Home Search Collections Journals About Contact us My IOPscience
Elliptic integrals, the forgotten functions
This article has been downloaded from IOPscience. Please scroll down to see the full text article.
2001 Eur. J. Phys. 22 119
(http://iopscience.iop.org/0143-0807/22/2/303)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 138.73.1.36
The article was downloaded on 09/05/2013 at 04:32
Please note that terms and conditions apply.
INSTITUTE OF PHYSICS PUBLISHING EUROPEAN JOURNAL OF PHYSICS
Eur. J. Phys. 22 (2001) 119126 www.iop.org/Journals/ej PII: S0143-0807(01)16834-0
Elliptic integrals, the forgotten
functions
R H Good
Department of Physics, California State University at Hayward, Hayward, CA 94542, USA
E-mail: rgood@csuhayward.edu
Received 5 September 2000, in final form 15 November 2000
Abstract
Ten simple and useful complete elliptic integrals are presented. Their
application is illustrated in elementary examples from electromagnetism. Some
background is provided involving their history and their relationship to elliptic
functions.
1. Introduction
Elliptic integrals represent the next step in complexity above elementary integrals.
Consequently they frequently arise in relatively simple geometries, yet they cannot be expressed
in terms of the elementary functions: algebraic, trig and inverse trig, log and exponential.
They were once accorded the respect they deserve; in particular, the great Legendre devoted a
significant part of his career to his three-volume work on the subject [1]. More recently they
have
fallen into disuse, for reasons unclear. For example, we are used to a function such as
dx (a 2 x 2 )1/2 = arcsin(x/a), and the forgetful student can easily look it up in a table of
integrals; but elliptic integrals also frequently arise in the calculations
of physics, and it must
often happen that a student wonders why such a simple form as 0 d(b + cos )1/2 is not
to be found in the integral table. It is most desirable that the student be able to recognize an
elliptic-integral form before invoking Mathematica or Maple.
Graduate-level books [2, 3] typically expect some student familiarity with elliptic integrals,
but undergraduate texts [4] typically do not supply the necessary preparation.
Our purpose here is to redress this state of affairs. We will present ten particularly simple
and useful forms, equations (1)(10), and we will illustrate their use. These formulae frequently
enable one to identify complete elliptic integrals as soon as they appear in a calculation;
furthermore they are far easier to deal with than series expansions tailored to particular
problems, or hypergeometric functions. They are involved in all fields of physics: mechanics
(even the simple pendulum), gravitation, electromagnetism, etc. Yet these formulae are not to
be found in any readily-accessible source (except for the first two or three).
We limit ourselves to the complete integrals because these include most of the elementary
physics applications; and we will not use elliptic functions (see appendix 1). This subject is
by no means advanced, and in fact we will need only first-year physics and first-year calculus.
We hope that students will copy equations (1)(10) and add them to their tables of integrals.
0143-0807/01/020119+08$30.00 2001 IOP Publishing Ltd Printed in the UK 119
120 R H Good
2. Ten useful elliptic integrals
Explanations will follow the list. For the most part our notation conforms to [5]:
/2
d
= K(m) with 0 < m < 1 (1)
0 (1 m sin2 )1/2
/2
(1 m sin2 )1/2 d = E(m) (2)
0
/2
sin2 d K E
= (3)
0 (1 m sin 2
)1/2 m
/2
d E
= (4)
0 (1 m sin2 )3/2 1m
/2
sin2 d E K
= (5)
0 (1 m sin 2
)3/2 m(1 m) m
d 2
= 2mK(m) with m = (6)
(b cos ) 1/2 1+b
0
4
(b cos ) d = E
1/2
(7)
0 2m
cos d 4 2 m
= E 2mK (8)
0 (b cos ) m
1/2
2m
d m
= 2mE (9)
0 (b cos ) 2 2m
3/2
cos d 2m
= 2mK 2mE. (10)
0 (b cos ) 2 2m
3/2
(Note that, in equations (8) and (10), the signs are either both plus or both minus.)
Equation (1) defines the complete elliptic integral of the first kind, K(m). It is called
complete because the limits of integration are 0 and /2. m is the parameter, sometimes
written as k 2 . (If the sign preceding m is positive, then the transformation = /2 is
required.)
Similarly, equation (2) displays the complete elliptic integral of the second kind, E(m).
(Elliptic integrals are so named because earlier they were studied in connection with the
ellipse; in particular, the circumference of an ellipse of semi-axes a and b < a is 4aE(m), with
m = 1 b2 /a 2 .)
The functions K(m) and E(m) are shown in figure 1. We see that K increases without
limit as m 1 (it approaches ln(4(1 m)1/2 ) [6]). When we subtract E from K, we obtain
equation (3) after a little algebra. Numerical tables of K and E are commonplace [79], and
correspondingly it is usual to express other complete elliptic integrals in terms of K and E.
Equations (1) and (3) typically involve a 1/r potential, with the distance r appearing
in the form of a radical in the denominator, for example in the case of the potential V in
electrostatics (and they also serve for the simple pendulum). For problems involving fields,
we may encounter r 2 in the denominator, and also a field component which entails another r
in the denominator, so in such cases the forms (4) and (5) may be useful (for their derivation,
see [14]). We will illustrate both cases in the two examples below.
For formulae (1)(5), the integrand must go through its entire repertoire in the interval
0 < < 90 . However, in many physical problems, bilateral symmetry instead involves an
angle from 0 to 180 and, correspondingly, the integrals are more readily recognized after a
Elliptic integrals, the forgotten functions 121
2
K(m)
/2
E(m)
1
0 m 1 Figure 1. Complete elliptic integrals.
half-angle substitution. So for equations (6)(10) we substituted
= and cos = 1 2 sin2
2
in the case of the plus signs in front of the cosines, or
= and cos = 2 sin2 1
2
for the minus signs; we also introduced b = (2/m) 1, with b > 1 (since 0 < m < 1).
Now we will demonstrate how much equations (1)(10) can simplify problems. We will
deal with only one geometry, a circular ring.
3. Applications
3.1. Example 1
First, we find the scalar potential V in the vicinity of a circular charged ring of charge density
and radius a, located in the xy plane, figure 2. For the scalar potential,
1 dl
V =
40 r
where dl is a length increment along the loop; its magnitude is a d. We pick a point in the
xz plane ( = 0). Using the law of cosines, we find
c = (a 2 + 2 2a cos )1/2
and so
r = (z2 + a 2 + 2 2a cos )1/2 .
Making use of bilateral symmetry, we integrate around half of the circle and multiply by
two. We obtain
1 a d
V =2 .
40 0 (z + a + 2 2a cos )1/2
2 2
As soon as we write the integral we recognize it as being of the form of equation (6):
a d
V =
20 2a 0 (b cos )1/2
122 R H Good
z V, B
y r
z
l dl
a c
f
r x
Figure 2. V for a charged loop.
where b = (z2 + a 2 + 2 )/(2a). Then from (6), with m = 2/(1 + b) = 4a/(z2 + (a + )2 ),
we find the potential V to be [10]
1/2
a
V = 2mK(m). (11)
20 2
Here are three checks on its correctness.
On the loop, z = 0 and a = ; so m = 1 and, from figure 1, V , as expected.
At the centre, z = 0 and is small, so m 4/a, and K(m) /2 (from figure 1); thus
V 2K(m) =
20 20
as expected.
As or z , m 0 and V 0.
To find the electric field E , it would be easier simply to set up the integrals, and use
equations (1)(10), than to differentiate V . We will now illustrate the solution of a vector field
using the magnetic field B for a current loop.
3.2. Example 2
Let us find the vector magnetic field B of the same loop, which now carries current I in the
direction of the vector dl, figure 2. We begin with the BiotSavart equation:
0 I dl r
B= .
4 r3
The hardest part is finding the components of the vectors dl and r . From figure 2, with
dl = a d,
dl = a sin d x + a cos d y r = (x a cos )x a sin y + zz .
Then
dl r = (az cos 0) d x + (0 + az sin ) d y + (a 2 sin2 ax cos + a 2 cos2 ) d z .
By symmetry we no longer need to restrict the field point to the xz plane; we can change x to
and y to , and the cross product becomes
dl r = az cos d + az sin d + (a 2 a cos ) d z .
Elliptic integrals, the forgotten functions 123
So the field components are
0 I az cos
B = 2 d
4 0 (z2 + a 2 + 2 2a cos )3/2
0 I az sin
B = 2 d = 0 by symmetry
4 0 (z + a + 2 2a cos )3/2
2 2
0 I a 2 a cos
Bz = 2 d.
4 0 (z2 + a 2 + 2 2a cos )3/2
(B is an elementary integral; you can integrate it directly and, sure enough, it gives zero.)
The problem is already over; the integrals for B and Bz are immediately recognized
as involving the forms of equations (9) and (10). With b = (z2 + a 2 + 2 )/(2a) and
m = 2/(1 + b) = 4a/(z2 + (a + )2 ), we can write
0 I az cos d 0 I az 2m
B = = 2mK + 2mE
2(2a)3/2 0 (b cos )3/2 2(2a)3/2 2 2m
0 I z m 1/2 2 m
= EK (12)
2 4a 2 2m
0 I a d cos d
Bz = a +
2(2a)3/2 0 (b cos ) 0 (b cos )
3/2 3/2
0 I a m 2m
= a 2mE + 2mK 2mE
2(2a)3/2 2 2m 2 2m
1/2
0 I m am (2 m)
= K + E . (13)
2 4a 2 2m
These results were obtained by Smythe [11] by the much more painful procedure of
differentiating the magnetic vector potential A. Griffiths [12] displays the B integrals but
neither he nor Jackson [3] attempts to solve them. Jackson provides some approximations.
I believe that all of these presentations would be improved with accessibility of equations
(1)(10).
Checks:
B = 0 at z = 0, by inspection (on the x axis; we omit the point at x = a);
Bz = 0 I /(2a) at the centre ( = 0, z = 0);
As or z , m 0 and B and Bz 0.
4. Conclusions
My purpose here is not really subversive as far as elliptic functions are concerned; in fact, I
expect that when students see how easy all of this is they will be motivated to look into them.
However, primarily, I would hope that the ten formulae presented should find their way into
tables and texts, in order to help bring elliptic integrals back into the memory of the current
generation of physicists.
Appendix A. Elliptic functions
In 1825, at the age of 73, Legendre published the first tome of his great work on elliptic integrals
[1]. Three years later, in the introduction to the supplementary third volume, he remarked [13]
(perhaps a little ruefully): . . . but hardly had my work seen the light of day, and hardly could its
title be known to foreign savants, when I learned, with as much astonishment as satisfaction,
that two young geometers, MM Jacobi (C. G. J.) of Koenigsberg and Abel of Christiania,
124 R H Good
had succeeded, by their own efforts, in considerably perfecting the theory of elliptic integrals
at its most advanced points. In fact, what the youngsters had done was to turn the theory
on its head. The crucial insight is attributed to Abel: in about 1823 he noted that Legendres
elliptic integrals had the character of inverse functions, like the inverse trigonometric functions.
After this, what Legendre had called fonctions elliptiques became known simply as elliptic
integrals, and the elliptic functions themselves were introduced as follows.
If we write the incomplete elliptic integral of the first kind as
d
u = F () =
0 (1 m sin )
2 1/2
(so F (/2) = K), then we may write
sin = sn u cos = cn u
and
(1 m sin2 )1/2 = dn u.
((1 c2 sin2 )1/2 was abbreviated to ! by Legendre; his c2 is our m.) The functions sn u,
cn u and dn u are Jacobi elliptic functions, and nine others may be defined in terms of these.
There are many interesting relationships between the functions, for example
sn2 u + cn2 u = 1.
Also, the incomplete elliptic integral of the second kind may be written as a function of the
first:
u
E(u) = dn2 w dw.
0
The understanding of elliptic functions is a prerequisite for further study of elliptic
integrals, especially if one needs to deal with incomplete integrals. However, it takes weeks
just to acquire a useful facility in the subject; and if all one wants is to evaluate a few integrals
in physics, one may not need this facility at all, as was demonstrated by Legendres treatise
(which, after all, pre-dated the subject of elliptic functions) and as we have seen in the examples
in this paper.
Appendix B. Other sources
Legendre [1] presented a list of 17 incomplete elliptic integrals with the comment, We will
finish by uniting in one table some of the simplest formulae which are encountered frequently
in the application of elliptic integrals. Five of them are represented in my equations (1)(5).
Since Legendres list is now seldom encountered [14] we will provide the remaining 12 here,
in their complete form:
/2
cos2 d 1 1m
= E K (14)
0 (1 m sin )
2 1/2 m m
/2
cos2 d K E
= (15)
0 (1 m sin )
2 3/2 m
/2
d 2(2 m) 1
= E K (16)
0 (1 m sin )
2 5/2 3(1 m) 2 3(1 m)
/2
4 2m 1m
(1 m sin2 )3/2 d = E K (17)
0 3 3
/2
2m 1 1m
sin2 (1 m sin2 )1/2 d = E+ K (18)
0 3m 3m
Elliptic integrals, the forgotten functions 125
/2
1+m 1m
cos2 (1 m sin2 )1/2 d = E K (19)
0 3m 3m
/2 /2
d 1 cos d
tan2 21 =
0 (1 m sin2 )1/2 0 1 + cos (1 m sin2 )1/2
= 2(1 m)1/2 + K 2E (20)
/2 /2
1 d 1 d
=
0 2 cos2 21 (1 m sin2 )1/2 0 1 + cos (1 m sin2 )1/2
= (1 m)1/2 + K E (21)
and in the last four formulae the integrals diverge:
/2
d
0 cos (1 m sin2 )1/2
2
/2
tan2 d
0 (1 m sin2 )1/2
/2
(1 m sin2 )1/2
d
0 cos2
/2
tan2 (1 m sin2 )1/2 d .
0
Modern sources [4, 5, 79] typically give only the first two or three of the formulae (1)(5).
([7, 8] also give equation (14), and [7] gives the last four above, which diverge.) This is
unfortunate because, as a consequence, the student (or even the teacher) may be unable to
recognize elliptic integrals in many problems which should be easy.
Mathematica 2.2 [15] is not necessarily much help in identifying the elliptic integrals.
It provides correct results for equations (1), (2), (4) and (6); but it will not integrate (7) at
all, and for the other five it provides results in terms of hypergeometric functions which will
be intractable for the student. For a numerical value of m, it provides correct values of the
integrals except in the case of (7). As for the other 12 formulae of Legendre, listed above, it
does none properly, but it does give hypergeometric functions and numerical values for the
first six of them.
Byrds remarkable collection [16] of over 3000 integrals and formulae is a necessary
resource for anyone seriously interested in elliptic integrals. However, considerable familiarity
with elliptic functions must be acquired in order to put the work to use. (To test this statement,
I suggest that the reader try looking up my formulae (1)(10) in Byrd without knowledge of
elliptic functions.)
I know of no current and readily accessible source for these formulae.
References
[1] Legendre A M 1825 Traite des Fonctions Elliptiques vol 1 (Paris: Huzard-Courcier) pp 2567
[2] Goldstein H 1980 Classical Mechanics 2nd edn (Reading, MA: Addison-Wesley) p 241
[3] Jackson J D 1999 Classical Electrodynamics 3rd edn (New York: Wiley) p 182
[4] Arfken G 1985 Mathematical Methods for Physicists 3rd edn (Orlando, FL: Academic) pp 3216
Boas M L 1983 Mathematical Methods in the Physical Sciences 2nd edn (New York: Wiley) pp 47481
Harper C 1999 Analytic Methods in Physics (Berlin: Wiley) pp 194, 241
[5] Abramowitz M and Stegun I A (eds) 1972 Handbook of Mathematical Functions (Applied Mathematics Series
55) (New York: Dover) (Reprinted from 1964 National Bureau of Standards, US Department of Commerce)
[6] Bowman F 1953 Introduction to Elliptic Functions (New York: Wiley) p 21
Newman F W 1889 Elliptic Integrals (Cambridge: Macmillan and Bowes) p 6 (Legendre [1] p 66 is unsound
on this point.)
[7] Dwight H B 1961 Tables of Integrals and other Mathematical Data 4th edn (New York: MacMillan) pp 1806
126 R H Good
[8] Jahnke E and Emde F 1938 Tables of Functions 3rd edn (Leipzig: Teubner) pp 7385
[9] Zwillinger D (ed) 1996 CRC Standard Mathematical Tables and Formulae 30th edn (Boca Raton, FL: CRC
Press)
[10] Smythe W R 1968 Static and Dynamic Electricity 3rd edn (New York: McGraw-Hill) p 149 (Smythe presents
V in the form of an infinite series of Legendre polynomials. See also p 235, with Bessel functions.)
[11] Smythe W R 1968 Static and Dynamic Electricity 3rd edn (New York: McGraw-Hill) pp 2901
[12] Griffiths D J 1999 Introduction to Electrodynamics 3rd edn (Upper Saddle River, NJ: Prentice-Hall) problem 5.48
[13] Legendre A M 1828 Traite des Fonctions Elliptiques vol 3 (Paris: Huzard-Courcier) p 1
Whittaker E T and Watson G N 1958 A Course of Modern Analysis 4th edn (Cambridge: Cambridge University
Press) pp 51221
[14] Hancock H 1954 Elliptic Integrals (New York: Dover) pp 5962
Newman F W 1889 Elliptic Integrals (Cambridge: Macmillan and Bowes) p 3
[15] Wolfram S 1993 Mathematica 2.2 (Champaign, IL: Wolfram Research)
[16] Byrd P F and Friedman M D 1954 Handbook of Elliptic Integrals for Engineers and Physicists (Berlin: Springer)
