Elliptic Function Jacobi Approach Complete
Elliptic Function Jacobi Approach Complete
Elliptic Function Jacobi Approach Complete
The great advantage of this approach is that there is vast literature for the
properties of the Jacobain elliptic functions and of their associated Jacobian
theta functions. Moreover, the trigonometric functions are the special cases
of the Jacobian elliptic functions when the (elliptic) modulus degenerates
to zero so that the properties and relations of trigonometric functions can
serve a guide, albeit an extremely crude one, to those of the Jacobian elliptic
functions and Jacobian theta functions. The second approach, which we
call the approach of Weierstrass, starts with the construction of an infinite
sum of partial fraction expansions, known as the Weierstrass ℘ function, and
then verifies that the function it represents is a doubly periodic meromorphic
function on C which satisfies a first-order differential equation so that it is
the inverse of an indefinite integral whose integrand is the reciprocal of the
square root of a cubic polynomial. The second approach of Weierstrass is
Math 213a (Fall 2021) Yum-Tong Siu 2
Both Approaches Are Treated in the Course. We will discuss both ap-
proaches in this course, first the inversion of the inverse Jacobian elliptic sine
function defined by an indefinite integral whose integrand is the reciprocal
of the square root of a quartic polynomial. Properties and relations of the
several Jacobian elliptic functions will be discussed. Then we will introduce
the Weierstrass ℘ function as an infinite sum of partial fractions and derive
its properties and its relation with the Jacobian elliptic sine function. Then
we will treat the Jacobian theta functions and their applications to the sums
of squares problem and quadratic reciprocity. We start with the motion of
the simple pendulum, with more details, for whose solution the Jacobian el-
liptic sine function with (elliptic) modulus k is introduced as the inverse of
the indefinite integral
Z ξ
−1 1
sn ξ = p .
x=0 (1 − x )(1 − k 2 x2 )
2
√
by using the substitution x = sin ϕ to get rid of 1 − x2 , one gets the integral
Z
dϕ
p
1 − k 2 sin2 ϕ
and in that integral there is the expression sin2 ϕ. In order to get an integral
equivalent to the form Z
dϕ
p ,
1 − k 2 sin2 ϕ
we use the substitution
sin 2θ
sin ϕ = .
sin α2
One gets the differential equation
2
dϕ g α
= 1 − sin2 sin2 ϕ .
dt a 2
So Z ϕ
a dψ
t= q .
g ψ=0 1 − sin2 α2 sin2 ψ
The final answer is r
θ α g
sin = sin sn t
2 2 a
with the (elliptic) modulus k equal to sin α2 .
because π
π 1 · 3 · 5 · · · (2n − 1)
Z
2
sin2n ϕdϕ = · ,
0 2 2 · 4 · 6 · · · (2n)
Math 213a (Fall 2021) Yum-Tong Siu 4
from
π
1 2π 2n 1 2π 1 iϕ 2n
Z Z Z
2
2n −iϕ
sin ϕdϕ = sin ϕdϕ = e −e dϕ
0 4 0 4 0 2i
Z 2n 2n
1 2π 1
n 2n 1 1 n 2n
= (−1) dϕ = 2π (−1) .
4 0 2i n 4 2i n
The purpose of the formula about the periodic time of the pendulum is to
show the order of the error involved when the trigonometric sine function is
used instead of the Jacobian elliptic sine function.
Inversion by Addition Formula by Abel. Abel in his 1827 paper used the
addition formula to invert the indefinite integral
Z ξ
−1 1
sn ξ = p .
x=0 (1 − x )(1 − k 2 x2 )
2
To understand the role of the additional theorem, let us consider first the
simpler case of trigonometric functions (also known as circular functions).
The inverse sine function is given by
Z ξ
−1 dx
θ = sin ξ = √ .
x=0 1 − x2
Because of monotonicity from the positivity of the derivative √ 1 , we can
1−ξ 2
π π
always invert for the range of −1 ≤ ξ ≤ 1 to get the range of θ in − 2 , 2 .
We know that the function θ = sin−1 ξ of the indefinite integral is not single-
valued.
For that reason, one would like to use the addition formula to express
sin(θ + ϕ) and sin(θ − ϕ) in terms of sin θ, cos θ, sin ϕ and cos ϕ. In 1827
Abel actually used this method to invert the indefinite integral which defines
an elliptic function and was the first person to succeed the inversion.
Math 213a (Fall 2021) Yum-Tong Siu 5
To understand how one can derive the addition formula from the indefinite
integral of the inverse function, one first looks at the simpler case of the
circular functions (i.e., the trigonometric functions). The addition formula
is
sin(θ + ϕ) = sin θ cos ϕ + cos θ sin ϕ
q p
= sin θ 1 − sin2 ϕ + 1 − sin2 θ sin ϕ.
Let ξ = sin θ and η = sin ϕ so that θ = sin−1 ξ and ϕ = sin−1 η. The addition
formula expressed in terms of ξ and η now becomes
sin−1 ξ + sin−1 η = θ + ϕ
q p
−1 2 2
= sin sin θ 1 − sin ϕ + 1 − sin θ sin ϕ
p p
= sin−1 ξ 1 − η 2 + 1 − ξ 2 η ,
The addition theorem tells us that the sum of two definite integrals is equal
to a third definite integral whose upper limit is expressed in terms of the
upper limits of the original two definite integrals. The upper term of each
integral is the variable for the inverse function in the addition formula.
more elegant. To help us understand how it works, let us first use the concept
of a Riemann surface to invert the circular function case of
Z ξ
−1 dx
θ = sin ξ = √ .
x=0 1 − x2
We already mentioned that the difficulty of choosing ξ > 1 is that we have
to go to complex numbers. So we consider the integration of the 1-form
dz
ω := √
1 − z2
with z ∈ C. The trouble is that the 1-form ω is multi-valued and, more
precisely, double-valued. Riemann’s idea to make it single-valued is to replace
a point z by two points above it to form a new structure so that a 1-form can
be defined on this new structure as single-valued. This new structure is a
Riemann
√ surface. What makes the 1-form ω double-valued is √its denominator
1 − z . We need only come up a new structure on which 1 − z 2 is single-
2
(with coordinate z). We are going to use this map to describe the projection
π : X → C To describe the map, we use the polar coordinates w = ρeiϕ for
Math 213a (Fall 2021) Yum-Tong Siu 7
Figure 1: Two red-colored edges identified and two blue-colored edges iden-
tified
the domain space but use the Cartesian coordinate z = x + iy for the target
space to get
1 1 1 1
x= ρ+ cos ϕ and y = ρ− sin ϕ.
2 ρ 2 ρ
The image of a fixed circle ρ = ρ0 > 1 in the w-plane traversed
in the
counter-
clockwise direction is an ellipse with semi-major axis 12 ρ0 + ρ10 and semi-
minor axis 12 ρ0 − ρ10 traversed in the counter-clockwise direction in the
z-plane. The image of a fixed circle ρ = ρ0 < 1 in the w-plane traversed
in
1 1
the counter-clockwise direction is an ellipse with semi-major axis 2 ρ0 + ρ0
1 1
and semi-minor axis 2 ρ0 − ρ0 traversed in the clockwise direction in the
z-plane.
The image of the fixed unit circle ρ = 1 in the w-plane traversed in the
counter-clockwise direction in the w-plane starting with w = 1 covers the
Math 213a (Fall 2021) Yum-Tong Siu 8
interval [−1, 1] twice in the z-plane, once from right to left and then once
from left to right. This description shows us that C−{0} (with coordinate w)
is precisely the Riemann surface X and the map z = 21 w + w1 is precisely
Now that we have seen the introduction of the notion of Riemann surface
and its role in inverting an indefinite integral the familiar setting of the
trigonometric sine function, we start to carry out the same procedure for the
more complicated new setting of the elliptic sine function.
Riemnn Surface for log z. With the branch-cut [0, ∞) the m-th branch of
log z with zeiϕ is defined with 2mπ < ϕ < 2(m + 1)π. The Riemann surface
X for log z is obtained by stacking up copies of C − [0, ∞) indexed by m ∈ Z
so that the lower edge of (0, ∞) of the m-th copy of C − [0, ∞) is glued with
the upper edge of (0, ∞) of the (m + 1)-th copy of of C − [0, ∞). On X the
logarithmic function log z is single-valued and holomorphic.
√
of w = 1 − z 2 in C2 . One disadvantage of using the graph is that it is in a
Math 213a (Fall 2021) Yum-Tong Siu 11
Any of the two branches take values of opposite sign on both edges of
each of the two branch-cuts. To get the Riemann surface X we can take two
copies of
−1 1
P1 − , −1 ∪ 1,
k k
−1
and join them by identifying
−1 the upper
edge of k
, −1 in one copy identified
with the lower edge of k , −1 in the other copy and at the same time
1
identifying
1the
upper edge of 1, k in one copy identified with the lower
edge of 1, k in the other copy.
Math 213a (Fall 2021) Yum-Tong Siu 13
Figure 4: To switch positions of two edges of horizonal slit along the real
axis, one rotates the plane with respect to the real axis so that the back
side of the plane becomes the front side. When the plane is compactified
to a sphere, the switching of its back side and front side means turning the
sphere inside out
Math 213a (Fall 2021) Yum-Tong Siu 15
Figure 8: Both loops of the fundamental group of the torus. First goes to
another sheet through one branch-cut (represented by dotted curve) and then
returns to the original sheet through another branch-cut.
Math 213a (Fall 2021) Yum-Tong Siu 19
1
The period from the second loop which goes around the line segment 1, k
(or equivalently around the line-segment − k1 , −1 ), all on one single sheet,
dz
p
(1 − z 2 )(1 − k 2 z 2 )
which we will integrate with some initial point to get a map and then use
the inverse of the map. The initial point we use is the point in X above 0
where the value of p
(1 − z 2 )(1 − k 2 z 2 )
is +1. p
After the construction of the Riemann surface X to make (1 − z 2 )(1 − k 2 z 2 )
single-valued, we have a well-defined 1-form
dz
ω := p .
(1 − z 2 )(1 − k 2 z 2 )
we obtain
dw p
= (1 − w2 )(1 − k 2 w2 ),
dz
√
which motivates the definition of the Jacobian elliptic function cn z = 1 − sn2 z,
with value 1 at√z = 0, analogous to the trigonometric functions. There is
another factor 1 − sn2 z without analogue in the theory of trigonometric
functions,
√ for which one introduces the Jacobian elliptic function dn z =
2 2
1 − k sn z, with value 1 at z = 0, known as the Jacobian delta amplitude
function.
d
The formula for the derivative of w = sn z is now simply dz sn z = cn z dn z
and is the same as the trigonometric case except for the factor dn z, which be-
comes 1√when the (elliptic) modulus
√ k degenerates to 0. From the definitions
cn z = 1 − sn z and dn z = 1 − k 2 sn2 z, both with value 1 at 0, it is not
2
Note that the variable w is in the lower limit of both definite integrals to
absorb the minus sign the derivatives of cn z and dn z. The choice of 1 in the
upper limit is to guarantee that cn 0 = 1 and dn 0 = 1.
Math 213a (Fall 2021) Yum-Tong Siu 22
We now discuss the addition theorem for the elliptic sine function.
Statement (i) clearly holds from sin 0 = 0 and cos 0 = 1. To check Statement
(ii), we let ξ = sin z and η = sin w with w = C − z. We use the notation
sin0 z = dz
d
sin z to rewrite
as
sin z sin0 w − sin0 z sin w = ξη 0 − ξ 0 η
so that
d 0
(sin z cos(C − z) + cos z sin(C − z)) = (ξη 0 − ξ 0 η) = ξη 00 − ξ 00 η,
dz
where the notation “prime” means differentiation with respect to z. State-
ment (ii) follows from ξ 00 = −ξ and η 00 = −η. In this proof of the addi-
tion formula for the trigonometric sine function is the identical vanishing of
(ξη 0 − ξ 0 η)0 .
Math 213a (Fall 2021) Yum-Tong Siu 23
For the case of the elliptic sine functions, with a complex constant C and
the setting of w = C − z and ξ = sn z and η = sn w, the crucial expression
(ξη 0 − ξ 0 η)0 is no longer identically zero. For the proof of the additional
formula for the trigonometric sine function, one can also use the identical
vanishing of
0
(log (ξη 0 − ξ 0 η))
instead of the identical vanishing of (ξη 0 − ξ 0 η)0 . It makes no difference in the
case of the proof of the additional formula for the trigonometric sine function,
but it makes all the difference in the exploration for the additional formula
for the elliptic sine function, because after reduction by using the differential
equations for the elliptic sine function, the non-identically-zero expression
0 ξη 00 − ξ 00 η
(log (ξη 0 − ξ 0 η)) =
ξη 0 − ξ 0 η
may show us how to modify the addition formula of the trigonometric sine
function to obtain the additional formula for the elliptic sine function.
and
2
η 0 = (1 − η 2 )(1 − k 2 η 2 ) = 1 − (1 + k 2 )η 2 + k 2 η 4
for the elliptic sine functions. Here again the notation “prime” means differ-
entiation with respect to z. To get the second-order derivatives of the elliptic
sine function, we can differentiate the first-order differential equation
2
df
= (1 − f 2 )(1 − k 2 f 2 )
dz
for the elliptic sine function f (z) = sn z once more with respect to z to get
2
df df df 2 2 2 df
2 2
= −2f (1 − k f ) + (1 − f ) −2k f .
dz dz dz dz
df
which, after the cancellation of 2 dz from both sides, is reduced to
d2 f
= −(1 + k 2 )f + 2k 2 f 3 .
dz 2
Math 213a (Fall 2021) Yum-Tong Siu 24
From
−ξη 00 + ηξ 00 = 2k 2 ξη(ξ 2 − η 2 ).
= ξ 2 − η 2 + k 2 ξ 2 η 2 (η 2 − ξ 2 ) = (ξ 2 − η 2 )(1 − k 2 ξ 2 η 2 )
so that
−ξη 00 + ηξ 00 −2k 2 ξη(ξ 0 η + ξη 0 )
= .
−ξη 0 + ηξ 0 1 − k2ξ 2η2
At this point the big surprise is that the numerator of the right-hand side is
equal to the derivative of the denominator of the right-hand side so that
0
1 − k 2 ξ 2 η 2 = −2k 2 ξη(ξ 0 η + ξη 0 .
log(−ξη 0 + ηξ 0 ) − log(1 − k 2 ξ 2 η 2 )
By integrating
0
log(−ξη 0 + ηξ 0 ) − log(1 − k 2 ξ 2 η 2 ) ≡ 0,
log(−ξη 0 + ηξ 0 ) − log(1 − k 2 ξ 2 η 2 )
ξη 0 − ηξ 0 sn z cn w dn w + sn w cn z dn z
= .
1 − k2ξ 2η2 1 − k 2 sn2 z sn2 w
To determine the constant, we set z = 0 and w = C − z = C to get
sn z cn w dn w + sn w cn z dn z sn 0 cn C dn C + sn C cn 0 dn 0
2 2 2
= = sn C
1 − k sn z sn w 1 − k 2 sn2 0 sn2 C
so that we end up with the addition formula
sn z cn w dn w + sn w cn z dn z
sn(z + w) =
1 − k 2 sn2 z sn2 w
for the elliptic sine function.
d2 = dn w. Then
= 1 + k 4 s41 s42 − s21 − s22 + 2s21 s22 − k 2 s21 s42 − k 2 s41 s22 + 2s1 s2 c1 c2 d1 d2
= (1 − s21 )(1 − s22 ) + s21 s22 (1 − k 2 s21 )(1 − k 2 s22 ) − 2s1 s2 c1 c2 d1 d2
= (c1 c2 − s1 s2 d1 d2 )2
to yield
c1 c2 − s 1 s 2 d 1 d 2
cn(z + w) = ± .
1 − k 2 s21 s22
The plus sign is determined by continuity and the evaluation at z = w = 0.
Of course, the addition theorems for the elliptic cosine and delta ampli-
tude functions can also be proved in the same way that we prove the addition
theorem for the elliptic sine function. We will leave such a proof for the el-
liptic cosine function to a homework problem.
when w1 = f (z1 ) and w2 = f (z2 ), because the above equation simply states
that
f −1 (w1 ) + f −1 (w2 ) = z1 + z2 = f −1 (f (z1 + z2 ))
= f −1 (Ξ(f (z1 ), f (z2 )) = f −1 (Ξ(w1 , w2 )) .
Math 213a (Fall 2021) Yum-Tong Siu 27
and
Z 1 Z 1
dx dx
p + p
x=ξ (x2 − 1)(1 − k 2 − x2 ) x=η (x2 − 1)(1 − k 2 − x2 )
Z 1
dx
= √
ξη− 12 (ξ2 −1)(1−k2 −ξ2 )(η 2 −1)(1−k2 −η 2 )
p .
k (x 2 − 1)(1 − k 2 − x2 )
x= 1
1− (1−ξ2 )(1−η 2 )
k2
Riemann Surfaces for Jacobian Elliptic Functions. Since the inverse func-
tion of the Jacobian elliptic function is defined by
Z 1
−1 dζ
cn w = p ,
ζ=w (1 − ζ )(1 − k 2 + k 2 ζ 2 )
2
(1 − ζ 2 )(1 − k 2 + k 2 ζ 2 )
q
2
which are ±1 and ± 1−k −k2
. To simplify expressions we introduce k 0 =
√ √
1 − k 2 . To make a choice of the sign of the square root k 0 = 1 − k 2
continuously, we take a branch-cut [−1, 1] with the value of 1 at k = 1.
Math 213a (Fall 2021) Yum-Tong Siu 28
to transform it to Z kw
1 dζ
q
k ζ=0 (1 − 1 2
ζ )(1 − ζ 2)
k2
and to end with the same elliptic sine function with rescaling in both the
independent and dependent variable. We come back to the indefinite integral
Z 1
−1 dζ
cn w = p ,
ζ=w (1 − ζ )(1 − k 2 + k 2 ζ 2 )
2
so that
r
0
p 1 ik 0
cn (K + iK ) = 1 − sn2 (K + iK 0 ) = 1 − 2 = −
k k
and r
p 0 1
dn (K + iK ) = 1 − k 2 sn2 (K + iK 0 ) = 1 − k2 = 0.
k2
0
cn (K + iK 0 ) = − ikk means that
Z 1
dζ
p = K + iK 0
0 2 2 2 2
(1 − ζ )(1 − k + k ζ )
ζ=− ikk
and one of the two primitive periods of cn z is 2K + 2iK 0 . Note that since
sn z is odd and has period 4K, we have
cn z cn w − sn z sn w dn z dn w
cn(z + w) =
1 − k 2 sn2 z sn2 w
Math 213a (Fall 2021) Yum-Tong Siu 30
and the other primitive period of cn z is 4iK 0 . Instead of the two primitive
periods 2K + 2iK 0 and 4iK 0 , one can also use 4K and 2K + i2iK 0 as periods.
Note that the lattice of periods Z(4K) + Z(2K + 2iK 0 ) of cn z is different
from the lattice Z(4K) + Z(2iK 0 ) of periods of sn, z.
Period
p Lattice of Delta Amplitude Function. The Riemann surface for
(ζ 2 − 1)(1 − k 2 − ζ 2 ) in the denominator of the indefinite integral
Z 1
−1 dζ
dn w = p
ζ=w (ζ 2 − 1)(1 − k 2 − ζ 2 )
are formed as quotients of ordered pairs from the three functions sin z, cos z, 1.
Analogously there are 12 possibilities of quotients of ordered triples from the
four functions sn z, cn z, dn z, 1. The naming convention is that the recip-
rocals of
sn z, cn z, dn z
are
ns z, nc z, nd z
with the two letters in the name of the original function reversed. The
quotient
sn z
cn z
is named sc z. The first letter of the name of the quotient function is the
first letter of the name of the numerator function. The second letter of the
name of the quotient function is the first letter of the name of the numerator
function.
sn K = 1, cn K = 0, dn K = k 0
ik 0
sn(K + iK 0 ) = 1, cn(K + iK 0 ) = − , dn(K + iK 0 ) = 0.
k
Addition formulas yield
1
sn(z + K + iK 0 ) = dc(z),
k
ik 0
cn(z + K + iK 0 ) = − nc(z),
k
0 0
dn(z + K + iK ) = ik sc(z).
Math 213a (Fall 2021) Yum-Tong Siu 35
Abel’s Elliptic Functions and Addition Formulas. For his elliptic function
ϕ(αi)
corresponding to the sine function, Abel used ϕ(α) which is given by y = i
with Z y
∂y
α= p
0 (1 − e y 2 )(1 + c2 y 2 )
2