ELLIPTIC INTEGRALS AND SOME APPLICATIONS

Jay Villanueva
Florida Memorial University
15800 NW 42nd Ave
Miami, FL 33054
[email protected]

1. Introduction
1.1 How they arose – perimeter of an ellipse
1.2 Standard forms from Legendre, Jacobi, and Landen
2. Properties and examples
2.1 Evaluating integrals
2.2 Elliptic functions
3. Applications
3.1 Calculus
3.1.1 Perimeter of an ellipse
3.1.2 Arclength of a lemniscate
3.2 Physics
3.2.1 Finite-amplitude pendulum
3.2.2 Perihelion of Mercury
4. Conclusion

References:

1. JW Armitage and WF Eberlein, 2006. Elliptic Functions. Cambridge

University Press.
2. HM Kean and V Moll, 1997. Elliptic Curves. Cambridge University Press.
3. DF Lawden, 1989. Elliptic Functions and Applications. New York: Springer
Verlag.
4. V Prasolov and Y Sovolyev, 1997. Elliptic Functions and Elliptic Integrals.
Providence, RI: American Mathematical Society.
5. HE Rauch and A Lebowitz, 1973. Elliptic Functions, Theta Functions, and
Riemann Surfaces. MD: Williams and Wilkins Co.
6. Schaum’s Outlines, 1996. Advanced Calculus. NY: McGraw-Hill.
Elliptic integrals arose from the attempts to find the perimeter of an ellipse:

𝑥2 𝑦2
+ 𝑏2 = 1, 𝑥 = 𝑎 cos 𝜃, 𝑦 = 𝑏 sin 𝜃
𝑎2

𝑎 2𝜋
𝐿 = ∫ 𝑑𝑠 = 2 ∫−𝑎 √1 + 𝑦′2 𝑑𝑥 = ∫0 √1 + 𝑏 2 𝑐𝑜𝑠 2 𝜃(−𝑎𝑠𝑖𝑛𝜃)𝑑𝜃,

which last integral cannot be evaluated by elementary functions, viz., trigonometric,

exponential, or logarithmic functions.

The incomplete elliptic integral of the first kind is defined as

𝜙 𝑑𝜃
𝑢 = 𝐹(𝑘, 𝜙) = ∫0 √1−𝑘 2 𝑠𝑖𝑛2 𝜃
, 0 < 𝑘 < 1,

where 𝜙 is the amplitude of 𝐹(𝑘, 𝜙) or u, written 𝜙 = am 𝑢, and k is the modulus,

𝑘 = mod 𝑢. The integral is also called Legendre’s form for the elliptic integral of the first
kind. If 𝜙 = 𝜋⁄2, the integral is called the complete integral of the first kind, denoted by
𝐾(𝑘), or simply K.

The incomplete elliptic integral of the second kind is defined by

𝜙
𝐸(𝑘, 𝜙) = ∫0 √1 − 𝑘 2 𝑠𝑖𝑛2 𝜃 𝑑𝜃, 0 < 𝑘 < 1,

also called Legendre’s form for the elliptic integral of the second kind. If 𝜙 = 𝜋⁄2, the
integral is called the complete elliptic integral of the second kind, denoted by 𝐸(𝑘), or
simply E. This is the form that arises in the determination of the length of arc of an
ellipse.

The incomplete elliptic integral of the third kind is defined by

𝜙 𝑑𝜃
𝐻(𝑘, 𝑛, 𝜙) = ∫0 , 0 < 𝑘 < 1, 𝑛 ≠ 0,
(1+𝑛𝑠𝑖𝑛2 𝜃)√1−𝑘 2 𝑠𝑖𝑛2 𝜃

also called Legendre’s form for the elliptic integral of the third kind.

If the transformation 𝑣 = sin 𝜃 is made in the Legendre forms, we obtain the following
integrals, with 𝑥 = sin 𝜙
𝑥 𝑑𝑣
𝐹1 (𝑘, 𝑥) = ∫0 ,
√(1−𝑣 2 )(1−𝑘 2 𝑣 2 )

𝑥 1−𝑘 2 𝑣 2
𝐸1 (𝑘, 𝑥) = ∫0 √ 𝑑𝑣,
1−𝑣 2

𝑥 𝑑𝑣
𝐻1 (𝑘, 𝑛, 𝑥) = ∫0 ,
(1+𝑛𝑣 2 )√(1−𝑣 2 )(1−𝑘 2 𝑣 2 )
called Jacobi’s forms for the elliptic integrals of the first, second, and the third kinds
respectively. These are complete integrals if 𝑥 = 1.

Using Landen’s transformation

1 sin 2𝜙
tan 𝜙 = 𝑘+cos 2𝜙 or 𝑘 sin 𝜙 = sin(2𝜙1 − 𝜙),
1

we can rewrite

2 2 𝜙1 𝑑𝜙1 2√𝑘
𝐹(𝑘, 𝜙) = 1+𝑘 𝐹(𝑘1 , 𝜙1 ) = 1+𝑘 ∫0 , where 𝑘1 = 1+𝑘, 𝑘 < 𝑘1 < 1.
√1−𝑘12 𝑠𝑖𝑛2 𝜙1

By successive applications of the transformation, we obtain a sequence of moduli

2√𝑘
𝑘𝑛 , 𝑛 = 1,2,3, …, 𝑘𝑖 = 1+𝑘𝑖−1 , lim𝑛→∞ 𝑘𝑛 = 1,
𝑖−1

from which we can write

𝑘1 𝑘2 𝑘3 … 𝜙 𝑑𝜃 𝑘1 𝑘2 𝑘3 … 𝜋 Φ
𝐹(𝑘, 𝜙) = √ ∫0 √1−𝑠𝑖𝑛2 𝜃
=√ ln 𝑡𝑎𝑛 ( 4 + 2 ) , Φ = lim𝑛→∞ 𝜙𝑛 .
𝑘 𝑘

In practice, accurate results are obtained after only a few applications of the
transformation.

Many integrals are reducible to elliptic type. If 𝑅(𝑥, 𝑦) is a rational algebraic function of
x and y, i.e., the quotient of two polynomials in x and y, then ∫ 𝑅(𝑥, 𝑦)𝑑𝑥 can be
evaluated in terms of the usual elementary functions (algebraic, trigonometric, inverse
trigonometric, exponential, and logarithmic) if 𝑦 = √𝑎𝑥 + 𝑏 or 𝑦 = √𝑎𝑥 2 + 𝑏𝑥 + 𝑐,
with a,b,c constants. If 𝑦 = 𝑃(𝑥), with 𝑃(𝑥) a cubic or quartic polynomial, the integral
R can be evaluated in terms of elliptic integrals of first, second, or third kinds, or for
special cases in terms of elementary functions. If P is a polynomial of degree > 4, then R
may be evaluated with the aid of hyper-elliptic functions.
2 𝑑𝑥
Example 1. ∫0 . Let 𝑥 = 2 sin 𝜃. The integral becomes
√(4−𝑥 2 )(9−𝑥 2 )

𝜋/2 𝑑𝜃 1 𝜋/2 𝑑𝜃 1 2 𝜋
∫0 √9−4𝑠𝑖𝑛2 𝜃
= 3 ∫0 4
= 3 𝐹 (3 , 2 ) .
√1− 𝑠𝑖𝑛2 𝜃
9

1 𝑑𝑥
Example 2. ∫0 . Let 𝑥 = tan 𝜃. The integral becomes
√(1+𝑥 2 )(1+2𝑥 2 )
 𝜋/4 𝑠𝑒𝑐 2 𝜃𝑑𝜃 𝜋/4 𝑑𝜃 𝜋/4 𝑑𝜃 1 𝜋/4 𝑑𝜃
∫0 = ∫0 = ∫0 = ∫ .
√1+𝑡𝑎𝑛2 𝜃√1+2𝑡𝑎𝑛2 𝜃 √𝑐𝑜𝑠2 𝜃+2𝑠𝑖𝑛2 𝜃 √2−𝑐𝑜𝑠2 𝜃 √2 0 1
√1− 𝑐𝑜𝑠2 𝜃
2

Let 𝜙 = 𝜋⁄2 − 𝜃. The integral becomes

1 𝜋/2 𝑑𝜙 1 1 𝜋 1 𝜋
∫ = {𝐹 ( , 2) − 𝐹 ( , 4 )}.
√2 𝜋/4 1
√1− 𝑠𝑖𝑛2 𝜙 √2 √2 √2
2

6 𝑑𝑥
Example 3. ∫4 . Let 𝑢 = √𝑥 − 3 or 𝑥 = 3 + 𝑢2 . The integral becomes
√(𝑥−1)(𝑥−2)(𝑥−3)

√3 𝑑𝑢
2 ∫1 . Let 𝑢 = tan 𝜃. The integral becomes
√(𝑢2 +2)(𝑢2 +1)

𝜋/3 𝑑𝜃 𝜋/3 𝑑𝜃 𝜋/3 𝑑𝜃

2 ∫𝜋/4 √2𝑐𝑜𝑠2 𝜃+𝑠𝑖𝑛2 𝜃
= 2 ∫𝜋/4 √2−𝑠𝑖𝑛2 𝜃
= √2 ∫𝜋/4 1
√1− 𝑠𝑖𝑛2 𝜃
2

1 𝜋 1 𝜋
= √2 {𝐹 ( , )−𝐹( , )}.
√2 3 √2 4

𝑑𝑥
In general, ∫ , where P is a 3rd or 4th degree polynomial, can be evaluated by elliptic
√𝑃
integrals.

The elliptic functions are defined via the elliptic integrals. The upper limit x in the Jacobi
form of the elliptic integral of the first kind is related to the upper limit 𝜙 in the Legendre
form by 𝑥 = sin 𝜙. Since 𝜙 = am 𝑢, it follows that 𝑥 = sin(am 𝑢). We define the elliptic
functions

𝑥 = sin(am 𝑢) = 𝑠𝑛 𝑢

√1 − 𝑥 2 = cos(𝑎𝑚 𝑢) = 𝑐𝑛 𝑢

√1 − 𝑘 2 𝑥 2 = √1 − 𝑘 2 𝑠𝑛2 𝑢 = 𝑑𝑛 𝑢
𝑥 𝑠𝑛 𝑢
√1−𝑥 2
= 𝑐𝑛 𝑢 = 𝑡𝑛 𝑢.

It is also possible to define inverse elliptic functions. For example, from 𝑥 = 𝑠𝑛 𝑢, we

define 𝑢 = 𝑠𝑛−1 𝑥, or 𝑢 = 𝑠𝑛−1 (𝑥, 𝑘) = 𝑠𝑛−1 𝑥, mod 𝑘, to show the dependence of u on
k.

These functions have many important properties analogous to those of trigonometric

functions. One special property is the fact that these functions are doubly periodic, one
period is real, the other is complex. If
 𝜋/2 𝑑𝜃
𝐾 = ∫0 √1−𝑘 2 𝑠𝑖𝑛2 𝜃
,

then 𝑠𝑛(𝑢 + 4𝐾) = 𝑠𝑛 𝑢

𝑐𝑛(𝑢 + 4𝐾) = 𝑐𝑛 𝑢

𝑑𝑛(𝑢 + 2𝐾) = 𝑑𝑛 𝑢

𝑡𝑛(𝑢 + 2𝐾) = 𝑡𝑛 𝑢.

These functions also have other periods, which are complex. If

𝜋/2 𝑑𝜃
𝐾 ′ = ∫0 √1−𝑘′2 𝑠𝑖𝑛2 𝜃
, 𝑘 ′ = √1 − 𝑘 2 ,

Then 𝑠𝑛 𝑢 has periods 4𝐾 and 2𝑖𝐾 ′ ; 𝑐𝑛 𝑢 has periods 4𝐾 and 2𝐾 + 2𝑖𝐾 ′ ; and dn u has
periods 2𝐾 and 4𝑖𝐾 ′ . For this reason, the elliptic functions are called doubly-periodic.

Application 1. Perimeter of an ellipse.

The ellipse 𝑥 = 𝑎 cos 𝜃, 𝑦 = 𝑏 sin 𝜃, 𝑎 > 𝑏 > 0, has length

𝜋/2 𝜋/2
𝐿 = 4 ∫0 √𝑑𝑥 2 + 𝑑𝑦 2 = 4 ∫0 √𝑎2 𝑐𝑜𝑠 2 𝜃 + 𝑏 2 𝑠𝑖𝑛2 𝜃𝑑𝜃

𝜋/2 𝜋/2
= 4 ∫0 √𝑎2 − (𝑎2 − 𝑏 2 )𝑠𝑖𝑛2 𝜃𝑑𝜃 = 4𝑎 ∫0 √1 − 𝑒 2 𝑠𝑖𝑛2 𝜃𝑑𝜃,

where 𝑒 2 = (𝑎2 − 𝑏 2 )⁄𝑎2 = 𝑐 2 /𝑎2 is the square of the eccentricity of the ellipse.
 𝜋
The result can be written as 𝐿 = 4𝑎𝐸 (𝑒, 2 ) = 4𝑎𝐸(𝑒). For the special case of a circle,
𝑎 = 𝑏 = 𝑟, i.e., 𝑒 = 0, and 𝐸(0) = 𝜋/2, and we recover the circumference of a circle:
𝐿 = 2𝜋𝑟. The term elliptic integral was coined by Count Fagnano (1682 – 1766) in 1750.
He discovered that the arclength of the lemniscate can be expressed in terms of an elliptic
integral of the first kind.

Application 2. Arclength of a lemniscate.

The lemniscate is the figure 8 curve: (𝑥 2 + 𝑦 2 )2 = 𝑎2 (𝑥 2 − 𝑦 2 )2 , or in polar form

𝑟 2 = 𝑎2 cos 2𝜃.

From 𝑑𝑠 2 = 𝑑𝑥 2 + 𝑑𝑦 2 = 𝑑𝑟 2 + 𝑟 2 𝑑𝜃 2 ,
𝜋/4 1 1
𝐿 = 4 ∫𝑟=0 𝑑𝑠 = 4𝑎 ∫0 𝑑𝑟, 𝑟 = tan 𝜃 ⇒
√1−𝑟 4

1 1 𝜋/4 𝑠𝑒𝑐 2 𝜃𝑑𝜃 𝜋/4 𝑑𝜃

∫0 𝑑𝑟 = ∫0 = ∫0 , (cos 2𝜃 = 𝑐𝑜𝑠 2 𝑢) ⇒
√1−𝑟 4 𝑠𝑒𝑐𝜃√1−𝑡𝑎𝑛2 𝜃 √cos 2𝜃

𝜋/2 𝑑𝑢 1 𝜋/2 𝑑𝑢 1 1
= ∫0 = ∫ = ∙ 𝐾( ).
√2−𝑠𝑖𝑛2 𝑢 √2 0 1
√1− 𝑠𝑖𝑛2 𝑢 √2 √2
2

1 1
Thus, 𝐿 = 4𝑎 ∙ 𝐾 ( ) = 𝑎 ∙ 2√2(1.85407) = 𝑎(5.244102).
√2 √2

(Historical note: The rectification of the lemniscate was first done by Fagnano in 1718.
The lemniscatus, L.‘decorated by ribbons’, was first studied in astronomy in 1680 by
Cassini, known as the ovals of Cassini (Figure 3), but his book was published in 1749,
many years after his death. The curves were popularized by the Bernoulli brothers in
1694.) Cassini considered more general forms of the lemniscate for whose points the
products of the distances to two foci is a constant:

𝑑1 𝑑2 = 𝑏 2 ,
𝑎2
𝑏4 = 𝑟 4 + − 𝑟 2 𝑎2 cos 2𝜃.
4

𝑎
When 𝑏 = , centered at the origin, we get the ribbon-shaped curve.
√2

Application 3. Finite-amplitude pendulum.

The equation of motion is:

𝑑𝑝 ̇ 𝑔
𝑚𝑙𝜃̈ = −𝑚𝑔 sin 𝜃. Let 𝑝 = 𝜃̇ → 𝑝 𝑑𝜃 = − 𝑙 sin 𝜃

𝑝2 𝑔
⇒ = 𝑙 cos 𝜃 + 𝐶.
2

𝑑𝜃 2𝑔
I.C.: At 𝑡 = 0: 𝜃 = 𝜃0 , 𝜃̇ = 0 ⇒ = −√ √cos 𝜃 − cos 𝜃0 .
𝑑𝑡 𝑙

The period, T, is given by

𝑇 𝑙 0 𝑑𝜃
=√ ∫ ,
4 2𝑔 𝜃 0 √cos 𝜃−cos 𝜃0

𝑙 𝜃 𝑑𝜃 𝑙 𝜃 𝑑𝜃
or, 𝑇 = 4√2𝑔 ∫0 0 = 2√𝑔 ∫0 0
√cos 𝜃−cos 𝜃0 √𝑠𝑖𝑛2 (𝜃0⁄2)−𝑠𝑖𝑛2 (𝜃⁄2)

𝑙 𝜋/2 𝑑𝑢 𝜃 𝜃0 𝜃
= 4√𝑔 ∫0 √1−𝑘 2 𝑠𝑖𝑛2 𝑢
, sin ( 2) = sin ∙ sin 𝑢, 𝑘 = sin ( 20 )
2

𝑙
∴ 𝑇 = 4√𝑔 ∙ 𝐹(𝑘), an elliptic integral.

For the special case of small oscillations, 𝑘 = 0, we get the classical result:

𝑙
𝑇 = 2𝜋√𝑔.

Application 4. Perihelion of Mercury.

According to Newton’s law, where the force on a particle per unit mass is 𝜇 ⁄𝑟 2 , a planet
moves around the Sun in an ellipse and, if there are no other planets disturbing it, the
ellipse remains the same forever. According to Einstein’s law, the curvature of space
appears as a correction term proportional to 1⁄𝑟 4 , so that the gravitational force per unit
mass is of the form

1 3ℎ2
𝜇 (𝑟 2 + 𝑐 2 𝑟 4 ),

where h is the angular momentum per unit mass of the planet about the Sun, and c is the
speed of light (cf. Lawden, 1989, or Armitage and Eberlein, 2006, for more details). The
prediction of Einstein is that the path is very nearly an ellipse, but it does not quite close
up. In the next revolution, the path will have advanced slightly ahead in which the planet
is moving, and so the orbit is an ellipse which very slowly precesses. The advance of
Mercury is the perihelion of Mercury, and is indeed one of the tests of the theory of
relativity. Elliptic functions help to explain all that.

Mercury’s orbit has period T = 88 days, its semimajor axis is a = 57,909,050 km, and
eccentricity e = 0.205630. In 1859, Leverrier of France calculated 574” of arc/century by
perturbations due to other planets; 531” were measured. To acccount for the
42.98”/century, Leverrier postulated the existence of an inner planet Vulcan inside
Mercury’s orbit, but which was never found. Einstein’s prediction came in 1915.

Figure 5. Perihelion of Mercury

The equations of motion, in polar coordinates (𝑟, 𝜃), are

1 1 ℎ2
(𝑟̇ 2 + 𝑟 2 𝜃̇ 2 ) − 𝜇 (𝑟 + 𝑐 2 𝑟3 ) = 𝐸, 𝑟 2 𝜃̇ = ℎ,
2

where E is the energy. Let 𝑢 = 1⁄𝑟 , the orbit equation becomes

𝑑𝑢 2 2𝜇 2𝜇 2𝐸
(𝑑𝜃 ) = 𝑢 − 𝑢 2 + 𝑐 2 𝑢 3 + ℎ2 .
ℎ2

For the planets in the Solar System, the 3rd term on the RHS is very small, define the
dimensionless variable v by

𝑢 = 𝜇𝑣⁄ℎ2 .

The orbit equation becomes

 𝑑𝑣 2
(𝑑𝜃) = 2𝑣 − 𝑣 2 + 𝛼𝑣 3 − 𝛽 = 𝑓(𝑣),

where 𝛼 = 2(𝜇 ⁄𝑐ℎ)2 , 𝛽 = − 2𝐸ℎ2 ⁄𝜇 2 .

Constraints on the stability of the orbit require

0 < 𝛽 ≤ 1, thus, 𝛽 = 1 + 𝜀, 𝜀 > 0.

The term 𝛼 is very small, the largest value being that for Mercury, 𝛼 = 5.09 × 10−8 .
Further constraints on 𝑓(𝑣) show that the zeros are real and positive and satisfy 0 < 𝑣1 <
1 < 𝑣2 < 2 < 𝑣3 . Accordingly,

𝑓(𝑣) = 𝛼(𝑣 − 𝑣1 )(𝑣 − 𝑣2 )(𝑣 − 𝑣3 ).

Since 𝑓(𝑣) ≥ 0, v must lie in the interval 𝑣1 ≤ 𝑣 ≤ 𝑣2. The case 𝑣 ≥ 𝑣3 would lead to
𝑣 → ∞ as 𝜃 → ∞ (so the planet would fall into the Sun) and so must be excluded.

For small 𝛼, the roots of the cubic 𝑓(𝑣) may be expanded in ascending powers of 𝛼:
𝛼
𝑣1 = 1 − 𝑒 − 2𝑒 (1 − 𝑒)3 + 𝑂(𝛼 2 ),

𝛼
𝑣2 = 1 + 𝑒 + 2𝑒 (1 + 𝑒)3 + 𝑂(𝛼 2 ),

1
𝑣3 = 𝑎 − 2 + 𝑂(𝛼),

with 𝑒 2 = 1 − 𝛽 = 1 + 2𝐸ℎ2 ⁄𝜇 2 .

The orbit is given by

𝑑𝑣
𝛼 1/2 𝜃 = ∫ .
√(𝑣−𝑣1 )(𝑣−𝑣2 )(𝑣−𝑣3 )

Let 𝑣 = 𝑣1 + 1⁄𝑡 2 →
2 𝑑𝑡
𝛼 1/2 𝜃 = − ∫ (𝑡 2 −𝑎2 )(𝑡 2 −𝑏2) ,
√(𝑣2 −𝑣1 )(𝑣3 −𝑣1 )

where 𝑎2 = 1⁄(𝑣2 − 𝑣1 ), 𝑏 2 = 1⁄(𝑣3 − 𝑣1 ).

The last integral is an elliptic integral

1
𝛼 1/2 𝜃 = 𝑠𝑛−1 (𝑡√(𝑣2 − 𝑣1 )),
√(𝑣3 −𝑣1 )

with modulus k given by 𝑘 2 = (𝑣2 − 𝑣1 )⁄(𝑣3 − 𝑣1 ).

Finally, the relativistic orbit is given by
1
𝑣 = 𝑣1 + (𝑣2 − 𝑣1 )𝑠𝑛2 (2 √𝛼(𝑣3 − 𝑣1 )𝜃).

In polar coordinates, the orbit is given by

1 𝜇
= ℎ2 (𝐴 + 𝐵𝑠𝑛2 𝜂𝜃),
𝑟

𝛼
where 𝐴 = 1 − 𝑒 − 2𝑒 (1 − 𝑒)3 + 𝑂(𝛼 2 ),

1
𝐵 = 2𝑒 + 𝛼 (3𝑒 + 𝑒) + 𝑂(𝛼 2 ),

1 1
𝜂 = 2 − 4 𝛼(3 − 𝑒) + 𝑂(𝛼 2 ), and

𝑘 2 = 2𝑒𝛼 + 𝑂(𝛼 2 ).
1
For the special case 𝛼 = 0: 𝐴 = 1 − 𝑒, 𝐵 = 2𝑒, 𝜂 = 2 , 𝑘 = 0, and we recover the
classical orbit
1
= 1 − 𝑒 cos 𝜃,
𝑟

with semilatus rectum 𝑙 = ℎ2 ⁄𝜇 and eccentricity e.

Perihelion occurs between 𝜃 = 𝐾 ⁄𝜂 and 𝜃 = 3𝐾 ⁄𝜂, or Δ𝜃 = 2𝐾 ⁄𝜂 vs 2𝜋 for the

𝜋/2
classical result. Thus, the advance of perihelion per revolution is, with 𝐾 = ∫0 (1 −
𝑘 2 𝑠𝑖𝑛2 𝜃)−1/2 𝑑𝜃:
1
2𝐾 𝜋(1+ 𝑘 2 +⋯ )
4
𝜖= − 2𝜋 = 1 1 = 3𝜋𝑎.
𝜂 − 𝛼(3−𝑒)+⋯
2 4

For Mercury: 𝛼 = 5.09 × 10−8 , 𝑇 = 88 days, and 𝜖 = 43"/century.

Conclusions:

 Elliptic integrals arose from the attempts to find the perimeter of an ellipse, akin
to the circumference of a circle.
 There are several standard forms of elliptic integrals, but they involve radicals of
polynomials of degree 3 or 4.
 Familiarity with elliptic integrals allows us to solve interesting problems in
mathematics and physics that we have heretofore avoided.