Engin

Field Guide to
Special
Functions for
Engineers
Larry C. Andrews
SPIE Field Guides

Volume FG18
John E. Greivenkamp, Series Editor
Bellingham, Washington USA

Library of Congress Cataloging-in-Publication Data
Andrews, Larry C.
Field guide to special functions for engineers / Larry C.
Andrews.
p. cm. (The field guide series; 18)
Includes bibliographical references and index.
ISBN 978-0-8194-8550-2
1. Engineering mathematicsFormulaeHandbooks,
manuals, etc. I. Title.
TA332.A53 2011
620.001051dc22
2011002631
Published by
SPIE
P.O. Box 10
Bellingham, Washington 98227-0010 USA
Phone: +1.360. 676.3290
Fax: +1.360.647.1445
Email: books@spie.org
Web: http://spie.org
Copyright 2011 Society of Photo-Optical Instrumentation

Engineers (SPIE)
All rights reserved. No part of this publication may be repro-

duced or distributed in any form or by any means without writ-
ten permission of the publisher.
The content of this book reflects the work and thought of the
author. Every effort has been made to publish reliable and ac-
curate information herein, but the publisher is not responsible
for the validity of the information or for any outcomes resulting
from reliance thereon.
Printed in the United States of America.
Introduction to the Series
Welcome to the SPIE Field Guidesa series of publications

written directly for the practicing engineer or scientist.
Many textbooks and professional reference books cover
optical principles and techniques in depth. The aim of
the SPIE Field Guides is to distill this information,
providing readers with a handy desk or briefcase reference
that provides basic, essential information about optical
principles, techniques, or phenomena, including definitions and
descriptions, key equations, illustrations, application examples,
design considerations, and additional resources. A significant
effort will be made to provide a consistent notation and style
between volumes in the series.
Each SPIE Field Guide addresses a major field of optical

science and technology. The concept of these Field Guides is a
format-intensive presentation based on figures and equations
supplemented by concise explanations. In most cases, this
modular approach places a single topic on a page, and provides
full coverage of that topic on that page. Highlights, insights,
and rules of thumb are displayed in sidebars to the main
text. The appendices at the end of each Field Guide provide
additional information such as related material outside the
main scope of the volume, key mathematical relationships,
and alternative methods. While complete in their coverage, the
concise presentation may not be appropriate for those new to
the field.
The SPIE Field Guides are intended to be living documents.

The modular page-based presentation format allows them
to be easily updated and expanded. We are interested in
your suggestions for new Field Guide topics as well as what
material should be added to an individual volume to make
these Field Guides more useful to you. Please contact us at
fieldguides@SPIE.org.
John E. Greivenkamp, Series Editor

College of Optical Sciences
The University of Arizona
Field Guide to Special Functions

The Field Guide Series
Field Guide to Geometrical Optics, John E. Greivenkamp

(FG01)
Field Guide to Atmospheric Optics, Larry C. Andrews (FG02)
Field Guide to Adaptive Optics, Robert K. Tyson & Benjamin W.

Frazier (FG03)
Field Guide to Visual and Ophthalmic Optics, Jim Schwiegerling

(FG04)
Field Guide to Polarization, Edward Collett (FG05)
Field Guide to Optical Lithography, Chris A. Mack (FG06)
Field Guide to Optical Thin Films, Ronald R. Willey (FG07)
Field Guide to Spectroscopy, David W. Ball (FG08)
Field Guide to Infrared Systems, Arnold Daniels (FG09)
Field Guide to Interferometric Optical Testing, Eric P. Goodwin

& James C. Wyant (FG10)
Field Guide to Illumination, Angelo V. Arecchi; Tahar Messadi;

R. John Koshel (FG11)
Field Guide to Lasers, Rdiger Paschotta (FG12)
Field Guide to Microscopy, Tomasz S. Tkaczyk (FG13)
Field Guide to Laser Pulse Generation, Rdiger Paschotta

(FG14)
Field Guide to Infrared Systems, Detectors, and FPAs, Second

Edition, Arnold Daniels (FG15)
Field Guide to Laser Fiber Technology, Rdiger Paschotta

(FG16)
Field Guide to Wave Optics, Dan Smith (FG17)
Field Guide to Special Functions for Engineers, Larry C.

Andrews (FG18)

Most of the material chosen for this Field Guide is condensed

from two textbooks: Special Functions of Mathematics for
Engineers by L. C. Andrews and Mathematical Techniques for
Engineers and Scientists by L. C. Andrews and R. L. Phillips.
Both books are SPIE Press publications.
Many modern engineering and physics problems demand a

thorough knowledge of mathematical techniques. In particular,
it is important to recognize the various special functions
(beyond the elementary functions) that may arise in practice
as a solution to a differential equation or as a solution to some
integral. It also helps to have a good understanding of the
functions basic properties. The functions treated in this Guide
are among the most important for engineers and scientists.
They commonly occur in problems involving electro-optics,
electromagnetic theory, wave propagation, heat conduction,
quantum mechanics, probability theory, and electric circuit
theory, among many other areas of application.
Because of the close association of power series and improper

integrals with special functions, a brief review of these
important topics is included in this guide. Useful engineering
functions like the step function, rectangle function, and delta
(impulse) function are also introduced.
Unfortunately, notation for various engineering and special

functions is not consistent among disciplines. Also, some special
functions have more than one definition depending on the
area of application. For these reasons, the reader is advised
to be careful when using more than one reference source. The
notation for the special functions adopted in this Field Guide is
that which the author considers most widely used in practice.
Larry C. Andrews
Professor Emeritus, University of Central Florida

vii
Table of Contents
Glossary of Symbols and Notation x
Engineering Functions 1
Step and Signum (Sign) Functions 2
Rectangle and Triangle Functions 3
Sinc and Gaussian Functions 4
Delta Function 5
Delta Function Example 6
Comb Function 7
Infinite Series and Improper Integrals 8

Series of Constants 9
Operations with Series 10
Factorials and Binomial Coefficients 11
Factorials and Binomial Coefficients Example 12
Power Series 13
Operations with Power Series 14
Power Series Example 15
Improper Integrals 16
Asymptotic Series for Small Arguments 17
Asymptotic Series for Large Arguments 18
Asymptotic Series Example 19
Gamma Functions 20
Integral Representations of the Gamma Function 21
Gamma Function Identities 22
Incomplete Gamma Functions 23
Incomplete Gamma Function Identities 24
Gamma Function Example 25
Beta Function 26
Gamma and Beta Function Example 27
Digamma (Psi) and Polygamma Functions 28
Asymptotic Series 29
Bernoulli Numbers and Polynomials 30
Riemann Zeta Function 31
Other Functions Defined by Integrals 32

Error Functions 33
Fresnel Integrals 34
Exponential and Logarithmic Integrals 35
Sine and Cosine Integrals 36
Elliptic Integrals 37

viii
Table of Contents
Elliptic Functions 38
Cumulative Distribution Function Example 39
Orthogonal Polynomials 40
Legendre Polynomials 41
Legendre Polynomial Identities 42
Legendre Functions of the Second Kind 43
Associated Legendre Functions 44
Spherical Harmonics 45
Hermite Polynomials 46
Hermite Polynomial Identities 47
Hermite Polynomial Example 48
Laguerre Polynomials 49
Laguerre Polynomial Identities 50
Associated Laguerre Polynomials 51
Chebyshev Polynomials 52
Chebyshev Polynomial Identities 53
Gegenbauer Polynomials 54
Jacobi Polynomials 55
Bessel Functions 56
Bessel Functions of the First Kind 57
Properties of Bessel Functions of the First Kind 58
Bessel Functions of the Second Kind 59
Properties of Bessel Functions of the Second Kind 60
Modified Bessel Functions of the First Kind 61
Properties of the Modified Bessel Functions of the First
Kind 62
Modified Bessel Functions of the Second Kind 63
Properties of the Modified Bessel Functions of the Second
Kind 64
Spherical Bessel Functions 65
Properties of the Spherical Bessel Functions 66
Modified Spherical Bessel Functions 67
Hankel Functions 68
Struve Functions 69
Kelvins Functions 70
Airy Functions 71
Other Bessel Functions 72
Differential Equation Example 73
Bessel Function Example 74

ix
Table of Contents
Orthogonal Series 75
Fourier Trigonometric Series 76
Fourier Trigonometric Series: General Intervals 77
Exponential Fourier Series 78
Generalized Fourier Series 79
Fourier Series Example 80
Legendre Series 81
Hermite and Laguerre Series 82
Bessel Series 83
Bessel Series Example 84
Hypergeometric-Type Functions 85
Pochhammer Symbol 86
Hypergeometric Function 87
Hypergeometric Function Identities 88
Hypergeometric Function Example 89
Confluent Hypergeometric Functions 90
Confluent Hypergeometric Function Identities 91
Confluent Hypergeometric Function Example 92
Generalized Hypergeometric Functions 93
Relations of p F q to Other Functions 94
Meijer G Function 95
Properties of the Meijer G Function 96
Relation of the G Function to Other Functions 97
MacRobert E Function 98
Meijer G Example 99
Bibliography 101
Index 102

x
Glossary of Symbols and Notation
(a)n Pochhammer symbol

Ai(x) Airy function of the first kind
ber p (x), bei p (x) Kelvins functions
Bi(x) Airy function of the second kind
B(x, y) Beta function
Bn Bernoulli numbers
B n (x) Bernoulli polynomials
cn u Elliptic function
comb(x) Comb function
C(x) Fresnel cosine integral
Ci(x) Cosine integral
Cn (x) Gegenbauer polynomial
dn u Elliptic function
DE Differential equation
(x) Delta or impulse function
erf(x), erfc(x) Error functions
E(a 1 , . . . , a p ; c 1 , . . . , c q ; x) MacRobert E function
E(m, ) Elliptic integral of the second
kind
Ei(x), E 1 (x) Exponential integrals
E p (x) Weber function
F(m, ) Elliptic integral of the first kind
1 F1 (a; c; x) Confluent hypergeometric
function of the first kind
2 F1 (a, b; c; x) Hypergeometric function
p Fq Generalized hypergeometric
function
Gaus(x) Gaussian function
a p

m,n
G p,q x Meijer G function
cq
Eulers constant
(a, x), (a, x) Incomplete gamma functions

xi
(x) Gamma function

h(1) (2)
p (x), h p (x) Hankel spherical Bessel functions
H n (x) Hermite polynomial
H (1) (2)
p (x), H p (x) Hankel functions
H p (x) Struve function
i n (x) Modified spherical Bessel function of the
first kind
I p (x) Modified Bessel function of the first kind
j n (x) Spherical Bessel function of the first kind
Ji p (x) Integral Bessel function
J p (x) Bessel function of the first kind
J p (x) Anger function
ker p (x), kei p (x) Kelvins functions
k n (x) Modified spherical Bessel function of the
second kind
K p (x) Modified Bessel function of the second
kind
li(x) Logarithmic integral
L n (x) Laguerre polynomial
L(a)
n (x) Associated Laguerre polynomial
L p (x) Modified Struve function
(x) Triangle function
M k,m (x) Whittaker function of the first kind
n! Factorial function
P n (x) Legendre polynomial
P nm (x) Associated Legendre function of the first
kind
(a,b)
P n (x) Jacobi polynomial
(x) Rectangle function
(m, , a) Elliptic integral of the third kind
Q n (x) Legendre function of the second kind
Q nm (x) Associated Legendre function of the
second kind
rect(x) Rectangle function
S(x) Fresnel sine integral

xii
Si(x), si(x) Sine integral

sgn(x) Signum (sign) function
sinc(x) Sinc function
step(x) Step function
tri(x) Triangle function
T n (x) Chebyshev polynomial of the first kind
U(x) Step function
U(a; c; x) Confluent hypergeometric function of the
second kind
Un (x) Chebyshev polynomial of the second kind
Wk,m (x) Whittaker function of the second kind
yn (x) Spherical Bessel function of the second kind
Y p (x) Bessel function of the second kind
Ynm (, ) Spherical harmonic
(x) Digamma (psi) function
(m) (x) Polygamma function
(x) Riemann zeta function
. . . is asymptotic to
|| Absolute value

a
Binomial coefficient
n

1
Engineering Functions
In the solution of engineering problems it is often helpful to
employ special notation to identify certain functions that may
arise over and over, particularly those that must be prescribed
in a piecewise manner. Some authors use a single letter to
denote these functions and others use some sort of abbreviation.
However, because the same letters or abbreviations are not
consistently used by all authors, some care must be exercised
when using various reference sources. Although some of
these functions are also defined in two or three dimensions,
the following discussion is limited to functions in only one
dimension.

2 Engineering Functions
Step and Signum (Sign) Functions
The unit step function was first introduced by the English

engineer Oliver Heaviside (18501925) in circuit analysis
problems to handle finite jump discontinuities, acting like a
switch. This function is defined by
step(x)
0, x < 0
U(x) = step(x) = 1/2, x = 0 1
1, x > 0,
x

where the value 1/2 at x = 0 is

optional. The value at x = 0 is often either omitted or taken to
be zero or one.
When the jump takes place at an arbitrary point x = a, the step

function is U(x a), or step(x a), and so on. A more general
version of the step function is given by U[(x a)/b], in which the
jump occurs at x = a and the constant b = 1 permits reflection
about the line x = a by selecting b = 1.
The signum or sign function, defined by

1, x < 0
sgn(x) = 0, x = 0
1, x > 0,

is used to reverse the polarity

of another function at some point. It is directly related to the
step function U(x) above by
sgn(x) = 2U(x) 1.

Rectangle and Triangle Functions
One of the most useful engineering functions is the rectangle

function, defined by

1, | x| < 1/2
rect(x) = (x) =
0, | x| > 1/2.
In some instances it is called

the window function and is
directly related to the step function by
rect(x) = U(x + 1/2) U(x 1/2).
In the time domain this function can be used to represent a

gating operation in an electrical circuit, for example. Like the
step function, this function is sometimes defined by the value
1/2 at | x| = 1/2. More generally, the rectangle function can be
written as rect[(x a)/b], in which case it is centered at x = a
and | b| is equal to both its width and area.
The triangle function,

1 | x|, | x| < 1
tri(x) = (x) =
0, | x| > 1.
is sometimes used as a model

for the correlation function of
a stochastic variable. Written as tri[(x a)/b], the width of the
base is 2| b|, its area is | b|, and it is centered at x = a.

Sinc and Gaussian Functions
The sinc function is defined by
sin x
sinc(x) = .
x
This function consistently shows

up in studies of linear systems.
The factor in the sine function is
introduced so that the zeros of the sinc function are at integer
values of x. The sinc function is directly related to the Fourier
transform of the rectangle function, and its square is directly
related to the Fourier transform of the triangle function. This
function can be defined more generally by sinc[(x a)/b], in
which the zeros occur at a nb, n = 1, 2, 3, . . ., and it is centered
at x = a.
The Gaussian function, described as
2
Gaus(x) = e x ,
is encountered in the field of

statistics and is a common model
used to characterize the output of
a laser operating in the lowest-
order mode TEM00 . To center this
function at x = a, it is written Gaus[(x a)/b], which still has a
height of unity and an area equal to | b|.

Delta Function
The Dirac delta function, or simply the delta function, is

named after Paul A. M. Dirac (19021984) who was awarded
a Nobel prize (with E. Schrdinger) in 1933 for his work in
quantum mechanics. The delta function is now widely accepted
as a standard tool in engineering analysis. In applications
involving linear systems it is customary to call it the impulse
function. In optics, the delta function is often used as a model
for a point source.
The defining properties of the delta function are given by
(x a) = 0, x 6= a,
Z x2
f (a), x1 < x < x2
f (x)(x a) dx =
x1 0, otherwise.
The second expression above is often

called the sifting property of this
function. Clearly, this is not a function
in the usual sense of the word. Nonetheless, it is common
practice to assign certain function properties to this symbol, the
primary ones being the following:
(a) ( x) = (x),
1
(b) (ax) = (x),
| a|
(c) f (x)(x a) = f (a)(x a),
d
(d) U(x a) = (x a).
dx
It is sometimes useful to introduce the notion of a derivative

of the delta function, defined by
Z
0 (x) f (x) dx = f 0 (0).


Delta Function Example
Show formally that
d
U(x a) = (x a),
dx
where U(x a) is the step function.
Solution:
The delta function has meaning only when it appears

under an integral. Hence, perform integration by parts on the
derivative of the step function to obtain

d
Z

U(x a) f (x) dx = f (x)U(x a)
dx
Z
U(x a) f 0 (x) dx,

where f (x) is any reasonably smooth function for which
f () = f () = 0.
Then, there is
Z d
Z
U(x a) f (x) dx = f 0 (x) dx = f (a).
dx a
However, from definition of the delta function, there is

Z
(x a) f (x) dx = f (a).

Comparing integrals on the left of these last two expressions

leads to the conclusion that
d
U(x a) = (x a).
dx

Comb Function
The comb function is an infinite series of delta functions that

is useful in representing sampled data. It is typically defined
by

X
comb(x) = (x n).
n=
Among other uses, both the delta

function and the comb function
are very expedient in the spec-
tral representations of periodic functions that commonly appear
in the Fourier analysis of various power and energy signals. To
place the delta functions above at a nb, n = 1, 2, 3, . . ., rather
than at the integers, the notation comb[(x a)/b] can be intro-
duced.
The comb function, also known as an impulse train

or sampling function in electrical engineering, is defined
in terms of the Dirac delta function. Neither one of
these functions is a function in the usual sense of the
word. In mathematics, they represent what are commonly
called generalized functions. Following more rigorous
lines, generalized functions can be defined as a limit of an
infinite sequence of well-behaved, or standard, functions.

8
Infinite Series and Improper Integrals

Infinite series are important in almost all areas of
mathematics and engineering. In addition to numerous other
uses, they are used to define certain functions and to calculate
accurate numerical estimates of the values of these functions.
Most of the infinite series that we encounter in practice are
known as power series.
In calculus, the primary problem is deciding whether a given

series converges or diverges. In practice, however, the more
crucial problem may actually be summing the series. If a
convergent series converges too slowly, the series may be
worthless for computational purposes. On the other hand, the
first few terms of a divergent asymptotic series in some cases
may give excellent results.
Improper integrals are used in much the same fashion as

infinite series, and, in fact, their basic theory closely parallels
that of infinite series. They are often classified as belonging to
improper integrals of the first kind or improper integrals
of the second kind.

Series of Constants
If a number S n can be associated with each positive integer

n, then the ordered arrangement S 1 , S 2 , . . . , S n , . . . is called
an infinite sequence. If limn S n = S (finite), the infinite
sequence is said to converge to S; otherwise it diverges.

P
Infinite series: u 1 + u 2 + + u k + = uk.
k=1
If S n S as n , where S n is the sequence defined by

S n = u 1 + u 2 + . . . + u n , then the series converges to S; otherwise
it diverges. The letter k used in the summation is called an
index; any other letter like n can also be used.
Geometric series: The special series

1 + r + r2 + + r k + = rk
X
k=0
is an example series where r is the common ratio. It has been

shown that
nX1 1 rn 1
rk = rk =
X
, r 6= 1 and , | r | < 1.
k=0 1r k=0 1r
Harmonic series: Although it diverges, the particular series

X 1
n=1 n
is quite important in theoretical work.
Alternating series: This series is of the form

(1)n u n , u n > 0, n = 1, 2, 3, . . . .
X
n=0
Alternating harmonic series: The special alternating series

defined by
X (1) n1
n=1 n
differs from the harmonic series above in that this series

converges.

10 Infinite Series and Improper Integrals
Operations with Series
The series u n is said to converge absolutely if the

related series | u n | converges.
If u n converges but the related series | u n | diverges, then
u n is said to converge conditionally.
In some applications the need arises to combine series by

operations like addition and subtraction, and multiplication.
The following rules of algebra apply:
1. The sum of an absolutely convergent series is independent

of the order in which the terms are summed.
2. Two absolutely convergent series may be added termwise,
and the resulting series will converge absolutely.
3. Two absolutely convergent series may by multiplied, and
the resulting series will also converge absolutely.
Forming the product of two series leads to double infinite

series of the form

X
X X
X n
X
X
am bk = A m,k = A nk,k .
m=0 k=0 m=0 k=0 n=0 k=0
Note that the result on the right is a single infinite series of

finite sums rather than a product of two infinite series. Another
product of series leads to the identity
X
X n
X
X [n/2]
X X
A m,k = A nk,k = A n2k,k ,
m=0 k=0 n=0 k=0 n=0 k=0
where
hni
n/2, n even
=
2 (n 1)/2, n odd.

Factorials and Binomial Coefficients
A product of consecutive positive integers from 1 to n is called

factorial n and is denoted by
n! = 1 2 3 n, n = 1, 2, 3, . . . .
The fundamental properties of the factorial are
0! = 1
n! = n(n 1)!, n = 1, 2, 3, . . . .
The binomial formula, or finite binomial series, is defined

by
n
n n k k
(a + b)n =
X
a b .
k=0
k
The special symbol
n!

n
= , n = 0, 1, 2, . . . , k = 0, 1, 2, . . . , n,
k k! (n k)!
is called the binomial coefficient. Basic identities for the

binomial coefficient include the following:

n n
(a) = =1
0 n

n n
(b) = =n
1 n1

n n
(c) =
k nk
r(r 1) (r k + 1)

r r
(d) = 1, = , k = 1, 2, 3, . . .
0 k k!

r r+k1
(e) = (1)k
k k

Factorials and Binomial Coefficients Example
Use the definition of the binomial coefficient and properties

of factorials to show that
(1)n (2n)!

1/2
= 2n , n = 0, 1, 2, . . . .
n 2 (n!)2
Solution: From the definition, one obtains
(1/2)(3/2) (1/2 n)

1/2
=
n n!
(1)n 1 3 5 (2n 1)
= .
2n n!
Multiplying the numerator and denomination on the right by

the product 2 4 6 (2n) and rearranging the numerator terms,
one finds that
(1)n 1 2 3 4 5 6 (2n 1) (2n)

1/2
=
n 2n 2 4 6 (2n)n!
(1)n (2n)!
= ,
2n 2n 1 2 3 n n!
which can be written as
(1)n (2n)!

1/2
= 2n , n = 0, 1, 2, . . . .
n 2 (n!)2
Binomial coefficients appear as coefficients in the binomial

expansion of (a + b)n (called the binomial theorem). In
addition to arising in algebraic expressions, the family of
binomial coefficients is commonly found in combinatoric
expressions.

Power Series
A power series is an infinite series of the form

c 0 + c 1 (x a) + c 2 (x a)2 + + c n (x a)n + = c n (x a)n ,
X
n=0
where c 0 , c 1 , . . . , c n , . . . are called coefficients of the series and

a is the center of the series.
Assigned to every power series is a number R, 0 R < ,

called the radius of convergence, with the property
that the series converges absolutely for | x a| < R and
diverges for | x a| > R. If R > 0, then for every such that
0 < R, it can be shown that the power series converges
uniformly on the interval | x a| .
A binomial series is defined by

n
(1 + x) =
X
x , 1 < x < 1.
n=0
n
If is a positive integer m, then the binomial series truncates

at n = m. Other useful power series include:
1
(1)n x n ,
X
= 1 < x < 1
1+ x n=0
xn
ex =
X
, < x <
n=0 n!
x2n+1
(1)n
X
sin x = , < x <
n=0 (2n + 1)!
x2n
(1)n
X
cos x = , < x <
n=0 (2n)!
(1) n1
(x 1)n ,
X
ln x = 0<x2
n=1 n

Operations with Power Series
A power series always converges to a continuous

function. Moreover, power series can always be integrated
and differentiated termwise. That is, if

c n xn ,
X
f (x) =
n=0
then
x n+1
Z Z
x n dx =
X X
f (x) dx = cn cn +C
n=0 n=0 n+1
d n X
0
c n nx n1 ,
X
f (x) = cn x =
n=0 dx n=1
where C is an arbitrary constant.
Two polynomials are equal if and only if their corresponding

terms are identical. The same applies to infinite series.
n n
Thus, if
P P
n=0 a n x and n=0 b n x both have nonzero radii of
convergence, and

a n xn = b n xn ,
X X
n=0 n=0
wherever the two series converge, then a n = b n , n = 0, 1, 2, . . ..
Two or more power series can be combined through termwise

addition of the terms and through multiplication, the latter
known as the Cauchy product:
1. Sum:

a n xn + b n xn = (a n + b n )x n
X X X
n=0 n=0 n=0
2. Cauchy product:
! !
X n X n
n n n
a k b n k x n
X X X X
an x bn x = a n k b k x =
n=0 n=0 n=0 k=0 n=0 k=0

Power Series Example
Use the identity 2 sin a cos a = sin 2a and the Cauchy product to
deduce that
n
2n + 1
= 22n ,
X
n = 0, 1, 2, . . . .
k=0
2k + 1
Solution:
Recall the power series expansions
(1)n 2n+1 (1) n

a2n ,
X X
sin a = a , cos a =
n=0 (2n + 1)! n=0 (2n)!
which, based on the Cauchy product, lead to
2 sin a cos a = sin 2a

(1)n 2n+1 X (1) n (1)n
a2n = (2a)2n+1
X X
2 a
n=0 (2n + 1)! n=0 (2n)! n=0 (2n + 1)!
n
X (1)k (1)nk
a2k+1 a2(nk) =
X
2
n=0 k=0 (2k + 1)!(2n 2k)!

X (1)n 2n+1 Xn (2n + 1)!
2 a =.
n=0 (2n + 1)! k=0 (2k + 1)!(2n 2k)!
Next, observe that
(2n + 1)!

2n + 1
= ,
2k + 1 (2k + 1)!(2n 2k)!
which permits
(1)n 2n+1 Xn X (1)n 22n 2n+1

X 2n + 1
a = a .
n=0 (2n + 1)! k=0
2k + 1 n=0 (2n + 1)!
Last, by comparing like terms, one can deduce that
n
2n + 1
= 22n ,
X
n = 0, 1, 2, . . . .
k=0
2k + 1

Improper Integrals
If the Riemann integral
Z b
f (t) dt
a
exists, it is said to be a proper integral. This generally means

that the interval of integration is finite and f (t) is bounded on
this interval. Integrals that have an infinite limit of integration
or an unbounded integrand between limits of integration are
called improper integrals.
An improper integral of the first kind occurs when one or both

limits of integration are infinite but the integrand is bounded
on a t < , < t < b, or < t < . If the following
limits exist, the given improper integral is said to converge, and
otherwise diverge:
Z Z b
f (t) dt = lim f (t) dt
a b a
Z b Z b
f (t) dt = lim f (t) dt
a a
Z Z c Z b
f (t) dt = lim f (t) dt + lim f (t) dt, < c < .
a a b c
R b f (t) is unbounded on the interval a t b, then the integral

If
a f (t) dt is an improper integral of the second kind. If f (t) is
unbounded at only one point t = c, a c b, and the following
limit exists, the improper integral is said to converge, and
otherwise diverge:
Z b Z c Z b
f (t) dt = lim f (t) dt + lim f (t) dt.
a 0+ a 0+ c+
R R
a f (t) dt is said to converge absolutely if
| f (t)| dt converges. a
R R
a f (t) dt is Rsaid to converge conditionally if a f (t) dt

converges but a | f (t)| dt diverges.

Asymptotic Series for Small Arguments
In computational work, it is common to represent a given

function f (x) by some simpler function, say g(x), that accurately
describes numerical values of f (x) in the vicinity of some point
x = a. If it is true that
f (x)
lim = 1, then
x a g(x)
f (x) g(x) as x a,
where the symbol means is asymptotic to. For the special

case x 0, a series representation may be sought where f (x)
P n
n=0 c n x , x 0, from which we can deduce small argument
formulas like
f (x) c 0 , x0
f (x) c 0 + c 1 x, x 0,
and so on. In general, the above series representation is said to

be an asymptotic series for f (x) as x 0 if and only if
| f (x) S n (x)|
lim = 0, where
x0 | x |n
n
c k xk .
X
S n (x) =
k=0
By the above limit, | f (x) S n (x)| must approach zero more

rapidly than | x |n as x 0.
Asymptotic relations are normally obtained from the Maclaurin

power series expansion of f (x). However, the asymptotic series
does not have to be a power series nor is there any requirement
for the series to converge! This means that if the series diverges,
the accuracy of computation for fixed x is tied to the number n.

Asymptotic Series for Large Arguments
Given a function f (x), it may be possible to find useful

approximations to it for large values of x by seeking series
representations of the form
a
X n
f (x) n
, x ,
n=0 x
called an asymptotic series or semiconvergent series. More

precisely, this is said to be an asymptotic series if and only if, for
every n,
lim x n | f (x) S n (x)| = 0,

x
where
n a
X k
S n (x) = .
k=0 xk
Asymptotic series of this nature are intriguing in that they often

diverge for all values of x. Nonetheless, they may still be useful
for numerical computation provided one does not take too many
terms. As a rule of thumb, one should stop just before the terms
begin to increase. The error incurred in most cases turns out to
be less than the first term omitted in the approximation.
Not all functions have an asymptotic series. For example,

neither e x nor sin x has such a series. If f (x) has no asymptotic
series, there may exist a suitable function h(x) such that an
asymptotic series exists for f (x)/h(x). In this case, it is written
as
f (x) X a
n
, x ,
h(x) n=0 x n
or, equivalently,
a
X n
f (x) h(x) n
, x .
n=0 x

Asymptotic Series Example
Find an asymptotic series for the function
e xt
Z
f (x) = dt, x > 0.
0 1 + t2
Solution:
First make the change of variable s = xt, which yields
1
1 s2
Z
f (x) = es 1 + 2 ds.
x 0 x
For x > s, there is the binomial series
1 2n
s2 X 1 s

1+ 2 = ,
x n=0
n x2n
which, together with termwise integration, leads to

1
Z
1
es s2n ds,
X
f (x) x .
n=0
n x2n+1 0
Through repeated integration by parts, one finds that

Z
es s2n ds = (2n)!, n = 0, 1, 2, . . . .
0
Then, by use of the identity

1 n
= (1)n = (1)n ,
n n
one can deduce that
(2n)!
(1)n
X
f (x) , x .
n=0 x2n+1

20
Gamma Functions
In the eighteenth century, L. Euler (17071783) concerned
himself with the problem of interpolating between the numbers
Z
n! = e t t n dt, n = 0, 1, 2, . . . ,
0
with noninteger values of n. This problem led Euler in 1729 to

what is now called the gamma function, a generalization of
the factorial function.
Because it is a generalization of n!, the gamma function has

been examined over the years as a means of generalizing
certain functions, operations, etc., that are commonly defined
in terms of factorials. In addition, the gamma function is useful
in the evaluation of many nonelementary integrals and in the
definition of other special functions like Bessel functions and
hypergeometric functions.
Another function useful in various applications is the related

beta function, often called the Eulerian integral of the first
kind. In 1771, some 43 years after discovering the gamma
function, Euler found that the beta function was actually a
particular combination of gamma functions. The logarithmic
derivative of the gamma function leads to the digamma (or
psi) function, while further differentiations produce the family
of polygamma functions, all of which are related to the zeta
function of G. Riemann (18261866).

Gamma Functions 21
Integral Representations of the Gamma Function
The gamma function most often arises in practice as the

improper integral
Z
(x) = e t t x1 dt, x > 0.
0
Other integral representations are given by

Z 2
(x) = 2 e t t2x1 dt, x>0
0
1 x1
1
Z
(x) = ln dt, x > 0.
0 t
However, the gamma function has several equivalent defini-

tions that do not involve an integral representation. One defi-
nition credited to C. Gauss (17771855) is
n!n x
(x) = lim , x 6= 0, 1, 2, . . . ,
n x(x + 1)(x + 2) (x + n)
which is not limited to only positive values of x. The gamma

function, however, is not defined at zero and the negative
integers, but exists for all other values of x.
(x)
-4 -2 2 4 x
-2
-4
-6

22 Gamma Functions
Gamma Function Identities
There are numerous identities associated with the gamma

function, some of which are given below.
(G1) : (1) = 1
p
(G2) : (1/2) =
(G3) : (x + 1) = x(x) (Recurrence formula)
(G4) : (n + 1) = n!, n = 0, 1, 2, . . .
1 (2n)! p

(G5) : n+ = n , n = 0, 1, 2, . . .
2 2 n!
1
(G6) : = 0, n = 0, 1, 2, . . .
( n)

(G7) : (x)(1 x) =
sin x
1 1

(G8) : +x x =
2 2 cos x
p
(G9) : (2x) = 22x1 (x)(x + 1/2)
(Duplication formula)
(a + 1)

a
(G10) : = , n = 0, 1, 2, . . .
n n!(a n + 1)
(Binomial coefficient)
k
(1) n!
(k n) (n k)! , (k, n non-negative integers)

(G11) : =
( n)
0kn
0, k > n

1 x x/n
(G12) : = xe x
n=1 1 + e
(x) n
(x + k)
(G13) : (x) = ,
x(x + 1)(x + 2) . . . (x + k 1)
x > k, k = 1, 2, 3, . . .
p
(G14) : (x + 1) 2 x x x e x , x (Stirlings formula)

Gamma Functions 23
Incomplete Gamma Functions
A function closely related to the gamma function is

Z x
(a, x) = e t ta1 dt, a > 0,
0
called the incomplete gamma function. It is customary to

also introduce the companion function
Z
(a, x) = e t ta1 dt, a > 0,
x
which is known as the complementary incomplete gamma

function. Thus, it follows that
(a, x) + (a, x) = (a).
These functions commonly arise in probability theory, particu-

larly in applications involving the gamma and chi-square dis-
tributions.
Although the gamma function doesnt have a series representa-

tion, the incomplete gamma functions have series represen-
tations given by
(1) n x n
(a, x) = xa
X
, a>0
n=0 n!(n + a)
(1) n x n
(a, x) = (a) xa
X
, a > 0.
n=0 n!(n + a)

24 Gamma Functions
Incomplete Gamma Function Identities
Some identities satisfied by the incomplete gamma func-

tions include the following:
(IG1) : (a + 1, x) = a (a, x) xa e x
(IG2) : (a + 1, x) = a (a, x) + xa e x
d a
x (a, x) = xa1 (a + 1, x)

(IG3) :
dx
d a
x (a, x) = xa1 (a + 1, x)

(IG4) :
dx
1 2 p

(IG5) : , x = erf(x)
2
1 p

(IG6) : , x2 = erfc(x)
2
xa
(IG7) : (a, x) = 1 F1 (a; a + 1; x)
a
(IG8) : (a, x) = e x U(1 a; 1 a; x)
(IG9) : (a, x) e x xa1 , x
As already mentioned, the incomplete gamma functions

are often met in probability and statistics, particularly
in calculating the cumulative distribution function. The
incomplete gamma function of the second kind (a, z), where
z = x + i y is complex, is an entire function whereas (a, z)
has poles at z = 0, 1, 2, . . . .

Gamma Functions 25
Gamma Function Example
Use properties of the gamma function to evaluate the

following integrals:
Z 3
(a) x4 e x dx.
Z 0
2
(b) e2ax x dx.
a
Solution:
(a) Let t = x3 , dt = 3x2 dx, which leads to
Z 3 1
Z 1 5

x4 e x dx = e t t2/3 dt = .
0 3 0 3 3
(b) Note that 2ax x2 = (x a)2 + a2 . By making this

substitution, one can obtain
Z 2 2
Z 2
e2ax x dx = e a e(xa) dx.
a a
The substitution t = x a yields the more familiar form

Z 2 2
Z 2 1 1 a2

e2ax x dx = e a e t dt = e ,
a 0 2 2
where the property

Z 2
(x) = 2 e t t2x1 dt, x>0
0
p
is used. Also, recognizing that (1/2) = , the final result is
Z 2 1 p a2
e2ax x dx = e .
a 2

26 Gamma Functions
Beta Function
A useful function of two variables is the beta function,

expressed as
Z 1
B(x, y) = t x1 (1 t) y1 dt, x > 0, y > 0.
0
Two other integral representations for the beta function are
t x1
Z
B(x, y) = dt
0 (1 + t) x+ y
Z /2
B(x, y) = 2 cos2x1 sin2y1 d ,
0
with the same restrictions on x and y.
The beta function is also directly related to the gamma function

by
(x)(y)
B(x, y) = , x > 0, y > 0.
(x + y)
From any of the above definitions, the symmetry property can

be readily deduced and expressed as
B(x, y) = B(y, x).
Some additional identities include:
(B1) : B(x + 1, y) + B(x, y + 1) = B(x, y)

y y
(B2) : B(x, y + 1) = B(x + 1, y) = B(x, y)
x x+ y
(B3) : B(x, x) = 212x B(x, 1/2)
(x)(y)(z)(w)
(B4) : B(x, y)B(x + y, z)B(x + y + z, w) =
(x + y + z + w)

Gamma Functions 27
Gamma and Beta Function Example
Use properties of the gamma and beta functions to show that

Z cos x
dx = , 0 < p < 1.
0 xp 2(p) cos(p/2)
Solution:
From properties of the gamma function, note that
1 1
Z
= e xt t p1 dt,
x p (p) 0
and, consequently,
Z cos x 1
Z Z
dx = cos x e xt t p1 dt dx
0 xp (p) 0 0
Z Z
1 p1
= t e xt cos x dx dt,
(p) 0 0
where the order of integration is interchanged in the last step.

By evaluating the inside integral, one obtains
Z cos x 1
Z Z
dx = t p1 e xt cos x dx dt
0 xp (p) 0 0
p1
1 t
Z
= dt.
(p) 0 1 + t2
If u = t2 , du = 2tdt, and the beta function is used, the last result

leads to
Z cos x 1

1+ p 1 p

1

1+ p

1 p

dx = B , = ,
0 xp 2(p) 2 2 2(p) 2 2
and, through property (G7) of the gamma function, one can

deduce that
Z cos x
dx = , 0 < p < 1.
0 x p 2(p) cos(p/2)

28 Gamma Functions
Digamma (Psi) and Polygamma Functions
Closely associated with the derivative of the gamma function is

the digamma function, defined by
d 0 (x)
(x) ln (x) = , x 6= 0, 1, 2, . . . .
dx (x)
This function is also called the psi function. A series

representation of this function is given by

X

1 1

(x) = + , x > 0,
n=0 n+1 n+ x
where Eulers constant = (1) = 0 (1) has the integral

representation
Z
= e t ln t dt = 0.577215 . . . .
0
Basic identities of the digamma function are:
1
(DG1) : (x + 1) = (x) +
x
(DG2) : (1 x) (x) = cot x
Xn 1
(DG3) : (n + 1) = + , n = 1, 2, 3, . . . .
k=1 k
Continued derivatives of the gamma function lead to the family

of polygamma functions,
d m+1
(m) (x) = ln (x), m = 1, 2, 3, . . . , or
dx m+1
(m + n)!
(m) (x + 1) = (1)m+1 (1)n (m + n + 1)x n ,
X
n=0 n!
1 < x < 1,
where the function (x) is the Riemann zeta function.

Gamma Functions 29
Asymptotic Series
It is often useful in practice to have asymptotic formulas or

asymptotic series to describe the behavior of the general family
of gamma functions with large arguments. The most important
of these for the gamma function is Stirlings formula:
p
(x + 1) 2 x x x e x , x .
For greater accuracy, one can use the more general result
p 1 1

x x
(x + 1) 2 x x e 1+ + + , x ,
12x 288x2
which is very accurate for all x 1.
For the complementary incomplete gamma function, there is the

asymptotic series
1
(a, x) (a)xa1 e x
X
, a > 0, x .
k=0 (a k)x k
The asymptotic series for the digamma function is given by
1 B
1 X 2n
(x + 1) ln x + , x ,
2x 2 n=1 n x2n
where the constants B n are the Bernoulli numbers, and,

similarly, for the polygamma function, the series is
1 1 B
2n
0 (x + 1)
X
2+ 2n +1
, x .
x 2x n=1 x

30 Gamma Functions
Bernoulli Numbers and Polynomials
Bernoulli numbers were originally introduced by Jacob

Bernoulli (16541705). Among other areas, they arise in the
theory of numbers and are defined by
dn t

Bn = , n = 0, 1, 2, . . . .
dt n t
e 1 t=0
The first few Bernoulli numbers can be expressed as
1 1 1
B0 = 1, B1 = , B2 = , B3 = 0, B4 = ,....
2 6 30
All Bernoulli numbers with odd index, except B1 , are zero. The
even-index numbers are defined in terms of the Riemann zeta
function by
2(2n)!
B2n = (1)n+1 (2n), n = 0, 1, 2, . . . .
(2)2n
The related Bernoulli polynomials B n (x) can be defined by

the generating function relation
te xt
X tn
= B n (x) ,
e t 1 n=0 n!
where B n = B n (0). The first few Bernoulli polynomials are
1 1
B0 (x) = 1, B1 (x) = x , B2 (x) = x2 x + ,
2 6
3 1 1
B3 (x) = x3 x2 + x, B4 (x) = x4 2x3 + x2 ,....
2 2 30
Some basic identities are
(Ber1) : B0n (x) = nB n1 (x), n = 1, 2, 3, . . .

(Ber2) : B n (x + 1) B n (x) = nx n1 , n = 0, 1, 2, . . .
(Ber3) : B n (1 x) = (1)n B n (x), n = 0, 1, 2, . . .
(Ber4) : (1)n B n ( x) = B n (x) + n x n1 , n = 0, 1, 2, . . .

Gamma Functions 31
Riemann Zeta Function
The Riemann zeta function, expressed as
X 1
(x) = x
, x > 1,
n=1 n
is closely related to the logarithm of the gamma function and

to the polygamma functions. Although known to Euler, it was
Riemann in 1859 who established most of its properties that
are now important in the theory of numbers.
An interesting series relation is given by
1
(x)(1 2 x ) =
X
.
n=1 (2n 1) x
The zeta function also has the integral representation
1 t x1
Z
(x) = dt, x > 1,
(x) 0 et 1
and satisfies Riemanns famous formula
21 x x
(1 x) = cos (x)(x).
x 2
Some additional identities involving the zeta function are:
1 1
(Z1) : (0) = ; 0 (0) = ln 2
2 2
(2)2n
(Z2) : (2n) = |B2n |, n = 1, 2, 3, . . .
2(2n)!
(1) n
X
(Z3) : = (n) (Eulers constant)
n=2 n

32
Other Functions Defined by Integrals

In addition to the gamma function, there are numerous other
special functions whose primary definition involves an integral.
The error function derives its name from its importance in
the theory of errors, but it also occurs in probability theory
and in certain heat conduction problems on infinite domains.
The closely related Fresnel integrals are fundamental in
the theory of optics. A special case of the incomplete gamma
function leads to the exponential integral and to the
logarithmic integral. The sine and cosine integrals are
important in Fourier transform theory.
Elliptic integrals first arose in the problem associated with

computing the arc length of an ellipse. Some early results
concerning elliptic integrals were achieved by Euler and
Landen, but essentially the entire theory of these integrals was
developed by A. Legendre (17521833) over a period spanning
40 years. The inverses of the elliptic integrals, called elliptic
functions, were independently introduced in 1827 by C. G. J.
Jacobi (18021859) and N. H. Abel (18021829). Nonetheless,
many of the properties of elliptic functions had already been
developed as early as 1809 by C. Gauss (17771855). Elliptic
functions have the distinction of being doubly periodic, with
one real period and one imaginary period. Among other areas
of application, elliptic functions are important in solving the
pendulum problem.

Error Functions
The error function and complementary error function

have integral representations, defined respectively by
2
Z x 2
erf(x) = p e t dt, < x <
0
2
Z 2
erfc(x) = p e t dt, < x < .
x
Special values are given by
erf(0) = 0; erf() = 1; erfc(0) = 1; erfc() = 0.
The error function also has the power series representation:
2 X (1) n x2n+1
erf(x) = p , < x < .
n=0 n!(2n + 1)
Additional identities include the following:
(E1) : erf( x) = erf(x)

(E2) : erfc(x) = 1 erf(x)
2
e x
(E3) : erfc(x) p , x
x
d 2 2
(E4) : erf(x) = p e x
dx

34 Other Functions Defined by Integrals
Fresnel Integrals
Closely associated with the error function are the Fresnel

integrals:
Z x
1 2

C(x) = cos t dt
0 2
Z x
1 2

S(x) = sin t dt.
0 2
From definition it follows that
C(0) = S(0) = 0.
Also, the derivatives

are
1 2

C 0 (x) = cos x ,
2
1

0 2
S (x) = sin x .
2
In the limit x it
has been shown that
1
C() = S() = .
2
Some basic identities for these functions are
(F1) : C( x) = C(x); S( x) = S(x)

(1) n (/2)2n
x4n+1
X
(F2) : C(x) =
n=0 (2n)!(4n + 1)
(1) n (/2)2n+1
x4n+3
X
(F3) : S(x) =
n=0 (2n + 1)!(4n + 3)
s !
1 i p
(F4) : C(x) iS(x) = p erf x , i = 1
2i 2

Exponential and Logarithmic Integrals
The exponential integral is defined by
x et
Z
Ei(x) = dt, x 6= 0.
t
Another definition often found in practice is
e t
Z
Ei( x) E 1 (x) = dt, x > 0.
x t
The exponential integrals are encountered in antenna theory

and in some astrophysical problems.
A relationship between this function and the incomplete gamma

function is
E 1 (x) = (0, x) = lim [(a) (a, x)] .

a0
Thus, properties of E 1 (x) can be deduced from those of the

incomplete gamma function (a, x).
A series representation for E 1 (x) is given by
X (1) n x n
E 1 (x) = ln x , x > 0,
n=1 n!n
and this function has the asymptotic formulas
e x
E 1 (x) ln x, x 0+ ; E 1 (x) , x .
x
The logarithmic integral is defined by

Z x dt
Li(x) = = E 1 ( ln x), 0 < x < 1.
0 ln t
Thus, the properties of Li(x) are readily deduced from the

properties of E 1 (x).

Sine and Cosine Integrals
Another set of special func-

tions related to the exponen-
tial integral includes the sine
integral and cosine inte-
gral defined, respectively, by
Z x
sin t
Si(x) = dt, x>0
0 t
Z x
cos t
Ci(x) = dt, x > 0.
0 t
In some applications it is con-

venient to introduce another
sine integral defined by
Z sin t
si(x) = dt = Si(x) .
x t 2
Special values include

Si(0) = 0; Si() =
2
Ci(0+ ) = ; Ci() = 0,
and their derivatives lead to
sin x cos x
Si0 (x) = ; Ci0 (x) = .
x x
Finally, their series representations are given by

X (1)n x2n+1
Si(x) =
n=0 (2n + 1)(2n + 1)!
X (1) n x2n
Ci(x) = + ln x + ,
n=0 2n(2n)!
where is Eulers constant.

Elliptic Integrals
There are three classifications of elliptic integrals (called the

elliptic integrals of the first, second, and third kinds),
defined, respectively, by
Z
d
F(m, ) = p , 0<m<1
0 1 m2 sin2
Z p
E(m, ) = 1 m2 sin2 d , 0<m<1
0
Z
d
(m, , a) = p ,
0 (1 + a2 sin2 ) 1 m2 sin2
0 < m < 1, a 6= m, 0.
The parameter is called the amplitude, and m is the

modulus. When = /2, the following are referred to as
complete elliptic integrals:
F(m, /2) = K(m) = K

E(m, /2) = E(m) = E
(m, /2, a) = (m, a) = .
Special values of the complete elliptic integrals include

K(0) = ; E(0) =
2 2
K(1) = ; E(1) = 1.
The series representations for the complete elliptic integrals are
2
X 1/2
K(m) = m2n
2 n=0
n

X 1/2 1/2 2n
E(m) = m .
2 n=0
n n

Elliptic Functions
Elliptic functions represent the inverse of the elliptic

integrals. In general, if u = F(m, ), then three elliptic functions
can be defined by the relations
sn u = sin
cn u = cos
q
dn u = 1 m2 sin2 .
These elliptic functions of Jacobi belong to the class of doubly

periodic functions with one real period and one imaginary
period. Thus, they have characteristics of both circular and
hyperbolic functions.
Special values include
sn (0) = 0; cn (0) = 1; dn (0) = 1.
These functions also satisfy the identities
sn2 u + cn2 u = 1; m2 sn2 u + dn2 u = 1; dn2 u m2 cn2 u = 1 m2 ,
and the derivative relations
d
sn u = cn u dn u
dx
d
cn u = sn u dn u
dx
d
dn u = m2 sn u cn u.
dx
Among others, these functions exhibit the following periodic

relations:
sn(u + 4K) = sn u,
cn(u + 4K) = cn u,
dn(u + 2K) = dn u,
where K is the complete elliptic integral of the first kind, K(m).

Cumulative Distribution Function Example
Given the normal probability distribution
1 2 2
f x (x) = p e(xm) /2 , < x < ,
2
where m is the mean value and 2 is the variance, calculate the

cumulative distribution function.
Solution:
From definition, the cumulative distribution function is
1
Z x 2 /22
F x (x) = p e(um) du
2
1
Z 2 /22 1
Z 2 /22
= p e(um) du p e(um) du.
2 2 x
The first integral on the right is unity (by definition),

p and by
making the change of variable t = (u m)/ 2 in the second
integral on the right, one obtains
1
Z 2
F x (x) = 1 p p e t dt
(x m)/ 2
1 xm

= 1 erfc p ,
2 2
which can also be written as
1 xm

F x (x) = 1 + erf p .
2 2
In electrical engineering applications it is common to express

the cumulative distributionR of a Gaussian variate in terms
2
of the function Q(x) = p1 x e t /2 dt rather than the error
2
function.

40
Orthogonal Polynomials
There are several special polynomial sets with properties that
distinguish them from ordinary polynomials. In particular,
each polynomial set satisfies a three-term recurrence relation
between consecutive polynomials, and each set is a mutually
orthogonal set of functions. A set of functions {n (x)}, n =
0, 1, 2, . . . is said to be orthogonal on an interval a < x < b
with respect to a weighting function r(x) > 0 if it is true for all
members of the set that
Z b
r(x)n (x)k (x) dx = 0, k 6= n.
a
Sets of orthogonal functions play an extremely important role

in applied analysis, primarily because functions belonging to a
very general class can be represented by series of orthogonal
functions, called generalized Fourier series.
A special class of orthogonal functions consists of the sets of

orthogonal polynomials {n (x)}, where n denotes the degree
of the polynomial. The Legendre polynomials are perhaps
the simplest set of polynomials belonging to this class.
Other polynomial sets important in a variety of applications
include Hermite polynomials, Laguerre polynomials, and
Chebyshev polynomials. More general polynomial sets are
the Gegenbauer and Jacobi polynomials, which include the
others as special cases.

Legendre Polynomials
The Legendre polynomial y = P n (x) represents a bounded

solution of the second-order differential equation (DE),
(1 x2 )y00 2x y0 + n(n + 1)y = 0, 1 < x < 1,
called Legendres equation. This DE often arises in problems

featuring spherical symmetry.
The generating function for the Legendre polynomials is
1
P n (x)t n ,
X
p = 1 x 1, | t| < 1,
1 2xt + t2 n=0
where
[n/2] (1)k (2n 2k)!x n2k

X n/2, n even
P n (x) = , [n/2] =
k=0 2n k!(n k)!(n 2k)! (n 1)/2, n odd
and n = 0, 1, 2, . . ..
For n = 0, 1, 2, 3, the Legendre polynomials are:
P0 (x) = 1
P1 (x) = x
1
P2 (x) = (3x2 1)
2
1
P3 (x) = (5x3 3x)
2
42 Orthogonal Polynomials
Legendre Polynomial Identities
Some basic identities for the Legendre polynomials include

the following:
(P1) : P n ( x) = (1)n P n (x), n = 0, 1, 2, . . .

(P2) : P n (1) = 1, P n (1) = (1)n , n = 0, 1, 2, . . .
n
(1) (2n)!
(P3) : P2n (0) = , P2n+1 (0) = 0, n = 0, 1, 2, . . .
22n (n!)2
1 dn
(P4) : P n (x) = [(x2 1)n ], n = 0, 1, 2, . . .
2n n! dx n
(P5) : (n + 1)P n+1 (x) (2n + 1)xP n (x) + nP n1 (x) = 0,
n = 1, 2, 3, . . . (Recurrence formula)
(P6) : P n0 +1 (x) 2xP n0 (x) + P n0 1 (x) P n (x) = 0, n = 1, 2, 3, . . .
(P7) : P n0 +1 (x) xP n0 (x) (n + 1)P n (x), n = 0, 1, 2, . . .
(P8) : xP n0 (x) P n0 1 (x) nP n (x) = 0, n = 1, 2, 3, . . .
(P9) : P n0 +1 (x) P n0 1 (x) = (2n + 1)P n (x), n = 1, 2, 3, . . .
(P10) : (1 x2 )P n0 (x) = nP n1 (x) nxP n (x), n = 1, 2, 3, . . .
1 h in
Z p
(P11) : P n (x) = x + x2 1 cos d , n = 0, 1, 2, . . .
0
Z 1
(P12) : P n (x)P k (x) dx = 0, k 6= n (Orthogonality)
1
Z 1 2
(P13) : [P n (x)]2 dx = , n = 0, 1, 2, . . .
1 2n + 1
1 x

(P14) : P n (x) = 2 F1 n, n + 1; 1; , n = 0, 1, 2, . . .
2
Legendre polynomials first arose in the problem of

expressing the Newtonian potential of a conservative force
field in an infinite series involving the distance variables of
two points and their included central angle. Similar problems
dealing with either gravitational or electrostatic potentials
can also lead to these polynomials, as do some problems in
steady-state heat conduction.

Legendre Functions of the Second Kind
The Legendre polynomial P n (x) represents a bounded solution

of Legendres DE on the interval 1 < x < 1. A second linearly
independent solution (not bounded) on this interval is the
Legendre function of the second kind:
[(n
X 1)/2] (2n 4k 1)
Q n (x) = P n (x)Q 0 (x) P n2k1 (x),
k=0 (2k + 1)(n k)
1 < x < 1,
where
1 1+ x
Q 0 (x) = ln , 1 < x < 1.
2 1 x
Because of the logarithm term in Q 0 (x), it is clear that all Q n (x)

have infinite discontinuities at x = 1. Nonetheless, within the
unit interval these functions are well defined.
Sometimes it is important to consider Q n (x) on the interval

x > 1. Although the above representation for Q n (x) is not valid
on this interval, the functions Q n (x) can be expanded in a
convergent asymptotic series defined by the hypergeometric
function relation
p
n! n n 1 3 1

Q n (x) = 2 F1 + 1, + ; n + ; 2 ,
(n + 3/2)(2x)n+1 2 2 2 2 x
where n 6= 1, 2, . . .. In addition to Q 0 (x) defined above, the first

few Legendre functions of the second kind are:
Q 1 (x) = xQ 0 (x) 1
3
Q 2 (x) = P2 (x)Q 0 (x) x
2
5 2
Q 3 (x) = P3 (x)Q 0 (x) x2 + .
2 3
The Legendre functions Q n (x) satisfy all of the same recurrence

relations as those that are satisfied for the Legendre
polynomials P n (x).

Associated Legendre Functions
In applications involving either the Laplace or Helmholtz

equation in spherical, oblate spheriodal, or prolate spheroidal
coordinates, it is not Legendres equation that arises but rather
the associated Legendre equation
m2

(1 x2 )y00 2x y0 + n(n + 1) y = 0, 1 < x < 1.
1 x2
The linearly independent solutions of this DE, called the

associated Legendre functions of the first and second
kinds are defined by (m = 0, 1, 2, . . . , n) and respectively
expressed as
dm
P nm (x) = (1 x2 )m/2 P n (x),
dx m
dm
Q nm (x) = (1 x2 )m/2 Q n (x).
dx m
Basic properties of these functions are
(AP1) : P n0 (x) = P n (x), n = 0, 1, 2, . . .

(n m)! m
(AP2) : P nm (x) = (1)m P (x), n = 0, 1, 2, . . .
(n + m)! n
(AP3) : (n m + 1)P nm+1 (x) (2n + 1)xP nm (x)
+ (n + m)P nm1 (x) = 0,
Z 1
(AP4) : P nm (x)P km (x) dx = 0, k 6= n (Orthogonalty)
1
Z 1 2(n + m)!
(AP5) : [P nm (x)]2 dx = , n = 0, 1, 2, . . .
1 (2n + 1)(n m)!

Spherical Harmonics
The term spherical harmonic often refers to any solution

of the more general form of the associated Legendre equation
given by
2

(1 x2 )y00 2x y0 + ( + 1) y = 0, 1 < x < 1,
1 x2
in which the parameters and can take on arbitrary
real or complex values. In many applications involving
Laplaces equation, Helmholtzs equation, or the Shrdinger
wave equation for central force fields, we define normalized
spherical harmonics by
s
m m (2n + 1)(n m)! m
Yn (, ) = (1) P n (cos ) e im ,
4(n + m)!
n m n, n = 0, 1, 2, . . . ,
which are the spherical coordinate angular solutions of

1 2 1 1 2

r + sin + + k2 f (r) = 0.
r2 r r sin r sin2 2
Some particular spherical harmonics are given by

1
Y00 (, ) = p
4
r
3
Y10 (, ) = cos
4
s s
3 i 3
1
Y1 (, ) = 1
sin e , Y1 (, ) = sin e i
8 8
r
5 3 1

0 2
Y2 (, ) = cos
4 2 2
r
5
Y21 (, ) = 3 sin cos e i
24
s
5
2
Y2 (, ) = 3 sin2 e2i , Y22 (, )
96
s
5
= 3 sin2 e2i
96

Hermite Polynomials
For a given value of n, the Hermite polynomial y = H n (x) is a

solution of the second-order DE
y00 2x y0 + 2n y = 0, < x < ,
called Hermites equation. The Hermite polynomials are

important in quantum mechanics and probability theory, among
other areas of application.
The generating function for the Hermite polynomials is
tn
exp(2xt t2 ) =
X
H n (x) , < x < , | t| < ,
n=0 n!
where
[n/2] (1)k n!
(2x)n2k ,
X
H n (x) = n = 0, 1, 2, . . . .
k=0 k!(n 2k)!
For n = 0, 1, 2, 3, the Hermite polynomials are
H0 (x) = 1
H1 (x) = 2x
H2 (x) = 4x2 2
H3 (x) = 8x3 12x

Hermite Polynomial Identities
Some basic identities for the Hermite polynomials include
(H1) : H n ( x) = (1)n H n (x), n = 0, 1, 2, . . .

(2n)!
(H2) : H2n (0) = (1)n , H2n+1 (0) = 0, n = 0, 1, 2, . . .
n!
2 d n x2
(H3) : H n (x) = (1)n e x (e ), n = 0, 1, 2, . . .
dx n
(H4) : H n+1 (x) 2xH n (x) + 2nH n1 (x) = 0, n = 1, 2, 3, . . .
(Recurrence formula)
(H5) : H n0 (x) = 2nH n1 (x), n = 1, 2, 3, . . .

Z
2
(H6) : e x H n (x)H k (x) dx = 0, k 6= n (Orthogonality)

Z 2 p
(H7) : e x [H n (x)]2 dx = 2n n! , n = 0, 1, 2, . . .

(2n)! 1 2

(H8) : H2n (x) = (1)n F 1 n; ; x , n = 0, 1, 2, . . .
n! 1 2
n (2n + 1)! 3 2

(H9) : H2n+1 (x) = (1) 2x 1 F1 n; ; x ,
n! 2
n = 0, 1, 2, . . .
Like the Legendre polynomials, the Hermite polynomials

form a classic orthogonal polynomial set. They appear
in the Edgeworth series of probability theory, in certain
combinatoric problems, and they describe the eigenstates of
the quantum harmonic oscillator. The Hermite polynomials
are also used to describe higher-order modes of a Gaussian
optical wave.

Hermite Polynomial Example
Use the generating function relations for the Hermite

polynomials, given by
tn 2
sk 2
H n (x) = e2xt t ; H k (x) = e2xss ,
X X
n=0 n! k=0 k!
to deduce that
Z 2
e x H n (x)H k (x) dx = 0, k 6= n

Z
2
h i2 p
e x H n (x) dx = 2n n! .

Solution:
Multiplying the two series above provides
X tn sk h i
H n (x)H k (x) = exp (t2 + s2 ) + 2x(t + s) .
X
n=0 k=0 n! k!
2
Next, multiply both sides by e x and integrate to find
X tn sk Z 2
e x H n (x)H k (x) dx
X
n=0 k=0 n! k!
Z
2
h i
= e x exp (t2 + s2 ) + 2x(t + s) dx

Z
2 2 2
= e(t +s ) e x +2x(t+s) dx

p 2ts
= e .
If the last exponential function is expanded in a series, the
result is
X tn sk Z 2 p X 2n t n s n
e x H n (x)H k (x) dx =
X
.
n=0 k=0 n! k! n=0 n!
Because the right-hand side has no terms involving t n s k , k 6= n,

one can conclude that
Z
2
e x H n (x)H k (x) dx = 0, k 6= n.

From the remaining series with k = n, the comparison of like

coefficients of t n s n yields the second result above.

Laguerre Polynomials
For a given n, the Laguerre polynomial y = L n (x) is a solution

of the second-order DE, expressed as
x y00 + (1 x)y0 + n y = 0, 0 < x < ,
called Laguerres equation. The standard Laguerre polynomi-

als occur in quantum mechanics and probability theory, among
other areas of application.
The generating function for the Laguerre polynomials is

1 xt

L n (x)t n ,
X
exp = 0 x < , | t| < 1,
1 t 1 t n=0
where
Xn (1) k n!x k
L n (x) = 2
, n = 0, 1, 2, . . . .
k=0 (k!) (n k)!
For n = 0, 1, 2, 3, the Laguerre polynomials are
L 0 (x) = 1
L 1 (x) = x + 1
1 2
L 2 (x) = x 4x + 2
2!
1 3
L 3 (x) = x + 9x2 18x + 6
3!
Laguerre Polynomial Identities
Some basic identities for the Laguerre polynomials include
(L1) : L n (0) = 1, n = 0, 1, 2, . . .
e d n n x
x
(L2) : L n (x) = (x e ), n = 0, 1, 2, . . .
n! dx n
(L3) : (n + 1)L n+1 (x) + (x 1 2n)L n (x) + nL n1 (x) = 0,
(L4) : L0n (x) L0n1 (x) + L n1 (x) = 0, n = 1, 2, 3, . . .
(L5) : xL0n (x) = nL n (x) nL n1 (x), n = 1, 2, 3, . . .

Z
(L6) : e x L n (x)L k (x) dx = 0, k 6= n (Orthogonality)
0
Z
(L7) : e x [L n (x)]2 dx = 1, n = 0, 1, 2, . . .
0
Among other applications, the Laguerre polynomials arise

in the hydrogen atom problem. That is, the radial part of the
wave function of an electron in a Coulomb potential is the
product of a Laguerre polynomial and an exponential with
the variable scaled by a factor depending on the degree. The
Laguerre polynomials are also used to describe higher-order
modes of a circularly symmetric Gaussian optical wave.

Associated Laguerre Polynomials
A generalization of the Laguerre polynomial, called the

associated Laguerre polynomial, is sometimes found in
applications. The more general polynomial can be derived from
the standard Laguerre polynomial L n (x) by
dm
L(m)
n (x) = (1)
m
[L n+m (x)], m = 0, 1, 2, . . . ,
dx m
or defined by its series representation as
n (1) k (n + m)!x k
L(m)
X
n (x) = , m, n = 0, 1, 2, . . . .
k=0 (n k)!(k + m)!k!
When m is neither zero nor a positive integer, the associated

Laguerre polynomial is defined by
n (1) k (n + a + 1)x k
L(a)
X
n (x) = , a > 1, n = 0, 1, 2, . . . .
k=0 (n k)!(k + a + 1)k!
Some basic identities satisfied by the associated Laguerre

polynomials are the following:
(AL1) : L(0)
n (x) = L n (x), n = 0, 1, 2, . . .
(n + m)!
(AL2) : L(m)
n (0) = , m, n = 0, 1, 2, . . .
n!m!
x m
e x d n n+ m x
(AL3) : L(m)
n (x) = x e , m, n = 0, 1, 2, . . .
n! dx n
(AL4) : (n + 1)L(m) (m)
n+1 (x) + (x 1 2n m)L n (x)
+ (n + m)L(m)
n1 (x) = 0, m = 0, 1, 2, . . . ,
(AL5) : L(m) (m1)
n1 (x) + L n (x) L(m)
n (x) = 0, m, n = 1, 2, 3, . . .
(m)0 (m+1)
(AL6) : L n (x) = L n1 (x), m = 0, 1, 2, . . . , n = 1, 2, 3, . . .
Z
(m)
(AL7) : e x x m L(m)
n (x)L k (x) dx = 0, k 6= n (Orthogonality)
0
Z
(n + m)!
(AL8) : e x x m [L(m) 2
n (x)] dx = , m, n = 0, 1, 2, . . .
0 n!

Chebyshev Polynomials
Chebyshev polynomials of the first kind are defined by
X (1)k (n k 1)!
n [n/2]
T0 (x) = 1; T n (x) = (2x)n2k ,
2 k=0 k!(n 2k)!
n = 1, 2, 3, . . . ,
and Chebyshev polynomials of the second kind by
[n/2]
nk
(1)k (2x)n2k ,
X
Un (x) = n = 0, 1, 2, . . . .
k=0
k
They also have the representations

h i
sin (n + 1) cos1 x
T n (x) = cos n cos1 x ;Un (x) =

p , n = 0, 1, 2, . . . .
1 x2
The Chebyshev polynomials have special importance because

series of Chebyshev polynomials converge more rapidly than
any other series of Gegenbauer polynomials and much more
rapidly than power series. For this reason they are widely used
in approximation theory.
For any n, the polynomial y = T n (x) is a solution of the second-

order DE, expressed as
(1 x2 )y00 x y0 + n2 y = 0, 1 < x < 1,
and y = Un (x) is a solution of
(1 x2 )y00 3x y0 + n(n + 2)y = 0, 1 < x < 1.
The first few of each polynomial type are given by
T0 (x) = 1 U0 (x) = 1
T1 (x) = x U1 (x) = 2x
T2 (x) = 2x2 1 U2 (x) = 4x2 1
T3 (x) = 4x3 3x U3 (x) = 8x3 4x

Chebyshev Polynomial Identities
Some identities satisfied by the Type I and Type II polynomials

are
Type I:
(T1) : T n (1) = 1, n = 0, 1, 2, . . .
(T2) : T2n (0) = (1)n , T2n+1 (0) = 0, n = 0, 1, 2, . . .
1 xt
T n (x)t n
X
(T3) : 2
=
1 2xt + t k=0
(T4) : T n+1 (x) 2xT n (x) + T n1 (x) = 0, n = 1, 2, 3, . . .
(T5) : T n (x) = Un (x) xUn1 (x), n = 1, 2, 3, . . .
Z 1
1
(T6) : p T n (x)T k (x) dx = 0,
1 1 x2
k 6= n (Orthogonality)
Z 1
1

, n = 0
(T7) : p [T n (x)]2 dx =
1 1 x2 /2, n = 1, 2, 3, . . .
Type II:
(U1) : Un (1) = n + 1, n = 0, 1, 2, . . .
n
(U2) : U2n (0) = (1) , U2n+1 (0) = 0, n = 0, 1, 2, . . .
1
Un (x)t n
X
(U3) : =
1 2xt + t2 k=0
(U4) : Un+1 (x) 2xUn (x) + Un1 (x) = 0, n = 1, 2, 3, . . .
(U5) : T n (x) = Un (x) xUn1 (x), n = 1, 2, 3, . . .
Z 1p
(U6) : 1 x2 Un (x)Uk (x) dx = 0,
1
k 6= n (Orthogonality)
Z 1p

(U7) : 1 x2 [Un (x)]2 dx = , n = 0, 1, 2, . . .
1 2

Gegenbauer Polynomials
The Gegenbauer (or ultraspherical) polynomials C

n (x) are
defined by the generating function
1 1
C n
X
= n (x)t , 1 x 1, | t| < 1, > ,
(1 2xt + t2 ) n=0 2
and are solutions of the DE

(1 x2 )y00 (2 + 1)x y0 + n(n + 2)y = 0, 1 < x < 1.
The Gegenbauer polynomials are closely connected with
axially symmetric potentials in n dimensions and contain the
Legendre, Hermite, and Chebyshev polynomials as special
cases.
Some of the identities associated with these polynomials include

the following:
(GB1) : C 1/2
n (x) = P n (x) (Legendre polynomial), n = 0, 1, 2, . . .
x

(GB2) : n! lim n/2 C n p = H n (x)

(Hermite polynomial), n = 0, 1, 2, . . .
n C (x)
(GB3) : lim n = T n (x)
2 0
(Chebyshev polynomial), n = 1, 2, 3, . . .
(GB4) : C 1n (x) = Un (x) (Chebyshev polynomial), n = 0, 1, 2, . . .

n 2 2
(GB5) : C
n (1) = ( 1) , C
n ( 1) = , n = 0, 1, 2, . . .
n n

(GB6) : C 2n (0) = , C 2n +1 (0) = 0, n = 0, 1, 2, . . .
n
(GB7) : (n + 1)C
n+1 (x) 2( + n)xC n + (2 + n 1)C n1 (x) = 0,
0 +1
(GB8) : C n (x) = 2C n+1 (x), n = 0, 1, 2, . . .
Z 1
(GB9) : (1 x2 )1/2 C
n (x)C k (x) dx=0, k 6= n (Orthogonality)
1
1 212 (n + 2)
Z
(GB10) : (1 x2 )1/2 [C 2
n (x)] dx = ,
1 (n + )[()]2 n!
n = 0, 1, 2, . . .

Jacobi Polynomials
The Jacobi polynomials, which are generalizations of the

Gegenbauer polynomials, can be defined by the generating
function relation
2 a+ b
(a,b)
(1 t + R)a (1 + t + R)b = P n (x)t n , a > 1, b > 1,
X
R n=0
p
where R = 1 2xt + t2 . Among others, these polynomials have
the following three series representations:
n
n + a n + b x 1 k x + 1 n k

(a,b) X
Pn (x) = ,
k=0
nk nk 2 2
n
n+a n+k+a+b x1 k

(a,b) X
P n (x) = ,
k=0
nk k 2
n x+1 k

(a,b) n k n + b n + k + a + b
X
P n (x) = (1) .
k=0
nk k 2
Some identities for these polynomials are given below:

(0,0)
(JP1) : Pn
(x) = P n (x), n = 0, 1, 2, . . .
2x

(a,b)
(JP2) : lim P n 1 = L(a)
n (x), n = 0, 1, 2, . . .
b b
( + 1/2)(n + 2) (1/2,1/2)
(JP3) : Pn (x) = C
n (x),
(2)(n + + 1/2)
n = 0, 1, 2, . . .
(a,b) (a,b)
(JP4) : Pn ( x) = (1)n P n (x), n = 0, 1, 2, . . .

(a,b) a+n+1
(JP5) : P n (1) = ,
n

(a,b) b+n+1
P n (1) = (1)n , n = 0, 1, 2, . . .
n
(a,b) (1)n
(JP6) : Pn (x) = (1 x)a (1 + x)b
2n n!
dn h a+ n b+ n
i
(1 x) (1 + x) , n = 0, 1, 2, . . .
dx n
Z 1
(a,b) (a,b)
(JP7) : (1 x)a (1 + x)b P n (x)P k (x) dx = 0, k 6= n
1

56
Bessel Functions
The German astronomer F. W. Bessel (17841846) is credited
with deriving the differential equation bearing his name
and carrying out the first systematic study of the properties
of its solutions, now called Bessel functions. Nonetheless,
Bessel functions were first discovered in 1732 by D. Bernoulli
(17001782), who provided a series solution for the Bessel
function representing the oscillatory displacements of a heavy
hanging chain. Other famous mathematicians like Fourier also
used Bessel functions in some of their published work.
Bessel functions are closely associated with problems possess-

ing circular or cylindrical symmetry. Because of this, the so-
lutions of Bessels equation are also called cylinder func-
tions. Bessel functions of the first and second kinds are special
cases of cylinder functions, as are modified Bessel functions
of the first and second kinds, Hankel functions, spherical
Bessel functions, Kelvins functions, Struve functions,
Airy functions, and Anger and Weber functions.

Bessel Functions 57
Bessel Functions of the First Kind
The Bessel function of the first kind y = J p (x) is a solution

of the second-order DE
x2 y00 + x y0 + (x2 p2 )y = 0, p 0,
called Bessels equation. The parameter p is called the order

of the function. This equation is often associated with problems
featuring circular or cylindrical domains. When p is not an
integer, the function y = J p (x) is a second linearly independent
solution.
The series representation for the Bessel function J p (x) is

expressed as
X (1) k (x/2)2k+ p
J p (x) = , p 0,
k=0 k!(k + p + 1)
where (x) is the gamma function. The most commonly

occurring Bessel functions in practice are those of integer order,
i.e., p = n, n = 0, 1, 2, . . .. For such functions there is
(1) k (x/2)2k+ n
; Jn (x) = (1)n Jn (x), n = 0, 1, 2, . . . .
X
Jn (x) =
k=0 k!(k + n)!
All Bessel functions of the first kind have an oscillatory behavior

somewhat like that of a simple sinusoid. Graphs of J0 (x), J1 (x),
and J2 (x) are shown below.

58 Bessel Functions
Properties of Bessel Functions of the First Kind
The Bessel functions satisfy a large number of identities and

integral relations, some of which are provided below:
(BJ1) : J0 (0) = 1; J p (0) = 0, p > 0

x 1

Jn (x)t n , t 6= 0
X
(BJ2) : exp t =
2 t n=
p
e ix cos = J0 (x) + 2 ( i)n Jn (x) cos n,
X
(BJ3) : i= 1
n=0

X
(BJ4) : Jn (x + y) = Jk (x)Jnk (y), n = 0, 1, 2, . . .
k=
(Addition formula)
p
Jk (a)Jk (b) e ik
X
(BJ5) : J0 a2 + b2 2ab cos =
k=
Z 2
1
(BJ6) : J0 (x) = e ix cos d
2 0
Z
1
(BJ7) : Jn (x) = cos(n x sin ) d , n = 0, 1, 2, . . .
0
s s
2 2
(BJ8) : J1/2 (x) = sin x, J1/2 (x) = cos x, x > 0
x x
d p
(BJ9) : [x J p (x)] = x p J p1 (x)
dx
d p
(BJ10) : [x J p (x)] = x p J p+1 (x)
dx
p
(BJ11) : J p0 (x) + J p (x) = J p1 (x)
x
p
(BJ12) : J p0 (x) J p (x) = J p+1 (x)
x
(BJ13) : J p1 (x) J p+1 (x) = 2J p0 (x)
2p
(BJ14) : J p1 (x) + J p+1 (x) = J p (x) (Recurrence formula)
x
Z
(BJ15) : x p J p1 (x) dx = x p J p (x) + C
Z
(BJ16) : x p J p+1 (x) dx = x p J p (x) + C
Z b
(BJ17) : xJ p (k m x)J p (k n x) dx = 0, J p (kx) = 0 (m 6= n)
0
(Orthogonality)
(x/2) p
(BJ18) : J p (x) , p 6= 1, 2, 3, . . . , x 0+
(1 + p)
s
2 1

(BJ19) : J p (x) cos x p + , p 0, x
x 2 2

Bessel Functions 59
Bessel Functions of the Second Kind
The Bessel function of the second kind y = Y p (x) is a

solution of Bessels second-order DE
x2 y00 + x y0 + (x2 p2 )y = 0, p 0,
which is linearly independent of the Bessel function J p (x). It is

defined by
(cos p) J p (x) J p (x)
Y p (x) = .
sin p
When p = n, n = 0, 1, 2, . . ., the above definition requires further

investigation. When p = 0, it has been shown that
2 x 2 X (1) k (x/2)2k 1 1

Y0 (x) = J0 (x) ln + 1 + + + ,
2 k=1 (k!)2 2 k
where x > 0 and is Eulers constant, and for n = 1, 2, 3, . . . we

use
2 1 (n k 1)!(x/2)2k n
x 1 nX
Yn (x) = Jn (x) ln ,
2 k=0 k!
1 X (1) k (x/2)2k+ n
[(k + n + 1) + (k + 1)] x > 0,
k=0 k!(k + n)!
where (x) is the digamma function. Graphs of Y0 (x), Y1 (x), and

Y2 (x) are shown below. These functions also have oscillatory
characteristics, but their asymptotic behavior near zero is like
that of a logarithmic function.
Y0(x)
Y1(x)
0.5 Y2(x)
2 14
0
6 10
x
-0.5
-1

60 Bessel Functions
Properties of Bessel Functions of the Second Kind
Bessel functions of the second kind satisfy many of the

same recurrence formulas as those that are satisfied for J p (x).
These and some additional properties of Y p (x) are listed below:
(BY 1) : Yn (x) = (1)n Yn (x)

d p
(BY 2) : [x Y p (x)] = x p Y p1 (x)
dx
d p
(BY 3) : [x Y p (x)] = x p Y p+1 (x)
dx
p
(BY 4) : Y p0 (x) + Y p (x) = Y p1 (x)
x
p
(BY 5) : Y p0 (x) Y p (x) = Y p+1 (x)
x
(BY 6) : Y p1 (x) Y p+1 (x) = 2Y p0 (x)
2p
(BY 7) : Y p1 (x) + Y p+1 (x) = Y p (x) (Recurrence formula)
x
2
(BY 8) : Y0 (x) ln x, x 0+

(p) 2 p

(BY 9) : Y p (x) , p > 0, x 0+
x
s
2 1

(BY 10) : Y p (x) sin x p + , p 0, x
x 2 2
The standard Bessel functions of the first kind occur in

a wide variety of applications, particularly those featuring
cylindrical symmetry. These functions occur in mechanics,
statistical communication theory, heat conduction, and
vibration phenomenon, to name a few areas of application.
Bessel functions of the second kind, also called
Neumann functions in much of the literature, are generally
less important in applications. Nonetheless, Bessel functions
of the second kind do appear in some problems like
electromagnetic waves in coaxial cables.

Bessel Functions 61
Modified Bessel Functions of the First Kind
The modified Bessel function of the first kind y = I p (x) is a

solution of the second-order DE, expressed as
x2 y00 + x y0 (x2 + p2 )y = 0, p 0, x > 0,
and called Bessels modified equation. The parameter p is

called the order of the function. This function is often associated
with probability density functions in probability theory. When
p is not an integer, the function y = I p (x) is a second linearly
independent solution.
The series representation for the Bessel function I p (x) is

X (x/2)2k+ p
I p (x) = , p 0,
k=0 k!(k + p + 1)
where (x) is the gamma function. The most commonly

occurring modified Bessel functions in practice are those of
integer order, i.e., p = n, n = 0, 1, 2, . . .. For such functions,
X (x/2)2k+ n
I n (x) = ; I n (x) = I n (x), n = 0, 1, 2, . . . .
k=0 k!(k + n)!
Unlike the standard Bessel functions J p (x), the modified Bessel

function I p (x) does not exhibit any oscillatory behavior. Graphs
of I 0 (x), I 1 (x), and I 2 (x) are shown below.

62 Bessel Functions
Properties of the Modified Bessel Functions of the First Kind
The modified Bessel functions satisfy a large number of identi-

ties and integral relations, some of which are provided below:
(BI1) : I 0 (0) = 1; I p (0) = 0, p > 0
x

1

I n (x)t n , t 6= 0
X
(BI2) : exp t+ =
2 t n=

(BI3) : e x cos = I 0 (x) + 2
X
I n (x) cos n
n=0

X
(BI4) : I n (x + y) = I k (x)I nk (y), n = 0, 1, 2, . . .
k=
Z 2
1
(BI5) : I 0 (x) = e x cos d
2 0
(x/2) p
Z
(BI6) : I p (x) = p e x cos sin2p d , p > 1/2
(p + 1/2) 0
s s
2 2
(BI7) : I 1/2 (x) = sinh x, I 1/2 (x) = cosh x, x > 0
x x
d p
(BI8) : [x I p (x)] = x p I p1 (x)
dx
d p
(BI9) : [x I p (x)] = x p I p+1 (x)
dx
p
(BI10) : I 0p (x) + I p (x) = I p1 (x)
x
0 p
(BI11) : I p (x) I p (x) = I p+1 (x)
x
(BI12) : I p1 (x) + I p+1 (x) = 2I 0p (x)
2p
(BI13) : I p1 (x) I p+1 (x) = I p (x) (Recurrence formula)
Z x
(BI14) : x p I p1 (x) dx = x p I p (x) + C
Z
(BI15) : x p I p+1 (x) dx = x p I p (x) + C
(x/2) p
(BI16) : I p (x) , p 6= 1, 2, 3, . . . , x 0+
(1 + p)
ex
(BI17) : I p (x) p , p 0, x
2 x

Bessel Functions 63
Modified Bessel Functions of the Second Kind
Modified Bessel functions I p (x) and K p (x) of the first and

second kind, respectively, are linearly independent solutions of
Bessels modified equation
x2 y00 + x y0 (x2 + p2 )y = 0, p 0, x > 0.
The modified Bessel function of the second kind is defined by
I p (x) I p (x)
K p (x) = .
2 sin p
When p = n, n = 0, 1, 2, . . ., the above definition requires further

investigation. When p = 0, it has been shown that
x X (x/2)2k 1 1

K 0 (x) = I 0 (x) ln + + 2
1+ ++ , x > 0,
2 k=1 (k!) 2 k
where is Eulers constant, and for n = 1, 2, 3, . . ., there is

1 (1) k (n k 1)!(x/2)2k n
x 1 nX
K n (x) = (1)n1 I n (x) ln + ,
2 2 k=0 k!
(1)n X
(x/2)2k+ n
+ [(k + n + 1) + (k + 1)] x > 0,
2 k=0 k!(k + n)!
where (x) is the digamma function. Graphs of K 0 (x), K 1 (x),

and K 2 (x) are shown below. Note the exponential decay of K n (x)
for increasing x.
10
K2(x)
6 K1(x)
4 K0(x)
2
0 0.5 1 1.5 2 x

64 Bessel Functions
Properties of the Modified Bessel Functions of the Second Kind
Bessel functions of the second kind satisfy many recurrence

formulas similar to those that are satisfied for I p (x). These and
some additional properties of K p (x) are listed below:
(BK1) : K p (x) = K p (x)

r
x
(BK2) : K 1/2 (x) = e
2x
d p
(BK3) : [x K p (x)] = x p K p1 (x)
dx
d p
(BK4) : [x K p (x)] = x p K p+1 (x)
dx
p
(BK5) : K 0p (x) + K p (x) = K p1 (x)
x
0 p
(BK6) : K p (x) K p (x) = K p+1 (x)
x
(BK7) : K p1 (x) + K p+1 (x) = 2K 0p (x)
2p
(BK8) : K p1 (x) K p+1 (x) = K p (x) (Recurrence formula)
x
(BK9) : K 0 (x) ln x, x 0+
(p) 2 p

(BK10) : K p (x) , p > 0, x 0+
2 x
r
x
(BK11) : K p (x) e , p 0, x
2x
Modified Bessel functions of the first and second kind are

prominent in probability applications, theory of waveguides,
and also occur in some diffusion problems. Because of
the distinction in behavior compared with the oscillatory
standard Bessel functions, the modified Bessel functions are
sometimes referred to as the hyperbolic Bessel functions.
Most of their properties are analogous to those of the
standard Bessel functions, much like the similarity in
properties associated with the hyperbolic functions and
circular functions.

Bessel Functions 65
Spherical Bessel Functions
Spherical Bessel functions j n (x) and yn (x) of the first and

second kind, respectively, are linearly independent solutions of
the second-order DE, defined as
x2 y00 + 2x y0 + [x2 n(n + 1)]y = 0, x > 0, n = 0, 1, 2, . . . .
These functions are defined in terms of the standard Bessel

functions of half-integral order by
r r

j n (x) = Jn+1/2 (x); yn (x) = Yn+1/2 (x), n = 0, 1, 2, . . . .
2x 2x
The spherical Bessel functions are also directly related to the

circular functions sin x and cos x, e.g.,
sin x
j 0 (x) =
x
sin x cos x cos x
j 1 (x) = 2 y0 (x) =
x
x x
3 1 3

cos x sin x
j 2 (x) = 3 sin x 2 cos x y1 (x) = 2
x x x x x
3 1 3

y2 (x) = 3 cos x 2 sin x
x x x
and so on. Their graphs are shown below:

66 Bessel Functions
Properties of the Spherical Bessel Functions
Recurrence relations and other properties of the spherical

Bessel functions can be derived directly from those of the
standard Bessel functions J p (x) and Y p (x). Also, the function
yn (x) satisfies many of the same identities, so they will not be
listed separately.
(SB j1) : j 0 (0) = 1; j n (0) = 0, n = 1, 2, 3, . . .

d n+1
(SB j2) : [x j n (x)] = x n+1 j n1 (x), n = 1, 2, 3, . . .
dx
d n
(SB j3) : [x j n (x)] = xn j n+1 (x), n = 1, 2, 3, . . .
dx
n+1
(SB j4) : j 0n (x) = j n1 (x) j n (x), n = 1, 2, 3, . . .
x
n
(SB j5) : j 0n (x) = j n (x) j n+1 (x), n = 1, 2, 3, . . .
x
(SB j6) : (2n + 1) j 0n (x) = n j n1 (x) (n + 1) j n+1 (x),
n = 1, 2, 3, . . .
2n + 1
(SB j7) : j n1 (x) + j n+1 (x) = j n (x), n = 1, 2, 3, . . .
x
n!(2x)n
(SB j8) : j n (x) , n = 0, 1, 2, . . . , x 0+
(2n + 1)!
1 n
(SB j9) : j n (x) sin x , n = 0, 1, 2, . . . , x
x 2
(SB y1) : yn (0) = , n = 0, 1, 2, . . .
2(2n)!
(SB y2) : yn (x) , n = 0, 1, 2, . . . , x 0+
n!(2x)n+1
1 n
(SB y3) : y(x) cos x , n = 0, 1, 2, . . . , x
x 2

Bessel Functions 67
Modified Spherical Bessel Functions
Modified spherical Bessel functions i n (x) and k n (x) of the

first and second kind, respectively, are linearly independent
solutions of the second-order DE
x2 y00 + 2x y0 [x2 + n(n + 1)]y = 0, x > 0, n = 0, 1, 2, . . . .
These functions are defined in terms of the modified Bessel

functions of half-integral order by
r s
2
i n (x) = I n+1/2 (x); k n (x) = K n+1/2 (x), n = 0, 1, 2, . . . .
2x x
They are also directly related to the hyperbolic functions sinh x

and cosh x, and to the exponential function e x , e.g.,
sinh x
i 0 (x) =
x
e x
sinh x cosh x k 0 (x) =
i 1 (x) = 2 +
x x
x 1 1 x

3 1 3 k 1 (x) = 2 + e
i 2 (x) = 3 + sinh x 2 cosh x x x
x x x
and so on.
In addition, they satisfy the following asymptotic relations:
n!(2x)n
(SBi1) : i n (x) , n = 0, 1, 2, . . . , x 0+
(2n + 1)!
ex
(SBi2) : i n (x) , n = 0, 1, 2, . . . , x
2x
(2n)!
(SBk1) : k n (x) , n = 0, 1, 2, . . . , x 0+
n!(2x)n+1
e x
(SBk2) : k n (x) , n = 0, 1, 2, . . . , x
x

68 Bessel Functions
Hankel Functions
The usefulness of the Euler formulas, which are expressed as
e ix = cos x i sin x,
leads to the next definition.
Linear combinations of the standard Bessel functions lead

to the Bessel functions of the third kind, or Hankel
functions, given by
H (1)
p (x) = J p (x) + iY p (x), p0
H (2)
p (x) = J p (x) iY p (x), p 0.
Analogously, spherical Hankel functions are defined as
h(1)
n (x) = j n (x) + i yn (x), n = 0, 1, 2, . . .
h(2)
n (x) = j n (x) i yn (x), n = 0, 1, 2, . . . .
Properties of these functions follow directly from the

corresponding properties of the standard Bessel functions and
spherical Bessel functions, respectively.
The standard Hankel functions are directly related to the

modified Bessel function of the second kind, viz.,
1 1+ p (1)
K p (x) = i H p (ix)
2
1 p
K p (x) = i 1 p H (2)
p ( ix), i= 1.
2
Asymptotic relations for small arguments are given by
2
H0(1) (x) x 0+
ln x,

(p) 2 p

H (1)
p (x) , p > 0, x 0+ .
x

Bessel Functions 69
Struve Functions
The Struve function y = H p (x) is a particular solution of the

nonhomogeneous second-order DE
4(x/2) p+1 1
x2 y00 + x y0 + (x2 p2 )y = p , p> .
(p + 1/2) 2
It can also be defined by the integral
2(x/2) p
Z /2
1
H p (x) = p sin(x cos ) sin2p d , p> ,
(p + 1/2) 0 2
and it has the series representation

X (1)k (x/2)2k+ p+1
H p (x) = , 0 x < .
k=0 (k + 3/2)(k + p + 3/2)
The related modified Struve function is defined by
(x/2)2k+ p+1
L p (x) = i (p+1) H p (ix) =
X
, 0 x < .
k=0 (k + 3/2)(k + p + 3/2)
These functions of half-integral order are related to standard

and modified Bessel functions by
H(n+1/2) (x) = (1)n Jn+1/2 (x), n = 0, 1, 2, . . .

L(n+1/2) (x) = I n+1/2 (x), n = 0, 1, 2, . . . .
Some relations satisfied by these functions include:

s s
2 2
(BS1) : H1/2 (x) = sin x; H1/2 (x) = (1 cos x)
x x
s
2
(BS2) : L1/2 (x) = sinh x
x
2p (x/2) p
(BS3) : H p1 (x) + H p+1 (x) = H p (x) + p
x (p + 3/2)
p
(x/2)
(BS4) : H p1 (x) H p+1 (x) = 2H0p (x) p
(p + 3/2)
70 Bessel Functions
Kelvins Functions
The modified Bessel functions I p (xi 1/2 ) and K p (xi 1/2 ) are
solutions of the second-order DE
p
x2 y00 + x y0 (ix2 + p2 )y = 0, i= 1.
The real and imaginary parts of these modified Bessel functions

can be used to define four real functions known as Kelvins
functions, viz.,
I p (xi 1/2 ) = ber p (x) + i bei p (x)

K p (xi 1/2 ) = ker p (x) + i kei p (x),
where
X cos[(3p/4 + k/2)] x p+2k

ber p (x) =
k=0 k!(p + k + 1) 2
X sin[(3p/4 + k/2)] x p+2k
bei p (x) = .
k=0 k!(p + k + 1) 2
When p = 0, the following special cases are obtained:
X (1) k x 4k
ber(x) = 2 2
k=0 [(2k)!]
X (1)k x 4k+2
bei(x) =
[(2k + 1)!]2 2
k=0
x X (1) k (2k + 1) x 4k
ker(x) = ln ber(x) + bei(x) +
2 4 k=0 [(2k)!]2 2
x X (1) k (2k + 2) x 4k+2
kei(x) = ln bei(x) ber(x) + ,
2 4 k=0 [(2k + 1)!]2 2
where (x) is the digamma function. Recurrence formulas and

other relations can be deduced from those of the modified Bessel
functions.

Bessel Functions 71
Airy Functions
Airy functions Ai(x) and Bi(x) of the first and second kinds,
respectively, are linearly independent solutions of the second-
order DE (Airys equation)
y00 x y = 0.
In terms of the modified Bessel functions of fractional order,

these functions can be defined as
r
1 x 2 3/2

Ai(x) = K 1/3 x ,
3 3
r
x 2 2

Bi(x) = I 1/3 x3/2 + I 1/3 x3/2 .
3 3 3
Some properties associated with the Airy functions are given

below:
1 1
(BAi1) : Ai(0) = ; Ai0 (0) =
32/3 (2/3) 31/3 (1/3)
x 2 3/2

0
(BAi2) : Ai (x) = p K 2/3 x
3 3
Z
1 1

(BAi3) : Ai(x) = cos t3 + xt dt, x 0
0 3
1 2

(BAi4) : Ai(x) p 1/4 exp x3/2 , x
x 3
1 1
(BBi1) : Bi(0) = ; Bi0 (0) = 5/6
31/6 (2/3) 3 (4/3)
x 2 2

(BBi2) : Bi0 (x) = p I 2/3 x3/2 + I 2/3 x3/2
3 3 3
1 2

(BBi3) : Bi(x) p 1/4 exp x3/2 , x
x 3

72 Bessel Functions
Other Bessel Functions
The integral Bessel function of order p is defined by

Z x J p (t)
Ji p (x) = dt.
t
For p = n, n = 0, 1, 2, . . ., it can be shown that
x X (1) k (x/2)2k
Ji0 (x) = ln ++
2 k=1 2k(k!)2
1 X (1) k (x/2)2k+ n
Jin (x) = + , n = 1, 2, 3, . . . .
n k=0 (2k + n)k!(n + k)!
The Anger function is a generalization of the standard Bessel

function J p (x) and is defined by
Z
1
J p (x) = cos (p x sin ) d , x 0,
0
which, for p = n, n = 0, 1, 2, . . ., reduces to Jn (x) = Jn (x).

Recurrence formulas for the Anger function include
J p1 (x) J p+1 (x) = 2J0p (x)

2p 2
J p1 (x) + J p+1 (x) = J p (x) sin p.
x x
A similar function is the Weber function, defined by
1
Z
E p (x) = sin (p x sin ) d , x 0,
0
which is related to the Anger function by
J p (x) sin p = E p (x) cos p E p (x)

E p (x) sin p = J p (x) J p (x) cos p.
Recurrence formulas for the Weber function include
E p1 (x) E p+1 (x) = 2E0p (x)

2p 2
E p1 (x) + E p+1 (x) = E p (x) (1 cos p).
x x
Bessel Functions 73
Differential Equation Example
A large number of differential equations (DEs) occurring

in physics and engineering problems are specializations of the
form
x2 y00 + (1 2a)x y0 + b2 c2 x2c + a2 c2 p2 y = 0,

p 0, b > 0,
with general solution
y = xa C 1 J p (bx c ) + C 2 Y p (bx c ) ,

C 1 , C 2 arbitrary.
Find the general solution of the following DEs:
x2 y00 + 5x y0 + 4x2 + 3 y = 0 y00 + x y = 0

(a) (b)
Solution:
(a) Here, 12a = 5, 2c = 2, b2 c2 = 4, a2 c2 = 4, a2 c2 p2 = 3, from

which a = 2, b = 2, c = 1, and p = 1 can be deduced. Hence, the
general solution is
1
y= [C 1 J1 (2x) + C 2 Y1 (2x)].
x2
(b) First multiply the DE by x2 to obtain
x2 y00 + x3 y = 0.
Thus, a = 1/2, b = 2/3, c = 3/2, and p = 1/3. In this case the

general solution is
p 2 2

y= x C 1 J1/3 x3/2 + C 2 Y1/3 x3/2 .
3 3
Because p is not an integer, an alternate form of the general

solution is
p 2 2

y= x C 1 J1/3 x3/2 + C 2 J1/3 x3/2 .
3 3

74 Bessel Functions
Bessel Function Example
Show that
1 b2
Z
a2 x2
xe J0 (bx) dx = 2 exp 2 , b > 0.
0 2a 4a
Solution:
First replace the Bessel function with its series representa-
tion and interchange the order of summation and integration.
This action leads to
Z 2 x2
(1) n b2n Z 2 2
xea x2n+1 ea x dx.
X
J0 (bx) dx = 2 22n
0 n=0 (n!) 0
However, the integral under the summation is recognized as a

form of gamma function, which yields the result
Z 2 x2 n!
x2n+1 ea dx = .
0 2a2n+2
Consequently,
Z 1 X (1) n b2 n
a2 x2
xe J0 (bx) dx =
0 2a2 n=0 n! 4a2
2
1 b

= 2
exp 2 ,
2a 4a
the last step of which follows from recognizing the power series
expansion of the exponential function.
In working with integrals involving nonelementary functions
like the Bessel function, it is common practice to expand the
nonelementary function in an infinite series and integrate
the result termwise (assuming this is justified!). In many
cases the final series after integration can be recognized as
in the example above. When not in an identifiable form,
the resulting infinite series can sometimes be expressed as
some hypergeometric-type series defining a function whose
properties may be known.

75
Orthogonal Series
This chapter introduces the Fourier series method. Among
other uses, this technique can be useful for the analysis of
periodic waveforms (e.g., power signals). The Fourier series
approach to periodic waveforms reduces the signal being
studied to a spectral representation in which the distribution
of power is found to be concentrated at specific frequencies that
are harmonically related to a fundamental frequency.
Standard Fourier series are those defined by simple sinusoids

leading to trigonometric series. More general forms of
Fourier series can arise in connection with various applications
for which the solution of a DE and/or its boundary conditions
depend on a parameter . Problems of this kind are
widely known as eigenvalue problems. The values of the
parameter that permit nontrivial (i.e., nonzero) solutions
are called eigenvalues, and the corresponding solutions are
the eigenfunctions. In applications the eigenvalues and
eigenfunctions have many different physical interpretations.
For example, in vibration problems the eigenvalues are
proportional to the squares of the natural frequencies of
vibration, whereas the eigenfunctions provide the natural
configuration modes. The eigenvalues denote the possible
energy states in quantum mechanics, and the eigenfunctions
are the associated wave functions. Of particular importance
here is that the eigenfunctions are mutually orthogonal,
and as such are useful in developing generalized Fourier
series, also called eigenfunction expansions. Two common
examples of such series are the Legendre series and Bessel
series.

76 Orthogonal Series
Fourier Trigonometric Series
A function f (t) is called periodic if there exists a constant T > 0

for which f (t + T) = f (t) for all t. The smallest value of T for
which the property is true is called the period.
A periodic function is an example of a power signal if its energy

over one period is finite. Trigonometric functions like sines and
cosines are the simplest examples of periodic functions. Observe
that all members of the set
{1, cos(n0 t), sin(n0 t)} , n = 1, 2, 3, . . . ,
have the same period T = 2/0 . Therefore if f (t) is any

periodic function with period T = 2/0 , look for series
representations of the form
1 X
f (t) = a0 + (a n cos n0 t + b n sin n0 t) ,
2 n=1
where
2 T/2
Z
an = f (t) cos n0 t dt, n = 0, 1, 2, . . .
T T/2
2 T/2
Z
bn = f (t) sin n0 t dt, n = 1, 2, 3, . . . .
T T/2
If f (t) is an even function [i.e., f ( t) = f (t)], then b n = 0, n =

1, 2, 3, . . ., and the series reduces to a cosine series, expressed
as
1 X
f (t) = a0 + a n cos n0 t.
2 n=1
However, if f (t) is an odd function [i.e., f ( t) = f (t)], then

a n = 0, n = 0, 1, 2, . . ., and the series reduces to a sine series,
defined as

X
f (t) = b n sin n0 t.
n=1

Fourier Trigonometric Series: General Intervals
If a Fourier trigonometric series converges, it must converge

to a periodic function f (t) having the same period as the
trigonometric functions. If f (t) is not periodic but defined only
on the finite interval [c, c + T], then it may still be represented
by a trigonometric series of the form
1 X 2n t 2n t

f (t) = a0 + a n cos + b n sin , c < t < c + T,
2 n=1 T T
where
2
Z c+T 2n t
an = f (t) cos dt, n = 0, 1, 2, . . .
T c T
2
Z c+T 2n t
bn = f (t) sin dt, n = 1, 2, 3, . . . .
T c T
In this case the interval of convergence is confined to the finite

interval [c, c + T].
If the constant c defining the start of the interval is zero, i.e.,

[0, T], then a suitable function f (t) defined on this interval can
be represented by either a cosine series, given by
1 X n t
f (t) = a0 + a n cos , 0<t<T
2 n=1 T
2 T n t
Z
an = f (t) cos dt, n = 0, 1, 2, . . . ,
T 0 T
or a sine series, expressed as

X n t
f (t) = b n sin , 0<t<T
n=1 T
2
Z T n t
bn = f (t) sin dt, n = 1, 2, 3, . . . .
T 0 T

Exponential Fourier Series
A sinusoidal steady-state solution to an ac circuit problem

is found most simply by inputting a complex time function
e i t , where is angular frequency. In circuit problems and
other related problems, it is often useful to represent a given
waveform or signal f (t) in a complex exponential series rather
than a trigonometric series. For example, the simple sinusoidal
signal f (t) = A cos(0 t + ) can be written as
h i
f (t) = Re Ae i(0 t+) ,
where Re means the real part of the complex quantity, 0 is the

angular frequency, and is the phase. In this formulation, the
signal is often called a phasor.
If the power signal f (t) is periodic with period T = 2/0 , then it

may be represented by a Fourier trigonometric series, or, using
the complex exponential function in place of the trigonometric
functions cos n0 t and sin n0 t, one obtains the exponential
Fourier series

c n e in0 t = | c n | e in0 t+ in ,
X X
f (t) =
n= n=
where c n = | c n | e in . The Fourier coefficients c n can be

defined by either
1
Z T/2
cn = f (t)e in0 t dt, n = 0, 1, 2, . . . ,
T T/2
or
1
Z T
cn = f (t)e in0 t dt, n = 0, 1, 2, . . . .
T 0
Finally, note that when the signal f (t) is real, there exists the
relation c n = cn , where * denotes the complex conjugate.

Generalized Fourier Series
A set of functions {n (x)}, n = 1, 2, 3, . . . is said to be orthogonal

on an interval a < x < b, with respect to a weighting function
r(x) > 0 if it is true for all members of the set that
Z b
r(x)n (x)k (x) dx = 0, k 6= n.
a
If {n (x)} denotes an orthogonal set of functions on the interval

[a, b] with weighting function r(x), and f (x) is a suitable
function on this interval, then one can consider generalized
Fourier series of the form

X
f (x) = c n n (x), a < x < b,
n=1
where
Rb
r(x) f (x)n (x) dx
c n = Ra b , n = 1, 2, 3, . . . .
2
a r(x)[ n (x)] dx
The following important theorem concerning convergence

applies to all orthogonal series of the type presented in this
chapter.
THEOREM: If {n (x)} is a complete set of orthogonal

functions on the interval [a,b] with weighting function r(x) >
0, and f (x) and f 0 (x) are piecewise continuous functions
on this interval, then the generalized Fourier series of f (x)
converges on the open interval a < x < b to f (x) at points of
continuity of f (x), and to the average value (1/2)[ f (x+ ) + f (x )]
of the left and right limits at points of discontinuity.

Fourier Series Example
A certain type of full-wave rectifier converts the input voltage

v(t) to its absolute value at the output, i.e., |v(t)|. Find the
exponential Fourier series of the rectified sine wave (see figure
below right)
f (t) = |v(t)| = A | sin t|.
Solution:
Here, T = 1 and f (t) = A | sin t|,

0 < t < 1. Thus, 0 = 2/T = 2, and
the series sought is

c n e2ni t ,
X
f (t) =
n=
where
Z 1 A
Z 1h i
cn = A sin()e2ni t dt = e(2n1)i t e(2n+1)i t dt,
0 2i 0
which reduces to
2A
cn = , n = 0, 1, 2, . . . .
(4n2 1)
The series sought is therefore

2A X e2ni t
f (t) = .
n= 4n2 1
Because the function f (t) is even, terms in this series can be

rearranged to obtain the cosine series
A 4A X cos 2n t
f (t) = .
n=1 4n2 1

Legendre Series
The orthogonality property of the Legendre polynomials

is
Z 1
P n (x)P k (x) dx = 0, k 6= n.
1
Because of the special properties associated with the Legendre

polynomials, it may be useful in some situations to represent
arbitrary polynomials as linear combinations of Legendre
polynomials. Thus, if q m (x) is a polynomial of degree m, expect
to find a representation of the form
q m (x) = c 0 P0 (x) + c 1 P1 (x) + + c m P m (x),
where each constant c 0 , c 1 , . . . , c m is defined by

1
Z 1
cn = n + q m (x)P n (x) dx, n = 0, 1, 2, . . . , m.
2 1
Note that if q m (x) is a polynomial of degree m, and m < r,

then (because of the orthogonality property of the Legendre
polynomials) it follows that
Z 1
q m (x)P r (x)dx = 0, m < r.
1
Moreover, if f (x) and f 0 (x) are piecewise continuous functions

on the interval 1 < x < 1, then f (x) can be represented in a
pointwise convergent Legendre series, expressed as

X
f (x) = c n P n (x), 1 < x < 1,
n=0
where the Fourier coefficients are defined by

1
Z 1
cn = n + f (x)P n (x) dx, n = 0, 1, 2, . . . .
2 1

Hermite and Laguerre Series
The orthogonality property of the Hermite polynomials is

Z
2
e x H n (x)H k (x) dx = 0, k 6= n.

Thus, if f (x) and f 0 (x) are piecewise continuous functions on

every finite interval, and
Z
2
e x f 2 (x) dx < ,

then f (x) can be represented in a pointwise convergent

Hermite series, defined as

X
f (x) = c n H n (x), < x < ,
n=0
where
1
Z 2
cn = p e x f (x)H n (x) dx, n = 0, 1, 2, . . . .
2n n!
The orthogonality property of the Laguerre polynomials is

Z
e x L n (x)L k (x) dx = 0, k 6= n.
0
Thus, if f (x) and f 0 (x) are piecewise continuous functions on

every finite interval on 0 < x < , and
Z
e x f 2 (x) dx < ,
0
then f (x) can be represented in a pointwise convergent

Laguerre series, shown by

X
f (x) = c n L n (x), < x < ,
n=0
where
Z
cn = e x f (x)L n (x) dx, n = 0, 1, 2, . . . .
0

Bessel Series
The orthogonality property of the Bessel functions of the first

kind is given by
Z b
xJ p (k m x)J p (k n x) dx = 0, m 6= n,
0
where k m and k n are distinct roots of
J p (kb) = 0, b > 0.
Thus, if f (x) and f 0 (x) are piecewise continuous functions on the

interval 0 < x < b, then f (x) can be represented in a pointwise
convergent Bessel series, expressed as

X 1
f (x) = c n J p (k n x), 0 < x < b, p> ,
n=1 2
where the Fourier-Bessel coefficients are defined by
2
Z b
cn = x f (x)J p (k n x) dx, n = 1, 2, 3, . . . .
b [J p+1 (k n b)]2
2
0
If k m and k n , appearing in the orthogonality property above,

are distinct roots of
hJ p (kb) + kJ p0 (kb) = 0, h 0,
then the Fourier-Bessel coefficients are defined by
2k2n b
Z
cn = 2 x f (x)J p (k n x)dx,
(k2n + h2 )b2 p2 J p (k n b)

0
n = 1, 2, 3, . . . .

Bessel Series Example
Find the Bessel series for

x, 0<x<1
f (x) =
0, 1 < x < 2,
corresponding to { J1 (k n x)}, where J1 (2k n ) = 0, n = 1, 2, 3, . . ..
Solution:
The series sought is

X
f (x) = c n J1 (k n x), 0 < x < 2,
n=1
where
Z 2
1
cn = x f (x)J1 (k n x) dx
2[J2 (k n )]2 0
Z 1
1
= x2 J1 (k n x) dx.
2[J2 (k n )]2 0
By making the change of variable t = k n x and using the integral

property (BJ15) as shown:
Z
t p J p1 (t) dt = t p J p (t) + C,
it follows that
J2 (k n )
cn = , n = 1, 2, 3, . . . .
2k n [J2 (2k n )]2
Consequently,
1 X J2 (k n )
f (x) = J1 (k n x), 0 < x < 2.
2 n=1 k n [J2 (2k n )]2

85
Hypergeometric-Type Functions
Because of the many relations connecting the special functions
to each other, and to the elementary functions, it is natural
to inquire whether more general functions can be developed
so that the special functions and elementary functions are
merely specializations of these general functions. General
functions of this nature have in fact been developed and
are collectively referred to as functions of hypergeometric
type. There are several varieties of these functions, but the
most common are the standard hypergeometric function
2 F 1 (a, b; c, x) and the confluent hypergeometric functions
1 F 1 (a; c, x) and U(a; c; x). Other generalizations exist for which
even generalized hypergeometric functions of the type p F q
are certain specializations where p and q are non-negative
integers.
The major development of the theory of the standard

hypergeometric function was carried out by C. F. Gauss
and published in 1812. A similar analysis on the confluent
hypergeometric function was developed by E. E. Kummer
and published in 1823. Hence, the confluent hypergeometric
functions are sometimes called Kummer functions.
In the first half of the twentieth century new theories

concerning generalized functions began to flourish. Most of this
work followed Barnes use of the gamma function in 1907 to
develop a new theory of the hypergeometric function. In the
1930s, both the E function of MacRobert and the G function
of Meijer were introduced in an attempt to give meaning to the
symbol p F q for the case p > q + 1.

86 Hypergeometric-Type Functions
Pochhammer Symbol
In dealing with the hypergeometric-type functions, it is useful

to introduce a symbol known as the Pochhammer symbol,
expressed as
(a)0 = 1, (a)n = a(a + 1) (a + n 1), n = 1, 2, 3, . . . ,
also defined in terms of the gamma function by
(a + n)
(a)n = , n = 0, 1, 2, . . . .
(a)
Below are some useful properties associated with this symbol in

which n and k are non-negative integers:
(PS1) : (0)0 = 1, (1)n = n!
(PS2) : (a)n+k = (a)k (a + k)n

(1)k (a)n
(PS3) : (a)nk =
(1 a n)k
(1)n
(PS4) : (a)n =
(1 a)n
n
(1) k! ,

(PS5) : ( k)n = (k n)! 0nk

0, n > k

(PS6) : (n + k)! = n!(n + 1)k = k!(k + 1)n

(1)n (a + 1)
(PS7) : (a + 1 n) =
(a)n
(1)n

a
(PS8) : = (a)n
n n!
(PS9) : (2n)! = 22n (1/2)n n!
(PS10) : (2n + 1)! = 22n (3/2)n n!

Hypergeometric Function
The hypergeometric function is defined by
(a) (b) x n
X n n
2 F 1 (a, b; c; x) = , 1 < x < 1,
n=0 (c)n n!
where c 6= 0, 1, 2, . . .. The subscripts 2 and 1 refer to

the number of numerator and denominator parameters,
respectively, in its series representation. This function is also
commonly designated by the symbol F(a, b; c; x), where the
subscripts are omitted.
The argument x in the hypergeometric function can be replaced

by the complex argument z = x+ i y, and the resulting expression
is analytic in various domains of the complex plane. For
example, for | z| > 1, it has been shown that
(c)(b a) 1

2 F 1 (a, b; c; z) = ( z)a 2 F 1 a, 1 c + a; 1 b + a;
(b)(c a) z
(c)(a b) 1

+ ( z)a 2 F 1 b, 1 c + b; 1 a + b; , |arg( z)| < .
(a)(c b) z
The hypergeometric function is one of 24 solutions of this type

that satisfy the second-order DE, given by
x(1 x)y00 + [c (a + b + 1)x]y0 ab y = 0, 1 < x < 1.
The six functions 2 F 1 (a 1, b; c; x), 2 F 1 (a, b 1; c; x), and

2 F 1 (a, b; c 1; x) are called contiguous functions. There exist
15 linear recurrence relations with coefficients at most linear
in x, each of which is between 2 F 1 (a, b; c; x) and two contiguous
functions.
Finally, if either a or b is zero or a negative integer, e.g., if

a = 0, 1, . . . , m, the series above truncates to a polynomial at
n = m.

Hypergeometric Function Identities
Some of the special properties associated with the hypergeo-

metric function include the following:
(HY 1) : 2 F 1 (a, b; c; x) = 2 F 1 (b, a; c; x)
dk (a)k (b)k
(HY 2) : 2 F 1 (a, b; c; x) =
dx k (c)k
2 F 1 (a + k, b + k; c + k; x), k = 1, 2, 3, . . .
(c)
(HY 3) : 2 F 1 (a, b; c; x) =
(b)(c b)
Z 1
t b1 (1 t) cb1 (1 xt)a dt, c>b>0
0
(c)(c a b)
(HY 4) : 2 F 1 (a, b; c; 1) =
(c a)(c b)
a
x
(HY 5) : 2 F 1 (a, b; c; x) = (1 + x) 2 F 1 a, c b; c;
1+ x
(1 + x)a 1
(HY 6) : 2 F 1 (1 a, 1; 2; x) =
ax
1 3 1 1+ x

(HY 7) : 2F1 , 1; ; x2 = ln
2 2 2x 1 x
1 3 1

(HY 8) : 2 F 1 , 1; ; x2 = tan1 x
2 2 x
1 1 3 1

(HY 9) : 2 F 1 , ; ; x2 = sin1 x
2 2 2 x
1 x

(HY 10) : 2 F 1 n, n + 1; 1; = P n (x), n = 0, 1, 2, . . .
2
1 1 x

(HY 11) : 2 F 1 n, n; ; = T n (x), n = 0, 1, 2, . . .
2 2
3 1 x 1

(HY 12) : 2 F 1 n, n + 2; ; = Un (x), n = 0, 1, 2, . . .
2 2 n+1

Hypergeometric Function Example
Use properties of the hypergeometric function to establish

the identity of
n 2
X n 2n
= .
k=0
k n
Solution:
From properties of the Pochhammer symbol, there is
(1)k (1)k

n
= ( n)k = ( n)k ,
k k! (1)k
which leads to
n 2 n ( n) ( n) 1 k
X n X k k
= = F ( n, n; 1; 1).
k=0
k k=0 (1) k k! 2 1
From identity (HY4), note that
(1)(2n + 1) (2n)!
2 F 1 ( n, n; 1; 1) = = .
(n + 1)(n + 1) n!n!
Then, using the definition of the binomial coefficient,
(2n)!

2n
= ,
n!n! n
the intended result is produced, i.e.,
n 2
X n 2n
= .
k=0
k n

Confluent Hypergeometric Functions
The series representation for Kummers function or the

confluent hypergeometric function of the first kind is
given by
(a) x n
X n
1 F 1 (a; c; x) = , < x < ,
n=0 (c) n n!
where c 6= 0, 1, 2, . . .. The subscripts 1 and 1 refer to

the number of numerator and denominator parameters,
respectively, in its series representation. This function is also
commonly designated by the symbol M(a; c; x) or (a; c; x),
where the subscripts are omitted.
The confluent hypergeometric function of the second

kind is a linear combination of functions of the first kind, i.e.,
(1 c)
U(a; c; x) = F (a; c; x)
(1 + a c) 1 1
(c 1) 1 c
+ x 1 F 1 (1 + a c; 2 c; x).
(a)
The confluent hypergeometric functions are linearly indepen-

dent solutions of the second-order DE, given by
x y00 + (c x)y0 a y = 0.
Closely associated with these functions are the Whittaker

functions of the first and second kinds,
M k,m (x) = e x/2 x m+1/2 1 F 1 (1/2 + m k; 2m + 1; x),

2m 6= 1, 2, 3, . . .
and
Wk,m (x) = e x/2 x m+1/2U(1/2 + m k; 2m + 1; x),

2m 6= 1, 2, 3, . . . .
These functions are linearly independent solutions of
1 k 1/4 m2

y00 + + + y = 0.
4 x x2
Confluent Hypergeometric Function Identities
Some useful properties associated with the confluent

hypergeometric functions are listed below:
dk (a)k
(CH1) : F (a; c; x) = F (a + k; c + k; x),
dx k 1 1 (c)k 1 1
k = 1, 2, 3, . . .
(c)
Z 1
(CH2) : F
1 1 (a; c; x) = e xt ta1 (1 t) ca1 dt,
(a)(c a) 0
c>a>0
(CH3) : 1 F 1 (a; c; x) = e x 1 F 1 (c a; c; x)
ax
1 c , | x| << 1

(CH4) : 1 F 1 (a; c; x) (c)
xa , x >> 1
(c a)

d
(CH5) : U(a; c; x) = aU(a + 1; c + 1; x)
dx Z
1
(CH6) : U(a; c; x) = e xt ta1 (1 + t) ca1 dt,
(a) 0
a > 0, x > 0
(CH7) : U(a; c; x) = x1 c U(1 + a c; 2 c; x)
(1 c) + (c 1) x1 c ,

| x| << 1

(CH8) : U(a; c; x) (1 + a c) (a)
a

x , x >> 1
(CH9) : Ei(x) = e xU(1; 1; x)
p 1

p x
(CH10) : K p (x) = (2x) e U p + ; 2p + 1; 2x
2
Hypergeometric and confluent hypergeometric func-

tions represent the most well-known of the generalized hy-
pergeometric functions. They have many properties that are
useful in applications, particularly their differentiation and
asymptotic forms. Most of the other special functions, as well
as many elementary functions, are simply special cases of one
of these hypergeometric functions.

Confluent Hypergeometric Function Example
Show that
1 1 1

2
erfc(x) = p e x U ; ; x2 .
2 2
Solution:
p
By introducing the substitution t = x 1 + s into the definition of
the complementary error function, one finds that
2
Z 2
erfc(x) = p e t dt
x
Z
1 x2 2
= p xe e x s (1 + s)1/2 ds.
0
Relating this last integral to that in identity (CH6),
1
Z
U(a; c; x) = e xt ta1 (1 + t) ca1 dt, a > 0, x > 0,
(a) 0
observe that a = 1 and c = 3/2; hence,
1 3

2
erfc(x) = p xe x U 1; ; x2 .
2
Next, from use of the identity (CH7), i.e.,
U(a; c; x) = x1 c U(1 + a c; 2 c; x),
it can also be written that
1 x2 1 1 2

erfc(x) = p e U ; ; x .
2 2

Generalized Hypergeometric Functions
During the last 90 years there has been considerable interest

in working with generalized hypergeometric functions,
of which the hypergeometric and confluent hypergeometric
functions are special cases. In general, a power series A n x n
P
is said to be a series of hypergeometric type if the ratio

A n+1 /A n is a rational function of n. A particular series of this
type is
(a ) (a ) x n
X 1 n p n
p F q (a 1 , . . . , a p ; c 1 , . . . , c q ; x) = ,
n=0 (c 1 ) n (c q ) n n!
where p and q are nonnegative integers and no c k (k =

1, 2, . . . , q) is zero or a negative integer. This function, which
is denoted by p F q , is called a generalized hypergeometric
function. Provided the series does not terminate, it can be
established by the ratio test of calculus that:
1. If p < q + 1, the series converges for all | x| < .

2. If p = q + 1, the series converges for | x| < 1 and diverges for
| x| > 1.
3. If p > q + 1, the series diverges for all x except x = 0.
Based on point 3, the series for p F q is therefore meaningful in

the case p > q + 1 only if it truncates. Attempts to give meaning
to p F q when p > q + 1 led to further generalizations known as
the Meijer G function and MacRobert E function.
The generalized hypergeometric functions include the

hypergeometric and confluent hypergeometric functions as
special cases. The use of such functions often facilitates
the analysis of complicated problems by permitting complex
expressions to be represented more simply in terms of some
generalized hypergeometric function.

Relations of p F q to Other Functions
Many functions are specializations of various generalized

hypergeometric functions. A few of these relations are listed
below; the absence of a parameter in p F q is emphasized by a
dash:
0 F 0 (; ; x) = ex
a
1 F 0 (a; ; x) = (1 x)
2
1 x

0F1 ; ; = cos x
2 4
x2

F
0 1 ; 1; = J0 (x)
4
3 x2 sin x

F
0 1 ; ; =
2 4 x
1 x2 x 1/2a 1

0 F 1 ; a + ; = a+ Ja1/2 (x), x>0
2 4 2 2
1 F 1 (a; a; x) = 0 F 0 (; ; x) = ex
1 2 n n!

1 F 1 n; ; x = (1) H2n (x), n = 0, 1, 2, . . .
2 (2n)!
3 2 n! 1

n
1 F 1 n; ; x = (1) H2n+1 (x), n = 0, 1, 2, . . .
2 (2n + 1)! 2x
n!
1 F 1 ( n; a + 1; x) = L(a) (x)
(a + 1)n n
p
1 3

2
F
1 1 ; ; x = erf(x)
2 2 2x
1

1F1 ; 1; x = e x I 0 (x/2)
2
1

F
1 1 ; 2; x = e x [I 0 (x/2) + I 1 (x/2)]
2
1 x
1 F 1 (1; 2; x) = (1 e )
x
1 x
1 F 1 (2; 3; x) = 2 (1 e xe x )
x

Meijer G Function
In 1936, C. S. Meijer introduced the G function

!
a 1 , . . . , a p
m,n
G p,q x

c1 , . . . , c q
m m 0
j =1
(c j c k )nj=1 (1 + c k a j )
xck
X
= q p
k=1 j=m+1 (1 + c k c j ) j=n+1 (a j c k )
p F q1
[1 + c k a 1 , . . . , 1 + c k a p ; 1 + c k c 1 , . . . , , . . . ,
1 + c k c q ; (1) pmn x],
where 1 m q, 0 n q 1, no two of the c k s (k = 1, 2, . . . , m)

differ by zero or an integer, and a j c k 6= 1, 2, 3, . . . for j =
1, 2, . . . , n and k = 1, 2, . . . , m. If p = q, restrict | x| < 1. The prime
in the product symbol 0 denotes the omission of the term when
j = k. Also, the parameter corresponding to 1 + c k c k (indicated
by ) is to be omitted. Last, an empty product is interpreted as
unity.
For notational convenience, the shortened symbol is often

introduced, expressed as

m,n a p m,n a 1 , . . . , a p
G p,q x = G p,q x .
cq c1 , . . . , c q
Because of its relation with the p F q functions, it is clear that the

G function incorporates a great many other functions as special
cases (see the next two pages). By doing so and possessing the
properties listed on the next page, the Meijer G function often
provides a powerful tool of analysis.
Meijer redefined the G function in 1941 in terms of a Barnes

contour integral, which ultimately led to an interpretation of
the symbol p F q when p > q + 1.

Properties of the Meijer G Function
Basic properties of the G function are numerous. Listed below

are a few of the most useful of these properties:

1 a p

m,n n,m 1 cq
(MG1) : G p,q = G q,p x
x cq 1 a p

m,n a p m,n1 a 1 , . . . , a j1 , a j+1 , . . . , a p
(MG2) : G p,q x
= G p1,q1 x

cq c 1 , . . . , c k1 , c k+1 , . . . , c q
where a j = c k for some j = 1, . . . , n and some k = m + 1, . . . , q

m,n a p m,n a p + r
(MG3) : x r G p,q x = G p,q x
cq cq + r

d

m,n a p
(MG4) : x c1 G p,q x
dx cq

m,n a1 , . . . , a p
= x1 c1 G p,q x

c 1 + 1, c 2 , . . . , c q

d

m,n a p m,n a 1 1, a 2 , . . . , a p
(MG5) : x G p,q x = G p,q x
dx cq c1 , . . . , c q

m,n a p
+(a 1 1)G p,q x , n1
cq
Z
m,n a 1 , . . . , a p
(MG6) : e x x G p,q x dx
0 c1 , . . . , c q

m,n+1 , a 1 , . . . , a p

= 1 G p+1,q ,
c1 , . . . , c q
> 0,
p + q < 2(m + n), c j > 1, j = 1, . . . , m
Z
m,n a 1 , . . . , a p
(MG7) : x2 J ( x)G p,q x2 dx
0 c1 , . . . , c q
!
22 m,n+1 4 , a 1 , . . . , a p ,

= 2+1 G p+2,q ,
2 c 1 , . . . , c q

1 1+
> 0, = , = , p + q < 2(m + n),
2 2
1 3
c j + + > , j = 1, . . . , m, a j + < , j = 1, . . . , n
2 2 4

Relation of the G Function to Other Functions
The Meijer G function is of such generality that it

incorporates many other functions as special cases. Some of
those cases are listed below:
G 10
01 (x|a) = x e
a x
xa (1 x)b

a + b + 1
G 10
11 x = , 0<x<1
a (b + 1)

1 a
G 11
11 x
0 = (a)(1 + x)a , 1 < x < 1
2
x 1 1

G 10
02 , 0 = p sin x, x 0
4 2
2
x 1 1

G 10
02 0, = p cos x
4 2
p
G 10
02 (x|a, b) = x
(a+ b)/2
Jab (2 x)
2
x p p

G 10
02 , = J p (x), x 0
4 2 2

1, 1
G 10
22 x 1, 0 = ln(1 + x), 1 < x < 1

(a)

1a
G 11
12 x 0, 1 c = (c) 1 F 1 (a; c; x)

p x/2 x

1/2
G 11
12 x p, p = e
Ip , x0
2
(1 a + b) b

a
G 11
12 x b, c = (1 c + b) x1 F1 (1 a + b; 1 c + b; x)

(a)(b)

1 a, 1 b
G 12
22 x = F (a, b; c; x), 1 < x < 1
0, 1 c (c) 2 1
p
G 20
02 (x| p, 0) = 2x
p/2
Kp 2 x

1 x

1/2
G 20
12 2x p, p = p e K p (x)


MacRobert E Function
In the late 1930s, T. M. MacRobert made an attempt to give

meaning to the symbol p F q when p > q + 1. His work led to the
E function, expressed as
p 0 m (a a )
k=1 k n
(a n )xa n
X
E(a 1 , . . . , a p ; c 1 , . . . , c q ; x) = q
n=1 j=1 (c j a n )
q+1 F p1
[a n , 1 + a n c 1 , . . . , 1 + a n c q ; a n a 1 , . . . , , . . . ,
1 + a n a p ; (1) p+ q x],
where | x| < 1 if p = q + 1.
MacRoberts E function never gained wide acceptance

because it was found to be a special case of the Meijer G
function, i.e.,

p,1 1, c 1 , . . . , c q
E(a 1 , . . . , a p ; c 1 , . . . , c q ; x) = G q+1,p x .
a1 , . . . , a p
Hence, all properties of the E function are simple consequences

of the properties of the G function.
Both the Meijer G function and the E function of

MacRobert were attempts to give meaning to the symbol
p F q for the case p > q + 1. These functions are more general
than even the generalized hypergeometric functions p F q and
contain the latter as special cases. The Meijer G function
is especially useful in developing asymptotic formulas for
m,n a
large arguments as a result of the property G p,q 1/x c qp =

n,m 1 c
G q,p x 1 a qp (i.e., MG1).

Meijer G Example
Use the Meijer G function to derive the asymptotic formula
(c)
1 F 1 (a; c; x) x a , x .
(c a)
Solution:
Using the relations
(a)

1a
G 11
12 x 0, 1 c = (c) 1 F 1 (a; c; x);

m,n 1 a p

n,m 1 cq
G p,q = G q,p x 1 a p ,

x cq
the result is
(c) 11 1 a (c) 11 1 1, c

1 F 1 (a; c; x) = G 12 x = G 21 .
(a) 0, 1 c (a) x a
Using the definition of the Meijer G function in terms of a series

of generalized hypergeometric functions, it can be deduced that
a
(c)(a) 1 1

1 F 1 (a; c; x) = F a, 1 + a c; ; .
(a)(c a) x 2 0 x
For x , the 2 F 0 function can be replaced with its first term

in a power series, which is unity. What remains is the desired
result of
(c)
1 F 1 (a; c; x) x a , x .
(c a)

101
Bibliography
Abramowitz, M., and I. A. Stegun, (Eds.), Handbook of

Mathematical Functions, Dover Publications, New York 1965.
Andrews L.C., Special Functions of Mathematics for En-
gineers, 2nd ed., SPIE Press, Bellingham, WA 1998.
[doi:10.1117/3.270709].
Andrews L.C., and R.L. Phillips, Mathematical Techniques for
Engineers and Scientists, SPIE Press, Bellingham, WA 2003.
[doi:10.1117/3.467443].
Arfken G., Mathematical Methods for Physicists, 3rd ed.,
Academic Press, New York 1985.
Bell W.W., Special Functions for Scientists and Engineers, Van
Nostrand, London 1968.
Erdelyi A., et al., Higher Transcendental Functions 3 vols.,
in Bateman Manuscript Project, McGraw-Hill, New York 1953.
Gradshteyn I.S., Ryzhik I.M., and A. Jeffrey, Table of Integrals,
Series, and Products, 4th ed., Academic Press, New York 1980.
Lebedev N.N., Special Functions and Their Applications, 1972.
R. A. Silverman, transl. and ed., Courier Dover Publications,
New York.
Luke Y.L., The Special Functions and Their Applications 2 vols.,
Academic Press, New York 1969.
Olver F.W.J., Asymptotics and Special Functions, Academic
Press, New York 1974.
Watson G.N., A Treatise on the Theory of Bessel Functions, 2nd
ed., Cambridge University Press 1944.

102
Index
Airy functions, 56, 71 elliptic functions, 32, 38

alternating series, 9 elliptic integrals, 32, 37, 38
harmonic, 9 error function, 3234
Anger function, 72 Euler formulas, 68
associated Laguerre Eulers constant, 28, 36, 59,
polynomials, 51 63
associated Legendre exponential Fourier series,
equation, 44, 45 78, 80
associated Legendre exponential integral, 32, 35,
functions, 44 36
asymptotic series, 1618, 28,
43 factorial, 11, 20
Fourier coefficients, 78, 81
Bernoulli numbers, 29, 30 Fresnel integrals, 32, 34
Bernoulli polynomials, 30
Bessel functions, 20, 5658, gamma function, 2023,
6071, 83 2529, 31, 32, 35, 57, 61,
Bessel series, 75, 83, 84 74, 85, 86
Bessels equation, 56, 57 Gaussian function, 4
Bessels modified equation, Gegenbauer polynomials,
61, 63 52, 54, 55
beta function, 20, 26, 27 generalized Fourier series,
binomial coefficient, 11, 12 40, 75, 79
binomial formula, 11 generating function, 30
binomial series, 11, 13, 19 geometric series, 9
Hankel functions, 56, 68

Cauchy product, 14, 15
harmonic series, 9
Chebyshev polynomials, 40,
Hermite polynomials, 40,
52, 54
4648, 82
comb function, 7
Hermite series, 82
converge absolutely, 10, 16
hypergeometric functions,
converge conditionally, 10,
20, 85, 8790, 93
16
cosine integral, 32 improper integrals, 8
cosine series, 76, 77, 80 infinite series, 8, 9
integral Bessel function, 72
delta function, 57
digamma function, 28, 29, Jacobi polynomials, 40, 55
59, 63, 70
double infinite series, 10 Kelvins functions, 56, 70

103
Index
Kummers functions, 85, 90 power series, 8, 1315, 17,

33, 52
Laguerre polynomials, 40, psi function, 28
4951, 82
Laguerre series, 82 rectangle function, 3, 4
Legendre functions, 43 recurrence formulas, 60, 64
Legendre polynomials, Riemann zeta function, 28,
4043, 81 30, 31
Legendre series, 75, 81
Legendres equation, 41, 44 signum (sign) function, 2
logarithmic integral, 32, 35 sinc function, 4
sine integral, 32
MacRobert E function, 85,
sine series, 76, 77
93, 98
spherical Bessel functions,
Meijer G function, 85, 93,
56, 68
9599
spherical Hankel functions,
normal probability 68
distribution, 39 spherical harmonics, 45
step function, 2, 3, 6
orthogonal functions, 40 Stirlings formula, 29
orthogonality property, Struve function, 69
8183
triangle function, 3, 4
periodic functions, 7, 38, 76,
77 Weber function, 72
Pochhammer symbol, 86, 89 Whittaker functions, 90
polygamma functions, 20,
28, 31 zeta function, 20, 31

Larry C. Andrews is Professor Emeritus
of Mathematics at the University of Central
Florida and an associate member of the
College of Optics/CREOL. He is also an
associate member of the Florida Space
Institute (FSI). Previously, he held a faculty
position at Tri-State University and was
a staff mathematician with the Magnavox
Company, an antisubmarine warfare (ASW)
operation. He received a doctoral degree in theoretical
mechanics in 1970 from Michigan State University. Dr. Andrews
has been an active researcher in optical wave propagation
through random media for more than 30 years and is the
author or co-author of twelve textbooks on topics of differential
equations, boundary value problems, special functions, integral
transforms, wave propagation through random media, and
mathematical techniques for engineers. Along with wave
propagation through random media, his research interests
include special functions, random variables, atmospheric
turbulence, and signal processing.

Engin

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Engin

Uploaded by

Copyright:

Available Formats

Field Guide to

SPIE Field Guides

John E. Greivenkamp, Series Editor

Bellingham, Washington USA

Copyright 2011 Society of Photo-Optical Instrumentation

All rights reserved. No part of this publication may be repro-

Welcome to the SPIE Field Guidesa series of publications

Each SPIE Field Guide addresses a major field of optical

The SPIE Field Guides are intended to be living documents.

John E. Greivenkamp, Series Editor

Field Guide to Special Functions

Field Guide to Geometrical Optics, John E. Greivenkamp

Field Guide to Atmospheric Optics, Larry C. Andrews (FG02)

Field Guide to Adaptive Optics, Robert K. Tyson & Benjamin W.

Field Guide to Visual and Ophthalmic Optics, Jim Schwiegerling

Field Guide to Polarization, Edward Collett (FG05)

Field Guide to Optical Lithography, Chris A. Mack (FG06)

Field Guide to Optical Thin Films, Ronald R. Willey (FG07)

Field Guide to Spectroscopy, David W. Ball (FG08)

Field Guide to Infrared Systems, Arnold Daniels (FG09)

Field Guide to Interferometric Optical Testing, Eric P. Goodwin

Field Guide to Illumination, Angelo V. Arecchi; Tahar Messadi;

Field Guide to Lasers, Rdiger Paschotta (FG12)

Field Guide to Microscopy, Tomasz S. Tkaczyk (FG13)

Field Guide to Laser Pulse Generation, Rdiger Paschotta

Field Guide to Infrared Systems, Detectors, and FPAs, Second

Field Guide to Laser Fiber Technology, Rdiger Paschotta

Field Guide to Wave Optics, Dan Smith (FG17)

Field Guide to Special Functions for Engineers, Larry C.

Field Guide to Special Functions

Most of the material chosen for this Field Guide is condensed

Many modern engineering and physics problems demand a

Because of the close association of power series and improper

Unfortunately, notation for various engineering and special

Field Guide to Special Functions

Glossary of Symbols and Notation x

Infinite Series and Improper Integrals 8

Other Functions Defined by Integrals 32

Field Guide to Special Functions

Field Guide to Special Functions

Field Guide to Special Functions

Glossary of Symbols and Notation

(a)n Pochhammer symbol

Field Guide to Special Functions

Glossary of Symbols and Notation

(x) Gamma function

Field Guide to Special Functions

Glossary of Symbols and Notation

Si(x), si(x) Sine integral

Field Guide to Special Functions

Field Guide to Special Functions

Step and Signum (Sign) Functions

The unit step function was first introduced by the English

where the value 1/2 at x = 0 is

When the jump takes place at an arbitrary point x = a, the step

The signum or sign function, defined by

is used to reverse the polarity

Field Guide to Special Functions

Rectangle and Triangle Functions

One of the most useful engineering functions is the rectangle

In some instances it is called

rect(x) = U(x + 1/2) U(x 1/2).

In the time domain this function can be used to represent a