Derivatives of any order of the confluent hypergeometric function 1F1(a,b,z) with respect to the parameter a or b

L. U. Ancarani and G. Gasaneo Citation: J. Math. Phys. 49, 063508 (2008); doi: 10.1063/1.2939395 View online: http://dx.doi.org/10.1063/1.2939395 View Table of Contents: http://jmp.aip.org/resource/1/JMAPAQ/v49/i6 Published by the American Institute of Physics.

Related Articles
Kernel functions and Bcklund transformations for relativistic Calogero-Moser and Toda systems J. Math. Phys. 53, 123512 (2012) The rigged Hilbert space approach to the Gamow states J. Math. Phys. 53, 102113 (2012) Indirect Coulomb energy for two-dimensional atoms J. Math. Phys. 53, 095213 (2012) Yet another position-dependent mass quantum model J. Math. Phys. 53, 072101 (2012) A different approach to obtain Mayers extension to stationary single particle Wigner distribution Phys. Plasmas 19, 072101 (2012)

Additional information on J. Math. Phys.

Journal Homepage: http://jmp.aip.org/ Journal Information: http://jmp.aip.org/about/about_the_journal Top downloads: http://jmp.aip.org/features/most_downloaded Information for Authors: http://jmp.aip.org/authors

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

JOURNAL OF MATHEMATICAL PHYSICS 49, 063508 2008

Derivatives of any order of the conuent hypergeometric function 1F1a , b , z with respect to the parameter a or b
L. U. Ancarani1 and G. Gasaneo2,a
1

Laboratoire de Physique Molculaire et des Collisions, Universit Paul Verlaine-Metz, 57078 Metz, France 2 Departamento de Fsica, Universidad Nacional del Sur and Consejo Nacional de Investigaciones Cientcas y Tcnicas, 8000 Baha Blanca, Buenos Aires, Argentina Received 19 March 2008; accepted 9 May 2008; published online 20 June 2008

The derivatives to any order of the conuent hypergeometric Kummer function F = 1F1a , b , z with respect to the parameter a or b are investigated and expressed in terms of generalizations of multivariable Kamp de Friet functions. Various properties reduction formulas, recurrence relations, particular cases, and series and integral representations of the dened hypergeometric functions are given. Finally, an application to the two-body Coulomb problem is presented: the derivatives of F with respect to a are used to write the scattering wave function as a power series of the Sommerfeld parameter. 2008 American Institute of Physics. DOI: 10.1063/1.2939395

I. INTRODUCTION

The conuent hypergeometric function F = 1F1a , b ; z or alternatively Whittakers function has been studied in great detail from its mathematical point of view.13 The Kummer function, as it is also named, is closely related to a fundamental problem of quantum mechanics: the two-body Coulomb problem. It is well known that the closed form solution of the two-body Coulomb Schrdinger equation both in spherical and parabolic coordinates are written in terms of the Kummer function F.4 To perform different types of physical studies see, e.g., Refs. 59 and applications,1013 it is necessary to know the mathematical properties of the function. Usually F is considered as a function of variable z; however, in some physical applications, the rst a or second b parameter may be the physical variable, as, for example, when Coulomb solutions are extended to the complex plane of the energy5,7 or the angular momentum.5 The rst and the nth derivatives with respect to the variable z is known in a compact form e.g., Ref. 1. The derivatives with respect to the rst or second parameter, on the other hand, have been less studied simply because the mathematical formulation is more difcult see below. These derivatives, however, are needed in different physical applications, the Coulomb Born series being probably the most well known example. Several forms of the rst derivatives have been given in the literature; none of them, however, is compact except possibly in some special cases. In this contribution we address this issue by providing a compact form not only for the rst but also for the nth derivatives. The usefulness of the investigation is illustrated by considering the two-body Coulomb wave function as a function of the Sommerfeld parameter. We shall use the following notation for the nth derivatives: d n 1F 1

G n = G n a , b ; z =

a ;z b dan

Electronic mail: ggasaneo@gmail.com. 49, 063508-1 2008 American Institute of Physics

0022-2488/2008/496/063508/16/$23.00

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-2

L. U. Ancarani and G. Gasaneo

J. Math. Phys. 49, 063508 2008

H n = H n a , b ; z =

d n 1F 1

a ;z b dbn

Let us recall the denition of the conuent hypergeometric function as a power series on the variable z:

F = 1F 1

a an n ;z = z , n b n=0 !bn

where an = a + n / a is the Pochhammer symbol dened in terms of the gamma function.1 A rst approach to get the rst derivative G1, discussed, for example, in Ref. 1, is based on the use of the derivative of the Pochhammer symbol: dan = ana + n a , da 4

and its denition in terms of the digamma function z.14 With the previous denition, the rst derivatives of F with respect to a or b are
1

n=0

an an zn zn a + n a = a + n a 1F 1 bn n! n=0 bn n!

a ;z , b

5a

H 1 =

n=0

an an zn zn b + n b = b + n + b 1F 1 bn n! n! n=0 bn

a ;z . b

5b

We have therefore an innite series containing the digamma function.14 An alternative form is obtained by using the recurrence formula 6.3.6 of Ref. 14: G 1 =

n=0

1 an zn , bn n! p=0 p + a
n1

n1

6a

n=0

1 an zn . bn n! p=0 p + b

6b

In one form or the other, the generalization to the nth derivative is particularly cumbersome. A second approach makes use of Whittaker functions: M , z = e
z/2 1/2+

1F 1

1 2

+ 1 + 2

;z .

In the book The Conuent Hypergeometric Function,3 Buchholz studied briey the rst derivative of the Whittaker function with respect to and gave expression in terms of the digamma function. This author also related the rst derivative with respect to and with inhomogeneous differential equations, but no explicit and compact expressions were given for the derivatives G1 and H1. Moreover, the generalization to the nth derivative was not provided. An algorithm for the computation of Gn was given recently by Abad and Sesma.15 They made use of a convergent expansion of Whittakers function in series of Bessel functions and of the properties of Buchholz polynomials pj2z j = 0 , 1 , 2 , . . . . The main result is

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-3

Derivatives of the Kummer function

J. Math. Phys. 49, 063508 2008

n M , z

pj2z d n M , z n +1/2+n = = 1 z 0F 1 j dn j=0 2 2 + 1 j+n

2 + 1 + j + n

; z .

The algorithm is based on this expansion, truncated at a conveniently large value jmax. In this paper explicit expressions for the nth derivatives Gn and Hn are given in Secs. II and III. The connection with Kamp de Ferit-like multivariable hypergeometric functions is given in Sec. IV, where different properties of these functions are presented. As application of these results, the scattering wave function for the two-body Coulomb problem is expressed as a power series of the Sommerfeld parameter up to order 2 Sec. V. A summary of the results is given in Sec. VI.
II. FIRST DERIVATIVE WITH RESPECT TO a OR b

With an approach based on the solution of inhomogeneous differential equations related to the conuent hypergeometric function F, we give here a compact form for the rst derivatives G1 and H1. The procedure will then be used in the next section to nd expressions for the higher derivatives Gn and Hn. We start from the differential equation satised by the conuent hypergeometric function F;

d2 d + b z a F = 0. dz2 dz

Since F is an analytic function of z and a,1 taking the derivative of Eq. 9 with respect to the rst parameter a, one nds

a m1 z m1 d2 d 1 z 2 + b z a G = F = dz dz m1=0 bm1 m1!

10

the use of m1 as index will become clear in the next section. Similarly, taking the derivative of the differential equation 9 with respect to b F has poles and is not dened for b = 0 , 1 , 2 , . . . Ref. 1, we have d2 a d dF = 1F 1 z 2 + b z a H 1 = dz dz dz b

a+1

a + 1 m1 z m1 a . ;z = b+1 b m1=0 b + 1m1 m1!

11

Now, according to Eq. 4.162 of Ref. 16, the solution of the inhomogeneous differential equation

d2 d a y = z m1 2 + b z dz dz

12

is given by Eq. 4.163: y = m1+1a, b ; z = zm1+1 F 1 + m1b + m1 2 2

1, a + 1 + m1 2 + m 1, b + 1 + m 1

;z .

13

Since the differential equation 10 is linear, the solution for G1 reads G 1 = that is to say, G 1 =
a m1 1 m1 z m1 z F b1 m1=0 b + 1m12m1 m1! 2 2

m +1a, b ; z , m =0 m1!bm
1 1 1

a m1

14

1, a + 1 + m1 2 + m 1, b + 1 + m 1

;z .

15

Similarly, for H1 we nd

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-4

L. U. Ancarani and G. Gasaneo

a + 1 m1 b m1 1 m1 z m1 a z = F b b1 m1=0 b + 1m1b + 1m12m1 m1! 2 2

J. Math. Phys. 49, 063508 2008

1, a + 1 + m1 2 + m 1, b + 1 + m 1

;z .

16

These expressions, to the best of our knowledge, have not been given before. Expanding the 2F2 hypergeometric function in series index m2 and after some algebraic manipulations, one nds G 1 =
1 m1 1 m2 a m1 a + 1 m1+m2 z m1+m2 z , b1 m1=0 m2=0 a + 1m12m1+m2b + 1m1+m2 m1!m2! 1 m1 1 m2 b m1 a + 1 m1+m2 z m1+m2 1a z . b b m1=0 m2=0 b + 1m12m1+m2b + 1m1+m2 m1!m2!

17a

H 1 =

17b

These double series could have been obtained also starting directly from expressions 6a and 6b using the property

n=0 k=0

Bk, n + k . Bk, n = n =0 k=0

18

However, as mentioned in the Introduction, this technique does not allow for an easy generalization to the nth derivatives. Moreover, these double series can be related to the following hypergeometric function in two variables:
1

a 1, a 2 b 1, b 2 c 1 d 1, d 2

; x 1, x 2 =

m =0 m =0
1 2

1 m2 a 1 m1 a 2 m2 b 1 m1 b 2 m1+m2 x m 1 x2 , c 1 m1 d 1 m1+m2 d 2 m1+m2 m 1! m 2!

19

which, as we shall see in Sec. IV, is a Kamp de Friet-like function. Hence, in terms of 1, the rst derivatives G1 and H1 read G 1 = z 1 b1

1,1a, a + 1 a + 12, b + 1

H 1 =

a z 1 b b1

1,1b, a + 1

b + 12, b + 1

; z, z ,

20a

; z, z .

20b

An alternative formulation of the solution to the inhomogeneous differential equation Eq. 12 is given by a nite sum of powers of z see Eq. 4.172 of Ref. 16:
m 1 m 1! b m1 m 1 m 1! b m1 1 a n z n = F 1 a, b, z . m1+1a, b ; z = a a + 1m1 n=0 bn n! a a + 1 m1

21

As showed notin Ref. 16 Eq. 4.175, the functions m1+1a , b ; z and m1+1a , b ; z are linked to each other through

m1+1a, b ; z = m1+1a, b ; z

1 m 1! b m1 F a, b ; z . a a + 1 m1 1 1

22

The function m1+1a , b ; z can be considered as an asymptotic solution of Eq. 12 as it is derived from a series in inverse powers of z; note that limm Fma , b , z = 1F1a , b ; z. Both m1+1a , b ; z and m1+1a , b ; z are closely related to the inhomogeneous Whittaker functions S and R dened by Buchholz.3 In terms of the nite sum Fm1, the derivative G1 given by Eq. 14 can thus be expressed as

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-5

Derivatives of the Kummer function

a m1 1 = F m1 a , b , z . a m1=0 a + 1m1

J. Math. Phys. 49, 063508 2008

23

After some algebraic calculations and the application of series rearrangement techniques,17 the G1 function can be reduced to G
1 a m1 1 = F a m1=0 bm1 3 2

1, a, a + m1 a + 1, a + 1 + m1

;1

z m1 . m 1!

24

For the particular argument of z = 1, the function the 3F2 simplies18 and an expression in terms of digamma functions, leading to Eq. 5a, can be obtained. Following the same procedure, a similar representation for H1 can be obtained.
III. nTH DERIVATIVE WITH RESPECT TO a OR b

Without much effort, one may generalize the result to the nth derivative with respect to a or b. Following the same procedure, we dene rst the inhomogeneous differential equation satised by each derivative order. Consider the second derivative G2. Differentiating with respect to a Eq. 10 satised by 1 G , we have

d2 d dG0 2 1 a = 2 G 1 , + b z G = G + dz2 dz da

25

where we used F = G0. Taking again the derivative with respect to a, we obtain for the third derivative z d2 d dG1 3 2 a = 3 G 2 , + b z G = G + 2 dz2 dz da

26

and, following the same procedure, we nd that the nth derivative satises the general differential equation

G
2

d2 d a Gn = nGn1 . 2 + b z dz dz

27

Proceeding as for G1, in Appendix A we give the detailed calculations that lead to the following explicit expressions for G2:
1 m1 1 m2 1 m3 z2 = b2 m1=0 m2=0 m3=0 3m1+m2+m3b + 2m1+m2+m3

a m1 a + 1 m1+m2 a + 2 m1+m2+m3 z m1+m2+m3 . a + 1 m1 a + 2 m1+m2 m 1! m 2! m 3!

28

It can be easily veried by induction that the general expression for Gn n 1 reads G n =
1m11m2 1mn+1 zn bn m1=0 mn+1=0 n + 1m1+m2++mn+1b + nm1+m2++mn+1

am1a + 1m1+m2 a + nm1+m2++mn+1 zm1+m2++mn+1 . m1!m2! mn+1! a + 1 m1 a + n m1++mn

29

Following a similar procedure, one nds the set of equations satised by Hn:

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-6

L. U. Ancarani and G. Gasaneo

J. Math. Phys. 49, 063508 2008

H
2

d2 d dF a H 1 = , 2 + b z dz dz dz

30a

d2 d dH1 2 a , + b z H = 2 dz2 dz dz ,

30b

d2 dHn1 d n . a + b z H = n dz2 dz dz

30c

The solution for 30a was given above by Eq. 20b. The solution for 30b reads
1 m1 1 m2 1 m3 2a = 1 2 z b b m1=0 m2=0 m3=0 2m1+m2+m3b + 1m1+m2+m3 2

b m1 b m1+m2 a + 1 m1+m2+m3 z m1+m2+m3 , b + 1 m1 b + 1 m1+m2 m 1! m 2! m 3!

31

and, by induction, the solution for the nth derivative Hn is H n = 1 n

1m11m2 1mn+1 n! a z n b b m1=0 mn+1=0 2m1+m2++mn+1b + 1m1+m2++mn+1

bm1bm1+m2 bm1+m2++mna + 1m1+m2++mn+1 zm1+m2++mn+1 . m1!m2! mn+1! b + 1 m1 b + 1 m1+m2 b + 1 m1+m2++mn

32

It is worth mentioning that the system of equations for the nth derivatives Eqs. 10 and 2527 for Gn and Eqs. 30a30c for Hn are all of order 2 and could be numerically solved with well known methods.
IV. CONNECTION WITH MULTIVARIABLE HYPERGEOMETRIC FUNCTIONS AND VARIOUS PROPERTIES

It is interesting to notice that the expressions given above for the derivatives of F with respect to a and b can be expressed in terms of generalizations of multivariable Kamp de Friet hypergeometric functions. In Sec. II we have written the rst derivatives G1 and H1 in terms of the two-variable hypergeometric 1. Similarly, for each nth derivative n 2, we may associate a different multivariable hypergeometric function which we shall name n. Several properties of these functions will be presented in this section.
A. The denition of Gn and Hn in terms of hypergeometric functions

The function 1, dened by Eq. 19, results from the application of the rule used by Appell19 to the product of the generalized conuent hypergeometric functions 3F3 and 2F2,
3F 3

a 1, a 2, a 3 c 1, c 2, c 3

; x1
2F 2

b 1, b 2

d 1, d 2

; x2 =

1 m2 a 1 m a 2 m a 3 m b 1 m b 2 m xm 1 x2 1 1 1 2 2 . m ! m ! c c c d d 2 1 m1 2 m1 3 m1 1 m2 2 m2 m1=0 m2=0 1

Using the replacements a 3 m1 b 1 m2 a 3 m1+m2 ,

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-7

Derivatives of the Kummer function

J. Math. Phys. 49, 063508 2008

c 3 m1 d 2 m2 c 3 m1+m2 , c 2 m1 d 1 m2 c 2 m1+m2 , the following coefcient is generated: a 1 m1 a 2 m1 a 3 m1 b 1 m2 b 2 m2 c 1 m1 c 2 m1 c 3 m1 d 1 m2 d 2 m2 a 1 m1 a 2 m1 b 2 m2 a 3 m1+m2 c 1 m1 c 2 m1+m2 c 3 m1+m2 .

According to the theory presented in Ref. 17, the fact that the functions we started from are conuent hypergeometric functions ensures that the function 1 given by Eq. 19 is also a conuent hypergeometric function whose convergency radius is innity in x1 and x2. The function 1 is a Kamp de Friet function in two variables. Similarly, for the second order derivatives G2 and H2, we may introduce the function 2

a 1, a 2, a 3 b 1, b 2, b 3 c 1, c 2 d 1, d 2

; x 1, x 2, x 3

33

m1=0 m2=0 m3=0

1 m2 m3 a 1 m1 a 2 m2 a 3 m3 b 1 m1 b 2 m1+m2 b 3 m1+m2+m3 x m 1 x2 x3 , c 1 m1 c 2 m1+m2 d 1 m1+m2+m3 d 2 m1+m2+m3 m 1! m 2! m 3!

which can be generated using Appells technique to the product of generalized conuent hypergeometric functions 4F4, 3F3, and 2F2. Generalizing to the nth derivative, we introduce the function n dened by n

a1, a2, . . . , an+1b1, . . , bn+1 c 1, . . . , c n d 1, d 2

; x1, x2, . . , xn+1

, 34

1 2 n+1 m =0 m1 !m2! mn+1! m =0

1 n+1

xm1xm2 xmn+1

a1m1a2m2 an+1mn+1b1m1b2m1+m2 bn+1m1+m2++mn+1 c1m1c2m1+m2 cnm1+m2++mnd1m1+m2++mn+1d2m1+m2++mn+1

which results from the application of Appells technique to the product of n+2Fn+2,. . ., 3F3, and 2F2. The way in which the coefcients are combined is easily induced from the series given here. In terms of these generalized Kamp de Friet-like hypergeometric functions, the derivatives Gn given in the previous section read G 1 a , b ; z = z 1 b1

1,1a, a + 1 a + 12, b + 1

G 2 a , b ; z =

z 2 2 b2

1,1,1a, a + 1, a + 2 a + 1, a + 23, b + 2

G n a , b ; z =

z n n bn

1,1, . . . ,1a, a + 1, . . . , a + n

a + 1, a + 2, . . . , a + nn + 1, b + n

; z, z , ; z, z, z ,

35a

35b

; z, z, . . . , z .

35c

Similar expressions can be obtained for Hn following the same procedure: H n a , b ; z = 1 n n ! a n z bn b

1,1, . . . ,1b, b, . . . , b, a + 1 b + 1, b + 1, . . . , b + 12, b + 1

; z, z, . . . , z .

36

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-8

L. U. Ancarani and G. Gasaneo

J. Math. Phys. 49, 063508 2008

With these derivatives one may thus provide Taylor expansions for the function F in power series of a around a0 or of b around b0, F=

n=0

a a 0 n n G a 0, b ; z , n! b b 0 n n H a, b0 ; z . n!

37a

F= Here we are using the notation G 0 a 0, b ; z = 1F 1

37b

n=0

a0 ; z and H0a, b0 ; z = 1F1 b

a ;z . b0

In Sec V, we shall use expansion 37a for a0 = 0.

B. Particular value: a = 0

The hypergeometric functions n dened by Eq. 34 depend on a large number of parameters. However, in the expression of the derivatives Gn and Hn Eqs. 35c36, only the parameters a and b of the initial function F are actually variable. Since in the calculations to be performed in Sec. V we will need the evaluation of Gna , b ; z with n 1 for a = 0, we now provide explicit formulas for this situation. The reduction formulas for the nth derivatives Gn in the case a = 0 can be related to those, presented in Appendix B, corresponding to the n functions. The results are G10, b ; z = z 1 b1

1,10,1

12, b + 1

G20, b ; z =

z 2 1 b2

Gn0, b ; z =

zn n1 bn

; z, z = z F b1 2 2 1,11,2 23, b + 2 1, . . . ,11, . . . , n 2,3, . . . , nn + 1, b + n

1,1 2, b + 1

;z ,

38a

; z, z ,

38b

; z, . . . , z .

38c

The nth derivatives Gn n 1 for a = 0, dened in terms of n, are thus expressed in terms of a function n1.
C. Recurrence relations for Gn 1. Relations for general a and b

Starting from the recurrence relations for the conuent hypergeometric function F, recurrence relations for the 1 function can be easily deduced. For example, consider the contiguous relation 13.4.4 of Ref. 14: b 1F 1

a ; z b 1F 1 b

a1 b ; z z 1F 1
1

a ; z = 0. b+1

39

If we take the derivative with respect to a, we nd bG1a, b ; z bG1a 1, b ; z zG1a, b + 1; z = 0, and hence, using Eq. 20a, a relation for : 40

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-9

Derivatives of the Kummer function

J. Math. Phys. 49, 063508 2008

1,1a, a + 1 a + 12, b + 1

; z , z 1

1,1a 1, a a2, b + 1

; z, z

z 1 b+1

1,1a, a + 1 a + 12, b + 2

; z, z = 0. 41

Similarly using relation 13.4.13 of Ref. 14, one nds a relation providing the derivative of the 1 function with respect to z, z d 1 b dz

1,1a, a + 1 a + 12, b + 1

; z , z = 1

1,1a, a + 1 a + 12, b

; z , z 2 1

1,1a, a + 1 a + 12, b + 1

; z, z . 42

Several other relations can be obtained in a similar way and easily generalized to n.
2. Relations for a = 0

Let us introduce also the following function: G n,m a , b ; z = z n n bn

1,1, . . . ,1a, a + 1, . . . , a + n a + 1, a + 2, . . . , a + nm + 1, b + n

; z, z, . . . , z .

43

By inspection of relation 38c, we see that 1 Gn,m0, b ; z = zGn1,m1, b + 1; z . b As a consequence, we have the following recurrence relation for n 2: 1 Gn0, b ; z = Gn,n0, b ; z = zGn1,n1, b + 1; z . b
D. Series and integral representations in terms of one-variable hypergeometric functions

44

45

In this section we provide series and integral representations for 1 in terms of well known one-variable hypergeometric functions generalization to n can be easily obtained. Using series rearrangement techniques and different properties for the Pochhammer symbols, it is easy to verify that the following series representations hold for the 1 function: 1

a 1, a 2 b 1, b 2 c 1 d 1, d 2

; x 1, x 2 =

m =0 c1m d1m d2m

1 1 1

1 a 1 m1 b 1 m1 b 2 m1 x m 1 1

m 1!

2F 2

a 2, b 2 + m 1 d 1 + m 1, d 2 + m 1

; x2 46a 46b

m =0 d1m d2m
2 2

2 a 2 m2 b 2 m2 x m 2 2

m 2!

3F 3

a 1, b 1, b 2 + m 2 c 1, d 1 + m 2, d 2 + m 2

; x1 .

Denitions for G1 and H1 result immediately from Eqs. 20a and 20b. Equation 46a is the formulation encountered in Eqs. 15 and 16. The rate of convergence of these series is faster than the one corresponding to the double series of Eq. 19 as illustrated in Fig. 1, where we plot 1 given in Eq. 35a as a function of x1 = x2 = ix for xed values of the parameters a = i / 0.75 and b = 1 the value b = 1 and the purely imaginary values of a and argument correspond to the conuent hypergeometric function appearing in the Coulomb problem, see Sec. V. We used m1 = m2 up to 20 to evaluate 1 with the double series of Eq. 19 and m1 up to 20 or 60 with Eq. 46a. As can be noticed from the gure, a much smaller number of terms are needed when the

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-10

L. U. Ancarani and G. Gasaneo

J. Math. Phys. 49, 063508 2008

FIG. 1. Color online The 1 function given in Eq. 35a is plotted vs the argument x1 = x2 = ix for xed values of the parameters a = i / 0.75 and b = 1: it is is calculated with the double series, Eq. 19 with m1 = m2 up to 20 solid line and squares and the series of Eq. 46a with m1 up to 20 solid lines and circles or 60 solid lines and triangles.

single series is used. This is a natural consequence of the fact that the whole series in m2 has been summed up. A numerically useful integral representation for 1 can be obtained starting from Eq. 46a. Combining the integral representations for the 2F2 see Ref. 20, p. 854:

2F 2

1F 1

, a ; x2 = , b

dt1 t1t1 1F1

a ; x 2t , b

47

where R 0, R 0, and the one corresponding to the 1F1 Ref. 1:

a b

;z =

b ab a

du1 uba1ua1ezu ,

48

where Rb a 0, Ra 0, we obtain

2F 2

, a b ;z = , b ab a

dt1 t1t1

du1 uba1ua1eztu . 49

The condition over the parameters can be removed by performing the integrations over contours on the complex plane.1 With this representation, the series of Eq. 46a becomes

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-11

Derivatives of the Kummer function

J. Math. Phys. 49, 063508 2008

a 1, a 2 b 1, b 2 c 1 d 1, d 2

; x 1, x 2
1

m =0 c1m d1m d2m

1 a 1 m1 b 1 m1 b 2 m1 x m 1 1

d 1 + m 1 d 2 + m 1 m 1! a 2 d 1 + m 1 a 2 b 2 + m 1 d 2 b 2 1

dt1 td1+m1a21ta21

du1 ud2b21ub2+m11ex2tu ,

50

which after algebraic manipulations converts into 1

a 1, a 2 b 1, b 2 c 1 d 1, d 2

; x 1, x 2 a 1, b 1

d 1 d 2 a 2 d 2 b 2 b 2 d 1 a 2

ex2tu 2F2

c 1, d 1 a 2

dt1 td1a21ta21

; x 1 1 t u .

du1 ud2b21ub21 51

This integral representation is particularly useful to numerically evaluate 1, and thus G1 and H1 through the use of Eqs. 20a and 20b. Following similar procedures, integral and series representations of this type can be obtained for n, and hence for Gn and Hn.
E. Special cases

When b = a, the function F reduces to an exponential F = ez, so that there is no parameter dependence and G1 = H1 = 0. Although the result is trivial, it should appear also through the 1. Indeed, we have d F da 1 1

a ; z = G 1 a , a ; z + H 1 a , a ; z , a

52

and using Eqs. 20a and 20b, we immediately nd 0. When b = a 1, the function F reduces to
1F 1

a ; z = e z 1F 1 a1

1 a1

; z = ez 1 +

z . a1

53

The derivative with respect to a is G1 = zez / a 12, a result which can be easily found by applying, for example, Eq. 20b. Another interesting situation is when b = a + 1 since the F function is related to the incomplete gamma function
1F 1

a ; z = a zaa, z . a+1

54

In this case the derivative of the conuent hypergeometric function with respect to a reads d F da 1 1

a z ; z = G1a, a + 1; z + H1a, a + 1; z = 1 a+1 a + 12

1,1a, a + 1 a + 22, a + 2

; z, z . 55

Hence, the rst derivative with respect to a of the incomplete Gamma function reads

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-12

L. U. Ancarani and G. Gasaneo

J. Math. Phys. 49, 063508 2008

1 za z d a, z = a, z ln z + 1 da a a a + 12

1,1a, a + 1 a + 22, a + 2

; z, z

and the nth derivative can be easily derived with the results given in the previous sections. Other special cases can be considered in a similar manner.
V. APPLICATION TO THE BORN-LIKE SERIES FOR THE TWO-BODY COULOMB SCATTERING WAVE FUNCTION

As we mentioned in the Introduction, a close connection exists between the Kummer function and the two-body Coulomb problem. Let r and k respectively represent the relative vector position and momentum between an electron and a heavy nucleus of charge Z placed at the origin of coordinates. The solution for the scattering problem in parabolic coordinates is given by4 + , k , r = N e ikr 1F 1

; ik , 1

56

where outgoing wave boundary conditions are considered. The parabolic coordinates are = r r, and tan = y / x, where the Cartesian coordinates x and y correspond to the r, = r k +k position of the particle relative to the reference center.4 Here atomic units = me = e = 1 are used, so that the electron charge is equal to 1; = iZ / k is the Sommerfeld parameter we have included i in its denition for convenience. The normalization factor N is dened in terms of the gamma function as N = e i/2 1 . 57

Let us consider the expansion of the scattering wave function + , k , r in power series of the Sommerfeld parameter : + , k , r = e ikr 0+ k , r + 1+ k , r + 2+ k , r where l+k , r is given by l+ k , r =

2 + , 2

58

d l + , k , r dl

.
=0

59

Such an expansion is interesting as it is related to the Born series for the Coulomb problem. In order to get analytical expressions for the different orders l+k , r of 58, we shall need the nth derivatives Gn with respect to the rst parameter of the conuent hypergeometric function F appearing in 56. Indeed, they appear in the Taylor expansion 37a of F around a0 = 0; we have F = G00,1; ik + G10, b ; ik +

2 2 G 0,1; ik + , 2

60

where G0 = 1F10 , 1 ; ik = 1. We also need to expand N in power series of : N = N 0 + N 1 + N 2 where N0 = 1, N 1 = i 62a

2 + , 2

61

, 2

62b

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-13

Derivatives of the Kummer function

J. Math. Phys. 49, 063508 2008

N 2 = 2 i

2 , 12

62c

and represents the Euler gamma constant.14 Combining the two series expansions 60 and 61 and comparing with 58, the following expressions are readily deduced for the rst three terms: 0+k, r = N0G00,1; ik , 1+k, r = N1G00,1; ik + N0G10,1; ik , 2+k, r = N2G00,1; ik + 2N1G10,1; ik + N0G20,1; ik . 63a 63b 63c

Using 62a62c and the reduction formulas 38a and 38b for, respectively, G1 and G2, we nd 0+k, r = 1, 1+ k , r = i 64a

+ ik 2F2 2

1,1 2,2

; ik ,

64b

2+ k , r = 2 i + ik2 1 2

2 + 2 i ik 2F2 12 2
1,11,2 23,3

; ik, ik .

1,1 2,2 ; ik 64c

The numerical evaluation of 1 can be easily performed using one of the representations given in Sec. IV D. The function 2F2 appearing in 1+ and 2+ can be reduced to simpler functions:21
2F 2

1,1 2,2

; ik =

log ik 0, ik , ik

where a , z represents the incomplete gamma function.14 Explicit expressions of n+k , r have been given here up to order n = 2; higher orders can be easily obtained in terms of the generalized hypergeometric functions n or of n1 if the reduction formula 45 is used. In order to reduce the difculties in their evaluations, further investigation of the properties of the multivariable hypergeometric functions n dened here is necessary.
VI. SUMMARY

We have given closed form formulas for the nth derivatives of the Kummer function with respect to its parameters a and b. These derivatives are expressed in terms of multivariable Kamp de Friet-like hypergeometric functions, named here n. For 1, which is related to the rst derivatives, various types of properties such as recurrence relations and series and integral representations are provided, and some special cases are discussed. The system of ordinary differential equations satised for the derivatives of order n is also given. The above mathematical study is applied to the physical case of the two-body Coulomb scattering wave function, for which the exact solution in parabolic coordinates is written in terms of the Kummer function. A power series expansion in terms of the Sommerfeld parameter is considered, and analytic closed form expressions up to order 2 were given for the terms, which

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-14

L. U. Ancarani and G. Gasaneo

J. Math. Phys. 49, 063508 2008

are functions only of the energy and the parabolic coordinate. Further studies of the hypergeometric functions n here introduced are necessary and this is the subject of our current investigations. Finally, an investigation of the derivatives of the generalized hypergeometric functions pFq with respect to its parameters would also be of interest. In particular, the 2F1 case can be of high interest because of its application to the study of many physical problems.
ACKNOWLEDGMENTS

One of the authors G. Gasaneo thanks the support by PICTR 03/0437 of the ANPCYT, PGI 24/F038 of the UNS Argentina, and PIP5595 of CONICET.
APPENDIX A: DEMONSTRATION OF FORMULA 28 FOR G2

To get the explicit formula 28 for G2, we start with the differential equation 25 it satises:

d2 d a G 2 = 2 G 1 , 2 + b z dz dz

A1

and the expression of G1 Eq. 17a: G

1 1 m1 1 m2 a m1 a + 1 m1+m2 z m1+m2 z . = b1 m1=0 m2=0 a + 1m12m1+m2b + 1m1+m2 m1!m2!

A2

Using solution 13 of the inhomogeneous Kummer equation 12, we have G 2 =

1 m1 1 m2 a m1 a + 1 m1+m2 2 1 m +m +1+1a, b ; z , b1 m1=0 m2=0 a + 1m12m1+m2b + 1m1+m2 m1!m2! 1 2

A3

which in terms of hypergeometric functions 2F2 becomes G 2 =

1m11m2am1a + 1m1+m2 zm1+m2+2 2 b1 m1=0 m2=0 a + 1m12m1+m2b + 1m1+m2 m1!m2!

1 F 2 + m1 + m2b + 1 + m1 + m2 2 2

1, m1 + m2 + 1 + a + 1 m1 + m2 + 1 + 2, m1 + m2 + 1 + b + 1

;z A4

Using the two identities 1 n 1 , = n + + 1n m+n = m + m n , and simplifying, we nd G 2 =

1 m1 1 m2 a m1 a + 1 m1+m2 1 b1b + 11 m1=0 m2=0 3m1+m2a + 1m1b + 2m1+m2

zm1+m2+2 F m 1! m 2! 2 2

1, a + m1 + m2 + 2 m1 + m2 + 3, b + m1 + m2 + 2

;z .

A5

Now, replacing the series expansion index m3 for the 2F2 function leads to a triple series:

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-15

Derivatives of the Kummer function

J. Math. Phys. 49, 063508 2008

1 m1 1 m2 a m1 a + 1 m1+m2 1 = b1b + 11 m1=0 m2=0 m3=0 3m1+m2a + 1m1b + 2m1+m2

zm1+m2+m3+2 , m 1 + m 2 + 3 m3 b + m 1 + m 2 + 2 m3 m 1! m 2! m 3!

1 m3 a + m 1 + m 2 + 2 m3

A6

which, after some algebraic manipulations, nally simplies into G

2 1 m1 1 m2 1 m3 z2 = b1b + 11 m1=0 m2=0 m3=0 3m1+m2+m3b + 2m1+m2+m3

a m1 a + 1 m1+m2 a + 2 m1+m2+m3 z m1+m2+m3 . a + 1 m1 a + 2 m1+m2 m 1! m 2! m 3!

A7

Similar calculations can be performed to obtain Gn for n 2.

APPENDIX B: REDUCTION FORMULAS FOR n

In this appendix, we provide the reduction formulas for the hypergeometric functions n dened by Eq. 34, in the case where b1 is set to zero. Starting from the n = 1 case, we see that in the double series expansion 19, only the m1 = 0 term survives in the summation, so that 1

a1, a20, b2 c 1 d 1, d 2

; x 1, x 2 = 2F 2 a 2, b 2 d 1, d 2

; x2 .

B1

Similarly from Eq. 33 for the n = 2 case and from Eq. 34 for the general n 1 case, one easily nds 2

a1, a2, a30, b2, b3 c 1, c 2 d 1, d 2

; x 1, x 2, x 3 = 1

a1, a2, . . . , an+10, b2, . . . , bn+1

= n1
1

c 1, . . . , c n d 1, d 2

a2, . . . , an+1b2, . . . , bn+1 c 2, . . . , c n d 1, d 2

a 2, a 3 b 2, b 3 c 2 d 1, d 2

; x1, x2, . . . , xn+1

; x2, . . . , xn+1 .

; x 2, x 3 ,

B2

B3

A. Erdelyi, W. Magnus, F. Oberhettinger, and F. G. Tricomi, Higher Trascendental Functions McGraw-Hill, New York, 1953, Vols. IIII. 2 L. J. Slater, Conuent Hypergeometric Functions Cambridge University Press, London, 1960. 3 H. Buchholz, The Conuent Hypergeometric Function Springer-Verlag, Berlin, 1969. 4 L. D. Landau and E. M. Lifshitz, Quantum Mechanics: Non-Relativistic Theory Pergamon, Oxford, 1965. 5 C. H. Greene, A. R. P. Rau, and U. Fano, Phys. Rev. A 26, 2441 1982. 6 A. Dzieciol, S. Yngve, and P. O. Frman, J. Math. Phys. 40, 6145 1999. 7 C. R. Garibotti, G. Gasaneo, and F. D. Colavecchia, Phys. Rev. A 62, 022710 2000. 8 L. U. Ancarani and M. C. Chidichimo, J. Phys. B 37, 4339 2004. 9 H. van Haeringen, Charged Particle Interactions Coulomb, Leyden, 1985. 10 Y. E. Kim and A. L. Zubarev, Phys. Rev. A 56, 521 1997. 11 R. Szmytkoswki, J. Phys. A 31, 4963 1998. 12 G. Gasaneo and F. D. Colavecchia, J. Phys. A 36, 8443 2003. 13 B. H. Bransden and C. J. Joachain, Physics of Atoms and Molecules, 2nd ed. Prentice Hall, Englewood Cliffs, NJ, 2003, Chap. 13. 14 M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions Dover, New York, 1972. 15 J. Abad and J. Sesma, Comput. Phys. Commun. 156, 13 2003. 16 A. W. Babister, Transcendental Functions Satisfying Nonhomogeneous Linear Differential Equations Macmillan, New York, 1967. 17 H. M. Srivastava and H. L. Manocha, A Treatise on Generating Functions Ellis Horwood, Chichester, 1978.

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

063508-16
18 19

L. U. Ancarani and G. Gasaneo

J. Math. Phys. 49, 063508 2008

http://functions.wolfram.com/07.27.03.0119.01 P. Appell and J. Kamp de Feriet, Funtions Hypergomtriques et Hypershriques; Polynomes dHermie GauthierVillars, Paris, 1926. 20 I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products Academic, New York, 1994. 21 http://functions.wolfram.com/07.25.03.0076.01

Downloaded 01 Feb 2013 to 128.210.126.199. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions