
)
+^-^(i-^.W=o;
if we define a function u (z) by the equation
where i*o, i, Ma, us> ... denote functions of z which are independent of v, by
substituting in (2) and equating to zero the coefficients of the various powers
of v on the left, we find that
iu
32(l-5 ) 2 7
1 + 353 _
( 2
\
4>z
2
+ z*
24i/|(i_^ j
16r> 2 (l-5 )3 2
1 ( 16 -1512s 2 - 3654^-3755" |
+ 57607 s
t (l-*8 ) 1 3
(3) j ( pz)=s
<?*y ex p y
&T (v + 1) (1
^- - *
exp (- F*>
- *s )}'
*)* {1 + V(l
where
4>z* + z*
<*>
*-tS-}-
(1-*)t 16i^(l-^) j
s
It is stated by Graf and Gubler* that the first approximation derivable from (3), namely
j (ve) *pW(r-/
(2irv)i(l-**)i{l+V(l -*)}''
is due to Duhamel but a search for the formula in Duhamel's writings has not been
;
successful,and it seems certain that, even if it had been discovered by Duhamel, his
discovery would have been subsequent to Carlini's.
Note. The reader should observe that (3) may also be written in the form
(5) (v
t/, v seen a)' = -
. .
* '
,
'
y/(2trv tanh o)
where
(6) ^=^^(2 + 3secha)-^^(4secha+sech*a)
- gsLrJ (16 - 1512 sech8 a - 3654 sech4 a - 375 sech8 a)
Ma
-C
^ 4
(32 sech* a+ 288 sech* a + 232 sech8 a + 13 sech*- a)
+ ....
will see, by making some modifications in 81 1 (5), that these solutions may
be written in the form
where*
-
^^ (32 sec8 + 288 sec4 + 232 sec 6
+ 13 sec8 )
+ 4242 sec 10
+ 103 sec 12
)
C
- - 1512 sec - 8
3654 - 375 sec 6
5 yg ^ (16 sec* 0)
+ ....
It follows that
/
where P and Q are defined by (1) and (2). It will appear subsequently
( 8 41) that these formulae are actually asymptotic expansions of Jv {y sec 0)
#
and F (y sec 0) when i> is large and is any assigned acute angle.
Formulae which are valid for small values of 0, i.e. asymptotic expansions
of J
(z) and F (z) which are
valid when z and v are both large and are nearly
equal, cannot easily be obtained by this method but it will be seen in 8*2 ;
* The reader will observe that the approximation has been carried two stages farther than in
the corresponding analysis of 8*11.
8*2] FUNCTIONS OF LARGE ORDER 229
Note. The dominant terms in the expansions (5) and (6), which may be written in the
form
(7) Jv (x)=Mv coa(Qv -br), Yv {x)=Mv &va{Q v -ir),
Wh6re *'-
(**-*) )*>
Qv ~ */(*? v 2 ) - \vir + v arc sin (v/x),
had been obtained two years before the publication of Meissel's paper by L. Lorenz in a
memoir on Physical Optics, K. Dantke Videnskabernes Selskabs Skrifter, (6) vi. (1890).
[Oeuvres Scientifiques, L (1898), pp. 421 436.]
The procedure of Lorenz was to take for granted that, as a consequence of the result
which has been proved in 7*51,
Jf -* Til+ 8'
1 y2
-*, 1-3 (*>-*). (*-f)
+ "I
*' t*L * 2.4 ** -J
y2
-JLTi
*#!_
-*T*
tf* J '
ix. (1911), pp. 6780; (2) XL (1913), pp. 104126. A result concerning Qv+1 -Q which
is closely connected with (8), has been published by A. Lodge, British Association Report,
1906, pp. 494498.
^[{*-*/(i)}]=0.
In the range (ji-a, /a+a), the expression m{x-tf(m)} is then replaced by the first
three terms of its expansion by Taylor's theorem, namely
m M
W2
"n/[-Mm/"(m) + 2/'(m)]'
f
cob {*//' 0) + !*} Ar
then U
~7r^/[-2t{hf"( H.) + 2f'(H.)}]
In the last integral the limits for a, which are large even though a be small, have been
replaced by - oo and + oo
It will be seen from the foregoing analysis that Kelvin's principle is, effectively, that in
the case of the integral of a rapidly oscillating function, the important part of the integral is
due to that part of the range of integration near which the phase of the trigonometrical
function involved is stationary \.
The form of Bromwich's theorem which will be adequate for the applica-
tions of the principle to Bessel functions is as follows
Let F{x)be a function of x which has limited total fluctuation when x^O;
let ybea function of v which is such that vy -* oo as v -* oo Then, ifl<fi<l, .
The method which has just been explained will now be used to obtain an
* This is the appropriate substitution when m{x-tf(m)} has a minimum at =/*; for a
maximum the sign of the expression under the radical is changed.
t A persistent search reveals traces of the use of the principle in the writings of Cauohy. See
e.g.equation (119) in note 16 of his TMorie de la propagation des Ondes, crowned Sept. 1815,
Mem. prisent&s par divers savants, i. (1827). [Oeuvres, (1) i. (1882), p. 230.]
t Proc. Camb. Phil. Soc. xix. (1918), pp. 4955.
Bromwich, Theory of Infinite Series, 174.
8*2] FUNCTIONS OF LARGE ORDER 231
approximate formula for J (v) when v is large and positive. This formula,
which was discovered by Cauchy*, is
r(*)
(1) J9 (v),
2^.3*7T^'
This formula has been investigated by means of the principle of stationary phase, com-
paratively recently, by Nicholson, Phil. Mag. (6) xvi. (1909), pp. 276 277, and Rayleigh,
Phil. Mag. (6) xx. (1910), pp. 10011004 [Scientific Papers, v. (1912), pp. 617620]; see
also Watson, Proc. Camb. Phil. Soc. xix. (1918), pp. 4248.
J( v) = 1 f'cos
tJo
[v (0- sin 0)} d$-^^ Jo ff
( V-ie-fsinh*) dt,
and obviously
smw / "
e^ {t+Anh() dt ^ 1 ["^t dt=Q ^i v)
Hence
J v () = 1 'cos
f [v (6 - sin 6)} d$ + (1/v).
T JO
Now let <f>
= sin 0, and then
cosv<l>
f 'cos { v (0
- sin 0)\ d0 - I' T
d<f>.
Jo Jo 1 -cos
But l im _i_ 2
*-ol -cos0 6*
and hence, if <f>*/(l cos 0) has limited total fluctuation in the interval (0, v),
it follows from Bromurich's theorem that
Jol-costf 6* Jo
r |
r ^
It still has to be proved that 0*/(l - cos 0) has limited total fluctuation ; to establish
this result we observe that
. f
4>* 1
_ <fr~*sinflg(0) '
<# (1-costfJ 3(1 -cos 6)*
2(1-008^ _
where o ^ 3 $ _ 8m 0)
* v ' sin ^
* Compte* Eendus, xxxvm. (1854), p. 993. [Oeuvret, (1) xn. (1900), p. 663.] A proof by
Cauchy's methods will be given in 8-21.
232 THEORY OF BjSSSEL FUNCTIONS [CHAP. VHI
(4) F.(,)~-?
2*7Tl/*
The integral just discussed has been used by Cauchyf and Meissel $ to
obtain the formal asymptotic expansion of Jn (n) when n is a large integer.
It willnow be explained how this expansion was obtained by Cauchy and (in
a more complete form) by Meissel the theoretical ; justification of the pro-
cesses employed will be investigated in 8*42.
let us write sin $* ; it then follows that, for sufficiently small values of t,
m=0
and \q = 1, \ = $, \2 = TlW ^* = 26&00> ^< ~ T7 llooo >
It follows that
When n is large, \ni? is large at the upper limit, and Meissel inferred that
t Comptet Rendu*, xxxvm. (1854), pp. 990998, 11041107. [Oeuvret, (1) xu. (1900),
pp. 161164, 167170.]
t Astr. Nach. cxxvn. (1891), col. 359 362; cxxviu. (1891), col. 145154. Concerning
formula (1), Meissel stated "Sohon vor dreissig Jahren war ich zu folgenden Forme! gelangt."
8*21, 8*22] FUNCTIONS OF LARGE ORDER 233
,m+}
(1) Jn (*) ~ 7TI I A T (fm + i) (l
\/)
cos (*m + *) .
in _
Meissel also gave an approximation for X m , valid when m is large; and this approxima-
tion exhibits the divergent character of the expansion (1).
be monogenic) are the points at which dirn and t=(12rir)*, where r +1, 2, 3, ...
and near* **= (12w) the dominant terms in the expansion of 8 are
2*- + (36r)*{l1
J
+ -41'
*
Xw ~2.(36,r)i *-M-<.->
(2w + l)!(12ir)*" +
,
1
2r(2m+)
3^r(|)r(2m + 2).(127r)*m
'
Now
cos [v (0 sec sin 0)} dd
Jo
r/-0 f-/5 + tan/5-] dd ,
and 0* as 0-/8.
V(2 tan 0)
Hence, i/^
<f>* (d0/d<f>) has limited total fluctuation in the range 0^0 ^ tt, it
1
cos \v (0 - sec sin 0)} d0 ~ 2 cos (v (< + - tan 0)} //
J* 2 <f>tan)
^i/ tan 0,
and so
n\
(1)
v '
ti
"\(v
/
o\
sec r'/
0) ~ cos {y tan 0-0)-
*
(
m
777
{ir)
.
V(i*"rtanJ)
The formula
* ^
v/ sin{*>(tan/9-#)-?r}
0) ~
/0 \ Y o\ 7
(2)' v(y sec #-/
i-v os
v(i/7rtan/8)
The reader will observe that these are the dominant terms in Meissel's
expansions 8*12 (5), (6).
To complete the rigorous proof of these formulae we have to shew that 0* (d$/d<f>) has
limited total fluctuation.
'
say. But
., cos ft cosecd (1 - sec ft cos fff- 2 (fl-secftsin - ft+tanft)
'~
^
cosftcosecd(l-secftcos0) 3
and so k(0) decreases steadily as $ increases from ir, and then increases steadily
to
as $ increases from \ir to n ; since k(0)=O wben 0=ft<r, it follows that h! (0)^0
when 0^0 <ft and A'(0) changes sign once (from negative to positive) in the range
ft^d^JT.
Hence |
yjh (ff) |
is monotonic (and decreasing) when < 6< and it has one stationary
ft,
through a point w at which /'(w) vanishes and the whole of the contour is ;
on it are
u, v, Rf(w).
If Rf(w) = z, and if the z-axis be supposed to be vertical, the surface has no
absolute maxima or minima except where /(w) fails to be monogenic; for, at
all other points,
du* dv 2
The points [u , v , R/(w )] are saddle points, or passes, on the surface, so that
the contour of integration the plan of a curve on the surface which goes
is
through one of the passes on the surface. This curve possesses a further
property derived from the equation of the contour for the rate of change of ;
f(w), at any given value of w, has a definite modulus, since /(w) is supposed
to be monogenic and since If(w) does not change as w traverses the contour,
;
it follows that Rf(w) must change as rapidly as possible ; that is to say, that
the curve is characterised by the property that any point of
its direction, at
it, is so chosen that it is the steepest curve through that point and on the
surface.
from the consideration that the curve must descend on both sides of the pass
for if the curve ascended, Rf(w) would tend to + oo (except in very special
cases) as w left the pass, and then the integral would diverge if R (p) > 0.
* French "Methode du Col," German "Methode der Sattelpunkte."
t Math. Ann. lxvii. (1909), pp. 535 558; Miltichener Sitzmigsberichte, xl. [5], (1910). The
method is to be traced to a posthumous paper by Biemann, Werke, p. 405 ; and it has recently
been applied to obtain asymptotic expansions of a variety of functions.
236 THEORY OF BESSEL FUNCTIONS [CHAP. VIII
The contour has now been selected* so that the integrand does not
oscillate rapidlyon it and so we may expect that an approximate value of
;
F(t)= I am ^ m^~\
when |
t j
^ a, r being positive ; also, let \
F (r) \
< lTe*
r
, where K and b are
positive numbers independent of t, when t is positive and r > a. Then the
asymptotic expansion
1 am T (m/r) v~ m lr
e- VT F{r)dT~
/,m=l
is valid in the sense of Poincare when \v\is sufficiently large and
I
m=l
whenever t ^ whether r ^ a or t > a and ; therefore
j e- vr F{r)dr= 2 I e^a^^^dr + RM ,
J m=l JO
= 0(v-x'r ),
* For
an account of researches in which the contour is the real axis see pp. 1343 1350 of
Burkhardt's article in the Encyclopadie der Math. Wits. n. 1 (1916).
t Cf. Proe. London Math. Soc (2) xvu. (L918), p. 133.
8*31] FUNCTIONS OF LARGE ORDER 237
8*31. The construction of Debye's contours* when the variables are real.
It has been seen in 6*2, 6*21 that the various types of functions associated
with J, (x) can be represented by integrals of the form
taken along suitable contours. On the hypothesis that v and x are positive,
we shall now examine whether any of the contours appropriate for the
method of steepest descents are of the types investigated in 6*2, 6*21.
x sinh w vw,
qua function of w, i.e. we have to solve the equation
(1) a; cosh w v 0;
and it is at once seen that we shall have three distinct cases to consider,
in which xjv is less than, greater than, or equal to 1, respectively. We con-
sider these three cases in turn.
(I) When xjv < 1, we can find a positive number a such that
(2) x v sech a,
w = a + 2mri.
It will be sufficient to confine our attention to the stationary points + a; at
f
these points the imaginary part of a? sinh w vw is zero, and so the equation
of the contour to be discussed is
/ (x sinh w vw) 0.
Write w = u + iv, where u, v are real, and this equation becomes
cosh u sin v v cosh a = 0,
so that v = 0, or
/ox
(o) coshw =
,
v cosh
siny
a
.
* The contours investigated in this section are those which were discussed in Debye's
earlier
paper, Math. Ann. lxyii. (1909), pp. 535 558, except that their orientation is different; cf. 6-21.
t The effect of taking stationary points other than a would be to translate the contour
parallel to the imaginary axis.
238 THEORY OF BESSEL FUNCTIONS [CHAP, vin
symmetrical with regard to the axes the shape of the part of the contour
;
TTl
>
-77
Fig. 16.
IlTlJ ao-vi
(4) a;=i'sec)9,
and the relevant stationary points, which are now roots of the equation
cosh w cos /8 = 0,
are w= i#.
When we take the stationary point if$, the contour which we obtain is
/
5\ cos h u=
sin ft
+ 1^_J^
-
(v
__
ft) cos ft
8*31] FUNCTIONS OF LARGE ORDER 239
The shape of the curves given by equation (5) is as shewn in the upper
half of Fig. 17; and if
^^ \
ni
S^3
-"""*^ * -ni
Fig. 17.
have obtained a curve from which [ 6'21 (4)] we can derive information con-
cerning v H
(#) when x and v are large and x/v > 1. The detailed discussion
{1)
If we had taken the stationary point ift, we should have obtained the
curves shewn in the lower half of Fig. 17, and the curve going from oo to
oo 7rt gives an integral associated with H v
{i)
(x); this also will be discussed
in 8*41. The two integrals now obtained form a fundamental system of
solutions of Bessel's equation, so that there is a marked distinction between
Fig. 18.
But sin(; /3)secv-(i>-/3)cos/3 has the positive derivative cos /3 tan 2 v, and hence it
follows that
sin (v- /9) - (v f3) cos v cos
has the same signt as v /3. Therefore since v /3 and u are both positive or both negative
fur the curve under consideration, dv/du is positive.
* Proc. Comb. Phil. Soc. xix. (1918), p. 105. Since, in the limiting case (Fig. 18) in which
3=0, the slope is on the left of the origin and is ^3 immediately on the right of the origin, no
better results of this type exist.
3^'*(*>)-iP(t;)+l>0.
Now the expression on the left (which vanishes when v=j8) has the derivate
v,
But
.
/o\
x
'
o
(v - 0) cos 0+
, 8in 2
^
vsin-3cost>sin(v-/3)
--5i
sin2 r +3 cos* v
-'
s
positive when <v< ir. Therefore, since y (v) has the same sign as v - 0, it follows that
2*' [3*'>) -*()]
has the same sign as v 0, and consequently
3^'()ip()+l
has =/3 for its only minimum between v0 and t>=r ; and therefore it is not negative.
This proves the result stated.
any positive value of t, there are two values of w, which will be called wx and
Wj the values of wx and w% differ only in the sign of their imaginary part, and
;
where x v sech a.
Next we discuss the expansions of /i and w% in ascending powers of t.
Since t and dr/dw vanish when w = a, it follows that the expansion of t in
powers of w a begins with a term in (w a)* by reverting this expansion, ;
Wl _= i _^! Ti(+D
=o> + l
Wa _ a= 2
=0
(-)'n+1 -^
ni+1
T*<"+1 >,
* The asymptotic expansions contained in this section and in 8*41, 8*42 were established
by Debye, Math. Ann. lxvii. (1909), pp. 535 558.
242 THEORY OF BESSEL FUNCTIONS [CHAP. VIII
and, by Lagrange's theorem, these expansions are valid for sufficiently small
values of Ti. Moreover
1 f(Q+.o+)/d t i;A dr
am
~2wJ Ut/t*< +1 >
dw
*
2vi] Ti(m+i)
The double circuit in the T-plane is necessary in order to dispose of the
powers of t and a single circuit round a in the w-plane corresponds
fractional ;
to a double circuit round the origin in the T-plane. From the last contour
integral it folio ws that am is the coefficient of l/(w o) in the expansion of
T-j(m+i) m ascending powers of w a; we are thus enabled to calculate the
coefficients a m .
The coefficients in this expansion will be called a (m), a, (m), Oa (m), ...,
and so we have
fa (wi) = c -* (wt+1) ,
ai(w) = Co-H-,|_^.|},
/
a l(TO)
\ i. f m + 1 .-Ca + (m + l)(m + 3)
_ <v *i*n|__
<,j.t> .
s
T 2 . 2!
(m + l)(m + 3) 2dc,
a,(m) = c -Hw|-||
Co 2s 21. ' c2
(1)
(ro+l)(i + 3)(wi + 5) c^
2*. 3! 'c*
(m +
(m+l)(m + 3)(ra + 5)
1) (m +
W
3) /2cj c8
3^
+
CoV
2s 3! .
'
c 8
I
4 - a* (4) - + (- i sinh a)"t T ^ - coth'a 4 fljfr cotha},
{ # ff
.
when t | j
is sufficiently small ; and since
Jo \dr dr)
has the asymptotic expansion
" o^IVm + l )
when x is large.
Since arg {(a/, -o)/t*}-- \ir as t -*- 0, it follows that, in ($5), the phase of a
has to be interpreted by the convention arg a = + ^7r, and hence
where
(A =l, ^^-^cotn'o,
1 ^ 2 = TlF-tVVcoth 'a + ^^coth*a,
(4) !
The formula (3) gives the asymptotic expansion of (v sech o) valid J when
a is any fixed positive number and v is large and positive.
The corresponding expansion for the function of the second kind, obtained by taking a
contour from oo to oo + irt, is
,_, v , . , e
K-tnha) r(m+i) (~)m A m
r,(,secha)~-
(5)
V(iffytanha) ^-T^T -(^vtanhar-
The position of the singularities of d^-w^ldr, qua function of the
complex variable t, should be noted. These singularities correspond to the
points where w fails to be a monogenic function of t, i.e. the points where
dr/dw vanishes. Hence the singularities correspond to the values a + 2ntri
of w, so they are the points where
(W( To) be a pair of corresponding values of (w, r), then, by Cauchy's theorem,
fdw\ = J_ /"('+) dw dr Jl_ f (*+> dw_
tt
'
\drJo 2irt J dr t-t 2tri J
where the contour includes no point (except w ) at which t has the value t .
244 THEORY OF BESSEL FUNCTIONS [CHAP. VIII
be supposed that
R(w )>Q, R(w )<0.
l 2
We then have
HJm (v sec 8) =
wi
:
Jo
e
XT
\
[dr
~- -r- \ dr,
dr)
where x = v sec 8. The analysis now proceeds exactly on the lines of 8*4
except that a is replaced throughout by %8, and the Bessel function is of the
third kind.
(1) H
{1)
(v sec 8) ~ _ X 1,
T ()
L-*l .
*
*
(2) HjHv*e*e)~^T*~;: 5 T tt h) .-
(3 >
j
^2 = Th + ih cot2 y3 + $& *,
(4) J (v8ec8)rsj
v
\pwtaa 8/ L ^ ^ * =o T(i) (i/tan/S)
m r(2m + 8)
_i_
/ * *> o i n ? (-) Atn+r "I
8'41, 8-42] FUNCTIONS OF LARGE ORDER 245
(5) F(i/sec)~
w r(2m l)
/
(
I
2
\*T sin(>tan-i//3-7r)
3/
Vi^rtan^/ L
/ a
a x
*.
2 (-) =-r + *-'.,-.
v
r<$)
i ilw,
?'.-
(|itanJ*
'-
=o
-cosCi/tan^-iz/g-iTr)
* '
|
)W =o
0-)"-r(2m + f)
r(i)
^
(^tan/9)2 ' +1
The dominant terms in these expansions are those obtained by the principle
of stationary phase in 8*21.
The formulae which have been established in 84, 8'41 obviously fail
to give adequate approximations when a (or ) is small, that is when the
argument and order of the Bessel function concerned are nearly equal. It is,
however, possible to use the same method for determining asymptotic ex-
pansions in these circumstances, and it happens that no complications arise by
supposing the variables to be complex.
#.(*), HJ*(i) t
We shall write
v = z(l- ),
We then have
where the contour is that shewn in Fig. 18; on this contour sinh w w is real
and negative.
We write
t= w sinh w,
and the values of w corresponding to any positive value of t will be called w Y
We then have
1 /(<>+, 0+.0+)
/-(0+.0+.0+) /dw v
rf T
<+!)
_j_r(o+) ^w
~ 6wt J
6XP ( ^W) (w - sinh /)* <' +1 >
*
The coefficients in this expansion will be called b (m% &i(w), b2 (m), ... so that
(b,(m) = &*"**,
bx (m) = &<*+*&,
w.).a- {-*}.
(3)
6w= 61(m+l)
tT--V-|'
o*{m) o --
j.
| 24 12Q 504()0
so that*
5 = l,
(e*) 4()-e* t
B% (m) - \ <?z*
- fa B (es) = *V - ^,
3
(4)
exp ^=
(jreto,)
ttT
ir~* 2
=0
ei<+u 6* <M+1 > TO (e*) t**
ezp(#e^)^*-iT-< 2 e^
m+1 ^Q^m+li Bm (e2)T^n t
\ (IT TO=0
and [exp (zew) (dw/dr)] . satisfies the conditions of the lemma of 8*3.
(6) h *{*)
9
AJ^-i( ^ +
5TO( ^)sini(m + i )w .r|i^i|)
We deduce at once that
(7) j^z
^Lh Bm{ez) ^ Hm+l)v '~m^
(8) F, W ~-^2 f
(->.^Min4(m + l).5jg^).
From the Cauchy-Meissel formula 821 (2), it is to be inferred that, when
m is large,
<) ^m(0)~ r() (w + i)J(127r)Jw ,
The values of B (0). Bt (0), ... B w (0) were given by Meissel, Astr. Naeh. czxvii. (1891),
*
col.359362; apart from the use of the contours Meissel's analysis (of. 8*21) is substantially
the same as the analysis given in this section. The object of using the methods of contour
integration is to evade the difficulties produced by using generalised integrals.
The values of B (tz), Bj (cz) and JB9 (ez) will be found in a paper by Airey, Phil. Mag. (6) xxxi.
(1916), p. 524.
t See the Jahrbueh liber die ForUchritte der Math. 1892, pp. 476 178.
I See the Jahrbueh itber die ForUchritte der Math 1899, pp. 420, 421.
:
drjdw = 0.
They are therefore the points
t = Intri,
If we confine our attention to real variables, we see that the solution of tbe problem is
not quite complete ; we have determined asymptotic expansions of Jv (x) valid when x and
v are large and (i) xjv <1, (ii) x/v>l, (iii) x- v not large compared with #*. But there
| |
are transitional regions between (i) and (iii) and also between (ii) and (iii), and in these
The failure of the formulae of 8*4 8*42 in the transitional regions led
Nicholson* to investigate second approximations to Bessel's integral in the
following manner
In the case of functions of integral order n,
OS -
f
"cos (n0 -x0 + \x0*) d0,
ir J a
* Phil. Mag. (6) xix. (1910), pp. 247249 ; see also Erode, Archiv der Math, und Phys. (3)
xxiv. (1916), pp. 239250.
:
and the last expression is one of Airy's integrals (6*4). It follows that,
when x<n,
(1) Jn (x) ~ - {-^^-} *1 3 ^ j
>
[
and, when a? > n,
(2)
^(^^Ij^iOp^H.^,
where the arguments of the Bessel functions on the right are {2 (x- n)] }/#*-
The corresponding formula for Yn (x) when a; > n was also found by
Nicholson ; with the notation employed in this work it is
(3) r.(.)~-p^(V-i-^-
The chief disadvantage of these formulae is that it seems impossible to
determine, by rigorous methods, their domains of validity and the order of
magnitude of the errors introduced in using them. -
Debye's integral for a Bessel function whose order v exceeds its argument
x(~ j/sech o) may be written in the formf
J v (v sech a) = ~
g(tanha a)
/
rx>-\-ni
e~XT dw,
\e-**dW,
where r = - \ W* sinh a - W s
cosh a,
and the contour in the plane of the complex variable is so chosen that t is W
positiveon it. If W=
U + iV, this contour is the right-hand branch of the
hyperbola
tftanha + Jtf^F 2
,
and this curve has contact of the third order with Debye's contour at the origin.
It therefore has to be shewn that an approximation to
r<x> +ni /ocexp(Jn-i)
I e~ ZT dw is I e'^dW.
J oo -ni J ooexp( Jirt)
i/:+/;H-^K
and so the problem is reduced to the determination of an upper bound for
|
{d(w W)fdr) j. And it has been proved, by exceedingly heavy analysis
which will not be reproduced here, that
d (w - W ) i
and so
6-7T
<
ifj>j:H-i* v
Hence
~
ZTTl J ao-ni
e-*r dw = J_l
ATI J , ex p (- Jiri)
e-*rdW + -l,
V
where 6l j j
< 1.
e*
ni
exp (^ tanh3 a) [ exp {- & v? - \ v%e$ ni tanh2 a} dl-
Jo
7ri tanh a exp (1/ tanh3 a) . [7~_j (1/ tanh 3 a) - /j {\v tanh 3 a)],
whore d ] x 1 < 1. This is the more precise form of Nicholson's approximation (1).
8*43] FUNCTIONS OP LARGE ORDER 251
It can be shewn
whether \v tanh8 o be small, of a moderate size, or
that,
a smaller order of magnitude (when v is large) than the
large, the error is of
approximation given by the first term on the right.
Next we take the case in which the order v is less than the argument
x (= v sec 0).
We then have
gvi(tanj30) foo+Hir-p)
H (v sec 0) = Tn
:
J -co-ip
e~*r div,
e-^dW,
/
where r = -\i W* sin - W 3
cos 0,
and the contour in the plane of the complex variable is such that t W is
positive on it. If WsU+iV, this contour is the branch of the cubic
d(w- W)
< 127r sec 0.
UiT
-.
TTIJ
e-^dw^.
irXJ- oo-i tan
e -xr dW + zl^
v
)
-*-ip fi
where |
& <
j
1.
To evaluate the integral on the right, modify the contour into two lines
meeting at W= i tan and inclined at angles Kir and ir respectively to the
real axis. On these lines, write
found that
(5) Jv (p sec /3) = tan cos {* (tan /3 - tan -)}. [J"_j + /j] 3
(6) F (i/ sec ) = | tan /3 sin {i> (tan - tan - )} [J"_ j + J]3
.
- 3"* tan cos {*/ (tan - tan -)}. [7"_ - </j] + 240 M
3
j 3
where the argument of each of the Bessel functions J>. on the right is
%v tan 3 and ;
2 and !
forms of Nicholson's formulae (2) and (3); and they give effective approxima-
tions except near the zeros of the dominant terms on the right.
It is highly probable that the upper limits obtained for the errors are
largely in excess of the actual values of the errors.
i rw
__ I gv(as8inhww) fly^
ttJo
* The results of this section are investigated in rather greater detail in Proc. London Math. Soc.
(2) xvi. (1917), pp. 150174.
*
1 20
r x sin
so that
,
log
x sin
e + y/ifr-X* sin 0)
-.
2
cot.a ya(tr
at
. a;
a
2
*m
sin 2 0).
(1) J {vx)=^{'<r'>de,
and by differentiating under the integral sign (a procedure which is easily
justified) it is found that
/o\
(2)'
K JTi
(vx) =
\ l
-
ttJo
f
w
fib*
e- *<.*< - Ja
fl- a'Binflcosfl
-
x */(&* x* sm* 0)
. dv.
JJ (vx) = s
1
.
raa+iri
\ e
(x*inhu>-w)
sinh w ^
Before proceeding to obtain further results concerning Bessel functions, it
< 3>
m^'^ ^-M)*^-^' ^ 6 -
w 4^.,-^a^
so that
(5) F(0,x)>F(O,x)>F(O,l) = O;
and also
ia\
d
V(<* \
fl-g'sinflcosfl n
* This function will not be confused with Schlafli's function defined in 4-15.
t It is supposed throughout the following analysis that O<.c^l,O^0<ir.
254 THEORY OF BESSEL FUNCTIONS [CHAP. VIII
It is clear that
g(0,x) = s'(\-x*)<<J(\+x*),
g(Tr,x)=\ <J(\+x%
so that, if g (0, x), qua function of 0, attained its greatest value at or tt,
1 - a;
2
cos 2fl _ (0 - xO
2
si n O cos o f_
(tfo'-^sin ^)* 2 2
(0o -a?sin 2 o
)i ~ '
and therefore
g {0, x)^g (8 ,
x) = V(l - a? cos 20 ) ^ V(l + *), o
so that, no matter where g(0,x) attains its greatest value, that value does
not exceed ^(l +# 2
)-
Hence
and so
^W-*w^^
a
V(l+a )
1
sin cos
2
dF X)
'
Thus, since
OV IT JO
the integrand being positive by (5), it follows that Jv (vx) is a positive de-
creasing function of v, in like manner, JJ {vx) is a positive decreasing function
of v.
Also, since
n
d {**<*> JAva:)\ = _ 1 (
SF{0> x) _ F ^ 0> x)} *-*>. *>-"*.*> d0 <0,
OV 7T.' o
T Jo
eF(0,x) roo
< exp {- i/0\/(l - x*)\ d0,
so that
p -vF(0,x)
(9) <L(vx)^
(1 - x*)l ^(Zttv)'
Then
dG ( >^
2xj;{vx)= - /V-.* {G(0,x)}-id0
fi
-vF(0,x) fir
7T I
expi-^GyVa+a^.G-MG
.'o
e -F(Q,x) fao
Texp {- J^(l+^)j . G-idG,
IT Jo
and so
(10) xj; (vx) ^ -'*<. ) ( 1 + *)VV(2 w).
The absence of the factor ^(1 2
a? ) from the denominator is remarkable.
* It is evident from Debye's expansion that the expression is in error by excess for sufficiently
large values of v.
If V(^ - a 2
sin2 0) = H{0, x), we shall first prove that
dF(0,x) dH(0,x) ,
d0 / d0
is a non-decreasing function of ; that is to say that
(\-0cot0)2 + i
-a?sm*0
a? sin cos
is a non-decreasing function of 0.
and every group of terms in this expression is positive (or zero) in consequence
of elementary trigonometrical inequalities.
To establish the trigonometrical inequalities, we first observe that, when < 6 ^ *r,
(i) d+sindcosd-2^- 1
sin 2
d>0,
(ii) 6 + sin 6 cos 6 - 26* cot 6 > 0,
(iii) sin - 6 cos - sin 3 0^0,
because the expressions on the left vanish wheu 60 and have the positive differential
coefficients
(}><>
where the. variables are understood to be and x, and primes denote differ-
ifiM-*{S>*
and, if we integrate this inequality from to 0, we get
F>(0,x)H{0x) _
+ j >Q
if (0, #)
and the truth of the lemma becomes obvious when we substitute the value
of H (0, x) in the last inequality.
8*52] FUNCTIONS OF LARGE ORDER 257
[The actual proof of the theorem be valid only when S^.r^l, (where 8 is an
will
arbitrarily small positive number), since some expressions introduced in the proof contain an
x in their denominators; but the theorem is obvious when ^ x < 8 since e vF > z Jv (vx) and >
will be seen in Chapter xvn, the theorem owes its real importance to the fact that it is
true for values of x in the neighbourhood of unity.}
OV 7T .
o
'
- L\/^ > <* <* *))-* or e vn$,x) de
"
W
V
[ \\C((l -rVh
dFie X ^>
IP/A \in(A ^W^1^L A,)
x e-J?ie,x)d0,
if we integrate by parts the former of the two integrals.
where
by using the inequality F(yfr, x) > ^(0, x) combined with the theorem of 8*51.
' ;
Since CI (0, yfr) is not negative, the repeated integral cannot be negative
that is to say, we have proved that
J ( vx
\ ^H^} _ dJAvx) dJ (vx ) >" '
dvdx dx dv
so that
d \dJv (vx) / i
~dJv (vx) I
T .
,"|*
A
so that
dJv (vx) I , dJv (v)
-I Jrv (VX) *Z J* (")
dv fo~ J
Since J v (vx) and Jv (v) are both positive, this inequality may be written
in the form
and this exhibits the result which was to be proved, namely that Jv (vx)jjv (v)
is a non -increasing function of v.
^~ -~ IP-*
*"(*> cot *)/*}
+ ^ ,,
Vi &2 _ siR2 ,.
V(^ - 6-* * '
sin* 0)
* It is to be understood that JV (v) means the value of dJv (x)/d.v when x has the particular
value v.
^
and that
d^ 1 -0 cot 0) _ fl
2
cosec 2 fl + flcotfl- 2
j
~
dd \ fr~~~) p '
d_ f
vW-sin a
fl))
= _ (ffcosec'fl + flcot 0-2) sin
2
d&\ 0*
J" "^VO^^-sin^)"'"
'
x (0 + sin
2>
- 20* cot 0)
cos
that is to say
^ [OF' (0) - 3F (0)} > 0.
If we integrate this inequality from to we get
(4) 0F'(0)-SF(0)>O,
and this is the condition that F(0)/03 should be a non-decreasing function of 0.
It follows that
s
*m * 9V3'
and therefore
Jv (v)< - \ exp j- (*
< 9lp |-
ii." "SI-
T(i)
2*3**'
so that Cauchy's approximation for / (v) is always in error by excess.
The truth of this may be seen by writing the expression on the left in the form
{&* - 2 sin 2 6 + 6 sin 6 cos 6) F' (0) + (0 - sin 6 cos 6) {BF' {$) - ZF{fi)\ %
(6)
f
l
Jv ( vt ) dt ~ J- - T ,
It has already been seen ( 8'5) that the functions and JJ (v) are J v (v)
decreasing functions of v. It will now be shewn that both v*Jv {v) and v*Jv '(v)
are steadily increasing* functions of v.
dj^ioi _
dv
r
Sir J o
-
e vFi e )de ^r tJo
F(d)e-.F W dd
>0,
since the integrated part vanishes at each limit and ( 8*53) the integrand is
positive.
To prove the second result, by following the same method we find that
dl' o7T |_ JO
>o,
by 8*53 (5), and so v*Jv'(v) is an increasing function of v.
Hence
(2) v* J: (v) < lim {*i JJ (v)} - 3* T (f )/(2*w) = 0-41085.
It is to be noted that
J t
'
(1 ) = 0-32515, 4J6 '
(8) = 0-38854.
A theorem which is slightly more recondite than the theorems just proved
is that the quotient
W(V)} + {!*/, (*)]
* It is not possible to deduce these monotonic properties from the asymptotic expansions. If,
as v-*-<z> , /(*)~ </>(), and if ${) is monotonic, nothing can be inferred concerning monotonic
properties of f() in the absence of further information concerning/^).
8-54, 8-55] FUNCTIONS OF LARGE ORDER 261
Taking the parametric variable in the first and third integrals to be >/r in
> ^^
2
{* (*) - *<*)) -
by 8'51. The function fi x (0, i/r) does not seem to be essentially positive
^# +*" <*> - *<+>
^V^-sin'^^-s^
+
^ + VrgJnVrC0B^-2gin'Vr U V
_ V "
^V(^ 2
-sin2 ^) '
fl + flsinflcosl-2sinl _
l5r vr ' vr/;
&)/(&- sin* 0)
Since 0"1 ^(0* - sin 8
0) and J &F'(0) - F(0) are both ( 8'53) increasing functions
of 6, the factors of the term in the sum on the right are both positive or
first
both negative; and, by 851, 853, the second and third terms are both
positive. Hence fl! {6, yfr) +- llj (yfr, 6) is positive, and therefore
x '
dv { Jv {v) J
HJV (z) = A
TTIJ _
e-^ w >
dw,
If(w) = //( 7 ).
If we replace w + iv, this equation may be written in the form
by u
(v - fi) cosh a cos & + (u a) sinh a sin - cosh u sin v + cosh a sin /8 = 0.
creases asw moves away from 7 on the first branch, while it decreases as w
moves away from 7 on the second branch. The increase (or decrease) is steady,
and Rf{w) tends to + 00 (or - x ) as w moves off to infinity unless the curve
has a second double-point t.
If (i) and (ii) denote the whole of the contours of which portions are
marked with those numbers in Fig. 19, we shall write
^ ^ tanh y))
s
i
r(t +
o r (i)
j)
(
_ ^ y tanh 7)w
am
,
(i)
Fig. 19.
The values of A A
, 1} A.2} ... are
8'61. The form of Debyes contours when the variables are complex.
where (u, v) are current Cartesian coordinates and < /S < it.
when it v
and it # are written for v and #, we shall also at first suppose
that 0</9^^7T, though many of the results which will be proved when y8 is
an acute angle are still true when /9 is an obtuse angle.
For brevity, the expression on the left in (1) will be called <f>(u, v). Since
d<f>(u,v)
= sinh . ,
o sin
.
p sinn
. ,
u sin v,
it follows that, when v is given, d()>/du vanishes for only one value of u, and so
the equation in u,
<f>
(u, v) = 0,
has at most two real roots ; and one of these is infinite whenever v is a multiple
Of 7T.
<f>
(a, v) = cosh a {(v /3) cos /3 sin v + sin /3} ^ 0,
< (u, v) = 0,
is less than a and the other is greater than a. both becoming equal when v = /3.
TTi
\
<z
)
J ^
TTi
y-2iri
Fig. 20.
* Since d<f> (a, u)/3t'=cosha(cos/3-coBv), and this has the same sign as v-/3, 4> (a, r) has a
minimum value zero at v = .
8-61] FUNCTIONS OF LARGE ORDER 265
responding domains for the complex viz cosh (a 4- ift) have the same numbers
in Fig. 22.
TTl
-1>
5 4
7b 1
66
5
la 6a
3 2
so that the curve never crosses the line v = ft. The only possibility therefore
is that the curve after crossing the real axis goes off to 00 as shewn by the
upper dotted curve in Fig. 20.
<f>
(u, mr) = (u a) sinh a sin ft + (mr ft) cosh a cos ft + cosh a sin ft,
;
and this is positive when u > a, so that the parts of the curve which go off to
infinity on the right must lie as shewn in the north-east corner of Fig. 23.
Fig. 23.
We now have to consider what happens when (a, ft) lies in the domain
numbered 6a in Fig. 21. In such circumstances
and ^ ( a, v) has a maximum at v = 2tt ft, the value of ( a, 2ir ft) being
negative. The curve, after crossing v = tt, consequently remains on the right of
u =a until it has got above v = 2rr ft.
Now <f>
( a, v) is increasing in the intervals
-oo +(2M+l)iri;
it cannot go to infinity lower than this, for then the complete curve would
meet a horizontal line in more than two points.
:
-oo +(2M+l)Tri,
where M is the smallest integer for which 1 tanh o + {(M + 1) tr /9} cot #
is positive.
H v (z) and
{1
>
v H
(z) when (o, fi) lies in the domains numbered 1, 2 and 6a in
(i)
Fig. 21 and by suitable reflexions we obtain their values for the rest of the
;
complete strip in which < yS < ir. The reader should observe that, so far as
the domain 1 is concerned, it does not matter whether /S is acute or obtuse.
is positive when cot is positive, and if N is the smallest integer for which
5, 76 - oc , - oo + 2m' 2e-^iJ_ v z ) {
The reader will find it interesting to prove that, in the critical case /3=r, the contours
pass from oo to oc + iri and from - oc + ni to oo , so that the expansions appropriate to
the region 1 are valid.
Note. The differences between the formulae for the regions da and 66 and also for the
regions 7a and 76 appear to have been overlooked by Debye, and by "Watson, Proc. Royal
Soc. xcv. A, (1918), p. 91.
This formula is less precise than Carlini's formula because the factor (2jrn)* (1 -s )* does
2
not appear in the denominator on the right, but nevertheless the inequality is sufficiently
powerful for the purposes for which it is required %.
Jn {nz) = ^. J
r 1
exp { \nz (t - 1/t)} dt,
be chosen subsequently.
If we write t eu+i8 we , get
is \p
w
{e cos (a + 0) - e~ u cos (a -
$)} - u,
|
Jn (npe im
) j
< exp [np v/(sinh 2 u + sin8 a) nit].
We now choose u so that the expression on the right may be as small as
possible in order to get the strongest inequality attainable by this method.
The expression
p V(sinh u
2
+ sin3 o) u
has a minimum, qua function of u, when u is chosen to be the positive root of
the equation*
sinh u cosh u 1
\/(sinh 2 u -f sin* o) p
2 V(l "8
s
) sinh u cosh u = (cosh 2 g2la ),
and, by taking z to be real, it is clear that the positive sign must be taken in
the ambiguity. Hence
2 {1 + V(l - z*)) sinh u cosh u= e2" - *,
and so
z exp \/(l -z) _ %
! , 2 V(sinh a u + sin2 a) . | exp \/(l -z) a
\
log
T+ V(l - z*)
V(i-* s
)
\~ g
I
e 3*
- e** I
= 5V(l-z')-
sinh 2 m + sin2 a
sinh u cosh w
= p \/(sinh 2 u + sin 2 a) u,
and it is now clear that
V( 1-^) |"y
rexpV(l-^)
fcexp
Jn (nz) $ I
j j
sinh u cosh u _ 1
"~ p
V(sinh 2 w + sin 2
a)
It follows that
2u
P*
= '
sin* a sinh u (u cosh u sinh u).
sinh 2 m
inside and on the boundary of an oval curve containing the origin. This curve
When the order of the Bessel function is positive but not restricted to be an integer we
take the contour of integration to be a circle of radius e* terminated by two rays inclined
+ it arc tan (coth u tan a) to the real axis. If we take 1 = ev on these rays, we get 1 1
; j^/l + |
s
2!L!^|
w
r e xv{-v(v-u)}dv\
[ I I
J u )
and so
I pex P V(l- g) )'
|Jr(v*)|< \l + \\
W t=z~ Jo (Z) (<) + 2Jl (2) l (<) + 2J* < 2 > * <*)+
- S en Jn (z)On (t).
n~0
From we
shall derive an explicit expression for the function,
this definition
and it will then appear that the expansion (1) is valid whenever \z\< \t\.
In order to obtain this expression, assume that z < t and, after expanding j \ j j
-1
t-z
=- +
1
< s=1
2
7*
** +1
- 1
s T (,\a. v 2 M v (s + 2m).(s + m-l)l r j
fr
m=0 *=1 & (,i = nl -
1
Assuming for the moment that the repeated series is absolutely convergent*.
Math, lxvii. (1867), pp. 310 314. Neumann's procedure, assuming the expansion (1), is to
after
derive the differential equation which will be given subsequently ( 9 12) and to solve it in -
series.
1 /(*+)/()"
(
,=.
f(z)r
si/
+ Cf. Pincherle's rather more general investigation, Rcndiconti R. 1st. Lomlardo, xv. (1882!,
(2)
pp. 224225.
272 THEORY OF BESSEL FUNCTIONS [CHAP. IX
m!
* * m=<\ n=l V =0 f
J
(4) en O(0*-^iT :
|i + _ + 2i4t(2n-2)(2M _ + ...|,
2(2n 2) 4)
and the series terminates before there is any possibility of a denominator factor
being zero or negative.
We have now to consider the permissibility of rearranging the repeated series for
l/(* - 1). A sufficient condition is that the series
% 2* % (<+2m).(+m-l)i .
, . \
* iTTTi J1 * TTi
*/
+2> \'/ If
.i * t i
' U-o .
should be convergent. To prove that this is actually the case, we observe that, by 2' 11
(4), we have
+
<3(J|*|) -{exp(i|*P)}/(2m)!
2
<(il*l)*e*P(il*l )-
Hence
_ ||eipQi|)
'
||(|*|-|*|)
When these inequalities are satisfied, the sum of the moduli of the terms does not
x
^_ J
(8 + 2m).( + m-l)!
*"
( /)*
+2ro exp (fr*) \ exp (frr )
8
Since the expression on the right is independent of z and t, the uniformity of the
convergence follows from the test of Weierstrass.
1 n* n ( -2 w a a s
)
8
(n2 -2 )(w
2 9
-4)
_ J. r= 1 i i -4- - '-
+ ...
(6) n (t) =
j JE o (1
__ 1Jiara=s (n odd)
-p + ^ + +....
t
,
* By analogy
* Jn <*)
,
t-z
=
-=,
n=o
r. (2n + l) Pn {z)Qn (t).
Cf. Modern Analysis, % 15*4.
^
and hence it may be shewn* that the points on the circle of convergence at
which either series converges^ are identical with the points on the circle at
which the other series is convergent. It may also be proved that, if either
series is uniformly convergent in any domains of values of z and t, so also is
the other series.
r -^
(-)g.(p + g)!: =
T - dPjn (z) diQn (t)
(10) y
(t-zf + * +1 y_ 2 w=0 dzP dV '
o.(0i/. o 1 () = i/< 2
,
The coefficients in the polynomial On (t), for =0, 1, 2, ... 15, have been calculated by
Otti, Bern Miltheilungen, 1898, pp. 4, 5.
(3) -0,()0.'(*)
The first of these was stated by Schlafli, Math. Ann. in. (1871), p. 137, and proved by
Gegenbauer, Wiener Sitzungsberichte, lxv. (2), (1872), pp. 33 35, but the other two were
proved some years earlier by Neumann, Theorie der BesseFschen Functionen (Leipzig, 1867),
p. 21.
Taking |
z \
< \
1 1, observe that, by 91 (1) and 2-22 (7),
* It is sufficient to use the theorems that, if 2fr n is convergent, so also is 2&/n, and that then
Sbjri* is absolutely convergeut.
t This was pointed out by Pincherle, Bologna Meviorie, \i) in. (1881 2), p. 160.
Cf. Modem Analysis, 5-33.
9'11] ASSOCIATED POLYNOMIALS 275
and hence
z I en Jn (z) n (0 =I en Jn (z) {tOn (0 - cos' \ntr)
n=0 =0
since tO (t)=l. If now we use the recurrence formula for Jn (z) to modify
the expression on the right, we get
If we notice that Jn+1 (z) {On (0 cos2 n7r }/n tends to zero as n-*>oo, it
is clear on rearrangement that
Now regard zasa variable, while remains constant if the coefficients of;
all the Bessel functions on the left do not vanish, the first term which does
not vanish can be made to exceed the sum of all the others in absolute value, by
taking \z\ sufficiently small. Hence all the coefficients vanish identically*
and, from this result, formula (1) is obvious.
/d d\ _i
\dt
+ dz) t-z~ '
5 *nJn (Z) On' () -J x (z) (t) - I \J^ (,) - Jn+1 (jr)} On (t)
=0 n=l
From the recurrence formulae 9*11 (6) and (7), it is clear that
+ n + 1) (* - n + 1) = -L;- (^ + n +
1) {- ntOn+1 (t) + n sin \nnr}
2
(* n (t)
=- Tit
-j
(^ + n + 2) n+ , (0 + n sin \mr 8
is y = On (0 + r 1
^(0,
and so the only solution of (1) which is expressible as a terminating series
is On (t).
where
lere
(w even)
( odd)
+z
l +zj Jo 6-*4 {z) {t)= i
2
enw2,7n {z) * (0
f z? '
00 00
00
(2) f
J m (z) On (z) dz = 0, (ma n>)
J c
The first result is obvious from Cauchy's theorem, because the only singu-
larity of m (z) n (z) is at the origin, and the residue there is zero.
The third result follows in a similar manner ; the only pole of the inte-
grand is a simple pole at the origin, and the residue at this point is l/ea .
./ W iifO|<f)i-, W ^),
*U' (z) + zll{z) + (m2 - w2 ) zJm (z) On (z) = z*g n (z) Jm (z),
* Theorie der BesseVschen Functionen (Leipzig, 1867), p. 19.
278 THEORY OF BESSEL FUNCTIONS [CHAP. IX
and hence
Two corollaries, due to Schlafli, Math. Ann. m. (1871), p. 138, are that
1
/(o
/> + )
Jn {*+y)= 2 /, +p (i)/.j(y),
p 00
and the second follows by making an obvious change of variable.
(i) o, (.)
= /;
J" +**+
^i+i:
- ^ + *
^du.
We shall now prove by induction the equivalent formula
/*> exp ia
(2) On (z) = UJo [{t + V(l + *
2
)} + {t- V(l + * )} w] *~* dt,
a
roo+ta
(3) 0 (z) = \\ w
{e * + (-)""} -*** cosh ^0.
Jo
To prove (2) we observe that
/"ooexpia racexpia
O (s) = e-*dt, 0,(z)= te-*dt;
Jo Jo
and so, by using the recurrence formula 911 (2), it follows that we may write
oo exp ia
0. <*)-[" <f>
gt
n (t)e~ dt,
Jo
where
(4) <f> +1 (0 - 2t<j> n (t) - <_, (0 = 0,
and
(5) <M0=1> *!<) = *.
* Theorie der BtsseVtchen Functionen (Leipzig, 1867), p. 16; Journal filr Math, lxvii. (1867),
p. 312.
9*14] ASSOCIATED POLYNOMIALS 279
This proof was given in a symbolic form by Sonine*, who wrote <p n (D) . (1/z) where we
/<*> exp ia
<t>n (0
*"*** dt> D standing for (djdz).
whose analysis is based on the expansion of 9*1 (1), which we now write in
the form
--=X
z n Jn (S)O n (z).
s *=o
When j f | < |
s j
, we have
= IT\
9
2 p
n Jn (0\ e-du,
2"
.
'">
oo
(n=-<*
V.n= j
if p be so chosen that
u 1
z
It follows that
We shall now shew that the interchange of summation and integration is justifiable; it
will be sufficient to shew that, for any given values of ( and z (such that | ( | < |
z |),
W f \ fu + J(u i 4-z )\ n 2
I
n=N+\ J I
3 I
\uJ{u*+z*)\^2(u + \z\),
andso {
Jn{0 \
r\^^p^l\
J |
2
e -u du ^J^ fV(+|,|)
I
z \
J
e -* t
1*1" J LI
<|(C/*)"|exp{|| + iim.
* Math. Ann. xvi. (1880), p. 7. For a similar symbolic investigation see 6 "14 supra.
M r- I
{uJW+f^jn() |
. If I *P {!! +* | <f}
(z)
.-l+i J. I
*= ^
* \*\{\'\-\i\r-
'
and the expression on the left can be made arbitrarily small by taking N sufficiently large
when z and ( are fixed.
?
= 2
w = - ee
/(f)
.'0
{w+^+f!)}
*
_ ^
e~"rfu
v
Z
'
y) tti
0 (,) = j"
\* + " + Wj- </(* + ** ^ du e
.
Let yjr (w) be an arbitrary function of w; and, if yfr (w) = x, let w = ^ ()>
so that <fr is the function inverse to
ty.
curve c surrounding the origin but not enclosing the point z. Then
+ 1
oo
-
/*oo r
gg(w)-
y%n
dwdx
Z n= oo
= i KJ(*){4 n + (-)A_n }.
=
Since
The series 2 (-^n^On (*), which is a generating function associated with On (z),
does not converge for any value of t except zero. Kapteyn* however, has "summed" the
series after the method of Borel, in the following manner
, v 1 * n n.(n + m-\)\t2n
2n +
_ (n + $).(n+m)l t
"
l
+
nlomlo (n-)!(i*)*
H
_1 1 + *
2 | n.(n + m-l)l t*
l+< 2 (-).m! tm
*(l-<*)m=o{*( 1 -< , )} m
'
/>
7j
e~"du
tot ^ >
an d this integral
is convergent so long as (1 - 1'*) z\t is not negative.
integral may be regarded as the generating function of n (z). Kapteyn has built up much
of the theory of Neumann's function from this result.
We have
1 r*
n < n *) = 2i^+i H + V(w + z )] n +{u>- V(/ 2 2 2
+ z*)} ]
n
e~nw dw,
J
j
n (nz) ! ^
I
~
z i
["
J o
|
[{, y/(vfi + *)} e-] | dw .
!
j
,
where that value of the radical is taken which gives the integrand with the
greater modulus.
[w + V(w + z )} e~ w 2 2
is V(l 2
-z ), and so
(!) I
On (nz) ^ |
JL_
|
Lt^^p Jl {"
+ VK + z>)} e-\.\dw\,
where the path of integration is one for which the integrand is greatest at the
stationary point.
(2)
|
1 + Vd-* 2
)
:
Hence, since a contour joining the origin to infinity can be drawn when (2) is
satisfied,and since the integral involved in (1) is convergent with this contour,
it follows that, throughout the domain in which (2) is satisfied, the inequality
is satisfied for some constant value of A ; and this is an inequality of the same
character as the inequality of 8'7.
9-17, 9-2] ASSOCIATED POLYNOMIALS 283
zv * z v+s
x 2 "+* (
(v + s + 2m).T (v + s + m ) r
. )
=
"
J iio
"
(<i 2" +n 2m
* -
(i/ + n)
~~^.
.
1" (v + n - m) ,
Jv+n {z)
. A
\
;
(2) .
Z
=2=0 A n>v {t)Jv + n {z).
t M
,, .,
= ( y + w _l)^' n> () +
^ v
2"(v + n)(v + n-l)r(v + ln-$)
rfln + i)
I -, 8in2l
i".
.
,
^sin'iuTr,
(7) 4^)2T(+l)/t
* Wiener Sitznngsberichte, lxxiv. (2), (1877), pp. 124 130.
284 THEORY OF BESSEL FUNCTIONS [CHAP. IX
where
(9) ^( .-A
r(J
L__.l WW
The general solution of (8) is -4 ,,() +^ _I ^+n (0-
Of these results, (3), (4), (8) and (9) are due to Gegenbauer; and he also
proved that
(11) 1
Lttx J
/"< 0+)
n (t) e
izt
dt = in cos {
n arc cos z)
(14) I
z~ v Jv+n (z) A n v (z) dz = 2irik,
.' c
where G is any closed contour, n = 0, 1,2, . .
. ,
and k is the excess of the number
of positive circuits over the number of negative circuits of G round the origin.
The first and third of these last results are proved by the method of 913 ;
~v
(m - n) (2v + m + n) \
z~ v Jv+m (z) A n,,( z ) dz = / z2 gn<v {z) Jv+m (z) dz = 0.
Jo J c
<n {n
(n _m_ 1 \
(1) Sn (t) = 2 A)
(iO-"
+MW
,
(n > 1)
(2) S.()-0.
(n - 1) #n _ (0 - 1 (n +
x 1) +, (0 = nSn '
(t) - nt- 8n (t) - 2*"
1 1
cos2 fnw.
If we multiply this by 2 and add the result to (4), we get
(5) Sn^{t)-Sn+1 (0 = 25/(0.
The formulae and
may, of course, be proved by elementary algebra
(4) (5)
by using the definition of Sn (t), without appealing to the properties of
Neumann's polynomial.
The definition of Schlafli's polynomial of negative order is
Other forms of the recurrence formulae which may be derived from (4)
and (5) are
(8) *n_, (0 - nSn (0 - tSn (0 = 2 cos8 ^nir, *
it follows that
(^
2
- n ) Sn (0 = t (^ + 1 - n) #n_, (0 + In cos8 \nir
8
= 2n + 2n(n
t- P
-2
i+
2n(n-- 2
t
)(w -4
^
+
i 2
)'
-
2
6
2 2
)
.
'
The polynomial (<)> for n = l, 2, ... 12, have been calculated by Otti,
coefficients in the
Ztera Mittheilungen, 1898, pp. 13
14 Otti's formulae are reproduced (with some obvious
;
errors) by Graf and Gubler, Einleitung in die Theorie der Bessel'schen Funktionen, II. (Bern,
1900), p. 24.
We
have already encountered two formulae connecting the polynomials of
Neumann and Schlafii, namely
\nSn (t) = t0n (t) - cos \ntr, 2
easily derivable from the formulae already obtained, and we shall now discuss
the more important of them.
When we eliminate cos2 |/wr from 93 (3) and either 93 (8) or (9), we
find that
<(i-D
(3) Sn (t) = -2 2 S'n^mr-i(t) + Bin*lnir.S l (t),
w=
and hence
n
(4) Sn (t) + Sn _ 1 (t) = -2 2 fi^-, (*) + &(*).
TO
* Comptes Rendus, cxxv. (1897), pp. 421423, 860863; BernMittheilungen, 1897, pp. 6196.
Y F
4Sn" (0 + 4^(0,
so that
(5) Sn" (t) + Sn (t) - O n _ x (t) + O n+1 (t).
This is the most interesting of the formulae obtained by Crelier.
Again, on summing formulae of the type of 911 (2), we find that
obviously true when n = 0, and also, by 363(12), when n=l. If now the
sum on the right be denoted temporarily by n (z), it is clear that <f>
+ irJn_ 1 (z) 2
m= n+l
Ym (z)-7rYn_ 1 (z) Y
m= +l
Jm (z)
= 7rJn+1 (Z) Yn+1 (Z) + Y_ n ^ (Z)} - 7T Fn+1 (z) {Jn+1 (z) + J_n_, ( Z)\
{
= 4.2 -1 eos 2
^wr,
by 363(12); and so n (z) satisfies the recurrence formula which
<f>
is satisfied
.'
eft
_ cos'^wtt _^ i"
00 * \t + \/(l 4- P)} - f t - V( 1 + * )} w 2
_rf
" + e
5 2zJ V(l+**)
Hence it follows that
This equation, which was given by Schlafli, Math. Ann. in. (1871), p. 146, in the form
We write temporarily
so that
y"+! _ ot 4- IT*
T T '
and therefore
7'
C + 2 + 2*+...+2e'
n
Tn+ l= K(2t,2t,...,2t)n
Tn K{2t,2t,...,2t) n ^
the suffixes n, n 1 denoting the number of elements in the continuants.
we have
Tn = 2</(l + t*).K(2t)n_ 1 ,
* Comptes Rendus, cxxv. (1897), pp. 421423, 860863 Bern Mittheilungen, 1897, pp. 6196.
;
and hence
fee exp ia
(3) Sn (z) = 2 K(2t)n _ e~^dt,
1
Jo
From this result it is possible to obtain all the recurrence formulae for
= 2 I e m Om (t)Jm (-z)
= 2/(* + z) = S, (t + z),
by 91(1) and 93 (7).
Now, if we assume the truth of (1) for Schlafli's polynomials of orders
0, 1, 2, ... n, we have
m= *>
and the induction is established ; to obtain the second line in the analysis,
we have used the obvious result that
The expansion was obtained by Schlafli by expanding every term on the right of (1) in
ascending powers of z and descending powers of t. The investigation given here is clue to
Sonine, Math. Ann. xvi. (1880), p. 7 ; Sonine's investigation was concerned with a more
general class of functions than Schlafli's polynomial, known as hemi-cylindrical functions
< 10-8).
This was proved directly by Gegenbauer, Wiener Sitzungsberichte, lxvi. (2), (1872),
pp. 220 223, who expanded n (t+z) in ascending powers of z by Taylor's theorem, used
the obvious formula [cf. $ 9-11 (2)]
(3) 2P
^r-)== * (-) m pCm .On - p+2m (t),
C1) ir-r,
t*-z
= # (*) Oo (0 + 2 ^ 8
<*) fli (0 + W <*) n. (0 + .
.
= 2 c/.'Wfl.^
n=0
* Math. Ann. xliii. (1893), pp. 141 142; see also Epstein, Die vier Rechnung*operationen
mit Bessel'schen Functionen (Bern, 1894). [Jahrbuch iiber die Fortschritte der Math. 18931894,
pp. 845846.]
t Leipziger Berichte, xxi. (1869), pp. 221256. [Math. Ann. in. (1871), pp. 581610.]
I
9*1, we have
1
| z 2*
a
-s ~,=oM+a
2
V> Un W ~ 4 ,r M+a
(n>l)
(n - *)! (2s)! (i '
(3) n i (o-i/'.
On reversing the order of the terms in (2) we find that
1 * n.(2n-tn-l)l{(n-m)!}
<*) nW ~4 mto ~ (n>l)
m!(2n-2/rc)!(02M 2w+2 '
Also
A* @44
(6) +
-
2.(2w-l)(2n-2)
2"
Hen* )
+ 1.2...n.(2n-l)(2n-2)...nf '
where
2 n-l
.= ~
2/i
@ = (2n-l)(2n-3)
4
2n(2n-2) '
_ (2-l)(2n-3)(2n -5)
W6_ "'
(7)
2n(2n-2)(2n-4)
(2m-1)(2w-3)...1
0 =
2n(2w-2)...2
Since < Bm < 1, it is easy to shew by the methods of 211 that
- 2- 2 (n 2 exp 2
(8) j
e n it) \ <* 2* j
t \ !) ( |
),
The reader should have no difficulty in verifying the curious formula, due
to Kapteyn*,
ao) -(')- s
5/r^(a^)-
9*41. The recurrence formulae for Qn (t).
K)
(1)
2
n'(t)
lln{t) = n -^ **<*> 2n 'Q
{n>4
< n >2)
t n-1 n+ 1 n-l' >
-2t/(t*-z>Y= 5 en Jn*(*)fV(0>
n=0
_ 2*i+ " + *(
^ r(/j.+s+l)r(v+%s + l)(fji+i>+s+2m)r(iJ. + v + s+m)
* *n swi + i
=0 i=0
(,m=0 &
it is supposed that z | \
< |
|, and then the rearrangement presents no greater
theoretical difficulties than the corresponding rearrangement of 91.
We consequently are led to consider the polynomial Bnillt (t), defined by
the equation
m
(-
1) R ;**(*)
-n
= 2+'+(m + i>+w)
^T+l
m!rV + v + n-2wi + l) W,
'
n\T(fi + v+l)
x aFa (-n,/ji + l,fj.+ v + n; f* + $* + $, \f- + l v + *J sin
2
<f>).
The result of eliminating Jv+1 (z), Jv+3 (z),... Jv + m_ x (z) from the system of
equations
0, 0, 0, 1
Jp('), 0, 0, -22- 1 (" + l)
Jv . 1 (z)-(2v/z)Jy (z), 0, 0, 1
-2?- 1 (v + m-l), i. 0, 0, o
1, -23~ 1 (v+m-2), 1, 0,
1, -2-i(i/+i-3), 0,
0, 0, _2*-J( + l), 1
0, 0, 1, -22-1
The effect ofsuppressing the last row and column of the determinant by
which Rmiy (z) is defined
is equivalent to increasing v and diminishing by m
unity ;
and so the cofactor of (-)m ~ ./_, (z) is (-) mr " 1 iC-i.r+i (z).
a
that is to say
1 1 1
2*rl(v +m~ 1>
>-2z->(v+m-2)-2z-*(v + m-3)-...-2z->v-
The function -R,,(z) was denned by Lommel by means of equation (1).
He then derived an explicit expression for the coefficients in the polynomial
by a somewhat elaborate induction ; it is, however, simpler to determine the
coefficients by using the series for the product of two Bessel functions in the
way which will be explained in 9'61.
It had been observed by Bessel, Berliner Abh., 1824, p. 32, that, in consequence of the
recurrence formulae, polynomials A n . x (2), Z?n _i (2) exist such that
4 2
'.<)- -2 '
'^- r"
2)
^n-l (*) JiW- *-! M* <*.
It should be noticed that Graft and Crelierf use a notation which differs from
Lommel's notation ; they write equation (1) in the form
1
Ja + > W-^CT <*> Ja <*) " K>-1 <*> '.-1 (*>
the same recurrence formulae as J,+m (z); and hence the analysis of 96 also
shews that
(1) (-)* J_,_ (z) = J. (z) Rm, v {z) + J.v+1 (z) R m- ltV+1 (5).
Multiply this equation by J^ (z) and 96 (1) by J"_+1 (z), and add the results.
It follows that
2 sin vtt
Rm,v{z),
TTZ
j v+m {z) j_ +1 w- z
opir{v + m+p+ l)r{ _ v + 2+p y
J- v - m (z) Jv - (z) = 2
X
n~l n\T(v- m + n + 1 V (v + ) n)
~
H
+ {\z)- m+tn
_ S (~) (- -)n
1
_
n =o n\ T( v m+n + 1) r(y + n)
+ i
(-v+*- P + i) (|*y+g+* mw
+ p + l)!r(-i/ + 2+p)r(i/ + w+p + l)'
P =o(r/i
when we replace ?? in the last summation by m-f jo + l. Now it is clear
that
(jo + l)m+p+i = (m + 2p + 1) = (m + p + !
2)p
(m-\p+l)\ p\(m+p+\)\ p\
and so, when we combine the series for the products of the Bessel functions, we
find that
2 sin vir
R __
(-) m+w (- m + n) n {\z)- m+m- 1
the terms for which n > \m vanish on account of the presence of the factor
( + n)n in the numerator.
in
N (\ n C - - - drZ\~ m +-n
The series (3) was given by Lommel, Math. Ann. iv. (1871), pp. 108111; an equi-
valent result, in a different notation, had, however, been published by him ten years earlier,
Archiv der Math, und Phys. xxxvn. (1861), pp. 354 355.
(3) Jv+m (z)Jm+1 _ v (z) + J_ v _ m (z)J_m _ i+v (z) = 2 (-)m sin vir. R^m v - m {z)\{TTz\ ,
In particular, we have
()
J> M + y_ W - 2 (i + + ),
298 THEORY OF BESSEL FUNCTIONS [CHAP. IX
Formula (5) was published in 1870 by Lommel*, who derived it at that time by a
direct multiplication of the expansions ( 3 4) #
[
(-)'" J-m -i (z) = ^
J cos z.Rm ,i (z) + ^
/
J
sin * . i^ lf , (*).
(8) R\ i
(z) + R*m-,, i (z) = (-)'" R, i-m (4
Finally, if, in 9*61 (2), we replace m by the odd integer 2ra + 1 and then
replace v by v m, we get
2 fm {z)
m=0
^ r (*)s0,
m=0
2 /, (*) i2m _!, + 1 (*) = 0,
by writing the postulated identity in the form
m=0
s /.w (/,(i)Ji;,()-/,.i(i)ii..i, t i(*)}^
and observing that, by 474 combined with 3*2 (3), the quotient Jv -i(z)lJv () is not an
algebraic function. Nielsen points out in this memoir, and its sequel, ibid. (3) vi. (1901),
pp. 331
340, that this result leads to many interesting expansions in series of Lommel's
polynomials; some of these formulae will be found in his Handbuch der Theorie der
Cylinderfunktionen (Leipzig, 1904), but they do not seem to be of sufficient practical
importance to justify their insertion here.
,/_! (Z) {i? m _ liH -i (*) + Rm+1,,-1 (*)} = {/", (*) + /,-(*)} Rm,u(z).
-z-*-*R mtV (z)={z-- m Jv+m (z)} \z-i F_, (*)}-{*--' Y9m (s)}[r-*J - w l (z)},
7T
(6) R' m< v (z) = Rmt y (z) R m-h +1 (z) Rm+ lt y (z),
z
The majorityof these formulae were given by Lommel, Math. Ann. IV. (1871), pp. 113
due to Nielsen, Ann. di Mat. (3) VI. (1901), p. 332; formula (2) has been used
116, but (6) is
by Porter, Annals of Math. (2) in. (1901), p. 66, in discussing the zeros of Rm> v (z).
It is evident that (2) may be used to define Rmv (z), when the parameter
m is zero or a negative integer; thus, if (2) is to hold for all integral values
of m, we find in succession from the formulae
4,v(v + l) p 3 )=
. JLv
R%v\ z) = ~i
1> ^i,A
that
(8) R 0<
y(z)=l, R_ltV (z) = 0, R^y(z) = -1,
and hence generally, by induction,
(9) R- m , y (Z) = (-)* Rm-^y (z).
This formula was given by Graf, Ann. di Mat. (2) xxiii. (1895), p. 59.
If we compare (9) with Graf's other formula, 9*61 (4), we find that
(10) Rm,y(z) = (-)" R-r^^y (z) = - R-.m^ v+m+1 {z) = (-)" Rm ,.y-m +i (z).
1
When the functions of negative parameter are defined by equation (9), all the
formulae (1) (7) are true for negative as well as positive values of m.
300 THEORY OF BESSEL FUNCTIONS [CHAP. IX
(2) Rm< (z) Rm< v+1 (z) - R m+hv (z) m- lt v+1 (z) = 1,
and so the value of the function on the left is unaffected by changing m into
m 1. It is consequently independent of m ; and since its value when m=n
is R n - i<v (z\ we find from 9*63(10) that
(4) Rm-\ v (z) #m _n-i, K+M+i (z) Rm, v (z) Rm-n-2, v+n+1 \ z )~ R n (Z ,v J-
If we rewrite this equation with p in place of n and eliminate Rm ^hv (z) be-
Rm, v (z) [ Rm-p-2, v+p+i (z) Rm-n-i, v+n+1 ( z ) ~~ Rm-n-l, v+n+i ( z ) Rm-p-^ v +p+i \Z)]
Rm, v \Z) Rn-pi, v+p+i \ z )>
by (3). If we transform the second factor of each term by means of 9*63 (10),
we obtain Crelier's result (loc. cit. p. 143),
This is the most general linear relation of the types considered by Crelier ; it
connects any three polynomials R m>v (z) Rn>v (z), RPiV (z) which have
)
the same
parameter v and the same argument z. The formula may be written more
symmetrically
(6) Rn, v {z) Rp-m-i, v+m+i (z) + Rp, v (z) Rm -n-\, v+n+i (z)
+ Rm v ( z ) Rn-p-i, K+p+i (z) = 0,
,
that is to say
"c+n \Z) = J+m \Z) K>nm,v+m \Z) Jy+mi \ z ) -">n m-i, p+m+i (*)>
^Jy+p \Z) Jy+m \Z) ltpm,v+m, \ z ) J v+m-i \Z) Rpm-\,t>+m+\ \Z),
we find that
the last expression is obtained from a special case of (5) derived by replacing
m, n, p, v by 0, n m, p m, v + m respectively.
It follows that
m,n,p
where ^ denotes any cylinder function.
The last two formulae seem never to have been previously stated explicitly, though
Graf and Gubler hint at the existence of such equations, Einleitung in die Theorie der
BmeVschen Funktionm, n. (Bern, 1900), pp. 108, 109.
[Note. If we eliminate J^-i (
z) from the equations
i
J+m () = J* (*) Rm, v () - Jv - 1 () Rm - 1. p+ 1 (*),
that say that the equation 9*6 (1), which has hitherto been considered only for
is to
positive values of the parameter wi, is still true for negative values.]
* It is supposed temporarily that m is the smallest of the integers m, n, p ; but since the final
result is symmetrical, this restriction may be removed. See also the note at the end of the section.
302 THEORY OF BESSEL FUNCTIONS [CHAP. IX
(1) lira
< **)^.^(*> = j (,).
V '
*-. r(i/ + m + l)
This result was applied by Hurwitz, Math. Ann. xxxm. (1889), pp. 250 252, to discuss
Jv (z) when v has an assigned real value ( 15*27). It has also
the reality of the zeros of
been examined by Graf, Ann. di Mat. (2) xxiii. (1895), pp. 49 52, and by Crelier, Bern
Mittheilungen, 1897, pp. 9296. "
Now write
(m n)\T (v + m n + 1) _ Q
(m-2n)!I> + m + l) = ( ' n) >
so that
n . . (m w)(ra n 1) ... (m - 2n + 1)
(m, n) = 7^ r-P TV
+ m-n +
.
-.v / >
numerator of 6 (ra, ri) is numerically less than the corresponding factor in the
denominator, provided that n > N.
Hence, when n> N, and m > 2N,
\0{m,n)\<l,
while, when n has any fixed value,
lim (m, n) = 1.
(->;(*'>+"
Since i
W
is it
lim t hZY+"
e <m, n) = i Jf^T-
and the theorem of Hurwitz is established.
Again,
since the convergence
of 2 rr is uniform in any bounded domain t
=o!
,
ru++i)
,
of values of z (by the test due to Weierstrass), it follows that the convergence of
f An arbitrarily small region of which the origin is an internal point must obviously be
excluded from this domain when R (v) < 0.
9*65, 9*7] ASSOCIATED POLYNOMIALS 303
^# = lim r^t^l
- lim L^-l/I^M-l,
by 963 (1). On carrying out the process of reduction and noticing that
R* v+m-Az) _ _ 1
=2 m_ z x
^i,+m(*) 2z l
(v + m)
we find that
^Hi^-2^ - , \
ttm,, + i (z) 2 (i/ + 1) z- - 1
2 (i/ + 2) r-
1
- . . . -2 + m) z~*
and hence
(2 ) ^W--
J v (z) 2(v
*
+ l)z- -2(v + 2)z- 1
J
1
-...
This procedure avoids the necessity of proving directly that, when m-*-ao, the last
element of the continued fraction
so that
(2) Rm ,r +1 (z) - (hz)~ gmA*)-
By making the requisite changes in notation in 9*63, 964, the reader
will easily obtain the following formulae
(3) 0+i,r (*) = (" + m + l)gm>v (z) - zgm_hv (z), [ 963 (2)]
(4) #,+,, r-i (z) = vgm (z) - zg^, +1 (z),
, [ 963 (1 )]
(5)
IF* bTz
P^'W = *9^A*) +gm+h ,- 1 (z), [J 9-63(7)]
(6 )
^m ^ (,-m-i gmv ( z ))
m+1 _, (*)
<, gm+1<v (z), [ 9-63 (4)]
(7) gm , , (z) gm+i< +1 (z) - gm+2t (z) gm _ 1% +1 (z) = z m g 0< (z) glt +m+1 (z).
* This notation differs in unimportant details from the notation used by Hurwitz.
:
These results will be required in the sequel ; it will not be necessary to write
down the analogues of all the other formulae of
9'6 964.
The result of eliminating alternate functions from the system (3) is of
some importance. The eliminant is
(f + m) gm+2> (z) = cm (z) gm% (z) - (v + m + 2) z*gm _ 2> (z),
(8) - {v + 2s + 2) z g^ {z),
(v + 2s) gv+ % y (z) = Ca, (z) <7M) (z)
3
t
zeros of gtm,v{z) are a ^ real; and also that they are all positive, except when
1 > v > 2, in which case one of them is negative.
To prove that the zeros alternate, it is sufficient to prove that the quotient
a monotonic function of the real variable z, except at the
9*m,( z )l9im--2,*(z ) is
We have f^ %9 (z) - j
w( *M - = - 389**, *-*,
dz {{gtm-^Kz))
The reason why the number of zeros is the excess and not the deficiency in that the
quotient gr^ a decreasing function, and not an increasing function of z,
(z)/2,_2, (z) is
as in the usual version of Sturm's theorem. See Burnside and Panton, Theory of Equations,
i. (1918), 96.
The arrangements of signs for the set of functions when z has the values
oo , 0, oo are as follows
2m 2m- 2 2m -4 2
00 + + + ... + +
+ + + ... +
~
00 (->* (~)m l (
_)m-2 ... - +
-2s>v>-2s-2.
It will now be shewn that*, when v lies between 2s and 2s - 1, gsm ,(z)
t
has no negative zero; but that, when v lies between 2* 1 and 2s 2, g-mtV (z)
has one negative zero. Provided that, in each case, m is taken to be so large
that v + 2m is positive.
It will first be shewn that the negative zeros (if any) of g^^z) alternate
with those of g*m -%,y{?)'
* This proof differs from the proof given by Hurwitz ; see Proc. London Math. Soc. (2) xix.
(1921), pp. 266272.
306 THEORY OF BESSEL FUNCTIONS [CHAP. IX
= ^2m-2, M zg*m-i, { M
+ 9*m+i, v-i( z )} - 9*m, ( z ) l^am-i, 1(2) ~ S'am-i, -(*)}
= (1/ + 2m)0 tM -i,r() + gmi~2, v {z)g*m,v-\(z )-g*m-i,v-i(z)gvmAz)
,
>o,
provided that i> + 2m is positive and z is negative. Therefore, in the circum-
stances postulated, the quotient
The existence of the system of equations 9*7 (8) now shews that the set
of functions
and there are s alternations of sign. When z is zero, the signs of the
functions are
, , -., , +,-,+, ,(-)*,
the upper signs being taken when 2s > v > 2s 1, and the lower signs
being taken when 2s l>v> 2s 2; there are s and s + 1 alternations
of sign in the respective cases. Hence, when 2s >v > 2s 1, #,() ^ as
no negative zero but when 2s 1 > v> 2s 2, gzm,v(z ) nas one negative
;
- 2s > v> 2s 2.
It will now be shewn* that when v lies between. 2s and - 2s 1, gtm, v (z)
has m 2s positive zeros; but that, when v lies between 2s 1 and 2s 2,
<Jvm v( z ) has m2sl positive zeros. Provided that, in each case, mis so large
In the first place, it follows from Descartes' rule of signs that, in each case,
9tm,v{z) cannot have more than the specified number of positive zeros. For,
when v lies between 2s and 2s 1, the signs of the coefficients of
1, z, z\ ...,z*>, z+\ z-+\ *+, ..., zm in ffsm v (z)
,
and since there are m 2s alternations of sign, there cannot be more than
m 2s positive zeros. When v lies between 2* 1 and 2* 2 the corre-
sponding set of signs is
_ _ _ _ _ a. _ ( V-
and since there are m 2sl alternations of sign there cannot be more
than w 2s 1 positive zeros.
Next, we shall prove by induction from the system of equations 9*7 (8)
that there are as many as the specified number of positive zeros.
When v lies between - 2* and 2s the coefficients in gVtV (z) have no
1,
alternations of sign (being all +) and so this function has no positive zeros.
On the other hand
it has no other positive zeros. Next, take <7+4,i,(s); from 9*7(8) it follows
that its signs at 0, a M + ao are +,, + hence it has two positive zeros, and by
, ;
the reasoning already given it has no others. The process of induction (whereby
we prove that the zeros of each function separate those of the succeeding
function) is now evident, and we infer that gm ,
y (z) has m 2s positive zeros,
and no more.
Again, when v lies between 2s 1 and 2s 2, the coefficients in
-
5 4+s,i'(^) have no alternations in sign (being all ), and so this function has
no positive zeros. On the other hand
and so g4a +4t ,(z) has one positive zero, and by the reasoning already given it
- 2s > v > - 2s - 2.
CHAPTER X
FUNCTIONS ASSOCIATED WITH BESSEL FUNCTIONS
This function obviously reduces to Jn (z) when v has the integral value n.
It follows from 6*2 (4) that, when
an integer, the two functions are
v is not
distinct. A function of the same type as J, (z) was studied by Anger*, but
he took the upper limit of the integral to be 27r; and the function J,(z) is
conveniently described as Anger's function of argument z and order v.
In connexion with this function reference should also be made to researches by Lommel,
Math. Ann. xvi. (1880), pp. 183-208.
2m
/"2"-
1 /"2t
cos(v0-2Sin0)cW=
1 fn
cos (p6 - z ain 6) dd + 7T-
If' cos(2vir-vd+zain8)dd
/ / I
*T J *w J 2,1t J
cos 2
vit . J (z) + sin vit cos vn . E (z).
* Neueste Schriften der Naturf. Ges. in Danzig, v. (1855), pp. 1 29. It was shewn by Poisson that
Vv I cos (vO - z sin 6) d0 = (z - v) sin v jt,
Additions a la Conn, des Temps, 1836, p. 15 (cf. 10*12), but as he did no more it seems reasonable
to give Anger's name to the function.
t ZUrich Vierteljahrsschrift, xxiv. (1879), pp. 33 76. Weber omits the factor 1/x in his defi-
nition of Ev (z).
10-1] ASSOCIATED FUNCTIONS 309
it . m ! sin ^yn-
Sin TO 0COSl>0d0=_-=7-; :
r : : .
But, evidently,
2* 1
sin2m
v
J(*)=- 2 /c ; :, 0cosv0d0+ - 2 hr rrr sin sin vddd,
Tr m =o (2m)! Jo 7r m=0 (2m + l)!Jo
so that
co (_)m/l^)m+i
+ Sill &V7T 2 =r- .
, , T Ti .
and similarly
w
(4) B(^) = w sinii/7r
*
2
TO -
^ "7
r(m-| /+l)r(m +
.->-
I
'
i
i/ + l)
1 AZ)
(
~ 2
+ 2 )~(2 -y ) + ,
' inr |_ 2 -i^ (2 -^)(4 2
-i/ 2 2 2
-i/ )(4 -i/ )(6 2 2 2 2
"'J
sin w[" 2 .2? 2 5
+ 12 -;2_ (l 2 -'2 )(3 2 -va ) + (l 2 -i'2 )(3 a -i/2 )(5 a -y2 )~'
7T L
(6)
1
E,(,)= -i:^
s
^[l-^L- +r 2 2
" ^ ,-...1
*"j
(2 -i/ )(4 -i/2)
2 2 2
VTT 2 -i>
_ 1 + cos virY z z 3
z*
~ (l - v ) (3 - + (l ~ '
tt" [l 2 -!/2 2 2 2
i/
2
)
2
-i/2 )(32 -i;a )(5 8 -^)
The formula corresponding to (5) was given by Anger (before the publication of his
memoir) in a letter to Cauchy which was communicated to the French Academy on July
17, 1854 ; see Comptes Rendu*, xxxix. (1854), pp. 128 135.
* M&m. sur les integrates difinies (Paris, 1825), p. 40. Cf. Modem Analysis, p. 263.
310 THEORY OF BESSEL FUNCTIONS [CHAP. X
For a reason which will be apparent subsequently ( 10*7), it is convenient
to write
Z Z3 2?_
1 z2 z*
(8) _,,(*) =-- + 2 yi a _ ^) ~ a(2 _ *) (4 - 2 2
+ ' "
^2
i/ )
(9)
-
J,
/
(z)
%
= sin vrr
s 0t v (z)
, . v sin vir
*_,, (*),
,
,,~x l+cosj/ir
- ,
i/(l cos i/7r) .
E (z) =
. .
(10)
/ ^
So, (^) - *-i. , (z).
have been examined by Unferdinger, Wiener Sitzungsberichte, lvii. (2), (1868), pp. 611620.
Also, Hardy, Messenger, xxxv. (1906), pp. 158166, has investigated the integral
/oc flO
I sin (vO-zsin 6) -r=-,
and similarly
(6) sin vir . E (z) = J_v (z) cos vir J (2). .
The recurrence formulae which are satisfied by the functions of Anger and
Weber have been determined by Weber.
It is evident from the definite integrals that
\* ^-{sm(v0-zam0))d0
irz J d0
_ 2 sin vir
irz
and
2 f'/
B_, (s) + E, +1 (z)
2i>
E (s) = - cos f - -iA sin (v0 -zain0)d0
J J
/i
(1 )
\ t
J,-! / \
(*> + Jt r+1 /(0)\ =
.
Z J,t
2l/ /
(z)
\ 2 sin w
,
TTZ
(^ 2
- v>) J (z) = O - v) \z J v _! (s) + (sin w)/ir}
= z(^ + l v) J v _! (*) {v sin vji)\ir
= z J v (z) + (z sin i>9r)/w - (v sin tnr)/7r,
2
so that
(9) v.J.(,)-^~.
We also have
(^
9
- v*) E (z) = (*-v) {*_! (z) + (1 - cos i*r)/w}
= Z (^ + 1 - v) E,., (2) (1 COS V7r)/7T J/
so that
Formulae equivalent to (9) and (10) were obtained by Anger, Neueste Schriften der
Naturf. Ges. in Danzig, v. (1855), p. 17 and by Weber, Zurich Vierteljahrsschrift, xxiv.
(1879), p. 47, respectively ; formula (9) had been discovered earlier by Poisson (cf. 10"1).
(2)
V '
-^
T e-"-i"h< dt = Sin {
l J (z) - Jv Or)},
J V7T
when jarg0|<^7r; the result is valid when jarg.r| = |7r, provided that
R(v)>Q.
Again, we have
J
10*13, 10*14] ASSOCIATED FUNCTIONS 313
(4) (* e~** inht cosh vtdt=\ir tan \ vir {J {z) - Jv (*)} - \ ir {E (z) + F (a)},
Jo
(5) f" e-zaiaht sinh itf d< = |tt cot i/tt [Jv (z) - J (z)} - %w {E (*) + Yv (z)\.
Jo
The integral [ e-zcosht cosh ^^ has already been evaluated (63); but
Jo
[" e-zco* ht smhvtdt
Jo
does not appear to be expressible in a simple form; its expansion in ascending
powers of z can be obtained from the formula of 6*22 (4),
2siny7r M -z^ht
I_ v (z) + Iv (z) = - PVcose cos v d6 +
7 ( e
sinh vtdt}
7r
7TJo 'O
but, since
[*
cosm
- . .. (-)m sinv7r - /
Fl {- m>
v +
"2~
m_ i/ -f ra \
The formulae (2) (5) are nugatory when v is an integer, but from 6*21,
9*33 we have
(7) I" e- M
- zainht
dt = \ (-)n+1 {Sn (z) + 7rE n (z) + ir Yn (z)}.
Jo
The associated integrals
08 tC
("e-^8m (xaiuh t)dt, l" e-* *(x cosh t)dt
Jo
v
Jo am
have been noticed by Coates, Quarterly Journal, xx. (1885), p. 260.
It follows from 1013 (2) that, in order to obtain the asymptotic expansion
of J v (z) when \z\ is large and |
args| <%tt, it is sufficient to obtain the
Jo
To carryout this investigation we shall first expand cosh i;/cosh t and
sinh i/t/cosh t in a series of ascending powers of sinh t.
* See Anding, Sechsstellige Tafeln der Besselschen Funktionen imaginaren Arguments (Leipzig,
1911) [Jahrbuch Uber die Fortschritte der Math. 1911, pp. 493494], ard
Takeuchi, Tdhoku Math.
Journal, xviii. (1920), pp. 295296.
11
W. B. V.
j
1 M+,l/+) (
ui i/M * )
so that
COshy<
cosh vt _ 1 f<+. !/+.!+) "~ fi"-*(f-l)d
cosh* ~ 2tti
2w7J (^l)2 -4fsinh
:
3
<
Now
(i+)f*+m-*d r(i/ + m + f)
27rtJ
$7rt J (^-l)am+1 r(|i/-m+).(2m)!
_ (-) cos^7r
OT
'
r(m-t-.^ + ^)r(m+|-^)
7T (2m)!
and, if we take p so large that R(p + $$v) > 0, and then take the contour
to be that shewn in Fig. 15 of 7 '4, we find that
If v and * are real, the last expression may be written in the form
<2p)I
+ ,, (zmi+M<irW (2 sinh , r]
(2p)!
For complex values of v and < this equation has to be modified by replacing
the condition < X ^ 1 by a less stringent condition, in a way with which the
reader will be familiar in view of the similar analysis occurring in various
sections of Chapter vn.
Similarly we have
- _L_
f( + ,l/M+)
u h _ u -* = _L
1
z
(
J
1 \ )
d,
2-iri-J * k-w -1/4
L
so that
sinh vt 1_ r(u+,ilu
+ ,i+) $.#
sinh 2t
~ 2iri J ( - I) - 4f sinh
2 2
*
1 r(+, i/+, i+) r p-i 22w* w sinh 2"1 t 2 p sinn 2* "I ,*.
"as J r L: f
(t-ir +a +
<r- 1>* {<r- ly -*r ***}]
whence it follows that, if we take p so large that R (p + 1 \v) > 0, then
^ ^2
(-)^( P + ui y)r( P + i-^) v sinh +1 1
(2p + l)! '
J
* r^.a~qgJg:
- )" r(m+ * (ro+ "^
cosh .
J o
<
;^
(
*
sin^Tr * (-)OT r (w +
2 1 + jQ r (m + 1 - & y)
sinh vt . e-* 8inh * cfe~ 2 2m+2
27T OT= o (f2)
It follows from 1013 (2) and (3) combined with 10*11 (6) that
(2) ii-
-J
;
_ 1 -cosier Vv _ y(2 2 -y ) 2
y (2 2 - v*) (4
2
- i/ ) _
8
"I
'"
irz [z z* z* "J
These results were stated without proof by Weber, Zurich Vierteljahrstchri/t, xxiv.
(1879), p.48 and by Lommel, Math. Ann. xvi. (1880), pp. 186188. They were proved as
special cases of much more general formulae by Nielsen, Handbuch der Theorie der
Cylinderfunktionen (Leipzig, 1904), p. 228. The proof of this section does not seem to have
been given previously.
Since the only singularities of cosh vtjcosht and sinh vt/cosht, qua functions
of sinh t, are at sinh t = i, it is possible to change the contours of integration
into curves in the tf-plane on which arg (sinh t) is a positive or negative acute
angle; and then we deduce in the usual manner (cf. 6*1) that the formulae
(1) and (2) are valid over the sector | arg z\<ir.
316 THEORY OF BESSEL FUNCTIONS [CHAP. X
10*15. Asymptotic expansions of Anger- Weber functions of large order and
argument.
We shall now obtain asymptotic expansions, of a type similar to the
expansions investigated in Chapter viii, which represent J (z) and E (z)
when j
v |
and z are both
| |
large.
Wo ""Jo
imaginary { iff), the singularities are on the imaginary axis and the origin is not one of
them since y is not equal to iti.
T m=
If
where a-=2^j
1 f<
0+ >
_ 9 - cosh 7
a _
1
_ 1
~l+cosh7'
ai ~ 2(1 + cosh 7 )4 ' * ~ 24 (1 + cosh 7 )
r '
* Expansions valid near y = iri are obtained at the end of this section.
10-15] ASSOCIATED FUNCTIONS 317
This expansion is valid when v/z is positive ; it has, so far, been established
on the hypothesis that |
arg z j
< \tr, but, by a process of swinging round the
contour in the T-plane, the range of validity may be extended to cover the
study the curve in which a > 0. It is evident that the curve has the origin as
its centre.
multiple of it.
and, when v = - rr, the maximum value of <1> (u, v), qua function of u, is at
u = a, the value of <J> (u, v) then being
- cosh a sin fi - a tanh a 4- (it - ft) cot ft}.
{
I
If this is negative, the equation (u, fi = has no real root, and so the
<I> -rr)
the continuous curve indicating the shape of the contour when a is positive
Fig. 25.
and the broken curve the shape when a is negative; the direction in which t
increases is marked by an arrow.
318 THEORY OF BESSEL FUNCTIONS [CHAP. X
It follows that the expansion (1) is valid when (a, 0) lies in any of the
domains I, 2, 3.
Next, we have to consider the asymptotic expansion when(or, /8) does not
lie inany of these domains. To effect our purpose we have to determine the
destinations of the branch of the curve 4> (u, v) = which passes through the
origin.
positive, it follows that <p (a, v) is positive when v is greater than tt.
Hence the curve cannot cross the line a = a above the point at which
v = 7T. and similarly it cannot cross the line u = a below the point at which
v tr. The branch which goes downwards at the origin is therefore confined
to the strip a < u < a until it gets below the line v 2Kir + tt #, where
K is the smallest integer for which
TTJo 7m=0 *
If, however, is obtuse and athe branch which goes below the
is positive,
IT Jo ~ m=0 z
(5) if r'--^ S.
'
Since formula (1) is the only one which is of practical importance, we shall
not give the other expansions in greater detail.
If
-
00
grt-tnnht fa _If* g-(-tcosh y +mnh* fa
I
j
The only difference between this and the previous integral is the change in
the sign of cosh 7; and so, when 7 lies in any of the regions numbered 1, 4,
(6)
K
1 rV-iainht^^l X J*)***'
2am+1
'
'
Jo
ir ir mZo
_,_ 1 /
2 ,_ 9 + cosh 7
l-cosh7' (1 -cosh 7)*' ^ 24 (1 - cosh 7)'
This expansion fails to be significant when 7 is small, just as the previous ex-
pansion (1) failed when 7 was nearly equal to iri.
If
i
00
If -00 dt
e,t-zinht dt= -l e" e~ tzt -T- dr
7rJo ttJ dr
A result equivalent to this has been given by Airey, Proc. Royal Soc. xciv. A, (1918),
p. 313.
:
The integral considered by Airy and Stokes ( 6'3) has been generalised
by Hardy* in the following manner:
If s = sinh <f>,
then
2 cosh = 4s + 2
2<f>
2
2sinh3=8s + 6s 3
and generally
T3 (t,a) = t* + 3at
T
= e +4at + 2a
Z 4 (,a) 4 2 2
T (<,a) = t +5at + 5a
6
6 s 2
of Bessel functions but when n is odd, the first only is so expressible, the
;
Gin (a) + iSin (a) = [" exp {%Tn (t, a)} dt.
Jo
* Quarterly Journal, xli. (1910), pp. 226240.
+ The sine-integral in the case n=3 was examined by Stokes, Camb. Phil. Trans, ix. (1856),
radius R
with centre at the origin (the arc being terminated by the points
with complex coordinates R, Re*11"), tends to zero as R-~cc .
And therefore
/ooexp(}irt/)
combination Cin (a)-iSin (a) may be treated in a similar manner, and the
result is then evident.
If
Cin (a) + iSin (a) -
we equate real
nsi ^
and imaginary
/w)
{e*>
parts,
J. Vn (2a*) - rW^ JVn <*)}
we have
^(.)- E
3=S^ 5
jlJ-*(W-)-^(^}.
In a similar manner,
Ein (a) = n
2et* f
J o
00
<*> Ci <"
" 2n inU) [J~ " (2
8
l ^ + ' ) (2/Si0)1 '
It follows from 431 (9) that, when n is even, the functions Ci* (a) and
Sin (a) are annihilated by the operator
In the case of the first two functions it is difficult to obtain this result*
directly from the definitions, because the integrals obtained by differentiating
twice under the integral sign are not convergent.
Cin (a) n
= 2a*
Jo
f
00
= 2a*cosOWn) 2ftin)
n
That is to say,
n& W
/9 is
<2 >
<- (J-' + *' <**"
It follows that the equation 10*21 (4) is true whether n be even or odd
and, whether n be even or odd, Cin (a) is annihilated by the operator
^+ (-)n 2 n ^
for all real values of a.
* It has been proved by Hardy, loe. eit., p. 229, with the aid of the theory of " generalised
integrals.
10-22] ASSOCIATED FTTNCTIOHS 323
Next we evaluate Ein (a) when a is positive making the usual substitution, ;
Ein (a) = n jo
exp (- 2a*n sinh u) cosh (u/n) du
Hence the series which represents Ein (a) when n is odd and a may have
any value is
Ein (a) -
(3)
n cos (i7r/r?) J o r (m + 3 _ i/n) r (m + f + 4/f|)
7T (-)^ttmw (-)m amM )
+ sin;(ir/n) [ g a g
T(m + 1 - 1/n) M=0 *! T(m + 1 + l/n)J
'
|m = w!
and hence it follows that
-2
na on Ein () = wofK""3'.
(4) jj^ + |
7ra*<
n+1 U" " amn
n cos (t>>) m=0 T (m + f - */n) r (m + f + /n)
7ra*
+ n sin /x (M* 7_
.,(-Tr/n) 1/n (2o*) - - W /1/B (2a*)},
and therefore
7ro* (w+1) " <*
mn
-
(5) Sin (a) -
wcos(Ww) J o r(m + t-i/n)r(TO + f + i/n)
2wcos($7r/w)
-1 - n ~2 = - not n -3
(7) |
2
an j Si n (a) <
>
The formulae of the preceding three sections are due to Hardy, though
his methods of obtaining them were different and he gave some of them only
in the special case n = 3.
/ 1\*
1\*/ IV
1 /'
(0+
/ 1\* / l\ m
(i)
JU*--5RJ <~'H) (") dt
* Comptes Rendu*, xi. (1840), pp. 473475, 510511; xn. (1841), pp. 9293; xm. (1841),
It follows that
^.*--2^TTi)/
rn
K) dA'-i)
dt
and similarly
(7) -^-n,k, 2 Cr
m= r=Q t . 2r- n+s -.2r, k-$, m>
(9)
s
^-. *. 2 -
m" r=0 ( )
r
k Cp - r m Cr
. ,
where p=\ (k+mn). This is proved by selecting the coefficient of n in the product
(t+l/0*x(-l/r-
<i) J,. W = ^ P *~ {*
+
))' P }'* (' - j)} *
where w is an integer, and k is a positive integer.
It follows that
1
A..
* (*) = 2,r
f*
I
ex P {~ * ( nB ~ z sin 0)\ (2 cos W de >
and therefore
[Note. An earlier paper by Giuliani, Giornale di Mat. xxv. (1887), pp. 198202,
contains properties of another generalisation of Bessel's integral, namely
1 i'
v
- / cos (nd - z*> sin p 6) dB,
(4) Jn ,
(z) = Jn (z).
Again from 10*3 (2) and (3) it is evident that
~ A*+i - *
(9) -A*+2 (*) = (*) {/-!, * (?) - /+,,*(*)},
V n Jn *(*)=-
ir
f"
.'o
4 {- (n + z cos 0) sin (nd - z sin
du
0)} (2 cos 0f d$
2k n
C
= (n + z cos 0) sin (n0 z sin 0) (2 cos 0)* -1 sin 0d0
7T Jo
= - 2kz J' nk (z) + [\ffi cos (n0 ~ z sin *)} (2 cos 0)*"1 sin ed6
= - 2kzJ'n k (z) - cos (n0 - * sin 0) ^ {(2 cos 0?-* sin 0} d0,
'
7T ./
o <W
and so
d*
Operating on this equation by -r-9 + 1 , and using (10), it follows that
n k
' -
k W -*(*-!) </.* CO.
this is verified by applying Fourier's rule (cf. 2*2) to the function on the
right.
A somewhat similar function J{z\ v, k) has been studied by Bruns, Atr. Nach. civ.
(1883), col. 1 8. This function is defined by the series
J(z;
(13)
*)-J^ w!r (, + 2 * +m )( v + 2*-2)( + 2A)(+2*+2 + 2)'
The most important property of this function is that
(1
>
H - <*> - r$m f.
(1 - ^ 8in
" *
" r^hfel) .C
sin ( " cos $) sin* m
provided that R(v) > - |.
By analysis similar to that of 3*3, we have
The function H (z) is defined by this equation for all values of v, whether
R (v) exceeds $ or not. It is evident that H (z) is an integral function of v
and, if the factor (\z) v be suppressed, the resulting expression is also an in-
tegral function of z.
where
and j
v +f |
is the smallest of the numbers |
v +f I, I
*> +f |> I
^ +$ l
* Mim. de VAcad. Imp. des Sci. de St P4tersbourg, (7) xxx. (1882), no. 8 ; Ann. der Physik,
(3) xvn. (1882), pp. 10081016. See also Lommel, Archiv der Math, und Phys. xxxvi. (1861),
p. 399.
t Programm, Luisenschule, Berlin, 1890. [Jahrbuch uber die Fortschritte der Math. 1890,
pp. 340342.]
J The Analytical Theory of Light (Cambridge, 1904), pp. 392 395. The results contained in
and (11), are there given.
this section, with the exception of (3), (4), (10)
' : + ' ))
and similarly
d _ _ S (-)m (2m + l)g2w
< t*
H " (
* )} =
, .o 2" + - +1 r (m + f ) r ( to +*
=
,r_i sF*"* 5 r (to + 1) r ( v + to + 4
=
2*r( + f)r(iT*~
,'
H +l( *
''
)-
HU + H +1 () = - H, (5) +
(5) (*)
z Y^\Wl\) '
* H " +1 ( * }
(
(8) (* - V ) h (*) = r( y *
+t)r(f ) "
In particular we have
so that
to
(10)
H (z)
The function
Jv {z) has been
satisfies
Ii (a)
the differential equation
V.H. W - r( ^
which bears the same relation to Struve's function as /
studied (in the case
(t)
(z) bears
xlii. (1911),
(11) L ()-
i r(. + | )P(r+(l+t)
Let us take
the contour crosses the real axis, and the contour does not enclose the point
* = -l.
If we suppose that R (v) > \ , we may deform the contour into the seg-
ment (0, 1) of the real axis, taken twice, and we find that
r(i+) ri
2
(t - 1)"-* sin zt . dt 2i cos wr I (1 - 2 )"-*
sin zt . dt,
Jo Jo
where the phase of t* is zero.
1
Hence,
(1)
when R (v) > \, we have
= r <* -
h, ( .)
in yr
1 (. )
Both sides of this equation are analytic functions of v for all* values of v
Jo
v - D-* *. ^ .
and so, by the general theory of analytic continuation, equation (1) holds for
all values of v.
r (1+)
(2) j
v (z) + <*) = ei
* (
'
lr * dt
^rVi)*^ /
To transform a> be any acute angle (positive or negative),
this result, let
and let between \ir + a> and \ir + o>. We then deform
the phase of z lie
the contour into that shewn in Fig. 26, in which the four parallel lines
make an angle a> with the imaginary axis. It is evident that, as the lines
parallel to the real axis move off to infinity, the integrals along them tend to
zero. The integral along the path which starts from and returns to 1 + oo ie~ iu
is equal to ffw ( z ) and on the lines through the origin we write t = iu, so
;
that on them
(P _ i)*-4 = e =F (r- (1 + u*y-l.
It follows that
* The isolated values J, f , |, ... are excepted, because the expression on the right is then an
undetermined form.
10-41] ASSOCIATED FUNCTIONS 331
This result, which is true for unrestricted values of v, and for any value of
z for which - ir < arg z < ir, will be applied immediately to obtain the asym-
ptotic expansion of H (z) when |
z |
is large.
Fig. 26.
If, as in 6*12, we replace a> by args /S, it is evident that (3) may be
written in the form
where \tt < /8 < \ir and \ir + $< arg z < \ir + (3.
t Proc. London Math. Soc. xix. (1889), pp. 504507. [Scientific Papers, m. (1902), pp. 4446.]
332 THEORY OF BESSEL FUNCTIONS [CHAP. X
If we write z=ix in (3), where x is positive, we see that, when R(v)<%>
a result given by Nicholson, Quarterly Journal, xlii. (1911), p. 219, in the special case in
which i>=0.
We shall now obtain an asymptotic expansion which may be used for tabu-
lating Struve's function when the argument z is large, the order v being fixed.
Since the corresponding asymptotic expansion of Y v (z) has been completely
investigated in Chapter vn, it follows from 10*41 (4) that it is sufficient to
determine the asymptotic expansion of
/ooexpi/3 /
t***\-*
As in | 7*2, we have
(p-.l)!*w V ' V W
We take p so large that 22 (v p ) < 0, and take $ to be any positive angle
for which
|/3|^tt-S, |args-/3|^7r-o\
so that z is confined to the sector of the plane for which
7T + 28 ^ arg z ^ ir 28.
We then have
( 1 J ^sinS, arg 1 1 --
J < tt,
so that
1
(l + ^V" " 1
j
^ gftrl/WI Sin ^)2iJW-2p-l
(
= Ap>
I
\ / j
v* ;*
where j
B. I $ =f l
M e *dtt
= (z-w).
+
r
w - y. "^-
( 2) b. (.)
1 n + ^S*r +
(z
The asymptotic expansion* was given by Rayleigh, Proc. London Math. Soc. xix. (1888),
p. 504 in the case v Q, by Struve, Mem. de I' Acad. Imp. des Sci. de St Pe'tersbourg, (7)
xxx. (1882), no. 8, p. 101, and Ann. der Phys. und C/iemie, (3) xvn. (1882), p. 1012 in the
case v = l; the result for general values of v was given by J. Walker, The Analytical
Theory of Light (Cambridge, 1904), pp. 394395.
Hj(*)=(^)*(l-cos*),
(3)
.-(s) K)-V*~0-
10*43. The asymptotic expansion of Struve s functions of large order.
We shall now obtain asymptotic expansions, of a type similar to the
expansions investigated in Chapter vm, which represent Struve's function
H (z) when |
v |
and z are both
j |
large.
/;s(-r:r.
see Rayleigh, Phil. Mag. (6) vm. (1904), pp. 481 487. [Scientific Papers, v. (1912), pp. 206211.]
.
f /- v dw
Accordingly we consider separately the cases (I) in which zjv is less than 1, and
(II) in which z\v is greater than 1.
The curve in the -plane, on which t is real, has for its equation
We now write
2 arc tan (tanh g tan tj)
v> V) - sm ^
cos yi
and examine the values of F(g, 17) as % traces out the rectangle whose corners
0, A, B, C have complex coordinates
As tges from to A, F(^, i/)is equal to 2 sinh /cosh 2 , and this steadily
increases from to 1.
V2 . arc tan ( J
. cosec 17,
Note. To establish this result, write tan i> = t *J2 and observe that
=
dt {r aroten
7 <v(i+2<*) li+^
arcten
7 ^'
because
-= -5- arc tan t, which vanishes with t, has the positive derivate ~ ^ .
When f is on BC, F(g, ij) is equal to ir sech , and this increases steadily
from tt/V2 to 7T as f goes from to C; and finally when is on CO, F(^, rj)
is zero.
Hence the curve, on which F(^, rj) is equal to sech a, cannot emerge from
the rectangle OABG, except at the double point on the side OA ; and so the
part of the curve inside the rectangle must pass from this double point to the
singular point G.
The contours in the w-plane for which a has the values 0, \ are shewn in Fig. 27 by
broken and continuous curves respectively.
Fig. 27.
d
l- h Tm
336 THEORY OF BESSEL FUNCTIONS [CHAP. X
then
* m!&
^ 2.
1=0 *
and hence, by 10 4 #
(1), we have
(1)
{TOT*) 3FJ
Hence, from 1041 (8) it follows that
provided that arg z\<\ir. This result can be extended to a somewhat wider
j
In particular, we can prove that, in the case of functions of purely imaginary argument,
[Note. If in (I) we had taken the contour from w=0 to w=*e~ and thence to w= - i,
a
p (lz) \mtri n +m
= v
= _r(|m+l)r(m + w + l)'
,
and
that is to say
<n " w
n
(i)\ e ^= 1
E-^> ^ 1>r
( ili<if)r! I _ Hw(2>
,,i
(z)
1
H
^i r(n + i-t)
In like manner, when n is a negative integer,
n+1 1
r (n - m - j) (js)-"*-^
(2)
/., _ (~)
B -(*) ^ Ic*'"" '
r(m + |)
.
H"w( ^
positive and v has any positive value greater than or equal to \. This result,
which was pointed out by Struve* in the case v = 1, is derivable from a
definite integral (which will be established in 1347) which is of con-
siderable importance in the Theory of Diffraction.
(h)"'
1
[*" d cos (x cos 6) ..,-,-
)jgi*~edB
"-W- rfr'tW
/ ^
.I.
(
-
l> tf)r(i) {[eos(xcos9)sin-J
- (2i/ - 1) |
'cos (x cos 0) sin 2'- 2 '*?
cos 0dd
Jo
*Y
^"p - (2v - 1) ^cos (* cos 0) sin2"- 3 cos 6dd\
.
1 1 f
V 1(
= ( i*? i^iV ./ f'sin2
*- 2 -'?
cos II - cos (x cos -9)1 <W
1 (i> + )r()
^0,
since the integrand is positive.
* Mem. de VAcad. Imp. des Sci. de St Ptersbour<j, (7) xxx. (1882), no. 8, pp. 100101. The
proof given here is the natural extension of Struve'- proof.
,
A comparison of the asymptotic expansion which was proved in 10*42 with that of
Yv (x) given in 7*21 shews that, when x is sufficiently large and positive, H (x) is positive
if v >
\ and that H
K (x) is not one-signed when v < ; for the dominant term of the
asymptotic expansion of H (x) is
F -
(ir) /2\* . . . , .
r(,+|)rd)
or
U> sin ^-ii-)
according asi*>ori<. The theorem of this section proves the more extended result
that Struve's function is positive for all positive values of x when v >\ and not merely
for sufficiently large values.
The theorem indicates an essential difference between Struve's function and Bessel
functions ; for the asymptotic expansions of Chapter vn shew that, for sufficiently large
values of x, Jv (x) and Yv (x) are not of constant sign.
*j {/o()-l(*-"f j^r'dt
and choose the contour to be the imaginary axis, indented at the origin*, and then write
+ttan |0, we find that
*
j {h (x) - Lo (*)} = f cos (* cot <f>)
log tan (i*r+|tf>)
^^
and so
a formula given by Theisinger, Monatshefte fiir Math, und Phys. xxiv. (1913), p. 341.
If we replace x by *' sin 6, multiply by sin 6, and integrate, we find, on changing the
order of the integrations in the absolutely convergent integral on the right,
iW
t E, (* tan log cot (<) ^$-r = *'
{/ (x sin 6) -L (x sin 6)} sin 6d6
<f>)
| f
so that
1 X
(2)
/o*'E (^tan0)logcot(^)^ = |. -^-
1 ,
on expanding the integrand on the right in powers of x. This curious result is also due to
Theisinger.
* The presence of the logarithmic factor ensures the convergence of the integral round the
indentation.
:
The integral
a* I e
izt
P v _i(t)dt,
7T
and then it is easy to verify the following recurrence formulae, either by using
the series (2), or by using recurrence formulae for Legendre functions
2ii-"z^e- iz
(3) W_ v 1 (z) + W ^{z)
v
2v
z
W ' (Z)
j(2Tr)r(-v)T(%+v)\'
2ii- v z~le- iz
(4) W_ a (z) - W r+1 (z) = 2W ,
(z) -
^2ir)r^-v)r^ + v) '
*"
(4 - *. 4 + " 5 1 ; 4 - 4') in ascending powers of 1 - 1, and integrating term-by-term.
S +
8 r(m-y + )r(m+i/ + ) - \m
P _ fcos*7r\
^<-*> i *\
/l
^-^"j .* (ml? VY)
x log
| \T~) ~ 2^ (m + 2 ) + ^ (w ~ v + ) + ^ ( + * + $)} >
and since
Some functions which satisfy equations of the same general type as (1) have been
noticed by Nagaoka, Journal of the CoU. o/Sci. Imp. Univ. Japan, iv. (1891), p. 310.
v / \ % 1
(n-m- 1)1 WM M ,
,
=0 "* j
+
?. ^(^)r <
2 10
<*'> ~ * <ro + x
>
- * < + m )
The series on the right may be expressed as the sum of four functions, each of
which has fairly simple recurrence properties, thus
(1) nrYn (z) = 2 {log <!)-* (1)} Jn (z) - n (z) + Tn (z) -2Un (z),
* Cf. Barnes, Quarterly Journal, xxxix. (1908), p. 111.
10*6] ASSOCIATED FUNCTIONS 341
where
w =o ra!(n + ro)!
and (cf. 3*582)
The functions 7^(2) and Un (z) have been studied by Schlafli, Math. Ann. m. (1871),
pp. 142 147, though he used the slightly different notation indicated by the equations
r ( ,\ - [l I (-ra*r*~ 1
jfnW
~La m >riw _ i r(m+i+6)r(n+m+i-6)J e o
_j_ra | (-)w (|^)w+wt /+> (i + ty*** , i
m+1+e
-&r* [&>-*-* (n + 2w)! J_, * J,.o
(-)w (jg)w+2OT
i_ - /
(o+)
(i w+2w log t
tj) ,
nie
(- 1> sin w+2w - tt)
__2 f' |
~iriJom>-in-i
e
(n
fl)
+ 2m)\
. (0
^
where t has been replaced by e *""r,i (
.
It follows that
Now
3 ( ta" sin 0)n +*l _ (cosh ( 1> sin 0) (n even)
_ ttt . t
m>in~i (n + 2m)\ (sinh (- iz sin 0) (n odd)
= |{ e
- sinfl
+ (_)n eiz sin } (
and so
yn (s) = 7rt
-L P(0 _ 1^) f
e
-8in# + em(e-r)+izsin} <&
Jo
t Einleitung in die Theori-e der Bessel'tchen Funktionen, iv (Bern, 1900), pp. 4269. Lommel'a
treatise, pp. 77 87, should also be consulted.
342 THEORY OF BESSEL FUNCTIONS [CHAP. X
If is replaced by ir in the integral obtained by considering only the
second of the two exponentials, the formula (4), which is due to Schlafli, is
obtained at once.
un ^ (-rq*)w+2w r (i + 6)
W = -fi ?
Ide Zo m T (n + + 1 + e) J'.-
!
i
=- I "(h
~ 0) sin (z sin - n0) . {2 cos - 2n/z} d0
7T J o
= -4 cos hmr
2
8
4
Jn (z),'
v
z z
Thus
(1) Tn. x (z) + Tn+1 (z) = (2n/z) Tn (z) + 4 {cos \mr - Jn (z)}/z. 2
Again Tn'
(z) =- !"($ ir-0) sin cos (s sin (9 - n0) d0,
7T Jo
and so
(2) r-iW-ZV+ito-MV^).
From these formulae it follows that
With the aid of these formulae combined with the corresponding formulae
for Jn (z),
Y(s) and Sn (z), we deduce from 10*6 (1) that
(6) Un_, (z) + Un+l {z) - (2ft/*) Un (z) - (2/z) Jn (z),
(7) Un _, (z) - Un+l (z) = 2 Un (z) - {21z) Jn (z), '
We now define U-H (z) by supposing 106 (I) to hold for all values of n ;
,
2, sin 2m0
2 7r a = Z
in the integral for Tn (z), and then integrate term-by-term. This procedure
needs justification, since the Fourier series does not converge uniformly near
= and = ir, and, in fact, the equation just quoted is untrue for these two
values of 0.
To justify the process*, let 8 and e be arbitrarily small positive numbers. Since the
series converges uniformly when 8 % $ ^7r-o\ we can find an integer m^ such that
I .. ...
M sin2m0|
(*tr-d)- 2 |<c,
I ro=l m I
* The analysis immediately following is due to D. Jackson, Palermo Rendiconti, xxxn. (1911),
pp. 257262. The value of the constant A is 1-8519...
r .
Ljt-0-
2 2 = / {l+2cos2< + 2cos4* + ... + 2cos2Jft}(fc
m =i i ye
_
i1* sin(2i/"-H)f
_J_ ,
'
J 9 t sin t
. (2JT+ l)t
f i* sin 1 . /"(if+J)T sin x
= An- / a= ,
Jjt /
,
eta:
for some value of between 8 and ^r, by the second mean-value theorem, since /sin t is
By drawing the graph of .r -1 sin# it is easy to see that the last expression cannot
f n sin x
exceed \n I dx in absolute value ; if this be called \ rrA, we have
*
/ o x
21 8
["{{ln _ 6 )- 2 -*^lsin(* S intf-tf)^|
,
<
-{/:
+
/r*aiK--i !!
^i-'---'-
T 2 J
- w0) dd
and that, in 3582, this definition was identified with the definition of Un (z)
as a power series given in 106 (3).
V
= J_
\
(\ir - 6) Jm (z) {e'(t8in-n9+me) _ e -*(^in9-n+)J ffl
771 '
o i=-oo
The pi-oof of the formula given by Graf, Math. Ann. xliii. (1893), p. 141, is more com-
plicated ; it depends on the use of the series of 10*62 combined with 2*4.
There seems to be no equally simple expression for Un (z + t).
10'7. The genesis of Lommel's functions $M> (z) and sM) (2).
(1) V v y = kz+\
where k and fi are constants. It is easy to shew that a particular integral of
this equation, proceeding in ascending powers of z beginning with z** is 1
,
+i
, r z* z*+*
(2) * + ]
+ 3)* - v*\
wt r^-^+m+Dr^+^+w+i)
For brevity the expressions on the right are written in the form
ks^ v {z).
is
. F. 12
346 THEORY OF BESSEL FUNCTIONS [CHAP. X
Next let us consider the solution of (1) by the method of "variation of
parameters." We assume as a solution*
y=A(z)J v (z) + B(z)J_ v {z),
where A (z) and B (z) are functions of z determined by the equations
Jv (z) A'(z) + J. = 0, v (z) B' (z)
A (z) =
^ V J- = - ^~i^- VJ
v '
-
2 sin w J
f v (z) dz, B(z)
2 sin vir J
[ v (z) dz.
it sin vir \_ J J
Similarly a solution of (1) which is valid for all values of i>, whether
integers or not, is
(8) ,^ r (,)*_[,/,(,)
l sin vir
\*z*J_ v {z)dz-J_ v (z) FzJ,(z)dz\ .
|_ j o jo J
(10) ^ %- (,)
y tl)(^> + l)
* Cf. Forsyth, Treatise on Differential Equations (1914), 66; it is supposed temporarily that
v is not an integer.
f The generalisation of this result, obtained by replacing zn+i in (1) by an arbitrary function
of z, was given by Cbessin, Comptes Rendus, cxxxv. (1902), pp 678679 and it was applied by ; ;
him, Comptes Rendus, cxxxvi. (1903), pp. 1124 1126, to solve a sequence of equations resembling
Bessel's equation.
10-71] ASSOCIATED FUNCTIONS 347
q (r \ * (-rr(ij*-iF+|)r(jj + ii' + l)
K '
m=o r(i + l)r(i; + m+l)
**'W- fr + iy-*
that is to say
(1) W.., (z) = **
+1 - {i> + 1) - 8
:*}
M (4
i
(4) {2v/z) SrtV (z) = (/M + v-l) s^ _j (z) - (fi - v - 1 ) $_i. m-i (*)>
( 5) 2s'M (0) = (/i + v - 1 ) v-j, .-i (*) + (/* - v - 1 ) Sp_ ,+1 (z).
, lt
The reader will find it easy to deduce from 10*71 (2) that the functions of
the type SntV (z) may be replaced throughout these formulae by functions of
the type #Mi (z)
(6)
(7)
;
W
so that
= **+1 - {(fi + 1) - #M( , (*),
(*)
These formulae may be transformed in various ways by using (1) and (6). They are
due to Lommel, Math. Ann. ix. (1876), pp. 429432, but his methods of proving them were
not in all cases completely satisfactory.
10'73. Lommel's functions Sfl>v (z) when fiv is an odd negative integer.
The formula 1071(2) assumes an undetermined form when fi v or fi + v
is an odd negative integer*. We can easily define Sv-vp-\ >v (z) in terms of
Sv-i,(z) by a repeated use of 10*72 (6) which gives
n\ <?
*-i
J.
(-fi^ +
(-)*&.,,,(*)
(1)
>*+*.,(*)- o 2^(- J ,)m+1 ( v -p)tn+1 2^! (1-,)/
* Since SMl (z) is an even function of v, it is sufficient to consider the case in which ix -v is an
odd negative integer.
10*72, 10*73] ASSOCIATED FUNCTIONS 349
We next define Sv-h ,(z) by the limiting form of 1072 (6), namely
pW
L 3/*
1
_U,-i
U + 1,n I
-i-iw
.> +
(-)"(^)"
l)ir( + m + 2)
m
x {2log z + yfr(l) + yfr (p + 1) - ^ (m + 2) - yfr(v + m + 2)}.
Also
(3) s^y-i'-rwsJrlfSli)
x {2 log \z - ty (p + m + 1) - yjr (m + 1)} - 2"- 2
ir V (v) Y (*),
and this formula, which appears to be nugatory whenever v is a negative
integer, is, in effect, nugatory only when v = ; for when v =n (where n is
-VW- (/t+1)2
whichgives 'Sr
-'.o(*)=^[p{^ +1 -^ +S.o()}] _i
W
Mss
From the descending series given in 1071 (1) it is evident that Neumann's
polynomial On (z) is expressible in terms of Lommel's functions by the equations
j
#J v (z) dz,
f
z" Y v (z) dz
{z Jv (z) . zl ~> 5M_ ]> _ 1 (*)} = zJv^ (z) /SM _ ,_ (g)+(jt-v- 1) zJv (z) SM_2,, (*)>
1 1
dz
d_
dz
{Z ~" /"_! (Z)
l
.
*-* SM> (*)} = - Zjv (Z) S^iz) + (fl+V-l) zj^ (Z) S^^ (z).
(3) P ** Jv (z) dz = (ii, + v-l) zJv (z) 8^,h _, (z) - zJv-x {z) Sp {z\ t
(4)
[* z Yv (z) dz = (p + v-l)zYv (z) /S^^-i (z) - zY^ (z) ,(*),
and, more generally,
(5) j* z ^
(
v (z)dz = ( i + v-l) z<@v (z) 8^ _, (*) - *#_, (*) 6V, (z).
f
Special cases of these formulae are obtained by choosing p and i> so that
the functions on the right reduce to Neumann's or Schlafli's polynomials, thus
Of these results, (1), (3), (4) contained in Lommel's paper, Math. Ann. ix.
and (6) are
(1876), pp. 425444; were given by Nielsen, Handbuch der Theorie der
(6) and (7)
Cylinderfunktionen (Leipzig, 1904), p. 100, but his formulae contain some misprints.
/n 2r r(y + m) v+ 2m Q
(8)
. .. 2 v+1 T(v+m + l) v+2m+l
\ III . a
It follows that the most general case in which the integral (5) is expressible
in terms of elementary functions and cylinder functions is given by the formula
v + 2m
The function denned by the series
* ,+l,lw
w r(v+2m+i) r(v-i)
has been studied in great detail by W. H. Young t this Function possesses many properties ;
analogous to those of Bessel functions, but the increase of simplicity over Lommel's more
general function seems insufficient to justify an account of them here.
The
f* J (t)
an integer) by H. A. Webb,
integral v I ,
'. dt has been studied (when v is
Jo t(z t)
Messenger, xxxin. (1904), p. 58; and he stated that, when /=, its value is On (z). This is
incorrect (as was pointed out by Kapteyn) ; and the value for general values of v is J
It follows that the expression given above is equal to the sum of the
residues of
T(j-1ui +lv-s)r(l-l H,-$v-s) nrQir
2
r(i-i * + jior<j-*A*-*i')
i
sin^TT '
at the points
0,-1,-2,. ..,-<_p-l),
1,2,3,...,
so that
_-i (-yr(i-^-i-iir+m)r(i-^-|y + w) . ,
sini>7r
'(*)
have been studied in great detail by Sonine*. They will be called hemi-
cylindrical functions.
Sn (s)=/(D).S (*),
combined with
and therefore
(3) Bm {M)-ll[-D + J(I* + l)}+[-D-</(l> + l)}*].B,(s).
If it is supposed that (1) holds for negative values of w, it is easy to see that
(4) S_ (*)=(-)"Sn
tt (*).
1+ 2! ? + ~4i * (even)
] re ,. re (re
2 - l2) .. ,,
'-.
fc
odd)
-jjf ^-31 (re
Hence
2 re
2
Cre
2
22 )
I
So <*) + |j
S" (#) + \, ' S (,)+... (re even)
(5) Sn (^)= *
n(n2 _ 12)
Take as contour a circle G with centre z such that S (f) has no singularity
inside or on the circle. Then
,, ( , +0 ifiM
%z
ZlTl] c t
Sw (s + t)= i % n Jn (I)
J
B* (0 On (t- s) %
f*
V di) Sm W = Sm_n ( z) + ( ~ )n &m+n ( ^'
and so the formula (1) is still valid. We thus have an alternative proof of
the formulae of 53, 91, 934 and 1063.
* Math. Ann. xn. (1880), pp. 48. See also Konig, Math. Ann. v. (1872), pp. 310 340; ibid.
xvh. (1880), pp. 8586.
10-81, 10-82] ASSOCIATED FUNCTIONS 355
where fv {z) and g (z) are given arbitrary functions of the variables v and z,
form an obvious generalisation of the pair of functional equations whereby
cylinder functions are defined. It has been shewn by Nielsen* that the
functions / (z) and g v (z) must satisfy the relation
and it has been proved f that, if this relation is satisfied, the system can be
reduced to a pair of soluble difference equations of the first order.
Again
<* - *2 ) F v (jr) - <* + *) [- zFv+l (z) - fi m (*)]
= - z*Fv (z) + za v+1 (z) -(* + p) &, (z).
We are thus led to the equation
(8) /_, (*) +/ +1 (*) - (2-/*)/r (*) = 9.-i (*) ~ 9*+i (*) " 2S^' (*)
It now has to be shewn that Nielsen's condition is sufficient for the exist-
ence of a solution of the given system. To prove this, we assume (8) to be
e
+ zYv ^ (z) \d + %tt J (t) (t) dtl + faz Y (z) J (z) TB V (z)
j
= zJr-, (z) c^ - \tt -j
J
Y^ (t) _, (t) dtl
+ zY _ x (z) <*_, + |tt l" J_, (t) w^, (0 dtl (z),
I
that is to say,
*/"_! (z) I
r - c w -WJ"{ Y* (t) v v (t) - Y^ (t) <*_, {t)\ dtl
{^U (z) &_, ( jr) - V. (z) a (jr)} - r (z) <& (s) - r r _, (z) V^ (z),
Tz
since (6) and (7) are satisfied ; and so (3) is satisfied if
and the reader will have no difficulty in verifying that, if these same two
difference equations (with v replaced by v + 1 throughout) are satisfied, then
the value of Fv {z) given by (9) is a solution of (4).
If, as z -* *>
/(#)-0 (**-), 5r,(*) = (**-'),
where ir^v) and 7r 2 (i/) are arbitrary periodic functions of v with period unity.
Note. Some interesting properties of functions which $ati*fy equation (2) only are to
be found in Nielsen's earlier paper, Ann. di Mat. (3) v. (1901), pp. 1731. Thus, from a
set of formulae of the type
F_v , (*) F v + !
(z) - (2v/) *T () = 2g y ()/,
it is easy to deduce that
* An account of various memoirs dealing with such equations is given by Barnes, Proc. London
It has been proved ( 4*73) that Bessel functions are not algebraic functions,
and obvious from the asymptotic expansions obtained in Chapter vn
it is fairly
that they are not simply periodic functions, and, a fortiori, that they are not
doubly periodic functions. Consequently, in accordance with a theorem due
to Weierstrass*, it is not possible to express JV (Z + z) as an algebraic function
of ,/ (Z) and Jv (z). That is to say, that Bessel functions do not possess
addition theorems in the strict sense of the term.
There are, however, two classes of formulae which are commonly described
as addition theorems. In the case of functions of order zero the two classes
coincide ; and the formula for functions of the first kind is
Heine J, who obtained the formula as a confluent form of the addition theorem
for Legendre functions.
zr = V( + * - 2* cos
2
0),
and all the variables are supposed to have general complex values.
which is valid for all (complex) values of vr and a, the integrand being a periodic
analytic function of 6 with period 2tt. We next suppose that a is defined by
the equations
ts sin ol Zz cos <f>,
ta cos a = z sin <f>,
*
J () = jj- f
exp {i ( - z cos 0) sin + iz sm < cos 6) dd
* [*
j | JM (Z)e w l efe " in( *"' )rftf
27T J _ ( m=-x )
n
= I Jm (Z)\ e
mie+izain <* 6)
dd
= J_ % Jm (Z) f
"
e
mi <+> - 8in<> ^
- X Jm {Z).Jm {z)e*,
m= <*>
The corresponding formulae for Beasel functions of order | were obtained by Clebsch,
Journal fiir Math. lxi. (1863), pp. 224227, four years before the publication of Neumann's
formula; see 11*4.
11
3. Grafs gen eralisation of Ne um arm's formula.
Neumann's addition theorem has been extended to functions of arbitrary
order v in two different ways.The extension which seems to be of more
immediate importance in physical applications is due to Graf*, whose
formula is
Z)Jm *"*'
{^-^Y - J_J'+<
(1) *<) ( {Z)
pp. 59 61. A special case of the result has also been obtained by Nielsen, Math. Ann. ui. (1899),
p. 241.
360 THEORY OF BESSEL FUNCTIONS [CHAP. XI
Graf's proof is based on the theory of contour integration, but, two years after it was
published, an independent proof was given by G. T. Walker, Messenger, xxv. (1896), pp. 76
80 this proof is applicable to functions of integral order only, and it may be obtained
;
To prove the general formula, observe that the series on the right in (1)
is convergent in the circumstances postulated, and so, if argJ?=a, we have
Now write
(Z - ze-* )t = vru, (Z- ze*)\t = m/u,
where, as usual, -st = \/(Z + z 2Zz cos
z 2
<f>),
and it is supposed now that that value of the square root is taken which makes
sr -*- + Z when z -*~ 0.
For all admissible values of z, the phase of ns\Z is now an acute angle,
positive or negative. This determination of r renders it possible to take the
u- contour to start from and end at ao exp ( i/3), where = arg m.
We then have
where yfr
-*- as z --*- (so that, for real values of the variables, we obtain the
relation indicated by Fig. 28), then Graf 3 formula may be written
Fig. 28.
If, in this formula, we change the signs of v and m, we readily deduce from
3-54 that
(6) ^o*)V=
sin
5
ifi= oo
^(^^w^^ Bin
The formula was given by Neumann in his treatise in the special case i>=0; see
(5)
also Sommerfeld, Math. Ann. xlv. (1894), p. 276; ibid, xlvii. (1896), p. 356. Some physical
applications of the formulae are due to Schwarzschild, Math. Ann. lv. (1902), pp. 177 247.
If we replace Z, z and r in these equations by iZ, iz and irsr respectively,
it is apparent that
(8) K.(v)"!*v+- 5
Sill TO= 00
K +m (Z)Im (z)*m<f>.
y
BUI
Of these results, (7) was stated by Beltrami, Atti della R. Accad. di Torino, xvi. (1880
1881), pp.201202.
the special case in which v an integer and the only functions involved are
is
of the first kind, the inequalities se*** < Z need not be in force.
| I | |
362 THEORY OF BESSEL FUNCTIONS [CHAP. XI
em+n
K }
ttr ~,t Zn zn d(cos<}>)n
MV(C+A)} _ * (-hhyjv+P m )
pro ,to
^.p\ql Z'+*
and z\
But, by 521,
Jy+P+g (Z)
hP+g(Z) _ S q\ v + p+2k T( v + p + k) T
Z kZ k\{q-k)\ 29 T(v + p + q+k + l)
and so
xt
<*'(-)* 2 w-2*r(i/+m-A;)cos", - 2*< .
of (1 2a cos <f>
+ a )~" in ascending powers of a. We have therefore obtained
2
the expansion
J
(2) ^ip = 2' T ()
W
2(v + m)
m-0
-^^ '^# Cw
It Z
' (cos <f>),
which is valid for all values of Z, z, and d>, and for all values of v with the
exception of 0, 1, 2,
(3)
In the special case in which v =
= . I (m + 4) ^>
, we have
Unlike the formulae of 11*4, the formula is true only when \z\ is so
p= P
J^>(?)
= I (- Zz cos <f>)P
^+ z, ri{v+p) j_ v y( Z , + ^2)}
v v
2*
v
2*
(-) p+1 (v +p + 2 k)T(-v- p- g - k) zp+* cos? <f>
J- V-V ^k (Z)
2*
v% qZ k =
{\ - v-p-k)
2n p \k\(q - k)\V
+1 {v
+p+ v - p - 2A;
2fc) T (- / 2k - n) z
zPv+*+ cosP <f> J-^-p-a (Z)
_ v t ? (-)p+^
(~)p (y 2k)
"",-0*0.-0
=
2k+M p\k\nlT(l-v-p-k)
2*+"p!ib!fi!r(l-v-i>-A)
2 '
t
=
00 00
(-) P+k 2 ,+p ^ +p + 2k) Tjv + p + k) cosp <ft J-r-.p_*(Z) Jy+p+2k (z)
?o *?o piu " * *
co <jm (_)m-fc 2 >-Hm-2fc ( + m ) f ( y + m - k) COS "*-* <ft
/-,-> (Z) J.+m (z)
~ Jo *-o ( " 2*)"1 A! *'
J-"-w
= 2*i 2 (-r (> + m)
m=0 "
;
(Z)
^^ cm Z
* (cos #,
|2^cos<!<|Z2 + *2 |, \z\<\Z\.
Now the last expression is an analytic function of z when z lies inside the
circle of convergence of the series*
"r(l-i'~TO)r(l+i' + m)
'
m.
and this circle is the circle of convergence of the series
2 CW(cos<).
(^J
Hence the given series converges and represents an analytic function of z
provided only that ze{* \<\Z\\ and, when this pair of inequalities is satisfied,
|
\ze*\<\Z\.
(2) ^^- )
2"*r(-v) 2 (-r(v-m)Z"Jv^n (Z)z"J_ v+m (z)Cnr v (eos<f>).
Again, if we combine (1) with 11*4 (2), we see that, for the domain of
values of z now under consideration,
(3) ^P = 2- T () ( + m) %^- %^ Cm }
* (cos *),
TO Tlt=0 ^ *
we find that
The formulae (1) and (2) have not been giveu previously; but (3) is due to Gegenbauer,
and (5) was given by Neumann in his treatise (save that the functions Ym were replaced
by the functions F<m >). The formula (3) with v equal to an integer has also been examined
by Heine, Handbuch der Kugelfunctionert, i. (Berlin, 1878), pp. 463 464. Some develop-
ments of (4) are due to Ignatowsky, Archiv der Math, und Phys. (3) xvm. (1911), pp. 322
327.
(6) ~ = 2" r 00 5
OT =
(-)- {v + m) %^- %& CV }
(cos 0),
(8 ) J^ = 2-r( y
)J o
(, + m) ^/ ^^Cm Z)
-(cos0).
Of these formulae, (8) is due to Macdonald, Proc. London Math. Soc. xxxn. (1900),
pp. 156 157; while (6) and were given by Neumann in the special case v=%.
(7)
366 THEORY OF BESSEL FUNCTIONS [CHAP. XI
The formulae of 11*4 and of this section are of special physical importance
in the case v = \. If we change the notation by writing ka, kr and for Z,
/
\
pm (cos $), .
,=o v v?
- 5 (-r(+j)i=^^i^>i>,(cos),
m=0 Va V?'
e xp {- A; yXr 2
+ ft
2
2ar cos 0)}
'
' v (^ + - 2 r
2 '
cos ^)
The following special cases of (4) were pointed out by Gegenbauer, and
are worth recording
If <f>
= 7r, we have
(12)
TZTW~~ ()
w -
( M ' & * 'm\P(2v)'
If <b
= \ir, we have
(13)
(z + ^^" 2
-o
( } (
*' *" ' i '
If Z= 2, = 0, and r
df is taken to be JVi
2 2 *r(i/)r(i/ + i) 2 .
, x
r (2j/ + 7> ra /x
7
y.i.W _ 2 .
r ( + m) j^C)l Ci/ (C03 ^.
2
(2s sin 9)* = I )
Gegenbauer, loc. cit. gives also special cases of this formula, obtained by taking
11*42] ADDITION THEOREMS 367
. [=0 (mp)
sin 2" <j> Cm " (cos <f>)
Cp ' (cos <f>) d<f> < _ -nY (2v + 7n)
J
~ p)
I 2i
"- 1
( v + m).m\{T(v)Y ^ l
(17)
J. ( y + ^-2^cos^ " < C0S *> Sm " d <f>
_ 7T T (2l> + Wi)
2"-1 .mlT(v)
^ +W1 (g) J,+w (2)
Z" z*
Asimple proof of this formula t, in the special case in which and the cylinder w=0
functions are functions of the first kind, was given by Sonine, Math. Ann. xvi.
(1880),
p. 37. Another direct proof for functions of the first kind is due to Kluyyer, Proc. Section
of Sci., K. Acad, van Wet. te Amsterdam, xi. (1909), pp. 749755. An indirect proof,
depending on 12-13(1), is due to Gegenbauer, Wiener Sitzungsberichte, lxxxv. (2\ (1882)
pp. 491 502.
[Note. An interesting consequence of (4), which was noticed by Gegenbauer, Wiener
Sitzungsberichte, lxxiv. (2), (1877), p. 127, is that, if \ze* \<\Z\ throughout the contour
of integration, then (cf. 9*2)
^^-^^ -
( +)
(18 >
2Tj rf 2 ^W-(^-)^9^^(cos0).
Special cases of this formula, resembling the results of
9-2, are obtainable by taking
equal to or r.]
The method used by Gegenbauer, Wiener Sitzungsberichte, lxx. (2), (1875), pp. 616,
to obtain the addition theorem of 11-4 is not quite so easy to justify as Sonine's
transformation. It consists in proving that Q is a solution of the partial differential
equation
3 2
Q 2i> + 1 do 1_ d*Q 2v cot dQ
+
dz* z dz
+ z*
dtf
+
J~ fy + o = -
Modern Analysis, 16-51 and Proc. London. Math. Soc. (2) xvn. (191S),
pp. 241246.
t Formula (16) has been given in the special case ,-=0 by Heaviside, Electromagnetic Theory, in.
(London, 1912), p. 267, in a somewhat disguised form.
368 THEORY OF BESSEL FUNCTIONS [CHAP. XI
of am in the expansion of (1 2a cos < + a2 ) " ". And then Bm qua function of z, satisfies the
,
differential equation
ozl z vz ( z J
so that Bm is a multiple of z~"Jv+m ()> the other solution of this differential equation not
being analytic near the origin.
From considerations of symmetry Gegenbauer inferred that Bm qua function of Z, is
Q= 2 bm ^ - lV(cos<),
m=0 > z
(2 ) e fcco8* =2 T(i/) i
m=0
(v + m)imJ^z ^Cm -(co8<j>);
Bauer's result is obviously the special case of this expansion in which v\.
In the limit when v -*- 0, the expansion becomes the fundamental expansion
of 2-1.
deducible from the expansion of 11 41 (4) by
Gegenbauer's expansion is
multiplying by Z" +i and making Z -*. co it is then apparent from -11*41 (9) ;
and (10) that the physical interpretation of the expansion is that it gives
the effect due to a train of plane waves coming from infinity on the axis of
harmonics in a form suitable for the discussion of the disturbance produced
by the introduction of a sphere with centre at the origin.
^ efecos* = 2 jri_^?l_*2''+
&!(m-2fc)!
m-*k (v + m)T (v + m-k)Jv+m (z)
m =o*=o
m=0
" +m
4) e -*cos* = 2" T (*) 5 (-)* (> + m) Gm" (cos
*y
( <f>),
m=0
"+2"? 1 ( **
(6) sin(*cos<f>) = 2"I 2 (-)" (y + . 2m + 1) 0Wi (cos ft),
i=0
n-0 * m,i
* r(>^)r(gr(, + )^ j^w.
(8) jV~*fi,()^-
The last is a generalisation of Poisson's integral, which was obtained by a
different method in 332. It is valid only when R{v)>-\.
These formulae are to be found on pp. 363365 of the first of Gegenbauer's memoirs
to which reference has just been made.
Equation (1) was obtained by Hobson, Proc. London Math. Soc. xxv. (1891), p. 59, by a
consideration of solutions of Laplace's equation in space of 2* + 2 dimensions, 2v+2 being
an integer.
-1
I Cm' (cos <f>
cos + sin <f>
sin <f>'
cos ifr) sin2 " yfr dty
<f>
-2-|r (
. i '^^)%Mc^(cos^)^
\Av-\-m)
A z
(cosf),
to-0
Cf. Gegenbauer, Wiener Sitzungiberichte, lxx. (2), (1874), p. 433; en. (2 a), (1893), p. 942.
370 THEORY OF BESSEL FUNCTIONS [CHAP. XI
and so
'
sin #)""*
exp T^ ,.
rL COg
<X
^ cog J/1J
^
<f>
The integral used in the proof converges only when R (v) > 0, but the final
result is true for all values of v, by analytic continuation.
This result was given by Bauer, Miinchener Sitzungsberichie, v. (1875), p. 263 in the
case v=\; the general formula is due to Gegenbauer, Monatshefte fur Math, und Phys. x.
(1899), pp. 189
192 see also Bateman, Messenger, xxxni. (1904), p. 182 and a letter from
;
Gegenbauer to Kapteyn, Proc. Section of Sci., K. Acad, van Wet. te Amsterdam, iv. (1902)
pp. 584588.
= 2" WK
J(2n) I ( - )m r (" + w )-(" + 27n) <?* + **(*) Jy + Zm(Z)
m=o m! 2" Z" '
so that the expression on the left is a symmetric function of z and Z\ this formula also
was given by Bauer in the case v \.
= cos* <f>
cos" O sin" <f>
sin" 4> 2 (-)"(/* +v + 2n + 1) J^+H-zn+i (*)
n-0
x
.Tro> + .. + i)(r(, + i))' r'<-^" + - 1 '
1; ' +1; ""'
which is valid for all values of /j, and v with the exception of negative integral
values.
Some of the results of 11*5 are special cases of this expansion, which was
discovered by Bateman* from a consideration of the two types of normal
solutions of the generalised equation of wave motions examined in 4'84.
We proceed to give a proof of the expansion by a direct transformation.
-
It is easy to deduce from the expansion ( 5 21) of a Bessel function as a
series of Bessel functions that
\zJ IL {z cos <fr cos <$) J (z sin < sin <I>)
+2m+1 (cos6cos<t>>t+2m
'5 (-) w,
(^y r , .
AX
[~(-)m cosaro
^
= cos* d> cos' 4> sin" 4
. . .
sin" 3>
^ "
2 tT1/
6 cos2"1
, ,,
<1>
x 2 -U/* + v + 2m + 2n + 1) ^- n\T{v
tr>/ ,>. /mh-2to+m-i(*)
-ol + l) ,
= cos* cos* <t> sin" < sin* 3> 2 m* + + 2n+l) JM+ +2n+
i' i (2)
-o L
( (-) cos'"* + v + n + m + l)
<ft cos2"1 &.r(fi
X
=ot mS(n-w)!r(i/ + l)r(/* + m+l)
x 2 F (m-n,/A + i'4-m + w + l;
1
v + l; sin2 <sin2 <&)U
^
= cos* <6 cos* <P sin"
.
<f>
.
sin* <P
. "
2 + v+2
r(/i
pru
+ l)r(/* + +
1x r/ ^ i\ ,
i' n + l) r
W.'+sn+i 1*7
/x
x^ 4 (-n,/* + + w + l;
i/ fi + 1, v + l; cos2 <cos2 <I>, sin2 <sin2 <I>) ,
formation we assume that R (fj.) > 0, though obviously this restriction may
ultimately be removed by using the theory of analytic continuation.
r -o-o slr\(jjk+l),(v+l)r
"r-o.-o
5 ^ (-nX^ + v + n + lU,
r\(v+l) r t-o
(-)W^y>
t\(s-t)l _
- (-)W-*<S>
!(/*+!).-
This transformation has not been previously noticed to exist except in the special case in
f-
which #=#, see Appell, Journal de Math. (3) x. (1884), pp. 407 428 ; some associated researches
are due to Tisserand, Annates {Mimoiret) de V Observatoire (Paris), xvni. (1885), inm. C.
372 THEORY OF BESSEL FUNCTIONS [CHAP. XI
= V *2
"r^O -0 <- r
T -r^K
2 (- n\
+
+ v + n + 1 ),+, (-)+" sin* ft sin8"
+t (^
l)r(t-r)\(r + S- t)l (u-r)\ (fl+ 1 ),+,_
,
<S>
= S I I V
(~ Vh (ft + y + w + l)r+< (-)'+ sin ft sin8"
-0 -0 r-0 ,-t-r r !( + !)r (t - r) \{r + s - (u - r) !(/* + l)r+,_ !
8'
<S>
tt
W+M
= I 5 i (- )t Q* + y + n + l)t (y + < + tt + !),.< (-) sin ft sin
2" 8f
<S>
x 8 F (-
1 ) /i + + + l; v +1 ; sin8 <).
x(-)n +1
r Tiv-^(- w>/* +,/+n + 1
1
' " J
sin2 ^)
DEFINITE INTEGRALS
The reader may here be reminded of the very important integral, due to
Sonine and Gegenbauer, which has already been established in 11*41, namely
(C S
,
x
.
,,
Sm ***
a
,
w *>
J i* + *-2XcoB*r
= *r(2y + m) <@v+m (Z) Jv+m (z) '
2*- l .m\r(v) Z" z*
The formula
which is valid when both R (/*) and R (v) exceed - 1, expresses any Bessel
function in terms of an integral involving a Bessel function of lower order.
(1884), p. 979.
^r~-^
l + 1) Jo
1 \V
V* (z sin 0) sin* I
+1 6 cos2 " +1 6d6
= 2
,^
n
2' i+,
.
'
,_
+2w m!r
rWr
(/*
m^
+ + l)T(v+l)j
'
tt I sin 2 ^ *
2 1
cos'-"* 1 0d0
() m (A 2V + " +2rw+1
= 2
1=0 m!r(/t + i + w + 2)
and the truth of the formula is obvious.
It will be observed that the effect of the factor sin'* +1 in the integrand
is to eliminate the factors T (/t + m + 1) in the denominators. If we had taken
1- **
sin 6 as the factor, we should have removed the factors ml. Hence, when
R (v) > 1 and fi is unrestricted, we have
J*7,
In particular, by taking i/ = , we have
(4) f
* M (* sin 0) J (, cos 0) tan *+i 0<Z0 - p^i'V^ff J, (*),
when R(v)> R (/*) > 1. This may be proved by expanding / (z cos 0) and
integrating term-by-term, and finally making use of Lommel's expansion
given in 5*21.
It will be obvious to the reader that Poisson's integral is the special case of (1) obtained
by taking /*=-.
For some developments of the formulae of this section, the reader should consult two
papers by Rutgers, Nieuw Archie/ voor Wiskunde, (2) vi. (1905), pp. 368373 ; (2) vn. (1907),
pp. 8890.
If (I, m, n) are the direction cosines of the line joining the centre of the
* Due to Rutgers, Nieuw Archie/ voor Wisknnde, (2) vn. (1907), p. 175.
1
*
r (fi + \) V (\) (|*)"+1 f J, (z sin 0) sin* +1 cos2* +1 0d0
Jo
JJl>0,m>0
= (i*>t+, +1 ff
JO
'
V 008 ' sin*+f,,+1 cos2* <f> sin2" +1 </><tyd0
2r<^+ y +t) Jo
= K(/*+*)r(z/+i)r(j)/M+ ,+1 (^),
and the truth of Sonine's formula is obvious.
An integral involving two Bessel functions which can be evaluated by the
same device* is
fin
J v {z sin2 0) J v (z cos2 0) sin2 "* 1 cos2 "+*0d0,
Jo
in which, to secure convergence, R(v) > \.
If we write
w 2 = sin4 + cos4 - 2 sin2 cos2 cos <f>
= 1 - sin2 20 cos2 f <f>,
(*)"/T J{zJ(l-n*)} 2
"n*v da>
2^ I> + ) r(J).//,
JJ^o,>o (1 -n )*" 2
l
* '
=
2+*r(*+i)ra) f J o
J-(zsinO ) s ^+l6cos '
2 ''
<
t
>coa,i '' dd0d<f>
f*Jw (z sin
2
Jw (z cos 2
0) sin "+>
2
cos2"+' 0<*0 -
(1) 0)
^M^lffi '
* This integral has been evaluated by a different method by Rutgers, Nieuw Archief voor Wis-
kunde, (2) vn. (1907), p. 400; cf. ateo 12-22.
, ' , , ;
Some integrals which resemble this, but which are much more difficult to evaluate,
have been the subject of researches by Bateman, Kapteyu and Rutgers; see 12 2.
-
As a simple example of an integral which may be evaluated by the same device, the
readermay prove that, when R(v)> &,
^ (v+ |)
by writing the integral on the left in the form
The formula
and, in fact, he obtained the formula of 1211 from it by dividing both sides
of the equation by Z" and then making Z -*-0.
A simple method of proving the formula is to expand the integral in powers of z and Z
and to verify that the terms of degree p + v + 2m on the left combine to form
(-y**z v (Z2+z2 m )
ml r(p+v + m + 2)
The proof by this method is left to the reader.
* r( + *) k"
a*wfiv-
J (z sin e) J" (z cos e) sin,t+1 e cos " +1
e dd
u*Y(hzy I
Jo . oJo
V
\ \\
'oii?n>0,n>0
eizl+iZn cos * m^ n2+i 8 Q [
i>>
yf,
da ^
J OJ J >0,J>0
JO.' J -Jir
= I
Jo Jo Jo
/ /
"^ 8in 8sin * ( * " * +Z sin w cos2 ''
^ S ^n ^ cosi" ^ sm2
1 " +2
^ ^ d<l> d0.
J
eim sin 0sin*co8* coa ^ *0dyfrd<f>d0,
Jo Jo Jo
and this integral may also be obtained from its preceding form by replacing
z by r and Z by zero. On retracing the steps of the analysis with these
substitutions we reduce the triple integral to
eiw sin cos * s in2M+i Cos2" +1 6 sin2** < sin 2 " yfr d<f> dyfr d$,
o
io Jo Jo
= r(y+|)r(j) rr giw? m^ w2 +1 dw
r(j/+l) JJ m >0,n>0
F(v + 1) JJl>Q,m>0
_ i!ii)_li} f'f
r<v+i) Joio
V^ CO9
'sin2* +2 " +2 COS2* <^m 2 " +1
<<&*> c0
= r(Ai + i)r(y+^)r(j)
r
f' e
iwcos<
sin^+2,+2 ^^
2r(/*+v + f) Jo
= ^r (/i+i)r(, + i)^^r) ,
expressions.
Sonine's own proof of this formula was based on the use of infinite dis-
continuous integrals, and the process of making it rigorous would be long and
tedious.
The formula may be extended to the domains in which - \> R (n) > - 1,
and \> R (v) > 1, by analytic continuation.
provided that
JR( M )>-1, R(v)>-$.
This result is also due to Sonine, ibid. p. 45. In connexion with the formulae of this
section the reader should consult Macdonald's memoir, Proc. London Math. Soc. xxxv
(1903), pp. 442, 443
W.B.F. 13
( ^ ^
r ) rm
r'
/ J
f*
e* s (cos ' s * +sin ' sin * cos * )
^ cos ^ sin2 "
sin2,"_1 ^ ^
= ^ffit?,7 0/ (n) m2- d
(
ff e te <*+*"to+) 1
= (^smjr-i
ff
If we retrace the steps of the analysis, using the last integral instead of its
immediate predecessor, we find that the original integral is equal to
(
i??ptC ff e
ia
Cr v (I cos yfr-m sin ^n'^dco
1{?) (?) * J- >o
=( ^Sm ff4
I (") A (i)
(I
JJwi>0
*
izn
0/ (n cos yfr- I sin &) m "" 2 1
da>
r(v)
/
I ()
2 1
^^^.
J ,'o
= r!
p( ^ ,^ 2"^
( )
(
S1
)
"*
C/(cOSVr) f V 008 '
0/(008 0)8^0^,
(1) I
Jo
'e^costfcos * j, (^ s n # i gi n ^ (7^ ( cos #) gin^+i Odd
v sin*-*
=
(y ) i ^ 0/ (cos f) Jv+r ().
If we equate real and imaginary parts, we obtain Gegenbauer's formulae
_ f(-)*
r
( )*
sin""1 1 C/ (cos -f Jp+r (z), ) (r even)
(0 (r odd)
and
2
.'0
{^ (-w, + i/+n+l;i/ + l; sin
/*
2
<)}
2
cos2" +1 <f>
sin 2 "+1 <cty
nir^ + n+ixr^+i)}*
(/* + + 2n + 1) T (fi + v + n + 1) T (v + n + 1)
1/
* Journal fUr Math. lvi. (1859), pp. 149175 [Werke, vi. (1891), pp. 184202].
/ ,
that is to say
2M
Jo
^ (0 / (, - y = J o {^-i (0 + J, + i (01 ^ (* - 1) dt -2 M+ (#),
so that
This formula is due to Bateman*; some special cases had been obtained
independently by Kapteyn f, who considered integral values of fi and v only.
It will be observed that we can deduce from (2), combined with 2*22 (2),
that
Jo t ' Z t \fi v) z
and therefore
u = zJ (z) 4- A cos z + B sin z,
* Proc. London Math. Soc. (2) in. (1905), p. 120. Some similar integrals occurring in the theory
of integral equations areexamined by the same writer, ibid. (2) iv. (190G), p. 484.
t Proc. Section of Sci., K. Akad. van Wet. te Amsterdam, vn. (1905), p. 499; Nieuw Archie/
voor Wiskunde, (2) vn. (1907), pp. 2025 ; M6m. de la Soc. R. des Sci. de Lige, (3) vi. (1906), no. 5.
The formula
(4) Psin (z - 1)
Jo
^ *
dt = -l
/*=<>
(-)" JM+an+1 (z),
which is valid when R (fi) > 0, is of a more elaborate character, and the result
of the preceding section is required to prove it.
= {' Jl ^r-Mt)dt+j;(z)
Jo zt
= \*J(z-t)lP-dt + j;{z)
Jo t
by 122.
By the method of variation of parameters (cf. 7*33), we deduce that
_ ,/,({)
v = A cos z + B sin z + /*
.
I
f* . .
sin (z - t)
.
*- ,
dt,
Jo t
- (** +8 ),
and, since v
^ T(fjb + 2) +
when z is small, it follows that, when R (fi) > 0,
A=B = 0.
(5) f
"cos (* - >^~ dt =- I (-)" 6W J"^^ (4
382 THEORY OF BESSEL FUNCTIONS [CHAP. XII
12*22. Hardy's method of evaluating finite integrals.
(!) l*W&t)J.Wt>&0^+*Mt-$$g^y ^,
which is valid when R (/*)>- and R (v) > -
The method is more elaborate than any other method described in this chapter, because
it involves the use of infinite integrals combined with an application of Lerch's theorem t
on null-functions.
iV
Let ( J^(zri 8w*e)Jv (si*coa*6)r* + ** +3 sin + i6coai + i0d03fl ''
(r) t
By changing from polar coordinates (r, 6) to Cartesian coordinates (x, y) and using
$ 13*2 (5) we see that, whenever t > / () | |, then
= j\xV (-r*t).Mr)dr,
and hence, by an obvious modification of Lerch's theorem, f
t (r) is identically equal to
V + 5+...
2
+ n+m - ; :
I ,
,*
dt;
I Jo i
if we substitute this result in the coefficients of the ascending series for T (), we obtain
the formula in question, namely
T. W -(r+*W.r.W-y ) 1) '
( '""r--
i 2Jn (z) - (<-*+<*) J {z sit)
-/ d
l-t
I must express my thanks to Professor Hardy for communicating the method to me before
the publication of his own developments of it. The method was used by Ramanujan to evaluate
many curious integrals ; and the reader may use it to evaluate the integrals examined earlier in
this chapter.
The form of the theorem required here is
t Acta Mathematics xxvn. (1903), pp. 339352.
that, if /(r) is a continuous function of r when r > 0, such that
exp(-J-2t)./(r)dr=0
/:o
for all sufficiently large positive values of t, then/(r) is identically zero.
CHAPTEK XIII
INFINITE INTEGRALS
(II) Replacing the Bessel function by Poisson's integral, changing the order
of the integrations, and then carrying out the integrations.
(IV) When two Bessel functions of the same order occur as a product
under the integral sign, they may be replaced by the integral of a single
Bessel function by Gegenbauer's formula (cf. 12*1), and the order of the in-
tegrations is then changed j\
(V) When
two functions of different orders but of the same argument
occur as a product under the integral sign, the product may be replaced by
the integral of a single Bessel function by Neumann's formula ( 5*43), and
the order of the integrations is then changed.
(VI) The Bessel function under the integral sign may be replaced by the
contour integral of Barnes' type ( 6*5) involving Gamma and the
functions,
order of the integrations is then changed ; this very powerful method has not
previously been investigated in a systematic manner.
Infinite integrals involving Bessel functions under the integral sign are
not only of great interest to the Pure Mathematician, but they are of extreme
importance in many branches of Mathematical Physics. And the various types
are so numerous that it is not possible to give more than a selection of the
most important integrals, whose values will be worked out by the most suitable
methods care has been taken to evaluate several examples by each method.
;
(i)
lyMJM dt ~7(^-V
where R (a) > 0, and, in order to secure convergence at the upper limit of in-
tegration, both the numbers R(a ib) are positive. That value of the square
root is taken which makes a + V(aa + &) >
| \
I I-
~~
ir Jo a ib cos &
= l/V(a 8
+fe2 ),
and the formula is proved.
re- at Jv {bt)t- l
dt.
Jo
|
b |
< a | |. If we expand the integrand in powers of 6 and integrate term-by-
term, we find that
I
e-**J {bt)tr-*dt= 2 \r/ lio.iv / ^ + ' +am_1 *~ at *
(-)*" (& bY +3m r (/x + v + 2m)
_ ^
"~
m .o m r (v + m +
! 1) a" + " +aw
(2) re-otJAtyP-'dt
(Wayr(fi + v) +v /* + +! !/ 6'\
fp .
+1
a"I>+l) M 2 ' 2 '" '
a*)'
The result has, as yet, been proved only when R (a) > and I b \
< a
\
|
(3) re-rtJrWV-'dt
Jo
+1 '" + 1
a*l>+l) V + a) M 2 '2 '
*)
(Wrfr+) //* + * i-/ + . .+ i. ft \
~ (& + >}<'+') r(v + l)* l
\ 2 ' 2 '
'
aa + 6V*
It is obvious that interesting special cases of the formulae so far discussed may be
(6)
These formulae were obtained by Gegenbauer, Wiener Sitzungsberichte, lxx. (2), (1875),
pp. 433
443; they were also noticed by Sonine, Math. Ann. xvi. (1880), p. 45; and Hardy,
Trans. Camb. Phil. Soc. xxi. (1912), p. 12 while Beltrami, Atti della R. Accad. delle Sci.
;
di Torino, xvi. (18801881), p. 203, and Bologna Memorie, (4) n. (1880), pp. 461505, has
obtained various special formulae by taking /*=1 and v to be any integer.
Other special formulae are
- /-,(*) dt
y(d?+W)- ay
(7)
J".
7- vb
(8)
(8) e-<j
e r
at) dt -
Jv {bt)dt w( a *+ b2 )-<*r
Jo 6V(a2 + ft2)
[Note. It was observed by Pincherle, Bologna Memorie, (4) viii. (1887), pp. 125 143,
that these integrals are derivable from the generalised form of Bessel's integrals ( 6*2) by
Laplace's transformation (cf. 9 15). This aspect of the subject has been studied by
#
Macdonald, Proc. London Math. Soc. xxxv. (1903), pp. 428 443, and Cailler, Mem. de la
Soc. de Physique de Geneve, xxxiv. (1902 1905), pp. 295368. The differential equations
satisfied by (5) and (6),qua functions of a, have been examined by Kapteyu, Archives
Ne'erlandaises, (2) VI. (1901), pp. 103116.]
The integral /
J o
-^j- J^bt)
sinn, itt
tdt was obtained by Neumann, Journal fur Math, lxxii.
(1863), p. 46, as a limit of a series of Legendre functions (cf. 14 64). The integral does -
not seem to be capable of being evaluated in finite terms, though it is easy to obtain a
series for it by using the expansion
A series which converges more rapidly (when b is large) will be obtained in 13*51.
Some same general type are given by Weber, Journal fur Math. lxxv.
integrals of the
(1873), pp. 92102; and more recently the formula
(9)
r Jy(bt)rdt = (2b)T( v +$ )
1
The special case of this formula in which i=0 had been given by Callandreau, Bull, des
Set. Math. (2) xv. (1891), pp. 121124, two years before Hobson published the general
formula.
(2)'
v f%-<^,g,(*8inha)P'<ft-
Jo sin ^^
+
{ft v) it
T f> - v + 1) Q/ (cosh a),
provided that R (p + 1) > R (v) j j
The modification of (1) which has to be used when the argument of the
Legendre function f is positive and less than 1 is
(3)
[%-<* jy (j s n ) p. dt = T (fi + v + 1) Pf (cos ),
i
Jo
r
-^(^ v + 1)-
(4) [V ""gy,
Jo
T^
(tain
V
. ~x
)*(&
'
= t
-
sin u7r
7 ,
,
Sin(/A+1>)7T
7T
(5)
(5)
[**'*** I
IA)
()
cfr
= &-j(cosh) '
Jo Vt V^tt)
1-
/" , cosvtt Q* * (cosh a)
(6) e- tt ^*IT.
9 (t)V- i
,
dt^-r/ r- ,'"\.
sm (m + y) 7T V(i7r) smh"-*1 , .
Jo
Jo sinn*-* a
and this formula is valid when R (fi) > R (v)| |
and R (cosh a) > 1.
If we take cosh a 0, we deduce that
(8)
p KAt)^dt^^r(^y(t^y
o
1
e-Pt J<uv/
i (zt)dt=
V(p +*8)'
2
/;o
It is suggested that, since e~# J
(zt) is a potential function, the integral on the left is a
potential function finite at all points of real space except the origin and that ont the plane
3=0 it is equal to 1/p, and so it is inferred that it must be the potential of a unit charge at
the origin. But such an argument does not seem to preclude the possibility of the integral
being a potential function with a complicated essential singularity at the origin, and so
this reasoning must be regarded as suggestive rather than convincing.
For various researches on potential theory with the aid of the integrals of this section,
the readermay consult Hafen, Math. Ann. lxix. (1910), pp. 517537. For some develop-
ments based on the potential function
These formulae, or special cases of them, have been examined by the following writers:
Beltrami, Bologna Memorie, (4) n. (1880), pp. 461 505 ; Atti delta R. Aecad. delle Sci. di
Torino, xvi. (1880 1881),
201205; Sommerfeld, Konigsberg Dissertation, 1891;
pp.
Gegenbauer, Monatshefte fur Math, und Phys. v. (1894), p. 55 and Macdonald, Proc. ;
(" - a Jll?dt = *
e t
[\/( a 2 + 2-2cos0)-a}(l+cos0)cty,
Jo t 2ir J o
so that the integral on the left, which was encountered by Rayleigh, Phil. Mag. (5) xlii.
(1896), p. 195 [Scientific Papers, iv. (1904), p. 260], is expressible as an elliptic integral.
(3) j
o
cos at I (bt) K (ct) dt =
jo v{a2 + (6 + c)2 _ 46csin2 }
-
This was discovered by Kirchhoff * as early as 1853 the reader should have ;
v
(4) ( e- at f- J^{bt)Jv {ct)dt
r(i/ + )r(|) Jo (a 2
+ 2mccos<^-c2 cos2 ^>+6 2 )'
t +i'
r e~
Jo
2at
J (bt) Jv (bt) t^" dt=-
"""Jo
rJof "e-^Jn+r (2bt cos<) 0*-+"cos (fx -v) <f>.d<f>dt
) M
r+ *
7T* Jo (a + 6coB 1 "
*'
(2) /." e
~ iatJ
> M J*M *~%j&+W t
(2 2 +6')*-2("2 +&2 )^
(3) f" c-~J*(b)3t
where the modulus of the complete elliptic integrals K and E is &/>/(* +&2 ). Beltrami's
corresponding formula is
where R(a)>\R(b)\, arid the modulus k t of the elliptic integrals is b/a. The formulae (3)
and (4) may be modified in a similar manner.
It was stated by Gegenbauer that the integrals in (2), (3) and (5) by
are expressible
means of elliptic integrals, but he did not give the results in detail; some formulae
deducible from the results of this section were given by Meissel, Kiel Programm, 1890.
[Jahrbuch liber die Fortschritte der Math. 1890, pp. 521 522.]
13*24. Weber's infinite integral, after Schafheitlin.
The formula
J v (t)dt T(^)
Jo
p-*+i 2"-* +i r (v - $n + i )
in which < R (u.) < R (v) + \, was obtained by Weber* for integral values of
The formula is of a more recondite type than the exponential integral formulae given
in 13*2 ; it may be established as a limiting case of these formulae, for, since the conditions
of convergence are satisfied, we have by 13 2 (3) -
rm
Jo t"-^ 1
lim
a Wo
r fV-H + l
A direct method of evaluating the integral is to substitute Poisson's integral for the
Bessel function, and then change the order of the integrations this is the method used by;
Schafheitlin, but the analysis is intricate because the result is established first for a
limited range of values of /* and v and then extended by the use of recurrence formulae
and partial integrations.
R (fi) < and R (v) > \, we then have (the integrals being absolutely con-
vergent)
r(o+)jr (t)dt /"( +> /"*"
L Fo^ =
2-r (,
1
+4)
2l Sin /47T. r (ft) COS /A7T
ra) j + .
<"*~ J. <)*''*
/"*"
_ 2i sin/x7r. T (^fi)
"^-^r^-i/i + i)'
By the theory of analytic continuation, this result is valid when p and v are
subjected to the single restriction R (fi) < R {v + ).
When R(fi) > 0, we deform the contour ir.to the positive half of the real
axis taken twice, and we at once obtain the Weber-Schafheitlin formula.
The integral*
/("+> H,(-<)cfc
may be treated in exactly the same manner; the only difference in the
analysis is that cos(tfcos0) has to be replaced by sin (2 cos 0), and so, by
Euler's formula (adapted for contour integrals), the factor cos \ fitr has to be
replaced by sin \pnr.
It is thus found that
f
(o +) H r (-Qcft _ 2t sin /iir. T (|/t) tan \fitr
(3)
Jo
-
T iTT.
W* *
2
~ir^**' 27r 8 ' 8 ^'2 '
was used by Struve, Ann. der Physik und Chemie, (3) xvn. (1882), p. 1014, to tabulate
1^ f
a>
H 1 (2t)dt
for both small and large values of x. The last integral is of importance in the Theory of
Diffraction.
[Note. By differentiating (1) under the integral sign we obtain Weber's result
(4) j / (01ogft--y-log2;
this formula has also been investigated by Lerch, MonaUhefte filr Math, und Phys. I.
f
Yv it) dt _ r (fr) r (fr - v) cos (|M - y ) w
provided that |
#(i>)| </?(/*- v)<. This result has been given by Heaviside, Electro-
magnetic Theory, in. (London, 1912), p. 273, when i>=0.]
(2)
Jo
r J (at) exp (-pH*)
v .
0*-1 dt
I^(\a\t).\ex V (-^)\.\t^\dt
Joo
/Jo
is convergent, it is permissible to evaluate the given integral by expanding
J, (at) in powers of t and integrating term-by-term.
* Journal filr Math. lxiz. (1868), p. 227. Weber also evaluated (2) in the case fi=y + 2, v being
an integer.
t Math. Ann. vni. (1875), p. 469. See also Gegenbaaer, Wiener Sitzungsberichte, lxzii. (2),
(1876), p. 346.
J This restriction may be disregarded if we replace the definite integral / by the contour
(0+)
+ m + l)J rx r
Jo m= orn\r(v '
)
=o m ! r (i/ + m + 1 " 2/>" +'l + 2m
(3) ("jriatJexyi-pH^.t^dt
Jo
= ex {i'-to+H'+iiij,)-
Vr(,ti) P("v)- fl
and so the integral is expressible in finite terms whenever fi v is an even
positive integer.
In particular, we have
^) ,
provided that R(v) > 1. This integral is the basis of several investigations
by Sonine, Math. Ann. xvi. (1880), pp. 35 38; some of these applications are
discussed in 1347.
*
(6) f F (a*) exp (- pH*) dt
j o
-" W exp
(" 2
) I/" (*?)
tan V7r +
* Kw ($)
sec w]
when |
fi (i) |
< ' ; and, if we make p -* 0, (a being now positive), we find that
(7)
Z
Jo
00
Y
2v
/
(at)dt
x 7
=
tan
u>
vir
,
when |
R (v) \
< %, by using 7'23 ; and, in particular,
(8) rY (t)dt =
Jo
0.
13*31] INFINITE INTEGRALS 395
Formulae (5) and (6) were given (when i/=0) by Heaviside, Electromagnetic Theory, III.
Another method of evaluating the integral on the left of (3) is suggested by Basset,
Proc. Camb. Phil. Soc. vm. (1895), pp. 122 128 ; the integrals have also been evaluated
with the help of Laplace's transformation by Macdonald, Proc. London Math. Soc. xxxv.
(1903), pp. 428443; see also Curzon, Proc. London Math. Soc. (2) xm. (1914), pp. 417
440 and Hardy, Trans. Camb. Phil. Soc. xxi. (1912), pp. 10, 27, for formulae obtained by
;
The result of applying the formula 11*41(16) to the integral just dis-
cussed is to modify it by replacing the Bessel function under the integral
sign by a product of two Bessel functions of the same order.
If r = V(a + 6- 2ab cos
2
<f>)
and if R (p) > - \ R (2v + fi) > 0,
,
|
arg|> |
< \ir,
we thus deduce that
Like the result of 13*3, this equation is due to Weber, Journal fUr Math. lxix. (1868),
p. Weber gave a different proof of it, as also did Hankel, Math. Ann. vin. (1875),
228;
pp. 469
470. The proof given here is due to Gegenbauer, Wiener Sitzungsberichte, lxxii.
(2), (1876), p.347. Other investigations are due to Sonine, Math. Ann. xvi. (1880), p. 40;
Sommerfeld, Konigsberg Dissertation, 1891 Macdonald, Proc. London Math. Soc. xxxv.
;
(1903), p. 438; and Cailler, Mem. de la Soc. Phys. de Geneve, xxxiv. (1902 1905), p. 331.
Some physical applications are due to Carslaw, Proc. London Math. Soc. (2), vm. (1910),
pp. 365374.
396 THEORY OF BESSEL FUNCTIONS [CHAP. XHI
13*32. Generalisations of Weber's second exponential integral.
When the Bessel functions in integrals of the type just considered are
not of the same order, it is usually impossible to express the result in any
simple form. The only method of dealing with the most general integral
is to substitute the series of 11*6 for the product of Bessel functions and
integrate term-by-term, but it seems unnecessary to give the result here.
In the special case in which X v /x, Macdonald* has shewn that the integral
is equal to
A. + 1* + iA
J o
J.(at)JAat)e^(-pH2 )t^dt^
2 ^ p^+vr(fi+1)r(v+1)
+ v+l + 2 \ + n+v
+
ffi fi v a*\
2 P+1 + 1 >P + P + 1 ~p)>
,
8 3 2 '
'
2 ' >
V >
\
by using the expansion of 541. Some special cases of this formula have
been investigated by Gegenbauerf.
r JMJ
+
v (t)
m _~2^'-(2ji~1)(2i--1)
.
x
f
J
r*- ri' mn{tmn0)an(ttdn
J J o t"
j) coa ^ _
Q cos2v 2 ^ gin $ sin ^ d0d4>dt
In view of the fact that tr 2 sin (t sin 6) sin (t sin <) does not exceed numeri-
cally the smaller of 1/tf- and sin sin <j>, the repeated integral converges
absolutely, and the order of the integrations may be changed.
Since *
00
j* sin (t sin 0) sin (t sin <) ,
_ \\ir sin 0, (0 ^ <f>)
J t*
~ I^tt sin <j>, {0 > <f>)
fin ri>
\tr I cos8*"8 cos8
""8
<f>
sin8 sin <f>d0d<f>
Jo Jo
-8 -8
+ $ir\ I cos8* cos8" <-6 sin sin8 <j>d<j>d0.
J J
(2v - cos8
"-8 <& sin cos*4
"8 sin8 0rf0[ cty
1) j <
j J
= - cos2"" 1
<f>
cos8*" 8 sin" d0 * + cos8*-1 </> . cos8*"8 <f>
sin8 <-W<
J J
r(j* + y -i)r(p
2IV+i> + *) *
whence the result stated is evident. The extension over the range of values
of fi and v for which merely R (/* + v) > is obtained by the theory of analytic
continuation.
It may be shewn in a similar manner that, when R (ji+v) is positive, then also
(2 )
r HM (QH,(Q Jf r(M +*)r(j)
Jo r+> 2* + " ro*+v+)r (/*+) r(*+$)'
This result was also obtained by Strove (ibid. p. 104) in the case /* = 1.
i
jo **+"
(2M -l)(2v-l)
/" /"** f|* {l-oos (< sin 0)} {1 - cos (< sin <f>)}
<*
"2'*+"-*-r( M +i)r(i/+i) Jo Jo Jo
x cos2* -2 cos2
"" 2
< sin sin <f>d$d<f>dt.
Now, if a and j8 are positive, it appears from a consideration of
round a contour consisting of the real axis and a large semicircle above it, that
"
r (1 - cos (at)} {1 - cos ($t)}
d _
C am (at) sin (pt)
t%
d{
} P Jo
Hence the triple integral under consideration is equal to the triple integral evaluated in
proving (1), and consequently (2) is established in the same way as (1).
The reader will prove in like manner that, if R (p) and R (v) both exceed \ then ,
( 3) /;^^= 2*+" 0* + v - 1) r 0* + ) r (v + )
and this may be extended over the range of values of p and v for which R (v) > \ and
R(n + v)>l.
The integrals
^ +|>+2
<*'
] ^++ i *
J^
may be evaluated in a similar manner, but the results are of no great interest*.
The integral
Jo t
The integral was evaluated, for all values of X, p. and v for which it is convergent, by
Soninet, Math. Ann. xvi. (1880), pp. 51 52; but he did not examine the integral in very
great detail, nor did he lay any stress on the discontinuities which occur when a and 6
become equal. Some years later the integral was investigated very thoroughly by Schaf-
heitlin but his preliminary analysis rests to a somewhat undue extent on the theory of
J,
linear differential equations.
The special case in which X=0 was discussed in 1895 by Gubler who used a very
elegant transformation of contour integrals unfortunately, however, it seems impossible
;
to adapt Gubler's analysis to the more general case in which X 0. The analysis in the #
special case will be given subsequently ( 13'44).
* Some related integrals have been evaluated by Siemon, Programm, Luisemchule, Berlin,
The first investigation which we shall give is based on the results of 13*2.
ry=V+l, [V =7 1.
It will be supposed that these relations hold down to the end of 13*41.
It is known that
Jo
r-J>(t)M>>t)
P
dt= lim
C+0J0
^W.(H^ t
since the integral on the left is convergent ; now, when c has any assigned
positive value, the integral on the right is convergent for complex values of b ;
_M J-fi(at)Jy-i(zt) dt
fy-^-'^zr^'
now
= e PV m\T(y + m)
Jo U=o J
(-)W
(~)w (j Y+2W
(^) '
T ^p+sm-1
. 2I e
-ot J (crf ) dt>
So V(y + m) Jo
w!Ti
provided thatf
P
m =o wi!r( 7 + m) Jo '
is absolutely convergent; and it is easy to shew that this is the case when
z\<c.
Hence, when z | |
< c,
+ m, - - m -+1
x ,FX (0 $ ; a :
^p)
* It follows from the asymptotic expansions of the Bessel functions that the conditions
R(p + v + l)>R(\)>-l
are sufficient to secure convergence when a = b, provided that /* -v is an odd integer.
r(a-^+l)r(-j) c
m++ il!_ Po m s. _\
+ r(i-y3-m)r(a + m)V(a8 + c)
, ^ (ra + m
j,
i ''*'<*
+ *)'
, .
sFiia + m^-p-m; J; x)
2 FX (a +m+h 1 fi m > $ * x)
Hence the terms in the infinite series which has been obtained do not
exceed in absolute value the terms of the series
|r(|-i9-)r(a + m)|J
where = (^/(a + c
a?
9
But this last series is absolutely convergent when
2
).
|
z |
< V(a* + c ) c, and it represents an analytic function of z in this domain.
2
/,
y ^(m-i(*o A
(-)* (j*)** 8- 1 (ja)*-*
v
T (2a + 2to)
_ |
"
m ,o m!r(7 + m) (a + c )*+T(a-/3 + l)
2 a
6<v (a + C )-(7,
/ a s
Then in the last integral formula we may take z = b, and when this has been
done, if we use fonctions majorantes just as before, we find that the resulting
series has its terms less than the terms of an absolutely convergent series
* (_-y(ft)y+8-i (| q)"-^r(2H-2m) [~ T(j )(l - JX)-* '**
m -om\T(y + m) " (a* + C2 a+m ) l\t(l - fi -m)T(a + m + iJ\
ir (-|)i(i-yx)-^-^-n
+,
c--oJo &
and therefore
Ja -p(at)Jy -i(bt) d
Jo
"
(-)m iy+zm-i p (2a + 2m) T (\)
i=o m! r (7 + m) 2-^ +2m- 1 a+*+* r (1 -
- w) f (a + + )
(1)
io
^^ rf
*-2r ^a^r^ra-^)- 2 * 1
^' 7 '^'
that is to say
provided that < b < a, and that the integral is convergent. This is the result
obtained by Sonine and Schafheitlin.
Ifwe interchange a and b, and also fi and v, throughout the work, we find
Now it so happens that the expressions on the right in (1) and (3) are not
the analytic continuations of the same function. There is consequently a
discontinuity in the formula when a = 6; and it will be necessary to examine
this phenomenon in some detail.
fiW*!^*. C+0J0
Jo
iim {"^iiMLMjt,
t
K t
K
Now consider
e ***
/. P=**
where z is a complex variable with R (z) positive.
m-OJ m! r (a - /3 + m + 1) - +7 + )
F (7+ m) T (a
m
(_) (| a
)a-^+y+2m-i p (2a + 2m) r (a - + 7 + 2m)
_
~~ l
m =o *""+"" mir(a-fi + m + l)T(y + m)V(a-/3 + y + m)'
Now the integral on the left is an analytic function of z when R (z) > 0,
and so its value, when z has the small positive value c, is the analytic con-
tinuation of the series on the right.
But, by Barnes' theory*, the series on the right and its analytic continua-
J_ (
xi
(ia)--*-*-*
41 "- 1
T (2a + 2s) T (a - /3 + 7 + 2) p ^
27rJ_ aci ^ + r(a-/tf + s + l)r(7 + s)r(a-/3 + 7 + *)
and this integral represents a function of z which is analytic when arg z <
| \
tr.
It is supposed that the contour consists of the imaginary axis with loops to
ensure that the poles of T (- s) lie on the right of the contour, while the poles
of T (2a + 2s) and of T (a - /9 + 7 + 2s) lie on the left of the contour.
When \z \
evaluate the integral by modifying the contour so
< 2a we may
as to enclose the poles on the left of the contour and evaluating the residues
* Proc. London Math. Soc. (2) v. (1907), pp. 59118. See also 6'5, 6-51 supra.
13*41] INFINITE INTEGRALS 403
at them. The sum of these residues forms two convergent series proceeding
in ascending powers of z ; hence, when R (z) > and \z\< 2a,
/"
Ja- fi
(at)Jy^{at) ,
Jo
_1 (_)m(^ a
)
^ r (7 - a - <8 - m) T (a + \m)
Y -a-g-m-i
~2 wl rn\ T(l --m) F( 7 -a - iw) T( 7 - - *m)
-1 .gY--*+
1 * (-) (^a)- w r (a + ft - 7 - t) T (^a + 7 ~ iff + 4w)
Now 22 (7 a ft) > 0, and so, when we make 2 assume the positive value
c and then make c -* 0, we deduce that
(2) f
J ^^dt
qq)*- 1 r (\) r (\n + \v - $\ + ^)
place of ft and i> in the preceding analysis ; if R (X) > 0, we then find that
-1
(|o ) to+M r (2a + 2s) T (2a + 2s - X)
__]_[-*
*- A r(o p( ,
^
"2wt.L. i ^ ,+
+ p + * + i)r(o-/) + )r(2o + )
T (X - m) T (a - |X + \m) "1 w
_1 (-)** (la^-"
1
*
" 2 mt w!r (;>- w + X+1)T (- p - \m + \\)T (a + \\ - $m)
and hence
~
f Ja+P (at) J^-x (at) aK T (X) T (a - j\)
1
w Jo *
A
}
_
2^r(J5 + i\+i)r(a+^\)r(i\- i>)'
unless \ = 0. This should be compared with the more general formulae
obtained from 13*4, namely that, when b < a,
a*-^r(a-p)r(p
2* a*-r-K r(a-p) U+l)
T (p + 1\ + 1)
* Fl \"- hX ' P ~ hX] '
a ~P
'
a>)
'
a+JT(a-iX)
yp-(/ -^ j.
+1n - iX;
,
j.
tt+y + , , as \
4 + i)r(ix- P
,
= 2>6^-^r (a )
' tt
i' 1;
p)-
Since X ^ 0, the functions on the right in (4) and (5) do not tend to limits
when a * b.
1 [** (| a
).+-i jr (2a + 2s)p r (- s)
U
according as 6 < a, 6 = a, 6 > a. Since
it is evident that the value of the integral when b =a is the mean of its limits
when b a and b = a + 0.
The result of taking Xl in (2) is
( 7) fy-coM^- ^j:-/' .
which is also easily obtained by inserting limits in 5*11 (13); this formula has been
discussed in great detail by Kapteyn, Proc. Section of Sei., K. Akad. van Wet. te
Amsterdam, iv. (1902), pp. 102103; Archives Nderlandaises, (2) vi. (1901), pp. 103
116.
13-42] INFINITE INTEGRALS 405
the value for b ^ a, the second for b > a and the values are correct for all ;
(1)
r
J^at)j^bt)
ai
a (b/ay/fj,,
Jo t (Ha/W/*- [R(ri>0]
sin{/*arcsin(6/a)},
Jp. (at) sin bt
dt = a* sin \iltt
(2)
F
Jo
[fi {b + V(6 2 - a )}" 2
[R(fj,)>-1]
(3) dt a cos
11
IfjLir [R(ri>0]
Jo I
[fi{b+^(b*~a2)}'
sin \fjj
arc sin (b/a)}
,, v( 2 b-)
v
V(6
(6,
r Jo (at) sin
/ fa . dt =
I
-j ,
/V(6
2
-a 2
).
fiM
2
-n
(?) I 7 (a) cos bt.dt = -: ,
Jo
la
These two formulae, which were given by Webert, Journal fiir Math. lxxv. (1873),
p. 77, are known as Weber's discontinuous factors ; they are associated with the problem of
determining the potential of an electrified circular disc J.
* Numerous other special cases are given by Nielsen, Ann. di Mat. (3) xiv. (1908), pp. 82 90.
The and (5) diverge for certain values of n when a = b.
integrals in (4)
t The former was known to Stokes many years earlier, and was, in fact, set by him as a
Smith's prize examination question in Feb. 1853. [Math, and Phys. Papers, v. (1905), p. 319.]
t Cf. Gallop, Quarterly Journal, xxi. (1886), pp. 230231.
406 THEORY OF BESSEL FUNCTIONS [CHAP. XHI
Another special formula is
The result of putting /* = in (8) is known as Dirichlet's discontinuous factor; see the
article by Voss, Encyclopadie der Math. Wisa. n. (1), (1916), p. 109.
Some other special formulae have been found useful in the theory of Fourier series by
W. H. Young, Leipziger Berichte, lxiii. (1911), pp. 369 387.
Another method of evaluating (5) has been given by Hopf and Sommerfeld, Archiv der
Math, und Phys. (3) xvin. (1911), pp. 116.
A consequence of formula (1) must be noted, When v > 0, we have, by 5*51 (5),
.'o *
= 1,
and so
(10) I .W |<1, |^ + 1 (*)j<l/V2,
provided only that v be positive; this an interesting generalisation of Hansen's inequality
is
-
( 2 5) which was discovered by Lommel, Miinchener Abh. xv. (1886), pp. 548 549.
The reader may find it interesting to deduce Bateman's integral *
In the special case in which the Bessel functions are of the same order,
Gegenbauerf found that by his method Weber's integral could be evaluated
in a simple manner.
If R (2v + 1) > R (\) > R (v + \) we have
f" dt
\^Jv (at)Jv (bt)fK
(\oby / Jv V(q + 6 - 2ab cos
2
Sm .,.,,
[t
"T f"
+ 1) r (J) J o Jo t (a* + fc*-2a& cos <*>)*"
(v
<fr)| .
Wdt
"
(ab)" T (v - \\ + ) ["
f* sin " <f>d<f> 2"
2* T (v + \) F (I) r (\ + ) J (a
Jo(a*
o
2
+6 --22a& cos 0)"-*A +i 2
* Messenger, xli. (1912), p. 101 ; for a proof of the formula by another method, see Hardy,
Messenger, xlii. (1913), pp. 9293.
t Wiener Sitzungsberichte, lxzxviii. (2), (1884), p. 991.
13-43] INFINITE INTEGRALS 407
b"r (v AX. + 1)
/b\ n C w
2 v-^r( + i)rft)raui) .?. U) J.
tv- (*-*
Now from the recurrence formulae
J* {(i
-*)-*(V(,)} = (n + 2/x- 1)(1 -^)-*"" 1 C^ 1 (5),
_ n+2fi-2v-l1 C
~ 2, + l /'/
+ ^TT
2v+n J_
1
1
-^ s| 1
(1 -'^ G
M dz
(n + 2p-2v-l)(n + 2n-l) r1
ii-*y-kC*n-i(s)dz,
2v +n+ l >J
so that
^
f
Jo
*M
(2v + 41 - A.
which is valid whether a > b or a < b. This result was given by Gegenbauer,
'
2v + i3 - \
;
v+1
-
>@TFf)-
4a 262 \
I J^(at)Jv (bt)dt
Jo
will now be investigated by the method due to Gubler*. It is convenient
first to consider the more general integral
[ J.(at)Jv (bt)
K
dt
.'o t
Fig. 29.
line R(z), = only at z = i; and then, for all the values of z and t under
consideration,
R\hbt(z-l/z)}^0;
and the repeated integral converges absolutely, since
K
Jo I
t J-
is convergent. The order of the integrations may therefore be changed, and
we have
+
r J(at)JAb t)^_
Jo ?
dt
1
~2frij. a>
f< >
]
f" Jy. (at)
*
exp
H- 1
;)}
z-*-1 dtdz.
If we write
b(z-l/s) a(e-l/Q,
* Math. Ann. xlviii. (1897), pp. 3748.
)
and suppose that that value of is taken for which | j ^ 1, we have, by 132,
r- J(at)jv {bt ) _ j_
/>+>
x
Qay r{p-\ + i)
dt ~ 2*i j _.
J
" ^ [\a (i + 1/0}*-"-' V(ji + l)
2 rv + i) .'-.({T+i/cr^
4,1
Z2
_ br + a t (6t + a)
~r(aT+6)' 6 ~ ttr +6 '
and the r contour is that shewn in Fig. 30; it starts from - b/a, encircles the
origin clockwise, and returns to b/a) where the contour crosses the positive
half of the real axis, we have arg t = 0.
_. dz 1 adr
Since
z(l + ?) 2r(bT + a)'
we find (on reversing the direction of the contour) that
(0+
[ ^ J,, (at) Jv (bt) dt = ^-.
*""* (
'
t**"-*-1 (6t'
+ a)-it"+*+l > (ar + &)*<"*- dr
J o Jb/a
=a
,
.
,1 W *(-H-i) (1 +-,2 , (1 dfO.
Ima** ] _x \ a J
* Journal fur Math. xv. (1836), p. 78, formula (57). See also Barnes, Quarterly Journal, xxxix.
(1908), pp. 115119.
t If X^fcO, the hypergeometric function does not in general reduce to an elementary function,
and the analysis becomes intractable.
w. n. f. 14
410 THEORY OF BESSEL FUNCTIONS [chap, xm
If we expand in ascending powers of b*/a? and substitute the values of the
Euler-Pochhammer integrals, then Gubler's result
I
J^ (at) Jv (bt) dt
yrUM + iv + i) F Lfi+ v +1 v n+l
is manifest.
Fig. 30.
The integral
*'<&*<& *.
which converges if R (a) > I (b) and R (v + 1 X) > R (fi) is expressible in
\ j |
\
,
MV '
Jo t
K
+ n+l)J
-ow!l (v
b"Y(\v-\X + \it + \)T(\v-\\-\. + .
\)
M+l nv A+l r> + i>
V \4-/4+l V \ fJb+1
^^ V+l '
a*)
and, in particular,
(2)
/; K^MH^U^^L^lR,
provided that R (v + 1) > R (/*) | |
and R (a) > \
I (b) |
Formula (1) was given by Heaviside* when /j, =v= and \ is and 1.
* Electromagnetic Theory, in. (London, 1912), pp. 249, 268, 275.
13-46, 13*46] INFINITE INTEGRALS 411
replace 6 by VX& 2 + c2 26c cos <), where b and c are positive, multiply by
sin 2 " <f>/(b2 + c2 - 26c cos <f>) iv and integrate. It is thus found that
f
l
-
o
where
,/M (a)
-Er
^
J, (bt)Jv (ct)
io
is
****
absolutely
convergent if
It follows that
/
(i) JM (at) Jv (6<) J, (c) t '* dt 1
Jo
( *6
4W-TTrp7n f
V " h ~ c + 26c cos ^"
*
'"
* , 1
sinS ^^
in which the value of A is
62 + c -a2 2
(2)
/J
J (at) J, (M) * (ct) f = ^, r
Multiply by a" +1 and differentiate under the integral sign with respect to
% )rW J." 8in" **
o
J v (at) Jv (bt) J v (ct) = dt
(a6c) ,
2
r(l/ + i)r(i)
>
when a, 6, c are the sides of a triangle of area A; but if a, 6, c are not sides
This formula is due to Sonine, Math. Ann. xvi. (1880), p. 46 other aspects of it have ;
been investigated by Dougall, Proc. Edinburgh Math. Soc. xxxvu. (1919), pp. 3347.
412 THEORY OF BESSEL FUNCTIONS [CHAP. XIII
It has been observed by Macdonald* that the integral on the left in (1) is
always expressible in terms of Legendre functions. The expression may be
derived from the integral on the right in the following manner:
sin $<f>
= sin \A sin 6
we have
= (26c)*-"- 1
(cos <f>
- cos Ay-"- 1
sin2" (j>d<j>
Jo
'r
==2 2*- 2 "- 1 (bcy-"-> sin 2*" 1 A . t (l-sin2 J/lsin2 ^)"-isin2"^cos2'*- 2 "- 1 ^^
Jo
T{V v)
= 4*-' (6c)* 1
sin*-* jj. ^i (* +"> i-; /* + *; sin2 M),
rfJ+%~
and therefore, if iJ (/a) and R (v) exceed , and a, b, c are the sides of a
triangle, we have
we have
2
- 6 2 - c2 + _1 sin2 " d>d<
[ (a 26c cos <f>)*-"
Jo
= (26ccosh^-"- 1 Ii^^i>
x.J^^^ + ^|^i;* + l;sech l,
2
^),
so that, when a > 2
(b + cf, we have
-1 -
/-\ C* t / *\ r /J^ r / .wix Tj (6c)* COS I>7T . sinh* *^
~i-M . . .
(o) J(at)J v (bt)Jv (ct)t -*dt=t Q ,(cosh<#).
/n\
(6)
f rr
Jo
r /i^ r
/
* (at) J. -x .v
(bt) J, <)
/ j-i
<
..
*j^
0* COS
(brf(hey^ An{
Z>7T . (X2 - -
l) ** *
;+9)w
^* + *
0,., (A )
, irs
R (a ib ic)
are positive ; this result is also due to Macdonald.
[Note. The apparent discrepancy between these formulae and the formulae of Mac-
donald's paper is a consequence of the different definitions adopted for the function Qn m ;
-
see 5 71.]
2Jp (xcos$)/xk
and integrating.
It is thus found that
cos*1 (f>
cos*1 <3> sin" < sin"<f> cosp 6
u== oL nir(/* + n + l)
X
r(^a+^ + |/j-^\ + w+l )
l\lfi+lv-ip + \\ + n + l)
.^,(
u+ v + pP X + n + 1, ^p X u v
p ^
^
~ n ; p + ,1 ;
cos2 6\
\
i
x-^-w, /a + + n + l;
i/ i/+l; sin 2 3>) I
,
Some special cases of this result have been given by Gegenbauer in a letter to Kapteyn,
Proc. Section of Sci., K. Acad, van Wet. te Amsterdam, iv. (1902), pp. 584 588.
Some extensions of formula (3) have been given recently by Nicholson*.
If cti, a2 , . . . am are positive numbers arranged in descending order of magnitude
it is easy to shew that, if
When a a ... am are such that they can be the lengths of the sides of
lt 2 ,
4
16 A 8 = IT (a1 + aa + a + ai -2a n s ),
The integral can be evaluated in a simple form only* when but i> = 0;
to deduce its value, it is simplest first to obtain an expression for the integral
when R(v) >\, and deduce the value for v = by analytic continuation; the
value of the integral assumes different forms according asf
i + a4 > fla+Gts.
i.e. according as A ^
2
Oj as a 8 a 4 .
ra
J
a t inrr
JJ/'^^-r^T^W.
dt
Jb^ i.
Jv ((M) J (<M) J*
^**
> *
* ^
#
where the lower limit is given by r =a x a 4 and the upper limit by m Gtj +a 4
We write
sr- (g - a y __ (a! + a y - (a - a )
2
; 4 4 t 4
2
,
w - (a a) (a, + a ) (a 2 a
2
2
2
4
2
3 )
2
so that the upper limit for a; is 1 or A/^(a 1 a.i a3 a i ): this expression will be
called 1/k.
when the integrals diverge at the upper limit if v 0), and we get
f" nJ (a n t)tdt
n=l
V{(1 - # ) (1 - k2 a?)\
2
Hence
/_1_ ^. / V(ai2 0sO\
fo 4 7T
2
A V A /'
(9) nJ (a n t) tdt =
.'0 n=l
v #
l7rV(G-iaa4) W(<
where If denotes the complete elliptic integral of the first kind, and that one
whose modulus is less than unity is to be taken.
* For other values of v it is expressible as a hypergeometric function of three variables,
f We still suppose that a 1 '^a^a 3 ^a 4 .
;
when R (v) > and a > 0. The simplest procedure is to regard the integral as
a special case of the last, so that it is equal to
<6 do
sin 2 "<M4>
+ C_^x>,-*
1 /"\o
/
,-,
{2a9* S * )] "'
^Wi^Wm-h
,
(1
f2a'(l -cos.fr) )*'
d hence*
""sr
a2 r ^>
no) r \j (atv* *- - < 2f/ >
f0, (a<b)
= |!^>p-v_iW(a! _ ls)) (o>J)
To secure convergence, a and b are taken to be positive and R(v) >R (/*)> 1
ifa = 6, then we take R (v) > R (fi + 1) > 0. The number z is an unrestricted
complex number, and the integral reduces to a case of the Weber-Schafheitlin
integral when z is zero.
The integrals involved being absolutely convergent ||, we see from 6'2 (8)
that, if c> 0, then
MV J
Jo (<
2
+ *-)*"
- ssJ.
1
b-
Too /YM-ooi
L,
c+coi
f
J- (bt) (' + '
""""' exp
r(a?b*)u
f"
L* l"
/
az*~] ,
- t* + z-\~\ dudt
)J
* An arithmetical error in Nicholson's work has been corrected. The result for values of R (v)
between and is obtained by analytic continuation.
t Math. Ann. xvi. (1880), p. 38 et teq.
X Wiener Sitzungsberichte, lxxxviii. (1884), pp. 990 1003.
This formula is also investigated by Cailler, Mem. de la Soc. de
phys. de Geneve, xxxiv.
When a < b the contour involved in the last integral maybe deformed into an
indefinitely great semicircle on the right of the imaginary axis, and the
integral along this is zero; but, when a > b, we have to apply 6*2 (8), and
then we obtain^ the formula stated*.
A related integral
From 6'22 (8) it follows that the integral on the left of (2) is equal to
Now make arg z-*- \ir. If we put z = iy, where y > 0, we find that
)rrre-i<--
j
V(tf + ) p-'
^_^ {y V(a +&2)} _ ,
^^ [y v(a2+ 6a)|]
the singularity t = y by
an indentation above the singular point, and that
interpretation is given to V(<2 y2 ) which makes the expression positive when
t>y.
Ifwe had put z - iy, we should have had the indentation below the real
axis and the sign of i would have been changed throughout (3).
In particular
(4) Jo<pt)
Jo (?=59 V(a8 +6)S
where the upper or lower sign is taken according as the indentation passes
above or below the axis of y.
* For physical applications of this integral, see Lamb, Proc. London Math. Soc. (2) vn. (1909),
pp. 122141.
t With certain limitations, a and b may be complex.
13-47] INFINITE INTEGRALS 417
The last formula (with the lower sign*) has been used in physical investigations by
Somnierfeld, Ann. der Physik ttnd Chemie, (4) xxviii. (1909), pp. 682683; see also
Bateman, Electrical and Optical Wave-Motion (Cambridge, 1915), p. 72.
provided that a > and R (| v ) > R (p) > 1 ; this might have been
established independently by the same method.
Similarly, from (2) we have
m
{)
r ^MV(< 3
+ ^) }
^ +1 ** _ r(^+i) f ^-^^sinfl)^ .
The integral on the right is easily expansible in powers of z; but the only
case of interest is when 2p = 2/a + 3, and we then have
so that
(9) j;3g(*--r**-J&$*.<w;
and these are valid if R (v) > \. The last formula was established in a
different manner (when i>=l) by Struvef and from it we deduce the important ;
theorem that J, when v > $ and x>0, H(#) is positive. Struve's integral is
of considerable value in the Theory of Diffraction.
f
Jo
^m+i (bt) ^ + ^ t* dt
the upper limit in the last integral being b or a, whichever is the smaller.
* My thanks are due to Professor Love for pointing out to me the desirability of emphasizing
the ambiguity of sign.
t Ann. der Physik und Chemie, (3) xvn. (1882), pp. 10101011.
t Cf. 10-45.
418 THEORY OF BESSEL FUNCTIONS [CHAP. XIH
If b < a, the integral on the right seems intractable, but, when b > a, we
put u a sin and deduce that
positive,
_ a"- b* /"*
1
fl""* /y _ M -i (re) cfo
~ ^-"-'Jo (w2 +6 2 )''
V(v) jo
by
-
13 3 (4), and thence we see that
(11)
Jo
J, W + ^ )iy+1 *
(ff
M+1 =
2> -! r (y) *m ( *)>
6
provided that a <b and i (v + 2) > R (/a) > - 1 the restriction ; that z is
(12) J M () ^)iv
t at- ^ gv
^+
,
Jo (
where i2 (v + 2) > .R (/*) > and 6 > a, has been generalised in two ways by
Gegenbauer*, by the usual methods of substituting Neumann's integral and
Gegenbauer's integral (cf. 13-1) for the second Bessel function.
(14)
Jo
P J*W (?Tfy
t at
_2*- T(ji)
1
^i: r.'^g& ~****
J v (az) J v (cz)
t
(15) /o J.M
v+ ,r. ~* M = -5^-' 7 *
L J
(16)
this has
iy,,.{bt)u[JAt)-\r-
Sci.,
?[f^Tj]
K. Akad. van Wet.
;
te Amster-
dam, xi. (1909), pp. 749755.
" A man starts from a point and walks a distance a in a straight line ;
he then turns through any angle whatever and walks a distance a in a second
straight line. He repeats this process n times.
RayleighJ gave the full details of the analysis of the problem (which had been
examined somewhat briefly by Kluyver), and then obtained the solution of the
corresponding problem for flights in three dimensions.
If sm is the resultant of a1} a 2 , . . . , am (m = 1, 2, . . . , n - 1), and if 6m is the
* Nature, lxxh. (1905), pp. 294, 342 (see also p. 318); Drapers' Company Research Memoirs,
angle between sm and a m+1 then, in the two-dimensional problem, , all values of
the angle m between ir and it are equally probable.
Now let Pn {f", a lt an ) denote the probability that after n stretches
2, ...,
the distance from the starting point shall be less than r, so that the probability
dr
It is then evident that
Now ( 1342)
rf^(r)j;(.)*=n' <'>
and so, i/ i^is discontinuous factor is inserted in the (n iy tuple integral, the
range of values of n_x may be taken to be ( ir, tr).
We change the order of the integrations with respect to n_1 and t, and,
remembering that
s2,, = s2n_i + a\ - 2*n _i an cos n_lt
we get
r\ I J1 (rt)Jo (s n t)dtd0n . 1 = 2irrf J, (rt) J (sn -x f) (an t) dt
J vJ Jo
by 11-41 (16). We next make the substitution
8
S _i = s n _2 + a2_j - 2s_2 an_j cos
2
n_2 ,
If P(r ; ay, 02, a \p) denotes the probability that the final distance
... ,
is
. 1
I" VVP
rtteL-pf '('...
~ 2) r ($)) .'c.'o
H
JoJnt-l
n\in*->em .dede^...d0,de l,
where the integration with respect to 6n- extends over the values of x n-i
which make sn < r.
The discontinuous factor which we now introduce is
(
W)b~ >
we infer that
P.(r; ,a if 1
...,a m |p)-r{r<fc^
When the displacements a^a^ a are all equal to a, and w is large, we may
approximate to the value of the integral by Laplace's* process. The important
part of the integrand is the part for which t is small, and, for such values of t,
* Hat)*'-1 ( aH\
so that (13-3)
is convergent if a and 6 are positive and i2 (ji + v) > 1 when ; a = b the last
* La thiorie analytique de$ ProbabiliUs (Paris, 1812), chapter in. The process may be
recognised as a somewhat disguised form of the method of steepest descents.
422 THEORY OF BESSEL FUNCTIONS [CHAP. XIII
f- J{a(Z+t)}Jp (bt)
i_co (Z+ty v
= (46)'
r ( ,+W(4j ,L
f
L
z^ty
f'J^[a(Z + t)\
(
co ^ btc08 ^ sm'^ d^ dt
= (kby f
00
r
n
J (at)
~*^ cos "^
'
{6 ( * cos * sin2v *****
r( y +i)r(f)J J o
]
= r(y + 4)ra) J I
~V cos ^ ht cos *^ cos ^ cos ^ sin2v
* d * dt
2 (4&y
- (2>ro.7i)i>+i) J.
r*
(a* - * cos' **"' cos <** ^ ***
by a special case of 13'4 (2).
(1) sma(z + t)
Jo {H) dt = wj 9 (bz)> (6 ^ a)
a
/ON f" sina(^ + ) r ,,,.,. f cosw^.dM .,
J (bt) dt
/ z+t
round the contour formed by the real axis and an indefinitely great semicircle above it ; it
has to be supposed that there is an indentation at z when z is real.
The integral
Z +t z+t'
13-5] INFINITE INTEGRALS 423
f- sin . ( + 1) _2 f sino(*-M)
rf<
.'
-oo ? + .'O Z +t
r oo ~ a /" oo
- oo
cos w (* + ^o (&0 ftdw
JO ' .'
ra roo
2s cos m (z + Jo (&0 dtdu
'o .'o
/oo ra />
(o)
J_
!
^ '-J (bt)dt = 0.
The appropriate contours are of two types. We take the first type when
(z) has no
<j>
upper half-plane; the contour is
singularities except poles in the
taken to be a large semicircle above the real axis with its centre at the origin,
together with that part of the real axis (indented at the origin) which joins
the ends of the semicircle.
from and ending at the indentation, one loop passing round each branch point,
so that the integrand has no singularity inside the contour.
Bessel function one of the integrals discussed in 65, and change the order of
*
taken round the first type of contour described in 13*5. In this integral, a
is positive, m is a positive integer (zero included), ris a complex number with
\R(v)\<R(p)<2m + ^.
The first inequality secures the convergence of the integral when the radius
of the indentation tends to zero; and (as a consequence of Jordan's lemma)
the second inequality ensures that the integral round the large semicircle
tends to zero as the radius tends to infinity.
The only singularity of the integrand inside the contour is the point r. It
follows that
1 [ of-
1
{Hv u (ax) - e* JT,w ^HV(az)dz
X_
(axe**)} ,
J_ f<"+>
(1) ["[(1
J o
+ e 0>-*)*) Jv (cue) + % (1 - e (p-")) Y. (ax)] -^
(tfr
xf>~*dx
(2)
J"
[cos \(p-v)tt.Jv (ax) + sin (p - v) tt Yv (ax)]
.
{ ^
(
-n(ra)
\M-i / J \tn
^ '.(*
* Hankel's work was published posthumously, Math. Ann. vm. (1875), pp. 458 461. A partial
investigation of the integral with v=n, p=2n + 2, m=2n was given bj Neumann, Theorie der
Bessel' schenFunctionen (Leipzig, 1867), p. 58.
t The evaluation of integrals of this character which contain only one of the two Bessel
functions is effected in 13*6.
A ! : , ^ .
The reader should notice the following special cases of this formula
(3)
J^
{cos vn.Jv (ax) - sin vtr . J (or)} ^-^^ = l-^i-jZ
#"v + / m
x _ a it"- '
f JAflx)dx
l
(a.r) <r A' _ m (ak)
(4)
~(^+i^ + i~~
J 2 .
The former is valid when 2m - f < 72 < and the latter when - 1 < R (v < 2m + 1
(i/) I , )
For an extension of (4) to the case when m is not an integer, see $ 13*6 (2).
The special formula
...
(5)
/""
Jo
xJ (ax)dx
~^^f
(a*),
.
has been pointed out by Mehler, Math. Ann. xvm. (1881), p. 194, and Basset, Hydro-
dynamics, n. (Cambridge, 1889), p. 19; while Nicholson, Quarterly Journal, xlii. (1911),
p. 220, has obtained another special formula
} o ( ax ) dx K Q (ak)
(6)
x* + i:*
'
/: k
by a complicated transformation of repeated integrals.
f*
J (ax) dx _ 2 f" fl" cos (ax cos 0) , - ,
j x* + k* ~Wo Jo aF+T*
1 [i*
fCJo
by 10*4 (11), provided that a and R (k) are both positive; so that
,. r Jo(ax)dx IT ,
T ,. T / ;xi
Iv (ak) - L (ak) =
T ( v + i)Y(l) ll"
~
e akC0S6 sin2 " W
and since, by a special form of (2),
J
.
_ (2fc) 6~
J oo * ~r A? ft*
ar' + fc
2
""tTJ-oo
= ^r
Jo
[(1 H- *-) ^ (a*) + (1 - *"*) H (a*)] ^ s
a -
f
xv K v (ax) dx
Jo x? + k*
'
rx K v
v {ax)dx _ Y(v + l) p f" (2a) cos xu.dudx
v
'~ 1
7r2A;,
{H_ (ak) - F_, (a&)},
4 COS V7T
Kv a
_
(10)'
v f"
Jo
}
x2 + kf>
2
xv
= v+1
4k cos vtt
{H (ak)' -
{
Yv (ak)}.
v Ji
These formulae (when v 0) are due to Nicholson, and the last has also
been given by Heaviside.
m,
The
,
integral
i
^f -^
J - v (ax) dx
T(v +^uln,
[*%-*>
\)Y($)kU
sin2 " Odd,
and so
formula are also incorrect. A similar error was made by Basset, Proc. Camb. Phil. Soc. vi. (1889),
p. 11. The correct result was given by Gubler, Zurich Vierteljahnschrift, xlvii. (1902),
pp. 422 424.
13-52] INFINITE INTEGRALS 427
psinha^ y + 1
J sinh irx
sinh tt2
(12)
J o
r*^^J (bx)a?*
sinh irx
v '
x
dx=-
it
2 (-)-+i
=i
'
sin na.Kv {nb).
An integral expressible as a similar series was investigated by Riemann, Ann. der Physik
und Chemie, (2) xcv. (1855), pp. 132135.
27TIJ
a>0, \R(v)\<R(p)<2R(H) + i.
The contour is chosen with a loop to exclude the point ik, as shewn in
Fig. 31, and then there are no poles inside the contour; and the integral round
the large semicircle tends to zero as the radius tends to infinity. Hence
1 ;"
+ k*)*+1
27riJo (z*
(*
2
+#r +1 * r^p-^r^ + i)-
As in 13-51, we take R(k)>0.
1
J^
[(l + 6 (P-r-U)H )t 7; (ax) +i (i _ ,o-r-*M) Yv (ax)]
( J+ ^+ 1
w-eKp-r-a^w W r
f,, x _,. .,_..
-2M-2
2 r (\p + %v 4- m) (fafr) .
2
V2 " /'
sinv7r. +
r(/*+ 1) L
'"
m=0 m\r(v + vi+l)r(^p + ^v- fi + m)
and therefore
x '~ dx
00
l 1
f
(!) [cos (/> - \v - fi) Tr.Jv {ax) 4- sin (\p -\v-fi) tt. F (ax)] /^ w-n
Jo +jfc2
TT^rP-aM-a
x
lf( v +i)r(is p+isv- H,y**{~ 2~>
1
2
"^ w + 1
TV
_ (ja*)*- r (j p - \v) fp-yp-y . . a'JrVI
Fig. 31.
K{bz)
*rij* (z*-r*T*
'
positive imaginary part, 9B^ denotes any cylinder function of order /i, and
|
R O) |
4- R (/jl)
I
< R (p) < 2m 4- 4.
Math. Ann. vm. (1875), pp. 461467.
C -
(1) ^ (M
[r
Hvil) (ax) " *" r ^ {hxelH) Ht,{l) (axeni)]
ds (f-lr^
1 id \ m
- strata?) [^"-o)ff. oh.
(2)
<L\l [^()^^
)
() + ^(^^K^](^^|-,
; it is
according as a $ 6.
The existence of the discontinuity in the expression for this integral was
pointed out by Hankel.
If we modify formula (3) we see that, if a >b > and R (k) > 0, then
f" x
(5)
J ^rp -^m (&*) [cos H/* - v) if A. (ax) + 8in$(fjk-v)Tr. Yv (ax)] dx
= I(bk)Kv (ak).
'
(6)
Jo
r ,,
^ [cos(p+
rcosiCo+ a
,_ /i z>) 7r. Jv (ax)
v) tt.J + sin i
J (p n v) it. Y
+ fi Fv (ax)] dx
= -I (bk)K (ak) !<?-*.
ll v
^-I^bfyl^cfyKriak)*?-*.
This process may be repeated as often as we please ; and we find that, if
a>b+1c, then
~
*P I to)
/+ KL " J ( c^) -> s i {p + + (^ - 1) *} t ^ (a*)
,t
( fi ) r
JO X M=i
* /*
+ sin \ [p + p. + (N- 1) v) ir Y .
v (ax)] dx
n=\
Again, by considering
SjJi^[n-W] *"<>*
round the contour previously used, where both b and p. differ in the different
If p + 2/i v
is an even integer, the integral on the left involves functions
of the first kind of this type was given by Gegenbauer, who overlooked the
necessity for this restriction (cf. 1351).
1
fy. ^IM^+ni HV(az)
"
= _J_ (
Ar r,^ j.i^<-r+! *, (or)
:
and, in particular,
=\ j" /*
-r^-p J> { V(p* + ** - 2pi cos *)) dfdp.
f (z* + bk y +ii
Pig. 32.
\R(v)\<R(p)<4>R(fi) + ^-.
* Quarterly Journal, xlii. (1911), p. '2 ,
24.
,
- w k V 2 )""*
(
-4
| (akW 2Y + m T(p + \ v + m ) p+y-4/n+2m "*
- f COS
Qg
2sini/7r.r(/t+l)Lm=oi!r(i/+m+l)r(i 3+ii/-/A+i?) /
4
If the series on the right are compared with those given in 5-41, it is seen
that the former is expressible as a product of Bessel functions if p 2 = i/= /* +
or if p 4 = v = p. + \, while the latter is so expressible Up 2 = v p. + \
or if p 4 = = /4 + .
i>
l r eUz+k)
2tti
|s| = 8, \z\ = R,
terminated by the lines arg ( z) = ir, and the lines which join the extremities
of these circular arcs*.
R(p)< m + $.
By Cauchy's theorem we have
2iriJ(^ + ^M+lV w! ^
* Cf. 3/odern Analysis, 6-2; or 7*4 supra.
1
(!) ^T^^-^-^M*)}
m! dkm
/
1 Too \(x+k\
i{x+k) ~a\1
= -L, 1
^_
of-
(# + &)
2ttiJ
e
*
[ep*HJ l
(#e'"')
v
'
- T**H m (xe- l "
)\ dx
we find that
cr--
1 r
-
p-iiz+h)
If we take p = 1, v = 0, we get
r/7x 2 fsin(a?+&) ,
//i\
(4) /(&) = - v '
, Jr ,
(x)dx,
.
T .
X -r K
o ...
Bywntmg ^
cos(x + k)
^ 1
- f
j
1
. .,
sm (. + *) A. ,
,vj.
and using the formulae (6) and (7) of 13*42, Sonine deduced from (5) that
(7) Y (*) = -- P ^?
irJox + k
efo +-
tt.'o
I
**
sin (fr cos 6?) dd,
* '
(8) Yn (k) = - - I 0 O + A) Jo (*) dx + -T I sin (k cos - \ mr) cos n0f/0.
TJO .'
* See alsoLercb, Monatshefte filr Math, und Phys. i. (1890), pp. 105112.
*
/'
af~ x J v (aw)
dx,
Jo
in which it is at first supposed that
When this is evaluated (by swinging round the contour so as to enclose the
poles on the right of the contour) we find that
,-v f
af- 1
J
v (ax) dx
^ }
!o (a? + krr +1
_ q"fr+-^r(ip + i y )r(/i+i-ip-iF) l2 (p + v.p + v +I ,'
o^-\
2*+i r(p + l)T(p+l) \ 2 2
'
' '*' '
4;
a^-pr(^ + p-M-l) / *-p. u+2-""^- a '*^
(2)
Jo (x*+kr +l >ro.+i)^" ( }'
a formula obtained by another method by Sonine, J/a*A. Jrm. xvi. (1880), p. 50; it is
valid when
-KR(v)<2R(lt) + $.
* This method is due to Lerch, Rozpravy, v. (1896), no. 23 [Jahrbuch iiber die Fortschritte der
Math. (1896), p. 233]; he shewed that
(5)
J (* + 4*rr ~ (2*r r ( + i) J ' w w * '
(6)
Jo (^ + 4*T + *
-
M2^=il^HT) J ^
(a *) ^^ ^> (
The former of these is valid when 72 (v) > , the latter when R (p) > , and,
in both, a > and arg k\<\rr. |
436 THEORY OF BESSEL FUNCTIONS [CHAP. XTII
f
x of'1 Jv (ax) dx
Jo (x + ky+i '
j / root
T(-s) of- QaxY
^ +2s
V2 ^^ /
1
^-T\ ~,
V
:
I ^./ 5+l) '
TT +1,/ <&<&
2tti .'<,./_.*
ITl.'o J _ooi T(i/ + (x + ky
_ y r (fi + m + 1) sin $ Q + y - - m) ir
(jafc>rH ~'>+w 1 fi
The first series reduces to J v (ak) when /* = 0, and the second series is then
expressible by Lommel's functions (cf. 107). In particular we have
/(_. f
00
x v Jv (ax)dx irk" r __ . ,. , ....
integral (5*43) and using the method just described, or else by replacing the
product J^ (x) Jv (x) by
in which the poles of V ( s) are on the right of the contour while those of
r (p, + v + 2,5 +1) are on the left this expression is easily derived from 5*41;
f p-i
(1) / x Jn {ax) Jv (b/x) dx
f da
due
lim ( p'a?) J
exp (-p*a?) (a) J
v (a?)
/ J x-
_ Vv (w)
{x) -^
&
f
/ \ f r / \ r / \ &X SIT! V7T ) d%
+ 4^"l"-
y)
{ (2) - * (' + 1) -* (2 - ') - 2 log 2p).
It is easily proved that
d#
f
jo
exp^^)^^)^^)-^^^-
/ i# . x f r / x t , \ sin i/tt)
_1 (2n + l)!(2p)-*
4 =i w.(n+ 1) ! T(v + n + 1) T(2 - */ + w)'
and this series is an integral function of 1/p.
small and |
argp \<i'ir, we observe that the last integrand has double poles
at and 1, and simple poles at f 2, - f 3, ... , , .
Hence we find
~2ttv{1-v) n 3
and so
+ 1) ~^ (2 " I,)}
+ 4^=l0 {1+2logi8 -^ (,,
]
^; ->^^
2
(X>
^^(2)-^(, + l)-t(2-.)-21og2p}^-
l)-yjr(2-v) v
(-)"(2j>)T(in)
+ 2 2H(n-2)r(i/+l-^)r( 2 -''-).(-2)!
t 3
|r(m + t)l'(2p)-*'
~ _i2^ + Mffo7r
, .
2
(2m + l) 2 (2m + 3).(2m + l)!'
"ski. L.i
sm *' exp i(
l
\
dtd *'
1 /"+'*
f, Z* + z*\ T (zZ\ dt
By using the results of 6*21 (4) and (5), we find in the same way that
The formula was obtained by Macdonald, Proc. London Math. Soc. xxxn. (1900),
(1)
A further modification of the integrals on the right in (2) and (3) was given by Sonine,
the object of the change being to remove the exponential functions.
For physical applications of these integrals, see Macdonald, Proc. London Math. Soc.
(2) xiv. (1915), pp. 410427.
K v, namely
(1)
ww-trM-*-*?]*'*
This formula is valid for all values of v when
\ArgZ\<7r, |
arg j
< 7T and \a,rg(Z + z)\ < \ir\
By 6-22(7) we have
*
K v (X) K v (x) =\
** \
J 00
f
J oo
e
-r+)-Xeodi *-*eoSh dtdu
~| | e-ivT-Xoo&MT+U)-xco*h(T-U) dlJdT
t J oo J -oo
e -Xco&(T+U)-xcoh(T-U) flJJ
/ -oe
r f If X 2
+ a? + 2Xocco8h2T\l
~
dv
it becomes I exp a\ v + *f
~^~>
and so we have
K. (X) K, (.) = I,
[ -~-M"* e exp
[
-
(|
-
^}] AT.
The formulae of this type have been developed by Nicholson*, and the
two which are most easily proved are
(1) KIIL
(z)Kv (z) = 2f K^tfz cosh i) cosh (p-v)tdt
Jo
,00
=2 K^ v (2z cosh t) cosh (/* + v) t dt,
.'o
K IL {z)K v (z) = -\
T* J 00 JJ 00
0-s (cosh H-coshM) cosn
^ cos h vudtdu.
The repeated integral is absolutely convergent, and it may be regarded as a
Kh (z) K v (z) = \\
"J
I
00 J 00
e-*>* r cosh u cos h ^(T+U) cosh v(T-U)dTdU.
The integrals corresponding to the last two of these four terms obviously
vanish; and, if we interchange the parametric variables Tand 6r in the integral
corresponding to the second of the four terms, we obtain the formula
The formula
(2) 7M (z) I v (z)=- j" /M+ (2s cos 9) cos (ti-v)0 d0,
T Jo
which is valid when R(fi+v) exceeds 1, is at once deducible from Neu-
mann's formula.
r ,
, .
, 2 T + |) fl* p (4s)" cos (u cos 0) cos i>0 .
a result of which a more general form has been given in 13*6, formula (3).
the contour, it is clear that when an exponential factor exp { \w-} is inserted,
the resulting integral converges uniformly with regard to \, and so it is a
continuous function of \. Hence f
* Phil. Mag. (6) xix. (1910), p. 234; Quarterly Journal, xlii. (1911), p. 221.
t Hardy, Quarterly Journal, xxxv. (1904), pp. 2266 ; Trans. Camb. Phil. Soc. xxi. (1912),
pp. 1 48. In this integral (as distinguished from those which follow) the sign lim is commuta-
tive with the integral sign.
w. B. F. 15
442 THEORY OF BESSEL FUNCTIONS [CHAP. XIII
we get
H (x) =
lim
W*
r-
-x
exp {- X (t + \-rri)') eixcoaht
- vt
dt
X-^+0 J
a\v-ai rx
- - vu
|
e l (t)ei * co* ht vt
dt x I ei {u)e- ixco *Ylu du
. X J CD
(in which e^t) and e2 (u) stand for the exponential factors) as a double
integral
"
/ ^ (t) % (u) e (cosh <-cohii)-v+) (dtdu).
.
I
-X J X
We thus find that
Tr
2
^' 1'
(a?) #< 2 (x)
> =G TX
J ./
[*
e <*(coBh*-corii )-(*+) (^^
with the implied exponential factor
[ exp {- 2\ (v + 7Ti') ]
2
e" 2* sinh Tcosh v dv
J -cc
= 2 exp (2\tt ) 2
J
exp (- 2\v*) cos 47r\v e~ 2xsinh rcosht .
'
<fo
= 2 exp (2XTT2 ) I
e-torinhroodidv
-'o
- ixsinh Tcoshv
- 2 exp (2\tt ) 2
r
Jo
{1 - exp (- 2\v 2 ) cos 4tt\v} e dv.
= 1 - exp (- 2\v ) + 2
2 exp (- 2\v2 ) sin2 27r\t> 2Xv2 + 8tt 2
X 2 v2 ,
f*
exp{-2X(i7+^7n")2 }e- 2 a!sinh:rsinh ^rf/'
J CO
On collecting the results and remembering that means the same thing
Jo
- (2\ + 8tt \ ) 2 2
exp (2\7ra ) j^^^sinh T)\
(o/x
1 e- 2 " r dr
J/*-oJ
A ++06-*+0J6 L Jm-oJ I
*/*
T
Jo
( \~K ll
(2xsinhT)\ e^ T dT
(cy* Jh-o
are convergent, the result of making \ -* is
2 2
1 7T {J, (x) + Y* (x)} = f
Jo
" K (2x sinh T) (e~ ivT + e
2" r
) rfT.
J, 2 (a?) + F 2
(a?) = t
I* K (2x sinh T) cosh IvTdT.
If we replace x by 2, both sides of this equation become analytic functions
of z, provided that E (z) > 0. Hence, by the theory of analytic continuation,
we have the result
(2) Jv (z )
d
I^- Y
OV
v (z)
d
-^4^=--
OV 7T
rK,(2zsm\it)e-^dL
.!
o
OV OV
- I yt,
2i
$H 2>
<
(x)
{x)
dH"{1)(x) -
dy
H (x) dH'*W
a v
dv i
* For the full details of the analysis, see Watson, Proc. Royal Soc. xciv. a, (1918), pp. 197
202.
((
where the constant implied in the symbol (\) is a function of T such that
its integral with respect to T from to x is convergent.
In like manner,
(""
(2 U + Trt) exp {- 2X U + $ irif] e" 2te sinh Tsinh * dU
J 00
/" fa) W dY ^ - F, W ^ -^ = - - fT
"
9l>
fa)
dv TTJo J -ao
e
-^sinhTcosh*-2,T rf ^r
= -- f "jfo^sinh T)e~ T dT.
7T.'o
The extension to the case in which the argument of the Bessel functions is
w ^(^Lf^,)^..
(8 ) v v ' cv Cv
-*
nz* J (o
'(zicoshZT-v^K^vBrnh^e-^dT.
Other formulae which may be established by the methods of this section are
(4) y.w/rW+FdWfrW
= - f K _n(2z8inht) .{eC+W + e-<*+")t cos(p-v) n}dt,
v
IT JO
* ain( *~ v) n
(5) J^z) Yv {z)-Jv {z) FM (z) = j~ *,- (2^inh t) -+>) dt;
Jf{x)+Y*{x)
is a decreasing function of x for any real fixed value of v, when x is positive.
It is clear that
^[x[J/(x) + Y*(x)}-]
8^
7T
2
K (2x sinh T) tanh T cosh 2vT
(2x
1S C*if n mJ m d
sinh T) [cosh 2vT - -^ {tanh Tcosh 2vT\
, . , ,
+ dT,
^ Jo
2 [
"
=~ f K (2x sinh T) tanh Tcosh 2vT {tanh T - 2v tanh 2iT} dT.
t' Jo
We twice integrate by parts the last portion of the second term in the
integral thus
2j>
2
J
sinh t K '
(2x sinh t) cosh 2vt dt
Jo
r d
= v \
-j- {sinh tK '
(2% sinh t)} sinh 2vtdt
Jo dt
+ x \
j- [sinh t cosh iT (2a; sinh )] cosh 2itf eft
Jo dt
= ( [x cosh 2lf (2x sinh ) + 2a? sinh t cosh 2
* J5T ' (2# sinh t)] cosh 2i/i eft;
Jo
the simplification after the second step is produced by using the differential
equation
zK " (z) + K: (z) - zK (z) = 0.
The integral under discussion consequently reduces to
|
[- 2x sinh tK (2x sinh
2
t) - 2x* sinh 3
1
'
(2x sinh*)] cosh 2vt dt
+x
r
I jfiT (2# sinh )
T
- 2 sinh 2
cosh 2i> +
rf
~=-
-j r-
(sinh't
cosh 2irt
)~1
eft
are 2/tt, it follows from the last two results that when x^.v^\,
An elementary proof of the last inequality (with various related inequalities) was
deduced by Schafheitlin, Berliner Sitzung*beric1de, v. (1906), p. 86, from the formula
(cf. 5-14)
(4r-l)JVr(0*
= i( a 2 + 62 )
_^Qf^ + ^; (A ) .
|
2
+ 2(l-^)^2 (a;) + ^; 2 (. t .-)],
The next consequence which we shall deduce from the integrals of 13'73
is that, when v is positive,
r , ^
dYv (v) v dJv {v)
To obtain this result, we observe that the expression on the left may be
written in the form
=1
irv
-^T K (2v sinh T)e-'" T dT
7T J o
-s[ 1
-/.Vjr-( 2 " inh s)*]-
But, for each positive value of t, 2v sinh (^tjv) is a decreasing function of v,
and so, since K (x) is a positive decreasing function of its argument, we see that
[MpY dv
1AP)
dv
> lim
= lim
by using the asymptotic expansions of 8'42 ; and this establishes the result
stated.
(2m)!
T "
cosh* m=0
where |
i2p |
<
cosiC(i>7r)| (2^)!
and, when v is real and p is so large that p + %>v, Mp lies between and
(2p)l
'
2
7r m=0 (2w)! Jo
j
arg z < 7r and, if v is real
|
; and z is positive, and p exceeds v \, the remainder
after p terms is of the same sign as, and numerically less than, the (p + l)th
term.
r. r (2cos^)^+- 2 eitY ,
y . >t , <7r ),
e ittd ( jf
= r 0* + v - x) > j
J ..ro. +B r(.- , s
.
( 0; ( <[>r)j
where t is any real number.
By expanding in ascending powers of x and y, and then applying this
formula, it is seen that
a) /1%^^-W
if 7r < < < 7r ; for other real values of t, the integral is zero.
In particular
In view of the researches of March, Ann. der Physik und Chemie, (4) xxxvn. (1912),
pp. 2950 and Rybczyriski, Ann. der Physik und Chemie, (4) xli. (1913), pp. 191208, it
seems quite likely that, in spite of the erroneous character of the analysis of these writers J,
these integrals evaluated by Ramanujan may prove to be of the highest importance in the
theory of the transmission of Electric Waves.
The difference between the subjects of this chapter and the last is more
than one of mere degree produced by the insertion of an additional integral sign.
In Chapter xiii we were concerned with the discussion of integrals of perfectly
definite functions of the variable and of a number of auxiliary parameters; in
the integrals which are now to be discussed the functions under the integral
Thus, in the first problem which
sign are to a greater or less extent arbitrary.
will be discussed, the integral involves a function which has merely to satisfy the
conditions of being a solution of a partial differential equation, and of having
continuous differential coefficients at all points of real three-dimensional space.
because many of the formulae of Chapter xiii may easily be derived from
them, and were, in fact, discovered by Weber as special cases of the results of
this section.
is
u = if"
I"
r (x + 2X^t,y+2Y</t, z + 2ZJt)
x exp {- (X 2 + F + Z )} dXdYdZ,
2 2
<
*
Journal fur Math. lxix. (1868), pp. 222 237.
La TMorie Analytique de la Chaleur (Paris, 1822), 372. The simpler equation with only
t
one term on the right had previously been solved by Laplace, Journal de VEcole poly technique,
vni. (1809>, pp. 235244.
_ ;
then
(2) u = exp (- kH) $ (x, y, z),
provided that <I> has continuous first and second differential coefficients, and
the integral converges in such a way* that transformations to polar coordinates
are permissible.
=-
1 r ao r v fir
operator
1 d ( d\ ,.
r2 dr\ drj
to vf (r) is
We proceed shew that the last integral is zero. If we make use of the
to
differential equation (1), which <S> satisfies, we find that
To avoid the difficulty! caused by the apparent singularity of the last inte-
grand on the polar axis, we consider the integral taken over the surface of a
pole
sphere with the exception of a small cap of angular radius 8 at each
since the integrand on the bounded at the poles, the integrals over the
left is
* A sufficient condition is that * should be bounded when the variables assume all real values,
infinite values of the variables being inoluded. Cf the corresponding two-dimensional investigation,
.
to <f>,
we see that its integral vanishes because 8<J>/9<^> is supposed to be a one-
-?/".[sin ^]r^'
and this can be made arbitrarily small by taking 8 sufficiently small since
d<P/(sin0d0)
is zero.
Consequently
(8)
l&a + *T.(r)-0,
, , ,
.
x A
=
sin kr + B cos kr
so that ts (r) ,
where A and B are constants since rst (r) and its derivate are continuous for
;
all values of r, A and B must have the same constant values for all values
of r.
= = exp ( - kH) * ( *
u
t&w !o
exp (~
3 sin kr rdr
- j y '
*>'
(4) f"
r r ^(X >
Y Z)fW{(X-xY+(Y-yy + (Z-zy}]dXdYdZ
>
= -y
-ii_' y r) sm &r . rdr.
with respect to k.
;
A somewhat simpler formula established at about the same time by Weber* is that,
if u(r, 6) isa function of the polar coordinates (r, 0) which has continuous first and second
differential coefficients at all points such that 0<r <a, whose value at the
origin is u ,
The formula
The formulae (2) and (3) are, naturally, much more easy to prove than (1)
and the proof of (3) is of precisely the same character as that of (2), the
arbitrariness of the order of the Bessel functions not introducing any additional
complications.
Following Hankel, many writers* describe the integrals (2) and (3) as
"Fourier integrals" or "Fourier-Bessel integrals."
On account of its greater simplicity, we shall give a proof of (3) before
proving (1) ; and at this stage it is convenient to give a brief account of the
researches of the various writers who have investigated the formulae.
As has already been stated, Hankel was the first writerf to give the
general formula (3). He transformed the integral into
and then applied the second mean-value theorem to the integrand just as
in the evaluation of Dirichlet's integrals. Substantially the same proof was
given by SheppardJ who laid stress on the important fact that the value
of the integral depends only on that part of the JJ-range of integration which
is in the immediate neighbourhood of r, so that the value of the
integral is
independent of the values which F(R) assumes when R is not nearly equal to r.
An attempt has been made by Orrf to replace the Bessel functions by any
cylinder functions, the w-path of integration being a contour which avoids the
origin but some of the integrals used by him appear to be divergent, so it is
;
difficult to say to what extent his results are correct. The same criticism
applies to the discussion of Weber's problem in Nielsen's treatise. It will be
shewn ( 14*5) that if, as Nielsen assumes, the two cylinder functions under
the integral sign are not necessarily of the same type, the repeated integral is
but it does not seem easy to construct a rigorous proof on these lines (cf.
14*64). A
more direct method of proof is given in a difficult memoir by
Du Bois Reymond|| on the general theory of integrals resembling Fourier's
integral. The proof which we shall give subsequently ( 14*6 et seq.) is based
on these researches.
Subsequently ErmakofflT pointed out that the formula is also derivable
from a result obtained by Du Bois Reymond which is the direct extension to
two variables of Fourier's theorem for one variable, namely
coordinates (x, y) and (X, Y) respectively, the formal result is fairly obvious
when we replace V(X, Y) by F(R, <>) but the investigation by this method
;
the repeated integral taken over an infinite rectangle in the (a, /3) plane may
be replaced by a repeated integral taken over the area of an indefinitely great
circle.
If the arbitrary function F(R, <I>) is not continuous, the factor F (r, <f>)
which occurs on the right in (1) must be replaced by the limit of the mean
value of F(R, 4>) on a circle of radius & with centre at (r, when 8 -*-0.
<f>)
This was, in effect, proved by Neumann in his treatise of 1881, and the proof
will be given in 146 14*63. The reader might anticipate this result from
what he knows of the theory of Fourier series.
has been examined by Bateman, Proc. London Math. Soc. (2) iv. (1906), p. 484 ; cf.
-
12 2.
\"F{R)s/R.dR
exists and is absolutely convergent; and let the order v of the Bessel functions
be not* less than \. Then
X
(1) {"udu \
F(R) Jv (uR) Jv (ur) RdR = \ {F(r + 0) + F(r - 0)1,
provided that the positive number r lies inside an interval in which F(R) has
limited total fluctuation.
The proof which we shall now give is substantially Hankel's proof, and it
is same general character as the proof of Fourier's theorem it will be
of the ;
set out in the same manner as the proof of Fourier's theorem given in Modern
Analysis, Chapter IX. It is first convenient to prove a number of lemmas.
* It seems not unlikely that it is sufficient for v to be greater than - 1 ; but the proof for the
more extended range of values of t> would be more difficult.
: ]
Letf |
F(R)*/R.dR exist, and {if it is an improper integral) let it be
a
It is convenient to divide the proof into three parts ; in the first part it is
assumed that F(R)\/R is bounded, and that b is finite; in the second part
the restriction that 6 is finite is removed ; and in the third part the restriction
that F(R) *JR is bounded is also removed.
(I) Let the upper bound of j
F{R)\/R be K. |
Divide the range of in-
tegration (a, b) into n equal intervals by the points xJt x2 , ... <r n _, {x =a,
xn = b), and choose n so large that
where e is an arbitrarily small positive number and Um and Lm are the upper
and lower bounds of F(R) \/R in the rath interval.
Write F(R) *JR = F{R m_ ) V#m-i + m (R), so that, when R lies in the mth
1
interval, |
w m (R) \^Um L m Now, when v ^ \, both of the functions of x,
.
Let A and B be the upper bounds of the moduli of these functions. It is then
clear that
*
F{R)Jv (\R)RdR
a
[*m
2 F{R m ^) V-R*,-, Jv (\R) V-K . dR
=1 - Xm-l
^ 2
[
-^
^R^W i
j**r
Jv {x)\Jx.dx + 2
/*
I
0> m (i)
AdR
2.Bwg Ae
* \*
+
VX-'
* Cf. Modern Analysis, % 9-41.
t The upper limit of the integral may be infinite ; and a^O. The apparently irrelevant factor
R preserves the analogy with 14-3 (3).
"
By taking X sufficiently large (n remaining fixed after e has been chosen) the
last expression can be made less than 2Ae/^X, and so the original integral
is o (1/^X).
r\F(R)\ s/R.dR<e,
J e
I
I
Ja
F(R) Jv (XR) Rd R U
\
I
\
f
J a
F (R) J v (XR) RdR\ + ^- r\F(R)\</R.dR;
\ \XJ C
then, proceeding as in case (I), we get
2BnK 2Ae
J F(R)J (XR)RdR
\
J a
v
still independent of X, and so we can infer that the integral (with upper limit
St \F(R)\^R.dR < .
8 J8
By applying the arguments of (I) and (II) to the parts of (a, b) outside these
intervals, we get
,
J a I
A, V^
where K is now the upper bound of |
F (R) *JR outside the intervals h. The
\
We shall next prove that, when v~^ \, and I F(R)*JR .dR exists and
Jo
is absolutely convergent, then
F(R)\JR.dR<
F
where A is the constant defined in 1441.
it is clear that*
I
["
j f%(22,
u)duldR- Tj P <j> (R, u) dRi\ du
T
^ji? CR)|v #.<^<R+J/
?-\F(R)\<JR.dRdi
^J j
J'0
Since this result is true for arbitrarily small values of e, we infer that
P [ <f>(R,u)dudR
Jo Jo
= r r<b(R, u)dRdu,
Jo Jo
the integral on the left existing because the integral on the right is assumed
to exist. If the integral on the left has a limit as X -* oo , it is evident from
the definition of an infinite integral that
Next we shall prove that, in Hankel's integral, the only part of the .R-range
of integration which contributes anything to the value of the integral is the
part of the path in the immediate vicinity of r, provided merely that F(R) \/R
has an absolutely convergent integral.
* The justification of the inversion of the order of integration for & finite rectangle whose sides
are X and /3 presents no great theoretical difficulties.
.
\
fb F(R) [RJy+1 (XR) Jv (Xr) - rJ^ (Xr) Jv (XR)] - TftlTf
i
*
= lim \J. (Xr) j* /,+. (XB) dR
J
- lim XrJv+1 (Xr) ^ </ (XB) <iR.
J*
Since both the integrals
fEgV^figgn
are ea; hypothesi absolutely convergent, it follows from the generalised Riemann-
Lebesgue lemma ( 14*41) that the last two limits are zero ; and so
to r as X -* oo
Let us first consider the integral obtained by taking the dominant terms
of the asymptotic expansions, namely
b K
2 f
(ur \vir $ir) du dR
f
I cos ( uR \vtr \ir) cos
7T\/r.J a .'0
- n 7T sjrj a |_
h ("sin X (R r)
R-r
cos (X (R + r) vir\ cos inr~
R+.r
dR
-
1
f
CHb-r) si n x
dx f
I
A(6+r) cos (a*
*
i>tt) ,
dx + cos vrr
.
log +-
b r*l
.
VrL.'A(-r) # Jx(*+r) a + rj
* It is permissible for 6 to be infinite.
14*44] MULTIPLE INTEGRALS 461
f I J (uR)J(ur)uR*dudR
Ja J
Now, of the integrals on the right, the first is the integral with respect to
R of an integral (with respect to u) which converges uniformly in any positive
domain of values of R and r, and a continuous (and therefore bounded)
so it is
function of r when r is positive and bounded.
J A(a-r) X
does so.
42
-
inr
_\/r
J rb
r-
} a J A
roo r
d
[_uK
j sin (uR hnr \ir) cos {ur \vtr \ir)
where ^ (\), < 2 (X) and 3 (X) are functions of \ and i2 which tend uniformly
to zero as \ -* oo .
462 THEORY OF BESSEL FUNCTIONS [CHAP. XIV
Now that all the preliminary lemmas have been proved, the actual proof
of Hankel's theorem is quite simple.
F(R)s/R = Xi(R)-X*(R)>
where X\ ( R ) an(^ X* ( R ) are nionotonic (positive) increasing functions.
and also
f
r+S X
f Xi
-'0
WJ " (w R ) J" ( Mr ) Us/B" dudR
fr+S r\
= Yj (r + 0) I J (uR) J v (ur) u JR dudR
.
Jr '
fr+S r\
+ (Xi ( r + 8 ) ~ Xi ( r + )1 J /
J" ( uR ) J (ur)u*JR. dudR.
r+( J
Afi
sin
Since
f a; ,
<tx
,
7r,
Jo x
as X - oo , 8 remaining fixed, it follows from 1444 that the first term on the
right tends to a limit as \ -*- oo while 8 remains fixed. And the second term
on the right does not exceed Ce in absolute value, where G is the upper bound
Hence, if
lim f f
Jv (uR)J(ur)uR i dudR = C / y/r, l
A-*. J r Jo
it follows that
+
lim P ( Xi( R ) J( uR ) J*( ur ) uRhdudR
*-*.*> J r JO
exists and is equal to X X\ (r + 0)/\/r.
rr+8 fK
then lim F(R) Jv (uR)Jv (ur)uRdudR
A-^eo J r -S J
exists and is equal to
C F(r+0) + C,F(r-0).
1
S/f Jo Jr
r+ *
f" f 1
= lim exp (p^u*) J v (uR) Jv (ur) uR* dRdu
i>-0 Jo J r
rr+S roo
= lim exp (- p*u 2 ) Jv (uR) Jv (ur) uR* dudR
p-^oJr .0
,. p+ fi
1 ( R* + r*) _ /flr\ Di ,_
by 1331(1).
as ^) -*- 0.
4p 2
- I dR
p _fc0 2jt>V7rJ r | j
1 ri//>
= lim -} I exp ( #2 ) <&e = ,
p-*o "J"* Jo
and similarly
1 r
C'2 = lim j I exp ( xz) dx = .
p-^0 v"* J -}6/p
The use of generalised integrals in the proof of the theorem seems to be due to
Sommerfeld, in his Konigsberg Dissertation, 1891. For some applications of such methods
combined with the general results of this chapter to the probleme des moments of Stieltjes,
see a recent paper by Hardy, Messenger, xlvii. (1918), pp.
8188.
The proof given by Hankel of his formula seems to discuss two points somewhat
inadequately. The first is in the discussion of
lim f |
F(R)Jv (uR)Jv (ur)uRdudR,
which he replaces by
In order to approximate to this integral, he substitutes the first terms of the asymptotic
expansions of the Bessel functions without considering whether the integrals arising from
the second and following terms are negligible (which seems a fatal
objection to the proof),
and without considering the consequences of \R vanishing at the lower limit of the path of
integration.
lim J J
Jv (uR)Jv (ur)uRdudR
*-.> J r+t J o
after proving by the method just explained that this is zero if tends to a positive limit
and is $ if {-=0, he takes it for granted that it must be bounded if -*-0 as X-^oo ;
and
this does not seem prima facie obvious.
f F(R) e (
v (XR) RdR = o (l/y/\),
Ja
rb
provided that \
F(R)*JR.dR
. <>.
fa^O if O^v^l,
[a >0 if v >\.
The theorem of 14*44 has to be modified slightly in form. The modified
theorem is that the repeated integral
[Note. If the two cylinder functions in the repeated integral were not of the same
type, i.e. if we considered the integral
A
<&9 ( R) <(c v
(r) u sjR du dR,
.
/:/:
it would be found that the convergence of this integral necessitates the convergence of the
integral
/A(6-r) 1-cos.r ,
dx;
(1)
(* udu r F(R) <&. (uR) <@v (ur) RdR
-, , n/ ^ / s > 2<r sin 2
o sin (a + vir) f
00
R riv 2
"
viv\ fjp
* This point was overlooked by Nielsen, Handbuch der Theorie der Cylinderfunktionen
(Leipzig, 1904), p. 365, in his exposition of Hankel's theorem.
466 THEORY OF BESSEL FUNCTIONS [CHAP. XIV
provided that the positive number r lies inside an interval in which F(R) has
limited total fluctuation.
( &,(itRy&y (ur)udu
Jo
1 r
uR <@v+1 (uR) c@v (ur) - ur t% v+1 (ur) ^ (uR)
R 2
-i*
X
[R <@, +1 (\R) <% (\r) - r <@y+1 (\r) <&, (\R)]
R -r 2 2
Hence we infer that, if r is not a point of the interval (a, b), then
K
f
F (R) \ <@, (uR) <@v (ur) uRdudR
n\ 2<r
2
sina sin (a + yTr) /-
6
R*-r*> v/ ^ p^ p
as X. -- oo ; and so the last repeated integral has a limit when \ -*- oo .
\\F(R)-F(r-0)\<e if r- 8 ^R<r.
(" *
Now take F(R) I # (uR) <@v (ur) ududR,
Jo Jo
and divide the 22-path of integration into four parts, namely
(0,r-8), (r-8,r), (r,r + 8), (r + 8,<x>).
Apply the second mean-value theorem as in 14*45, and we find that
K<
(* F(R)\ @ (uR) <@ v (ur) uRdudR
Jo Jo
2a*sinasin(a + *-7r)jf'- /* R - r* ^ /pw
| ^>
+ V,
14-51] MULTIPLE INTEGRALS 467
The integrals on the right converge to limits when \ -*- oo and, so, by
making e -* after \ -* oo , we infer that
X
r udu Jof
Jo
F(R) <$ v (uR) <@, (ur) RdR
IT Sin VJT
rco fr+6
r
+ F(r + 0) Vr lim <@v (uR) @ v (ur) uR*dRdu
-*0.'0 Jr
r
+ F(r-0)<Jr.\im I I* tfv (uR) <@v (ur) uR* dRdu,
To prove that the limits exist and to evaluate them simultaneously, take
F(R) = R when r<R<r+B and F(R) = for all other values of R.
We thus find that
--0.'o Jr-S
+ dRdu
I" f &* (uR) ^r (ur) uR' >
Jo
P -*.+oJo
u
since the second of these three integrals is convergent, and the third is abso-
lutely convergent when < p < 1 v.
Now the last expression can be replaced by a combination of the four
integrals of the types
and these are all absolutely convergent. They may be evaluated as cases of
Weber's discontinuous integral of 13 "4, and hence we find that
|
{6-+ 1 K+ i (ah) - f+ ^ +1 (ur)\ <& (wr)
%
_ <r 2
r* sin (a + pir) sin (a + vrr ) .T(v + lp)
2* sin pir sin vir T (v + rJT (p + 1)
.
W
=
t-J
7T Sin J/7T
L \ ^ / "^ V
after some algebra ; and the limit of the last expression, when b -* r + 0, is
simply ^^r".
In like manner it may be shewn that
r
lim (" ( <@v (uR) <&p (ur) uR*+' dRdu = \<x*r\
6++0J0 J r-S
and so we have proved that
[
udu F (R) <&. (uR) <@ v (ur) RdR
j
Z^{F{r + 0) + F(r-0)}-
n
2 ^ Sma
x
^ (a+ ^r- f ~^ , F(R)dR, v
Trsini/TT Br-^ilfr-r*)
Jo '
provided that %v%\, F(R) is subject to the conditions stated in 144, and
<@ (z) = er {cos aJ (z) + sin aF (z)}
and this is the general theorem stated at the beginning of the section.
(1) lim [
F{R) [R<@v+1 (\R) V. (Xr) -
=
r<@v+1 (Xr) * (Xi2)] ^^
0) + F(r-0)}, $<r*{F(r +
provided that r lies inside an interval in which F(R) has limited total
fluctuation and F(R) is defined to be zero when ^R < a, if the order of the
cylinder functions lies between and \.
Y ,
We shall now establish the truth of this formula for cylinder functions of
unrestricted order.
Let [R<&p+l (XR) <w (Xr) - r%% +1 (Xr) r$v (XR)] ^-^ = <*> (R, r ; X).
<D (22, r ; X) - <,,_> (R, r ; X) = -- OVi (\R) <$ v (Xr) + %Vi (Xr) ? (Xft)]
and so, by the analogue of the Riemann-Lebesgue lemma ( 14*41), we have
Choose n so that one of the integers v n lies between $, and then from (1)
This result is practically due to Weber*, and it was obtained by the method
indicated in 14*46.
To obtain the result in Weber's form, let
r/(R)RdR
J a
Carry out the integration with respect to u, and replace the integrated part
by the sum of the four terms written above, divided by R* r2 .
J,(B)F.W-J.(r)r.(B)
Since
is bounded near r, and has limited total fluctuation in any bounded interval
containing r, it follows that the integrals corresponding to the second group
of terms tend to zero as \->- <x> by the generalised Riemann-Lebesgue lemma.
,
When we use (1), we are therefore left with the result that
A
<& (ur) 9$ v (uR) uR dudR
lim
p / (R) f .
"
(6) f
uduT f(R)<@ v (ur)9$ v (uR)RdR
We shall now state precisely the theorem which will be the subject of
discussion in the sections immediately following. It is convenient to enunciate
the theorem with Du Bois Reymond's* generalisation, obtained by replacing
the Bessel function by any function which satisfies certain general conditions.
The generalised theorem is as follows
g{t)sJ (t), and the conditions (I) (III) are substantially those given in Neumann's treatise
published in 1881.
,
(III) If Q(R, <I>) denote the total fluctuation of F(R, <J>) in the interval
( 0, R), let fl(R, 3>) tend to zero uniformly with respect to 4> as R-*-0,
throughout the ivhole of the interval ( it, ir), with the exception* of values of <I>
in a number of sectors the sum of whose angles may be assumed arbitrarily
small.
Since \
F( R, <I>) - F(+ 0, <J>)
!
rR /
dt
Let g(t)tdt = O(R), and let j G(t) be convergent.
It will be observed that this is a theorem of a much weaker character than the
theorem of 14*41, in view of hypothesis (II) of 14*6. The reason of this is the fact that
G (\R) may be* (,/A) for certain values of R, and this seems to make arguments of the
-
type used in 14 41 inapplicable.
To prove the lemma, suppose first that T is bounded. Then, for any value
of <t>, F(R, <>) may be expressed as the difference f of two (increasing) mono-
tonic functions Xi (-^> )> %2 (-^ )> wnos e sum is the total fluctuation of
F(R, 3>) in the interval (0, R).
If R and R 1 are the extreme values of R for any particular value of <J>, it
follows from the second mean- value theorem that, for some value of R% between
R and R lf
rKR kR ,.,dt
*,.,dt , Ax f <
= Xl (,<*>) 0(*)y
.
+ X (*i.*) G(t)
J kR l J kRt t
C
x dt .
IttCJR,*)!** X
(< >,<t>) + $F(+0,<t>) + $\F(+0,)\,
whence it follows that
(dRd)
F(R,<&)G(\R)
Jf R
<2eT {*<>, *) + *(ao,<I^^(+0,*)l}d*
J -ir
= 2e r {n(oo,<t>) + \F(+0,<?>)}d<!>;
and, since F(+ 0, bounded, this can be made arbitrarily small by taking
3>) is
e sufficiently small, it is independent of the outer boundary of T.
and Hence
we may proceed to the limit when the outer boundary tends to infinity.
can be made arbitrarily small by taking \ sufficiently large ; and this is the
theorem to be proved.
We shall next prove that the existence and absolute convergence of the integral
r r (dXdT)
<t< x t)
JO J CD J
K
- lira
f"
A-.oo J oo J oo
(X,T Y)\ g{u>J{X*+Y*j\udu{dXdY),
./
For any given value of \ and any arbitrary positive value of e, there exists
a number /9 such that
,
JO J-irio I
= i
I
J
r-vJr JOr F R fi
( > ^ ^r o^) (<^*&)
-f f
*
f" ^(ii, <&)g(uR) R (dRd)udu
Jo J -it J/3 !
< ji f*
[" /"*!
-F(#, <&) w*dw !
itf (dRd<P)
J -irJ fi Jo
+ aT i'-n r
J J J fi
!
F(R, 4>) I
R* (dRd^) u* d
< .
Jo
f !'-wor F(R, <5>)g(uR) R (dRd<P) udu
J
exist.
w. b. f. 16
474 THEORY OF BESSEL FUNCTIONS [CHAP. XIV
f* dR r* fJJ?
-Xk(+0,*)/ O(XB)^+to(8,*)- Xl (+0,*)}J^(xfi)^ f
o
where ^ f ^ 8.
f* dR /"** dii
Now
I t I
G(\R)^\ = G(u)-\<2G.
Jt R \
\
I J a* I
Hence
f
" l'F(R, ) Q (MJ)^*> _ 2,JJ(+ 0,*) f<? () *
< 27r.2etf-l-e.250
= 2eC {2tt + }.
that
Hence,
is
/" /* *<* )
./ -WO
to say
for large values of X,
(Mi) ^ -tt
- ftriW( + 0, )
<2eC(2ir +
f (.)
JO
2?)
*
U
+ o(l),
Urn"
f
I
K-^colJ-wJo
'
f JF(#, 4>) (\R) ( ^?> -
Jti
2,r JWF<+ 0, 0) f G () -
Jo **
Km
A-xJ-irJO f f *(**) 0(XS)<-^ -"-
Applying the result of 1462, we see that Neumann's theorem has now
been proved.
In the special case in which g (u) = J (w), we have
ru
I tg(t)dt = uJ x (w),
Jo
so that G (u) = uJ x (u),
(3) ruduT
J-wJ-aa
r F(R,<P)J [u</(R? + r*-2Rr cos (<P - <f>)]
RdRd
Jo
= 27rJ*LF'(r,0),
where 0lF(r, <f>) now means the mean of the values of F(R, <E>) when (R, 3>)
traverses the circumference of an indefinitely small circle with centre (r, ^).
f(0 ) = =0
t
5 ?!L2 f ' f" /(S, <l>) P (cos 7 ) sin @d4>de
**?r Jo J -n
by a limiting process ; in this formula
* Math. Ann. v. (1872), pp. 135137 ; cf. Lamb, Proc. London Math. Soc. (2) n. (1905), p. 384.
R = k%, r = k0,
so that R, r are substantially cylindrical coordinates of the points with polar
coordinates (, , 3>), (k, 6, <j>) ; the function of position /(, <E>) is then de-
noted by F(R, <l>), and Pn (cos7) becomes approximately equal to J (nvr/ic),
where
ra-
2
= R? + r2 - 2Rr cos (4> - <f>).
2n Kn
*>,<) = Km 1 lf
*Tr
r F(R,^)J (n^/ K )^d^.
-*.> =0 JO J -n K*
But this procedure can hardly be made the basis of a rigorous proof, be-
cause there are so many steps which require justification.
It is possible that the method which has just been described is the method
by which Neumann discovered his integral formula in 1862. Concerning his
method he stated that "Die Methode, durch welche ich diese Formel so eben
abgeleitet habe, ist nicht. vollstandig strenge."
CHAPTER XV
THE ZEROS OF BESSEL FUNCTIONS
15*1. Problems connected with the zeros of Bessel functions.
There are various classes of problems, connected with the zeros of Bessel
functions, which will be investigated in this chapter. We shall begin by proving
quite general theorems mainly concerned with the fact that Bessel functions
have an infinity of zeros, and with the relative situations of the zeros of different
functions. Next, we shall examine the reality of the zeros of Bessel functions
(and cylinder functions) whose order is real, and discuss the intervals in
which the real zeros lie, either by elementary methods or by the use of Poisson-
Schafheitlin integrals. Next, we shall consider the zeros of Jv (z) when v is not
necessarily real, and proceed to represent this function as a Weierstrassian
product. We then proceed to the numerical calculation of zeros of functions
of assigned order, and finally consider the rates of growth of the zeros with
the increase of the order, and the situation of the zeros of cylinder functions of
unrestrictedly large order. A
full discussion of the applications of the results
Except in 15*4 15*54, it is supposed that the order v, of the functions
under consideration, is real.
The zeros of functions whose order is half an odd integer obviously lend
themselves to discussion more readily than the zeros of other functions. In
and by Rayleigh*; and more recently Hermitef has examined the zeros of
Jn+l (a?). The zeros of this function have also been the subject of papers by
RudskiJ who used the methods of Sturm; but it has been pointed out by
Porter and by Schafheitlin that some of Rudski's results are not correct, and,
in particular, his theorem that the smallest positive zero of Jn ^(x) lies
between J (n + 1) tr and \ (n + 2) tt is untrue. Such a theorem is incompatible
with the inequality given in 153 (5) and the formulae of 15*81, 15*83. >
It was stated by Daniel Bernoulli* and Fourierf that J (z) has an infinity
of real zeros and a formal proof of this result by an analysis of Parseval's
;
of m, (0, 2, 4, ...), and is negative for odd values of m, (1, 3, 5, ...). Since
Jv (x) is a continuous function of x when x > 0, it is obvious that Jv (x)
has an odd number of zeros in each of the intervals (\ir, it), (7r, 2ir),
(fir, Sir),....
Some more precise results of a similar character will be given in 15-32 1536.
To prove Lommel's theorem, let x = (wi + \6) ir where < ^ 1; then, by
obvious transformations of Poisson's integral, we have
M _~r(i/ + |)r(|).(2m
*(** p+ cosfrm
r
t/ " w + ^J {(2m + ey-u*}*->
au '
^m du.
J \m {(2m+ey-u*}i-'
If now we write u = 2r 1 + U, and then put
{(2m + 0y-(2r-l + UYy-i-{(2rn+OY-(2r-l-Uy}'-*=fr (U),
it is clear that
tV = Cf^^sin^irU.dU,
Jo
and, 8ince|| v^ t fr(U) is a positive increasing! function of r.
* Comm. Acad. Sci. Imp. Petrop. vi. (17323) [1738], p. 116.
+ La TMorie Analytique de la Chaleur (Paris, 1822), 308.
% Berliner Abh., 1824, p. 39.
Studien iiber die BeueVsehen Functionen (Leipzig, 1868), pp. 6567.
This is the point at which the condition r ^J is required ; the condition v > - & ensures the
||
It follows that
and so
sgn Jv (nnr + 10tt) = sgn [(-) m {*'
-J- (, - ^-0 + (m_, - w_ ) + ...}] i* 8
= sgn(-l)"-,
since vm' is obviously not negative.
The zeros of Ji(x), as well as those of J (x), have been investigated by Baeh'r, Archives
Nlerlandaises^vu. (1872), pp. 351 358, with the help of a method which resembles the
Bessel-Lommel method. Baehr's result for Ji (x) is that the function is positive when x
lies in the intervals (0, n), (fw, 3r), (-, 5tt), ..., and that it is negative when x lies in
the intervals (7r, 2a-), ($-, 4jr), (Qir, 6n-), .... The function Ji(x) has also been investi-
gated in this way by C. N. Moore, Annals of Math. (2) ix. (1908), pp. 156162.
The results just stated are of a less exact nature than the results obtained with the aid
of slightly more refined analysis by Schafheitlin ( 15*33 15-35).
Itwas noted by Whewell, Trans. Camb. Phil. Soc. ix. (1856), p. 156, that J (x) has a
zero between 2 and 2 ^2, and that the function Hq (*) has some real zeros.
This result is sometimes expressed by saying that the positive zeros of J, (x)
are interlaced with those of J,+ x (x).
j
[*-Jp (x)} = - x-"J,+1 (x), ^ {tf"
+* J^ (x)} - ar J v (x) ;
the first of these shews that between each consecutive pair of zeros of
a?-' J (a,) at least one zero of x~" J, +1 (x), and the second shews that
there is
between each consecutive pair of zeros of x*+1 J, +1 (x) there is at least one zero
of x*+x J (x)\ and the result is now obvious.
* This is a special case of a theorem proved by Sturm, Journal de Math. i. (1836), p. 109.
480 THEORY OF BESSEL FUNCTIONS [CHAP. XV
If i>^ 1, the zeros are obviously still interlaced but the smallest zero of Jv +i(x) is
Jr/\ r
v (x) + Jv+2 (x)
= 2(i/ +
/ \ l) r
/ +1 (X),
. .
at any positive zero of Jv (x) the functions Jv+1 (x) and Jv+2 (x) have the
same sign but at successive zeros of Jv (x) the function J +1 (x) alternates in
;
sign, and so there are an odd number of zeros of Jv+i (x) between each con-
secutive pair of positive zeros ofJv (x) interchanging the functions /+2 (x) and
\
Jv (x) throughout this argument, we obtain Porter's theorem that the positive
zeros of Jv+Q (x) are interlaced with those of Jv (x).
Jo
dx dx
for the integral is positive when x is positive and the expression on the right
would vanish at a repeated zero of A J, (x) + BxJv f
(x).
Again, since
<@ {x), x<@v\x)
Let ^?r (x) and ^P (x) be any distinct cylinder functions of the same order
we shall prove that their positive zeros are interlaced*.
Now it is known that, at consecutive positive zeros of #r (x), <$,' (a?) has opposite
signs, and therefore, from the last equation, <& (x) has opposite signs; that is
to say & 9 (x) has an odd number of zeros between each consecutive pair of
<
positive zeros of 9^ (x); similarly <@ {x) hasan odd number of zeros between
each consecutive pair of positive zeros of 9% 9 (x) and so the zeros must be ;
interlaced.
, =0 i!r(i> + ? + 1)
would be a series of positive terms.
j
tMt)JAa t)dt = ^-^[ Max)_^_ ^(c^-gj ,
15*26. The analogue of Lommel's theorem for functions of the second kind.
It is not possible to prove by the methods of 15*25 that Yv (z) has no
complex zeros in the region|| in which arg2 < ir. But it has been proved | |
by SchafheitlinU that Y (z) has no zeros with a positive real part, other than
the real zeros.
* La TMorle Analytique de la Chaleur (Paris, 1822), 308 ; see also Steam, Quarterly Journal,
Y,(z) = e T
'"ri
I!
When arg*=ir, Yy {-z)2i cosi'T7(-), and hence, by 3-63 (1), Yv (z)
cannot vanish unless v is half of an odd integer. This type of reasoning is due to Macdonald,
Proc. London Math. Soc. xxx. (1899), pp. 165179.
**I Archiv der Math, und Phys. (3) i. (1901), pp. 133137. In this paper Schafheitlin also sub-
jects the complex zeros of Yi (z) to a similar treatment.
15*25-15'27] ZEROS OF BESSEL FUNCTIONS 483
For let /9 be a complex zero of Y (z), and let j3 be the conjugate complex,
so that ft is also a zero of Y (z). Then, by 511 (8) and 351 (1),
l't7.(/3t)Y,(frt)dt
= I,(A * )
/s^ft2 L
Fo(/3ar)
-is "r-J-^09--^ u g'
and so, if /9 = pe*", we have
and the expression on the left is positive while the expression on the right is
n 1n (y +
!
/ \
yn
ii -r
yr then the function
i)
,
f ()
has at least as many complex zeros as g<m,*(Z)' After Hurwitz, we write
Again, it is not difficult to deduce from the recurrence formulae ( 9*7) that
fora discussion of the Bessel coefficients. The analysis of this section differs in some respects from
that of Hurwitz ; see Watson, Proc. London Math. Soc. (2) xix. (1921), pp. 266272.
484 THEORY OF BESSEL FUNCTIONS [CHAP. XV
Hence, for sufficiently large values of m (i.e. those for which v + 2m is positive),
the curve m<j> (, 17) = lies in the finite part of the plane, and <f> m (ff, rj) is
negative when <f> m+1 (f , rj) is zero so the curve $+,. (f, rj) = lies wholly inside
one or other of the closed branches which compose the curve m (f
<j> , rj) 0.
Hence asm-^oo, the complex zeros of g^, (f) lie in bounded regions of
the f-plane, and consequently have limit-points.
Jv ^ }
r(!/+2m + l)
can be made arbitrarily small in any bounded domain of the f-plane, by taking
m sufficiently large, it follows from Lagrange's expansion * that the number of
zeros off, () in any small area number of zeros ofg2mv ()
is at least equal to the
in that area when m is sufficiently large
and so / () has 2s complex zeros.
;
None of these zeros is real, for if one of them were real it would be a limit
point of two conjugate complex zeros of gtmiV (), and so it would count as a
double zero of/(); and/() has no double zeros.
Again, from the series for/ (f ) it is seen that, when v lies between - (2s + 1)
and (2s + 2),/ () must have one negative zero, and it cannot have more than
one negative zero, for then gm> v (f) could be made to change sign more than once
as X varied from to 00 [since g*m,,{C) can be made to differ from /() by
an arbitrarily small number], and this is impossible.
For similar reasons /(") cannot have more than 2s complex zeros.
It seems that this theorem was not proved before 1929, except in some
simple special cases, such asm = l and 2 (cf. 15'22). m
The formula
Jv+m (z) Jv (Z) Rm, v (Z) Jv -\ (z) Bm-i, H-l (z)
* Cf. Modern Analysis, 7 '32.
t Ann. Sci. de VEcole norm. sup. in. (1866), pp. 55 95.
15*28, 15*3] ZEROS OF BESSEL FUNCTIONS 485
shews that, since Jv (z) and Jv _x (z) have no common zeros, the common zeros
of J(z) and /+, (z) must satisfy the equation
Rm-i,y +1 {z) = 0,
i.e. they must be algebraic numbers.
been proved by Siegel* that J,(z) is not an algebraic
It has, however,
number when number and z is an algebraic number other than
v is a rational
zero; hence follow theorems which include Bourget's hypothesis as a special case.
pfo y)
The reader will find a more systematic investigation of these theorems in various
papers by Schafheitlin, notably Journal fur Math. cxxn. (1900), pp. 299 321 Archiv ;
der Math, una" Phyt. (3) I. (1901), pp. 133137; Berliner Sitzungsberichte, in. (1904),
pp. 8385.
For brevity, the smallest positive zeros of J (#), JJ (a), J" (x), . . . will be
"
called j , j,', j The smallest positive zeros of Y (x), F' (x), Y " (x),
v ...
"
will similarly be called yv , yv\ y v
Therefore, so long as < x < v, both Jv (x) and xJJ (x) are positive increasing
functions so that J,,
and jj cannot* be less than v.
Again, from the differential equation
vJv"{v) = -J '(v)<0,
v
and so J "(x)
v has become negative before x has increased to the value v from
zero. Hence, when! v>\,
(2) j,"<p.
Next, since
., 2(i/ + l)L v(v + 2)\ ,. L ~ 2(,/ +1)0/ +2)1
JT ,
= v
- T
Jv - ^ -A
+2 (x)
x
>
|1
-^ J (*) |1 }
(),
the expression on the right is positive so long as x < v + 2. Now, if ^V were less
than V{" (" + 2 )} the expression on the right would be negative when x is equal
(
to jj (which, from a graph, is obviously less than jy ), and this is not the case.
Therefore
(3) jJ><J{v{v + 2)).
JJ <j v ,
so that /(>) and Jv+*{j*) are both positive. If now we put x=j v
'
in the
formula
it is obvious that
(4) jv'<j\2v(v + l)}.
(/ + 3) Jv (x) + 2 {v + 2) |l - 2 {V +
^ + 3)
|
</,+, () + ( + 1) J* () = 0,
we deduce that
jr <V{2( + 1)^ + 3)},
and therefore
(5) */{v(v+2)}<j r <</{2(v + l)(v + 3)}.
* Cf. Riemann, Partielle Differentialgleichungen (Brunswick, 1876), p. 269.
4J'h-4 () - /r (*) {R*. v (#) + R 3, ,+a (#) - ^s, 0) - ^1, M-2 (*)}
,
Hence
(10) jJ<y v <j,-
This inequality (withj/ replaced by v + \) was established by Schafheitlin
with the aid of rather elaborate analysis.
* Berliner Sitzungsberichte, in. (1904), p. 83.
^r(*)(^ycoB(* + a-W-iir)+o(^).
This indicates the possibility of proving inequalities consistent with
i^,(/*)| = 0(l/vV)
when /in is large.
fa, ...form an increasing sequence whose members are less than *J(2/tt).
(II) The values assumed by x^\9^,(x)\ when x takes the values fa, fa+i,
A*r+2 form a decreasing sequence whose members are greater than \f(2jir)
provided that
(i) v> V3, (ii) fa' > * {4i>2 + 4 + ^(48^ + 13)}/(4i* - 3).
Consider the function
A (x) <@S (x) + 2B (x) <@, (x) <&,' (x) + G (x) <$;* (x) = @ (*),
where A (x), B(x), G(x) are to be suitably chosen. We have
' (x) = [A' (x) -2(x*-v*)B (a-)/* ^V 8
}
8
(*)
+ 2 {R (x) + A(x)-B (x)/x - (a3 - v*) G ()/*} # (x) #' (x)
+ {C (x) + 2B (x) - 2G(x)/x] <&'* (x)
= D(xY$J*(x),
where D (x) = G' (x) + 2B(x)-2C (x)/x,
* Cf. Proc. London Math. 80c. (2) xvi. (1917), pp. 170171.
15*31, 15*32] ZEROS OF BESSEL FUNCTIONS 489
provided that A (x) is chosen arbitrarily and that B (x) and C(x) are then
defined by the equations
2B (x) = a? A' (?)/(** - v%
C(x) = [B' (x) +A(x)-B (as)/*} /(a? - v*).
ID {x) (a? v )* = 2
a?
3
(3a* + 14a? a + 4*4) > 0,
z/
<0,
provided that 4n? > 3 and x exceeds the greatest root of the equation
(4v -3)ar*-8i/2
a
(^ + l)a* + i/*(4/
8 - 1) = 0.
In this case (a?) is a decreasing function and we can apply arguments, similar
to those used in theorem (I), to deduce the truth of theorem (II).
By means of the integrals which have been given in 6*12, it has been
shewn by Schafheitlin* that the only positive zeros of J (x) lie in the intervals
(m7T + f tr, mir +^ir) and the only positive zeros of Y (x) lie in the intervals
(mir + \ir, mir + 7r), where m = 0, 1, 2, ....
Jo( 6 '
'"ttJo sinflVcostf
it is obvious that, when mir <x< mir + \ir,
sgn {sin (a + \G)) = sgn (- l)m ,
Since <f><^ir, the second of these intervals is the longer; and the function
23 cote
sin ycos
is an increasing function* of when x >^ir and is an acute angle.
Hence to each value of between and 20 there corresponds a value
between 20 and \ir which sin {\0 0) has the same numerical value, but
for
has the positive sign, and the cofactor of sin {\0 <f>) is greater for the second
set of values of than for the first set. The integral under consideration is
consequently positive, and so J (x) cannot have a zero in any of the intervals
(mir + fair, mir + ir). Therefore the only positive zeros of Jo (x) are in the
intervals (mir + ir, mir + %ir).
r
and so, for such values of x, $ Consequently the only zeros
v (x) is not zero.
of 9y (x) lie in the specified intervals, and there are an odd number of zeros
in each interval, with the possible exception of the first if a > f ir+ \vir.
Next we obtain the more precise result that the only positive zeros of <$ (x)
lie in the intervals
Next we obtain the more precise result that the only positive zeros of 9BV (x)
lie in the intervals
(mir tt + \vrr-a., mir \tt + %vir a),
where m = 0,
1, 2, ..., except that, if a is sufficiently near to ir, there may be
a zero in the interval (0, %vir ^ir a), and there may be one in the interval
(tt a, ^7r 4- \vir a).
We use the same notation and reasoning as in 15 33 ; only now, if
f(2<f>-ir)<f(24>+ylr).
To obtain this result, observe that
4l log T^r^rl - ** {
cose2 ( 2 <*> + *) + cose2 ( 2 * ~ *)}
r v-\ 2i> +i I
^ Lcos (20 + +) cos (20 - +) J
sin (20 + +) sin (20 - ^ )
But [cosec 2 (20 + f) + cosec2 (20 - VO] sin (20 + +) sin (20 - $)
is an increasing function of ijr, and therefore, a fortiori,
[cosec2 (20 + ip) + cosec2 (20 - ^)] cos (20 + ^) cos (20 - ^)
is an increasing function, since tbis function exceeds tbe former by an increasing function
We shall now prove that, if v > J, those zeros of the cylinder function
Jy (x) cos a Yv (x) sin a
which exceed (2i/ +1) + S)/ir lie in the intervals
(2v
cot dtf -
r (v f )T (|) J o 33 1 cos*+i e )
2y-l 7T . 2y + 2 ir
< ^1<
2i/+3'2 2i/ + 3'2'
so that 0! is a positive angle less than 2, provided that
v +h v
arc tan <
x 4i> +6
We suppose now that
x>(2v + l)(2v + 3)/ir,
so that X is certainly less than 2.
Then, by the second mean- value theorem, there exists a number O, between
and X) such that
{ +%)
!o sin2 "^ e atf
J,d0\ cos'+itf J
-
-(cot *-"*,! (cosJww + ^)
foot** 1 ox .b -
cos(x + a-v0o -l0 o) )
+ a - v0 - 0)/cos K+ *
cos {x
is stationary when sin (x + a v0 + \&) = 0, and for such values of the
-
fraction is equal to 1 /cos* * 0.
Hence
nenCe
co S (x + a-v0
cob*+0,
o -W
cannot exceed numerically the greatest value of l/cos*"""**? in the interval (0, 0^),
and therefore
cos ( m7r +ft) _ cos (g + - "fl> -*fl>) | _ gM/_ ! vm
}- sSm
(
Sgn
SDT1
1>
) cos-** cos"+*0o
Therefore, since the sign of sin (x + a v0 + \0) is the sign of ( l)m when
lies between X
and ^7r, we see that, for the values of a; under consideration,
sgn 9B9 (x) = sgn (- l) m .
Hence, when x exceeds (2v+ 1) (2v+ S)/ir, 9$v {x) has no zeros in intervals of
the type
(rrnr a + vir \ir, mir a+ \ vir + \ir),
and so the only zeros of <$* (x) which exceed (2j/ + 1) (2i/ + 3)/tt lie in intervals
of the type
(mir a + \vir + \tr, mir a + \vtt + J ir),
and this reduces to Schafheitlin's result* when a = and v is an integer.
The reader will observe that this theorem gives no information concerning
the smaller zeros of ^v {x) when v is large it will be apparent in 15*8 that ;
there are a large number of zeros less than (2i + l)(2i/ + S)/tt, and that
interesting information can be obtained concerning them by using Debye's
integrals.
* Schaf heitlin gives (2v + 3) (2r + 5)jr as the lower limit of the values of x for which the zeros
lie in the specified intervals.
t Bulletin American Math. Soc. v. (1899), pp. 385388.
X Cf. Modern Analysis, 3-63.
15-36, 15*4] ZEROS OF BESSEL FUNCTIONS 495
du
I u
cv
as = \ v I _
,
CV
J J ,
coefficients inside the closed curve , and d/dv indicates differentiation along the normal.
By taking v=J {J(a?+y )} and the curve to be x2 +tf2 =*jo2 Weber* deduced that u
2
,
Bocher inferred from this result that since ^n (r) cos n6 satisfies the requisite conditions
except at the origin, if a circle of radius j is drawn with centre on the axis of x and
r
subtending an angle less than ir/n at the origin, $n (r) must vanish somewhere on the
circle. Hence the positive zeros of wn {r) are such that consecutive zeros are at a distance
apart less than 2/ , and the distance from the origin of the smallest of them does not
exceed j -ll + cosec 5- J-
These results are of interest on account of the extreme simplicity of the methods used
to prove them.]
following theorem, which gives some indication as to the situation of those zeros
of Jv (z) which are of large modulus. In this investigation it is not supposed
that v is restricted to be a real number, though it is convenient to suppose
that v is not a negative integer. When v is real the results of 15*2 to some
extent take the place of the theorem which will now be proved.
Since z~* J (z) is an integral function of z, the number of its zeros inside
v
Cis
1_[ d\og{w-"Jr ('w)} dw _ 1
[ JrlM dw .
where ij
lt (w) and r)
2< (w) are (l/w) when |
w ||
is large.
Next take the integral along the upper horizontal side of C ; this is equal to
. tB+l+Jnr+Jir
1 fiB+mTT+Ji/ir+iir Q i _ /^Y)
^ + ^-
. .
log +Q(i/J)J
/(l/)
-iw + iflfcO + i,
as B oo.
Similarly the integral along the lower side tends to the same value, and
so the limit of the integral along the three sides now considered is m + \v + .
Lastly we have to consider the integral along the fourth side, and to do
this we first investigate the difference
riB+mv+iw+ir
Now I tan(w \vir ^7r)dw = 0,
J tB+Hr+J>w+J
and so
1_ riB+mw+ivw+i* jr+i ( W )
dw
2iri J -u
iB+mv+iv*+\w ^kW
T
~-i(2i/ + l)+0(l/m).
Hence the limit of the integral round the whole rectangle is m+ (l/m).
* Allowance is made for the indentations, just specified , in the first step of the following analysis.
15-41] ZEROS OF BESSEL FUNCTIONS 497
That is to say, the number of zeros of z~*Jv {z) between the imaginary axis
and the line on which
R (z) = W7T + {$ R (v) + 1} ir
is exactly ra.
Note. The approximate formulae quoted for the functions of the third kind shew
that the large zeros cannot have a large imaginary part and so all the zeros of / () lie ;
inside a strip whose sides are parallel to the real axis and at distances from it which are
bounded when |
v \
is bounded.
jy<3 , ... being all unequal (15*21). We draw a (large) rectangle D, whose
vertices are where A and B are positive, and we suppose that j>m
A iB,
are the zeros of highest rank which are inside the rectangle.
We now consider
z /,+, (w)
dw,
2irtJ*J)W(W
2 Z)' J (w)
where z is any point inside the rectangle, other than a zero of J (w), and v is
.Jv,m'
[2 + Jv,n Jy,n)
It follows that
= 2-rri ]j)W(w z)
* J*+1 (W) dl.
J ( w)
v
* Acta Soc. Seient. Fennicae, xi. (1880), pp. 273293. Cf Modern Analysis, 74.
.
t If R ( jy>n )=0 for any value of n, we choose jvn to have its imaginary part positive.
498 THEORY OF BESSEL FUNCTIONS [CHAP. XV
We next shew that, by giving A and B suitable sequences of values which
increase without limit, Jv+1 (w)/J,(w) can be taken to be bounded on D.
Since this function is an odd function of w, it is sufficient to consider the
right-hand half of D.
We takeA = Mir + M (\v \) where M a positive integer and then
4- ir, is ;
also to be so large that we can take the functions iy r ,i(w), *7 f2 (w), defined in
L
and when the expression does not lie within
is* less than \ or greater . han 2 ;
That is to say Jv+l (w)jJ v (w) is bounded on the whole of the perimeter of
the rectangle D as B and M tend to infinity.
Hence
if _ JjtiW,,,,^
2itijdw(w z) J p {w)
and therefore f
1
K
' Jy(z) n=l l*->.n jv,n) =1 \*
When we integrate, we find that
This formula was assumed by Euler, Acta Acad. Petrop. v. pars 1, (1781) [1784], p. 170,
when v=0, aud subsequently by various writers for other values of v ; cf. 15*5, 15'51.
T/ueorie der BetseVschen Funktionen, I. (Bern, 1898), pp. 123 130, and it was given
explicitly by Kapteyn, Monatshefte fUr Math, und Phyt. xiv. (1903), pp. 281282.
* Because **
TTa 5
t If we take the rectangle to have its vertices at AiB, -A'iB, we see that the two series
on the right converge separately.
: , '
The expansion on the right of (1) is evidently expansible in a power series; the
such a series have tieen expressed as determinants by Kapteyn, Proc. Section
coefficients in
of Sci., K. Acad, van Wet. te Amsterdam, vm. (1905), pp. 547 549, 640 642 Archives ;
Neerlandaises, (2) xi. (1906), pp. 149168. Some associated formulae have just been
published by Forsyth, Messenger, L. (1921), pp. 129 149.
15*42. The Kneser-Sommerfeld expansion.
An expansion which, in some respects, resembles the partial fraction
formula obtained in 15*41 is as follows
The integral round the indentation tends to zero with the radius of the
indentation, whether v be an integer or not ; and the integrals along the two
parts of the imaginary axis cancel.
remains bounded on the other three sides of the rectangle when B -*- oo and
when A -*-oo through the values specified in 1541.
Hence the limit of the integral round the rectangle is zero, and so the
limit of the sum of the residues of the integrand at the poles on the right of
the imaginary axis is zero.
j (2v*)= n(i-).
w
As has already been stated ( 15'41), this formula was assumed by Euler; if it
is differentiated logarithmically, then
-log/
dz
(2\/*)=i *
z
w=1 an
oo oo
- 2 X
n=l m^O^n
provided that |
z \
<a t ; and the last series is then absolutely convergent.
00
Put 2 l/am+1 = orOT+1 and change the order of the summations ; then
p+p 2s
"~
1* . 2s 3 s
.
+ ' "
* Acta Acad. Petrop. v. pars 1, (1781) [1784], pp. 170 et teq. A paper by Stern, Journal /fir
Math, xxxiii. (1846), pp. 363 365 should also be consulted.
t From 15*25 it follows that the zeros are positive and unequal.
15-5] ZEROS OF BESSEL FUNCTIONS 501
multiply out the product on the right, and equate coefficients of the various
powers of z in the identity; we thus obtain the system* of equations
= <T \ = <r <r
1 lt 2 x ,
1 ,
whence
o-i = 1, <r2 = J, a = , <r4 = ^, <r = 1^
3 8 y , o-6 = ^$r,
Since < 0^ < 02 < a3 < . . . , it is evident that
l/cti
m < <rm , <rm+1< <rml*i,
and so o"m
_1/m < <*! < <rm /<rm+1 .
m ^ml^m+l
l/On^d-',
-2
and then using the inequalities
l/a2m < am , <x' m+1 < <r' m /a.z ,
Euler deduced that a 2 = 7*6658, and hence that o3 =1863, by carrying the
process a stage further.
These results should be compared with the values
ax = 1-445796, a2 = 76178, a3 = 18-72,
derived from the Tables of Willson and Peirce, Bulletin American Jiath.,Soc. in. (1898),
pp. 153155.
The value of a, is given by Poissont as 1 446796491 (misprinted as 1 46796491);
according to Freeman J this result was calculated by Largeteau for Poisson by solving the
quartic obtained by equating to zero the first five terms of the series for J (2 Jz) the ;
magnitude of the sixth term is quite sufficient to account for the error.
* This system is an obvious extension of Newton's system for an algebraic equation,
t M4m. de I'Acad. R. des Sci. xn. (1833), p. 330.
t Proe. Camb. Phil. Soc. ra. (1880), pp. 375 377. Of. Freeman's translation of Fourier's La
Thiorie Analytique de la Chaleur, p. 310, footnote.
502 THEORY OF BESSEL FUNCTIONS [CHAP. XV
15*51. Rayleigh' s extension of Eider's formula.
x i
and writing 2 -^
=i J" v,n
The smallest positive zeros of J (z) and /i (z) are deduced to be 2*404826
and 3831706.
Immediately afterwards Cayley f noticed that o-( r can be calculated rapidly when r is >
a power of 2 by a process which he attributed to Encke J, but which is more usually known
as Graeffe's method of solving an equation.
The method consists in calculating <r( r when r is a power of 2 by starting with the )
given equation and forming from it a sequence of equations each of which has for its roots
the squares of the roots of its predecessor; and aJ?) then rapidly tends to a ratio of
equality with l//2r,i.
where Br is the rth Bernoullian number ; this formula is an evident consequence of the
equation
J *w-(s)* 8ini -
Extensions of some of these results to the zeros of zJv (z)+hJv (z), where h '
is a constant,
have been made by Lamb, Proc. London Math. Soc. xv. (1884), p. 273.
The smallest zero of Jv {z\ for various values of v between and 1, has recently been
tabulated by Airey, Phil. Mag. (6) xli. (1921), pp. 200205, with the aid of the Rayleigh-
Cayley formulae.
* Proc. London Math. Soc. v. (1874), pp. 119124. [Scientific Papers, i. (1899), pp. 190195.]
t Proc. London Math. Soc. v. (1874), pp. 123124. [Collected Papers, ix. (1896), pp. 1920.]
t Journal fur Math. xxn. (1841), pp. 193248.
Die Aufibsung der hoheren numerischen Oleichungen (Zurich, 1837).
I!
Einleitung in die Theorie der Bessel'schen Funktionen, i. (Bern, 1898), pp. 130 131.
15-51, 15-52] ZEROS OF BESSEL FUNCTIONS 503
[Note. The procedure of calculating the sum of the rth powers of the roots of an
equation in order to obtain the numerical value of its largest root seems to be due to
Waring, Meditationea Analyticae (Cambridge, 1776), p. 311 other writers who were ;
acquainted with such a method before Graeffe are Euler (cf. 15*5); Dandelin* Mem. de
VAcad. R. des Sci. de BruxeUes, in. (1826), p. 48; Lobatschevsky* Algebra, or Calculus
of Finite* (Kazan, 1834), 257.]
zw
P( m
0) ~ X
, 1.9 1.9.2.5.49
-
*'
-27(8^ + ^!(8^
nt *' ax 1 1.9.25
Q( 0)
~-TT^ + 3W~----
For sufficiently large values of x, P (x, 0) is positive, Q (x, 0) is negative
and the quotient Q(x, 0)jP(x, 0) a negative increasing function of x.
is
where 6 and 1 are certain functions of # and m which lie between and 1.
{1-P2-Q2}/P2,
where P, Q stand for P (x, 0), Q (x, 0) ; and, by the asymptotic expansions, this is ultimately
positive.
504 THEORY OF BESSEL FUNCTIONS [CHAP. XV
Now consider the equation
Wl-l
2 V, + Vm
cot(a?-^7r) = ^2 ur + 0um
x
r=0
in which it is temporarily supposed that and 1} instead of having their actual
values, are any numbers which lie between and 1.
The equation now under consideration involves no functions more compli-
cated than trigonometrical functions. If a? were supposed complex, there
would be a number of contours in the a?-plane each of which enclosed one of
the points nir \ir and on which cot (a? \ir) exceeded the modulus of the
| \
By Biirmann's theorem * the modified equation would have one root inside
the part of the contour which surrounds mr tt, and this root can be ex-
panded in descending powers of mr \tt.
We thus obtain an expansion for the root of the equation in the form
* '
r=0 (nir - Sir)
in which the coefficients r (0, 0i) are independent of n but depend on
f and X ;
and it is readily perceived that the first in of the coefficients are actually
independent of and lt so that, when r < m, we may write
f r {0 t x )=f r.
as and 1 vary between and 1 and it is clear that the upper bound of the
;
and X are given their actual values which they have at the zero under con-
sideration, the sum of the terms after the with is still (w~ am_1 ).
That is to say, it has been proved that there exists one zero (nearly equal
to nir ^7r), and its value may be written
= (mr-%Tr)+ 2 J
x ,
\ w+1 + (n--*)-
Xf^nir \ir 2 ^~
,=o (nir ^iryr+1
It remains to calculate the first few of the coefficients fr . If
, ,
1 33 3417
tan y oo 8a? 512a? '
16384a?6
* Cf. Modern Analysis, % 7'31.
.
1 25 , 1073
^~8^-384^ + 5l20^"-*
,
sothat
1 25 1073
-(nir-i7r)~Q--KQTZ,+ T
8a? 384a 5120a?
31 3779
t^ + 8(w7r _ i7r) - 384 (n7r _ i7r) + i5360(n7r-i7r) _ -'
1
-(
x-{mr i
.
This series is adequate for calculating all the zeros of J {pe), to at l eas* nve
places of decimals, except the smallest zero, for which n 1.
cot(s - \vrr - i^ + ) =
p^ v y
It seems unnecessary to prove the existence of such zeros (with large
positive real parts) or the fact that they may be calculated as though the
series for P
(z, v) and Q (z, v) were convergent, because the proof differs from
The expression for the large zeros of a cylinder function of any given order
was calculated after the manner of Stokes by McMahon*; but the subsequent
memoirs of Kalahnef and Marshall J have made the investigation more simple
and have carried the approximation a stage further with no greater expendi-
ture of work in the calculation.
* AnndU of Math. ix. (1895), pp. 2325 ; see also Airey, Proe. Phys. Soc. 1911, pp. 219224,
225232.
t ZeitschriftfUr Math, und Phy*. lit. (19Q7), pp. 5586.
t Annals of Math. (2) xi. (1910), pp. 153160.
Cf. Nioholson, Phil. Mag. (6) xix. (1910), pp. 228249.
W. B. F. 17
506 THEORY OF BESSEL FUNCTIONS [CHAP. XV
Again
{j$} -z-hw-l'*-*,
aTC tan
and, when we differentiate this equation, and use 363 (3), we find that
2/(irz)
cty _
dz ~JS(z) + Y/(z)'
so that, by 7*51,
When the expression on the right is expanded as far as the term involving
ljz 9, we find that
(n - 1) (5/r> -s-
1535/* 2 + 54703/a - 375733) '"'*
2 15 zs
_rr ~ /*-! ,
(/i-DQ*-25) Qi- 1X^-114^ + 1073)
23 s 3.2 7 23 5.2 10 * 5
lOS^/S7
Therefore the large zeros of J, (z) cos a Yv (z) sin o are given by the
asymptotic expansion
-l 4i/
2
(4p-1)(28i/ 2 -31)
(n + \v \)ir a
8 {(n + \v - \) ir - a} 384 {(n + %v- \) ir - aj
3
* This equation (in the case v = \) was given by Gauss in his notebook with the date Oct. 16,
1797, but no clue is given concerning the method by which he obtained it. [Cf. Math. Ann:
[Note. The fact that J* (z) + F2 (z) has a simple asymptotic expansion shortens the
analysis in a manner which was not noticed by Marshall he used the equations ;
l+
dz M*' |_<fc2 z* J^ J/3'
and he solved the latter by assuming a descending series for M.]
fi +3 8
7/x, + 82fi - 9
^ 8ft 384& 8
zeros ot * ? *-* is
fi +7 7/a2 + 154/a + 95
A- 8& 384ft
s
order is varied.
The equation in z
has an infinite number of roots, the values of which depend on v since / (z) ;
*'o->| +
P^ = 0.
* Annals of Math. ix. (1895), pp. 2529.
(2) |=^|V,'()*
dp jj\
jJ\+i(j)Jo
(j)j
wr +1
"
This formula shews that when v is positive, the positive zeros of Jv (x) in-
crease as v is increased.
Equation was stated without proof by Schlafli, Math. Ann. x. (1876), p. 137 and the
(2) ;
R. des Set. de Liege, (3) n. (1900), no. 3, in the case of the smallest zero of Jv (x).
{Jv (c)
is constant, so that
dc[d .
\Tv (z)Y] A* , (Y,(2
= 0,
W'JJ
and therefore
vx v
ircdp [_
' dp ' dp ] z=0
Hence, by 1373 (2), we have
~ = 2c "
(3)
dp f
Jo
K (2c sinh t) r**dt.
sin (o pit) = 0.
* The reader should note that the analysis in the latter part of Gegenbauer's memoir is
vitiated by his use of Rudski's erroneous results ( 15*1).
t Berliner Sitzungsberichte, v. (1906), pp. 82 93 ; Jahretberieht der Deutachen Math. Vereini-
gung, xvi. (1907), pp.272279.
15*6] ZEROS OF BESSEL FUNCTIONS 509
and they can only come into existence or disappear when c fails to be an
analytic function of v, i.e. when c = 0.
It follows that the positive zeros of p (z) are derived from those of W (z) ^
by a process of continuous variation as v varies, except that one positive zero
disappears whenever v passes through one of the specified negative values.
gives the nth positive zero, when the positive zeros are regarded as arranged
in order of magnitude.
If, however, v has varied so that it finally lies between (a/7r) k and
(a/Tr) k 1, where & is a positive integer, k zeros have disappeared, and so
This type of argument is due to Macdonald, Proc. London Math. Soc. xxix. (1898),
pp. 575 584 ; it was applied by him to the discussion of the zeros of Bessel functions of the
first kind of order exceeding 1.
If we take any point with positive coordinates (<>, y<>) and draw from it a
line to the right and a line downwards terminated by the #-axis, it is evident
that the curve ^x (y) ^ meets each of the lines in the same number of
points. It follows that the number of zeros of $* (y ), qua function of v, which
exceed v is equal to the number of positive zeros of ^, (y) qua function of y
which are less than y This is a generalisation of a theorem due to Macdonald -f>
.
who took v = and the cylinder function to be a function of the first kind.
Pig. 33 illustrates the general shape of the curves Jx (y) = 0, the length
of the sides of the squares being 2 units. A much larger and more elaborate
diagram of the same character has been constructed by Gasser j, who has also
constructed the corresponding diagram for F(y) = 0. The diagram for
<$x (y) = is of the same general character as that for Jx (y) 0, except that
* This does not lead to any real loss of generality.
t See a letter from Macdonald to Carslaw, Proc. London Math. Soc. (2) xni. (1914), p. 239.
t Bern Mittheilungen, 1904, p. 135.
510 THEORY OF BESSEL FUNCTIONS [CHAP. XV
the portions of the curves below the axis of x consist merely of a number of
isolated points on the lines on which 2x is an odd integer.
y
o I I
L Fig. 33.
[Note. The reader will find it interesting to deduce from 13*73 (3) that, if d is a zero
of 9BJ (2), then
dd 2d C*
(4) ^ = ~-^ J
(c' cosh 2t - r 2) K (2d sinh t)e~ ** dt,
The sign of dd/dv has also been discussed (by more elementary methods) by Schafheitlin,
Jahresbericht der Deutschen Math. Vereinigung, xvi. (1907), pp. 272 279; but the analysis
used by Schafheitlin is extremely complicated.]
the circle r = a, and if the straight edges of the membrane are fixed, the dis-
placement in a normal vibration is proportional to
Jv (rp/c) sin vd cos {pt + e),
where If the circular boundary
c is the velocity of propagation of vibrations.
of the membrane is fixed, the values of apjc are the zeros of Jv (x), while if
the boundary is free to move transversely they are the zeros of Jv (x). '
The K
v (z), where v is a given positive
zeros of the function number (zero
included), and z lies in the domain in which axgz < ir, have been studied | |
qualitatively by Macdonald*.
from a consideration of the integral given in 1371 ; for, if z = reia were such
a zero (r > 0, \ir < a < tt), then z = re~ia would be another zero but the ;
K
rr
v
/
{re-) K
isTjr /
v (re-~)
^ = If" exp {- v^
^ Jo
>0,
f
^cos2a) /r*\dv
j* -)
In the first place, there are no zeros on the lines arg z ir, unless v \
* Proc. London Math. Soe. xxx. (1899), pp. 165179.
t This is not the number given by Macdonald.
K
it is evident that the number of zeros of v (z) in the pair of quadrants under K
consideration is equal to the number of zeros of zv Kv (z) inside the contour,
and this is equal to l/(2rr) times the change in phase of z v K v (z) as z traverses
the contour.
Now the change in phase is
K
:
. . , . .
being numerically less than two right angles and having the same sign as the
sign of cos vir.
Hence the total number of zeros of K v (z) in the pair of quadrants J in which
R (z) is negative and arg z < it is
| \
and the reader will find it easy to verify that this number is the even integer
which is nearest to v \.
When v $ is an integer, K v (z) is a polynomial in z multiplied by a
function with no zeros in the finite part of the plane, and so the number of
zeros for which R (z) < is exactly v \.
Next consider the portion of the plane for which ir < argz ^ 2w.
If we write z ^e***, we have
K v {z) = - ^--[F(D + *(l + 2e 8) /()],
and so K {z)
has a sequence of zeros lying near the negative part of the
v
imaginary axis. The zeros of large modulus which belong to this sequence
are given approximately by the roots of the equation
it may be verified that they are ultimately on the right or left of the imaginary
a
axis in the s-plane according as cos i>7r is less than or greater than ; i.e.
according as v differs from the nearest integer by more or less than \. The
sequence does not exist when e 2vni = 1, i.e. when v is half of an odd integer.
where t = sinh w w + i tan (3 (cosh w 1), and the contour in the plane of
the complex variable w is chosen so that t is positive on it.
e-^du, e~ VT dv
J oo J 3
are both positive.
values of "9 which make "9 + a equal to an odd number of right angles.
^
and so arc .tan [
<
Jy ( V )
: .
\
>
<X < ? 7r 1
Jv (v)=-Y(v)ta.ny v ;
The following Table gives the sexagesimal measure (to the nearest half
minute) of the angle whose circular measure is yv it exhibits the closeness of \
1 2 3 4 6 8 10 12 00
V t f
30
yy 28 39' 29 23 29 38' 29 45' 29 51' 29 54' 29 56' 29 57$' 29" 58' 2958'
J* {v) dv Yv (v)
~^r >
and this inequality has already been established in 1374.
We are now in a position to infer that <@v {v sec fi) has a zero for which
and so, if < a < tt, the number of zeros of <WV {X) in the interval (0, x) is the
greatest integer contained in (" + a)lir + .
A simple theorem which may be noted here is that, when vs > , it follows from 1374
+
that, if 8 is any positive number, P *^| rf* decreases as 8 is increased. This result
shews that the interval between consecutive zeros of <$y (x) decreases when the rank of
the
zeros is increased.
tan(f-f7r)=-Q(,i)/P(,),
the value of so obtained is the value of i>tans # at the zero with an error
which is (i>~ J ). The value of is approximately 2*383447, and hence the
smallest positive zero of Jv (x) is
v + v*x 1855757 + 0(1).
tan(-i7r) = -Q(,)/P(M),
of which the smallest root is approximately = 0*847719, we find that the
smallest zero of Yv (x) is
v + v^x 0-931577 + 0(1).
The formula for the smallest zero of Jv (x) has been given by Airey, Phil. Mag. (6)
xxxiv. (1917), p. 193. Airey's formula was derived by using Debye's asymptotic expansion
of 8*42 for Jv (x) when x has a value such that
x-v = 0{v*) + o(yi).
For such values of the variables, it has not been proved that Debye's expansion is valid,
and although Airey's method gives the two dominant terms of the smallest zero of / (x)
correctly, the numerical result which Airey gives for the smallest zero of J (x) is not the
'
same as that of 15-83. The reason why Airey's method gives correct results is that Jv (v + f)
is expansible in powers of so long as { is o (v), and in this expansion it is permissible to
substitute Debye's formulae for Jv (v), Jv ' iy\ Jv" (v), ....
A formula for the smallest zero of J- V (x) was given by Airey. This zero may lie any-
where between and the smallest zero of Jv (x), according to the value of v.
15-81, 15*82] ZEROS OF BESSEL FUNCTIONS 517
It does not seem to be possible to make further progress by the methods used in this
section. We now make a digression to explain the methods of Sturm (which have
shall
been applied to BessePs equation by various mathematicians), and we shall then give an
investigation which leads to the fascinating result that the two expressions which, in this
-1 so that approximations are obtained
section, were proved to be 0(1) are in reality (i ),
for the smallest zeros of Jv (x) and Tv (x) in which the errors are 0(v~*), i.e. the errors
become negligible when v is large.
[Note. An elementary result concerning the smallest zero of Jv (x) has been obtained
from the formula of 5-43 by Gegenbauer, Wiener Sitmngsberiehte, cxi. (2a), (1902), p. 571;
if /* = v + e where <
e < 1, then the smallest zero of Jiv + (x) is less than twice the smallest
zero of Jv (x), because, for the latter value of x the integrand cannot be one-signed.]
in which the invariant / is positive, then the greater the value of /, the more
rapidly do the solutions of the equation oscillate as x increases.
dx*
+ x* /
\
1,2 A c2_2_i.l
and that, when x > c, 1 -$= > ^ - .
A slightly more abstruse result is due to Porter, American Journal of Math. XX. (1898),
pp. 196198, to the effect that, if v 2 > \ and if the zeros of <@ (x), greater than ^(v 2 - ),
in ascending order of magnitude are clf c2 c3 ... then cn+1 cn decreases as n increases.
, ,
d2 u T _ d2 u2 r
zt + i^-o,
x
^? + /..-o
such that, when x = a,
Wj (a) = u2 (a), Mi' (a) =m 2
'
(a),
x lies between a and the first zero of u x (x) in the interval, so that the first zero
of u x
(x) in the interval is on the left of the first zero of u 2 (x).
Further, if w/ (a) has the same sign as u x (a), the first maximum point of
i
u (x) in the interval
x |
is on the left of the first maximum point of'j u 2 (x) \,and,
moreover
max I
Uj (x) |
< max |
u 2 (x) j
To prove the theorem f, observe that, so long as u x (x) and u 2 (x) are both
positive,
d2 u2 ~ d2 u^
U2 =(T ~, T .
UlU*
~dxv
Ul
dxJ ^
1
*'
'
[du 2 du 1
~\ x .
Since the expression now under consideration vanishes at the lower limit,
we have
n \
du 2 du x
(1) -
^dx-'^dx-^
Hence we have
djujuj)
dx
and therefore
W%0,
LMiJa
that is to say,
M2 (a:) u 2 (a)
(2) > = 1
u (x) " Ui (a)
x
* To simplify the presentation of the proof of the theorem, it is convenient to change the signs
of u, (x) and u 2 (x), if necessary, so that u x (x) is positive immediately on the right of x = a; the
signs indicating moduli may then be omitted throughout the enunciation.
t The theorem is practically due to Sturm, Journal <}e Math. i. (1836), pp. 125 127, 145 147.
15*83] ZEBOS OF BESSEL FUNCTIONS 519
It follpws that just before 1*1 (a?) vanishes for the first time Ma(#) is still
positive, and it has remained positive while x has increased from the
value a.
Again, if w/(a) is positive, as well as v^ (a), then v^{x) must have a maximum
before it vanishes, and at this point, fi lf we have from (1)
so that 2'(/*i) is positive and v^(x) must be positive in the interval (0, /*,).
Therefore the first maximum point /*2 of u^ (x) must be on the right of /t^.
Finally we have
max %! (x) = w, (jj^) < w2 (/Hi) $ u 2 (ft*) = max u^ (x),
and the theorem is completely proved.
When two functions, Uy, (x) and it, (x), are related in the manner postulated
in this theorem, it is convenient to say that i*j (x) is more oscillatory* than
i/ 2 (x) and that w2 (x) is less oscillatory than v^ (x).
In the first place we reduce Bessel's equation to its normal form by writing
x= ve; we then have
(3)
[^ + "* (** " 1 >] *' W = -
(4) [^+2^]u-0,
since e* - 1 > 20 when 6 > 0:
The general solution of (4) is
<x /2*i/0*>
.-**,(?*);
and the constants implied in this cylinder function have to be adjusted so
that u and its differential coefficient are equal to <@ (ye) and its differential
coefficient at 6 = 0.
*
The reason for the use of these terms is obvious from a consideration of the special case in
which 2} and I2 are positive constants.
f These results supersede the inequalities obtained by Watson, Proe, London Math. Soc. (2)
xvi. (1917), pp. 166169.
520 THEOEY OF BESSEL FUNCTIONS [CHAP. XV
It follows that a function which is (slightly) less oscillatory than 90 {x),
when x ^ v, is
The formula for the less oscillatory function, combined with the result
stated in 8*43, suggests that we should construct a function of the type*
sin s #
(5) 3(tan/3-)^
cos 2
/3V(l+itan2 /9)
'
sin3 /3
Now ^3(tan-/3)
cos 3
V(l + tan2 )_
is negative .when tan 2 /3 < V24 3, and it is positive for greater values
of tan2 /9.
* The multiple of the cylinder function is taken so that the product satisfies a differential
equation in its normal form cf. 4*31 (17).
;
15*83] ZEROS OF BESSEL FUNCTIONS 521
is a certain angle between arc tan V{V24 3} and \ir The sexagesimal
measure of O is 59 39' 24"-27.
V{3 (1-/9 cot 0)} [r (|) (*)* . 9W (?) J_j [v (tan fi - /3)}
Since, by Sturm's theorem, the zero of 9v (x) lies between two expressions of
this form, we see that the value of the zero of *@v (x) which is next greater than v
is expressible in the form
v + $\M + 0(v~i).
When ^?r (x) is equal to /" (x) it is easy to verify from a Table of Bessel
functions of orders that
\, = T926529 + (ir*),
The maximum of the function Yv (x) cannot be treated in this manner because its
first
first maximum is on the right of its first zero this follows at once from 15*3, because
;
* This restriction is trivial because we have to consider values of j3 for which */3
3 is bounded;
i.e. small values of /3.
t The permissibility of this follows from the second part of Sturm's theorem just given.
X This value of is approximately 0*685548.
CHAPTER XVI
NEUMANN SERIES AND LOMMEL'S FUNCTIONS OF TWO VARIABLES
16*1. The definition of Neumann series.
2 a n Jv+n (z)
n=0
is called a Neumann series,although in fact Neumann considered* only the
special type of series for which v is an integer the investigation of the more
;
mann kind has been suggested by Nielsen, Math. Ann. lv. (1902), p. 493.
series of the first '
=o ni
TO =o " z
For a very general discussion of generalisations of all kinds of series of Bessel functions,
the reader may consult memoirs by Nielsen, Journal fiir Math, cxxxn. (1907), pp. 138
146; Leipziger Berichte, lxi. (1909), pp. 3361.
Various expansions of types which resemble Neumann's (other than those given in this
chapter) are due to H. A. Webb, Phil. Trans, of the! Royal Soc. cciv. (1905), p. 487 and
Nielsen, Atti delta R. Accad. dei Lincei, (5) xv. (1906), pp. 490497.
Now, by 9-1,
-i U0.(*)./.W;
t z M=0
and this expansion converges uniformly on the contour. It follows at once
that
/(*)- 5 b n z,
we see that
and so
fOo =b ,
(>1)
This result shews that the Neumann series corresponding to a given function
assumes a simple form whenever a simple expression can be found for the sum
=0 W*
Let f(z) be a function of z which is analytic and one- valued in the ring-
shaped region defined by the inequalities
r^\z\^R.
Let C and c be the contours formed by the circles
\z\ = R, \z\ = r,
both contours being taken counter-clockwise; then, if z be a point of the region
between the circles, we have
i //(*)& 1 /()*
f()=
JK
,
f
' 2-TriJc t-z 2ttiJ c z-t
^^ ^M^f/(t)On(t)dt+X^On (z)j
o c
f(t)Jn (t)dt.
/<,)-
*=0
ik*+ n=l
i .
*
we have, as in 1611,
J x ' 1m Jo t z
and so
(1884), pp. 298 816 for some special cases of the expansion.
16*12-16'14] NEUMANN SERIES 525
where
/(*) = S Kz\
=o
then
r{v + n - m)
(3) On = (v + n)t2'+- K
m=0 W!
- -= 2 B
J> Z n =0
n .^ (t)JIIL+in (z)Jp+ ^n (z) )
an =
(2)
^jc f(t)B ni ^{t)dt,
(4) a^ =
^
and Q it) is Neumann's second polynomial ( 9 "4), is valid provided that f(z)
c
tf{f)Cln {t)dt,
Gegenbauer's formula has been investigated more recently by Nielsen, Nouv. Aim. de
Math. (4) II. (1902), pp. 407410.
z- = 2>r(l v + l)2, -
{
r-Jiv+n (z),=0 ni
^-m + l) On-2m>
_<! 2 >+"-wt r(|i/ +
(6) an 2, t
Let 5 an Jy+niz) be any Neumann series, and let the function defined by
n=0
this series and its analytic continuations be called f{z).
Let also
The function defined by f(z)N and its analytic continuations will be called
the associated power series off(z).
and, in fact, it can be proved that f(z) has no singularities which are not also
singularities oif{z)N .
2
i. r> + , + t)r(tr * w -
pp. 493496.
16*2, 16*3] NEUMANN SERIES 527
^
cos {W(l -)} 0(^)^3^.
From the theory of analytic continuation it follows that, if <f>
(z) is analytic
forany value of z, so also is f(z), provided that the path of integration is suitably
chosen and so all the singularities off(z) must be singularities of $ (z).
;
2 S an T(v + n+l) n
then all the singularities of F s (z) are expressible in the form 8y, where /3 is
some singularity of F t (z) and y is some singularity of F2 (z).
g
Tfr+n + l)
n=0 r(v + n + %)
is at the point z=l,it follows that all the singularities of (z) are singularities (f>
of f(z)x ; and therefore all the singularities of f(z) are singularities cff(z)N ;
special interest.
* It is assumed that R(r + ) is positive; if not, the several series under discussion have
to be truncated by the omission of the terms for which R (v + n + %) is negative, but the general
argument is unaffected.
t Acta Math. xxn. (1899), pp. 5564 Hadamard, La Serie de Taylor (Paris, 1901), p. 69.
;
t For such theorems concerning the expansion of 16-14, see Nielsen, Math. Ann. m. (1899),
p. 230 et seq.
:
^p{^(<-l/Q}^
^I Jm
J (z)P f cos 20)
(*) /* (COS m =- JL
^lj ^ _ +
f
(0+)
2ti C08 2 ^ 1}
,
5 JU.()P.(co.M)-^,j*
V{(*'+l)(w'+sin0)}'
has been given by Jolliffef, who proved that the series on the right satisfied
the same differential equation as J,* ($z). This expansion is easily derived as
a special case of 11*6 (1), but the following direct proof is not without
interest
By Neumann's formula (
5 '43) we have J
J" (
^-7ri V{<(l-0}'
and, if we expand J (z VO into the series
r / u\ 2t " *
2
(2i/ + 2w + l)r(2i/ + w + l)
J* <* VO -
Ti o m!r(2* + l)
x & (- m, 2v +m+ 1 ; 2v +1 ; <) /,++! (s),
* Bologna Memorie, (4) vm. (1887), pp. 125143. Pincherle used elliptic functions of modulus
cosec 6 in his result.
t Messenger, xlv. (1916), p. 16. The corresponding expansion of z* J- j (J*) Jv ($z) was obtained
by Nielsen, Jfyt Tidsskrift, rx. b, (1898), p. 80.
J If B + i) < 0,
(j we use loop integrals instead of definite integrals.
16-31] NEUMANN SERIES 529
where
r
f '^ (*
a> = W9 - *>~* F (-rn,2v + m + l; 2, + 1 *)<**
\ZVt*i\
> >
7T .
If
ml i
l + 1) Jo ;
"
= i=J=.
TT.WilJo
I
V- (1 _ tr d
[<r
'~!,^- tri]
dm
*
by w partial integrations.
we find that
- ! 3 ...(2n 1) 1
v+ft _ i /i _ \_| j
f
aan "^* 2.4... (2n) Jo
( '
_1 1.3...(2w-l)r(i/ + rt + ^)r(^)
~tt' 2.4...(2w) T(i/ + n + l) '
=o
=o
We adopt this value offm (v) and then it is found by the same method that
(2) - M*, m
z
(z) fm^ l {v) J^ (z) + 2^, m_ 1 (z).
=o
Therefore, since
we have
- j w |j o r(1 _ n+ j-)
(**) } >
/r+l( *) ( ** )
t.-r(^- + *) }'
* Recent applications of Neumann series to the solution of equations of mixed differences are
due to Bateman, Proc. Int. Congress of Math. 1. (Cambridge, 1912), pp. 291294.
16*32] NEUMANN SERIES 531
were also investigated by Lommel, but the results are not so interesting. As examples of
his expansions the reader may notice that
These results were given by Lommel, though his formulae contain numerical errors.
r
a
Mz-v) j(a _ v)dv .
iz
2
Jo z-v
*
the sums of the alternate terms of the series have been expressed by Kapteyn
in the form of integrals from which this integral may be deduced, and
conversely Kapteyn's formulae may be deduced from this integral.
We proceed to establish this result by a simplified form of Kapteyn's
methods.
The series may be written in the formf
\z r (0+ > 0+) Z
\z 2 {J (z) + J-n+1 (#)} Jn (a) = j^-j-
\&TT%) J
f<
2 <- +
M= o
r ) I*-"
n=0 J
ST
* A'jeuip
+ m/ = 53
P 3 (?
+
1) * {*" ( " D] da - * [J
Archief voor Wiskunde, (2) vn. (1907), pp. 2025 Proc. Section of ScL, K. Akad.
;
(a) " ^ (a)/ '>-
van Wet. te Amsterdam, vn. (1905), pp. 494500 Kapteyn has subsequently summed other series,
;
Therefore, on integration,
I = Ce-^ + 1
&
f
.'
V"
o
-"'
{Jo (v) - J, (v)/t) dv,
Hence we have
that is to say
aJ-li^!lj
(1) ! nJn (z)Jn (a) = l ( (a-v)dv.
n=l * J * u
If we select the odd and even parts of the functions of z on each side of
this equation, we find that
-5[ Mz ~ v)
+ ^^^1 /. ( - ) *.
4 J
\
( z- z+v )
4J I
*- -* + *> i
1V
'
4j ( *+o *- I
VZ f a J v (z v) T , .,
intricate integral equation. In fact, the functions which are amenable to the
theory seem to be included in the functions to which the complex theory is
applicable, and simple functions have been constructed to which the real
variable theory is inapplicable.
J
(1) /() = 2 (4 + 2) Jm+1 (*) f" -^^f{t) dt ;
and we shall establish the truth of this expansion under the following
conditions
Jo
exists and is absolutely convergent.
(II) The function f(t) has a continuous differential coefficient for all
1 f* f
- j J, (t - v) J, (t + v)\ ,
=
I fI
Jo (* - V)
f*
V
^ {f(t + V) +f(t - V)} dtdv
+
Jo
f (x - v) \
Jo
f(v - 1) ^P dtdv.
t
We now transform the last integral by using 12*2, and then we have*
f
Jo
7 V (x - v)f(v -
; o
1) ^P dtdv
*
U
= I" f Jo (u - t)f(x - u)^P-dtdu
J oJ o t
= I J (u)f(x u) du
Jo
Hence
J
(3) I (4i + 2) Jm+1 (x) f
" -^^-f{t) dt
=0 .0 c
Now
so that
write
F(v) is
Z '
a continuous function of
Ji(0
{/(+*)+/<*-)}* 3
v, since Ji(t)/t has
m an absolutely
convergent integral.
If then we are to have S=f(x) when x has any value in such an interval
as (0, X), we must have
X
f J (x-v)F(v)dv = 0,
Jo
throughout this interval ; and, differentiating with respect to x,
\F(x)\^fjF(v)\dv,
F{xy
* hat \
^W^V
where A is the upper bound of |
F(x) |
in the interval and n is any positive
integer.
(2) is established.
The sufficiency of equation (2) for the truth of the expansion* is evident
from (3).
[Note. If f{x) is not an odd function, we expand the two odd functions
separately ; and then it is easy to prove, by rearranging the second expansion, that
2 /
n
f(z)=X cn z ,
f(z) B = 2 cn .n\zn ,
=0
supposed convergent for sufficiently small values of \z\ % may be represented
by the integral
jo
The function f(z)i, maybe termed the reduced function (la reduite) oif(z).
If the Neumann series which represents f(z) is
f(z)= 2 an Jn (z),
tt=0
then we have, formally,
1 * |V(l+^)-l) w
2, an
Now put
z
-t
and we see that
1 +? .( 2
Bf/G^).-i^-
Hence, if the Neumann series for f(z) is 2 an Jn (z), then the generating
tt=0
function of 2 an n is
=o
i-r / u-r/'
provided that this function is analytic near the origin.
More generally, iff(z) has a branch-point near the origin of such a nature
that
(2 >
.l-^-r^Ai^.),-
* Mini, de la Soc. de Phys. de Geneve, xxxit. (19021905), pp. 295368.
.
In like manner, if
then I ,=
26z ( 1 + g2 )
nil (l-2az-z*?+4b*z*-
This result, which is immediately deducible from Gainer's theory, was set as a problem
in tlie Mathematical Tripos, 1896.]
(w -
z2 mr\
==C08 +
{2 2w-^-)^
(4) tfn+i (w, z) - F_M+1 (w, z) = sin (^ + ~ -^ J
The last equation may be derived from the preceding equation by replacing
n by n + 1.
There is no difficulty in extending (1) to define functions of non-integral
order ; for unrestricted values of v we write
in it differs from that adopted subsequently by the factor ( - l) m Much of Lommel's work is repro-
.
duced by J. Walker, The Analytical Theory of Light (Cambridge, 1904). The occurrence of Lommel's
functions in a different physical problem has been noticed by Pocklington, Nature, lxxi. (1905),
pp. 607608.
W. B. F. 18
538 THEORY OF BESSEL FUNCTIONS [CHAP. XVI
It is evident that
which are closely associated with Lommel's functions, have been studied by Kapteyn,
Proc. Section of Set., K. Akad. van Wet. te Amsterdam, vn. (1905), pp. 375 376, and by
Hargreavcs, Phil. Mag. (6) xxxvi. (1918), pp. 191199, respectively.
and hence
d2 z* 1
^ JJ V (w, z) = - U. +2 (w, z) -- Uv+l (w, z),
and consequently
16*51, 16*52] lommel's FUNCTIONS 539
2
Z* z
A cos s 4- B sin ^ ,
where A and B are independent of z, it is clear from 165 (6) that F_ K+2 (w, z)
is also a particular integral. Therefore V v (w, z) is a particular integral of
{6) + J- +z(z) -
dz->-;dz- w>-\w)
These equations are due to Lommel, Miinchener Abh. xv. (1886), pp. 561 563.
16*52. Recurrence formulae for LomnieVs functions of two variables.
~ U.
OW
(w, z) = 2 (-)" (v + 2m) (/*)*+* J, +2m (z)/z
m -Q
(2) 2
L Uv (w> z) = u~* {w> z) ww) * u +i (w
" > z) -
(3) ^Mttf.sH-^F^M;,*),
Hence it follows that Uv {cz, z), and similarly V_ v+2 (cz, z), are particular
integrals of the equation
and so 7,, (w, 0) and F_ +2 (/, 0) are expressible in term's of Lommel's functions
of one variable by the equations
HO)
Of these
V. y+2 (w ,0)
The formulae
Jo
J^ x
{zt) . sin {\w (1 - i*)} . *<fc,
which are valid when R (v) > 0, may be verified immediately by expanding
the integrands in powers of w and then using the result of 12*11 (1) in
16*53] lommel's functions 541
(3) ^( w^) = - 27 _ r
_
I _^_J i
J^i-MQ.cnllwil-ty.i-tYdt,
J
1
J^ (zt) exp { \iw (1 - **)} p dt. .
* JO
/_! () exp {+ \iw (1 - *>)} .
*>
<&
The integral converges at the lower limit when R (v) > and at the upper
limit when R(v)<, if w and z are restricted to be positive.
To evaluate the last integral, swing the contour round until it coincides
with the ray arg= +
ambiguity in sign being determined by the
tt, this
ambiguity in sign in the integral; such a modification in the contour is
permissible by Jordan's lemma.
w" f
J o
/,_, (zt) exp \iw (1
{
- 1*}} V dt .
2 ^. mir ;+ m) ( J,
2 * +2W
' "exp(? }M)A
that is to say
(8) ^^(*)ntt(l-lO}.ri-rin(|+^-^).
: ,
It follows at once from (1) and (2) combined with 16*5 (6) that
(11) Vv (w, *) - -
z v 1
w^
-
fao
ji
/,_, {zt) cos {\w (1 - f)) J*
,
(12) V (to, z) = - ^_
ji
A_, (jf) sin {| W (1 - *>)}
provided that w and -2 are positive and R (v) > \.
The following special formulae are worth mention
Um (z,z) ^
(13) p j^ {zt) cos ^ (1 __ <8)}
<9n
^
- - <)} 2n~ 1
f /*-* (*0 sin {|* (1 . <ft,
.'o
(14)
ff+i(*,*)
=p ^ (ft) C08 n, (1 _
<s)
| .
| ^
= f
!
Jm-i (zt) sin ft* (1 - t*)} .
* eft.
. o
(16)'
v
fV.w^WI ^).
Jo sin \ 2 /
8
wcos\2iW
The last results should be comparod with 13 '3 ; see also Hardy, Trans. Cumb. Phil.
Soc. xxi. (1912), pp. 10, 11.
The formulae of this section (with the exception of the contour integrals) are all to be
found in one or other of Looomel's two memoirs.
= sin (}*) T- w * tf -
(^ *) wtU^ ( ztj]
. , ,
(1) ~ f
^ JAzt) sin (Iwt^t'-'dt
a
= cos L.^(^,)-sinK.^ +1 (l ,,) + ^g, ),
and, similarly,
=r cos (\wt*) *- dt
(2) x
f\ (zt)
w
f1
( 3)
z==;
Ji- (^) cos [\w (1 - * 2
)} . t" dt
j
(4)
^ J
J,- r (zt) sin { w (1 - *)] V dt .
and these integrals differ from the corresponding integrals of the preceding
section only in the sign of the order of the Bessel function.
The reader will find some additional formulae concerning Lommel's functions in a
paper by Schafbeitlin, Berliner Sitzungsberichte, vm. (1909), pp. 6267.
Some very curious formulae have been obtained by Lommel, which connect
functions of the type TJV (w, z) with functions of the same type in which the
second variable is zero, provided that v is equal to \ or
f
When we write v -\ in 16-53 (5), we get
Now write
(W 4- z)2 e (w zf
2w ' 2w
and we find that
n ,4
U h
(w, z) iU% (w, z) = eT* (-) f exp { <n (1 - f )} d
+ eiz (-)*
\7T/ Jf-v{z*/(2to8)>
eX P ( 8i (-
1 ~ P)l d &
in the respective integrals, awci V"> a/8 are * oe interpreted by the conventions*
V "
. w+z
= ~77o\> v
v
/S>
= w z
V(2w) ' V(2w)
Hence we have
U (w, z) i U$ (w, *) = \e* k
(2a, 0) + %Ut (2<r, 0)}
iz
{
+ ie *{O i (2,0)tt7,(28,0)}
r
+ e fe () exp {+ ft (1 - )} <*
J
When we take o-f2 and Sf 2
as new variables in the last two integrals respectively,
these integrals are seen to cancel ; and so we have the two results combined
in the formula
(1) U k
(w, z) iU (w, z) - te**{Ui(2*. 0) iUt {2*, 0)}
,
+ **{17(2S,O)0 I (2M)},
and, as a corollary,
(2) U k
{g,s)xU% (s,s)-l*r*[Ui (4*,0)iU% (4M t O)}.
These formulae are due to Lommel, Miinchener Abh. xv. (1886), pp. 601 605; they are
reproduced by Walker, The Analytical Theory of Light (Cambridge, 1904), pp. 401402.
It is easy to see from 1653 (1) and (2) that, when R(v) > 0,
so that
( X)
ovJrv \ f
<2 "
_1
cos $ wt^ d ^^(w, 0) cos J w +
* J7 +1 (w, 0) sin | /,
(2) _1
\* v I t
2 " -1
sin (A wt2 ) eft = TJV (w, 0) sin $w - Uy+l (w, 0) cos \w.
2" i (v)Jo
see that
(3) ^oo.(i^A-ij;Q*oo.i.yV^()A
- [ Uk (2z, 0) cos z + U% (2z, 0) sin z]/</2
(4) ^^(I^A-l^^rinlA-l/V^)*
= [ 0"j (2s, 0) sinz-Ui (2s, 0) cos s]/V2
= i ~ [ ^* (2*, 0) cos * - V% (2s, 0) sin s]/V2.
<> ,, ( ,,)=
(^f _ ^. + _^__....}
1
i >
(8) W,--(^-L-
Tables of Fresnel's integrals were constructed by Gilbert, Mem. couronnees de
+ ...}.
I' Acad.
R. des Set. de BruxeUes, xxxi. (1863), pp 152, and Lindstedt, Ann. der Physik und Chemie,
(3) xvn. (1882), p. 720; and by Lommel in his second memoir.
(9) (V ()ft-2
JO
r S Jy+m+1 (z)
n=0
= N Z n+l J^ n (z)
n~o(vi-l)( + 'd)...(v+2n + l)'
(10) ('JAW" =o
2 (" + l)( + 3).,.( y + 2-l) jr
Jz zn +n (*).
(12) JV (*)cft-2in|#^2(--)^
t +l (i#)]
f
\
Jo
cos
(
(
V
.
at tb\ =
t/lP'
dt
,
Jo
f*
/
.
Bin [at
/
\
, b\
r) *
tjl
tdt
tdt
/ b\ dt r* / tdt
/ . .
, b\
r + *-irF
,
<> i>(-+Di!*-*J>Mr!.
f>(- + ) *|L-
j;* (- +
-.
( 2)
i
l
)*-/xtt+? ) i
~~x~Ji (^ + t )V(t-1)* 2
Now consider
1 f e^rcfr
,
27^^J ^ (^+T)^/(Ta -l)
where T is a contour consisting of the real axis and a large semicircle above
it, the real axis being indented at t =* 1.
The only pole of the integrand inside the contour is at id, and so
Thus we have
~ 6)
f* 08
/ . &\ <fo tre~ (
a+b [* sin (a; cos <ft
) . cos <f> d<f>
'
(3)
Jo
cos(a* + -) =-^ wl^J^i*).
Similarly it is found that*
f\
(4)
F Bn /(* + h-)_,
\ tdt * .
r<-6
g
>
2 _ r
TTt^J^z), .
.
Jo
and
/ b\ dt
(5) Pjr cos^ + -^=j7rsin(a + 6)-7r 2 (-^(JV^*),
fl -J r
1
gi, 5 , w,i..^f (0+) .__/'. - <y <*<
u. <, *>
=
s (-r (*)-"- J _ ^ ex P U
\ \
u)
t t
v+ *n+l "
When the contour is so chosen that it lies wholly outside the circle on
which 1 J
1
=\ w \
I
, we may change the order of summation and integration
and get
and so
ir i x
1 f
(0+)
(WO" (4 *\dt
/o\
(2)
m
F^*)-Ssi.. x
X f
( +)
^/w >"
iTW^ exp V2-2l)T'
/< *V*
and, in this integral, the points + iw lie outside the contour.
and hence
(4) V v (w, 0) = w
sinvTr f^exp^w-ie-iwtr
-
1 +w 2
du,
Jo
provided that R(i/)>0 and a is an acute angle (positive or negative) such
that
Ja + argw|< J-t.
16-58, 16*59] LOMMEL'S FUNCTIONS 549
Formula (4) was obtained by Lommelt from the formula of 16*53 (11) by a transforma-
tion of infinite integrals.
From (4) it is clear that, when v and w are positive, Vv (w, 0) has the same sign as, and
is numerically less than
sinyTr
C uv-\ c
e
-kuw du _ *
Jo -r(i- v ).(fror
A similar but less exact inequality was obtained by Lommel.
The reader will also observe that Vv (w, 0)/sin vir is a positive decreasing function of w
when w is positive.
oo exp ia *)
M F+2p-lg-JW^M l
-{/;
= 0(w-'-'),
when \w\ is large and, as in the similar analysis of 7*2, |
arg w |
< tt.
Hence
(i)
~i, r(i-,->:).(t
under consideration.
^ i
When v+ 1p and w are both positive, ( ) p Vv+2p (w, 0) has the same sign
as, and is numerically less than
sin vtt r ( )p
7" w +2P-lg-i'^ =
,
r(l-i/-2p).(|w;) +^'
lt
,'
7T Jo
so that the remainder after p terms in (1) has the same sign as, and is
* Mint. courorme de VAcad. R. des Sci. de Bruxelles, xxxi. (1863), pp. 1 52.
t MUnchener Abh. xv. (1886), pp. 582585.
550 THEORY OF BESSEL FUNCTIONS [CHAP. XVI
It is evident from 16*5 (6) that the corresponding formulae for Uv (w, z)
are
K {ex, jc) -- | /
-^ j cos (t+Jwr- }*-) sin j
J c* (1 - * )}
2 .
If, however, c<l, then %cx(l -t2 ) + (xt+ \vir \tr\ qua function of t,
has a maximum at 1/c; and hence, by the principle of stationary phase ( 8*2),
it follows that
2 nJ*+{( + n)*},
in which * u and the coefficients an are constants, is called a Kapteyn series.
Such series owe, their name to the fact that they were first systematically
investigated, qua functions of the complex variable z, by Kapteyn f in an
important memoir published in 1893. In this memoir Kapteyn examined the
question of the possibility of expanding an arbitrary analytic function into
such a series, and generally he endeavoured to put the theory of such series
into a position similar to that which was then occupied by Neumann series.
The semi-major axis, semi-minor axis, and the eccentricity of the ellipse
are denoted by a, b, and e. The axes of the ellipse are taken as coordinate
* It will, for the most part, be assumed that r is zero.
t Ann. tei. de VEcole norm. sup. (8) x. (1893), pp. 91 120.
J Hist, de VAcad. R. de$ Sci. de Berlin, xxv. (1769) [1770], pp. 204233. [Oeuvrei, in. (1869).
pp. 113138.]
Berliner Abh. 18167 [1819], pp. 4955.
il Electromagnetic Radiation (Cambridge, 1912).
552 THEORY OP BBSSEL FUNCTIONS [CHAP. XVII
axes, the direction of the axis of x being from the centre of the ellipse to the
centre of force. The taken as origin of polar coordinates,
centre of force is
the radius vector to the particle being r, and the true anomaly, namely the
angle between the radius vector and the axis of x, being w. The eccentric
anomaly, namely the eccentric angle of the particle on the ellipse, is denoted
by E. The time which has elapsed from an instant when the particle was at
the positive end of the major axis is called t.
The mean anomaly M is defined as the angle through which the radius
vector would turn in time t if the radius vector rotated uniformly in such a
way as to perform complete revolutions in the time it actually takes to perform
complete revolutions.
a(1 ~
(1) r =
1
6
+ e cos w
T = a (1 - coos E\ '
are deducible ; and an integrated form of the equations of motion (the ana-
lytical expression of Kepler's Second Law) supplies the equation
(4) M = E-e8mE.
Kepler's problem
is that of expressing the various coordinates r, w, E,
which determine the position of the particle f, in terms of the time t, that is,
effectively, in terms of M. It is of course supposed that the variables are real
and, since the motion is elliptic (or parabolic, as a limiting case), < et. 1.
The more complete solution given by Bessel depends on the fact that (4)
defines E as a continuous increasing function of M
such that the effect of
increasing M by 2-n- is E by 2ir.
to increase
In particular e sin E is an odd periodic function of M, and so, for all real
where An =
j
2 r
" / esin E sin nMdM
ttJo
=
_ 2
n7rj
ir dE-dM ,,,
p cos nil -* dM tvt
=
/'"
2
mr '
cos nM.dE
= - Jn (ne).
n
Hence it follows that
and this result gives the complete analytical solution of Kepler's problem con-
cerning the eccentric anomaly. The series on the right is a Kapteyn series which
converges rapidly when e < 1, and it is still convergent when e = 1 cf. 8*4, 8*42. ;
- =B + 2 Bn cos nM,
/*"
1
where B = TT- (* e cos) dM
Jo
= l + e*,
while, when n 0,
2 f*
Bn =
IT
i (1 e cos E) cos nM dM
Jo
_ r(i-.j)..if[- _2f- sin nJf rf(^g) dJf
+ wiry "o
L 7T Jo o <3
2c f*
= sin Esin (nE - ne sin ^) d-E
mr J
*A'(~X
so that
The expansion of the true anomaly is derived from the consideration that
w - M is an odd periodic function of M, and so
00
L
Jo
,[_ 8(,-JQc,Un -
nrr Jo
_>
nnJa
r-
.(*
\d>M
^ /
= --[o cosnJf da
-j-fidE .
= 2 V(l -e 2
) f cos (nff - nc sin E ) ,
W7r J 1 ecos.E'
This expression not such a simple transcendent as the coefficients A n
is
and Bn . The most effective method of evaluating it is due to Bessel*, who
used the expansion
V(l -e ) 2
1 COS i?
1 2fco8E
y + ^ cos2ff + 2/cos3ff
2f*
+...,
where /i
1 + V(l - )
On making the substitution, we find at once that
71
L =i J
detail the analysis, which the reader will easily construct for himself.
First, we have
a(l-(?)-r
r cos w = x ae =
so that
rcosw = 2
/i\ - fa3 e + ?. - r// %g
v
(1) i J
n (ne) cos
'
nM,
a w=1 n
and next
rsinw b "> v2 -J
r = V(l 2 r / \
/aN
(2) = -smJs .
vr
n (ne)smnM,
a a e n =i
Berliner Abh. 1824 [1826], p. 42.
t A$tr.
Nneh. era. (1884), col. 17 28. Various expansions bad also been given by Plana, Mem.
delta R. Acead. delle Sei. di Torino, (2) x. (1849), pp. 249332. In connexion with their convergence,
see Cauchy, Compte$ Rendu*, xviu. (1844), pp. 625643. [Oeuvres, (1) vin. (1893), pp. 168188.]
17-21, 17*22] KAPTEYN SERIES 565
while
(3) cos
E ae = - \e + 2
i
- /'(ne) cos nM.
n
Next, if m is any positive integer*,
(4) cos mE = m 2 -
n
{/_ (ne) t7"w+OT (ne)} cos nM,
*=i
" 1
(5) sin mE m 2 - {/"_ (ne) + /+ (ne)} sin nM.
a n
The expansion of a/r is particularly simple, namely,
The reader will observe that, in the case of the expansions of even functions
of M, the results simplify when we take the particle to be at one of the ends
of the major axis, because then the three anomalies are all equal to or to ir,
n=i n n=i n
(2)
5- = 2
1 ~ *
I* n=l
Jn (ne), i^-S
+ 1
(-)-> Jn
e n=i
(ne),
(3)
(T
i^ - jr.' (ne), ^L = J (-)- J.' (ne).
* It is seen from (3) that, when m is equal to 1, the expansion (4) has to be modified by the
insertion of a constant term. These two formulae were given by Jaoobi, Attr. Naeh. xxvni. (1849),
col. 69. [Gen. Math. Werke, tu. (1891), p. 149.]
556 THEORY OF BESSEL FUNCTIONS [CHAP. XVn
More generally we find by differentiating 17*21 (6) that
\m r (Ism ~j oo
L
( J
< 4)
2 lM^TZ7^\ M-_rnl nimJ {ne) >
(_)-i r d* m
r d? i x
(5) __^ ~i
= 2 ( - )"- 1
n J (we).
1
^_ cosE dE
de\isin E (1 - e cos E)) ~~
sin 8 E (1 e cos 2?) 87
#(1 -ecosE?
w
cos Z?
" ^
4- c
e sin
x "
E
" =-
sin 911 9e
_ esini?
(1 -ecos^)3
_a_ i
=
9ilf 1 -ecos^'
so that, by 17'21 (6)
d { 1 )
(6)
v
'
-. = - - = - 2
=rp:
sm.(l-ecos2?) sinif
1 1
=- %
n sin nM .
['
j
Jo
Jn (nx) doc.
=i
The last two expansions do not appear to have been published previously.
Expressions resembling those on the right of (6) and (7) have occurred in the researches
of Schott, Electromagnetic Radiation (Cambridge, 1912) passim.
Thus, as cases of (4) and (5), Schott proved (ibid. p. 110) that
= ^lL^4j J^JV^Sii*)**.-^^.
(8)
JiW2j^ (2ne)
(1)
9 Z JmVne) e
+ (!*
*?
n ~i n* +? l"+f + ?)(* + )
<S\
V
2 g f K2-1)) ^
f
e'f
' nti (2-l^ + f l + "(l + f)(8 + p)
+ ft +
have been stated by Meissel * who deduced various consequences from them
it is to be supposed at present f that < e $ 1, and f; is real.
The simplest method of procedure to adopt in establishing these ex-
pansions is to take the Fourier series J
cos2yiif _ 7r cosh (ir 1M) 1
~ 2|"a
wti n +
'
2
* 2sinh7r
(which is valid when ^ M^ it), replace MbyEe sin E, and integrate from
to ir. It is thus found that
,c r '
by a formula due to Cauchy||; and the truth of Meissel's first formula is now
evident.
The second formula follows in like manner from the Fourier series
| cos (2n
= -1)M tt sinh (|tt- M)%
w=l (2n-l)*+f ~ 4fcosh^ '
-i ,ro '
n% (2n-l)
are polynomials* ine; the former is an even polynomial of degree 2m, and
The values of the former polynomial were given by Meissel in the cases to 1, 2, 3, 4, 5
the values for to 1, 2, 3 are
2 */ (2ne),
2 6J_1 K2n-l)},
* It is to be noted that the coefficients of c** and c**-1 in the respective polynomials are not
sero; they are
t-^ and
<- r" .
2.(m!) 2.1.8...(2m-l)
17-3] KAPTEYN SERIES 559
and the domain is smaller than the former thus, when the limit is 1,
latter ;
the first domain is the interior of the curve shewn in Fig. 24 of 8*7, in
which the longest diameter joins the points i 1, while the shortest joins
the points itx 0*6627434 while the second domain* is only the interior of
;
Hence, when we are dealing with Kapteyn series, if we use the method of
expansion into double series we succeed, at best, in proving theorems only for
a portion of the domain of their validity; and the proof for the remainder of
the domain either has to take the form of an appeal to the theory- of analytic
continuation or else it has to be effected by a completely different method.
(1) , = 1 + 2 1 Jn (nz),
provided that z lies in the open domain in which
z exp V(l -
<1.
l + vXl-**)
This domain occurs so frequently in the following analysis that it is con-
venient to describeit as the domain K; it is the interior of the curve shewn
in Fig. 24 of 87.
1 + 2 2 Jn (nz) = 8(z),
and then it has to be proved that 8(z) = 1/(1 - z).
Since J^^f^^pmj^
* For an investigation of the magnitude of this domain, see Paiieoz, Journal it Math. ziv.
(1849), pp. 3389, 942246.
t Nieuw Areme/voor Wiikunde, xx. (1898), pp. 128126; Ann. $d. de YficoU norm, tup, (8)
z. (1898), pp. 96102.
560 THEORY OF BESSEL FUNCTIONS [CHAP. XVII
we see that, if we can find a circle T with centre at the origin of such a radius
that on it the inequality
(2)
I *P {*(- 1/Pi L !
is true, then
+ r'exp{jjr (-!/<)
W
/ox /.v
V
'
= J_
2wt
( l
*
*
where p, u, a, are all real (/> and m being positive), then (2) is satisfied for
all values of if
a
p V(sinh u + sin 8
a) -u< ;
and when u is chosen so that the last expression on the left has its least
value, this value is ( 87)
g
1+VU-* 8
) '
which is negative when z lies in the domain K. Hence, when z lies in the
domain K, we can find a positive value of u such that the inequality (2) is
satisfied when 1
t |
= eu .
+ * exp {-*(*-
W
(4i\
w
S(z>>=J-[
2iri J (y+) 1
1
-*exp{-J (*-
1/0}
1/*)}
<ft
t
'
28 (si = w
t
+ xp { *(- 1/0} d*
and so 2S(z) is the sum of the residues of the integrand at its poles which lie
inside the aunulus bounded by V and 7.
We next prove that there only one pole inside the annvlus*, and, having
is
proved this, we notice that this pole is obviously t = \.
[
2m'J (r+>y _ )
<* log [l-rexp{fr (*-!/*))]
dt
dt
Lf <nog[l-rexp{fr(t-l/t)} ]
dt
+
W
2inJ (r+)
27riJ(r+ )
{
dlog[l-texp{-lz(t-l/t)})
dt
dt
dt
= J. T d\og[l-tr>ex V \$z(t-l/t)}-\
mJ (fjj.) dt
dt
dlos[-t^{-(t-i/t)}]
+-L.f dt
liri J (T+)
i
T+ ) dt
* The corresponding part of Kapteyn's
investigation does not seem to be quite so convincing
as the investigation given in the text.
17-31] KAPTEYN SERIES 561
then |
U <1 |
on T, and so the expression under consideration may be written
in the form
) dU
ti J r +) U=o
< j
J
dt
and the integral of each term of the uniformly convergent series involved is
zero.
>W
2iriJir+ NK)>-
It follows that 28 (z) is equal to the residue of
t + exy{$z(t-l/t)}
t-exp{fr(t-Lft)\
at t 1 ; and this residue is easily calculated to be 2/(1 z).
It has therefore been shewn that S (z) is equal the to 1/(1 z) throughout
domain K, i.e. throughout the whole of the open domain in which the series
defining S(z) is convergent.
[Note. It is possible to prove that S () converges to the sum 1/(1 - z) on the boundary
of K, except at = 1, but the proof requires an appeal to be made to theorems of an Abelian
type; cf. 17'8.]
+ f!
(i+^)(3+r)(5'+r)
_ +
which are valid when z lies in the domain and is a complex variable K
which is unrestricted apart from the obvious condition that ft must not be an
integer in (1) nor an odd integer in (2). These results are the obvious
extensions of Meissel's formulae of 17*23.
<f)
= yjr z sin ^r.
The singularities of yfr, qua function of <f>,
are given by cos^r =* \jz, that is
<f>
= arc sec z VC^ !) 2
None of these values of < is real* if z lies in the domain K ; and, as <f>
in-
creases from to oo through real values, ifr describes an undulating curve
which can be reconciled with the real axis in the ^r-plane without passing
over any singular points.
<nog(l-2ffcos<fr +78) ,
(r+) dt
_lf
+ 2irij
dlogjPexpi-zjt-l/t)}]
^
(r+> dt
= _!.(
/nj<r+) L=o
\% Un co6(n + l)<^\~dt-r2
a *
J
= 2,
* It is easy to shew that such values of <f>
satisfy the equation
=-=** <1.
so that ! |
17-31] KAPTEYN SERIES 563
the integral of each term of the uniformly convergent series vanishing, just
as in 173.
(3)
K
=
r = 1 + 22 Jn (nz)coBn<f>,
' 1^COS^r H=1
w
(4)
Jo
,
f r*i^'
l
"^-J+2i^
+ f =i
)
l
where the path of integration is the undulatory curve in the -^-plane which
corresponds to the real axis in the ^-plane ; and, by Cauchy's theorem, this
undulatory curve may be reconciled with the real axis.
Now, when the path of integration is the real axis, the integral on the left
in (4) is an integral function of z\ and this function may be expanded in the
form
m =o ml Jo
By changing the sign of z throughout the work we infer the two formulae
v '
=i4n+f m=1 (2m)! J
-i (fr-lf+p
which are now established on the hypotheses that z lies in the domain K and
that (?)><>.
By dividing the paths of integration into the intervals (0, ir), (w, 27r), ...
Jo
e~** sin"* ^frcM 1
= . .
sinh$7r./o
t I cosh 0 cos"* Odd
.
1
564 THEORY OF BESSEL FUNCTIONS [CHAP. XVII
and that
all other values of has already been explained. The required generalisations
of Meissel's expansions are therefore completely established.
In the case of an even power, zm we take the equation given by 17*31 (1)
,
in the form
2J,f|-. + f 1
(+)
m
When $ w, there are no poles outside the contour, and so the contour may
be deformed into an infinitely great circle, and the expression is seen to be
equal to unity; but when m > n, the poles im are outside the circle and the
expression is equal to unity minus the sum of the residues of the integrand
at these two poles, i.e. to
(m + w)!
m+r (m n
.
1)!'
Next we evaluate
r(n + l+it)T(n + l-iZ) dl;
27rtj| _ n+i
(
r (m + 1 %o r ( + i - #) *-*+*
17*32] KAPTEYN SERIES 565
When m^ri, the origin is the only pole of the integrand, and, if we take the
contour to be an infinitely great circle, the expression is seen to be equal to 1.
Hence we have
(3)'
V
** = 2n* 2 (w + n- !)./ (2m* )
mZ n m^.(m-M)! '
X dZ
Zi ? + (2m-iy
H3 +r2 }-..{(2n-i)
2 2 2 2
{i +-r +?}
-aisL
ATI I J
|f|=2W
x
Ji {i +r U3 +n--K2m-i) +r
8 a 2 a i
}
^
and we deduce in a similar manner that
The formulae (3) and (6) may be combined into the single formula
(7)
jn t ( + 2mr.m! '
which is obviously valid throughout the domain K when n has any of the
values 1, 2, 3,
This formula was discovered by Kapteyn*; the proof of it which has
just been given, though
somewhat artificial, seems rather less so than Kapteyn's
proof.
* Ann. sci. de VEcole norm. sup. (3) x. (1893), p. 103.
566 THEORY OF BESSEL FUNCTIONS [CHAP. XVII
which were proved for real variables in 17*22 and for complex variables in
17'3 ; it is, of course, supposed that, if z is real, then 1 < z< 1, and, if z is
complex, then z lies in the domain K.
The induction which will be used depends on the fact that when the sum,
ao
f(z), of the Kapteyn series 2 am Jm (mz) is known, then the sum F(z) of the
M=1
series
r=l
% m WV
W&8
- can be obtained by two quadratures, if the former series
* T^ + Z
d
^ =
}- 1
CLm [Z
*
Jm" {mz) + (z/m) Jm
= (1-* ) 2 2
'
{mz)]
a m Jm (mz),
m=1
so that (z)F(z) = (l-z')f(z);
it follows at once that F(z) can be determined in terms off(z) by quadratures.
Now, from (1), we have
I J^(%nz) = ^- 2 ,
A=B = 0.
Consequently
jP-2 I J"< 2w *)
(2)
m=i rn*
sn = n 2 2 bmn Jm (mz),
and consider the function (s) defined by the equation
<f>
(z) = (n + 2f X "HL-JL bm>n Jm (mz).
m-i fnr
-* >*"
=< +2
>^S+*SS-< b+2
,
>'< i
= (n + 2) s
*n+a.
*+* = ( W + 2) 1 a
6WjB+a JTO (mz),
where V n+a = * bm , n-
and so fan' i 7=
=i (2wi)^-
:rn=n
1
T (m -n +TT\
.?n a
1)
That is to say
=2 r(m + n)J(2iiu)
2n a
J
ntnm^rim-n + l)'
and this is equation (3) of 17'32. The expansion of zm is obtained
~ 1
in the
same way from the expansion of z the analysis in this case is left
;
to the
reader.
568 THEORY OF BESSEL FUNCTIONS [CHAP. XVII
f 4) <**) = -
(K + 2OTr,, m ,
>
which is the expansion obtained by other methods in 1732; and the ex-
pansion is valid throughout the domain K.
n+1
mZo(n + 2m) .m\
is absolutely convergent (being comparable with % l/m ),a
the expansion (4)
converges uniformly throughout K and its boundary. The expansion is there-
fore valid (from considerations of continuity) on the boundary of K, and in
particular at the points z = 1, as well as throughout the domain K.
Assuming that 1
1 \
> j
z |, we have
_1__1
+ JL =!+ . | ?!!i
2
| r (n + to) Jn+zn {(n + 2m) z\
t-z n+1 t ^t t nZ i *
w+1
m- (n + 2m) n+1 .ml
z exp V(l - g 2 )
Now, if
j
= jr
1+V(1-* ) 2
'
is convergent. But the terms in this series are less than the terms of the
double series
2*rw+2m _ 2Fex pF 2
* *
.-i*-om!|r ~|*Ki*l-2iT i
(1) r
t
= S <& {t)Jn(nz),
Z <,(*) + 2 n i n
* 4nn. sci. de VAcole norm. sup. (3) x. (1893), pp. 113120.
17-34, 17*35] KAPTEYN SERIES 569
where *
(2) *.(*) -1/fc
(3) < n (0 2
1 9
-
w (Mn_2m+1
,
From the last formula we may deduce a very remarkable theorem discovered
by Kapteyn ; we have
K w-m-1)! _ 1 <* (rt-2m)2 .(w-m-l)
it - S* (
!
| !
The expansion (1) is therefore valid throughout the domains in which both
of the inequalities (5) are satisfied.
[Note. This result gives a somewhat more extensive domain of values of t than was
contemplated by Kapteyn he ignored the theorem proved in 9*17, and observed that
;
(since the coefficients in the series for n (t) are positive) when t ^ 1, j |
|MOI<*.(l'l)<(i)=i.
by (4) ; so that Kapteyn proved that (1) is valid when
0(*)<0(1),. |*|^1.]
Now that explicit expressions have been obtained for the coefficients in the expansion
L
Z
=<fl (() +2 | n (t) Jn (nz\
t n=l
it is possible to verify this expansion in various ways. Thus, if < (t) be defined as
n (1 - *2 ) n (nt) +sin \mr
2 +t cos2 \nn,
the reader will find it an interesting analysis to take the series
7
^S
+ T1 S* +
2 (
1 -2) * nOn (nt)Jn (nz),
t n =i
W. B. F. 19
570 THEORY OF BESSEL FUNCTIONS [CHAP. XVII
substitute suitable integrals for the Bessel coefficients and Neumann polynomials, and
reduce the result to l/(tz) after the manner of 9'14.
and then, dividing by 1 z\ and making use of 17'3 (1), we find that
2
(< -l)(<-2)3 (P-\?{t-Z? (?-\f(t-zf
= I e nW (nz)\ (t)
C+ft'+l)'*** * (*+!) corf jnr )
whence the differential equation for n (t) is easily constructed in the form
(<) (1 - <2 ) 0 () + sin 2 \nir + < cos2 \nn +t~! {A n Jn (nt)+Bn Yn (nt)},
where A n and 2?n are independent of t ; but it does not seem easy to prove that A n Bn =0.
(2)
*.-%z JG.V)f(t) *,
and the path of integration is the curve on which fi () = a.
This result is obvious when we substitute the uniformly convergent
expansion
(<) +2
nl
< n (t)Jn (nz)
since D, (t) =a on the contour, while both fl(s) < 1 and D, (z) < fl (*) when s
is inside the contour.
/(*)- 2 an zn ,
n=0
then
(3) Oo=Oo,
% am Jr+m{(v + m)z\ i
(1)
M-'L i.Zw.mS"-*'***-
throughout the domain K.
It is easy enough to establish
expansion* when \z\< 0*6627434 but
this ;
From this result, we can prove that, under the conditions specified in 1 7*4,
And the reader will easily prove that if/ (2) satisfies the conditions specified
in 17-4, then
(4) *"/(*) = Sv
n=0
J*+n {(v + n)z],
where
W
(fo n '
"
=1
2 m
X*
t
(v
(%v + n)+-" m
to
. !
where a , a lf ... are the coefficients in the Maclaurin series for f{z).
[2sote. Jacobi in one of his later papers, Aatr. Nach. xxvm. (1849), col. 257 270 [Ges.
Math. Werke, vn. (1891), pp. 175 188] has criticised Carlini for stating that certain
expansions are valid only when \z\ <0'663.... But Carlini had some excuse for his state-
ment because the expansions are obtained by rearrangements of repeated series which are
permissible only in this domain, although the expansions are actually valid throughout the
domain K.~\
have been studied in some detail by Nielsen f. But the only series of this
type which have, as yet, proved to be of practical importance J, are some
special series with fi v, and with simple coefficients. The results required in
the applications just specified are obtainable by integrating Meissel's expansion
of 1731 (1) after replacing z by z sin#. It is thus found that, throughout
the domain K,
2 sin ""'
n=i ?+? "=i(i +r)(2 2 +H...( +r )-Wo
2 2 2
so that
m
yy) g I JnHnz}
nti n*+?
I
21 2 +
z*
t
2 "
1
1.3
"2.4 ,(P ,
*g
+ ^)(2 + t ) 2 2
tm is a polynomial in z2 of degree m;
and hence we deduce that 'S,Jn 2 (nz)/iv
while the sum of series of the typel n^n Jn2 (nz) may be found t in a similar
manner from the corresponding expansion 2n2m J 2n (2n2).
* Cf. Nielsen, Ann. sci. de VEcole norm. sup. (3) xvin. (1901), p. 60.
t Ann. sci. de VEcole norm. sup. (3) xvm. (1901), pp. 3975.
(2)
l/^-sTobr^
(3) j^M-^,.
A general theory resembling that of 16 14 -
is deducible from the ex-
pansion
If lim |
y/an \
= 1,
n-oo
we have seen that the Kapteyn series %an Jn (nz) represents an analytic
function throughout the domain K. is real, \Jn (nx)\<But since, when x 1,
the series may converge along the whole of the real
although when \z\ > axis, 1
the series does not converge at points which are not on the real axis.
Now, when x > 1, < decreases as -\jr increases from to arccos(l/#) and
then increases to ir as yjr increases from arc cos(l/#) to it. If m be the integer
such that the minimum value of < lies between 2mir and 2 (m -f 1) it, let
the values of yjr corresponding to the values
of <f>
be 7, 7j, ... ym , 8 m> &m-i, $i> S ,
and then
Now when yjr lies in the intervals (7,., 7r+ i) and (8 r , 8r+1 ) the sum of the
series under the integral sign is
2
< - %Tr<f> + %ir* + r (r + 1) tt + (r + 1) 2
ir<f>,
and, since
v f
*expV(l- .s
8
)
)*
6
and that of two similar series f with *Jn*, \/n written for >Jn.
Again,since
J^\{v + )x)
/+ (v + n)
where jlt jif js , ... denote the positive zeros of J (z) arranged in ascending
order of magnitude.
The necessity of expanding an arbitrary function in this manner arises
also in Daniel Bernoulli'sproblem of a chain oscillating under gravity and in
Euler's problem of the vibrations of a circular membrane with an initial arbi-
trary symmetrical displacement ( 1'3, 1*5).
xJ ' +n,
{jm x)J (jn x)dx=\
J
Jo KtJi(jm), m = n,
and hence Fourier inferred that
denote the positive zeros of the function J (z), arranged in ascending order
of magnitude, then
t The omission of the suffix y, associated with jj , jj, js, ..., should cause no confusion, and it
Dini, however, remarked that he was unable to deal with the range
- \<v< i,and limited himself to the ranges \. Several subsequent writers,
while proving theorems for the latter range, asserted that the extension to the
former range was merely a matter of detail; but it was not until after 1922 that
anyone took the trouble to supply the detail which is tedious and of no great
interest.In the exposition given here, it will be supposed that v> \.
The first attempt at a rigorous proof of the expansions (1) and (3) i&
contained in some notes compiled by Hankel* in 169 and published post-
humously. A
more complete investigation was given by Schlaflif a year
and an important paper by Harnack*
after the publication of Hankel's work;
contains an investigation of the expansion (3) by methods which differed
appreciably from those of earlier writers.
A few years after the appearance of the researches of Hankel and Schlafli.
the more general expansion
00
formally vanish at x 1.
% Leipziger Berichte, xxxix. (1887), pp. 191214; Math. Ann. xxxv. (1889), pp. 4162.
Serie di Fourier (Pisa, 1880), pp. 190277.
578 THEORY OF BESSEL FUNCTIONS [CHAP. XVIII
It was pointed out by Dini that the expansion (5) must be modified* by
the insertion of an initial term when H+v= ; and, although Dini's analysis
contains a numerical error, this discovery seems to make it advisable to
associate Dini's name rather than Fourier's with the expansion.
The researches which have now been described depend ultimately on a set
of lemmas which are proved by Cauchy's theory of residues. The use of com-
plex variables has, however, been abandoned, so far as possible, by Kneser f
and HobsonJ, who have constructed the expansion by using the theory of
integral equations as a basis.
* Details of necessary modifications when H + v <0 will be given in 18-3. The modification
was also noticed by Kirchhoff, Berliner Sitzungsberichte, 1883, pp. 519 524.
t Archiv derMath.undPkys. (3)vn. (1903), pp. 123 133; Math. Ann: lxiii. (1907), pp. 477524-
% Proc. London Math. Soc. (2) vn. (1909), pp. 359388.
Ibid. (2) xvm. (1920), pp. 163200.
Ibid. (2) iv. (1906), pp. 396430. Cf. 19 21 19 24.
||
In the exposition which will be given in this chapter, the methods of the
calculus of residues will be used to a far greater extent than has been usual
in recent researches; this is a reversion to the practice of Hankel and
Schlafli and (in the special case of Fourier series) of Cauchy. The advantage
of this procedure is that it results in a great simplification in the general
appearance of the analysis throughout the whole theory. And, although it
seems impracticable to prove certain theorems (notably those* relating to
fractional orders of summability) with the help of complex variables, the gain
in simplicity is so marked that it has been possible to include in this chapter
very many more theorems than would have been possible if the methods
of the theory of functions of real variables had been used more exclusively.
Wiener Sitzungsberichte, lxxxviii. (2) (1884), pp. 975 1003 Alexander, Trans. Edinburgh ;
Royal Soc. xxxm. (1888), pp. 313320; Sheppard, Quarterly Journal, xxm. (1889),
pp. 223260; Volterra, Ann. di Mat. (2) xxv. (1897), p. 145; Stephenson, Phil. Mug. (6)
xiv. (1907), pp. 547549; Messenger, xxxm. (1904), pp. 7077, 178182; Rutgers,
Nieuw Archie/, (2) vin. (1909), pp. 375380; Orr, Proc. R. Irish Acad. xxvn. A, (1910),
pp. 233248; and Dinnik, Kief Polyt. Inst. {Engineering Section), 1911, no. 1, pp. 8385.
[Jahrbuch iiber die Fortschritte der Math. 1911, p. 492.]
The investigations by Alexander are mainly based on operational methods, while Orr
dealt with expansions in which functions of the second kind are involved.
am = If' I
f(t) cos mtdt, bm =-If". f(t) sin mtdt.
7T J v IT JJ -.jf
2 am J v (jm x),
m=l
in which the coefficients am merely form a given sequence of constants, will
be called a series of Bessel functions.
If, however, the coefficients in this series are expressible by the formula
1
2 f
a = 75 TT~\ f (*) J "
f ( J ^ dt >
v v+iKjm) J
2 bm Jv (\ m x) t
there are certain distinctions between the two expansions which make this
view somewhat misleading (cf. 18*26, 18*34, 18"35).
* The greater part of the terminology is due to Young, Proc. London. Math. Soc. (2) xvm.
(1920), pp. 167168.
t It is supposed that the integral is convergent for all positive integral values of m.
J It is supposed that the series is modified, as in 18*34, when H + v ^ 0.
18-12, 18*2] FOURIER-BESSEL SERIES 581
investigated in 1542. Another is x', which gives rise to the formal expansions
2Jv{j x
(i) x"= ? \ ,
m=l (*-m
S
- V*) J v - {X m ) + A. m /' (Xm)
2 2
It will be seen subsequently that (1) is valid when ^x < 1, and (2) when
0$a;<l, if H
+ v>0. Cf. 1822, 18-35.
t
\n2
J
1
r + 2 + Jv (X m <) dt =
i
(v + 2n) Jv (X m )-Xm Jv '
(X m ) - 4 (+n)
J
V + *-> J"
r (X m ) rf<
positive integer. The Dini expansion of x v+2n + l may similarly be determined; in this
case the general coefficient is expressible in terms of known functions and
"i
t
v Jv (\ m t)dt.
/:
In order to calculate this when v is an integer, McMahon* has proposed to tabulate the
function
K
('
| J (t) dt=*Jx (x) + Js {x) + Jb (.r) + ...,
which is a special form of one of Lommel's functions of two variables ( 16*5, 16*56).
/" ir
2\ n*J (\ m x) J {\n t) v
(\\ rn /. x
_ v x
*"* v \.lrn ) Jy \Jmt)
K**) 4 n [l,X)- ~ -j-2 -r^p- ,
v + \> 0.
The method which will be used is due to Hankel* and Schlaflif, though
many of the details of the analysis are suggested by Young's J recent memoir.
In order to obtain the formulae connected with Tn (t, x) which are sub-
sequently required it is necessary to express the mth term of the sum for
Tn (t, x) as the residue at
jm of a function, of the complex variable w,
which has poles at jlt j2 js ...jn * , , When this has been done, we express
* Math. Ann. vm. (1875), pp. 471494.
Tn (t, x) as the integral of this function round a rectangle of which one of the
sides lies along the imaginary axis while the opposite side passes between
jn and jn+1 . The sides parallel to the real axis are then moved off to infinity
in opposite directions, so that, in order to secure the convergence of the integral,
it is necessary to prescribe the behaviour of the integrand as j
I(w) |
-* oo .
The first of these was the integrand studied by Schlafli; the other two are
suggested by the work of Kneser and Carslaw which was described in 15 42.
A study of the asymptotic values of these integrands indicates that (2) is
suitable for discussions in which xt and <x+t< 2 ; (3) when* < t< x < 1
and (4) when ^x< t < 1.
We proceed to verify that the integrands all have the same residue, namely
2
2/ ( jm x) Jy (jm t)jJ y +1 (jm ),
at tv=jm . Tn the case of (2), we define the function f g(w) by the formula
cient of 0* on the right vanishes; and hence the residue of g (w)/{wJv2 (w)}
at jm is
9'(jm)l{jm Jy*(jm)},
and this is easily reduced to
by 3*63, and this is the expression required ; the integrand (4) is dealt with
in the same way.
* This is most easily seen by writing the integrand in the form
\iriw {HyV) (w) HyV) (xw) - HvP) (xw) HrW (w) } Jv (tw)\o v (w).
f The results obtainable by using the integrand (2) are discussed in great detail by Graf and
Gubler, Einleitung in die Theorie der BesseVschen Funktionen, i. (Bern, 1898), pp. 131 139.
'
Now it is easy to verify that the three integrands are odd functions of w,
and so the three integrals along the left sides of the rectangles vanish*.
Again, if w = u + iv, it may be verified that when v is large, and either
positive or negative, while w>0, then the three integrands are respectively
(e-fe-s-t'M), (e-<*-'>l t'i), -x) W),
(e~ (t
and so, for any assigned value of An , the integrals along the upper and lower
sides of the rectangle tend to zero as B -* oo when x and t have the relative
values which have already been specified.
(0<t<x< 1)
An+CCt
(8) Tn (t, x) =I f w {J, (w) F (tw) - J (tw) Yv (,)} Appp
J
m
(0<x<t<l)
From equation (6) it is easy to obtain an upper bound for Tn (t, x) j; for j
tive (or zero) and bounded, there exist positive constants cx and c2 such that
|<*-<>r.ft.)|
;s? ^J_p{-<S-.->!.||*
i
so that
**'
(10) \Tn (t,x)\^ '
ttc 2
2
1
1* - #* 1
(2 - x - t) *J(xt)
This inequality gives the upper bound in question.
* It is necessary to make an indentation at the origin, but the integral round the indentation
tends to zero with the radius of the indentation.
:
f t"
+l
Tn (t,a;)(t*-a?)dt
Jo
fv+l rAn+aai J
= "^7 1* J" (xw "> / "+ 2
(^ ~ xJ >- ( tw ) J*+i ( aw)} ,.>-/,, ,
7n J An-ooi V)J,-(W)x
and hence
(11) f t"
+1 Tn (t, x) (JP -x*)dt\$ x t) *J{xt)
Ijo I
irc^A n (2
constant Cx in equation (9) is independent of t in the interval 0<t < 1. Thus (11) may be
written
valid when 0^#^1, 0^i<l. This extension is not so easily effected when integrals
involving functions of the second kind have been used because the simplest inequality
corresponding to (9) is
\
ry (tw)\^ Cl '
{\tw\-"\oS \tw\ + \tw\-i}exio{\ l(tw)\},
and it is a somewhat tedious matter to obtain a simple upper bound to the integrand in
(8) from this inequality.]
Equation (6) was used by Schlafli to prove that, when n is large, then
^n(g-a?) sin
_ *A n (t + x)l
|""
l
TnV(tx\~
' 2v
'
(<rf)|_wniw(*-)
f
sm$-jr(t + oc)j'
but, since the order of magnitude of the error in this approximation is not
evident, we shall next evaluate some integrals involving Tn (t, x) by means of
which difficulties caused by the unknown error may be evaded.
(3) lim
f\''+
1
Tn (t,x)dt = lx. (0<a;<l)
In the course of proving (1) it will be apparent that
a* [ t'
+1 Tn (t, w) dt -* x* + l
Jo
uniformly as n -* oo when x lies in the interval
^x^ 1 A,
where A is any positive number.
'
r t' +1 Tn (t,x)dt
Jo
in the interval in which < t ^ 1.
Of these was given by Young, Proc. London Math. Soc. (2) xvin. (1920),
results, (1)
pp. 173 174, and the proof of it, which will now be given, is his. Formula (2) seems to be
new, though it is contained implicitly in Hobson's memoir.
It is evident that
Since the integrand is an odd function of w, the value of the first integral
reduces to iri times the residue of the integrand at the origin, so that
fW
/:
(tx)dt=x--- f^ 2 ^(>^
Now
4ci
c2 A n (1 x) six
and, from this result, (1) is evident; it is also evident that
a* f t>
+1
Tn (t,x)dt-x + l
Jo
tends uniformly to zero as n -* oo so long as ^#^ 1 A.
It will be observed that the important expansion
2J - (
(4 **+-* I y> (0<tf<l)
m=\ JmJ v+i \Jm)
which was formally obtained in 1812, is an immediate consequence of (1).
Formula (2) can be proved in a somewhat similar manner (though the
details of the proof are rather more elaborate) by using an integrand involving
functions of the second kind. It is easy to see that
=1 +l
Jm*f v+i\Jm)
u~*)
= hm -^ {
J (w) F (xw) - Jy (xw) Y (to)} V, . .
A . ~ '
is large (7-21). The error produced thereby in the integrand is, at mosti
0(l/w*) when 0<#<1 ; and, as n ^oo, we have
J Jn-ooi v? i An \n)
Now the result of substituting these dominant terms is
A * +m s iD w 0-- x
- lim [ .
) sin (** ~ - i*") 7 \
= lim
b-*. 2iriJ
.
[
An+m C0S w ~ OTr ~
H -Bi
( * I*) cos ( 2*w -w-\inr-\ir)
w cos (w vir ^tt)
-*f- lim I*'*"
008(2^-^-^^-^)
B-^aaJ^-Bi wcos(tt/-w tt)
We shall have to discuss, almost immediately, several integrals of this general
type; so it is convenient at this stage to prove a lemma concerning their
boundedness as n -*- oo
f^** (* -*-* *)
lim
B-**>JA
9-+-<*>
J n -BiV>coa(v>-$vv\ir)
j.
2;Ucos(A-l)A.
L r "%'*
tA
/o 04*+^) cosh v
-sin(A-lM n
v
'jof"J2^* .
(A *+v*) cosh n
1
_|
When 1 < X < 1, the modulus of this does not exceed
+1
Tn (t, x)dt = \x* +
f o
t" (1/n),
I t
+1 Tn (t,x)dt,
Jn
and we shall prove that, when < t 1 and <x% 1, to integral is a bounded
function ofn,x and t, as n -* oo .
It is easy to shew by the methods which have just been used that, when
1 x + 1 ^ 1, i.e. when t^x, then
f f +1 Tn (t,x)dt
Jo
%t +1 J v (jm x)J v+l (jm t)
= I
- im
l
B-*-<*>
f^V. <>*<>- A (~)r.<))^*
*l J A n -Bi '>v\'")
These integrals are of the type examined in the Lemma given earlier in this
section ; and sO the original integral is bounded when l<#+ 1^1
and 1 < 1 + x^l, i.e. when < t^. x ^ 1.
To prove that the integral is bounded when 0<x^t^l,vre first shew that
f
t
v+1
Tn (t,x)dt
Jo
= hm -^t- (J (w) Fh-i (tw) - J v+l (tw) Yv (w)\ \ ,
and then apply the arguments just used in order to approximate to the
integral on the right the details of the analysis are left to the reader.
;
f**f(t)dt
J a
are sufficient to ensure that, as n -*- oo
6
[ tf(t)Tn (t,x)dt = o(l),
wherej 0<#<1.
The reader will observe that this theorem asserts that the only part of the
path of integration in
F tf(t)Tn {t,x)dt
Jo
which is of any significance, as n -* oo , is the part in the immediate vicinity of
the point x.
that t*/(0 is bounded and that the origin is not an end point
of (a, b). In
the second stage we remove the restriction of boundedness, and in the third
stage we remove the restriction concerning the origin.
so that |
a> m (t) $
|
Um - Lm in (tm_ 1 , t m ).
It is then evident that
+ 1 /
t^Tn (t,x)(t*-a*) a)m (t)dt 1
Sc^Kp
\\f{t)Tn {t,x)dt x b) \/x
J a TTC<?A n (2
4c 2 p fm
(2 - - 6) V#m=l
2
7rc 2 a; Jtm -i
that is to say
Now the choice of e fixes p ; when e (and therefore ^>) has been chosen, we
are at liberty to choose An so large that A n > 2Kp/e. That is to say, by a
suitable choice of A n we may make
,
the integral on the left less than
2
8 Cl 6
2
7TC2 (2 X b) *Jx
which is arbitrarily small.Consequently the integral iso(l)as^l n -*- oo , and
this is the theorem to be proved.
(II) When F (t) is not bounded throughout (a, b), let it be possible to
choose r intervals fi, such that F(t) is bounded outside these intervals and
such that
2 ! \F(t)\dt<e.
the expression on the left) arbitrarily small, and this is the result which had
to be proved.
rb
(III) If I
&f{t)dt exists and is absolutely convergent, we can choose i
Jo
so small that
8d2 \r + l)Kp 3e
\\f(t)Tn (t,x)dt
j o 7rc 2
2
(2 x b) >Jx An 2
where K is the upper bound of \F(t)\ in (v, b) when the intervals ^ are
omitted.
Hence it follows that the expression on the left can be made arbitrarily
small by taking n sufficiently large, and so the analogue of the Riemann-
Lebesgue Lemma is completely proved.
Let "m = /8 , . ,
f tf(t)J(jmt) dt,
** v+i \jm) ~
where v + > 0.
Let x be any internal point of an interval (a, b) such that < a < b < 1 and
such thatf(t) has limited total fluctuation in (a, b).
00
8* (*) s
j
V* {IT' f(t) - x-f(x - 0)} Tn (t, x) dt
*" +1 - r*/(* + 0)} Tn (t, x) dt,
+ f
{-/()
2 0>mJv (jmX)
We now discuss
1
V+ .
1
[tr"f(t) - xrf(x + 0)1 Tn (t, x) dt
iX
J
in detail, and the reader can then investigate the other integral involved in
such that
Xi(*+0)-#(* + 0)0.
Hence, when an arbitrary positive number e is chosen, there exists a
positive number 8 not exceeding b x, such that
We then have
'V+i it- f(t) - ar*f{a + 0)} Tn (t, x) dt
X+S
ra;+8
fa;+a
x, (t) Tn (t, x) dt.
1
+ ' <" +1
X! (0 Tn (t, x)dt- ***
J x > X
We now obtain inequalities satisfied by the three integrals on the right.
from the analogue of the Riemann-Lebesgue lemma that the
It follows
modulus of the first can be made less than e by taking n sufficiently large.
Next, from the second mean- value theorem it follows that there is a number
between and 8 such that
+S ('*'
r t+> Xl (t) Tn (t, x) dt = xx (* + 8) J *" +1 Tn <* *> dt >
J X *+f
and, by 1822, the modulus of this does not exceed 2Ue; and similarly the
modulus of the third integral does not exceed 2Ue. By treating the integral
between the limits and *ina similar manner, we deduce that, by taking n
sufficiently large, we can make the difference between
P^Tnit^dt,
Jo
when x and t are small, without using approximations for Bessel functions of
the second kind.
to 18-55.
v+1
~ *"/(*)} Tn (t, x) dt.
fx J {'""/CO
provided that
t+i{t-f(t)-x-''f(x)}dt
F X
J x-\-6
is a bounded function of x.
Now
11*1 l
l^r^^f^-^'f^)} dt U I
\o
/
\t h f(t)\dt+\x-*f{x)\ [\+idt,
Jo
and this is bounded in (a + A, b - A) since f(x) is continuous and therefore
bounded in this interval.
* Cf. Modern Analysis, 9-44.
t Cf. Modern Analysis, 3 61. It is now convenient to place an additional
(trivial) restriction
on s, namely that it should be less than A, in order
that the interval Ix- d, x + d) may lie inside
the interval (a, b).
594 THEORY OF BESSEL FUNCTIONS [CHAP. XVIII
tends to zero uniformly as w-oo , and this proves the theorem stated.
/(l-0) =
is necessary in order that the convergence of the Fourier-Bessel expansion
associated with/(j5) may be uniform near x = l.
We shall now prove is to be continuous in
that the conditions that f(x)
(a, 1) and that /(l) is zero, combined with the conditions stated* in 18*24,
are sufficient for the convergence to be uniform throughout (a + A, 1).
The analysis is almost identical with that of the preceding section; we take
just as before, and we then divide the interval (0, 1) either into three parts
(0, x 8), (x 8,x + 8), (a* + 8, 1), if x ^ 1 8, or into two parts (0, x 8),
(x 8, 1),. if x ^ 1 8. And we then prove that the three integrals (or the two
integrals, as the case may be) tend uniformly to zero.
\x~>f(x)\<6, \f(x)\<
when 1 o\ $x$ 1.
f{x)-x-f{x){\'^Tn {t,x) dt
Jo
expression does not exceed (U+l)e for any value of n when x lies in
(1-8,, 1).
* The interval (a, b) is, of course, to be replaced by the interval (a, 1).
f(x)-x-f{x)\\^Tn (t,x) dt
.'o
\\f(t)Tn (t,x)dt
Jo
to the sum/(ar) in (a + A, 1) in the postulated circumstances.
Now write **/(*) = >M0 ~ ">M0>where fr(t) and ^ % (t) are monotonic
in (a, 6); and then a number exists such that
| /
J a
*i (0 ** <L (M) dt = ^(a) f
*
<* Jv (\t) dt + ^ (b) f V /, (Xt) dt
'
J a Jf
<2c{|+i(a)| + |*i(6)|}X-
= 0(\-*).
A similar result holds for yfr2 (t), and hence the theorem stated is evident.
If it is known merely that
rh
f t*f(t)dt
J a
2 {t m m _i) ( Um L m) < ,
r=l
where Um and Lm are the upper and lower bounds of t*f{t) in (t m -i, tm ).
Next let tlf(t) = F(t), F(t) = F(tm^) + oy m (t),
and then
m tm
I \\f{t)Jv (\t)dt\<K I I P tU v (M)dt\+ \tiJ v (\t)a} m (t)\dt
\Ja I m = l> J tm-i I m=l J['m-i
where c is the upper bound of \t*J v (t)\ in the interval (0, oo ). Hence, by
reasoning resembling that used in 1-8-23, the integral on the left is o (Ar J ),
and this is the theorem to be proved.
The theorems of this section can be made to cover the closed interval
Hence the general term in the Fourier-Bessel series associated with fix)
tends to zero {after multiplication by Jx) throughout the interval (0 ^ x^ 1)
if x*f(x) has an integral which is absolutely convergent ; and, if this function
has limited total fluctuation, the general term tends to zero as rapidly as l/jm .
2 bmJrCkmX),
bm
f
tJ* (Xnt) dt = [\f(t) Jv (\J) dt
. Jo Jo
so that
l
2\m*j tf(t)J(Km t)dt
If H+ v = 0, an initial term
(1) 2( J / + 1)#" [\+if(t)dt
Jo
has to be inserted on account of the zero at the origin.
If H+ v is negative and the purely imaginary zeros are ftX,, then an
initial term
K
' (V + P)/. 1
(5l.)-V/r (^)io
/, /()y, (V)Ctt
'
These initial terms in the respective cases will be denoted by the common
symbol <%(#), so that the series which will actually be considered is
# (*)+ m2 bm J (K 1
v n x),
2Jy(jm x)Jy{jm t)
The residue at \ m is
2XTO J, (Xmiv) Jv (\mt)
J v (Xm ) \\n JJ' (Xw) + JJ (Xm ) + H JJ (\m)}
2 \ m2 J y (\fnOC) J p (Xmt)
X^ - VZ) JS (Xm ) + X Jy'^ (X^)
( *
2X 2 /y (X a?)/y (XoQ
+ ir-) /,* (x ) - x 2 // (x
(xa )
2J
Let Sn (t,x; H) = 2
m=l
^^~'^-S4
i>+i \Jm)
a (x,t)
2\^Jy(\m X)Jy(\m t)
m=l (Xw -I^)^ (^m)
2 2
+ KfJJ'iKd'
where 4 (x, t) is defined to be 0, 2 (*/.+ l)xv t* or
2X 2 /p (A ar)/> (X Q
(V + v ) L 2 2
CK) ~X 2
//" (,Xo)
according as H+ i> is positive, zero or negative.
Then, evidently,
We shall now prove a number of theorems leading up to the result that, when
< x< 1, the existence and absolute convergence of
Jo
are sufficient to ensure that, as n -* oo ,
where the symbol P denotes Cauchy's principal value.' The integrand being '
m
K)
nK(t v
'
.
'
m _~2TriJ
)
J_ [+"* 2wJ (xw)J,(tw)dw
J (w){wJ '(w) + HJ
J)n - aDi l>
v
r ,(w)}-
l
Also
and hence
4
n (t,x;H)dt\^
t"+*Sv.
!/.o, /-j (2-x-t)Dn ^/x'
where c4 is independent of n, x and t.
"fif{t)dt
I
J a
provided that < x < 1. And, if b< 1, the theorem is valid when < x $ 1.
The proof has to be divided into three stages just as in the corresponding
theorem ( 1823) for Tn (t, x). We shall now give the proof of the first stage,
when it is supposed that t*f(t) is bounded and a > 0. The proofs of the re-
maining stages should be constructed by the reader without difficulty.
Divide (a, 6) into p equal parts by the points t u t2 , ... , tp^. (t = a,tp = b),
and, after choosing an arbitrary positive number e, take p to be so large that
p
1 ( Um - Lm) (tm - m ~!) <
t e,
where Um and L m are the upper and lower bounds of F(t) in (<*_!, TO ).
Then
tf(t)Sn (t,x:H)dt
/,
W
= I Fit^f** t "Sn (t,x;H)dt + I f
J
t+*a> m (t)Sn (t,x;H)dt.
Hence, by 1831,
/" 6
^7 1 [2Kpc< 1
| J
*/() s ft * ff ) ;
* |
<
(g _*.V
and the expression on the right is arbitrarily small. Hence the integral on
\*f(t)dt
/,o
tends to zero as n -*-oo when < a; < 1 ; and this sum (multiplied by >Jx) tends
uniformly to zero when ^ x ^ 1 A.
Now, since the numbers \ m and jm which exceed v are interlaced | |
( 1523), it follows that Dw may be chosen so that n has the same value N
for all values of n after a certain stage.
Therefore, since
n
x* 2 am Jv (jm x)~*0,
Hence, if for any particular value of x in the interval (0, 1 A), the series
00
Q^a<b< 1,
then also the Dini series (multiplied by \/x) is uniformly convergent (or uni-
formly summable) throughout (a, b).
m=l
converges to the sum
* {/(* + <>)+/( -0)}
at all points x such that a + A^x^b A, where A is arbitrarily small ; and
the convergence is uniform, if f(x) is continuous in (a, 6).
Hobson*, that, if/(a) has limited total fluctuation in the interval (a, 1), the sum
of the Dini expansion at x = 1 is /(l 0).
We first write
Tn {t,x; H)=Tn (t,x)-Sn {t,x; H)
2\wt2 J y (\m x) J v {\m t)
= <J (r A 4- *?
where
</> (w, x) = 7T [{w j; (w) + #/, (/)} 7, (#w) - {w Y: (w) + HYV (v>)} J, (xw)].
The former representation of Tn (t,x\ H) is valid when 0< t < x^ 1, the
latter when 0< x< t^ 1.
tions of Tn (t, x) given by 18*21 (7) and 1821 (8) the fact that there is no formula for
;
Tn (t, x H) analogous
;
-
to 18 21 (6) is the reason why Dini series were discussed in 18-33
with the help of the theory of Fourier-Bessel series.]
'
v+1
Tn (t,l;H)dt
f
Jo
t
For any given positive value of 8, it follows from 1821 that this is a
bounded function of t in the interval (S, 1). When 8 ** t *S 1 h, it is (1/Dn).
And when t = 1, it has the limit 1 when n-*oo .
It follows that
S (1) +=i
I bm J v (\ m ) -/(l - 0) = I
Jo
V' {r"/(0 -/(l - 0)} Tn (*, 1 ; H) dt.
Xi(i-o)=x*(i-<>)=o.
Hence, given an arbitrary positive number e, we can choose a positive
number B, not exceeding 1 a, such that
whenever 1 8 ** $ 1.
We then have
f
V+ 1
[t-"f(t) -/(l - 0)} Tn {t,l;H) dt
Jo
Ji- Ji-s
It follows that
lim f
n-^at>J
V +1
{f'f(t) -/(l - 0)} Tn (t, l;H)dt = 0,
m=l
converges to the sum /(l 0).
This discrepancy between the behaviours of Dini series and of Fourier -
Bessel series ( 18*26) is somewhat remarkable.
\\*f(t)dt,
Jo
and the condition that f(x) has limited total fluctuation in (a, 1), may be
sufficient to ensure the uniformity of the convergence of the Dini expansion
in (a + A, 1).
The reason for the failure in the uniformity of the convergence of the
Fourier- Bessel expansion ( 18*26) near x 1 was the fact that
v+ *T (t,x)dt
\ t n
Jo
does not converge uniformly to xv in (A, 1), as was seen in 1822. We shall
prove that, on the contrary,
t t"
+l
Tn (t,x;H)dt
Jo
does converge uniformly to xv in (A, 1), and the cause of the failure is
removed.
A consideration of 1826 should then enable the reader to see without
difficulty that the Dini expansion converges uniformly in (a + A, 1).
It is easy to see, from 18*34, that
\ t*+*Tn (t,x\H)dt
Jo
is the sum of the residues of
7rt
v+l
[{wj; (w) + HJ V (w)} F+1 (too) - {w YJ (w) + HYV (w)} J+1 (tw)]
x J (xw)l{wJJ (w) + HJ V (/)},
at \ lt Xjj, ..., \, plus half the residues at or + iko if H + v < 0.
Hence f
t
v+1 Tn (t, x; H) dt
Jo
and so the integral on the right converges uniformly to zero like l/(Dn Jx)
when A ^ x $ 1. That is to say
l
t"^Tn {t,x;H)dt
converges uniformly to xv in (A, 1); and we have just seen that this is a suf-
ficient condition the uniformity of the convergence of the Dini series
for
associated with/(<) to the sum f{x) in (a + A, 1) under the conditions postu-
lated concerning f{t).
Ford also investigated the differentiability of Dini's expansion when H=-v, but his
method is not applicable to other values of H.
( 3) / (*) - ^/(*)
x
= - 2 am jm>y J
m-l
v+1 (jmtV x);
inside any interval in which the function has limited fluctuation, provided that
fy{f'(t)-j/(t)\dt
exists and is absolutely convergent.
* The term in wJv'(w) is more important than the term in Jv (w) except in the Uviit when H
is infinite; this shews clearly the reason for the difference in the behaviour of the Dini expansion
from that of the Fourier-Bessel expansion (cf. 18-26).
t Trans. American Math. Soc. vr. (1908), pp. 178184. % Cf. Modern Analysis, 9-31.
606 THEORY OF BBSSEL FUNCTIONS [CHAP. XVIII
Jm,v wu
Q,
The expression which will be taken as the 'sum' of the series by the
method of Riesz is
lira I (l-if)am Jv {jm xy,
and when this limit exists, the Fourier-Bessel series will be said to be sum-
mable (R).
It is evident that
l
where
(2) Tn (t,x\R)= i (i
- ip) -^ (}?*>/: [* m l) ,
manner of 1822 before we make further progress with the main problem.
ZlJAn->i\ AJ V
Jp (w)
|
Jp (tw)jjy (w)
4> (w, x)
|
* When t $. .r, the integrals taken along the lines joining iB to A n ill do not tend to zero
as B - ao . There is no need to make an indentation at the origin, because $ (w, x) is analytic at
the origin.
608 THEORY OF BESS EL FUNCTIONS |CHAP. XVm
We first deal with the factor Jv (tw)/Jy (w). We observe that*
J, (tw)
(1) <^exp{-<l-t) |
I(w)\}
Jy{W)
when w is on either contour; this follows from inequalities of the type
18*21 (9) when w is not small, and from the ascending series
| |
when |
w |
is
(2)
for all the values of w and x under consideration when \^v^\ . Hence
(5) iW'WK^- 1
, |ff,.()|<*gp.
If we now combine the results contained in formulae (3), (4) and (7) we
deduce that, whether ^%v^\ or v^^,
(8) 1
4> (w, *) |
< h (ar* + x) exp {(1 - x) / (w)\ j },
* It is supposed that the numbers Jtj, k^, is , ... are positive and independent of w, x and t;
when w is any point of either contour and < x ^ 1. Hence, by (1), it follows
that
J
(9) \^{w,x)
I
^\<hr^(^ + x-')^\-{x-1^\I{^\\,
O v \W) i
We now return to the integral formula for Tn (t, x R). If we replace w by\
ir.ft.|ji)i<^(^ + ^)j;^.*-*g+^.
We have consequently proved the two inequalities
(ii) |j.. tt .| a)
|<gffi^ <o<.<i).
r-< t "' B
)-si( -2;)*<" '>-^(5r'
1 i
or on the vertical sides of this rectangle and the factor 1 (iv/A n ) does not ;
(13) \
Tn (t,x\R)\<^!^^.
\ ftf{t)dt
Jo
are sufficient to ensure that the Fourier-Bessel series associated with f(t) is
summable (R) at all points x of the open interval (0, 1) at which the two limits
/(# 0) exist. And the sum (R) of the series is
*{/<* + Q)+/(*-0)}.
This theorem is obviously the analogue of Fejer's theorem* concerning Fourier
series.
X
lira [ t>
+l Tn (t,x\R)dt= lim f
V* 1
Tn (t, |
R) dt
I 1
m=l (
\
- 2p) a m Jv (jm x) -
-"/
!
JO
VT 1
n (t, x R) x~"f(x - 0) dt
|
- f
t*+ l Tn (t, x R) x~"f(x + 0) dt,
|
J x
and we shall prove that it can be made arbitrarily small by taking n sufficiently
large.
. The sum 8n (x R) \
is equal to
f
jo
V {t~'f(t) - x->f(x - 0)} Tn (t, x\R)dt
l
+ f t
v+l
{tr"f(t) - ar'f(x + 0)} Tn (t, x\R)dt.
J X
Now, on the hypothesis that the limits f(x 0) exist, if we choose an arbitrary
positivenumber e, there exists a positive number J 8 such that
f *-7(0
- *-v f{x + 0) < e, I
(x^t^x + j 8),
We now choose a positive function of n, say o- (n), which is less than 8 for
sufficiently large values of n, and divide the interval (0, 1) into six parts by
the points x 8, x a (n), x.
Zk6 +*
+ -, *
+ - *-/(* + 0)}
A-*V_ (*' {*"/(*> dt.
&*Jx}f
1 I
n x +*
For any given value of e (and therefore of 8), the first and last terms in this
expression can be made arbitrarily small by taking n sufficiently large, on
account of the convergence of
n
\t*f(t)\dl
/;
has been done, the first and last terms can be made as small as we please by
taking n sufficiently large.
lim Sn (x\R) = Q.
Hence
Since each of the limits on the right is equal to \xv , it has now been
proved that
2 am Jv (jm x)
issummable (R) with sum l\f(x + 0) +f(x - 0)} provided that the limits
f{x 0) exist and this is the theorem to be established.
;
As a corollary, the reader should be able to prove without difficulty that, if f(t) is
continuous in (a, b), the summability (R) is uniform throughout the interval in which
a + A^x^.b-A, where A is any positive number. Cf. 18*25.
Sn (x\R) =
J
V> \r*f{t) - x~"f(x)} Tn (t, x\R)dt
Jo
In this formula any of the limits of integration which are negative are supposed
to be replaced by zero.
18*53, 18*54] FOUBIER-BESSEL SERIES 613
at f F+ 1
Tn (t,x)dt
Jo
is uniformly convergent in (0, 1 A), and so, since uniformity of convergence
involves uniformity of summability,
xi [ tf(t)Tn (t,x\R)dt
Jo
2 am x*Jv (jm x)
TO = 1
.'o
fV<o dt
exists and is absolutely convergent, and that t~ v f(t) is continuous in (0,6).
2 a m xlJv (j m x)
m=0
on the hypotheses (i) that the limits /(a,- + 0) exist, (ii) that
j\if(t)dt
exists and is absolutely convergent, and (iii) that the series is summable (R).
It conduces to brevity to write fm {x) in place oia m x*Jv {j w x), so that/,,, (x)
* The factor x^ is inserted merely in order that the discussion may cover the investigation of
uniformity of summability near the origin.
614 THEORY OF BESSEL FUNCTIONS [CHAP. XVIII
w--ac \ Jn / m=l
and therefore
lim I (l-k)fm{x)= li m | (i_is) /m(a ); .
the limit on the right exists in consequence of the hypotheses made at the
beginning of the section.
Again, since
Jn n W
whether m be o{n) or 0(n), it follows that
lim 2,(f-)/.W=o.
and so
Jo
exists and is absolutely convergent are sufficient to ensure that
00
2 am xljv (jm x)
m=l
is summable (C 1) with sum #* {f(x + 0) +/(.# 0)}.
By the same reasoning, if /(a?) is continuous in (a, 6), the summability (C 1)
is uniform in (a + A, 6 A) ; and, if a and i~*f{t) has a limit as -* 0, the
1 am x*Jv (jm x)
JB=1
eo
w=l
co
m=l
is uniformly convergent throughout (0, 6 A), with sum x*f(x).
Again
2 (am -aJJJ. (jm x) = ** [ V() Tn ft a) <fc
and this tends uniformly to zero in (0, b A) as n-*-oo by an analogue of the
Riemann-Lebesgue lemma ( 18*23).
n
Hence 2, am x*Jv (jm x) tends uniformly to the sum x*f(x) in (0, 6 A) as
dv __, (d 2v 1 dv
dt~ \dr 2+ rdr
with the boundary condition
is the Dini series associated with the initial temperature f(r). It is evident
that the initial temperature is expressible as
00
and this limit exists when the Dini series is summable (R).
It has been shewn by YoungJ that the existence and the absolute con-
vergence of
Jo
are sufficient to ensure that if all the coefficients a m of the Dini series (or the
Fourier- Bessel series) associated withf(t) are zero, then the function (t) must
f
be a null-function.
* La Theorie Analytique de la Chaleur (Paris, 1822), 306320. Cf Rayleigh, Phil. Mag. {6)
.
xii. (1906), pp. 106107 [Scientific Papers, v. (1912), pp. 338339] ; and Kirchhoff, Berliner
Sitzungsberichte, 1883, pp. 519 524.
t In this physical problem, H
> 0, and so there is no initial term to be inserted.
t Proc. London Math. Soc. (2) xvni. (1920), pp. 174175.
18-6] FOURIER-BESSEL SERIES 617
<3C 00
(where the coefficients am and bm are any constants) which resembles Riemann's
theory of trigonometrical series f.
Such a theory is, however, more directly associated with Schlbmilch's
series of Bessel functions,which will be discussed in Chapter xix ; and it
seems convenient to defer the examination of the series
ao oo
* Lerch, Acta Mathematica, xxvn. (1903), pp. 345347 ; Yoang, Messenger, xl. (1910)
pp. 3743. Cf. 12-22.
t Cf. Modern Analysis, 9*6 9632.
CHAPTER XIX
SCHLOMILCH SERIES
jm = (m + \v - i) 7T + (1/m).
That is to say, the argument of the Bessel function in a term of high rank in
2 OmJyimx).
It will appear subsequently that it is convenient to add an initial term
(19-11; cf. 1833); and the analysis is simplified by making a slight modi-
CCr
where
o = 2/(0) + - ['
f uf(u sin <j>) d<bdu,
7TJ0 JO
(2) -
2 [ n [**
m = - / / uf (u sin <f>)
cos raw d<f>du; {m > 0)
and this expansion is valid, and the series is convergent, throughout the closed
interval (0, tt).
2x ft" ft*
-/() + J ] f (x sin 6 sin <f>)
sin0 d<f>dd.
2x [*" [ 9 1, x
sin cos v d% d0
/(a;Smx) V(cos^-cos
V.U x V0)
= 2x
- f* , .
J . /cos 0\~|** ,
J^
/ (x sin X ) |- arc sin
( J J^
cos X dX
= #/ /'(a;sinx)cosxax
Jo
and so, when <jr (x) is defined by (4), g (x) is a solution of (3).
with limited total fluctuation in the interval (0, ir), so also is g{x); and
therefore, by Fourier's theorem, g (x) is expansible in the form
00
g(x) = %a + 2 am cosmx,
m-l
where
2 [*
am = \ g (u) cos mudu
and this series for g (x) converges uniformly throughout the interval (0, tt).
/<*)-!fV**)<W
irJ o
= |a + 2 am J (mx),
m=l
and this is the expansion to be established. It is easy to verify that the
values obtained for the coefficients am are the same as those given by
equation (2).
When the restriction concerning the limited total fluctuation of /' (x)
is removed, the Fourier series associated with g(x) is no longer necessarily
convergent, though the continuity oif(x) ensures that the Fourier series
19 * 2 ] SCHLOMILCH SERIES 621
is uniformly summable (C 1) throughout (0, tt); and hence, by term-by-term
integration, the series
00
a + X am J (mx)
am = (l/0n)
is sufficient to ensure the convergence of the Schlomilch series to the sum
f(x) when x lies in the half-open interval in which < x % ir.
For further theorems concerning the summability of Schlomilch series,
the reader should consult a memoir by Chapman
f.
[Note. The integral equation connecting f(x) and (x) is one which was solved
g in 1823
by Abel, Journal fur Math. i. (1826), p. 153. It has subsequently been investigated*
by
Beltrami, 1st. Lombardo Rendiconti, (2) xm.
(1880), pp. 327, 402 Volterra, Ann. di Mat. ;
(2) xxv. (1897), p. 104 C. E. Smith, Trans. American Math. Soc. vm.
;
(1907), pp. 92 106.
The equation
2x fhw rj
rr
/ 1 / (x sin 6 sin <p) sin 6 <fy d6 / (x) -/(0)
* Cf.
Modern Analysis, 8 5.
t Quarterly Journal, xliii. (1911), p. 34.
t Some interesting applications of Fourier's integral theorem to the integral equation have
been made by Steam, Quarterly Journal, xvn. (1880), pp. 90104.
.
Bessel functions of the second kind and Struve's functions; and the types of
series to be considered may be written in the forms*:
a , m Jv (mx) + bm Y, (mx)
a am J* (x) + bm H (ma?)
I> + 1) =! (\rnxy
Series of the former type (with i/= 0) have been considered by Coatesf;
but his proof of the possibility of expanding an arbitrary function f(x) into
such a series seems to be invalid except in the trivial case in which f(x) is
Series of the latter type are of much greater interest, and they form a
direct generalisation of trigonometrical series. They will be called generalised
Schlomilch series.
function into such a series; and the second is the problem of determining the
properties of such a series with given coefficients and, in particular, the
construction of analysis (resembling Riemann's analysis of trigonometrical
series) with the object of determining whether a generalised Schlomilch
series, in which the coefficients are not all zero, can represent a null-function.
K. Danske VidensJkabernes Selskabs, 1899, pp. 661665 1900, pp. 5560 1901, pp.
; ;
Nielsen J has given the forms for the coefficients in the generalised
Schlomilch expansion of an arbitrary function and he has investigated with
great detail the actual construction of Schlomilch series which represent
null-functions, but his researches are of a distinctly different character from
those which will be given in this chapter.
* The reason for inserting the factor x" in the denominators is to make the terms of the second
This is analogous to the procedure which is adopted in* Dirichlet's proof of Fourier's
theorem in the expansion
:
GO
We then take the series in which the coefficients have these values, namely
2^1' f(t)dt+~ I
j f (t) cos m(x-t)dt,
and prove that it actually converges to the sum f(x).
Jy (mx) i H (mx)
'
(I mx)'
instead of with the pair of functions
We shall write
(1)
Jv (z) + il3..(z) _
* Apart from details of notation, the following analysis is due to Filon; it was given by him,
in the memoir just cited, for the special case *=0, but the extension to values of v between
presents no difficulty
) .
so that* <f> v
(z) is analytic and uniform for all finite values of the complex
variable z; and evidently
J (mx)
v
i
t'H (mx)
i ' s .
<f>
,
(
,
+ mx).
.
w<f> v (xz)
&\WKZ
where m = 0, 1, +2, . .
.
; and so we shall consider the integral
7r< (xz)
~dz,
S7TI J c SI
and be analytic at infinity (cf. 19'24); its only singularity in the finite
to
part of fche plane is an essential singularity at the origin.
=-
2m
1 r
J
(0+)
ET/
F(z)
.ir<t>*(xz) ,
T
sin ttz
dz.
?" y
-s
2m .
J
F(z)
sin vrrz
' dz
* The insertion of the factor (J z) v in the denominator makes # (z) amenable to Cauchy's
2m J ' smirz
and, in order to calculate it in a simple manner, we shall suppose that F(z) is
expansible in a series of Filon's type*
(1) F(z) = i
w=l
?4#,*
where yfrn (z) denotes the sum of those terms in the expansion of ir- 1 sinirz
whose degree does not exceed n, and the coefficients n will be defined later.
p
The reader will observe that
M*)=-M*)=*-*iry,
With this definition of F{z), it is evident that, for small values of \z\,
^"sin^^/^sin^ j
*-<">
- y i
X Tn+1 cos in7r + 0(^)1 ^ , .
=i (s
n+1
(n+1)! sin-w^ jt\ /
_ rgyrgn+v + i)
r -**+-*-**
= - *ru-l) fm (0)
)
(1
/!
r( *> ^-f-^r^ ^/W] dt
-
- ^r(i-,)Jo
fa
(1
d* U*' 4
and so we obtain the symbolic formula
<">
where D stands
* - - *m~) /> -
for dfdu.
^ it ['"
*/>] *
and therefore
and we are able to effect our purpose by making use of formula (4), whence
we find that
(6)
ii'"(n+T)r
Wehave now obtained symbolic expressions for all the coefficients in the
generalised Schlbmilch expansion of /(#), but it is necessary to transform
these expressions into more useful forms, by finding the significance to be
and
-jjyzT\ ^ fcn g6116'8,1 values
attached to the symbolic operator '
for of *
r^e /(<B) i
\_tt(D-iz) j ^o .
e-^fi^dv,
i:
where a is a constant of integration ; and therefore
sinh irD sinh-wi) T u ,. ,
"1
.._, . .
f . .
7T ) a
_*Z /(ftf
/V >]
= r sinh^(D-f^) f-
yV
I
f
L7T(i)-t^)- 7T Ja
'J w==0 [ J-0
- e- I* 'e-^fitv) dv
Ja J-0
that is to sayf
The second term on the right has simple zeros at all the points at which
* = 0, 1, 2, ....
ir<f>, (xz)
F(z)
sin irz
at 0, + 1, + 2, ... is concerned, we may omit the second term on the right in (1),
and calculate the residues of
sin trz
<
2
> *- ww^km F.
(1 - eyi"Jt [*' /". e
^ /(tv) dv dt.
Again, from 1922 (6) and equation (1) of this section we have
j.gcoBinir
= r(i)/(0) f
_ py^dr
(1
vx
(3)
ti (n + 1)! r(i-).l "'" A eft
The first
2r(* ^W^^W^*
term on the right in (3) is equal to T (v+ l)/(0), except when*
dt
In the special case in which v = 0, the modified form of (3) shews that an
additional term /(0) must be inserted on the right in (4).
When we change the notation to the notation normally used for Bessel
functions and Struve's functions, the expansion becomes
(5) /W "^(y + l)
+
.t 1
"- ,
(\rnxY
where f
(6) )
am
-T{\-v)T{\)), O-^-a J
I
IT
f (tv) cos mv dv dt,
We now prove that, when the function f{x) is restricted in a suitable manner,
shall
the function F(z) is bounded when [z|--oo, whatever be the value of arg. The reader
will remember that the assumption that F(z) is bounded was made in 19"21 to secure
the convergence of the contour integral.
We take the series of 1922 (1), by which F{z) was originally denned, namely
* When v is negative it is necessary to use a modified expression for the integrals ; cf. 19*3.
t When p = 0, the expression for a flas * he modified by tbe insrtion of the term 2/(0), in
consequence of the discontinuity in value of the expression on the right of (3).
:
and divide it into two parts, namely the first N terms and the remainder of the terms,
where N
is the integer such that
N^ it j z |
< N+ 1.
'
!*l=i (n-l) ir|*|^t n ?o |*|* '
sinh
Since
\z\" +l
tends to zero as \z \
00 , it is evident that a sufficient condition for F(z) to be bounded
as z I
j
-*- 00 is that the series
2 \Pn\
2 " + i|/()(0)|
is convergent.
Let v be a number such that \ < v < \; and letf(x) be defined arbitrarily
in the interval (ir, ir), subject^ to the following conditions:
(II) The function h(x) has limited total fluctuation in the interval ( ir, it).
ri*i\*r
J O
if (*)-/(<>)}]**
is absolutely convergent when A is a (small) number either positive or negative.
1904), p. 348; but the formulae which he gives for the coefficients in the expansion seem to be
quite inconsistent with those given by equation (2).
t The effect of conditions (I) and (II) is merely to ensure the uniformity of the convergence
of a certain Fourier series connected with h (x).
X If v is positive, this Lipschitz condition is satisfied by reason of (II).
630 THEORY OF BESSEL FUNCTIONS [CHAP. XIX
where
K=L w j r<F^r(J) 2?
[sin2v
^ { f( U8in V-f( M sin ^*>
wAen m > 0; the value of a is obtained by inserting an additional term
2I> + l)/(0)
on the right in the first equation of the system (2).
sin
+ r^7> * o% [sin2" * [f{x *> /(0)1] d<t> -
[Note. The (absolute) convergence of the integral contained in this formula is secured
by condition (III). It should be observed that the aggregate of terms containing /(0) in
equation (4) may be omitted when v is positive in view of the formula
r * r( ^
/:o cUf>
2cosy7r in
[* [ C082* sec^+i *
<j> [gin*- {/(* sin sin <f>) -/(0)}] d<f>d0
IT Jo Jo <*9
+/(0).
Hence we have to prove that
^^ IT
f
JO
**
f ^cos
Jo
2"
$ sec 2" +1 <f>
^r
d<P
[sin" <f> {f(x sin sin <f>) -/(0)}] d<f>d0
=/(*)-/(0).
Replace <f>
on the left by a new variable % defined by the equation
=
n* rh*
io i x
'.Jit
sin cos 2 "
(cos
d r
X - cos 0)"+* d
rjir
2
*
[Sm" {/(a; Sm X) " /(0)1]
.
, .
^
= j'jjj tan2r *<**
and it is easy to verify that when f(x) is a continuous function with limited
total fluctuation in the interval ( ir, 7r) so also is g (a:), and therefore the
expansion for g (x) is uniformly convergent when - ir + 8 ^ x ^ it - 8, where 8
is an arbitrarily small positive number.
Replace a; by a; sin in the expansion of g (x), multiply by cos2 * 0, which
has an absolutely convergent integral, and integrate term-by-term ; we deduce
at once that
f( \
r{x) ^ a _i_ V aJ* ( ma! ) + ^tH y (mx)
~~
-T( v + 1) +.* W^f ;
[Note. Just. as in 19*11, if condition (II) concerning the limited total fluctuation of
vf{x)-\-xf'(x) is not all statements made in this section up to this point
satisfied, then
about convergence of series have to be replaced by statements about sumtnability (Cl).]
There is one important consequence which follows from the fact that a m
and bm are both (1/m) when 2vf{x) + xf (x) has limited total fluctuation
in ( it, 7r), namely, that in the neighbourhoods of it and it, the general
term of the Schlomilch expansion is (l(m v+ %), and so the expansion repre-
sents a continuous function ; hence the expansion converges (uniformly) to
the sum/(a;) throughout the interval ( it, it).
There are a few problems of Mathematical Physics (other than the problem
mentioned in 19*1) in which Schlomilch series occur in a natural manner,
and we shall now give an account of various researches in which Schlomilch
series are to be found.
A very simple series is
1+1 e~ mz J (mp)
this series is convergent when p and z are positive, and, if p and z denote
cylindrical-polar coordinates, it is a solution of Laplace's equation at all points
of space above the plane z = 0.
Various transformations of the series have been given by Whittaker*;
thus, by changing to Cartesian coordinates (x, y, z) and using 2*21, we have
v/
(1) 1+ %
,
00
n= i
e- mz J (mp)
v
If'
r/ = z- :
2ir J _ 1 exp >
{ (z
.
du
. :
+ %x cos u + xy sin u)\
:
-.
(2)
v
1+2
7W=l
CO
e~ mz Jo (mp)
r/ = 5-
1
2-7T
fir
J _..
+
z
.
ix cos u
fill
~r~. +
+ %y sin u 52
1
+7T 2 -7s
27T OT=! (2m)l
-r-^ I
J_,r
(z
v + ix cos VI
u + iy sin m) 2**" 1 du
^L_ = I + l4. i f
1
\t - 2miri
I
l
} '
\-e- 2 1
t w=1 t + 2miri)
whence we deduce that
(3) 1+ I= e-Jo(mp) = i + s
1M 1 T &
?! \jj{(%nviri
due to a
It follows that the series represents the electrostatic potential
set of unit charges (some positive and some negative) at the origin and at a
set of imaginary points.
1
(4) -r-z ^ = " + *" 2 J (wwr) cos withe,
Multiply by e""* and integrate, and we then obtain the formula (also due
to Nagaoka)
<3 >
I e%
Jj
goaii
w=*>
fa = ~b igoxi n
+
2
?.
Jr '
,
(mv)
^^
.a cos mirx lairi sin mmsc\
J
The series on the right in (5) converges uniformly throughout the interval
fractions*.
Various representations of the integral on the left of (5) were obtained by Nagaoka
the formula quoted seems to be the most interesting of them.
?, (2m-iy -g(*-MX x
then, if/(#) denotes the sum of the Fourier series, we see that
= + \f(xsmd)d0
^
T (JO J arc sin {nix))
O
'
/"arc sin (nix)
^(Tr-2x8m6)d0
TTJ o o
-/"""
+ (2# sin - 3tt) d0,
'w arc sin (ir/x)
"
so that, when -n- < a? < 2-7T, we have
S J".[(2m-1)} ... /tt if
(9) ^ To rts
!
= v (^ t ) *a; 7r arc cos +
2
8"
(1) \+2
* m-l
(-rJ.(m)-0,
provided that <x< ir ; the series oscillates when x = and diverges to
+ oo when x = ir.
This theorem has no analogue in the theory of Fourier series, and, in fact,
it is definitely knownj that a Fourier cosine-series cannot represent a null-
function throughout the interval (0, ir).
7r Jo cos (# sin )
_ (~) M [ x cos (^ + h) u du
ir J cos \u ' ^(x2 v?)
= 0(1),
as itf -*- oo , by the Riemann-Lebesgue lemma*, which is applicable because
the integral
du
f
Jo cos \u
'
o . \Z(ar* a
it )
(i) when < v $ J and < a; < 7r, (ii) when v > and < x ^ tt.
By using Poisson's integral we have (since v> \)
"
\ (-)*Jp(imB)
, g
r<v+i) "",.-! {\mxy
" cos" tdt
- ro+Lr^ r {^ + .?i r C08 ( (rw * 8in
(
M '
r(* + *)r<*)Wo cos** <*
= o(l),
as Af -* co ,
provided that the integral
*(a?-u*y
dw
.'o cosu
exists and is absolutely convergent; and this is the case when x and v satisfy
the conditions stated.
The truth of (2) is now evident.
may be applied n times to equation (2). The effect of applying the operator
once to the function Jv {mx)j{\mx) v
is to multiply the function by w s
; and
therefore, when < x < ir,
(-jm )" (-)*"/,, (ms) =
2
J {)
m=i (Jwa?)"
that
l<\\
(3)
is to say,
y (-)"'/>
J5i (i^r-
H _
- '
same time, Nyt Tidsskrift, x. B (1899), pp. 73 81 ; Ooersigt K. Danske Yidenskabernes
Selskabs, 1899, pp. 661 665. In the first two of these three papers integral values of v
only were considered, the extension to general values of v being made in the third paper.
Shortly afterwards t Nielsen gave a formula for the sum of the series in (2) when x> ir ;
M f* cos (M+$)u ^_ ur -j d
; cos^it
It is thus found that, when x is positive and q is the integer such that
(2q-l)ir<x<(2q + l)ir,
then
(~)m Jv (mx) 2r(J) (2tt-l) 2 7ry-*
W w^TTT+
r(i/ + l)
2
m=1 (%mxy xT{*
J
The importance of Nielsen's formulae lies in the fact that they make it
evident that, when a function f{x) is defined for the interval ( ir, ir), if the
function can be represented by a Schldmilch series throughout the interval
number of points) the representation is not unique
(except possibly at a finite
and there are an unlimited number of Schlomilch series which are equal to
the function f(x) throughout the interval, except at a finite number of points,
namely the points already specified together with the origin and (when
< v K. i) the end-points ir.
The converse theorem, that the only Schldmilch series with non-vanishing
coefficients which represent null-functions at all points of the interval
ir<x<ir, (whenl \ < v ^.\) except the origin are constant multiples of
h | (-) m J,(mx)
I> + 1) ^-x ($mxy '
any Schlomilch series other than the one given can represent a null- function when J<.r<|.
19*5] SCHLOMILCH SERIES 637
is, of course, of a much deeper character, and it seems that no proof of it has
yet been published. We shall now discuss a series of propositions which lead
up to this theorem ; the analysis which will be used resembles, in its main
features, the analysis*, due to Riemann, which is applicable to trigonometrical
series.
ticular case is to avoid the loss of clearness due to the greater complication
in the appearance of the formulae in the more general case. With a few
exceptions, the complications in the general case are complications in detail
only; those which are not matters. of detail will be dealt with fully in
19-6 1962.
The series now to be considered is
ao
(1) a + 2 am J (ma;),
m=\
in which the coefficients a m are arbitrarily given functions of m.
We shall first prove the analogue of Cantor's lemma f, namely that the
condition that am J (mac) -*Oosm->oo,aiaW points of any interval of values
of X, is sufficient to ensure that
am = o ( *Jm).
[Note. If the origin is a point of the interval in question, then the theorem that
. , . . n
m=o(l)
is obviously true.]
Take any portion^ of the intervalwhich does not contain the origin, and
let this portion be called It . Let the length of J, be L t .
and, asm-*-,
P (ma?, 0)^1, Q(mco, 0)^0.
Hence, for all sufficiently large values- of m, (say all values exceeding ra,,)
P(mx,0)>$, \Q(mx,0)\^$
at all points of/,.
Now suppose that am is not o (^/m); we have to shew that this hypothesis
leads to a contradiction.
I
am >|
e slm
|
cos (w/a? far) > | V3, |
sin (ra/a? \ir) < | j
Then cos (w8'a; ^7r) goes through all its phases in 72 and so there must
,
j
cos (rn^'x \ir) ^ \/3, | |
sin (tfij'a? ^7r) < |
3-1 //2\
>6 V
4 vUi'
and this is inconsistent with the hypothesis that a^ J (mx) tends to zero at
all points of /
x .
The contradiction which has now been obtained shews that Om must be
o {\Jm).
The next theorem which we shall prove is that, if the Schlomilch series
converges throughout any interval, then the necessary and sufficient condition
that the series should converge for any positive value of x (whether a point of
the interval or not) is that the series
This theorem is evident from the fact that the general term of the
trigonometrical series differs from a m J (mx) by a function of m which is
%a + 2 am J (mx),
m-l
at any point at which the series is convergent, be called f(x).
Let
a J< mc>
(1) F(*)- o,*- 2 .
Then P (x) will be called the function associated with the Schlomilch
series whose sum is f(x).
It is easy to see that, if the series defining f(x) converges at all points
of any interval, then the series defining P (x) converges for all real values
of x.
lJ.(*)ki;
and consequently
am J (mx)
^ o / 1\
Since
*)
is convergent, it is obvious that the series on the right in (1) must be
convergent.
It is evident, moreover, not only that the convergence is absolute, but also
that it is uniform throughout any domain of values of the real variable x.
640 THEORY OF BESSEL FUNCTIONS [CHAP. XIX
19*52. Lemma I.
We shall now prove that, if Y(x) is the function associated with the
Schlomilch series whose sum is f(x\ and if
,i\
(1)
-*/
G(x,a)x = (x + a)T(x-r2a) + (x-a)T(a;-2a)-2xT(a;) ^ ,
then
(2) lim G(x,a) = xf(x)
a-fcO
at any point x at which the series defining f(x) is convergent, provided that*
am = o (Jm).
G- (x, a) = \a x 2
,. l 4o2 m3
lim .
\(x + a) J (mx + 2ma) + (x a) J (mx 2ma) 2xJ (mx)~\
2
a +o4>m cP
By observing that
-V{(2)} % tg + V fc (g2g)}
3
< |a /^
and that the series
S a Jo (mx 2ma)
m i w2 [x a + *J{x(x 2a)}]
* Since we are not assuming more than the convergence of / (x) at a single point, it is not
*".*, W?
[V( + 2a) Jo (mx + 2wa)
. + */(x 2a) J (mx - 2ma) -2*Jx.J
. (mx)]
by the sum of two series, each of which is uniformly convergent.
-
^^ V(# + [ 2a) . J {mx + 2wa) + ftx - 2a) J (mx - 2ma) - 2 sjx J
. . (mx)]
To deal with the fourth series, we observe that, by the first mean-value
theorem, numbers f and 1 exist such that
-1<0<1, -1<0,<1,
* Cf. Modern Analysis, 9'62, 9-621. It has been the general (bat not invariable) custom to
obtain various properties of the series without establishing the uniformity of their convergence.
The convergence of the series for f(x) is required to deal with the first serits the second series ;
can be dealt with in consequence of the less stringent hypothesis that a m =o\ >/m).
t The number is a function of a variable t which will be introduced immediately.
642 THEORY OF BESSEL FUNCTIONS [CHAP. XIX
for which
24> (mx) - <l> (mx + 2ma) (nix 2ma)
= 2ma I {<' (mx - 2mat) - <t>' (mx + 2mat)} dt
Jo
m=l
is uniformly convergent with respect to a.
Hence G (x, o) is expressed as the sum of six series each of which converges
uniformly with respect to a when \x < a < \x\ and therefore
lim G (x, a)
a--0
is equal to the sum of the limits of the terms of the series for G (x, a), i.e. it
isequal to xf(x), provided that the series for/O) is convergent ; and this is
We shall next prove that, with the notation of% 19*51, 19*52, the condition
Hence
lim [a G (x, a)]
a0
= lim(W)- \im-^~[(x + a)J (mx + 2ma)
+ (x - a) Jo (mx - 2ma) - 2xJ {mx)\
= 0,
and this is the lemma to be proved.
We can now prove that, if two Schlomilch series of the type now under
consideration (i.e. with v = 0, and with Struve's function absent) converge
**
Cf. Modrrv Analysis, 963.
19-53, 19*54] SCHLOMILCH SERIES 643
and have the same sum-function throughout the interval (0, ir), then corre-
sponding coefficients in the two series are equal. The formal statement of
the theorem is as follows
Two Schlomilch
series, of the special type, which converge and are equal at
of the closed interval (0, tr), with the possible exception of a finite
all points
number of points, must have corresponding coefficients equal, unless the end-
points and 7T are both exceptional points.
If these points are exceptional points, the two series may differ by a constant
multiple of the series
+2 (-T (nix).
$a + 2 am J (mx),
m-l
and let the sum of this series be f(x), so that f(x) converges to zero for all
If = 1, orif0 = -l,andif
*()-^[F()-F(ft)-^^{F(6)-F(ft)j]
*(fc)-*(&)-o.
If the first term of <f>
(x) is not zero J throughout the interval (&, 2), there
will be some point c at which it is not zero. Choose the sign of so that the
first term of <f>
(c) is positive at c, and then choose h so small that <f>
(c) is
still positive.
But <f>
(Cj 4- 2a) ^ <f>
(cj), <f>
(Cj 2a) ^ tf> (d), so the limit on the left must be
negative or zero. term of (cc) must
This contradiction shews that the first
be zero throughout (f1} f2), that is to say that F(#) must be a linear
function of log x and this is the theorem to be proved.
;
y= A log x + B.
Now, by 1951, F(#) is continuous in (0, ir), and so these logarithmic curves
are connected at the exceptional points; and the curve y = 'F(x) cannot have
an abrupt change of direction at an exceptional point, because, by Lemma II,
a-*-o {_
la 2a J
even when is an exceptional point ; that is to say
fF'(+0) = F'(f-0).
Hence the constants A and B cannot be discontinuous at the exceptional
points, and so they have the same values for all values of x in the
interval (0, -n-).
H* 2
- 2 -*-^ '-
= A\ogx + B.
"'*
r=l
m=\ if*
and the series on the left converges uniformly throughout (0, ir), so integrations
term- by- term are permissible.
Replace x by x sin 0, multiply by sin 0, tt nd integrate from to \tt. Then,
by 12 11,
&a x*-B= 2. ~ I
"j (mx sin sin 0d0
0)
s am sin ma?
f(x) = a
w=l J
From the results contained in 19*41 concerning the behaviour of the
series on the right at a? = and at #= ir, it is evident that f(x) cannot be a
convergent Schlomilch series at either point unless a is zero; and this proves
the theorem stated at the beginning of this section.
fro,
" am Jv (mx) + w H, (mx)
fr
We call this interval /, ; and, at all points of /,, we have, by 1041 (4),
am Jv (mx) + bm H r (mx)
* = 71
cm rD/
[P (w*,
.
v) cos
,
(mx - \vir
,
-\ir- n m )
, v
.\ 5 yii
+i */'jr
($mxy (%mxy /
We now suppose that a m and bm are not both o(w + *); we have to shew
that this hypothesis leads to a contradiction.
If am and b m are not both o (m" + *), a positive number e must exist such
that
cm >em v+i
whenever m is given any value belonging to a certain unending sequence
m x , wio, m 3 ,....
646 THEORY OF BBSSEL FUNCTIONS [CHAP. XIX
are satisfied whenever w has any value belonging to a sequence (mr ) which
is a sub-sequence of the sequence (mr ).
a m Jv (ml) + bm H (mX)
qmxy >m&-***>
and, since v \ is negative, the expression on the right cannot be arbitrarily
small. This is the contradiction which is sufficient to prove that am and bm
must both be o(ra"+*) if the (m + l)th term of the Schlomilch series tends to
zero at all points of J,.
The reader may now prove (as in 19 5) that, when v <\,tf the generalised
Schlomilch series converges throughout any interval, the necessary and sufficient
condition that it may converge for any positive value of x {whether a point of
the interval or not) is that the series
2 a \
Hcos(ra# kwr \ir vm )
4i/
2
Smx
-l
Bin (mx - W l-rr - Vm )\ + - .
r(l + i)r(i) ]
,
Let us take < v< \ , and let the sum of the series
a | am Jv (mx) + bm H , (mx)
r(i/+l) =i ($mxy
at any point at which the series is convergent be called f, (x).
Let
m
{)
*"
v_
W ~8I>-|-2)
/ floff
2
mti
5 a m J, (mx) + bm H (mx)
m\(tmxy
Then F(a?) will be called the function associated with the Schlomilch series
whose sum isf (x).
It is easy to prove that, if the series defining r (x) converges at all points f
of any interval, then the series defining F (x) converges for all real values of x.
The only respect in which the proof differs from the analysis used in
19*51 is that the additional theorem that H r (x^x" is a bounded function of
the real variable x has to be used.
19-61, 19-62] SCHLOMILCH SERIES 647
Again, let
at any point x at which the series defining fy (x) is convergent, provided that
ao
am and bm are both o (m -"'*) and that the series 2 bm/m is convergent.
provided only that am and bm are both o (to" + *), whether the series X bmfm is
m=l
convergent or not.
We can now prove that, if two generalised Schl&milch series of the same
order v {where \ < v < ^) converge and have the same sum-function at all
points of the closed interval ( ir, ir) with the possible exception of a finite
number of points supposed that the origin and the points tt are not all
(it is
exceptional points), and if the coefficients of the terms containing Struve's func-
tions in the two series are sufficiently nearly equal f each to each, then all
corresponding coefficients in the two series are equal.
Let the difference of the two series be
_Joo__ | am Jv (mx) + bm H.,(mx)
I> + 1) ti ($mxy
and let the sum of this series be f (x), so that the series for / (x) converges
to zero at all points of the interval ( tt, tt) with a finite number of exceptions.
The convergence of the series for / (x) nearly everywhere in the interval
( 7r, 7r) necessitates the equations
am = o (m"+*), bm = o (m"+i).
The statement that the coefficients of the terms containing Struve's func-
tions in the two be sufficiently nearly equal
series are to is to be interpreted to
L.
mean that 6m --Oasm-*oo in such a way that i. is convergent.
We now discuss the function F(a") associated with the Schlomilch series
for / (x). It can be proved J that if the interval (,, f2 ) is such that the origin
* The presence of the series on the right is due to the lack of homogeneity in the differential
equation satisfied by Struve's function.
f This statement will be made definite immediately.
t It seems unnecessary to repeat the arguments already used in 19*54.
648 THEORY OF BESSEL FUNCTIONS [CHAP. XIX
and the exceptional points (if any) are not internal points of the interval, then
F,(#) is a linear function* of a? - 2" in the interval. It may then be shewn "
that the exceptional points do Dot cause any discontinuity in the form of Fr (x),
and hence we deduce that
iT v (x) = Ax-*" + B, (0<x%ir)
\Fv {x) = A'\x\-" + B ,
> (0>x>-tt)
where A, B, A\ B' are constants.
p m = 8I>a^+
" K
> 2)
1 aw Jv (mx) + bm H (mx)
mZx m\{\mxy
replace x by #sin0, multiply by sin2"* 1
0/cos2v 0, (which has an absolutely
convergent integral) and integrate from to \ir. The series for Fr (#sin0)
converges uniformly in this interval of values of 0, so term-by-term integra-
tions are permissible.
f**-, ,
F (X Sin 0) -
^ sin "g-
+1 2
T/
0,0 =
, cue* fi'8m iv+3
^
5^=-^ d0
Jo cos "0 2
(i/ + 2)J
8r(i/ cos2 "0
when 0^#^7r ; and a similar equation may be obtained when O^x^ tt.
If we differentiate, twice oronce as the case may be, the resulting series
on the left tends to a limit as x -* 0, but the resulting expression on the right
fails to do so unless A is zero.
We infer that A = 0, and in like manner A' must be zero the continuity ;
(2)
Jr -^-BxTiv+l)
when 7r $ x ^ ir.
* When v is zero x~iv has to be replaced by logx.
19*7] SCHIidMELCH SERIES 649
a +
lr(r+i) mt 1 a^y J-
Now at least one of the points 0, - 7r
not an exceptional point and
ir, is ;
the series for/r (a?) cannot converge at that point unless a* is zero, so that am
is also zero.
We
have therefore proved that, if the series 2 bm/m is convergent all the
am and bm must vanish that is to say, the two Schlomilch series
coefficients ;
with which we started must have corresponding coefficients equal. And this
is the theorem to be proved.
We shall now give a very brief sketch of the method by which the series
discussed in Chapter xviii, namely
(in which v > ) may be investigated after the manner of Riemann's investi-
gation of trigonometrical series.
The method
is identical with the method of investigation of Schlomilch
given in 19 6 1962, though there are various points of detail*,
series just
which do not arise in the case of Schlomilch series, due to the fact that m
j
and Xm are not linear functions of m.
These points of detail are very numerous and there is no special difficulty in discussing any
*
of them but it is a tedious and lengthy process to set them out in full, and they do not seem to
;
be of sufficient importance to justify the use of the space which they would require. The reader
who desires to appreciate the details necessary in such investigations may consult the papers by
C. N. Moore, Trans. American Math. Soc. x. (1909), pp. 391435; xn. (1911),
pp. 181206;
xxi. (1920), pp. 107156.
650 THEORY OF BESSEL FUNCTIONS [CHAP. XIX
In the first place, it is easy to prove by the method used in 195 that
if the series
00 00
2 OmJ.ijmti), 2 bmJyQ^x)
f(x) = 2 am Jv (jm x\
m=l
and then the function associated with /(a?) is defined by the equation
It may be proved that, when x has any positive value for which the series
defining /(a?) is convergent, and if the expression
^-s l(x + 2va + a) F(x + 2a) - 2x F(x) + (x~ 2v* -)F(x- 2a)]
isarranged as a series in which the rath term has am for a factor, then the
latter series is uniformly convergentwith respect to a in an interval containing
the point o = 0, and that its limit when a -* is ^""/(a?).
It may also be proved that, whether the series for f(x) converges or not,
the condition that am =*o (*/m) is sufficient to ensure that
The proofs of these theorems depend on a number of lemmas such as the lemma* that
?2* - rt ain2/
2 8 2
m=l! Jm+1"' fm<**
!
is a bounded function of a ;
proofs of the lemmas can be constructed on the lines of the
It now follows in the usual manner (cf. 19'54) that, when/() is a null-
function throughout the interval (0, 1), then F(x) satisfies the differential
equation
For any assigned value of n multiplying the series for F(x) by x+1 J,(j x)
n
does not destroy the uniformity of its convergence; and, when we integrate,
we find that
o J? 0) = j 2
J
{Ax* + Bar 9 ) x Jv (jn x) dx
J
=jn B ~ (A + B) J: n)
[ '^rW ]
*
i
j''->i-V(),
and so the formula just obtained for an is inconsistent with the equation
n = o (00 when v > unless both A +B and B are zerof; and then an is
zero.
Dini's series
00
may be dealt with in the same manner. The associated function is defined
by the equation
F(x) = i ^J^zl,
* When *>}, the convergence of the series for /(a)/*" at.r=0 is
sufficient to ensure the
uniformity of the convergence.
t An exception might occur when v - 1= - ; but this is the trigonometrical case.
% The series divided by x" then has to converge at the origin.
It is supposed for the present that H+v>0, so that no initial term is inserted.
652 THEORY OF BESSEL FUNCTIONS [CHAP. XIX
and it is inferred that when f(x) is a null-function throughout (0, 1), then
constants A and B exist such that
F(x) = A + Bar*,
bn (l - ) /* (\n) + JJ 2 (X.)] = v/^ (A*P + Bar*) xJv (X) efo
and 5= when v ^ 0.
- \ r^^+1 (\) + b
f^p - ^. <*)}] 5
-
) ^ + J* (Xn)] = 4\ J,+1 (X).
2
6n [(l 2
(X.)
unless A is zero, since /+, (\) is not zero and then bn is zero. ;
b Xp + 2 bmJyCKnX),
w=l
oo
6 / (\ x) + 2 bm J, (\m x),
respectively.
In the second of the two cases the previous arguments are unaffected by
the insertion of an initial term the first of the two cases needs more careful
;
bn JS (\) = X 2
J
[Ax* + ^f^ + Bar*} xJ {\nx) dx
19-7] SCHLOMILCH SERIES 653
VIII and XV that a large part of the theory of Bessel Functions has been con-
structed expressly for the purpose of facilitating numerical computations
connected with the functions. To the Mathematician such computations are
of less interest and importance* than the construction of the theories which
make them possible; but to the Physicist numerical results have a significancef
which formulae may fail to convey.
The reader will not be concerned with the monotony and technical irrele-
vance of this Chapter when he realises that it can be read without the efforts
required to master the previous chapters and to amplify arguments so ruth-
lessly condensed.
The first Tables of J (*) and J (x) were published by Bessel himself in
x
The function Ji{x)jx was subsequently tabulated to six places of decimals by Lommel,
und Phy. xv. (1870), pp. 164167, with a range from x-^0 to #=20*0
ZeiUchrift fiir Math,
with interval 0*1 this Table, with a Table of JJ (a?)/^2 was republished by, Lommel,
; ,
* Cf. Love, Proc. London Math. Soe. (2) xrv. (1915), p. 184.
" in formulas unless I feel their arithmetical
f Cf. Lord Kelvin's statement I have no satisfaction
magnitude at all events when formulas are intended for definite dynamical or physical problems."
Baltimore Lectures (Cambridge, 1904), p. 76.
% Tram. Camb. Phil. Soc. v. (1835), p 291. A Table of 2JX (x)/x and its square, to four or five
places of decimals, in which the range is from to the circular measure of
1125 (with interval 15) :
/"i (x) to six places of decimals with a range from x = to x = 10-0 with
interval 01
was published in his Ermittelung der absoluten StOrungen
; this
in Ellipsen von beliebiger Excentricitdt
und Neigung (Gotha, 1843). Hansen's
Table was reprinted by Schlomilch* and also by Lommelf who extended it
toa: = 20.
A Table of J (ntr) to six places of decimals for n = l, 2, 3, ..., 50 has been computed by
Nagaoka, Journal of the Coll. of Sci. Imp. Univ. Japan, iv. (1891), p. 313.
The value of J (40) was computed by W. R. Hamilton from the ascending series, Phil.
Mag. (4) xiv. (1857), p. 375.
A Table of Ji(x) to six places of decimals from ^=201 to # = 41 with interval 0*1 or
0-2 has been published by Steiner, Math, und Naturwiss. Berichte aus Ungarn, xi. (1894),
pp. 372373.
were given to ten places of decimals for a range from x = to x = 6'0 with
interval 0*2 and from x = 6*0 to x= 160 with interval 05.
Shortly after the appearance' of Smith's Table, an elaborate table was con-
structed by Aldis, Proc. Royal Soc. lxvi. (1900), p. 41, of Heine's functions
O (x) and Gx (x) to twenty-one places of decimals; the reader should be
* Zeitschrift
far Math, und Phijs. n. (1857), pp. 158165.
t Studien iiber die Bessel'schen Functionen (Leipzig, 1868), pp. 127 135.
X Meissel's Table contains a misprint, the correct value of J (0-62) being +0-90618..., not
+ 0-90518.... An additional misprint was made in the reprint of the Table by Gray and Mathews.
These functions were also tabulated by B. A. Smith, Phil. Mag. (5), xlv. (1808), pp. 122
123 ; the scope of this table is the same as that of his Table of I'M (x) and I'*
1
) (,c).
656 THEORY OF BESSEL FUNCTIONS [CHAP. XX
reminded that these functions are equal to \irY (x) and |w Y t (x) re-
spectively. The range of Aldis' Table is from x = to x = 6'0 with interval 0*1.
The functions Y (x) and Yt (x) are tabulated to seven places of decimals
in Table I infra ; this table has an appreciable advantage over the British
Association Tables*, in that the auxiliary tables make interpolation a trivial
matter; in the British Association Tables interpolation is impracticable.
By means of the recurrence formulae combined with the use of the tables
which have now been described, it is an easy matter to construct tables of
functions whose order is any integer. Such tables of /", (x) were constructed
by Hansen and reprinted by Schlomilch and Lommel after their Tables of
J (x) and J (x). Subsequently Lommel, Miinchener Abhandlungen, XV. (1886),
x
pp. 315
316, published a Table of Jn (x) to six places of decimals, in which
n = 0, 1, 2, ..., 20, and x - 0, 1, 2,..., 12; this Table is reprinted in Table IV
infra, pp. 730
731. A Table of Jn (x) of practically the same scope was also
published by Meissel, Astr. Nach. cxxviii. (1891), col. 154 155.
A much more extensive Table of Jn (x) was computed by Meissel, but it
seems that he never published it. He communicated it to Gray and Mathews
for publication in their Treatise, pp. 267
279. This table gives Jn (x) to
eighteen places of decimals when n = 0, 1, 2, ... , 60, and x = 0, 1, 2, .... 24.
Some graphs of Jn (x) were constructed, with the help of the last-mentioned
table, by Hague, Proc. Phys. Soc. xxix. (1917), pp. 211 214.
The corresponding Tables of functions of the second kind are not so ex-
tensive.
* In the coarse of computing Table I, a small part of the British Association Table of G (x\
and 6, (x) was checked, and the last digits in it were found to be unreliable in about 5 / of the
entries checked.
t For the larger values of n the functions are not tabulated for small values of x.
20,1 J TABULATION OF BESSEL FUNCTIONS 657
Functions of the first kind with imaginary argument have been tabulated
in the British Association Reports. The Report for 1896,
pp. 99149, con-
tained a Table of I (x) to nine places of decimals, its range being from x=0
to x = 5-100 with interval 0001. A
Table of /, (x) of the same scope had been
published previously in the Report for 1893, pp. 229 279; an abridgement
of this (with interval 001) was given by Gray and Mathews in their Treatise,
pp. 282284.
I (x) and 1^ (x) to twenty-one places of decimals have been con-
Tables of
structed by Aldis, Proc. Royal Soc. lxiv. (1899), p. 218. The range of these
Tables is x = to x = 60 with interval 01; Aldis also gave (ibid. p. 221) the
values of I (x) and I (x) for x = 7, 8, 9, 10, 11.
x
Extensive tables connected with I (x) and I\ (x) have been published by
Andmg,Sechsstellige Tafeln der Bessel 'schen Funktionen imagindren Arguments
(Leipzig, 1911). These tables give log ]0 1 (x) and log 10 /, (x)/x] from #=0 {
to x= 1000 with interval 001. They also give the values of the functions
vX27nr) <r*. I (x), ^(2irx) e~*Ix . (x), log 10 [Jx IQ (x)) and log10
. {y/x /, (*)}
.
Table II infra, pp. 698 713, gives the values of e~ x IQ (x) and er* J, (a?);,
these have been computed, for the most part, by interpolation in Aldis' Table.
A brief Table of ber (x) and bei (x), computed by Maclean, was published
by Kelvin, Math, and Phys. Papers, ill. (1890), p. 493. Tables of J (xy/i)
and y'2 J (x \fi) to twenty-one places of decimals have been constructed
. x
by Aldis, Proc. Royal Soc. lxvi. (1900), pp. 42 43; their range is from
x= to x = 60 with
These are extensions of the Table of
interval O'l.
J (x *Ji) to nine places of decimals for the range from x = to x = 60 with
interval 02 published in the British Association Report, 1893, p. 228, and
reprinted by Gray and Mathews in their Treatise, p. 281.
Tables of ber(#), bei (a;), ker(#) and kei (x) to four significant figures for
x= 1, 2, 3, ...,30, have been published by Savidge, Phil. Mag. (6) xix. (1910),
p. 53.
The functions ber (x), bei (x), ber' (x) and bei' (x) are tabulated to nine
places of decimals, from x = to x = 10*0 with interval 01 in the British
Association Report, 1912, pp. 57
68; and a Table of ker (x), kei (x), ker' (x)
and kei' (x) of the same scope (except that only six or seven significant figures
were given) appeared in the Report for 1915, pp. 36 38. Tables of squares
and products of the functions to six significant figures from x = to x= 100
with interval 02 were given in the Report for 1916, pp. 118 121.
The functions </ (+$)(#) have been tabulated to six places of decimals by
Lommel, Milnchener Abh. xv. (1886), pp. 644 647, for n 0, 1, 2, ...,6 with
x = 1 2, 50, and (in the case of functions of positive order) n = 7, 8,
, . .
.
, 14 . . . ,
20-1] TABULATION OF BESSEL FUNCTIONS 659
Table V infra, pp. 740 743. A Table of the same functions to four places of
decimals with n =0, 1, 2 and from a; = to x = 80 with interval 02 is given
by Dinnik, Archiv der Math, und Phys. (3) xx. (1913), pp. 238240.
Functions related to J(n +h)( x) have recently been tabulated in the British
Association Reports. The notation used is
s/Qirx) . Jn+i (x) = Sn (x), i-Y^irx) . J_ n . h (x) = Cn (x),
En (x) = \Cn (x) + iSn (x)\,
and the functions tabulated are Sn (x), Cn (x), En (x), 2
Sn'(x), Cn'(x), En'*(x),
and their logarithms. In the Report for 1914, pp. 88 102, the functions are
tabulated to seven significant figures for n = 0,1, 2, ...,17 and x = 1,2,3, ...,10,
and in the Report for 1916, pp. 97107, for n = 0, 1, 2, ..., 10 and x =11,
12,..., 19.
Functions of order + J, , have been tabulated by Dinnik, Archiv der
Math, und Phys. (3) xvni. (1911), pp. 337 338, to four places of decimals;
the functions tabulated are T(l + J) Ji (x), T(l %)J$(x) from = to
x = 8*0 with interval 0*2; and Dinnik has also tabulated I(x), I$(x), ibid.
(3) xxii. (1914), pp. 226227 and Ji (x), Ji (x), ibid. (3) xxi. (1913),
pp. 326. All these
324 have the range x = x = 8 with
tables to interval
02. The Tables of T(l + $)Ji(x) are extensive than Table III less infra,
pp.714 729 with the exception of Dinnik's
; but, there tables, exist no
tables of functions of orders f, J and f
In connexion with functions of order + , Airy's Table of his integral
of the zeros of J (x) and */, (x) is due to Stokes, Trans. Camb. Phil. Soc. ix.
(1856), p. 186 [Math, and Phys. Papers, n. (1883), p. 355]. Stokes gave the
values of the first twelve zeros (divided by if) of J (x) and J, (x) to four places
of decimals. In the same memoir he gave the first fifty zeros of Airy's integral,
and the first ten stationary points of this integral.
The first nine zeros of J (x), J (x), ...,J (x)
t S were computed by Bourget,
Ann. sci. de Vficole norm. sup. ill. (1866), pp. 82
87. Bourget's results are
given to three places of decimals ; some corrections in his Tables have recently
been made by Airey*.
* Phil. Mag. (6) xxxii. (1916), pp. 714.
660 THEORY OF BESSEL FUNCTIONS [CHAP. XX
Bourget's Tables have been reprinted so frequently that their authorship has been
overlooked by the writers of the articles on Bessel Functions in the Encyclopiidie der
Math. Wiss. and the Ewrgclopedie de.i Set. Math.
The first five zeros of J (x) and J2 (x) were given to six places of decimals
x
by Lommel, Zeitschrift fur Math. und. Phys. xv. (1870), p. 167 and Munchener
Abhandlungen, XV. (1886), p. 315.
The first ten zeros of J (x) were computed to ten places of decimals bv
Meissel, Berliner Abhandlungen, 1888.
The first fifty zeros (and their logarithms) of J (x) were given to ten places
of decimals by Willson and Peirce, Bulletin American Math. Soc. Hi. (1897),
pp. 153 155; they also gave the values of J (x) and log \Ji{x)\ at these
x
values 0, , 1, f, 2, f.
* Jahrbitch iiber die Fortschritte der Math. 1890, p. 521. In consequence of the inaccessibility
of Meissel's table, the zeros of J\ (x) were recomputed (to ten places of decimals) for insertion in
Table VII, p. 748.
20*2] TABULATION OF BESSEL FUNCTIONS 661
Ji(t)dt;
J o
a different form was tabulated earlier by Lindstedt, Ann. der Physik und
Ch&mie, (3) xvn. (1882), p. 725.
J X
and writing x = {(y + ) tt}^ Lindstedt tabulated M(x) and N (x) to six places
of decimals from y = to y = 9 with interval O'l.
-
if Hi 2(2) rf*
t
has been tabulated to four places of decimals by Struve, Ann. der Physik und
Chemie, (3) xvn. (1882), pp. 10081016, from x = to 4'0 (interval 01), from
x = 40 to 70 (interval 02) and from x = 7*0 to 11-0 (interval 0'4).
J #
in which the limits are consecutive zeros (up to the forty-ninth) of Jx
(x), has
been published by Steiner, Math, und Naturwiss. Berichte aus Ungarn xi.
(1894), pp. 366
367 this integral occurs in the problem of Diffraction by
;
a Circular Aperture:
No Tables of Struve's functions- seem to have been constructed before the
Table of Ho (x) and Hi (x) which is given on pp. 666 697.
20*2. Description of the Tables contained in this book.
(I) That space did not usually admit of the inclusion of more than seven
places of decimals in the tables.
an error of a hundredth of a second does not affect the value of the sine by
more than 0*00000005. Hence, for seven-figure accuracy, it was considered
adequate to compute to nine places of decimals the sines (or cosines) of the
angles tabulated and then to compute the angles from Gifford's Natural Sines
(Manchester, 1914) ; these are eight-figure tables with an intervalf of 1".
The angles tabulated may consequently frequently be in error as to the
last digit, but, in all probability, the error never exceeds a unit (i.e. a
hundredth of a second of arc).
were computed partly by interpolation in Aldis' Table of G (x) and Gx (x) and
partly from the asymptotic expansions of J 2 (x) + F 2 (x) and J^ (x) + Y? (x)
given in 7*51.
The auxiliary tables give the values of \H n w(x)\ and arg {1)
n (x) for H
n = and n=l. In these tables the first differences are sufficiently steady
(except for quite small values of x) to enable interpolations to be effected
with but little trouble on the part of the reader ; thus, when x is about 10
the second differences of |
H W (x) do not exceed 0*0000009.
\
The values j|
of |
Hn
and arg H (x) can consequently be computed
ll)
(x) j
{1)
by the reader for any value of x less than 16, with the exception of quite
small values. The corresponding values of Jn (x) and Yn (x) can then be
calculated immediately by the use of seven-figure logarithm tables.
* The and it was found that the
tables were di. T erenced before removing the last three figures,
ten-figure results were rarely in errorby more than a unit in the tenth place so it is hoped that the ;
number of errors in the last figure retained does not exceed about one in every thousand entries.
tNo tables with a smaller interval have been published; the use of any tables with a larger
interval and a greater number of decimal places would have very greatly increased the labour of
constructing the auxiliary tables of angles, and the increased accuracy so obtained would be of no
advantage to anyone using the auxiliary tables for purposes of interpolation.
X I must here express my cordial thanks to the Preussische Akademie der Wissenschaften zu
Berlin for permitting me to make use of this Table.
The idea of constructing the auxiliary tables grew out of a conversation with Professor Love,
in the course of which he remarked that it was frequently not realised how closely Bessel functions
of any given order resemble circular functions multiplied by a damping factor in which the rate
of decay is slow.
I| The remarks immediately following of course presuppose that n is or 1.
20*2] TABULATION OF BESSEL FUNCTIONS 663
of a point in a plane ; then the Cartesian coordinates of this point are Jn (x)
and 7 (x). Thus, from the entry for x = 800,
+ 01716508 = 0-2818259 cos 412 28' 40"60,
+ 02235215 = 02818259 sin 412 28' 40"60.
Table I also contains the values of Struve's functions (x) and H H x (x).
These functions are included in Table I, instead of being contained in a
separate Table, to facilitate interpolation; by 10'41 (4), the difference
Hn (x) Yn (x) is a positive monotonic function, and it varies sufficiently
steadily for interpolation to be easy when x is not small.
The Tables of Struve's functions were computed by calculating the values of (x) H
and Ho'(ar) directly from the ascending series when x=\, 2, 3,..., and then calculating
Ho"(#), Ho"' (#),... for these values o( x from the differential equation 10-4(10) and the
equations obtained by differentiating it.
A few differential coefficients are adequate to calculate Ho (x) and Ho' (x) by Taylor's
theorem for the values 0-5, 0'6, 07, ... of x. Interpolation to fiftieths of the unit is then
effected by using Taylor's theorem in the same manner. This process, though it seems at first
sight to be complicated and lengthy, is, in reality, an extremely rapid one when a machine*
is used. It is very much more effective than the use of asymptotic expansions or the
process suggested in the British Association Report, 1913, p. 116. As an example of the
rapidity of the process, it may be stated that thewalues ofe~ x I (x) and e~ x Ix (x) in Table II
took less than a fortnight to compute of course the time taken over this tabulation was
;
Table II consists of Tables of e-* I (x), e~x Ij (x), e? (x), and e x A", (.r), K
and a Table of e* is inserted, in case the reader should require the values of
the functions'^ I (x), Ix (x), (x) and K K
(x); the functions are tabulated
x
Trans. Camb. Phil. Soc. XIII. (1883), pp. 145241. Unlike the other Tables
in this book, the Table of e* is given to eight significant figures J, and care
has been taken that the last digit given is accurate in every entry. Interpo-
lation in this Table is, of course, effected by multiplying or dividing entries
by exponentials of numbers not exceeding O'Ol ; such exponentials can be
calculated without difficulty.
* The machine on which the calculations were carried out is a Marchant Calculating Machine,
10 x 9 recording to 18 figures.
t These functions were not tabulated because tables of them are uusuited for interpolation.
X Nine figures are given in parts of the Table to avoid spoiling its appearance.
664 THEORY OF BESSEL FUNCTIONS [CHAP. XX
Newman's Table gives e~ x to a large number of places of decimals, but the actual
number of significant figures in the latter, part of the Table is small and less than half of ;
the Table of e* was constructed by the process of calculating reciprocals the rest was con- ;
structed from Newman's Table by using the values of e13 and e 16 given by Glaisher* and
the value of e -15,6 given by Newman in a short table of e~ x with interval 0-1. These ex-
ponentials were employed because the tenth significant figures in all three and the eleventh
significant figures in the first and third are zero.
of the same scope as Table I, and interpolations are effected in the manner
already explained. A Table of e* K$ (x) is also included. These Tables are
of importance in dealing with approximations to Bessel functions of large
order ( 8'43), and also in the theory of Airy's integral.
The reader can easily compute values of J_ 4 (x) from this table by means
of the formula
J_j (x) = |
H^ (x) |
cos {60 + arg H^ (x)}.
Table IV gives the values of Jn (x), Yn (x), e~x In (x) and Kn (x) for various
values of x and n. The values of Jn (pc) are taken from Lommel's Table t, with
some but the remainder of Table IV, with the exception of some
corrections,
values of K
n (x) taken from Isherwood's Table f, is new they have been con- ;
structed in part by means of Aldis' Tables of functions of orders zero and unity.
Table V
Lommel's Tablet of J (w+i )(a;) and Fresnel's integrals with
is
Table VII gives the first forty zeros of Jn (x) and Yn (x) for various values
of n; part of this Table is taken from the Tables of Willson and Peircef.
Forty zeros of various cylinder functions of order one-third are also given.
Table VIII gives the values of
X X
li J (t)dt, I J Y (t)dt,
from x = to 50 with interval 1, together with the first sixteen maxima and
minima of the integrals. The former table of maxima and minima can be
used to compute the coefficients (cf. 18'12) in certain Fourier-Bessel series
for which v = 0.
* Trans. Camb. Phil. Soc. xni. (1883), p. 245.
t I must here express my cordial thanks to the Bayerische Akademie der Wissenschaften zu
Miinchen, to the Manchester Literary and Philosophical Society, and to the American Mathe-
matical Society for permitting me to make use of these Tables. The non-existence of adequate
trigonometrical tables of angles in radian measure has made it impracticable to check the last
digital in the entries in the greater part of Table V.
TABLES
OF
BESSEL FUNCTIONS
1
X ;.w Y9 (x) (
|tf ?(*)l
(
arg H J{x)
{1
HW X
X /(*) Y (x)
(
\H l\x)\ axgH (x) H (*) X
398
-
-
0-4004218
03994854
03983943
- 0-0009343
- 0-0089594
0-4004853
03994865
03984951
180 V
181 17' 17*86
2*36 +
+
0-1524412
0-1437027
394
396
3-98
4-00 - 03971498 - 0-0169407 o-3975"o 182 26' 33*05 + 0-1350146 4-00
1
X /oW Y {*)
(
- + 0-I263794
4-02
4-04
- 0-3957530
- 0-3942053 -
0-0248755
0-0327610
0-396534
0-3955643
^o 35 47*96
',
$
4-68
-
-
-
-
0-2856866
0-2803228
0-2748700
-
-
-
-
0-2341813
0-2393683
o-2444376
0-3694018
0-3686164
0-3678359
219 20'
220 29'
221 38'
222 47'
31*03
38*75
46*29
- 0-1018790
- 0-1075998
- 0-1131987
- 0-1186742
1%
4-68
470 0-2693308 0-2493876 03670603 53*65 4-70
492 - 031 75386 + 0-1746449 0-3623971 151 11' 21*79 + 0-8353045 4-92
494 - 0-3202495 + 0-1680031 0-3616418 *5 2 *?', 6 "A7 + 0-8284820 4-94
4-96 - 0-3228269 + 0-1613238 0-3608913 153 26 5 J *6i + 0-8216241 4-96
4-98 - 0-3252702 + 0-1546097 0-3601456 154 34; 37*22 + 0-8147332 4-98
5-00 - 0-3275791 + 0-1478631 0-3594045 155 42 23-28 + 0-8078119 5-00
676 TABLES OF BESSEL FUNCTIONS
X 7o(*) Y (x)
{
\H l\x)\ arg //(*) H(*) X
X /iW Y l
(x) i^J'wi argH^x) Ht(*) X
X /(*) Y (x)
K'(*)l arg H"H( X ) Ho(*) X
X /iW Y 1
(x)
{
\H \\*)\ &rgH t
(x) H,W X
HoW X
z 9
8
8-oo + 0-1716508 + 0-2235215 0-2818259 4 1 2 28 40*60 + 0-3019881 8-00
TABLES OF BESSEL FUNCTIONS 681
X /iW Y 1
(x) i^i'wi argtf, (x) H^*) X
AW Yt(*) \rf
l
M\ arg H l\x)
{
H,W
9-02 + 0-2429143 + 0-1090553 0-2662713 384 io' 38*94 + 0-7532504 9-02
904 + 0-2404263 + 0-1137425 0-2659740 3|5 19' 5-56 + 0-7579051 904
9-06 + 0-2378489 + 0-1183744 0-2656776 386 27' 32*26 + 0-7625048 9-06
908 + 0-2351833 + 0-1229495 0-2653823 387^35'59'o 4 + 07670477 908
9-10 + 0-2324307 + 0-1274659 0-2650879 388 44' 25*90 + 07715322 9-10
X 7iW i\0) \H 1
(x)\ arg H j
(x) HW X X
X /oW nw 1*0*) 1
argH {x) Ho(*) X
/iW Y l
(x) 1*0*) I
arg H {
\
]
(x) HiW
15-02 + 0-2045057 + 0-0251476 0-2060460 727 ' 37*25 + 0-6645540 I5-02
15-04 + 0-2038267 + 0-0292062 0-2059085 728 9'i5 7 74 + 0-6606052 I5-04
15-06 + 0-2030675 + 0-0332478 0-2057713 729 1 7 54*25 + 0-6726395 15-06
15-08 + 0-2022286 + 0-0372707 0-2056344 73o 26 32^77 + 0-6766552 I5-08
15-10 + 0-2013102 + 0-0412735 0-2054977 73i 35 n
? 3i + 0-6806508 I5-IO
15-12 + 0-2003130 + 0-0452546 0-2053613 732 43' 49-87 + 0-6846246 I5'I2
15-14 + 0-1992373 + 0-0492124 0-2052252 733 52' 28*45 + 0-6885753 15-14
15-16 + 0-1980838 + 0-0531454 0-2050893 735 1 7-04 + 0-6925011 I5-l6
15-18 + 0-1968530 + 0-0570521 0-2049537 736 9 45-65 + 0-6964006 I5-I8
15-20 + 0-1955454 + 0-0609309 0-2048184 737 18' 24-28 4- 0-7002724 I5-20
15-22 + 0-1941618 + 0-0647803 0-2046834 738=27' 2*93 + 0-7041148 I5-22
15-24 + 0-1927027 + 0-0685990 0-2045486 739 35 4**59 + 0-7079263 I5-24
15-26 + 0-1911688 + 0-0723853 0-2044141 740 44 20*27 + 0-7117056 I5-26
15-28 + 0-1895608 + 0-0761370 0-2042798 741*52 58*97 + 07I545I2 I5-28
1530 + 0-1878794 + 0-0798551 0-2041459 743 1 37-69 + 0-7191616 I5-30
15-32 + 0-1861255 + 0-0835358 0-2040121 744 10' 16*42 + 0-7228353 1532
15-34 + 0-1842998 + 0-0871785 0-2038787 745 i8' 55*17 + 0-7264711 r 5-34
15-36 + 0-1824032 + 0-0907816 0-2037455 746 27 33*94 + 0-7300674 15-36
15-38 + 0-1804363 + 0-0943440 0-2036125 747 36' 12*72 + 0-7336229 I5-38
1540 + 0-1784003 + 0-0978642 0-2034799 748 44 51-52 + 07371363 15-40
15-42 + 0-1762958 + 0-1013409 0-2033475 749 53' 30 -34 + 0-7406061 I5-42
15-44 + 0-1741239 + 0-1047727 0-2032153 751 2 9*17 + 07440312 I5'4^
15-46 + 0-1718855 + 0-1081584 0-2030834 752 io' 48*02 + 0-7474101 15 -4
15-48 + 0-1695816 + 0-1114960 0-2029518 753 19' 26*89 + 0-7507416 I5-48
15-50 + 0-1672132 + 0-1147861 0-2028204 754 28' 5 "77 + 0-7540245 15-50
15-52 + 0-1647812 + 0-1180258 0-2026892 755 36' 44 -67 + 0-7572574 1552
15-54 + 0-1622868 + 0-1212142 0-2025584 756 45 23*59 + 0-7604392 15-54
I5-56 + 0-1597310 + 0-1243503 0-2024278 757 54 2*52 + 0-7635687 I5-56
I5-58 + 0-1571149 + 0-1274329 0-2022974 759 2 4i*47 + 0-7666447 I5-58
15-60 + O' 1 544396 + 0-1304608 0-2021673 760 11 20*43 + 0-7696660 15-60
15-62 + 0-1517062 + 0-1334329 0-2020374 761 19' 59*41 + 0-7726316 15-62
15-64 + 0-1489160 + 0-1363480 0-2019078 762 28' 38 -41 + 07755402 15-64
15-66 + 0-1460700 + 0-1392052 0-2017784 763 37' 17-43 + O7783909 15-66
15-68 + 0-1431695 + 0-1420033 0-2016493 764 + 0-7811825
45 56-46 15-68
15-70 + 0-1402157 + o-i4474i3 0-2015204 765 54 35*5<> + 0-7839140 15-70
15-72 + 0-1372099 + 0-1474181 0-2013918 767 3'i4''56 + 0-7865845 1572
15-74 + 0-1341533 + 0-1500329 0-2012634 768 11' 53 ''64 + 0-7891929 15-74
15-76 + 0-1310471 + 0-1525847 0-2011353 769 20' 32 "73 + 0-7917383 15.76
15-78 + 0-1278927 + 0-1550724 0-2010074 770 29' 11-84 + 0-7942197 15-78
15-80 + 0-1246913 + o-i5 74953 0-2008798 77* 37' 50-96 + 0-7966362 15-80
15-82 + 0-1214444 + 0-1598524 0-2007524 772 46' 30*10 + O7989870 15-82
15-84 + 0-1181532 + 0-1621429 0-2006252 773 55' 9*25 + O-0OI27I2 15-84
15-86 + 0-1148192 + 0-1643659 0-2004983 775 3' 48 -42 + 0-8034880 15-86
15-88 + 0-1114436 + 0-1665207 0-2003717 776 12' 27*61 + 0-8056365 15-88
15-90 + 0-1080279 + 0-1686064 0-2002452 777 21' 6-8i + 0-8077161 15-90
15-92 + 0-1045735 + 0-1706224 0-2001190 778 29' 46 -03 + 0-8097259 15-92
15-94 + 0-1010818 + 0-1725680 0-1999931 779 38' 25*26 + 0-8116653 15-94
I5-96 + 0-0975542 + 0-1744424 0-1998674 7o 47' 4*50 + 08135336 15-96
15-98 + 0-0939922 -
+ 0-1762450 0-1997419 7i 55 43*76 + 0-8153300 15-98
16-00 + 0-0903972 + 0-1779752 0-1996167 +
j 783 4 23-04 0-817054! 16-00
23
698 TABLES OF BESSEL FUNCTIONS
Table II. Functions of imaginary argument, and ex
8-
32 0-1405488 0-1318101 0-4283748 0-4534238 4105-1600 832
34 0-1403744 0-1316683 0-4278747 0-4528359 4188-0897
8-36 0-1402006 0-1315269 0-4273763 0-4522504 4272-6948
8- 8 0-1400274 0-1313859 0-4268797 0-4516670 4359-oo89 8-38
3
8-40 0-1398549 0-1312454 0-4263848 0-4510859 4447-0667 8-40
8-82 - I
363737 0-1283929 0-4163747 0-439374 6 6768-2646 8-82
8-84 0-1362144 0-1282615 0-4159155 0-4388392 6904-9926 8-84
8-86 0-1360556 0-1281306 -4 I 54578 0-4383058 7044-4827 8-86
8-88 o-i358974 0-1280001 0-4150016 o-4377743 7186-7907 8-88
8-90 o-i357397 0-1278699 0-4145468 0-4372448 7331-9735 8-90
12-22 o- 1
153497 0-1105243 0-3550176 0-3692681 202804-96 12-22
12-24 o- 1
152533 0-1104400 o-3547329 0-3689492 206901-89 12-24
12-26 0-1151572 0-1103559 0-3544489 0-3686311 211081-59 12-26
12-28 0-1150613 o-i 102720 0-3541656 0-3683138 2I5345-72 12-28
I2-30 0-1149656 0-1101883 0-3538830 o-3679973 219695-99 1 2 30
1502 0-1038291
** An
0-1003109 03207019 0-33I3037 3335055'9 1502
1504 0-1037588 0-1002478 0-3205819 0-33I073I 3402428-5 15-04
1506 0-1036087 0-1001847 03203723 0-3308429 3471162-1 15*06
15-08 0-1036186 0-1001218 0-3201631 0-3306I32 3541284-2 15-08
15-10 0-1035488 0-1000590 o-3i99543 03303839 3612822-9 1510
15-12 0-1034791 0-0999964 0-3I97459 0-330I552 3685806-8 15-12
15-14 0-1034095 0-0999338 0-3I95379 0-3299269 3760265-0 15*14
15-10 0-1033400 0-0998714 0-3193303 0*3296990 3836227-4 15-10
15-18 0-1032707 0-0998090 0-3191231 0-3294717 3913724-4 15*18
15-20 0-1032016 0-0997468 0-3189164 0-3292448 3992786-8 15-20
o O-OOOOOOO - 00 00 -90 00
/1/3W YutW 1
tfSw arg H m (x) e*K ll3 (x)
4-22
4-24
o-3759693
0-3769073
+
+
0-0946539
0-0868845
0-3877012
0-3867919
65
6Z 1' K 8*07
8*27
0-6011045
o-5997234
4-22
4-24
+ 0-0791172 0-3858890 68 10' ? 0-5983518 4-26
426 0-3776914
8*28
34
4-28
4-28 0-3783220 + 0-0713550 03849923 69 19' 0-5969897
4-30 o-3787997 + 0-0636006 0-3841018 70 28' 8*09 0-5956368 4-3o
432
4-34
0-3791248
0-3792981
+
+
0-0558570
0-0481270
0-3832175
0-3823392
K <
71
46
7*78
7*34
0-5942930
0-5929584
4-32
4'34
4'3
4-38
0-3793201
0-3791916
+
+
0-0404134
0-0327191
0-3814669
0-3806006 K u
75 u 4
55,
6*78
6*10
0-5916327
05903159
4-36
4'38
4-40 0-3789I3 1 + 0-0250469 0-3797400 7 13 5*29 05890079 4-40
522 -
0-2537034 - 0-2393685 0-3488018 223 20' 5*OI -54 I 8952 5-22
5-24 - 0-2483807 -
0-2439417 03481386 224 29' 0-27 0-5408834 5-24
5-26 -
0-2429794 - 0-2484004 o-347479i 225 37' 55 ;46 0-5398773 5-26
5-28 "
0-2375020 -
02527435 0-3468234 226 46 50*59 0-5388767 5-28
530
02319509 - 0-2569696 0-3461714 227 55' 45 T 6 4 0-5378818 5-30
532 02263285 -
0-2610776 O-3455230 229 4' 40-62 0-5368923 532
5'34
0-2206371 -
0-2650664 0-3448782 230 13' 35 ? 53 o-5359o82
0-2148793
534
5-3 -
0-2689349 0-3442370 231 22 30*37 o-5349296 536
538 02090575 -
0-2726820 0-3435994 232 31' 25*14 o-5339563 538
5*40 0-2031741 0-2763068 0-3429653 233 40' 19*85 0-5329883 540
5-42 0-1972317 0-2798083 o-3423346 234 49' 14*49 0-5320256 5-42
5 44 0-1912327 0-2831855 0-34I7075 235 58 9*07 0-5310681 5*44
5H6 0-1851797 0-2864378 0-3410837 237 7' 3-58 0-5301157 5-46
5-48 01790751 0-2895642 03404634 238 15' 58-02 0-5291685 5-48
5'50 0-1729216 0-2925640 03398464 23924'52-40 0-5282263 5-5o
X /i/j() *W*) 1
ff Siw 1
afg HZJ W (*)
l, 3
, *
e?K*z(x) X
8-02 + 0-2572095 + 0-1146236 0-2815942 384 1' 11*30 o-43 8 9564 8-02
8-04 + o- 2 545467 + 0-1196012 0-2812445 385 10' 0*94 0-4384185 8-04
8-o6 + 0-2517890 + 0-1245187 0-2808961 386 18' 50-56 0-4378827 8-06
8-o8 + 0-2489377 + 01 293743 0-2805489 387 27' 40*16 o-4373487 8-o8
8-io + 0-2459942 + 0-1341664 0-2802030 388 36' 29*74 0-4368168 8-io
8-12 + 0-2429598 + 0-1388931 0-2798584 389 45' I9 ? 29 0-4362867 8-12
8-14 + 0-2398360 + 0-1435528 0-2795151 39o 5 4' 8*83 o-4357586 8-14
8-i6 + 0-2366242 + 0-1481438 0-2791730 392 2' 58*35 04352324 8-i6
8-i8 + 0-2333259 + 0-1526645 0-2788322 393 11' 47*84 0-4347081 8-i8
8-20 + 0-2299425 + 0-1571133 0-2784927 394 20 37*32 0-4341857 8-20
X /1/3M ^i/sto 1
H%{*) 1
*#,(*) X
X /1/3W Vi/sW 1
H, >) I *Kv{*) X
12-12
12-14
- 00433371
- 0-0387947
- 0-2249971
-
0-2291327 619 5' '
I + 0-000021 + 0-000002
2 + 0-OOI202 + 0-000175 + 0-000022 + 0-000002 I
2
3 + 0-011394 + 0-002547 + 0-000493 + 0-000084 + 0-000013 + 0-000002 3
4 + 0-049088 + 0-015176 + 0-004029 + 0-000939 + 0-000195 + 0-000037 4
5 + 0-131049 + 0-053376 + 0-018405 + 0-005520 + 0-001468 + 0-00035I S
b + 0245837 + 0-129587 + 0-056532 + 0-02II65 + 0-006964 + 0-002048 6
+ 0-000006 + o-oooooi _
4
5 + 0-000076 + 0-000015 + 0-000003 4
6 + 0-000545 + 0-000133 + 0-000030 + 0-000006 + o-oooooi 5
6
+ o-oooooi _ _
I + 0-000005 + o-oooooi I
9 + 0-000030 + 0-000007 + 0-000002 9
IO + 0-000152 + 0-000043 + 0-OOOOI2 IO
ii + 0-000628 + 0-000199 + 0-000059 II
12 + 0-002152 + 0-000759 + 0-000251 12
X VfW Y e (x) Y 7
(x) V.to Y 9 (x) X
The numbers in [ ] are the numbers of digits between the last digits given and the decimal
points. For example, the integral part of F10 (o-i) is a number containing 19 digits of which the
first 14 are given.
1 I
X iW K 9 (x) * 10 (*)
X
49
I0539552
0-8664794
3-836264
3099405
15 439946
12-252049
u
4-9
50 0-7143624 2-512278 9-758563 50
The numbers in [ ] are the numbers of digits between the last digits given and the decimal
points. For example, the integral part of K10 (o-i) is a number containing 19 digits of which the
first 14 are given.
740 TABLES OF BESSEL FUNCTIONS
Table V. Values of J /
n + i) (x)
I -1
0-671397 + 0-431099 + 0-240298 - 1-102496 + 0-049497 + 2-876388 I
2 + 0-513016 - 0-234786 + 0-491294 - 0-395623 + 0-223925 + 0-828221 2
3 + 0-065008 - 0-456049 + 0-477718 + 0-087008 + 0-412710 + 0-369041 3
4 - 0-301921 0-260766 + 0-185286 + 0-367112 + 0-440885 - 0-014568 4
5 - 0-342168 + 0-101218 - 0-169651 + 0-321925 + 0-240377 - 0-294372 5
X AW J -I w /.(*)
2
/-#(*> X
X
7-V (*)
/-V W hi- (*) /-vW X
5 + 0-000007 + o-oooooi 5
+ 0-000005 + o-oooooi 7
8 + 0-001846 + 0-000551 + 0-000152 + 0-000039 + 0-000009 + 0-000002 8
9 + 0-006568 + 0-002261 + 0-000718 + 0-000212 + 0-000058 + 0-OOOOI5 9
10 + 0-018837 + 0-007421 + 0-002683 + 0-000898 + O-OOO20O + 0-000082 10
11 + 0-044768 + 0-020106 + 0-008237 + 0-003108 + 0-001086 + 0-000355 11
12 + 0-089695 + 0-045914 + 0-021263 + 0-0090I7 + 0003532 + 0-001288 12
13 + 0-152818 + 0-089532 + 0-046907 + 0-022324 + 0-009760 + 0003955 13
14 + 0-221379 + 0-149989 + 0-089312 + 0-047774 + 0-023297 + 0-010469 14
*5 + 0-269315 + 0-215531 + 0-147378 + 0-089050 + 0048533 + 0-024193 15
dt dt
*/>-i> />i<"
Maxima and minima
0-02 + 0-1128334 + 0-0007522 of Fresnel's integrals
0-04 + 0-1595514 + 0-0021274
o-oo + 0-1953707 + 0-0039078
0-08 + 0-2255314 + 0-0060153 #=(-)-
o-io + 0-2520611 + 0-0084044 //-i><"
0-12 + 0-2759976 + 0-0110444
0-14 + 0-2979565 + 0-0139124 1-570796 + 0-779893
O-IO + 0-3183378 + 0-0169904 4712389 + 0-321056
0-18 + '3374 I 86 4-0-0202639 7-853982 + 0-640807
0-20 + 0-3554002 + 0-0237204 IO-995574 + 0-380389
14-137167 + 0-605721
0-22 + o-3724338 + 0-0273496
0-24 + 0-3886365 + 0-0311421 17*278760 + 0-404260
O-20 + 0*4041012 + 0-0350898 20*420352 + 0*588128
0-28 + 0-4189028 + 0-0391853 23-56I945 + 0-417922
0-30 + 0-4331026 + 0-0434218 26703538 + 0-577121
29-845130 + 0-427036
0-32 + 0-4467517 + 0-0477932
o-34 + 0-4598932 + 0-0522937 32-986723 + 0-569413
0-36 + 04725635 + 0-0569181 36-128316 + 0433666
0-38 + 0-4847941 + 0-0616612 39-269908 + 0-563631
0-40 + 0-4966121 + 0-0665185 42-411501 + 0-438767
45 553093 + 0-559088
0-42 + 0-5080410 + 0-0714853
0-44 + 0-5191018 + 0*0765575 48-694686 + 0-442848
0-46 + 0-5298125 + 0-0817309
048 + 0-5401895 + 0-0870016
050 + 0-5502472 + 0*0923658
0-52 + 0-5599985 + 0-0978198 x = n*
o-54 + 0-5694551 + 0-1033602 */>!><<
0-56 + 0-5786275 + 0*1089835
0-58 + 0-5875253 + 0-1146863
o-6o + 3-I4I593 + 0-713972
0-5961571 4- 0-1204654 6-283185 + 0-3434I5
0-62 + 06045308 9-424778 + O62894O
+ 0-1263176 12-566371
.
+
0-64 + 0-6126537 + 0-1322398 0-387969
o-66 + 0-6205324 + o- 1382290
15*707963 + 0-60036l
o-68 + 0-6281731 + 0-1442820
0-70 + 0-6355815 + OT503961 18-849556 + 0-408301
21*991149 4- 0-584942
0-72 + 0*6427627 + 25-132741 + 0-420516
0-1565683
0-74 + 0-6497217 + 0-1627958 28274334 + o-574957
0-76 + 0-6564631 + 31*415927 + 0-428877
0-1690757
+ 0-6629910 + o -1 754054
+ 0-6693095 + 0-1817820 34-557519 + 0-567822
37*699112 + 0-435059
0-82 + 0-6754224 + 0-1882030 40-840704 + 0-562398
0-84 + 0-6813330 + 0-1946656
43*982297 + 0-439868
o-86 + 0-6870448 + 47-123890 + 0558096
0-2011673
o-88 + 0-6925609 + 0-2077055
0-90 + 0-6978843 + 50*265482 + 0-443747
0-2142775
0-92 + 0-7030179 + 0-2208809
o-94 + 07079643 + 0-2275131
0-96 + 0*7127261 + 0-2341717
0-98 + 07173059 + 0*2408543
I-OO + 0*7217059 + 0*2475583
' 1 '
For values of n exceeding 50, the following approximations may be used with seven-figure
accuracy: 000586 92885 r
0-44730 73184 F 1 ~\ 1213 -|
"'
J L 225 2J * "
L i4625n 2 J
0-41085 01939 r 23 1 0-08946 14637 r 947 "I
J,,' (w)
J L 3i5onaJ Hi L "69300aJ
TABLES OF BESSEL FUNCTIONS 747
For values of n exceeding 50, the following approximations may be used with seven-figure
accuracy.
^ 077475 90021 [~ 1 ~[ 001016 59059 f_ 1213 "1
+
,
ni L 225*J _* L l4625^_r
071 161 34100 r 23 0-15495 18004 rt
YM '(n). + [
947 "|
6930on8J
The coefficients are numerically equal to
V3 times the corresponding coefficients in Jn (n) and
748 TABLES OF BESSEL FUNCTIONS
n Ji.n y, Ja n y*,n n
v'3- {JWa;)cos30 - ri/3 (a;)sin3o }, ^3. ;J 1/8 (a;)cos 120 - ri/3 (x)sin 120 }
respectively.]
J 11*,' ym SH d
- % = yro,n y W (It
0-52 + 0-2542004 0-2851974 */l
o-54 + O2635IO3 - 0-2892124
0-56 + 0-2727682 - 0-2929516
0893577O - 0-320929I
0-58 + 0-28I9722 - 0-2964234
+ 0-29II206 - 3-9576784 + 0-I920I49
o-6o 02996358 7-0860511 - 0-1474447
0-62 + 0-3002II7 - 0-3025960 IO-2223450 -;- 0-1237411
0*64 + 0-3092437 - 0-3053111 I3-36I0975 - 01 085949
o-66 + 0-3182150 - 0-3077877
- 16-5009224 + 0-0978827
o-68 + 0-3271238 0-3100319 - 0-0898033
0-70 + - 0-3120498 19-6413097
0-3359684 22-7820280 + 0-0834339
25-9229577 - 0-0782474
0-72 + 0-3447472 - 0-3138471
- 29-0640303 + 0-0739188
0-74 + 0-3534586 0-3154290
0-76 + 0-362I0I0 - 0-3168010
32-2052041 - 0-0702357
078 + 0-3706727 - 0-3179678
- 35-3464523 + 0-0670523
o-8o + 0-379I722 0-3189344 - 0-0642652
38-4877567
- 03197054 41-6291045 + 0-0617985
0-82 + 0-3875979 - 0-0595953
0-84 + 0-3959484 - 0-3202852 44-7704866
o-86 + 0-4042220 - 0-3206781
- 03208884 47-9118963 + 0-0576118
o-88 + 0-4124174
0-90 + 0-4205330 - 0-3209201
For values of x between o and 16, the integrals may be calculated with the help of Table I
from the formulae (cf. 10-74)
X
BIBLIOGRAPHY*
ADAMOFF, A.
Onthe asymptotic representation of the cylinder functions Jv (2) and /' (2) for large
values of the modulus of z. Petersburg, Ann. Inst, polyt. 1906, pp. 239265. [JahrbuZh
itber die Fortschritte der Math. 1907, pp. 492493.]
AICHI, K.
Note on the Function Km (x), the Solution of the Modified Bessel's Equation. Proc.
Phys. Math. Soc. of Japan, (3) n. (1920), pp. 819.
AIREY, J. R.
The Roots of the Neumann and Bessel Functions (Dec. 29, 1910). Proc. Phys. Soc. xxm.
(1911), pp. 219 224.
The Vibrations of Circular Plates and their Relation to Bessel Functions (Feb. 15, 1911).
Proc. Phys. Soc. xxm. (1911), pp. 225232.
The Oscillations of Chains and their Relation to Bessel and Neumann Functions.
Phil. Mag. (6) xxi. (1911), pp. 736742.
Y
Tables of Neumann Functions G n (x) and n (x). Phil. Mag. (6) xxu. (1911), pp. 658
663.
The Asymptotic expansions of Bessel and other functions. Archiv der Math, und Phys.
(3) xx. (1913), pp. 240244.
The Vibrations of Cylinders and Cylindrical Shells. Archiv der Math, und Phys. (3)
xx. (1913), pp. 289294.
Tables of the Neumann functions or Bessel functions of the second kind. Archiv der
Math, und Phys. (3) xxn. (1914), pp. 3043.
Bessel and Neumann Functions of Equal Order and Argument. Phil. Mag. (6) xxxi.
(1916), pp. 520528.
The Roots of Bessel and Neumann Functions of High Order. Phil. Mag. (6) xxxil.
(1916), pp. 714.
Bessel Functions of Equal Order and Argument. Phil. Mag. (6) xxxn. (1916), pp. 237
238.
J
The Numerical Calculation of the Roots of the Bessel Function n (x) and its first
derivative Jn (x). Phil. Mag. (6) xxxiv. (1917), pp. 189195.
'
The Addition Theorem of the Bessel Functions of Zero and Unit Orders. Phil. Mag.
(6) xxxvi. (1918), pp. 234242.
The Lommel- Weber 12 Function and its Application to the Problem of Electric Waves
on a Thin Anchor Ring (Dec. 7, 1917). Proc. Royal Soc. xciv. A
(1918), pp. 307314.
Bessel Functions of small Fractional Order and their application to problems of Elastic
Stability. Phil. Mag. (6) xli. (1921), pp. 200205.
Dec. 24, 1917). Comptes Rendus, clxiii. (1916), pp. 2629; clxv. (1917), pp. 2325,
11001103.
* In the case of a few inaccessible memoirs, references are given to abstracts in the Juhrbuch
iiber die Fortschritte der Math, or elsewhere.
754 THEORY OF BESSEL FUNCTIONS
ALDIS, W. S.
Tables for the Solution of the Equation
d2 u 1 dy /, n2 \
(June 16, 1898). Proc. Royal Soc. lxiv. (1899), pp. 203223.
On the numerical computation of the functions G^^x), Gi(x) and JH (x Kfi) (June 15,
1899). Proc. Royal Soc. lxvi. (1900), pp. 3243.
ALEXANDER, P.
Expansion of Functions in terms of Linear, Cylindric, Spherical and Allied Functions
(Dec. 20, 1886). Trans. Edinburgh Royal Soc. xxxin. (1888), pp. 313320.
ANDING, E.
Sechsstellige Tafeln der Besselscken Funktionen imaginaren Arguments (Leipzig, 1911).
ANGER, C. T.*
Untersuchungen Function Ikh mit Anwendungen auf das Kepler'sche Problem.
iiber die
Neueste Schriften der Naturforschenden der Ges. in Danzig, v. (1855), pp. 1 29.
ANISIMOV, V. A.
The generalised form of Riccati's equation. Warsaw University, 1896,
Proceedings of
pp. 1 33. [Jahrbuch iiber die Fortschritte der Math. 1896, p. 256.]
APPELL, P. E.
Sur l'inversion approchee de certaines integrates reelles et sur 1'extension de l'equation
de Kepler et des fonctions de Bessel (April 6, 1915). Comptes Rendus, clx. (1915),
pp. 419423.
AUTONNE, L.
Sur la nature des integrates algdbriques de l'equation dn Riccati (May 7, 1883). Comptes
Rendus, xcvi. (1883), pp. 13541356.
Sur les integrales algebriques de l'equation de Riccati (Feb. 13, 1899). Comptes Rendus,
cxxvin. (1899), pp. 410412.
BACH, D.
De Pintegration par les series de l'equation
d2 y n \ dy __
x dx~ y
'
dx*
Ann. set. de VEcole norm. sup. (2) in. (1874), pp. 47 68.
BAEHR, G. F. W.
Sur les racines des equations I cos (.r cos >)<&> =0 et I cos(#cos)sin2 G>rfo>=0
(April, 1872). Archives Neerlandaises, vii. (187L), pp. 351 358.
BALL, L. DE.
Ableitung einiger Formeln aus der Theorie der Bessel'schen Functionen (June 6, 1891).
Astr. Nach. cxxvin. (1891), col. 1 4.
BARNES, E. W.
Onthe homogeneous linear difference equation of the second order with linear coeffi-
cients. Messenger, xxxiv. (1905), pp. 52 71.
On Functions denned by simple types of Hypergeometric Series (March 12, 1906).
Trans. Camb. Phil. Soc. xx. (1908), pp. 253279.
The asymptotic Expansion of Integral Functions denned by generalised Hypergeometric
Series (Dec. 3, 1906). Proc. London Math. Soc. (2) v. (1907), pp. 59116.
BASSET, A. B.
On a methodof finding the potentials of circular discs by means of Bessel's functions
(May 10, 1886). Proc. Camb. Phil. Soc. v. (1886), pp. 425443.
On the Potentials of the surfaces formed by the revolution of Limacons and Cardioids
about their axes (Oct. 25, 1886). Proc. Camb. Phil. Soc. vi. (1889), pp. 219.
A Treatise on Hydrodynamics (2 vols.) (Cambridge, 1888).
On the Radial Vibrations of a Cylindrical Elastic Shell (Dec. 12, 1889). Proc. London
Math. Soc. xxi. (1891), pp. 5358.
On a Class of Definite Integrals connected with Bessel's Functions (Nov. 13, 1893).
Proc. Camb. Phil. Soc. vin. (1895), pp. 122128.
* See also under Bourget and Cauchy.
BIBLIOGRAPHY 755
BATEMAN, H.
Certain definite integrals connected with the Legendre and Bessel functions. Messenger
xxxiii. (1904), pp. 182188.
A
generalisation of the Legendre polynomial (Jan. 1, 1905). Proc. London Math. Soc
(2) in. (1905), pp. 111123.
The inversion of a definite integral (Nov. 8, 1906). Proc. London Math. Soc. (2) iv.
(1906), pp. 461498.
On an expansion of an arbitrary function into a series of Bessel functions. Messenger,
xxxvi. (1907), pp. 3137.
The Solution of Linear Differential Equations by means of Definite Integrals (Jan. 25,
1909). Trans. Camb. Phil. Soc. xxi. (1912), pp. 171196.
The History and Present State of the Theory of Integral Equations. British Association
Report, 1910, pp. 345424.
Notes on integral equations. Messenger, xli. (1912), pp. 94 101, 180 184.
Some equations of mixed differences occurring in the Theory of Probability and the
related expansions in series of Bessel's functions. Proc. Int. Congress of Math. i. (Cambridge,
1912), pp. 291294.
Electrical and Optical Wave-motions (Cambridge, 1915).
BAUER, G.
Von den Reihen von Kugelfunctionen einer Variablen. Journal fur
Coefificienten der
Math. lvi. (1859), pp. 101121.
Bemerkungen iiber Reihen nach Kugelfunctionen und insbesondere auch iiber Reihen,
welche nach Producten oder Quadraten von Kugelfunctionen fortschreiten mit Anwendun<*
auf Cylinderfunctionen (July 3, 1875). Miinchener Sitzungsberichte, v. (1875), pp. 247 272
BECKER, J.
Die Riccatische Differential-Gleichung. Programm Karlsbad, 1908 (25 pp.). [Jahrbuch
iiber die Fortschritte der Math. 1908, p. 395.]
BELTRAMI, E.
Intorno ad un teorema di Abele e ad alcune sue applicazioni. R. 1st. Lombardo Rendi-
conti, (2) xin. (1880), pp. 327337.
Intorno ad alcune serie trigonometriche (June 17, 1880). R. 1st. Lombardo Rendiconti,
(2) xin. (1880), pp. 402413.
Sulle funzioni cilindriche (Jan. 14, 1881). Atti della R. Accad. delle Sci. di Torino, xvi
(188081), pp. 201205.
Sulla teoria delle funzioni potenziali simmetriche (April 28, 1881). Bologna Memorie
(4) ii. (1880), pp. 461505.
BERNOULLI, DANIEL.
Correspondence with Leibniz 16971704. [Published in Leibnizens Oes. Werke, Dritte
Folge (Mathematik), in. (Halle, 1855).]
Notata in praecedens schediasma 111. Co. Jacobi Riccati, Actorum Eruditorum quae
Lipsiae publicantur Supplementa, vm. (1724), pp. 73 75.
Solutio problematis Riccatiani propositi in Act. Lips. Suppl. Tom. vm. p. 73. Acta
Eruditorum publicata Lipsiae, 1725, pp. 473 475.
Exercitaticm.es quaedam mathematicae (Venice, 1 724), pp. 77 80.
Theoremata de oscillationibus corporum filo flexili connexorum et catenae verticaliter
suspensae. Comm. Acad. Sci. Imp. Petrop. vi. (1732 33) [1738], pp. 108 122.
Demonstrationes Theorematum suorum de oscillationibus corporum filo flexili connex-
orum et catenae verticaliter suspensae. Comm. Acad. Sci. Imp. Petrop. vn. (1734 35)
162179.
[1740], pp.
BERNOULLI, JOHN.
Methodus generalis construendi omnes aequationes differentiales primi gradus. Acta
Eruditorum publicata Lipsiae, 1694, pp. 435437. \0pera, I. Lausanne aud Geneva, 1742,
p. 124.]
BERNOULLI, NICHOLAS (the younger).
Correspondence with Goldbach. See under Fuss.
BESSEL, F. W.
Analytische Auflosung der Keplerschen Aufgabe (July 2, 1818). Berliner Abh. 181617
[1819], pp. 49 55. [Abhandlungen, herausgegeben von R. Engelmann, I. (1875), pp. 17
20.]
Ueber die Entwickelung der Functionen zweier Winkel u und u' in Reihen welche nach
den Cosinussen und Sinussen der Vielfachen von u und u' fortgehen (June 21, 1821).
Berliner Abh. 182021 [1822], pp. 5660. [Abhandlungen, n. (1876), pp. 362364.]
756 THEORY OF BESSEL FUNCTIONS
Untersuchung des Theils der planetarischen Storungen welcher aus der Bewegung der
Sonne entsteht (Jan. 29, 1824). Berliner Abh. 1824 [1826], pp. 152. [Abhandlungen, I.
(1875), pp. 84109.]
Beitrag zu den Methoden die Storungen der Kometen zu berechnen (Sept. 24, 1836).
Astr. Nnch. xiv. (1837), col. 148. [Abhandlungen, I. (1875), pp. 2954.]
BINET, J. P. M.
~
Note sur l'integrale u dye~y i qy prise entre des limites arbitraires (May 24, 1841).
/
y
J a
Comptes Rendus, xn. (1841), pp. 958 962.
BOCHER, M.
On kind (Jan. 1892). Annals of Math. vi. (1892), pp.
Bessel's functions of the second
8590.
On some applications of Bessel's functions with pure imaginary index (Feb. 11, 1892).
Annals of Math. vi. (1892), pp. 137160*.
On certain methods of Sturm and their application to the roots of Bessel's functions
(Feb. 1897). Bulletin American Math. Soc. ill. (1897), pp. 205213.
An elementary proof that Bessel's functions of the zeroth order have an infinite number
of real roots (Feb. 25, 1899). Bulletin American Math. Soc. v. (1899), pp. 385388.
Non-oscillatory linear differential equations of the second order (Feb. 4, 1901). Bulletin
American Math. Soc. vn. (1901), pp. 333340.
BOHMER, P. E.
Uber die Zylinderfunktionen (Nov. 26, 1913). Sitz. der Berliner Math. Ges. xin. (1913),
pp. 3036.
BOHREN, A.
Uber das Airysche Integral t (Oct. 6, 1902). Bern Mittheilungen, 1902, pp. 236239.
BOOLE, G.
On the transformation of Definite Integrals. Camb. Math. Journal, ill. (1843), pp. 216
224.
On a general method in analysis^Jan. 18, 1844). Phil. Trans, of the Royal Soc. 1844,
pp. 225282.
A Treatise on Differential Equations (London, 1872).
BOURGET, J.
Note sur une formule de M. Anger (Aug. 7, 1854). Comptes Rendus, xxxix. (1854),
p. 283.
Memoire sur les nombres de Cauchy et leur application a divers problemes de inecanique
celeste. Journal de Math. (2) vi. (1861), pp. 33 54.
Memoire sur le mouvement vibratoire des membranes circulaires (June 5, 1865). Ann.
sci. de VEcole norm. sup. in. (1866), pp. 55 95.
BRAJTZEW, J. R.
Uber Fourier-Besselschen Funktionen und deren Anwendung zur Auffindung
die
asymptotischer Darstellungen von Integralen der linearen Differentialgleichungen mit
rationalen Koeffizienten. Warschau Polyt. Inst. Each. 1902, nos. 1, 2. [Jahrbuch uber die
Fortschritte der Math. 1903, pp. 575 577.]
BRASSINNE, E.
Sur diverses equations differentielles du premier ordre analogues a
l'equation de Ricatti
(sic). Mem. de
I'Acad. R. des Sci. de Toulouse, (3) iv. (1848), pp. 234 236.
Sur des equations differentielles qui se rattachent a l'equation de Riccati. Journal de
Math. xvi. (1851), pp. 255256.
BRENKE, W. C.
Summation of a series of Bessel's functions by means of an integral (Nov. 27, 1909).
Bull. American Math. Soc. xvi. (1910), pp. 225 230.
BRIDGEMAN, P. W.
On a Certain Development in Bessel's Functions (July 22, 1908). Phil. Mag. (6) xvi.
(1908), pp. 947948.
BIBLIOGRAPHY 757
BRUNS, H.
Ueber die Beugungsfigur des Heliometer-Objectivea (Oct. 15, 1882). Astr. Nach. civ.
(1883), col. 18.
BRYAN, G. H.
On the waves on a viscous rotating cylinder (June 4, 1888). Proc. Camb. Phil. Soc. vi.
(1889), pp. 248264.
Wave Motion and Bessel's Functions. Mature, lxxx. (1909), p. 309.
BURKHARDT, H. F. K. L.
Trigonometrische Reihen und Integrale (bis etwa 1850). Encyklopadie der Math. Wiss.
II. 1 (Leipzig, 190416), pp. 8191354.
LUTTERWORTH, S.
On the Evaluation of Certain Combinations of the Ber, Bei and Allied Functions (May
30, 1913). Proc. Phys Soc. xxv. (1913), pp. 294 *97.
CAILLER, C.
Sur les fonctions de Bessel. Archives des Sci. (Soc. Helvetique), (4) xiv. (1902), pp. 347
350.
Note sur une operation analytique et son application aux fonctions de Bessel (March,
1904). M4m. de la Soc. de phys. et cPhistoire naturelle de Geneve, xxxiv. (1902 5), pp. 295
368.
CALLANDREAU, 0.
Calcul des transcendantes de Bessel
.
nK) n
1.2...L
(rr (it
l.(n + l)" ".1.2.(+l)(+2)
,
t
(it
'"J'
i
pour les grandes valeurs de a au moyen de series semiconvergentes. Bidletin des Sci. Math.
(2) xiv. (1890), pp. 110114.
Sur le calcul des polynomes Xn (cos 6) de Legendre pour les grandes valeurs de n. Bulletin
des Sci. Math. (2) xv. (1891), pp. 121124.
CARLINI, F.
Ricerche sulla convergenza della serie che serva alia soluzione del problema di Keplero*
(Milan, 1817).
CARSLAW, H. S.
Some Multiform Solutions of the Partial Differential Equations of Physical Mathematics
and their Applications (Nov. 10, 1898). Proc. London Math. Soc. xxx. (1899), pp. 121 161.
The Green's function for a wedge of any angle and other Problems in the Conduction
of Heat (Oct. 30, 1909). Proc. London Math. Soc. (2) vm. (1910), pp. 365374.
The Scattering of Sound Waves by a Cone. Math. Ann. lxxv. (1914), pp. 133 147.
The Green's function for the equation V 2 u 2 =0 (April 28, 1913; March 20, 1916).
+
Proc. London Math. Soc. (2) xm. (1914), pp. 236257 (2) xvi. (1917), pp. 8493. ;
Translated into German by Jacobi, Astr. Nach. xxx. (1850), col. 197 254. [Get. Math.
Werke, vn. (1891), pp. 189245.]
758 THEORY OF BESSEL FUNCTIONS
Considerations nouvelles sur la theorie des suites et sur les lois de leur convergence
(April 20, 1840). Comptes Rendus, x. (1840), pp. 640656. [Oeuvres, (1) v. (1885), pp. 180
198.]
Methode simple et generate pour la determination numerique des coefficients que ren-
ferme developpement de la fonction perturbatrice (Sept. 14, 1840). Comptes Rendus, XI.
le
(1840), pp. 453475. [Oeuvres, (1) v. (1885), pp. 288310.]
Note sur le developpement de la fonction perturbatrice (Sept. 21, 1840). Comptes Rendusr
xi. (1840), pp. 501511. [Oeuvres, (1) v. (1885), pp. 311321.1
Methodes propres a simplifier le calcul des inegalites periodiques et seculaires des
mouvements des planetes (Jan. 11, 1841). Comptes Rendus, xil. (1841), pp. 84 101.
[Oeuvres, (1) vi. (1888), pp. 1634.]
Note sur une transcendante que renferme le developpement de la fonction perturbatrice
relative au systeme planetaire (Oct. 4, 1841). Comptes Rendus, xin. (1841), pp. 682687.
[Oeuvres, (1) vi. (1888), pp. 341346.]
Note sur la substitution des anomalies excentriques aux anomalies moyennes, dans le
developpement de la fonction perturbatrice (Oct. 25, 1841). Comptes Rendus, xni. (1841),
pp. 850854. [Oeuvres, (1) vi. (1888), pp. 354359.1
Nouveau Memoire sur le calcul des inegalites des mouvements planetaires (April 8.
1844). Comptes Rendus, xvm. (1844), pp. 625643. [Oeuvres, (1) vm. (1893), pp. 168
188.
Sur la transformation des fonctions implicites en moyennes isotropiques, et sur leurs
developpements en series trigonometriques (May 22, 1854). Comptes Rendus, xxxvm.
(1854), pp. 910913. [Oeuvres, (1) xn. (1900), pp. 148151.]
Sur la transformation des variables qui ddterminent les mouvements d'une planete ou
m^me d'une comete en fonction explicite du temps, et sur le developpement de ses fonctions
en series convergentes (June 5, 1854). Comptes Rendus, xxxvm. (1854), pp. 990 993.
[Oeuvres, (1) xn. (1900), pp. 160164.]
Sur la resolution des equations et sur le developpement de leurs racines en series con-
vergentes (June 26, 1854). Comptes Rendus, xxxvm. (1854), pp. 1104 1107. [Oeuvres, (1)
xn. (1900), pp. 167170.]
Sur une formule de M. Anger et sur d'autres formules analogues (July 15, 1854).
Comptes Rendus, xxxix. (1854), pp. 129 135. [Oeuvres, (1) xil. (1900), pp. 171177.]
CAYLEY, A.
Sur quelques formules du calcul integral. Journal de Math. xn. (1847), pp. 231 240.
[Collected Papers, I. (1889), pp. 309316.]
On Riccati's equation (Sept. 29, 1868). Phil. Mag. (4) xxxvi. (1868), pp. 348351.
[Collected Papers, vii. (1894), pp. 912.]
Note on the integration of certain differential equations by series. Messenger (Old
Series), v. (1869), pp. 7782. [Collected Papers, vm. (1895), pp. 458462.]
Proc. London Math. Soc* v. (1874), pp. 123124. [Collected Papers, IX. (1896), pp.
1920.]
CHALLIS, H. W.
Extension of the Solution of Riccati's Equation (Oct. 5, 1864). Quarterly Journal, vn.
(1866), pp. 5153.
CHAPMAN, S.
On the general theory of summability with applications to Fourier's and other series.
Quarterly Journal, xliii. (1911), pp. 1 52.
CHESSIN, A. S.
Note on the General Solution of BessePs Equation. American Journal of Math. xvi.
(1894), pp. 186187.
On the expression of BessePs Functions in Form of Definite Integrals. Johns Hopkins
Univ. Circulars, xiv. (1895), pp. 2021.
Note on Cauchy's Numbers. Annals of Math. x. (1896), pp. 12.
On the relation between Cauchy's numbers and BessePs Functions (July 1, 1898).
Annals of Math. xn. (1899), pp. 170174.
On some relations between Bessel functions of the first and of the second kind (Oct.
20, 1902). Trans. Acad. Sci. of St Louis, xn. (1902), pp. 99108.
Sur Pequation de Bessel avec second membre (Oct. 27, 1902). Comptes Rendus, cxxxv.
(1902), pp. 678679.
Sur une classe d'equations differentielles redactibles a Pequation de Bessel (May 11,
1903). Comptes Rendus, cxxxvi. (1903), pp. 11241126.
* See under Lord Rayleigb.
BIBLIOGRAPHY 759
CHREE, C.
Longitudinal vibrations of a circular bar. Quarterly Journal, xxi. (1886), pp. 287298.
On the Coefficients in certain Series of Bessel's Functions. Phil. Mag. (6) xvn. (1909),
pp. 329 331.
CHRISTOFFEL, E. B.
Zur Abhandlung: "Ueber die Zahler und Nenner der Naherungswerte von Ketten-
briichen" pag. 231 des vorigen Bandes* (March, 1860). Journal fur Math, lviii. (1861),
pp. ye) y^,
CINELLI, M.
Diffrazione per aperture fatte sopra superfici curve. II Nuovo Cimento, (1895),
pp. 141 155. (4) i.
CLEBSCH, R. F. A.
(Jeber die Reflexion an einer Kugelflache (Oct. 30, 1861). Journal fiir Math. lxi. (1863),
pp. l7u ^u2.
CLIFFORD, W. K.
On Bessel's Functionst. Mathematical Papers (London, 1882), pp. 346 349.
COATES, C. V.
Bessel's functions of the Second order. Quarterly Journal, xx. (1885), pp. 250260.
Bessel's functions of the Second order. Quarterly Journal, xxi. (1886), pp. 183 192.
COCKLE, SIR JAMES.
On Linear Differential Equations of the Second Order (Dec. 24, 1861 Jan. 15, 1862: ;
May 21, 1862). Messenger (Old Series), I. (1862), pp. 118124, 164173, 241247.
COTTER, J. R.
A New Method of Solving Legendre's and Bessel's Equations and others of a similar
type (May 27, 1907). Proc. R Irish Acad. % xxyii. A
(1909), pp. 157161.
CRAWFORD, L.
A proof of Rodrigues' Theorem sin nx= .
1 .
_
.
.
a -
3 5...(2n
n
. -r(~x ^-V"
dxj
1) \sin
1
sin**-* x and
some expansions derived from it (Dec. 13, 1901). Proc. Edinburgh Math. Soc. xx. (1902),
CRELIER, L.
Sur quelques proprtetes des fonctions Besseliennes tirees de la theorie des fractions
continues (June, 1895). Ann. di Mat. (2) xxiv. (1896), pp. 131163. {Dissertation, Bern,
L ' '
1895.]
Sur
la fonction Besselienne de e espeee n
n (x) (Dec. 1896).
S Bern Mittheilunaen.
y 1897
[1898], pp. 61
96.
S
Sur les fonctions besseliennes 0*(x) et n {x) (Sept. 6, 1897 ; Nov. 29, 1897). Comptes
Rendus, cxxv. (1897), pp. 421423, 860863.
CURTIS, A. H.
Onthe integration of Linear and Partial Differential Equations (Nov. 24, 1854) Camb.
and Dublin Math. Journal, ix. (1854), pp. 272290.
CURZON, H. E. J.
Generalisations of the Hermite functions and their connexion with Bessel functions
(Nov. 10, 1913). Proc. London Math. Soc. (2) xin. (1914), pp. 417440.
DATTA, A.
On a generalisation of Neumann's Expansion in a Series of Bessel Functions (Feb. 29,
1920). Bulletin Calcutta Math. Soc. xi. (1921), pp. 2334.
On an extension of Sonine's Integral in Bessel Functions|| (Jan. 6, 1921). Bulletin
Calcutta Math. Soc xi. (1921), pp. 221230.
DENDY, A. AND
NICHOLSON, J. W.
On the Influence of Vibrations upon the Form of Certain Sponge Spicules (May 11,
1917). Proc. Royal Soc. lxxxix. B (1917), pp. 573587.
DINI, U.
Serie di Fourier e altere rappresentazioni analitiche delle funzioni di una variabile reale
(Pisa, 1880).
DINNIK, A.
die Darstellung einer willkiirlichen Funktion durch BessePsche Reihe. Kief. Polyt.
Uber
Inst. (Engineering Section), 1911, no. 1, pp. 83
85. [Jahrbuch iiber die Fortschritte der
Math. 1911, p. 492.]
Tafeln der Besselschen Funktionen J\ und /
. Archiv der Math, und Phys. (3)
DONKIN, W. F.
On the Equation of Laplace's Functions, etc. (Dec. 11, 1856). Phil. Trans, of the Royal
Soc. cxlvii. (1857), pp. 4357.
DOUGALL, J.
The determinationof Green's function by means of Cylindrical or Spherical Harmonics
(March 9, 1900). Proc. Edinburgh Math. Soc. xvni. (1900), pp. 3383.
A
Theorem of Sonine in Bessel Functions with two Extensions to Spherical Harmonics
(Dec. 13, 1918). Proc. Edinburgh Math. Soc. xxxvn. (1919), pp. 3347.
DU
BOIS REYMOND, P. D. G.
Die Theorie der Fourier'schen Integrale und Formeln (June 26, 1871). Math. Ann. iv.
(1871), pp. 362390.
EARNSHAW, S.
Partial differential equations. An essay towards an entirely new method of integrating
them (London, 1871).
ELLIS, R. L.
On
the Integration of certain Differential Equations (Nov. 1840 and Feb. 1841). Camb.
Math. Journal, n. (1841), pp. 169177, 193201.
On the Method of Least Squares (March 4, 1844). Trans. Camb. Phil. Soc. vm. (1849),
pp. 204219.
EMDE, F.*
Zur Berechnung der reellen Nullstellen der Besael'schen Zylinderfunktionen. Archiv
der Math, und Phys. (3) xxiv. (1916), pp. 239250.
ENNEPER, A.
Ueber ein bestimmtes Integral. Math. Ann. vi. (1873), pp. 360 365.
EPSTEIN, S. S.
Die vier Rechnungsoperationen mit Bessel'schen Functionen nebst einer geschichtlichen
Einleitung (Bern, 1894, 58 pp.). [Jahrbuch Uber die Fortschritte der Math. 1893 94, pp.
845846.]
* See also under Jahnke.
BIBLIOGRAPHY 761
ERMAKOFF, W.
Ueber die Cylinderfunctionen (May, 1872). Math. Ann. v. (1872), pp. 639640.
ESCHERICH, G. VON.
Zur Bessel'schen Differential-Gleichung. Monatshefte fur Math, und Phys. m. (1892),
p. 142.
Uber eine Naherungsformel. Monatshefte fur Math, und Phys. in. (1892), p. 234.
EULER, L.
Lettre de M. Euler a M. de la Grange (Jan. 1, 1760). Misc. Taurinensia, n. (176061),
pp. 110.
Recherches sur l'integration de l'equation
ddz _ ddz b dz c
d(l ~ dx2 x dx xx
Misc. Taurinensia, ill. (1762 65), pp. 60 91.
De integratione aequationum differentialium. Novi Gomm. Acad. Petrop. vin. (1760 61)
[1763], pp. 154169.
De resolutione aequationis dy+ayydx = bxm dx. Novi Comm. Acad. Petrop. ix. (1762
63) [1764], pp. 154 169.
Deniotu vibratorio tympanorum. Novi Comm. Acad. Petrop. x. (1764) [1766], pp.
243260.
Institutionum Calculi Integralis, II. (Petersburg, 1769).
De oscillationibus minimis funis libere suspensi. Acta Acad. Petrop. v. pars I (1781)
[1784], pp. 157177.
De perturbatione motus chordarum ab earum pondere oriunda. Acta Acad. Petrop. v.
pars 1 (1781) [1784], pp. 178190.
Analysis facilis aequationem Riccatianam per fract.onem continuam resolvendi. Mem.
de I' Acad. R. des Sci. de St. Petersbourg, vi. (1818), p) 1229.
FALKENHAGEN, J. H. M.
Ueber das Verhalten der Integrale einer Riccati'schen Gleichung in der Nahe einer
singularen Stelle. Nieuw Archief voor Wiskunde, (2) vi. (1905), pp. 209 248.
FAXEN, H.
(tt ~'i)
Expansion in series of the integral / e~ x Vdt (April 14, 1920). Arkiv for Mat.
J v
Astr. och Fysik, xv. (1921), no. 13.
FELDBLUM, M.
The theory of Riccati's equation and applications of the function which satisfies it.
Warsaw University, 1898, nos. 5 and 7 1899, no. 4. \Jahrbuch iiber die Fortschritte der
;
FORD, W." B.
On
the possibility of differentiating term-by-term the developments' for an arbitrary
function of one real variable in terms of Bessel functions (June, 1902). Trans. American
Math. Soc. iv. (1903), pp. 178184.
w. B. F. 25
762 THEORY OF BESSEL FUNCTIONS
FORSYTH, A. R.
On linear differential equations : in particular that satisfied by the series
, ,
a<30 M| a. g+1. + 1. 6. 0+1
fl.fl ,
ye 1.2.y.y + l.f. e +l ^ , "
Quarterly Journal, xix. (1883), pp. 292 337.
The expression of Bessel functions of positive order as products, and of their inverse
powers as sums of rational fractions. Messenger, l. (1921),
pp. 129149.
FOURIER, J. B. J.
La Thdorie analytique de la Chaleur (Paris, 1822). [Translated by A. Freeman, Cam-
bridge, 1878.]
FREEMAN, A.
Note on the value of the least root of an equation allied to Ja (z) = (April 19 1880)
Proc. Camb. Phil. Soc. in. (1880), pp. 375377.
FRESNEL, A. J.
Memoire sur la diffraction de la lumiere [July 29, 1818 ; crowned 18191. Mem. de VAcad.
R. des Sci. v. (182122), pp. 339476. [Oeuvres, I. (1866), pp. 247382.]
FRULLANI, G.
Sopra dipendenza fra i differenziali delle funzioni e
la gli Integrali definiti (Feb. 4,
1818). Mem. soc. ital. (Modena), xvm. (1820), pp. 458517.
FUSS, P. H.
Correspondance mathe'matique et physique de quelques ce'lebres geometres du xviii*'"
Steele*, n. (Petersburg, 1843).
GALLOP, E. G.
Thedistribution of electricity on the circular disc and the spherical bowl. Quarterly
Journal, xxi. (1886), pp. 229256.
GASSER, A.
Ueber die Nullstellen der Besselschen Funktionen (July, 1 904). Mittheilunqen der Naturf.
Ges. in Bern, 1904, pp. 92135.
GEGENBAUER, L.
Note tiber die Bessel'schen Functionen zweiter Art (Feb. 8, 1872). Wiener Sitzunqs-
berichte, lxv. (2) (1872), pp. 3335.
Zur Theorie der Bessel'schen Functionen zweiter Art (July 4, 1872). Wiener Sitzunqs-
berichte, lxvi. (2) (1872), pp. 220223.
Note uber bestimmte Integrale (Feb. 6, 1873). Wiener Sitzunqsberichte, lxvii. (2) (1873).
\ /\
pp. 202204. />
X
Uber die Functionen nm (June 13, 1873). Wiener Sitzungsberichte, lxviii.
pp. 357 367. (2) (1874),
Uber die Bessel'schen Functionen (March 19, 1874). Wiener Sitzunqsberichte, lxx.
(2)
(1875), pp. 616.
Uber einige bestimmte Integrale (June 18, 1874). Wiener Sitzunqsberichte, lxx. (2)
(1875), pp. 433443.
* w
Uber einige bestimmte Integrale (June 17, 1875). Wiener Sitzunqsberichte, lxxii. (2)
v '
(1876), pp. 343354.
Uber die Bessel'schen Functionen (June 22, 1876). Wiener Sitzunqsberichte, lxxiv (2)
(1877), pp. 124130.
'
'
w
Zur Theorie der Bessel'schen Functionen (Jan. 18, 1877). Wiener Sitzunqsberichte,
lxx v. (2) (1877), pp. 218 222. *.
Uber die Functionen Cn v {x) (April 12, 1877). Wiener Sitzungsberichte, lxxv. (2) (1877),
pp. 891
905. \ / \ />
Das Additionstheorem derjenigen Functionen welche bei der Entwicklung von "* nach
den Naherungsnennern regularer Kettenbruche auftreten. Wiener Sitzunqsberichte, lxxxv.
(2) (1882), pp. 491502.
Uber die Bessel'schen Functionen (Oct. 11, 1883). Wiener Sitzungsberichte, lxxxviii. (2V
(1884), pp. 9751003. V
* This work contains a number of letters from Nicholas Bernoulli
(the younger) to Goldbach,
in which Bernoulli's solution of Biccati's equation is to be found. The reader should notice that in
Daniel Bernoulli's letters to Goldbach (ibid. pp. 254, 256, 259), the equation described in the table
of contents as Riccati's equation is really the linear equation; Biccati's equation is mentioned
on p. 260.
BIBLIOGRAPHY 763
Zur Theorie der Functionen Cn" (x). Wiener Akad. Denkschriften, xlviii. (1884),
pp. 293316.
Uber die Bessel'schen Functionen (March 10, 1887). Wiener Sitzungsberichte, xcv. (2)
(1887), pp. 409410.
Einige Satze iiber die Functionen CH'(x). Wiener Akad. Denkschriften, lvil (1890),
pp. 425 480.
Uber die Ringfunctionen (June 4, 1891). Wiener Sitzungsberichte, c. (2 o) (1891), pp. 746
766.
Bemerkung zu der von Herrn Elaas gegebenen Theorie der elektrischen Schwingungen
in cylindrischen Drahten. Monatshefte fiir Math, und Phys. iv. (1893), pp. 379 380.
Uber die zum elektromagnetischen Potentiale eines Kreisstromes associierte Function.
Monatshefte fur Math, und Phys. iv. (1893), pp. 393401.
Eine Integralrelation. Monatshefte Mr Math, und Phys. v. (1894), pp. 53 61.
Bemerkung iiber die Bessel'schen Functionen. Monatshefte fiir Math, und Phys. vni.
(1897), pp. 383384.
Notiz iiber die Bessel'schen Functionen erster Art. Monatshefte fiir Math, und Phys. x.
(1899), pp. 189192.
Quelques proprtetes nouvelles des racines des fonctions de Bessel. Mem. de la Soc. R.
des Set. de Liege, (3) n. (1900), no. 3.
[Letter on a paper by H. M. Macdonald.] Proc. London Math. Soc. xxxii. (1901),
pp. 433436.
tJber eine Relation des Herrn Hobson (May 22, 1902). Wiener Sitzungsberichte, cxl (2 a)
(1902), pp. 563572.
On integrals containing functions of Bessel. [Letter to Kapteyn.] Proc. Section of ScL,
K. Akad. van Wet. te Amsterdam, iv. (1902), pp. 584588.
GENOCCHI, A.
Studi intorno ai casi d' integrazione sotto forma finite. Mem. delV Accad. delle Sci. di
Torino, xxni. (1866), pp. 299362.
Sur Pequation de Riccati (Aug. 13, 1877). Comptes Bendus, lxxxv. (1877), pp. 391394.
GIBSON, G. A.
A
Proof of the Binomial Theorem with some Applications (Dec. 12, 1919). Proc.
Edinburgh Math. Soc. xxxviil (1920), pp. 69.
GILBERT, L. P.
Recherches analytiques sur la diffraction de la lumiere. Mem. couronnes de VAcad. R.
des Sci. de BruxelUs, xxxi. (1863), pp. 1 52.
GIULIANI, G.
Sopra la funzione P (cosv) per n infinite. Oiornale di Mat. xxn. (1884), pp. 236 239.
Sopra alcune funzioni analoghe alle funzioni cilindriche. Oiornale di Mat. xxv. (1887)
pp. 198202.
Alcune osservazioni sopra le funzioni spheriche di ordine superiore al secondo e sopra
altre funzioni che se ne possono dedurre (April, 1888). Giornale di Mat. xxvi. (1888),
pp. 155171.
GLAISHER, J. W. L.
On (March 15, 1871). Quarterly Journal, xi. (1871), pp. 267 273.
Riccati's equation
On
the Relations between the particular Integrals in Cayley's solution of Riccati's
Equation (May 12, 1872). Phil. Mag. (4) xliii. (1872), pp. 433 138.
On
the Evaluation in Series of certain Definite Integrals. British Association Report,
1872, pp. 1517.
Notes on definite integrals. Messenger, n. (1873), pp. 72 79.
On a Differential Equation allied to Riccati's (Oct. 11, 1872). Quarterly Journal, xil.
(1873), pp. 129137.
Sur une Propriete' de la Fonction e^ x Nouvelle Corr. Math. n. (1876), pp. 240 243,
349350.
.
On a Formula of Caucby's for the Evaluation of a class- of Definite Integrals (Nov. 6,
1876). Proc. Camb. Phil. Soc. in. (1880), pp. 512.
On certain Identical Differential Relations (Nov. 9, 1876). Proc. London Math. Soc.
vm. (1877), pp. 4751.
A Generalised Form of Certain Series (May 9, 1878). Proc. London Math. Soc. ix. (1878),
pp. 197202.
On the Solution of a Differential Equation allied to Riccati's. British Association Report,
1878, pp. 469470.
764 THEORY OF BESSEL FUNCTIONS
Example illustrative of a point in the solution of differential equations by series.
Messenger, vin. (1879), pp. 2023.
On a symbolic theorem involving repeated differentiations (May 19, 1879). Proc. Camb.
Phil. Soc. in. (1880), pp. 269271.
On Riccati's Equation and its Transformations and on some Definite Integrals which
satisfy them (June 16, 1881). Phil Trans, of the Royal Soc. 172 (1881), pp. 759828.
[Proc. Royal Soc. xxxn. (1881), p. 444]
GORDAN, P.
See under Hermite.
GRAF, J. H.
Ueber die Addition und Subtraction der Argumente bei BesseFschen Functionen nebst
einer Anwendung (March, 1893). Math. Ann. xliii. (1893), pp. 136 144.
Ueber einige Eigenschaften der Bessel'schen Function erster Art, insbesondere fur ein
grosses Argument. Zeitschrift fur Math, xxxvm. (1893), pp. 115 120.
Beitrage zur Auflosung von linearen Differentialgleichuugen zweiter Ordnung mit linearen
Coefficienten so wie von Differentialgleichungen zweiter Ordnung denen gewisse bestimmte
Integrale geniigen (March, 1894). Math. Ann. xlv. (1894), pp. 235262.
Relations entre la fonction Besselienne de l re espece et une fraction continue (May, 1894).
Ann. di Mat. (2) xxm. (1895), pp. 4565.
Ableitung der Formeln fiir die Bessel'schen Functionen bei welchen das Argument ein
Distanz darstellt (Aug. 4, 1896). Verhandlungen der Schweiz-Naturf. Ges. lxxix. (1896),
pp. 5962.
Einleitung in die Theorie der Besselschen Funktionen. Von J. H. Graf und E. Gubler
(2 HefteBern, 1898, 1900).
;
Beitrag zur Auflosung von Differentialgleichungen zweiter Ordnung denen gewisse be-
stimmte Integrale geniigen (May, 1902). Math. Ann. lvi. (1903), pp. 423 444.
GRAY, A.
A Treatise on Bessel Ftmctions. By Andrew Gray and G. B. Mathews (London, 1895).
Ja {x)e- hx xc ~ l
dx.
/:
Inaugural-dissertation, Zurich, 1894 (38pp.). [Graf and Gubler, Einleitung in die Theorie
der Besselschen Funktionen, n. (Bern, 1900), pp. 110 135, 156.]
Ueber ein discontinuierliches Integrale (Dec. 1895). Math. Ann. xlviii. (1897),.pp. 37 48.
Beweis einer Formel des Herrn Sonine (Dec. 1896). Math. Ann. xlix. ( 1897), pp. 583 584.
Ueber bestimmte Integrale mit Bessel'schen Functionen (Oct. 1902). Zurich Viertel-
jahrsschrift, XLVll. (1902), pp. 422428.
GUNTHER, S.
Bemerkungen iiber Cylinderfunctionen. Archiv der Math, und Phys. lvi. (1874),
pp. 292297.
* See also under Sir Joseph John Thomson.
t See also under Graf.
BIBLIOGRAPHY 765
GWYTHER,
R. F.
The employment of a geometrical construction to prove Schlomilch's series, and to aid
in its development into a definite integral. Messenger, xxxm. (1904), pp. 97 107.
HADAMARD, J.
Sur l'expression asymptotique de la fonction de Bessel. Bulletin de la Soc. Math, de
France, xxxvi. (1908), pp. 77 85.
HAENTZSCHEL, E.
Ueber die functionentheoretisehen Zasammenhang zwischen den Lame'schen, Laplace'-
schen und Bessel'schen Functioned. Zeitschrift fiir Math. xxxi. (1886), pp. 25 33.
Ueber die Fourier Bessel'sche Transcendeute (Nov. 20, 1887). Zeitschrift fiir Math.
xxxm. (1888), pp. 185186.
HAFEN, M.
Studien iiber einige Probleme der Potentialtheorie. Math. Ann. lxix. (1910), pp. 517 537.
HAGUE, B.
A Note on the Graphs of the Bessel Functions of Integral Order. Proc. Phys. Soc.
xxix. (1917), pp. 211214.
HALL, A.
The Besselian Function. The Analyst, I. (1874), pp. 8184
HAMILTON, SIR WILLIAM ROWAN*.
On Fluctuating Functions (June 22, 1840). Trans. R. Irish Acad. xix. (1843), pp. 264321.
On the Calculation of the Numerical Values of a certain class of Multiple and Definite
Integrals (Sept. 29, 1857). Phil. Mag. pp. 375382.
(4) xiv. (1857),
HANKEL,
H.
Die Cylinderfunctionen erster und zweiter Art (Dec. 15, 1868). Math. Ann. I. (1869),
pp. 467501.
Bestimmte Integrale mit Cylinderfunctionen +. Math. Ann. vm. (1875), pp. 453 470.
Die Fourier'schen Reihen und Integrale fiir Cylinderfunctionen t (May 16, 1869). Math.
Ann. viu. (1875), pp. 471-494.
HANSEN, P. A.
Ermittelung der absoluten Storungen in Ellipsen von beliebiger Excentricitat und
Neigung, I. Schriften der Stermoarte Seeberg (Gotha, 1843). [Memoire sur la determination
des perturbations absolues dans les ellipses d'une excentricite et d'une inclinaison quel-
conques. Par M. Hansen. Traduit de l'AUemand par M. Victor Mauvais (Paris, 1845).]
Eutwickelung des Products einer Potenz des Radius Vectors mit den Sinus oder Cosinus
eines Vielfachen der wahren Anomalie in Reihen. Leipziger Abh. II. (1855), pp. 181 281.
HANUMANTA RAO, C. V.
On a certain definite integral. Messenger, xlvii. (1918), pp. 134 137.
HARDY, G. H.
General theorems in contour integration with some applications. Quarterly Journal,
:
HARTENSTEIN, J. H.
Integration der Differentialgleichung
^ + ^=i / s fur elliptische und parabolische
Coordinaten. Archiv der Math, und Phys. (2) xiv. (1896), pp. 170 199.
HATTENDORF, K.
See under Riemann.
HAVELOCK, T. H.
Mathematical Analysis of Wave Propagation in Isotropic Space of p Dimensions
(March 16, 1904). Proc. London Math. Soc. (2) n. (1904), pp. 122137.
HAYASHI,
T.
On a definite integral for cylindrical function. Nyt Tidsskrift, xxiu. B
Neumann's
(1912), pp. 8690. [Jahrbuch fiber die Fortschritte der Math. 1912, p. 555.]
8 8
On the Integrals Texp Lr? p6 Y?
q6d6 (Dec. 1920). Tdhoku Math. Journal, xx.
(1922), pp. 107114.
HEAVISIDE, O.
Electrical Papers, i., n. (London, 1892).
On Operators in Physical Mathematics (Dec. 15, 1892 ; June 8, 1893). Proc. Royal Soc.
Lll. (1893), pp. 504529; Liv. (1893), pp. 105143.
Electromagnetic Theory +, n., in. (London, 1899, 1912).
HEINE, H. E.
Ueber die Zahler und Nenner der Naherungswerthe von KettenbruchenJ (Sept. 1859).
Journal fur Math. lvii. (1860), pp. 231 247.
Die Fourier-Besselache Function (June, 1868). Journal fur Math. lxix. (1869), pp. 128
141.
Handbuch der Kugelfunctionen: Theorie und Anwendunqen (2 Bande) (Berlin, 1878,
1881).
HERMITE, C.
Sur E
transcendante n Ann. di Mat. (2) in. (1870), p. 83.
la .
Extrait d'une lettre de Monsieur Ch. Hermite a Monsieur Paul Oordan (June 9, 1873).
Journal fUr Math, lxxvi. (1873), pp. 303 311.
Extrait d'une lettre a M. E. Jahnke (Nov. 25, 1900). Archiv der Math, und Phys. (3)
I. (1901), pp. 2021.
HERTZ, H.
Uber die Induktion in rotierenden Kugeln. Dissertation, Berlin, March 15, 1880. [Oes.
Werlee, I. (Leipzig, 1895), pp. 37 134]
Uber das Gleichgewicht schwimmender elasticher Platten. Ann. der Physik und Chemie,
(3) xxii. (1884), pp. 449455. [Oes. Werlee, I. (Leipzig, 1895), pp. 288294.]
HERZ, N.
Bemerkungen zur Theorie der Bessel'schen Functionen (Sept. 21, 1883). Astr. Nach.
cvn. (1884), col. 1728.
Note, betreffend die Entwicklung der storenden Krafte (Mar. 30, 1884). Astr. Nach.
evil. (1884), col. 429432.
HILB, E.
Zur Theorie der Entwicklungen willkurlicher Funktionen nach Eigenfunktionen (Sept.
11, 1917). Math. Zeitschrift, I. (1918), pp. 5869.
Uber die Laplacesche Reihe (March 15, 1919 ; Nov. 17, 1919). Math. Zeitschrift. v.
(1919), pp. 1725 ; vm. (1920), pp. 7990.
HILL, C. J. D.
De radicibus rationalibus aequationis Riccatianae
dxV + a + by+cy^O,
ubi a, b, c functiones rationales ipsius x (May 24, 1840). Journal fur Math. xxv. (1843),
pp. 2237.
HOBSON, E. * W
Systems of Spherical Harmonics (June 11, 1891). Proc. London Math. Soc. xxn. (1891),
V
pp. 431449. "
On the Evaluation of a certain Surface- Integral, and its application to the Expansion,
in Series, of the Potential of Ellipsoids (Jan. 12, 1893). Proc. London Math. Soc. xxiv
(1893), pp. 8096.
On Bessel's Functions, and Relations connecting them with Hyper-spherical and
Spherical Harmonics (Dec. 14, 1893). Proc. London Math. Soc. xxv. (1894),
pp. 4975.
On the most general solution of given Degree of Laplace's Equation (May 9, 1895)
Proc. London Math. Soc. xxvi. (1895), pp. 492 494.
Note on some properties of Bessel's Functions (Jan. 14, 1897). Proc. London Math. Soc.
xxvni. (1897), pp. 370375.
On the representation of a function by series of Bessel's functions (Dec 10 1908)
'
Proc. London Math. Soc. (2) vn. (1909), pp. 359388.
HOPF,
L. UND SOMMERFELD,
A. J. W.
Uber komplexe Integraldarstellungen der Zylinderfunktionen. Archiv der Math, und
Phys. (3) xviii. (1911), pp. 116.
HORN, J.
Ueber lineare Differentialgleichungen mit einem veranderlichen Parameter (Dec 30.
1898). Math. Ann. lii. (1899), pp. 340362.
HURWITZ, A.
Ueber die Nullstellen der Bessel'schen Function (June 2, 1888). Math. Ann xxxm
(1889), pp. 246266.
Ueber die Wurzeln einiger transcendenten Gleichungen. Hamburqer Mittheilunqen,
II. (1890), pp. 25- 31. [Jahrbuch fiber die Fortschritte der Math. 1890, p. 115.]
HYMERS, J.
Treatise on Differential Equations, and on the Calculus of Finite Differences (Cambridge
loO?j
IGNATOWSKY, W. VON.
Uber die Reihenentwicklungen mit Zylinderfunktionen (May 13, 1911). Archiv der
Math, und Phys. (3) xviii. (1911), pp. 322327.
Uber Reihen mit Zylinderfunktionen nach dem Vielfachen des Argumentes (Dec 1913)
Archiv der Math, und Phys. (3) xxm. (1915), pp. 193219.
ISELI, F.
Die Riccati'sche Gleichung. Dissertation, Bern, 1909 (42 pp.). [Jahrbuch Uber die Fort-
*r , L
schrUteder Math. 1909, p. 369.]
ISHERWOOD, J. G.
Tables of the Bessel Functions for pure imaginary values of the argument (April 26.
1904). Manchester Memoirs, xlviii. (1903
4), no. 19. .
JACKSON, F. H.
Generalised forms of the series of Bessel and Legendre. Proc. Edinburgh Math. Soc.
XXI. (1903), pp. 6572.
On Generalised functions of Legendre and Bessel (Nov. 16, 1903). Trans. Edinburgh
Royal Soc.xh. (1905), pp. 128. y
* See also under Gegenbauer.
;
JACOBI, C. G. J *
Formula transformationis integralium definitorum (July 9, 1835). Journal fur Math.
xv. (1836), pp. 126. [Ges.Math. Werke, vi. (1891), pp. 86 118.] Formule pour la trans-
formation d'une classe d'integrales definies, Journal de Math. I. (1836), pp. 195 196.
Versuch einer Berechnung der grossen Ungleichheit des Saturns nach einer strengen
Entwicklung. Astr. Nach. xxvm. (1849), col. 6580, 8194. [Ges. Math. Werke, vn. (1891),
pp. 145174.]
Uber die annahernde Bestimmung sehr entfernter Glieder in der Entwickelung der
elliptischen Coordinaten, nebst einer Ausdehnung der Laplacefschen Methode zur Bestim-
mung der Functionen grosser Zahlen. Astr. Nach. xxvm. (1849), col. 257270. [Ges. Math.
Werke, vn. (1891), pp. 175188.]
JAHNKE, P. E. E.+
Funktionentafeln mit Formeln und Kurven. Von E. Jahnke und F. Emde (Leipzig, 1909).
Uber einige, in der elektromagnetischen Strahlungstheorie auftretende, bestimmte In-
tegrate. Archie der Math, und Phys. (3) xxm. (1914), pp. 264267.
JAMET, E. V.
Sur les equations anharmoniques (Sept. 13, 1901). Comptes Rendus de V Assoc. Fran-
caise,xxx. (Ajaccio) (1901), pp. 207228.
Sur les equations anharmoniques. Ann. de la Fac. des Sci. de Marseille, xn. (1902),
r 121.
pp.
JEKHOWSKY, B.
Les fonctions de Bessel de plusieurs variables exprimees par des fonctions de Bessel
d'une variable (Feb. 28, 1916). Comptes Rendus, clxii. (1916), pp. 318319.
Sur la fonction generatrice des fonctions de Bessel a plusieurs variables. Bulletin des
Sci. math. (2) xli. (1917), pp. 5860.
les fonctions de Bessel a deux variables
Sur (May 30, 1921). Comptes Rendus, clxxii.
(1921), pp. 13311332.
JOHNSON, W. W.
On the Differential Equation
g +y + py+<2 =0.
2
JULIUS, V. A.
Sur les fonctions de Bessel de deuxieme espece (Dec. 1893). Archives Neerlandaises,
xxvm. (1895), pp. 221225.
Sur les ondes lumineuses sphenques et cylindriques (Dec. 1893). Archives Neerlandaises,
XXVIII. (1895), pp. 226244.
KALAHNE, A.
tJber die Wurzeln einiger Zylinderfunktionen und gewisser aus ihnen gebildeter Gleich-
ungen (March 1, 1906). Zeitschrift fur Math, und Phys. liv. (1907), pp. 5586.
KAPTEYN, W*
Nouvelles formules pour representer la fonction Jn _Ax). Bulletin des Sci. Math. (2)
xvi. (1892), pp. 4144.
Over BESSEL'sche Functien (March 12, 1892). Nieuw Archief voor Wiskunde, xx. (1893).
V
pp. 116127. r
Recherches sur les fonctions de Fourier- Bessel. Ann. de I'Ecole norm, sun
sci.
r (3) x
\ /
(1893), pp. 91120. '
Sur quelques integrates definies con tenant des fonctions de Bessel. Archives Ne'er-
landaises, (2) vi. (1901), pp. 103116.
A definite integral containing Bessel's functions (June 29, 1901). Proc. Section of Sci.,
K. Akad. van Wet. te Amsterdam, iv. (1902), pp. 102103.
Sur un developpement de M. Neumann. Nieuw Archief voor Wiskunde, (2)i vi. (1905).
*\ h
pp. 4955. \
Einige Bemerkungen iiber Bessel'sche Functionen. Monatshefte fur Math, und Phys.
xiv. (1903), pp. 275282.
The values of some definite integrals connected with Bessel functions (Nov. 26, 1904).
Proc. Section of Sci., K. Akad. van Wet. te Amsterdam, vn. (1905), pp. 375 376.
On a series of Bessel functions (Dec. 24, 1904). Proc. Section of Sci., K. Akad. van Wet
te Amsterdam, vn. (1905), pp. 494500.
A definite integral of Kummer (Sept. 30, 1905). Proc. Section of Sci., K. Akad. van
Wet. te Amsterdam, vyi. (1906), pp. 350 357.
On an expansion of an arbitrary function in a series of Bessel functions. Messenger,
xxxv. (1906), pp. 122125.
Sur la sommation d'une serie infinie. Nieuw Archief voor Wiskunde, (2) vn. (1907), pp.
The quotient c-f two successive Bessel Functions (Dec. 30, 1905 ; Jan. 27, 1906). Proc.
Section of Sci., K. Akad. van Wet. te Amsterdam, viii. (1906), pp. 547 549, 640 642.
Sur le quotient de deux fonctions besseliennes successives. Archives Neerlandaises, (2)
xi. (1906), pp. 149168.
Recherches sur les fonctions cylindriques. Mem. de la Soc. R. des Sci. de Liege, (3) vi.
(1906), no. 5.
Sur le calcul numerique de la serie
l
I .
q
*~
(a2 +0V)2
Mem. de la Soc. R. des Sci. de Liege, (3) vi. (1906), no. 9.
On some relations between Bessel's Functions (Feb. 24, 1912). Proc. Section of Sci. ' K.
Akad. van Wet. te Amsterdam, xiv. (1912), pp. 962 969.
KELVIN, LORD.
On
the waves produced by a single impulse in water of any depth or in a dispersive
medium (Feb. 3, 1887). Proc. Royal Soc. xlii, (1887), p. 80; Phil. Mag. (5) xxin. (1887)
"pp.352 255. -{MaiK. and Phys. Paper*, iv. (1910), pp. 303^306.]
Ether, Electricity and Ponderable Matter (Jan. 10, 1889). Journal of Inst,
of ElectHcal
Engineers, xvni. (1890), pp. 437. [Math, and Phys. Papers, in. (1890),
pp. 484515.]
KEPINSKI, S.
TJber die Differentialgleichung
o 22 m+1 dz n dz _
dx* x cvr x dt~
(Jan. 1905). Math. Ann. lxi. (1906), pp. 397406.
Integration der Differentialgleichung
w. b. F. 26
770 THEORY OF BESSEL FUNCTIONS
KIRCHHOFF, G.
Ueber den inducirten Magnetisrnus eines unbegrenzten Cylinders von weichem Eisen
(June, 1853). Journal fiir Math, xlviii. (1854), pp. 348 376.'
Zur Theorie der Bewegung der Elektricitat in unterseeischen oder unterirdischen Tele-
graphendrahten (Oct. 29, 1877). Berliner Monatsberichte, 1877, pp. 598 611.
Ueber die Transversalsohwingungen eines Stabes von veranderlichem Querschnitt
(Oct. 27, 1879). Berliner Monatsberichte, 1879, pp. 815
828; Ann. der Physik und C/iemie,
(3) x. (1880), pp. 501512.
Ueber die elektrischen Strdraungen in einera Kreiscylinder (April 26, 1883). Berliner
Sitzungsberichte, 1883, pp. 519 524.
KLUYVER, J. C.
A local probability problem (Sept. 30, 1905). Proc. Section of Sci., K. Akad. van Wet.
te Amsterdam, vm. (1906), pp. 341350.
An integral theorem of Gegenbauer (Feb. 27, 1909). Proc. Section of Sci., K. Akad.
van Wet. te Amsterdam, xi. (1909), pp. 749 755.
KNESER,
J. C. C. A.
Die Entwicklung willkiirlicher Funktionen in Reihen die nach Bessel'schen Funktionen
fortschreiten (July, 1903). Archie der Math, und Phys. (3) vn. (1903), pp. 123133.
Die Theorie der Integralgleichungen und die Darstellung willkiirlicher Funktionen in
der mathematischen Physik. Math. Ann. lxiii. (1907), pp. 47s7 524.
KNOCKENHAUER, K. W.
Ueber die Oerter der Maxima und Minima des gebeugten Lichtes nach den Fresnel'schen
Beobachtungen. Ann. der Physik und Chemie, (2) xli. (1837), pp. 103 110.
KONIG, J.
Ueber die Darstellung von Functionen durch unendliche Reihen (Sept. 1871). Math.
Ann. v. (1872), pp. 310340.
Ueber Reihenentwicklung nach Bessel'schen Functionen (1880). Math. Aim. xvn.
(1880), pp. 8586.
KOESTLER, W.
Beitrage zu Reihenentwickelungen nach Besselsche Zylinderfunktionen. Dissertation,
Bern, 1907 (110 pp.). [Jahrbuch fiber die Fortschritte der Math. 1908, p. 535.]
KOPPE, M.
Die Ausbreitung einer Erschutterung an der Wellenmaschine darstellbar durch einen
neuen Grenzfall der Bessel'schen Functionen. Programm (96). Andreas Realgymn. Berlin,
1899 (28 pp.). [Jahrbuch Uber die Fortschritte der Math. 1899, pp. 420 421.]
E. E. KUMMER,
Sur l'integration de ^equation de Riccati par des integrales definies. Journal fiir Math.
XII. (1834), pp. 144147.
Ueber die hypergeometrische Reihe
1
1 |
a.g
+ l.y J .
|
"*"
a(" + l)/8 09 + l)
1.2.y(y+l)
Journal fur Math. xv. (1836), pp. 3983, 127172.
"
ra ,
a(a+l)(a + 2)j8Q3 + l)03+2)
1 2 3 y (y+1) (y + 2)
. . . ^ + --
De integralibus quibusdam definitis et seriebus infinitis (April, 1837). Journal fiir
Math. xvn. (1837), pp. 228242.
Note sur l'integration de l'equation T-f=^m -y- Journal fiir Math. xix. (1839), pp.
286288.
Sur l'iutegration de l'equation -^=3?" .y. Journal de Math. iv. (1839), pp. 390 391.
LAGRANGE, J. L. DE.
Sur le probleme de Kepler (Nov. 1770). Hist, de V Acad. R. des Sci. de Berlin, xxv.
1,
(1769) [1771], pp. 204233. [Oeuvres, in. (1869), pp. 113138.]
LAMB, H.
On the Oscillations of a Viscous Spheroid (Nov. 10, 1881). Proc. London Math. Soc.
XIII. (1882), pp. 5166.
On the Vibrations of an Elastic Sphere (May 11, 1882). Proc. London Math. Soc. xm.
(1882), pp. 189212.
BIBLIOGRAPHY 771
On Electrical Motions in a Spherical Conductor (March 14, 1883). Phil. Trans, of the
Royal Soc. clxxiv. (1883), pp. 519549;
On the Induction of Electric Currents in Cylindrical and Spherical Conductors (Jan.
10, 1884). Proc. London Math. Soc. xv. (1884), pp. 139149.
Note on the Induction of Electric Currents in a Cylinder placed across the Lines of
Magnetic Force (June 12, 1884). Proc. London Math. Soc. xv. (1884), pp. 270274.
On the Motion of a Viscous Fluid contained in a Spherical Vessel (Nov. 13, 1884).
Proc. London Math. Soc. xvi. (1885), pp. 2743.
A problem in Resonance illustrative of the Theory of Selective Absorption of Light
(Jan. 8, 1900). Proc. London Math. Soc. xxxn. (1901), pp. 1120.
Problems relating to the Impact of Waves on a Spherical Obstacle in an Elastic Medium
(March, 8, 1900). Proc. London Math. Soc. xxxn. (1901), pp. 120150.
On Boussinesq's Problem (Feb. 13, 1902). Proc. London Math. Soc. xxxrv. (1902), pp.
276284.
On Deep-water Waves (Nov. 10, 1904). Proc. London Math. Soc. (2) n. (1905), pp. 371
400.[Presidential Address.]
On the theory of waves propagated vertically in the atmosphere (Dec. 6, 1908). Proc.
London Math. Soc. (2) vii. (1909), pp. 122141.
LAMBERT, A.
See Wangerin, A
LANDAU, E. G. H.
Uber
''.#*
die Gitterpunkte in einem Kreise (May 8, 1915 ; June 5, 19i5 ; June 4, 1920).
Nachrichten von der K. Get. der Wis*, zu Obttingen, 1915, pp. 148160, 161 171 ; 1920,
pp. 109134.
Zur analytischen Zahlentheorie der definiten quadratischen Formen (Uber die Gitter-
punkte in einem mehrdimensionalen Ellipsoid) (May 20, 1915). Berliner Sitzungtberichte
LEBESGUE, V. A
Remarques sur l'equation y"+ -y'+ny*=Q.
+y=0. Journal
Jour de Math. xi. (1846), pp. 338340.
LEFORT, F.
Expression numerique des integrates d^finies qui se presentent quand on cherche les
termes generaux du developpement des coordonnees d'une planete, dans son mouvement
elliptique. Journal de Math. xi. (1846), pp. 142152.
LEIBNIZ, G. W.
See Daniel Bernoulli.
LE PAIGE, C*
Note sur l'equation xy"+iy'-xy^0. Bull, de VAcad. R. de Belgique, (2) xli. (1876),
pp. 10111016.
LERCH, M.
Mittheilungen aus der Integralrechnung. Monatshefte fttr Math, und Phys. I. (1890),
pp. 105112.
Betrachtungen uber einige Fragen der Integralrechnung. Rozpravy, v. (1896), no. 23
'Uber Fortschritte der Math. 1896, pp. 233234.]
(16 pp.). [Jahrbuch
Sur l'integrale
J
cos i(u-x sin u) du. Journal de Math. vi. (1841), p. 36.
Sur une formule de M. Jacobi (March, 1841). Journal de Math. vi. (1841), pp. 6973.
LIPSCHITZ, R. O. S.
Ueber ein Integral der Differentialgleichung ^+ - ~-+/=0 (July 14, 1858). Journal
LODGE, A.
Note on the Semiconvergent Series of Jn (x). British Association Report, York, 1906,
pp. 494 498.
LOMMEL, E. C. J. VON.
Beitrage zur Theorie der Beugung dea Lichts. Archiv der Math, und Phys. xxxvi.
(1861), pp. 385419.
Methode zur Berechnung einer Transcendenten. Archiv der Math und Phys. xxxvn.
(1861), pp. 349360.
Zur Integration linearer Differentialgleichungen. Archiv der Math, und Phys. XL.
(1863), pp. 101126.
Studien fiber die BesseVschen Functionen (Leipzig, 1868).
Integration der Gleichung *" + i jr^n +y=0. Math. Ann. II. (1870), pp. 624635.
Ueber der Anwendung der Bessel'schen Functionen in der Theorie der Beuguug. ^Zeit-
schriftfiir Math, und Phys. xv. (1870), pp. 141169.
Zur Theorie der Bessel'schen Functionen (Dec. 1870). Math. Ann. in. (1871), pp. 475
487.
Zur Theorie der Bessel'schen Functionen (Jan. 1871). Math. Ann. iv. (1871), pp. 103
116.
Ueber eine mit den Bessel'schen Functionen verwandte Function (Aug. 1875). Math.
Ann. ix. (1876), pp. 425 444.
Zur Theorie der Bessel'schen Functionen (Oct. 1878). Math. Ann. xiv. (1879), pp. 510
536.
Zur Theorie der Bessel'schen Functionen (Oct. 1879). Math. Ann. xvi. (1880), pp. 183
208.
Die Beugungserscheinungen einer kreisrunden Oeffnung und eines kreisrunden Schirm-
chens theoretisch und experimentell bearbeitet. Miinchener Abh.xv.(l 884 86), pp. 233 328.
Die Beugungserscheinungen geradlinig begrenzter Schirme. Miinchener Abh. xv.
(188486), pp. 531663.
* The headings of the pages of this paper are " Zur Theorie der Bessel'schen Functionen."
BIBLIOGRAPHY 773
LORENZ, L.
Sur developpement des fonctions arbitraires au moyen de fonctions donnees. Tids-
le
skriftfor Mathematik, 1876, pp. 129144. [Oeuvres Scientifiques, II. (1899), pp. 495513.]
Theorie de la dispersion. Ann. der Phi/sik und Chemie, (3) xx. (1883), pp. 1 21. [Oeuvres
Scientifiques, I. (1898), pp. 371396.]
Sur la lumiere reflechie et refractee par une sphere transparente. K. Danske Videnska-
bernes Selskabs Skrifter, (6) vi. (1890), pp. 1 62. [Oeuvres Scientifiques, I. (1898), pp. 405
502.]
LOVE, A. E. H.
The Scatteringof Electric Waves by a Dielectric Sphere (Feb. 9, 1899). Proc. London
Math. Soc. xxx. (1899), pp. 308 321.
The Transmission of Electric Waves over the Surface of the Earth (Dec. 19, 1914).
Phil. Trans, of the Royal Soc. ccxv. A
(1915), pp. 105 131.
MACDONALD, H. M.*
Note on Bessel Functions (Nov. 11, 1897). Proc. London Math. Soc. xxix. (1898),
pp. 110115.
Zeroes of the Bessel Functions (April 7, 1898 Jan. 12, 1899). Proc. London Math. Soc.
;
Note on the evaluation of a certain integral containing Bessel's functions (Dec. 6, 1908).
Proc. London Math. Soc. (2) vn. (1909), pp. 142149.
The Diffraction of Electric Waves round a Perfectly Reflecting Obstacle (Aug. 13, 1909).
Phil. Trans, of the Royal Soc. ccx. A
(1910), pp. 113144.
The Transmission of Electric Waves around the Earth's Surface (Jan. 10, 1914). Proc.
Royal Soc. xc. A
(1914), pp. 5061.
Formulae for the Spherical Harmonic n ~ m (/*) when 1 - ft is a small quantity (Feb. 6,
P
1914). Proc. London Math. Soc. (2) xm. (1914), pp. 220221.
[Extract from a letter to Prof. Carslaw, Oct. 17, 1912.] Proc. London Math. Soc. (2) xm.
(1914), pp. 239240.
A
class of diffraction problems (Jan. 14, 1915). Proc. London Math. Soc. (2) xiv. (1916),
pp. 410427.
The Transmission of Electric Waves around the Earth's Surface (May 23, 1916). Proc.
Royal Soc. xcn. A (1916), pp. 433437.
McMAHON, J.
On the Descending Series for Bessel's Functions of Both Kinds. Annals of Math. vm.
(1894), pp. 5761.
On the Roots of the Bessel and certain Related Functions. Annals of Math, ix. (1895),
pp. 23 30.
The Expression for a Rational Polynomial in a Series of Bessel functions of nth Order
as required in certain Cases of Dirichlet's Problem. Proc. American Assoc. 1900, pp. 42
43.
MacROBERT, T. M.
The Modified Bessel Function Kn (z) (Feb. 13, 1920). Proc. Edinburgh Math. Soc.
xxxviii. (1920), pp. 1019.
Asymptotic Expressions for the Bessel Functions and the Fourier-Bessel Expansion
(Jan. 14, 1921). Proc. Edinburgh Math. Soc. xxxix. (1921), pp. 1320.
MAGGI, G. A.
Sulla storia delle funzioni cilindriche. Atti della R. Accad. dei Lincei, Ser. 3 (Transunti),
vol iv. (1880), pp. 259263.
Sopra un problema di elettrostatica (June 3, 1880). Rend, del R. 1st. Lombardo, (2) xm.
(1880), pp. 384390.
MALMSTEN, C. J.
cos ax dx
/* -jz -^ K' Svenska V. Akad. Handlingar, LXn. (1841), pp. 65 74.
m
De liquation differentielle yt"+^- + Aaf y=0 (March 18, 1849). Journal fur Math.
Thebremes sur l'integration de l'Equation -r\ + ~j ='( ^m +Zi) V- Gamb. and Dublin
Math. Journal, v. (1850), pp. 180182.
MALTEZOS, C.
Sur la chute des corps dans le vide et sur certaines fonctions transcendantes. Nouvelles
Annates de Math. (4) xi. (1902), pp. 197204.
MANFREDIUS, G.
De construetione aequationum differentialium primi gradus (Bologna, 1707).
MARCH, H. W.
Uber die Ausbreitung der Wellen der drahtlosen Telegraphic auf der Erdkugel (Oct. 21,
1911). Ann. der Physilc und Chemie, (4) xxxvn. (1912), pp. 29 50.
MARCOLONGO, R.
Alcuni teoremi sulle funzioni cilindriche di prima specie (April 6, 1889). Napoli Ren-
diconto, (2) m. (1889), pp. 9699.
MARSHALL, W.
On a New Method of Computing the Roots of Bessel's Functions (Oct. 1909). Annals
of Math. (2) xi. (1910), pp. 153160.
MATHEWS, G. B.
See under Gray.
MAY
ALL, R. H. D.
On the Diffraction Pattern near the Focus of a Telescope (Feb. 22, 1897). Proc. Camb.
Phil. Soc. IX. (1898), pp. 259269.
MEECH, L. W.
Integration of Riccati'a Equation. Annals of Math. I. (1886), pp. 97103 ; in. (1887),
pp. 4749.
MEHLER, F. G.
Ueber die Vertheilung der statischen Electricitat in einem von zwei kugelkalotten be-
grenzten Korper (July, 1867). Journal fur Math, lxviii. (1868), pp. 134 150.
Ueber eine mit den Kugel- und Cyhnderfunctionen verwandte Function und ihre An-
wendung in die Theorie der Electricitatsvertheilung. '[EUnng Jahresbericht, 1870.] Math.
Ann. xviii. (1881), pp. 161194.
Ueber die Darstellung einer willkurlichen Function zweier Variablen durch Cylinder-
function* (Dec. 1, 1871). Math. Ann. v. (1872), pp. 135140.
Notiz uber die Dirichlet'schen Integralausdrucke fur die Kugelfunctionen P" (cos S)
und iiber eine analoge Integralform fur die Cylinderfunction J(x)^ (Dec. 2, 1871). Math.
Ann. v. (1872), pp. 141144.
MEISSEL, D. F. E.
Iserlohn Programm, 1862. [Nielsen, Nouvelles Annates de Math. (4) II. (1902), pp. 396
410J
Tafel der Bessel'schen Functionen 7 und i"*1 von -0 bis =15-5 (Nov. 8, 1888).
fc
[Jahrbuch Uber die Fortschritte der Math. 1890, pp. 521 522.]
Einige Entwickelungen die Bessel'schen /-Functionen betreffend (May 3, 1891). Astr.
Nach. cxxvu. (1891), col. 359362.
Beitrag zur Theorie der allgemeinen Bessel'schen Function (June 11, 1891). Astr. Nach.
CXXVIH. (1891), col. 145154.
Abgekurzte Tafel der Bessel'schen Functionen Ikh (July 10, 1891). Astr. Nach. cxxvin.
(1891), col. 154155.
Beitrag zur Theorie der Bessel'schen Functionen (Oct. 7, 1891). Astr. Nach. cxxvm.
(1891), col. 435438.
* The headings of the pages of this paper are "Ueber die Cylinderfunction J(x)."
t The headings of the pages of this paper are "Notiz iiber die Functionen P" (cos a) und J (x)."
BIBLIOGRAPHY 775
Neue Entwickelungen uber die Bessel'schen Functionen (Jan. 25, 1892). Astr. Nach.
cxxix. (1892), 281284.
col.
Weitere Entwickelungen uber die Bessel'schen Functionen (May 2, 1892). Astr. Nach.
cxxx. (1892), col. 363368.
Uber die Absoluten Maxima der Bessel'schen Functionen. Programm, Oberrealschule,
Kiel, 1892 (11 pp.). [Jakrbuch iiber die Fortschritte der Math. 1892, pp. 476 478.]
MELLIN, R. HJ.
Abriss einer einheitlichen Theorie der Gamma und der hypergeometrischen Funktionen
(June, 1909). Math. Ann. lxviil (1910), pp. 305337.
MICHELL, J. H.
The Wave Resistance of a Ship* (Aug. 9, 1897). Phil. Mag. (5) xlv. (1898), pp. 106123.
MOLINS, H.
m
Sur l'integration de l'equation differentielle -j^c =ax y (March 6, 1876). Mem. de VAcad.
des Sci. de Toulouse, (7) vm. (1876), pp. 167189.
MOORE, C. N.
Note on the roots of Bessel functions. Annals of Math. (2) IX. (1908), pp. 156 162.
The summability of the developments in Bessel functions with applications (Sept. 11,
1908). Trans. American Math. Soc. x. (1909), pp. 391
On the uniform convergence of the
135.
developments
in Bessel functions (Oct. 30, 1909).
Trans. American Math. Soc. xn. (1911), pp. 181 206.
A
continuous function whose development in Bessel's functions is non-summable of
certain orders (Sept. 4, 1916). Bulletin American Math. Soc. xxiv. (1918), pp. 145 149.
On the summability of the developments in Bessel's functions (Sept. 4, 1917). Trans.
American Math. Soc. xxi. (1920), pp. 107 156.
MORTON, W. B.
The Value
of the Cylinder Function of the Second Kind for Small Arguments (Oct. 25,
1900). Nature, lxiii. (1901), p. 29.
MURPHY, R.
On the general properties of definite integrals (May 24, 1830). Trans. Camb. Phil. Soc.
ill. (1830), pp. 429443.
MYLLER-LEBEDEFF, W.
See under Lebedeff.
NAGAOKA, H.
Diffraction Phenomena produced by an Aperture on a Curved Surface. Journal of the
Coll. of Sci. Imp. Univ. Japan, iv. (1891), pp. 301 322.
pp. 331340.
Evaluation nouvelle des integrales ind^finies et des series infinies contenant une fonc-
tion cylindrique (May 23, 1900). Ann. di Mat. (3) vi. (1901), pp. 43115.
Sur une classe de series infinies analogues a celles de Schlomilch selon les fonctions
cylindriques (Aug. 17, 1900). Ann. di Mat. (3) vi. (1901), pp. 301329.
Recherche sur les des fonctions cylindriques dues a MM. C. Neumann et
series,
W. Kapteyn. Ann. de VEcole norm. sup. (3) xvm. (1901), pp. 3975.
sci.
Note sur la convergence d'une sene neumannienne de fonctions cylindriques (Feb. 10,
1901). Math. Ann. lv. (1902), pp. 493496.
Recherches sur une classe de series infinies analogue a celle de M. W. Kapteyn
(April 13, 1901). Oversigt K. Danske Videnskabernes Selskabs, 1901, pp. 127 146.
Sur les series de factorielles (Jan. 20, 1902). Comptes 7fenrfiw,cxxxiv.(1902),pp.l57 160.
Theorie nouvelle des series asymptotiques obtenues pour les fonctions cylindriques et
pour les fonctions analogues (March 16, 1902). Oversigt K. Danske Videnskabernes Selskabs,
1902, pp. 117177.
* See also under Datta and under Dendy. t See also under Sonine.
BIBLIOGRAPHY 777
OLBRICHT, R.
Studien iiber die Kugel- und Cylinderfunctionen (June 2, 1886). Nava Acta Acad.
Caes. Leop. {Halle), 1888, pp. 148.
OLTRAMARE, G.
Note sur l'integrale
""
^y^d*.
/:
Comptes Rendu* de V Assoc. Francaise, xxiv. (1895), part I. p. 182 ;
part n. pp. 167171.
ONO, A.
On the First Root of Bessel Functions of Fractional Order (April 2, 1921). Phil. Mag.
(6) xml (1921), pp. 10201021.
ORR, W. McF.
On Divergent (or Semiconvergent) Hypergeometric Series (May 16, 1898; April 3,
1899). Trans. Camb. Phil. Soc. xvn. (1899), pp. 171199, 283290.
On the product Jm (x)Jn (x) (May 15, 1899). Proc. Camb. Phil. Soc. x. (1900), pp.
93100.
Extensions of Fourier's and the Bessel-Fourier Theorems (Dec. 14, 1908). Proc. R. Irish
Acad, xxvii. A (1910), pp. 205248.
OSEEN, C. W.
Neue Losung des Sommerfeldschen Dinraktionsproblems (Jan. 10, 1912). Arlciv for
Mat. Astr. och Fysik, vn. (1912), no. 40.
OTTI, H.
Eigenschaften Bessel'scher Funktionen n16 ' Art. Bern Mittheilungen, 1898 [1899], pp.
6196.
PAOLI, P.
Sopra gl' integrali definiti (Oct. 8, 1827). Mem. di Mat. e di Fis. della Soc. Italiana delle
Sci. (Modena), xx. (1828), pp. 161182.
Sull' integrazione dell' equazione
PERRON, 0.
Uber die Kettenbruchentwicklung des Quotienten zweier Bessel'schen Functionen
(Dec. 7, 1907). Miinchener Sitzungsberichte, xxxvii. (1907), pp. 483 504.
PETZVAL, J.
Integration der linearen Differential-Qleichungen, I. (Vienna, 1851).
PICARD, C. E.
Application de la theorie des complexes lineaires a l'etude des surfaces et des courbes
gauches. Ann. sci. de VEcole norm. sup. (2) vi. (1877), pp. 329 366.
PINCHERLE, S.
Sopra alcuni sviluppi in serie per funzioni analitiche (Jan. 12, 1882). Bologna Memorie,
(4) in. (1881), pp. 151180.
Alcuni teoremi sopra gli sviluppi in serie per funzioni analitiche (March 23, 1882).
Rend, del R. 1st. Lombardo, (2) xv. (1882), pp. 224225.
Delia trasformazione di Laplace e di alcune sue applicazioni (Feb. 27, 1887). Bologna
Memorie, (4) vm. (1887), pp. 125143.
PLANA, G. A. A.
d?v
Note sur integration de Pequation -r^| + <7*'" y=0 (Jan. Mem. Accad
,
20, 1822). della R.
POCHHAMMER, L.
Ueber die lineare Differentialgleichung zweiter Ordnung mit linearen Coefficienten.
Math. Ann. xxxvi. (1890), pp. 84 96.
Ueber einige besondere Falle der linearen Differentialgleichung zweiter Ordnung mit
linearen Coefficienten (Sept. 1890). Math. Ann. xxxvin. (1891), pp. 225 246.
Ueber eine binomische lineare Diflerentialgleichung nter Ordnung (Sept. 1890). Math.
Ann. xxxvm. (1891), pp. 247262.
Ueber die Differentialgleichung der allgemeineren .F-Reihe (Jan. 1891). Math. Ann.
xxxvm. (1891), pp. 586-597.
Ueber eine specielle lineare Differentialgleichung 2 ter Ordnung mit linearen Coefficienten
(May, 1891). Math. Ann. xli. (1893), pp. 174178.
Ueber die Differentialgleichungen aer Reihen ${p, <r x) und $(p, a, r x) (June, 1891).
; ;
BIBLIOGRAPHY 779
Suite du Memoire sur lea integrates denies et sur la sommation des series insere" dans
les precedens volumes de ce Journal. Journal de VHcole Polytechnique, xii. (cahier 19)
(1823), pp. 404509.
Memoire sur le calcul des variations (Nov. 10, 1831). Mem. de VAcad. R. des Sci. xn.
(1833), pp. 223331.
Sur le developpement des coordonnees d'une planete dans son mouvement elliptique et
la fonction perturbatrice de ce mouvement. Connaissance des Terns, 1836 [1833], pp. 3-31.
La Theorie de la Chaleur (Paris, 1835).
PORTER, M. B.
Note on the roots of Bessel's Functions. Bulletin American Math. Soc. iv. (1898),
pp. 274275.
On the Roots of the Hypergeometric and Bessel's Functions (June 3, 1897). American
Journal of Math. xx. (1898), pp. 193214.
On the roots of functions connected by a linear recurrent relation of the second order
(Feb. 1901). Annals of Math. (2) m. (1902), pp. 5570.
PUISEUX, V.
la convergence des series qui se presentent dans la theorie du mouvement elliptique
Sur
des planetes. Journal de Math. xiv. (1849), pp. 33 39.*
Seconde note sur la convergence des series du mouvement elliptique. Journal de Math.
xiv. (1849), pp. 242246.
PURSER, F.
On
the application of Bessel's functions to the elastic equilibrium of a homogeneous
isotropic cylinder (Nov. 11, 1901). Trans. R. Irish Acad. xxxn. (1902), pp. 31 60.
Some applications of Bessel's functions to Physics (May 14, 1906). Proc. R. Irish Acad.
xxvi. A(1907), pp. 2566.
RAFFY, L.
Une lecon sur l'equation de Riccati. Nouv. Ann. de Math. (4) n. (1902), pp. 529 545.
RAMANUJAN, S.
A class of definite integrals. Quarterly Journal, xlviii. (1920), pp. 294 310.
RAWSON,
R.
On Cognate Riccatian Equations (1876). Messenger, vn. (1878), pp. 69 72.
Note on a transformation of Riccati 's equation. Messenger, xii. (1883), pp. 34 36.
RAYLEIGH (J. W. STRUTT), LORD.
Onthe Vibrations of a Gas contained within a Rigid Spherical Envelope (March 14,
1872). Proc. London Math. Soc. iv. (1873), pp. 93103.
Investigation of the Disturbance produced by a Spherical Obstacle on the Waves of
Sound (Nov. 14, 1872). Proc. London Math. Soc. iv. (1873), pp. 253 283.
Notes on Bessel's Functions. Phil. Mag. (4) xliv. (1872), pp. 328344.
Note on the Numerical Calculation of the Roots of Fluctuating Functions (June 11,
1874). Proc. London Math. Soc. v. (1874), pp. 112194. [Scientific Papers, I. (1899),
pp. 190195.]
The Theory of Sound (2 vols. London, 1877, 1878; 2nd edition, 1894, 1896).
On
the Relation between the Functions of Laplace and Bessel (Jan. 10, 1878). Proc.
London Math. Soc. IX. (1878), pp. 6164. [Scientific Papers, I. (1899), pp. 338 341.]
On Images formed without Reflection or Refraction. Phil. Mag. (5) xi. (1881), pp. 214
218. [Scientific Papers, I. (1899), pp.
513 517.]
On the Electromagnetic Theory of Light. Phil. Mag. (5) xii. (1881), pp. 81101.
[Scientific Papers, i. (1899), pp. 518536.]
On the Vibrations of a Cylindrical Vessel containing Liquid. Phil. Mag. (5) xv. (1883),
pp. 385
389. [Scientific Papers, n. (1900), pp. 208211.]
On Point-, Line-, and Plane Sources of Sound (June 14, 1888). Proc. London Math. Soc.
xix. (1889), pp. 504507. [Scientific Papers, in. (1902), pp. 4446.1
On the Theory of Optical Images, with special reference to the Microscope. Phil. Mag.
(5) xlii. (1896), pp. 167195. [Scientific Papers, iv. (1904), pp. 235 260.]
On the Passage of Waves through Apertures in Plane Screens, and Allied Problems.
Phil. Mag. (5) sun, (1897), pp. 259272. [Scientific Papers, iv. (1904), pp. 283296.]
On the Bending of Waves round a Spherical Obstacle (May 23, 1903). Proc. Roy. Soc.
lxxii. (1903), pp. 4041. [Scientific Papers, v. (1912), pp. 112114.]
On the Acoustic Shadow of a Sphere (Jan. 21, 1904). Phil. Trans, of the Royal Soc.
ccra. A (1904), pp. 87110. [Scientific Papers, v. (1912), pp. 149161.]
780 THEORY OP BESSEL FUNCTIONS
On the Open Organ Pipe Problem in two Dimensions. Phil, Mag. (6) vm. (1904),
pp. 481487. [Scientific Papers, v. (1912), pp. 206211.]
On the Experimental Determination of the Ratio of the Electrical Units. Phil. Mag.
(6) XII. (1906), pp. 97108. [Scientific Papers, v. (1912), pp. 330340.]
On the Light dispersed from Fine Lines ruled upon Reflecting Surfaces. Phil. Mag. (6)
xiv. (1907), pp. 350369. [Scientific Papers, v. (1912), pp. 410418.]
The Problem of the Whispering Gallery. Phil. Mag. (6) xx. (1910), pp. 10011004.
[Scientific Papers, v. (1912), pp. 617620.]
Note on Bessel's Functions as applied to the Vibrations of a Circular Membrane. Phil.
Mag. (6) xxi. (1911), pp. 5358. [Scientific Papers, VI. (1920), pp. 15.]
On a Physical Interpretation of Schlomilch's Theorem in Bessel's Functions. Phil. Mag.
(6) xxi. (1911), pp. 567571. [Scientific Papers, vi. (1920), pp. 2225.]
Problems in the Conduction of Heat. Phil. Mag. (6) xxn. (1911), pp. 381 396.
[Scientific Papers, vi. (1920), pp. 5164.]
On the propagation of Waves through a stratified Medium (Jan. 11, 1912). Proc. Royal
Soc. lxxxvi. A (1912), pp. 207226. [Scientific Papers, vi. (1920), pp. 111120.]
Electrical Vibrations on a thin Anchor Ring (June 27, 1912). Proc. Royal Soc. lxxxvii. A
(1912), pp. 193202. [Scientific Papers, vi. (1920), pp. 111120.]
Further Applications of Bessel's Functions of High Order to the Whispering Gallery
and Allied Problems. Phil. Mag. (6) xxvu. (1914), pp. 100109. [Scientific Papers, vi.
(1920), pp. 211219.]
Further remarks on the Stability of Viscous Fluid Motion. Phil. Mag. (6) xxvin. (1914),
pp. 609619. [Scientific Papers, vi. (1920), pp. 266275.]
On the Stability of the simple Shearing Motion of a Viscous Incompressible Fluid.
Phil. Mag. (6) xxx. (1915), pp. 329338. [Scientific Papers, vi. (1920), pp. 341349.]
On Legendre's Function* Pn (0), when n is great and 6 has any value (April 27, 1916).
Proc. Royal Soc. xcn. A (1916), pp. 433
437. [Scientific Papers, vi. (1920), pp. 393397.]
On Convection Currents in a horizontal Layer of Fluid, when the higher Temperature
is on the under side. Phil. Mag. (6) xxxn. (1916), pp. 529
546. [Scientific Papers,
* vi.
(1920), pp. 432446.]
On the Problem of Random Vibrations and of Random Flights in one, two, or three
dimensions. Phil. Mag. (6) xxxvu. (1919), pp. 321 347. [Scientific Papers, vi. (1920),
pp. 604626.]
REINECK, A.
Die Verwandtschaft zwischen Kugelfunktionen und Besselschen Funktionen. Disserta-
tion, Bern (Halle, 1907). (72 pp.) [Jahrbuch iiber die Fortschritte der Math. 1907, p. 495.]
RICCATI, COUNT J. F.
Animadversationes in aequationes differentiales secundi gradus. Actorum Eruditorum
quae Lipsiae publicantur Supplementa, vm. (1724), pp. 66 73.
Com. Jacobi Riccati Appendix ad Animadversationes in aequationes differentiales
secundi gradus, editus in Actis Eruditorum quae Lipsiae publicantur Tomo vm. Supple-
mentorum, Sectione II. p. 66. Acta Eruditorum publicata Lipsiae, 1723, pp. 502 510.
RIEMANN, G. F. B.
Zur Theorie der Nobili'schen Farbenringe (March 28, 1855). Ann. der Physik und
Chemie, (2) xcy. (1855), pp. 130139.
Partielle Diferentialgleichungen und deren Anwendung auf physikalische Fragen. Von
Bernhard Riemann. FiiY den Druck bearbeitet und herausgegebenen von Karl Hattendorf
(Brunswick, 1876).
ROHRS, J. H.
Spherical and Cylindric Motion in Viscous Fluid (May 14, 1874). Proc. London Math.
Soc. v. (1874), pp. 125139.
RUDSKI, P.
Ueber eine Klasse transcendenter Gleichungen. Prace Mat. Fit.
[Jahrbuch iiber die Fortschritte der Math. 1892, pp. 107 108.]
ill. (1892), pp. 69
81.
Note sur la situation des racines des fonctions transcendantes JH +t(#)=0 (May 10,
1891). Mem. de la Soc. R. des Set. de Liege, (2) xvm. (1895), no. 3.
RUSSELL, A.
The and Inductance of a Concentric Main, and Methods of Com-
Effective Resistance
puting the Ber and Bei and Allied Functions (Jan. 22, 1909). Phil. Mag. (6) xvn. (1909),
pp. 524552.
* This is the function which other writers denote by the symbol P_ (cos 0).
BIBLIOGRAPHY 781
RUTGERS, J. G.
pp. 368373.
Over eene reeks met Besselsche Functies. Nieuw Archie/ voor Wiskunde, (2) VII. (1907),
pp. 8890.
Over reeksen van Besselsche functies en daarmede samenhangende bepaalde integraleu,
waarin Besselsche functies voorkomen. Nieuw Archie/ voor Wiskunde, (2) vn. (1907), pp.
164181.
Sur les functions cylindriques de premiere espece. Nieuw Archie/ voor Wiskunde, (2)
vii. (1907), pp. 385405.
Over eenige toepassingen der Fourier'sche ontwikkeling van een willekeurige functie
naar Bessel'sche functies. Nieuw Archie/ voor Wiskunde, vm. (1909), pp. 375380.
RYBCZYNSKI, W. VON.
die Ausbreitung der Wellen der drahtlosen Telegraphie auf der Erdkugel (March
Uber
5, 1913). Ann. der Physik und Chemie, (4) XLI. (1913), pp. 191208.
SASAKI, S.
On the roots of the equation ^4S " "^ /f"x =a TOoku Math. Journal, v. (1914),
pp. 45 47.
SAVIDGE, H. G.
Tables of the ber and bei and ker and kei Functions with further formulae for their
Computation (Nov. 12, 1909). Phil. Mag. (6) xix. (1910), pp. 4958.
SCHAFHEITLIN, P.*
die Darstellung der hypergeometrische Reihe durch ein bestimmtes Integral
Ueber
(Feb. 1887). Math. Ann. xxx. (1887), pp. 157178.
Ueber eine Integraldarstellung der hypergeometrischen Reihen (Jan. 1888). Math.
Ann. xxxi. (1888), p. 156.
Ueber die Gattss-sche und Bessel-aehe Differentialgleichuug und eine neue Integralform
der letzteren (Feb. 16, 1894). Journal /ur Math. cxiv. (1895), pp. 3144.
Die Nullstellen der Besselschen Functionen. Journal /Ur Math. cxxn. (1900), pp. 299
321.
Uber die Nullstellen der Besselschen Funktionen zweiter Art (Jan. 10, 1901). Archiv
der Math, und Phys. (3) i. (1901), pp. 133137.
Uber den Verlauf der Besselschen Funktionen (May 18, 1904). Berliner Sitzungsberichte,
III. (1904), pp. 8385.
Die Lage der Nullstellen der Besselschen Funktionen zweiter Art (June 27, 1906).
Berliner Sitzungsberichte, v. (1906), pp. 82 93.
Uber den Verlauf der Besselschen Funktionen zweiter Art. Jahresbericht der Deutschen
Math. Vereinigung, xvi. (1907), pp. 272 279.
Die Theorie der Besselschen Funktionen. Von Paul Schafheitlin (Leipzig und Berlin,
1908; Math. Phys. Schri/ten/ilr Ingenieure und Studierende, 4).
Beziehungen zwischen dem Integrallogarithmus und den Besselschen Funktionen (Feb.
24, 1909). Berliner Sitzungsberichte, vm. (1909), pp. 62 67.
Die semikonvergenten Reihen fur die Besselschen Funktionen (Sept. 1, 1909). Jahres-
bericht der Deutschen Math. Vereinigung, xix. (1910), pp. 120 129.
SCHEIBNEll, W.
the asymptotic values of the coefficients in the development of any power of the
On
radius vector according to the mean anomaly (March, 1856). Goidds Astronomical Journal,
IV. (1856), pp. 177182. [Math. Ann. xvn. (1880), pp. 531544.]
Ueber die asymptotische Werthe der Coefficienten in den nach der mittleren Anomalie
vorgenommeuen Entwickelungen (May 31, 1856). Leipziger Berichte, vm. (1856), pp. 40
64. [Math. Ann. xvn. (1880), pp. 545 560.
SCHERK, H. F.
Uber die Integration der Gleichung t-^ ={a+&x)y. Journal /ur Math. x. (1833),
pp. 92 97.
* See also under Sonine.
782 THEORY OF BESSEL FUNCTIONS
SCHLAFLI, L.
Sulle relazioni tra diversi integrali definiti che giovano ad esprimere la soluzione generale
della equazione di Riccati. Ann. di Mat. (2) I. (1868), pp. 232
242.
Einige Bemerkungen zu Herrn Neumann's Untersuchungen iiber die Bessel'schen
Functionen (May 4, 1870). Math, Ann. in. (1871), pp. 134149. [With note by C. N.
p. 149.]
Sopra un teorema di Jacobi recato a forma piu generale ed applicata alia funzione
cilindrica (Aug. 1871). Ann. di Mat. (2) v. (1873), pp. 199205.
SulP uso delle linee lungo le quali il valore assoluto di una funzione e constante (Oct. 4,
1872). Ann.di Mat. (2) VI. (1875), pp. 120.
Ueber die Convergenz der Entwicklung einer arbitraren Function f(x) nach den
Bessel'schen Functionen Ja (^x), J* (fax), Ja (j8j.tr), ..., wo ft, 32, fts, ... die positiven
Wurzeln der Gleichung Ja (ff) vorstellen (Jan. 17, 1876). Math. Ann. x. (1876), pp. 137142.
SCHLOMILCH, 0. X.
Note sur la variation des constantes arbitraires d'une integrate dennie. Journal fur
Math. xxxm. (1846), pp. 268280.
Analytische Studien, n. (Leipzig, 1848).
Ueber die Bessel'schen Function. ZeiUchrift fiir Math, und Phyt. n. (1857), pp. 137
165.
SCHONHOLZER, J. J.
Ueber die Auswerthung bestimmte Integrate mit Hiilfe von Veranderungen des Integra-
tionsweges (Dissertation, Bern, 1877). [Graf and Gubler, Einleitung in die Theorie der
Bessel'schen Funktionen, n. (Bern, 1900).]
SCHOTT, G. A
Electromagnetic Radiation (Cambridge, 1912).
SCHWARZSCHILD, K.
Die Beugung und Polarisation des Lichts durch einen Spalt. Math. Ann. lv. (1902),
pp. 177247.
SCHWERD, F. M.
Die Beugungserscheinungen aus den Fundamentalgesetzen der Undulationstheorie (Mann-
heim, 1835).
SEARLE, J. H. C.
On the propagation of waves in an atmosphere of varying density. Quarterly Journal,
xxxix. (1908), pp. 5166.
SEGAR, H. W.
On the roots of certain continuants. Messenger, xxn. (1893), pp. 171 181.
SERRET, J. A.
Note sur quelques formulas de calcul integral. Journal de Math. vm. (1843), pp. 1 27.
Memoire sur ^integration d'une equation differentielle a l'aide des differentielles a indices
quelconques (Sept. 4, 1843). Comptes Bendus, xvn. (1843), pp. 458475; Journal de
Math. ix. (1844), pp. 193216.
SHARPE, H. J.
On the Reflection of Sound at the Surface of a Paraboloid (Nov. 1874). Quarterly
Journal, xv. (1877), pp. 1 8.
On a diflerential equation. Messenger, x. (1881), pp. 174 185 XI. (1882), pp. 41 44.
;
On a transcendental differential equation. Messenger, xi. (1882), pp. 56 63.
On a differential equation. Messenger, xill. (1884), pp. 6679.
Note on Legendre's coefficients. Quarterly Journal, xxiv. (1890), pp. 383 386.
On The Reflection of Sound at a Paraboloid (June 20, 1899). Proc. Camb. Phil. Soc. x.
(1900), pp. 101136.
SHEPPARD, W. F.
On some expressions of a function of a single variable in terms of Bessel's functions.
Quarterly Journal, xxin. (1889), pp. 223 260.
SIACCI, F.
Sulla integrazione di una equazione differenziale e sulla equazione di Riccati (April 13,
1901). Napoli Rendiconto, (3) vn. (1901), pp. 139143.
Bibliography 783
sibirani, f.
Sopra equazione di Riccati. Riv. fis. mat. xix. (1909), pp. 216220. [Jahrbuch iiber die
1*
SIEMON, P.
Ueber die Integrale einer nicht homogenen Differentialgleichung zweiter Ordnung. Pro-
gramme Luisenschule, Berlin, 1890, 22 pp. [Jahrbuch fiber die Fortschritte der Math. 1890,
pp. 340342.]
SMITH, B. A *
Table of Bessel's Functions F and Ft Messenger, xxvi. (1897), pp. 98101.
.
Arched Dams. Proc. American Soc. of Civil Engineers, xlvi. (1920), pp. 375 425.
SMITH, CLARA E.
Atheorem of Abel and its application to the development of a function in terms of
Bessel's functions (June 8, 1906). Trans. American Math. Soc. vni. (1907), pp. 92106.
SMITH, O. A.
Quelques relations integrates entre les fonctions spheriques et cylindriques. Giornale di
Mat. (2) xil. (1905), pp. 365373.
SOMMERFELD, A. J. W.|
Die willkiirlichen Functionen in der mathematischen Physik. Dissertation, Konigsberg,
1891 (75 pp.). [Jahrbuch fiber die Fortschritte der Math. 1891, pp. 519523.]
Zur analytischen Theorie der Warmeleitung. Math. Ann. xvv. (1894), pp. 263277.
Mathematische Theorie der Diffraction (Summer, 1895): Math. Ann. xlvii. (1896), pp.
317374.
Uber die Ausbreitung der Wellen der drahtlosen Telegraphic (Jan. 15, 1909). Ann. der
Physik und Chemie, (4) xxviii. (1909), pp. 665736.
Die Greensche Funktion der Schwingungsgleichung. Jahresbericht der Devtscher Math.
Vereinigung, xxi. (1912), pp. 309 353.
SONINE, N. J.
Onthe resolution of functions into infinite series (Jan. 17/29, 1870). Mathematical
Collection published by the Moscow Math. Soc. [MaTenaTHiecKift C6opHHKt], y. (1870),
pp. 271302, 323382.
Recherches sur les fonctions cylindriques et le developpement des fonctions continues en .
SPITZER, S.
Integral der Differentialgleichung xy"~y=0. Zeitschrift fiir Math. undPhys. n. (1857),
pp. 165170.
Entwickelung von e***^x in unendliche Reihen. Zeitschrift fflr Math, und Phys. in.
(1858), pp. 244246.
Darstellung des unendlichen Kettenbruches
1
2a? + l, +
2x+3 +
2#+5-
2* + 7 + ...
in geschlossener Form. Archiv der Math, und Phys. xxx. (1858), pp. 331 334.
Ueber die Integration der Differentialgleichung
d*y
'
dxn
(1859). Journal fUr Math. lvii. (1860), pp. 82 87.
STEARN, H.
On some cases of the Varying Motion of a Viscous Fluid. Quarterly Journal, xvn.
(1880), pp. 90104.
* See also under Michell. t See also under Hopf.
784 THEORY OF BESSEL FUNCTIONS
STEINER, L.
Intensitats-verhaltnisse der Beugungserscheinung durch eine kreisformige Offnung
(Jui)e 19, 1893). Math, und Naturwiss. Berichte aus Ungarn, XL (1892 93) [1894], pp
362373.
STEINTHAL, A. E.
On the Solution of the Equation
. cPu _ du , ,
'
' dar ax
Quarterly Journal, xvm. (1882), pp. 330 345.
STEPHENSON, A.
An extension of the Fourier method of expansion in sine series. Messenger, xxxiii. (1904),
pp. 7077.
A more general case of expansion in sine series. Messenger, xxxin. (1904), pp. 178
182.
On Expansion in Bessel's Functions (June, 1907). Phil. Mag. (6) xiv. (1907), pp. 547
549.
STERN, M. A.
Ueber die Anwendung der Sturm'schen Methode auf transcendente Gleichungen.
Journal fiir Math, xxxiii. (1846), pp. 363 365.
STIELTJES, T. J.
Recherches sur quelques series s^mi-convergentes. Ann. sci. de VEcole nonn. sup. (3)
in. (1886), pp. 201258.
STOKES, SIR GEORGE GABRIEL.
Onthe numerical Calculation of a Class of Definite Integrals and Infinite Series
(March 11, 1850). Trans. Camb. Phil. Soc IX. (1856), pp. 166187. [Math, and Phys.
Papers, II. (1883), pp. 329357.]
On the Effect of the Internal Friction of Fluids on the Motion of Pendulums (Dec. 9,
1850). Trans. Camb. Phil. Soc. IX. (1856), pp. [8] [106]. [Math, and Phys. Papers, in.
(1901), pp. 1141.]
On the Discontinuity of Arbitrary Constants which appear in Divergent Developments
(May 11, 1857). Trans. Camb. Phil. Soc. x. (1864), pp. 106128. [Math, and Phys. Papers,
77109.]
iv. (1904), pp.
Supplement to a paper on the Discontinuity of Arbitrary Constants which appear in
Divergent Developments (May 25, 1868). Trans. Camb. Phil. Soc. xi. (1871), pp. 412 425.
[Math, and Phys.
Papers, iv. (1904), pp. 283 298.]
On the Communication of Vibration from a Vibrating Body to a surrounding Gas*
(June 18, 1868). Phil. Trans, of the Royal Soc. clviii. (1868) [1869], pp. 447 463. [Math,
and Phys. Papers, iv. (1904), pp. 299 324.]
Smith's Prize Examination Papers, Fet>. 1853 and Jan. 29, 1867. [Math, and Phys.
Papers, V. (1905), pp. 319, 347.]
Note on the Determination of Arbitrary Constants which appear as Multipliers of
Senii-convergent Series (June 3, 1889). Proc. Camb. Phil. Soc. vi. (1889), pp. 3652 366.
[Math, and Phys. Papers, v. (1905), pp. 221225.]
On the Discontinuity of Arbitrary Constants that appear as Multipliers of Semi-con-
vergent Series t (April 23, 1902). Acta Math. xxvi. (1902), pp. 393397. [Math, and Phys.
Papers, v. (1905), pp. 283287.]
STRUTT, J. W.
See Rayleigh (Lord).
STRUVE, H.
Ueber den Einfluss der Diffraction an Fernrbhren auf Lichtscheiben (May 25, 1882).
Mem. de I'Acad. Imp. des Sci. de St. Petersbourg, (7) xxx. (1882), no. 8.
Beitrag zur Theorie der Diffraction an Fernrohren (Aug. 1882). Ann. der Physilc und
Chemie, (3) xvn. (1882), pp. 1008 1016.
STURM, J. C. F
Sur les Equations differentielles lineaires du second ordre (Sept. 28, 1833). Journal de
Math. l. (1836), pp. 106186.
* An abstract will be found in Proc. Royal Soc. xvi. (1868), pp. 470471.
f This, the last published paper which Stokes wrote, contains a summary of his researches
composed for the Abel centenary volume.
BIBLIOGRAPHY 785
SUCHAR, P. J.
Sur equations differentielles lineaires reciproques du seco. d ordre (Nov. 18, 1903).
les
Bull, de la Soc. Math, de France, xxxn. (1904), pp. 103116.
SVANBERG, A. F.
De integralibus definitis disquisitiones. Nova Acta R. Soc. Sci. Upsala, x. (1832),
pp. 231288.
TAKEUCHI, T.
Gn the integral e*
cwp0 (imq6, cos q6d6 (May 17, 1920). TShoku Math. Journal,
J
xvin. (1920), pp. 295296.
THEISINGER, L.
Bestimmtc Integrate. Monatshefte fiir Math, undPhys. xxiv. (1913), pp. 328346.
THOMAE, J.
Die Brauchbarkeit der Bessel-Fourierechen Reihe. Berichte der Math. Seminar zu Jena,
191213, pp. 810. [Int. Catalogue of Set. Lit. xin. (1913), p. 100.]
/go
THOMSON, SIR JOSEPH JOHN.
Note on / ""f* llt dx. Quarterly Journal, xvin. (1882), pp. 377381.
THOMSON, W.
See Kelvin (Lord).
TODHUNTER, I.
VALEWINK, G. C. A.
Over asymptotische ontwikkelingen {Dissertation, Haarlem, pp. 138). [Jahrbuch iiber
die Fortschritte der Math. 1905, p. 328.]
VAN VLECK, E. B.
On the Roots' of Bessel- and P-functions. American Journal of Math. xix. (1897),
pp. 75 85. ,
VERDET, E.
Lecons dOptique Physique, I. (Paris, 1869).
VESSIOT, E.
Sur quelques equations differentieltes ordinaires du second ordre. Annates de la Fac.
des Sci. de Toulouse, ix. (1895), no. 6.
VOLTERRA, V.
Sopra alcune questioni di inversione di integrali definiti. Ann. di Mat. (2) xxv. (1897),
pp. 139178.
VORONOI, G.
Sur une fonction transcendante et ses applications a la sommation de quelques senes.
Ann. sci. de VtkoU norm. sup. (3) xxi. (1904), pp. 207268, 459534.
Sur le developpemeut, a l'aide des fonctions cylindriques, des sorames doubles
If (pm +2qmn + rn )
2 2
oh pmt + Zqmn+rn* est une forme positive a coefficients entiers. Verh. des dritten Int.
Kongresses in Heidelberg (1904), pp. 241245.
786 THEORY OF BESSEL FUNCTIONS
WAGNER, C.
Beitrage zur Entwicklung der Bessel'schen Functioni. Art. Bern Mittheilungen, 1894,
pp. 204266.
Uber die Darstellung einiger bestimmten Integrate durch Bessel'sche Funktionen
(June 12, 1895; Aug. 5, 1895). Bern Mittheilungen, 1895, pp. 115 119; 1896, pp. 53 60.
WALKER,
G. T.
Some formulae for transforming the origin of reference of Bessel's functions. Messenger,
xxv. (1896), pp. 76 80.
WALKER, G. W.
The scattering of electromagnetic waves by a sphere. Quarterly Jburnal, xxxi. (1900),
pp. 36
49.
WALKER, J.
The Analytical Theory of Light (Cambridge, 1904).
WALLENBURG, G.
Ueber Riccatfsche Differentialgleichungen hoherer Ordnung. Journal fiir Math. cxxi.
(1900), pp. 196199.
Die Differentialgleichungen deren allgemeines Integral eine lineare gobrochene Function
der willkiirlicheu Constanten ist. Journal fiir Math. cxxi. (1900), pp. 210 217.
Sur l'equation diff&rentielle de Riccati du second ordre (Dec. 14, 1903). Comptes Rendus,
cxxxvii. (1903), pp 1033 1035.
WANGERIN, A.
Cylinderfunktionen oder Bessel'sche Funktionen. Encyclopddie der Math. Wiss. Bd. n.
Teil 1 (Leipzig, 1904
16), pp. 742
757. [Encyclopedic des Sci. Math. Tome II. vol. 5
(Paris et Leipzig, 1914, pp. 209 229), translation by A. Lambert.]
WATSON, G. N.
On a certain difference equation of the second order. Quarterly Jotirnal, xli. (1910),
pp. 5055.
Bessel Functions and Kapteyn Series (April 26, 1916). Proc. London Math. Soc. (2)
xvi. (1917), pp. 150174.
Bessel Functions of Equal Order and Argument (June 17, 1916). Phil. Mag. (6) xxxn.
(1916), pp. 232237.
Simple types of Kapteyn Series. Messenger, xlvi. (1917), pp. 150157.
Bessel functions of equal order and argument (Nov. 13, 1916). Proc. Camb. Phil. Soc.
xix. (1918), pp. 4248.
The limits of applicability of the Principle of Stationary Phase (Nov. 22, 1916). Proc.
Camb. Phil. Soc. xix. (1918), pp. 4955.
Bessel functions of large order (June 14, 1917). Proc. Camb. Phil. Soc. xix. (1918),
pp. 96 110.
The Zeros of Bessel Functions (Aug. 17, 1917). Proc. Royal Soc. xciv. A (1918), pp.
190206.
Bessel Functions of Equal Order and Argument. Phil. Mag. (6) xxxv. (1918), pp. 364
370.
The Diffraction of Electric Waves by the Earth (May 29, 1918). Proc. Royal Soc. xcv.
A (1919), pp. 8399.
The Transmission of Electric Waves round the Earth (Jan. 13, 1919). Proc. Royal Soc.
xcv. A (1919),
pp. 546563.
On Nielsen's functional equations. Messenger, xlviii. (1919), pp. 49 53.
The zeros of Lommel's polynomials (May 15, 1919). Proc. London Math. Soc. (2) xix.
(1921), pp. 266272.
WEBB, H. A.
The expansion of an arbitrary function in a series of Bessel functions. Messenger, xxxm.
(1904), pp. 55-58.
On the Convergence of Infinite Series of Analytic Functions (Nov. 10, 1904). Phil.
Tram, of the Royal Soc. cciv. A
(1904), pp. 481- 497.
WEBER, H.
Uelier einige bestimmte Integrale (Jan. 1868). Journal fiir Math. lxix. (1869), pp.
222237.
Ueber die Integration der partiellen Difierentialgleichung : ~2 + x^ +1^11 = (July,
Ueber die Besselxvhcu Functionen und ihre Anwendung auf die Theorie der elektrischen
Strbme (May, 1872). Journal fUr Math. lxxv. (1873), pp. 75105.
Ueber die stationaren Stromungen der Electricitat in Cylindern (Sept. 1872). Journal
fiir Math, lxxvi. (1873), pp. .120.
Ueber erne Darstellung willkiirlicher Functionen durch Bessel'sche Functionen (Oct.
1872). Math. Ann. vi. (1873), pp. 146161.
Zur Theorie der Bessersohen Functionen (July, 1890). Math. Ann. xxxvn. (1890),
pp. 404416.
WEBER, H. F.
Die wahrc Theorie der Fresnel'aehen Interferenz-Erscheinuugen. Zurich Vierteljahrs-
schrift, xxiv. (1879), pp. 33 76; Ann. der Physik und Chemio, (3j vm. (1879), pp. 407 444.
WEBSTER, A. G.
Application of a definite integral involving Bessel's functions to the self-inductance of
solenoids (Dec. 29, 1905). Bulletin American Math. Soc. xiv. (1907), pp. 16.
WENDT, CACILIE.
Eine Vcrallgemeinerung des Additionstheoremes der Bessel'schen Functionen erster
Art. Monatshe/te fiir Math, und Phys. xi. (1900), pp. 125131.
WEYL, H.
Singulare Integralgleichungen (April, 1908). Math. Ann. lxvi. (1909), pp. 273324.
WEYR, E.
Zur Integration der Differential-Gleichungen erster Ordnung. Abh. bbhm. Oes. Wiss.
(Prag), (6) vm. (187576), Math. Mem. 1.
WHEWELL, W.
Of the Intrinsic Equation of
a Curve and its Application (Feb. 12, 1849). Trans. Camb.
Phil. Soc. vijr. (1849), pp. 659671.
Second Memoir on the Intrinsic Equation of a Curve and its Application (April 15,
1850). Trans. Camb. Phil. Soc. IX. (1856), pp. 150156.
.WHIPPLE, F. J. W.
Diffraction by a wedge and kindred problems (Nov. 8, 1915). Proc. London Math. Soc.
(2) xvi. (1917), pp. 94111.
WHITE, F. P.
The
Diffraction of Plane Electromagnetic Waves by a Perfectly Reflecting Sphere (June
9, 1921). Proc. Royal Soc. C. A
(1922), pp. 505525.
WHITEHEAD, C. S.
On the functions ber x, bei x, ker x, kei x. Quarterly Journal, xlii. (1911), pp. 316 342.
WHITTAKER, E. T.
On
the General Solution of Laplace's Equation and the Equation of Wave Motions and
on an undulatory explanation of Gravity. Monthly Notices of the R. A. S. lxii. (1902),
pp. 617620.
On the partial differential equations of mathematical physics. Math. Ann. lvii. (1903),
pp. 333 355.
On a new Connexion of Bessel Functions with Legendre Functions (Nov. 13, 1902).
Proc. London Math. Soc. xxxv. (1903), pp. 198206.
WIGERT, S.
Sur quelques fonctious arithm&iques (March, 1913). Acta Math, xxxvn. (1914), pp.
113140.
WILLIAMSON, B.
On the Solution of certain Differential Equations (March 5, 1856). Phil. Mag. (4) xi
(1856), pp. 364371.
WILLSON, R. W. AND
PEIRCE, B. O.
Table of the forty roots of the Bessel equation
first J
(x)=*0 with the corresponding
values of J\{x). Bulletin American Math. Soc. in. (1897), pp. 153 155.
WILTON, J. R.
A
continued fraction solution of the linear differential equation of the second order.
Quarterly Journal, xlvi. (1915), pp. 318334,
788 THEORY OF BESSEL FUNCTIONS
WIRTINGER, W.
Zwei Bemerkungen zu Airy's Theorie des Regenbogens. Berichte des natur.-med.
Vereins in Innsbruck, xxni. (1897), pp. 7 15.
WORMS DE ROMILLY, P.
Note sur l'integration de l'e'quation
*J
ax*
+ e!
x ax
+ r-a
Journal de Math. (3) iv. (1878), pp. 177186.
YOUNG, W. H.
On infinite integrals involving a generalisation of the sine and cosine functions (Oct. 6,
1911). Quarterly Journal, xliii. (1911), pp. 161 177.
On series of Bessel functions (Dec. 6, 1917). Proc. London Math. Soc. (2) xviii. (1920)
pp. 163200.
ZELINSKIJ, I. I.
On
the integration of Riccati's equation (Jan. 27, 1890). Proc. Phys. Math. Section,
Naturalist^ Soc. Imp. Univ. Kazan, vm. (1890), pp. 337 342.
INDEX OF SYMBOLS
[The numbers refer to the pages on which the symbols are defined.]
Dn 598
, h\ 13 JV, 661
(*), 74 h(z)> 77
0,197
#,6 J",(5), 10, 16 0(0, 271, 272
E n (z), 56 Jn (z),Q, 14, 19 0- n (0,276
Ein (a), 320 ./-(*), 16 < (0, 569
E(*), 308 ^,, 10
e, er , 111 J8 (*),30 P, 547
./,(*), 38 P 227
P(0), 258 */"(/!., a?), 49 P(#),156
F(a, 0, 90 J^ip). 129 P m-*(*), 295
P(0, a), 253 Jn,k(z )> 326 P-^(#),-51
P(<r), 77 /(*; it), 327
i/, P(cos<p), 129
790 THEORY OF BESSEL FUNCTIONS
Pn (r; a lt a.,.. , o), 420 2U(jr),343 *, 22
P(r;a 1 ,a.2 ,...,t 'n|p),421 7/n (*,a),320 k<*),56
P(x,v),205 ?;,(*,*), 582 B, 112
(a, b, c, rw (,;; #), 602 @, 27; 291
"J
^(4 46 F<">(*), 70 v, 51 83 ;
^, 33 V39 n, 1,103
Tn (s),340 e,6
LIST OF AUTHOES QUOTED
[The numbers refer to the pages. References are not given to entries in the bibliography,
pp. 753- -788.]
664 189, 191, 202, 203, 214, 230, 231, 234, 279,
Alexander, P.. 579 302, 349, 360, 385, 391, 393, 399, 574, 575,
Anding, E., 313, 657 601
Anger, C. T., 21, 22, 308, 309, 310, 311, 312 Bruns, H., 327
Anisimov* V. A., 92 Bryan, G. H., 127, 480
Appell, P. E., 146, 371 Burkhardt, H. F. K. I/., 236, 280
Autonne, L., 94 Burnside, W. S., 305
Bach, D., 96, 102, 110 Cailler, C, 44, 149, 386, 395, 415, 437, 455,
Bachmann, P., 197 536, 537
Baehr, G. F. W., 479 Callandreau, O., 196, 208, 387
Ball, L. de, 157 Cantor, G. F. L. P., 637
Barnes, E. W., 83, 100, 102, 104, 105, 156, Carlini, F., 6, 7, 194, 225, 226, 227, 249, 255,
190, 192, 196, 196, 220, 221, 340, 351, 357, 268, 572
367, 383, 387, 402, 409 Carslaw, H. S., 177, 366, 395, 499, 500, 507,
Basset, A. B., 76, 77, 78, 80, 172, 173, 180, 509, 583
385, 388, 395, 425, 426, 454 Catalan, E. C, 21, 22, 27, 96, 173, 188
Bateman, H., 130, 131, 367, 370, 372, 373, Cauchy, (Baron) A L., 7, 15, 16, 21, 150,
376, 379, 380, 389, 406, 417, 437, 456, 530, 183, 230, 231, 232, 233, 247, 249, 259, 309,
533, 535 319, 324, 449, 545, 554, 557, 579
Bauer, G., 50, 128, 368, 370 Cayley, A., 88, 90, 96, 102, 103, 109, 188,
Beltrami, E., 51, 358, 361, 374, 386, 389, 390, 502
391, 500, 579, 621 Challis, H. W., 91
Bernoulli, Daniel (1700-1782), 2, 3, 4, 9, 85, Chapman, S., 621
86, 87, 88, 111, 123, 478, 576 Chessin, A. S., 135, 175, 325, 346, 382
Bernoulli, James (1654-1705), 1, 2, 3, 88, 90, Chree, C, 6, 597
92 Christoffel, E. B., 154
Bernoulli, John (1667-1748), 1, 2, 3 Chrystal, G., 102, 288, 295
Bernoulli, Nicholas, (1687-1759), 2 Cinelli, M., 633
Bernoulli, Nicholas, (1695-1726), 2 Clebsch, R. F. A, 359, 363
Bessel, F. W., 1, 3, 4, 5, 9, 13, 14, 15, 18, 19, Clifford, W. K., 90, 91
21, 24, 25, 38, 84, 140, 148, 153, 160, 295, Coates, C. V., 173, 180, 313, 622
308, 478, 551, 554, 654 Cotter, J. R., 41
Binet, J. P. M., 183 Crawford, L., 27
Bocher, M., 46, 57, 64, 376, 494, 495, 517 Crelier, L., 286, 287, 288, 295, 300, 301, 302
Bohmer, P. E., 142 Curtis, A. H., 96, 110
Boole, G., 27, 47, 110,627 Curzon, H. E. J., 395
792 THEORY OF BESSEL FUNCTIONS
D'Alembert, J. le Rond, 3 Gegenbauer, L. von, 50, 51, 129, 138, 151,
Dandelin, G. P., 503 274, 283, 284, 290, 293, 351, 362, 363, 365,
Darboux, J. G, 233 366, 367, 368, 369, 370, 373, 378, 379, 383,
Darwin, C. G., 437 384, 385, 386, 389, 390, 391, 393, 395, 396,
Debye, P., 225, 235, 237, 240, 241, 247, 249, 398, 406, 407, 413, 414, 415, 418, 426,430,
250, 251, 255, 262, 263, 268, 513, 516 438, 439, 480, 508, 517, 522, 524, 525, 579
De la Vallee Poussin, Ch. J., 53, 160, 189 Genocchi, A, 119
De Morgan, A., 188, 190 Gibson, G. A., 197
Dendy, A., 107 Gifford, E., 662
Dini, U., 10, 577, 578, 597, 600, 616, 651 Gilbert, L. P., 545, 548, 549
Dinnik, A., 579, 659, 660 Giuliani, G., 155, 156, 324, 326, 327
Dirichlet, P. G. Lejeune, 157, 230, 406, 581, Glaisher, J. W. L., 89, 96, 102, 103, 108, 109,
B23 140, 171, 173, 183, 664
Dixon, A. C, 35, 480, 481, 482 Gordan, P. A., 55
Donkin, W. F., 109 Goursat, Edouard J. B., 120
Dougall, J., 65, 411 Graeffe, C. H., 502, 503
Du Bois Reymond, P. D. G., 183, 455, 470 Graf, J. H., 32, 64, 75, 145, 153, 160, 165,
Duhamel, J. M. C, 38, 49, 59, 68, 227 175, 197, 215, 227, 286, 287, 290, 295, 296,
299, 301, 302, 303, 341, 344, 345, 359, 360,
Eamshaw, S., 108 362, 398, 498, 502, 583
Ellis, R. L., 95, 109, 110, 173 Gray, A., 64, 65, 78, 194, 206, 454, 480, 655
Emde, 248
F., 656, 657, 658, 660
Encke, J. F., 342 Green, G, 124
Enestrom, G, 92 Greenhill, Sir A. George, 91, 96
Enneper, A., 173 Gregory, Duncan Farquharson, 391
Epstein, S. S., 145, 290 Gregory, Walter, 224
Ermakoff, W., 455 Grunert, J. A., 27
Escherich, G. von, 165 Gubler, E., 32, 64, 145, 160, 165, 177, 197,
Euler, L., 3, 4, 5, 6, 24, 49, 53, 60, 62, 87, 88, 215, 227, 286, 287, 301, 329, 341, 351, 398,
92, 93, 123, 133, 183, 410, 498, 500, 501, 408, 410, 426, 498, 502, 583
503, 576, 659 Gunther, S., 153
Gwyther, R. F., 621, 636
Falkenhagen, J. H. M., 94
610
Fejer, L., Hadamard, J., 204, 205, 527
Feldblum, M., 92 Haentzschel, E., 71, 96, 159
Fields, J. G, 110 Hafen, M., 389
Filon, L. N. G, 51, 578, 622, 623, 625, 629 Hague, B., 656
Ford, W. B., 578, 605 Hall, A, 15
Forsyth, A. R., 42, 57, 107, 109, 117, 346, Hamilton, Sir William Rowan, 12, 195, 655
358, 400, 499 Hankel, H, 10, 38, 57, 58, 61, 62, 63, 65, 73,
Fourier, (Baron) J. B. Joseph, 4, 9, 10, 13, 75, 76, 77, 160, 163, 164, 165, 167, 175,
22, 84, 135, 449, 450, 454, 455, 456, 478, 195, 196, 203, 206, 208, 211,' 384, 386, 390,
482, 483, 501, 576, 577, 578, 616 393, 395, 424, 427, 428, 429, 430, 434, 453,
Freeman, A., 501 454, 456, 457, 458, 459, 462, 464, 465, 471,
Frenet, F., 27 488, 513, 514, 577, 579, 581, 582, 633
Fresnel, A. J., 544, 545 Hansen, P. A., 14, 20, 30, 31, 37, 152, 154,
Frobenius, F. G, 57 155, 158, 195, 292, 406, 655, 656
Frullani, G, 14, 19 Hanumanta Rao, C. V., 437
Hardy, G. H., 8, 111, 180, 183, 188, 189, 200,
Gallop, E. G, 405, 421, 422 309, 320, 321, 322, 324, 373, 382, 386, 395,
Gaskin, T., 109 406, 421, 422, 437, 441, 442, 463, 464, 542,
Gasser, A., 509, 517 546, 547, 573, 575, 579, 606, 615, 621
Gauss, C F. ( Johann Friedrich Carl), .191, 506 Hargreave, C. J., 88, 170, 171
LIST OF AUTHORS QUOTED 793
393, 395, 410, 426 Laplace, P. S. de, 6, 7, 8, 9, 53, 280, 395, 421,
Heine, H. E., 4, 56, 65, 66, 84, 154, 155, 156, 450
157, 181, 358, 363, 365 Largeteau, C. L., 501
Hermite, C, 55, 477 Laurent, Paul Mathieu Hermann, 157
Hertz, H., 80, 81 Laurent, Pierre Alphonse, 100
Herz, N., 554, 555 Lebedeff, Wera Myller-, 99
94
Hill, C. J. D., Lebesgue, Henri, 457
Hobson, E. W., 10, 33, 5458, 125, 128, 129, Lebesgue, Victor AmeMee, 110, 123
149, 172, 174, 280, 353, 363, 369, 385, 386, Lefort, F., 6
387, 480, 485, 578, 586, 591, 602 Legendre, A. M., 52, 90, 183, 204, 485, 557
Hopf, L., 178, 406 l'Hospital, G. F. A. (Marquis de St Mesme),
Horn, J., 225, 526 134
Hurwitz, A., 9, 297, 302, 303, 304, 305, 306, Leibniz, G. W., 1, 2, 3
307, 483, 484 Le Paige, C, 96
Lerch, M., 382, 393, 433, 434, 617
Ignatowsky, W. von, 365 Lindner, P., 484
Isherwood, J. G., 657, 658, 664 Lindstedt, A., 545, 661
Liouville, J., 27, 28, 87, 111, 112, 116, 117,
Kepler, J., 551, 552 225, 229, 233, 365, 377, 385, 386, 389, 395,
Kirchhoff, G., 107, 196, 203, 389, 578, 616 396, 412, 413, 439, 440, 482, 509, 511
Kluyver, J. C, 367, 419, 420 Maclean, M., 658
Kneser, J. C. C. A., 499, 578, 583 McMahon, J., 64, 195, 200, 505, 507, 581
Schlomilch, O. X., 14, 18, 33, 34, 35, 36, 153, Thomson, Sir Joseph John, 65, 173
173,183,617,618,619,621,622,628,655,666 Thomson, Sir William (see Kelvin)
Schonholzer, J. J., 145 Tisserand, F., 371
Schott, G. A., 551, 556, 572, 573 Todhunter, I., 27, 157, 199
Schwarz, K. H. A, 358, 643 Turriere, E., 15
Schwarzschild, K., 361
Schwerd, F. M., 477, 654 Unferdinger, F., 310
Searle, J. H. C, 199
Segar, H. W., 483 Valewink, G. C. A., 196
Serret, J. A, 171, 173, 188 Vandermonde, A., 102
Sharpe, H. J., 105, 157 van Vleck, E. B., 480
Sheppard, W. F., 199, 454, 579, 595, 615 Verdet, E., 477
Siacci, F., 92 Vessiot, E., 94
Siegel, C. L., 485 Volterra, V., 579, 621
Siemon, P., 328, 398 Voronoi, G., 200
Smith, Bernard A., 655 Voss, A., 406
Smith, Clara E., 621
Smith, Otto Andreas, 50 Wagner, C, 13, 142
Sommerfeld, A. J. W., 56, 57, 178, 361, 389, Walker, Gilbert Thomas, 360, 361
395, 406, 417, 464, 499 Walker, James, 328, 331, 333, 537, 544
Sonine, N. J., 82, 83, 132, 137, 139, 143, 169, Wallenburg, G., 94
170, 171, 175, 176, 177, 180, 279, 280, 281, Waring, E, 503
290, 353, 354, 362, 363, 367, 373, 374, 375, Watson, G. N., 11, 105, 125, 158, 226, 231,
376, 377, 378, 383, 386, 391, 394, 395, 398, 249, 268, 355, 444, 483, 485, 513, 519, 566,
401, 411, 415, 417, 418, 431, 432, 433, 434, 575
439, 454 Webb, H. A, 351, 523, 533, 536
Spitzer, S., 68, 71, 153 Weber, Heinrich, 63, 64, 67, 75, 165, 167, 195,
Stearn, H., 482, 621 196, 210, 211, 212, 386, 391, 392, 393, 394,
Steiner, L., 655, 661 395, 396, 398, 402, 405, 406, 408, 421, 450,
Steinthal, A. E., 171, 387 451, 452, 453, 454, 455, 468, 469, 470, 495
Stephenson, A., 579 Weber, Heinrich Friedrich, 308, 309, 310,311,
Stern, M. A., 500 312, 315, 320
Stieltjes, T. J., 195, 196, 207, 208, 209, 213, Weierstrass, C. T. W., 358
214, 464 Wendt, Cacilie, 363
Stirling, James, 7, 8, 214 Weyl, H., 189, 454
Stokes, Sir George Gabriel, 8, 12, 53, 55, 68, Weyr, E, 93, 94
69, 70, 80, 95, 97, 188, 189, 195, 201, 202, Whewell, W., 479
225, 229, 238, 320, 324, 336, 391, 405, 503, Whipple, F. J. W., 177, 313, 387
505, 507, 605, 659 Whitehead, C. S., 81, 82, 132, 148, 203
Strutt, J. W. (see Rayleigh) Whittaker, E. T., 44, 50, 124, 125, 173, 197,
Struve, H., 328, 329, 333, 337, 392, 396, 397, 339, 503, 632, 633
417, 661 Wigert, C. S., 200
Sturm, J. C. F., 304, 477, 479, 517, 518, 521 Williamson, B., 110
Suchar, P. J., 90 Willson, R. W., 501, 660, 664
Svanberg, A. F., 173 Wilton, J. R., 154
Wirtinger, W., 189, 190
Takeuchi, T., 313
Tannery, J., 11, 156, 302 Young, W. H., 10, 351, 406, 578, 579, 580,
Theisinger, L., 184, 185, 338 582, 586, 596, 616, 617
GENERAL INDEX
[The numbers refer to the pages.]
Addition theorems, 358-372 (Chapter xi) for Bessel coefficients of order zero, 128, 359; for Bessel
;
coefficients of order n, 29; for Bessel functions of the first kind (Gegenbauer's type), 362, 367;
for Bessel functions of the first kind (Graf's type), 130, 143, 359 ; for Bessel functions or cylinder
functions of any kind (Gegenbauer's type), 363 for Bessel functions or cylinder functions of any
;
kind (Graf's type), 143, 361 for hemi-cylindrical functions, 354; for Lommel's functions of two
;
variables, 543; for Schlafli's function fn (z), 344; for Schlafli's polynomial, 289; integrals de-
rived from, 367 ; physical significance of, 128, 130, 361, 363, 366; special and degenerate forms
of, 366, 368
Airy's integral, 188 expressed in terms of Bessel functions of order one-third, 192 generalised by
;
;
Hardy, 320; Hardy's expressions for the generalised integral in terms of the functions of Bessel,
Anger and Weber, 321 references to tables of, 659
;
Analytic theory of numbers associated with asymptotic expansions of Bessel functions, 200
Anger's function 3 v {z), 308; connexion with Weber's function, 310; differential equation satisfied
by, 312; integrals expressed in terms of, 312; recurrence formulae for, 311; representation of
Airy's integral (generalised) by, 321 with large argument, asymptotic expansion of, 313 with
; ;
of large order (Carlini), 6, 7 (extensions due to Meissel), 226, 227, 232, 247, 521 (in transitional
; ;
regions), 248; to functions of large numbers (Darboux), 233; (Laplace), 421 to Legendre func- ;
tions of large degree, 65, 155, 157, 158 to remainders in asymptotic expansions, 213; to the sum
;
of a series of positive terms, 8. See also Asymptotic expansions, Method of stationary phase and
Method of steepest descents
Arbitrary functions, expansions of, see Neumann series and Kapteyn Beries (for complex variables);
Dini series, Fourier-Bessel series, Neumann series and Schlbmilch series (for real variables)
Argument of a Bessel function defined, 40
Asymptotic expansions, approximations to remainders in, 213; conversion into convergent series,
204; for Bessel coefficients of order zero with large argument, 10, 12, 194; for Bessel functions
of arbitrary order with large argument, 194-224 (Chapter vn) (functions of the first and second
;
kinds), 199 (functions of the third kind), 196; (functions of the third kind by Barnes' methods),
;
220; (functions of the third kind by Schlafli's methods), 215; (functions with imaginary argu-
ment), 202; for Bessel functions with order and argument both large, 225-270 (Chapter vni);
(order greater than argument), 241 (order less than argument), 244 (order nearly equal to argu-
; ;
ment), 245 (order not nearly equal to argument, both being complex), 262; for combinations of
;
squares and products of Bessel functions of large argument, 221, 448; for Fresnel's integrals,
545 for functions of Anger and Weber (of arbitrary order with large argument), 313 (with order
; ;
and argument both large), 316; for Lommel's functions, 351; for Lommel's functions of two
variables, 549; for Struve's function (of arbitrary order with large argument), 332; (with order
and argument both large), 333 for Thomson's functions, ber (z) and bei (z), 203; for Whittaker's
;
function, 340; magnitude of remainders in, 206, 211, 213, 236, 314, 332, 352, 449; sign of
remainders in, 206, 207, 209, 215, 315, 333, 449. See also Approximations
tion of), 5, 12, 59, 60; expressed as limit of a Legendre function, 65, 155, 157; oscillations of a
uniform heavy chain and, 3, 4 Parseval's integral representing, 9 with large argument, asymp-
; ;
totic expansion of, 10, 12, 194; zeros of, 4, 5. See also Bessel coefficients, Bessel functions and
Bessel's differential equation
Bessel coefficients J (z), 5, 6, 13, 14-37 (Chapter n); addition theorem for, 29; Bessel's integral
for, 19; expansion in power series of, 15; generating function of, 14, 22, 23; inequalities satisfied
by, 16, 31, 268; notations for, 13, 14; order of, 14; (negative), 16; recurrence formulae for, 17;
square of, 32; tables of (of orders and 1), 662, 666-697; (of order i), 664, 730-732; (with equal
order and argument), 664, 746; tables of (references to), 654, 655, 656, 658. See also Bessel
coefficient of order zero, Bessel'B differential equation and Bessel functions
Bessel functions, 38-84 (Chapter m) argument of, defined, 40 differential equations of order
; ;
higher than the second satisfied by, 106 expressed as limits of Lame functions, 159 expressed
; ;
as limits of P-functions, 158 history of, 1-13 (Chapter i) (compiled by Maggi and by Wagner),
;
;
13 indefinite integrals containing, 132-138 ; order of, defined, 38, 58, 63, 67, 70 rank of, de-
; ;
fined, 129 ; relations between the various kinds of, 74 representation of cylinder functions in
;
terms of, 82; solutions of difference equations in terms of, 83, 355; solutions of Laplace's
;;;
three-
equation containing, 83, 124 ; solutions of the equation of wave motions containing, 125 ;
term relations connecting, 300 ; with negative argument, 75. See also the two preceding
and ten
following entries, and Cylinder functions
367, 368
Bessel functions of the first kind, JV (*), 38 ; addition theorems for, 143, 359, 362, 363,
cut in
Barnes' type of integral representing, 190 ; Bessel's type of integral representing, 176 ;
plane to render uniform, 45 differential equation (Bessel's) satisfied by, 38
;
expansion of, ;
m
expressed
ascending series, 40 ; expansion of, in descending series, see Asymptotic expansions ;
limit of a hypergeometnc
as a generalised hypergeometric function, 100, 101 ; expressed as the
significance of), 155;
function, 154 ; expressed as the limit of a Legendre function, 156; (physical
generalisations
expressed as the limit of a Lommel polynomial, 302 functional properties of, 45 ; ;
55 (notations for), 55, 80 ; Poisson's integral representing, 47, 48 ; (modifications of), 161, 163,
;
of two, expressed as a
164, 169, 170 ; products of, see Products of Bessel functions ; quotient
continued fraction, 153, 154, 303 ; recurrence formulae for, 45, 294 ; relations with
Lommel s
polynomial, 297; represented by integrals containing Legendre functions, 173, 174; symbolic
formulae for, 170; tables of (of orders and 1), 662, 666-697 ; (of order n), 664, 730-732 ; (of
order n+4), 664, 740-745; (of order $), 664, 714-729; (of order - *, method of computing), 664;
(with equal order and argument), 664, 746 ; (zeros of), 664, 748-751 ; tables of (references
to),
654, 655, 656, 658, 659, 660 ; Weierstrassian product representing, 497
with large argument, see ;
(after Neumann), 67; Yv (z) (after Weber-Schlafli, the canonical form), 63; addition theorems
for, 144, 361, 365, 368 Bessel's type of integral representing, 177 component parts of, 71, 72,
; ;
340; continuity of (qua function of their order), 63; differential equation (Bessel's) satisfied by,
59, 63 expansion of, in ascending series, 59, 60, 61, 69, 72 expansion of, in
;
descending series,
;
Poisson's type of integral representing, 68, 73, 165; (modifications of), 169, 170; products of,
149; (represented by infinite integrals), 221, 441,' 446 ; (asymptotic expansions of), 221, 448;
recurrence formulae for, 66, 71 represented by integrals containing Legendre functions, 174
;
symbolic formulae for, 170 tables of (of orders and 1), 662, 666-697 (of order n), 664, 732-
; ;
735 ; (of order J), 664, 714-729; (of order -J, method of computing), 664; (with equal order
and argument), 664, 747 (zeros of), 748-751 tables of, references to, 655, 656, 658 ; with large
;
;
argument, see Asymptotic expansions ; with negative argument, 75 zeros of, set. Zeros of ;
fications of), 168, 169, 170; represented by integrals containing Legendre functions, 174;
symbolic formulae for, 170 ; tables of (of orders O'and 1), 662, 666-697 ; (of order ), 664, 714-
729 ; tables of (references to), 657 with large argument, asymptotic expansions of, 199, 210,
;
215 with large argument and order, asymptotic expansions of, 244, 245, 262 ; with negative
;
argument, 75
Bessel functions whose order and argument are equal, approximations to, 229, 231, 232, 259, 260,
448, 515 ; asymptotic expansions of, 245 ; integrals representing, 258 ; tables of, 658, 664, 746,
747 ; tables of (references to), 658
Bessel functions whose order fraction. Of orders J (and Airy's integral), 190; (and the
is a
stability of a vertical pole), 96 tables of, 664, 714-729 ; tables of (references to), 659 ; zeros of,
;
751. Of orders |, tables of (references to), 659. Of orders J, j . tables of (references to),
659. Of small fractional orders, tables of zeros of (references to), 502, 660. See also Bessel
funetionswhose order is (n + )
Bessel functions whose order is large, 225-270 (Chapter vm) asymptotic expansions of, 241, 244,
;
245, 262 Carlini's approximation to, 6, 7 ; (extended by Meissel), 226, 227 ; Horn's (elementary)
;
approximation to, 225 ; Laplace's approximation to, 7, 8, 9 ; method of stationary phase applied
to, 232 ; method of steepest descents applied to, 237 ; miscellaneous properties of, 252-261
tables of (reference to), 658 ; transitional formulae for, 248 ; zeros of, 513, 516, 517, 518. See
also Bessel functions whose order and argument are equal
Bessel functions whose order is =(n t J), 10, 52, 80 ; expressible in finite terms, 52 ; notations for,
55, 80 ; tables of, 664, 740-745 ; tables of (references to), 658, 659
Bessel functions with Imaginary argument, I{z), K
K (z), K v (z), 77, 78; differential equation
satisfied by, 77 integrals representing (of Bessel's type), 181 ; (of Poisson's* type), 79, 171, 172;
;
(proof of equivalence of various types), 185-188 monotonic property of, 446 ; of order (, + 1),
;
80 ; recurrence formulae, 79 tables of (of orders and 1), 663, 698-713 ; (of order J), 664, 714-
;
729 ; (of various integral orders), 664, 736, 737-739 ; tables of (references to), 657, 658; with large
argument, asymptotic expansions of, 202 ; zeros of, 511 ; (computation of), 512 ; (references to),
660
Bessel's differential equation, 1, 19 ; (generalised), 38 ; for functions of order zero, 5, 12, 59, 60;
for functions with imaginary argument, 77 ; fundamental system of solutions of, 42, 75 ; has no
798 THEORY OP BESSEL FUNCTIONS
algebraic integral, 117 ; soluble in finite terms when and only when the functions satisfying it
are of order n + %, 52, 119; solution of, in ascending series, 39, 40, 57, 59-61 ; solution of, in
descending series, see Asymptotic expansions ; symbolic solution of, 41 ; transformations of, 94,
97. See also Bessel coefficients and Bessel functions
Bessel's integral representing Bessel coefficients, 19, 21; generalisations and extensions of, see
Anger's function, Bourget's function, Brans' function and Weber's function ; modifications of,
to represent Bessel functions of arbitrary order, 175, 176, 177, 178, 181 ; Theisinger's transforma-
tion of, 184 ; used in theory of diffraction, 177 ; used to obtain asymptotic expansions, 215. See
also Parseval's integral
Bounds, upper, see Inequalities'
Bourget's function </,* (z), 326 ; differential equation satisfied by, 327 ; recurrence formulae for,
326
Brans' function J(z;v, k), 327
Carlini'sapproximation for Bessel functions of large order, 6, 7 ; extended by Meissel, 226, 227
Cauchy's numbers i^_ n i OT 324 ; recurrence formulae for, 325
, ,
,
higher than the second by, 106 three-term relations connecting, 300. See also Bessel functions
;
solutions of, 518 satisfied by the product of two Bessel functions, 145, 146 solved by elemen-
; ;
tary transcendants, 112; symbolic solutions of, 41, 108. See also under the names of special
equations, such as Bessel's differential equation, and under the names of various functions and
polynomials satisfying differential equations, such as Anger's function
Differential equations (partial), solution of by an integral containing Bessel functions, 99 see also ;
Dlni series, 577, 580, 596-605, 615-617 (Chapter xvra), 651-653 ; expansion of an arbitrary func-
tion of a real variable into, 580, 600 ; methods of theory of functions of complex variables applied
to, 596, 602 ; Riemann-Lebesgue lemma, analogue of, 599 ; Riemann's theorem, analogue
of,
649; summability of, 601, 615 ; uniformity of convergence of, 601, 604 ; uniqueness of, 616, 651
value at end of range, 602
Dirichlet's discontinuous factor, 406
Discontinuity of arbitrary constants (Stokes' phenomenon), 201, 203, 238, 336
Discontinuous factor (Dirichlet's), 406; (Weber's), 405
Discontinuous integrals, 398, 402, 406, 408, 411, 415, 421
Domain K (Kapteyn's), 559 ; diagram of, 270
Du Bois Reymond's integrals with oscillatory integrands expressed in terms of Bessel functions, 183
112
Equal order and argument, Bessel functions with, 231, 232, 258, 260 ; tables of, 746, 747 ; tables
of (references to), 658, 664
Buler's solution of Riccati's equation, 87
Exponential function, tables of, 698-713 tables referred to, 663, 664
;
Factors, discontinuous (Dirichlet's), 406 ; (Weber's), 405 ; Neumann's en (=1 or 2), 22 ; expression
of Bessel functions as products of Weierstrassian, 497
Fejer's theorem, analogue of, for Fourier-Bessel expansions, 610
Finite terms, Bessel functions of order Ht (n+i) expressed in, 52 ; Bessel functions of other orders
not so expressible, 119 ; solutions of Riccati's equation in, 85, 86, 89 ; the solution of Riccati's
equation in, not possible except in Daniel Bernoulli's cases and their limit, 123
Flights, problem of random, 419
Fourier-Bessel expansion, 580. See also Fourier-Bessel series
Fourier-Bessel functions, 4, 84
Fourier-Bessel integrals, see Multiple infinite integrals
Fourier-Bessel series, 576-617 (Chapter xvm), 649-651 ; expansion of an arbitrary function of a
real variable into, 576, 580 ; Fejer's theorem, analogue of, 610 ; Kneser-Sommerfeld expansion
of a combination of Bessel functions into, 499 ; methods of theory of functions of complex vari-
ables applied to, 582, 607 ; order of magnitude of terms in (Sheppard's theorem), 595 ; Riemann-
Lebesgue lemma, analogue of, 589 ; Riemann's theorem, analogue of, 649 ; summability of, 578,
606, 613 ; term-by-term differentiation of, 578, 605 ; uniformity of convergence of, 593, 594
(near origin), 615 ; uniformity of summability of, 612 ; uniqueness of, 616, 649 ; value at end of
range, 594, 603
Fractional differential coefficients, 107, 125
Fresnel's integrals, 544 ; asymptotic expansion of, 545 ; tables of, 744, 745 ; tables of maxima
and minima of, 745 ; tables of (references to), 660, 661, 664
Functional equations denning cylinder functions, 82 ; generalised by Nielsen, 355
Functions of large numbers, approximations due to Darboux, 233 ; approximations due to Laplace,
8, 421 . See also Approximations, Asymptotic expansions, Method of stationary phase and Method
of steepest descents
Fundamental system of solutions of Bessel's differential equation, 42, 75, 78
Gallop's discontinuous infinite Integrals, 421
Gamma functions, representation of Bessel functions by integrals containing, 190, 192, 221 ; appli-
cations to determination of asymptotic expansions, 220, 223 ; applications to evaluation of infinite
integrals, 383, 434, 436
Gamma functions, representation of Lommel's functions by integrals containing, 351 ; applications
to determination of asymptotic expansions, 352
Gegenhauer's addition theorem for Bessel functions, 362, 363, 367
Gegenbauer's discontinuous infinite integrals, 415, 418
Gegenbauer's function Cn " (z), 50, 129, 363, 365, 367, 368, 369, 378, 407
Gegenbauer's polynomial A nyV (t), 283 ; contour integrals containing, 284, 524 ; differential equa-
tion satisfied by, 283 equivalence with special" forms of Lommel's function, 351 ; recurrence
;
Hankel's infinite integrals, 384, 386, 389, 390, 393, 395, 424, 427, 428, 434
Hansen's upper bound for Jn (x), 31 generalised, 406
;
Hardy's functions Ci n (a), Si n (a), Ein (a), (generalisations of Airy's integral), 320; expressed in
terms of functions of Bessel, Anger and Weber, 321, 322
Hardy's integrals representing Lommel's functions of two variables, 546
Hardy's method of evaluating definite integrals, 382
Heat, conduction of, 9, 10, 450, 576, 577, 616
Hemi-cylindrical functions S w (z), defined, 353 ; expressed in terms of the function of order zero,
353 addition theorem for, 354
;
notations for, 100 relations between (Rummer's formulae), 101, 102 Sharpe's differential equa-
; ;
Imaginary argument, Bessel functions with, gee Bessel functions with imaginary argument;
Struve's functions with, 329, 332
Indefinite integrals containing Bessel functions under the integral sign, 132-138, 350, 581 ; tables
of, 744, 745, 752 ; tables of (references to), 660, 661, 664
Inequalities satisfied by Bessel functions, 16, 31, 49, 255, 259, 268, 406; by Neumann's poly-
nomial, 273, 282 ; by Struve's function, 328, 337, 417 ; by zeros of Bessel functions, 485, 489,
490, 492, 494, 515, 516, 521
Infinite Integrals containing Bessel functions under the integral sign, 383-449 (Chapter xm) ; dis-
continuous, 398, 402, 406, 408, 411, 415, 421 ; generalised, 441 ; methods of evaluating, described,
383 ; Bamanujan's type (integrals of Bessel functions with respect to their order), 449. See also
under the names of various integrals, e.g. Lipschitz-Hankel infinite integral
Infinity of the number of zeros of Bessel functions and cylinder functions, 4, 478, 481, 494, 495
Integrals, expressed in terms of Lommel's functions of two variables, 540 ; expressed in terms of
the functions of Anger and Weber, 312 ; Fresnel's, 544, 545, 660, 661, 664, 744, 745 ; Gilbert's,
548, 549 values of, deduced from addition theorems, 367 ; with oscillatory integrands, 183
;
with the polynomials of Neumann and Gegenbauer under the integral sign, 277, 285. See also
Definite integrals and Infinite integrals
Interference, 229
Interlacing of zeros of Bessel functions and of cylinder functions, 479, 480, 481
Irrationality of *-, 90, 485
Jacobl's transformation connecting sinnl with the (n - l)th differential coefficient of sin 2B_1 with
respect to cos d, 26 ; erroneously attributed to Bodrigues, 27 ; various proofs of, 27, 28
Kapteyn series, 6, 13, 551-575 (Chapter xvn) ; connexion with Kepler's problem, 551 ; expansions
into, derived from Kepler's problem, 554, 555 expansion of an arbitrary analytic function into,
;
570; fundamental expansions into, 557, 559, 561, 564, 566, 568, 571 ; Kapteyn's domain K, of
convergence of, 559 (diagram of), 270 nature of convergence outside and on the boundary of
; ;
approximations to, 158 relation between two kinds of, 174 Whipple's transformation of, 387.
; ;
of Gegenbauer, Neumann and Schlfifli, 350; special cases with /jlv an odd negative integer,
348 with large argument, asymptotic expansion of, 351
;
Lommel's functions of two variables, Uv (w, z), Vv (w, z), 537, 538 addition formulae for, 543 ;
integrals representing, 540, 546 reciprocation formulae, 542 recurrence formulae, 539 special
; ; ;
case of, 581, 752 ; tables of, 752 ; tables referred to, 660 with large argument, asymptotic expan-
;
9m, v (z), 303 limit of, expressed as a Bessel function, 302 of negative order, R^ m v (z), 299
; ; ,
recurrence formulae, 298 recurrence formulae in Hurwitz' notation, 303 ; relations with Bessel
;
functions, 295, 297, 302 ; three-term relations connecting, 300, 301 zeros of, 304, 305, 306 ;
Magnitudes of remainders in asymptotic expansions, 206, 211, 213, 236, 314, 332, 352, 449
MaTJitia of Bessel functions, 488; of FresnePs integrals, table of, 745 of integrals of Bessel func- ;
752
tions, table of,
Mean anomaly, expansions of elements of an orbit in trigonometrical series of, 6, 13, 552, 554, 556
Mehler-Dirichlet integral representing Legendre functions, limiting form expressed as Poisson's
integral, 157
Mehler-Sonine integrals representing Bessel functions, 169, 170
Meissel's approximations to Bessel functions of large order, 226, 227, 232, 247, 521 ; types of
Kapteyn series, 557, 561, 564, 566
Membrane, vibrations of a circular, 5, 576, 618 ; vibrations of a sectorial, 510
Method of constant phase (Schlafli's), 216
Method of stationary phase, 225, 229 applied to Bessel functions, 231, 233
;
Method of steepest descents, 235 applied to Bessel functions, 237, 241, 244, 245, 262 applied to
; ;
functions of Anger and Weber, 316; applied to Struve's function, 333; connexion with Laplace's
method of approximation, 421
Minima of Bessel functions, 488 of Fresnel's integrals, table of, 745 of integrals of Bessel func-
; ;
tions, 752
Monotonic properties of Jv (vx)jjv (*), 257; of Jv (v) and ,TV '
(v), 260 ; of K (.r), 446
Multiple infinite Integrals, 450-476 (Chapter xrv) ; investigated by Neumann, 453, 470 (generalised ;
by Hankel), 453, 456, 465; (generalised by Orr), 455; (modified by Weber), 468; Riemann-
Lebesgue lemmas, analogues of, 457, 471 Weber's type of, 450 ;
Neumann series, 522-537 (Chapter xvi); expansion of an arbitrary analytic function into, 523;
generalised, 525 ; (special series), 30, 31, 36, 69, 71, 151 Laurent's expansion, analogue of, 524;
;
Pincherle's theorem on the singularities of, 526; special series, 18, 23, 25, 33, 34, 35, 128, 130,
138, 139, 140, 527, 581 ; Webb-Kapteyn (real variable) theory of, 533. See also Addition theorems
and Lommel's functions of two variables
Neumann's factor eB (=1 or 2), 22
Neumann's Integral for <7m2 (z), 32 for Neumann's polynomial, 278, 280
;
Neumann's polynomial On (t), 271, 272, 273 connected with Kapteyn's polynomial, 569 connected
; ;
with Neumann's polynomial O n [t), 292 connected with Schlafli's polynomial, 285, 286 contour
; ;
integrals containing, 277 differential equation satisfied by, 276 expressed in terms of Lommel's
; ;
functions, 350 ; formerly called a Bessel function of the second kind, 67, 273 generalised by ;
Gegenbauer, see Gegenbauer*s polynomial A n<v (t) generating function of, 281, 282 inequali-
; ;
278, 280; of negative order defined, 276; recurrence formulae for, 274
Neumann's polynomial On (t), 290, 291 ; expressed as integral containing Neumann's polynomial
On (t), 292 Gegenbauer's generalisation of, see Oegenbauer's polynomial B n p, v (t) recurrence
;
,
;
Order of a Bessel function denned, 38, 58, 63, 67, 70 integrals with regard to, 449
;
Oscillatory integrands, Du Bois Beymond's integrals with, expressed in terms of Bessel functions,
183
439, 440, 441, 445, 446, 448 ; series of, 30, 151, 152 with large argument, asymptotic expansions
;
Radius vector of an orbit, expansion as trigonometrical series of the mean anomaly, 6, 13,552,553,554
Ramanujan's integrals of Bessel functions with respect to their order, 449
Bamanujan's method of evaluating definite integrals, 382
Random flights, problem of, 419
Rank of Bessel functions and cylinder functions, 129
Real variables, expansions of arbitrary functions of, see Dini series, Fourier-Bessel series,
Neumann series (Webb-Kapteyn theory), and 8chl6mllch series
Reality of zeros of Bessel functions, 482, 483, 511
Reciprocation formulae for Lommel's functions of two variables, 542
Recurrence formulae for Anger's functions, 311 ; for Bessel coefficients, 17 for Bessel functions ;
of the first kind, 45 ; for Bessel functions of the second kind, 66, 71 for Bessel functions of the
;
third kind, 74 for Bessel functions with imaginary argument, 79 for Bourget's functions, 326
; ;
for Cauchy's numbers, 325 ; for cylinder functions, 82 for Gegenbauer's polynomials, 283 for
; ;
tions, 329 ; for Weber's functions, 311 ; for Whittaker's functions, 339. See also Functional
equations, Heml-cylindrlcal functions and Three-term relations
Reduced functions, Cailler's, 536
Remainders in asymptotic expansions, magnitudes of, 206, 211, 236, 314, 332, 352 ; signs of, 206,
207, 209, 215, 315, 333 ; Stieltjes' approximations to, 213
Repetition of zeros of Bessel functions and cylinder functions, impossibility of, 479
Riccati's differential equation, 1, 2, 85-94; connexion with Bessel's equation, 1, 90} equation
cognate to, 91; limiting form of, 86; soluble cases .of (D. Bernoulli's), 85; soluble cases of,
exhausted by D. Bernoulli's formula and its limit, 123 ; solutions by various mathematicians
(D. Bernoulli), 2, 85, 89 ; (Cayley), 88 ; (Euler), 87 ; (Schlafli), 90 ; solved by means of infinite
series by James Bernoulli, 1 ; transformations of, 86
Riccati's differential equation generalised, 3, 92, 94 ; cross -ratio of solutions. 94 ; equivalence
GENERAL INDEX 803
with the linear equation of the second order, 92 ; singularities of, 94 ; soluble by various num-
bers (two, one or none) of quadratures, 3, 93
Riemann-Lebesgue lemma, analogues of the, 457, 471, 589, 599
Riemann's theorem on trigonometrical series, analogues for Schlomilch series, 642, 647 ; analogues
for series of Fouricr-Bcsscl and Dini, 649
Rodrigues' transformation, see Jacobl's transformation
Schafheitlln's discontinuous Infinite integral, 398, 402, 405, 406, 408, 411
Bcbafheitlln's integrals representing Bessel functions and cylinder functions, 168, 169, 490, 491, 493
Bchlafli's functions Tn (z) and Vn (z),
71, 340, 343; addition theorems for, 344, 345; differential
equations satisfied by, 342, 343; of negative order, 343 ; recurrence formulae for, 71, 342, 343
8chlafli's hypergeometric function, 90
Schlafli's polynomial .S' (t), 284, 286 addition theorem for, 289 connexion with Neumann's
; ;
polynomial On (t), 285, 286; Crelier's integral representation of, 288; differential equation
satisfied by, 285 expression by means of Bessel functions, 287; expression in terms of Lommel's
;
function, 350 integrals evaluated in terms of, 350 recurrence formulae for, 285
; ;
pansion of an arbitrary function of a real variable into, 619, 623, 629 ; nature of convergence of,
637, 645 null-functions expressed by, 634 Riemann's theorem on trigonometrical series (ana-
; ;
logue of), 642, 647; special cases of, 632; symbolic operators in the theory of, 626, 627; theory
of functions of complex variables connected with, 623 uniqueness of, 643, 647
;
Series containing Bessel functions, see Dini series, Fourier-Bessel series, Kapteyn series, Neumann
series and Schlomilch series
Series of Bessel functions, definition of, 580
Series of positive terms, approximation to the sum of (greatest term method), 8
Sharpe's differential equation, 105 solution by generalised hypergeometric functions, 105
;
Sign of remainders in asymptotic expansions, 206, 207, 209, 215, 315, 333, 449 ; of Struve's func-
tion, 337, 417
Sine-Integral expressed as a series of squares of Bessel coefficients, 152
Singularities of functions defined by Neumann series (Pincherle's theorem), 526 of the generalised ;
Riceati equation, 94
Smallest zeros of Bessel functions, 5, 500, 516
Sommerfeld's expansion, see Kneser-Sommerfeld expansion
Sonine-Mehler integrals representing Bessel functions, 169, 170
Bonlne's definite integral, 373 ;
generalised, 382
Bonlne's discontinuous Infinite Integrals, 415
Bonlne's infinite integrals, 432
Spherical geometry used to obtain transformations of integrals, 51, 374, 376, 378 ; used to express
Bessel functions as limits of Legendre functions, 155
Sound, Sharpe's differential equation in the theory of, 105
Squares of Bessel functions, see Products of Bessel functions
Stability of a vertical pole associated with Bessel functions of order one-third, 96
Stationary phase, method of, 225, 229 ; applied to Bessel functions, 231, 233
Steepest descents, method of, 235 ; applied to Bessel functions, 237, 241, 244, 245, 262 ; applied
to functions of Anger and Weber, 316 ; applied to Struve's function, 333 connexion with ;
equal1 order and argument), 664, 746 of Bessel functions of the first kind (of orders n + J,
i ; -
- n - ),
664, 740-741 (of order i), 604, 714-729 of Bessel functions of the second kind (of orders
;
;
and
1), 662, 666-697 (of order n), 664, 732-735
;
(of order ^), 664, 714-729 ; (with equal order and
;
argument), 664, 747 ; of Bessel functions of the third kind (of orders and 1), 662, 666-697 (of
order.Jj),
"'
" "" ' ** "
" ;
698-7i3;
integrals,
tions (of orders and 1), 663, 666-697 ; of zeros of Bessel coefficients and functions of integral
order n and of order $, 664, 748-751
Tables (references to) of Airy's integral, 659 of Bessel coefficients and iunctions derivable from
;
them, 654, 655, 656, 658 of Bessel functions (of orders n + , - n - h), 658, 659 (of orders I,
; ;
3), 659; (of orders J, ), 659; of Bessel functions of the second kind, 655, 656, 658; of
Bessel functions of the third kind, 657 ; of Bessel functions with imaginary argument, 657, 658; of
Fresnel's integrals, 661 of integrals of Bessel functions and Struve's functions, 661 ; of LommePs
;
functions of two variables, 660 ; of Thomson's functions ber r and bei x, etc., 658 ; of zeros of
Bessel coefficients, functions and associated functions, 659, 660
Theisinger's integral representation of Bessel functions, 184 ; of Struve's and Weber's functions,
338
Thomson's (Sir William) functions, ber 2, beiz, 81 ; connexion with Bessel functions, 81 ; generali-
sations, 81 ; references to tables of, 658 ; squares and products of, 82, 148 ; with large argument,
asymptotic expansions of, 203
Three-term relations connecting Bessel functions, cylinder functions and Lommel's polynomials,
300, 301
Transcendants, elementary, definition of, 111 order of, 111 ; solutions of differential equations
;
by, 112
Transitional regions associated with Bessel functions of large order, 248
Uniformity of convergence of Dini series, 601 of Fourier-Bessel series, 593, 594 of Kapteyn series,
; ;
Wave-motions, equation of, general solutions, 125 generalised to p dimensions, 128 ; used to ;
function, 336 ; differential equation satisfied by, 312 integrals expressed in terms of, 312 re- ; ;
currence formulae for, 311 representation of Airy's integral (generalised) by, 321 tables of, see
; ;
Struve's function Theisinger's integral for, 338 with large argument, asymptotic expansion of,
; ;
313 ; with large argument and order, asymptotic expansion of, 316
Weierstrassian products, expression for Bessel functions as, 497
Whipple's transformation of Legendre functions, 387
Whittaker's function v (z), 339 W
differential equation satisfied by, 339
;
recurrence formulae for, ;
Zeros of Bessel functions, 477-521 (Chapter xv) computation of (various methods of), 142, 500, 502,
;
503, 516 inequalities connected with, limits of, rates of growth of, 485, 489, 490, 491, 494, 507,
;
513, 516, 518 infinity of, 4, 478 interlacing of, 479, 480, 481 ; non-coincidenee of (Bourget's
; ;
hypothesis), 484 non-repetition of, 479 ; number of, in a strip of arbitrary width, 495 reality of,
; ;
482, 483; tables of, 664, 748-751 tables of (references to), 659 ; values of, 4, 5, 512, 516 with
; ;