18.

783 Elliptic Curves Fall 2023

Lecture #15 11/2/2023

15 Elliptic curves over C (part 2)

Last time we showed that every lattice L ⊆ C gives rise to an elliptic curve

EL : y 2 = 4x3 − g2 (L)x − g3 (L),

where X 1 X 1
g2 (L) = 60G4 (L) := 60 , g3 (L) = 140G6 (L) = 140 ,
∗
ω4 ∗
ω6
L L

with L∗ := L − {0}, and we defined a map

Φ : C/L → EL (C)
(
(℘(z), ℘0 (z)) z 6∈ L
z 7→
0 z∈L

where
1 X  1 1

℘(z) = ℘(z; L) = 2 + −
z ∗
(z − ω)2 ω 2
ω∈L

is the Weierstrass ℘-function for the lattice L, and

X 1
℘0 (z) = −2 .
(z − ω)3
ω∈L

In this lecture we will prove two theorems. First we will prove that Φ is an isomorphism
of additive groups; it is also an isomorphism of complex manifolds [3, Cor. 5.1.1], and of
complex Lie groups, but we won’t prove this right now.1 Second, we will prove that every
elliptic curve E/C is isomorphic to EL for some lattice L; this is the Uniformization Theorem.

15.1 The isomorphism from a torus to the corresponding elliptic curve

Theorem 15.1. Let L ⊆ C be a lattice and let EL : y 2 = 4x3 − g2 (L)x − g3 (L) be the
corresponding elliptic curve. The map Φ : C/L → EL (C) is a group isomorphism.

Proof. We first note that Φ(0) = 0, so Φ preserves the identity, and for all z 6∈ L we have

Φ(−z) = (℘(−z), ℘0 (−z)) = ℘(z), −℘0 (z) = −Φ(z),

since ℘ is even and ℘0 is odd, so Φ is compatible with taking inverses.

Let L = [ω1 , ω2 ]. There are three points of order 2 in C/L; if L = [ω1 , ω2 ] these
are ω1 /2, ω2 /2, and (ω1 + ω2 )/2. By Lemma 14.31, ℘0 vanishes these points, hence Φ
maps points of order 2 in C/L to points of order 2 in EL (C), since the latter are the
points with y-coordinate zero. Moreover, Φ is injective on points of order 2, since ℘(z)
maps each point of order 2 in C/L to a distinct root of 4℘(z)3 − g2 (L)℘(z) − g3 (L), as
shown in the proof of Lemma 14.33. The restriction of Φ to (C/L)[2] defines a bijection of
∼
(C/L)[2] −→ EL [2] ' Z/2Z ⊕ Z/2Z with Φ(0) = 0, which must be a group isomorphism.
1
This is not difficult to show, but it would distract us from our immediate goal. We will see an explicit
isomorphism of complex manifolds in a few lectures when we study modular curves, and in that case we will
take the time to define precisely what this means and to prove it.

Lecture by Andrew Sutherland

 To show that Φ is surjective, let (x0 , y0 ) ∈ EL (C). The elliptic function f (z) = ℘(z) − x0
has order 2, hence it has two zeros in the fundamental parallelogram F0 , by Theorem 14.18.
Neither of these zeros occurs at z = 0, since f has a pole at 0. So let z0 6= 0 be a zero of f (z)
in F0 . Then ℘(z0 ) = x0 , which implies Φ(z0 ) = (x0 , ±y0 ) and therefore (x0 , y0 ) = Φ(±z0 );
thus Φ is surjective.
We now show that Φ is injective. Let z1 , z2 ∈ F0 and suppose that Φ(z1 ) = Φ(z2 ).
If 2z1 ∈ L then z1 is a 2-torsion element and we have already shown that Φ restricts
to a bijection on (C/L)[2], so we must have z1 = z2 . We now assume 2z1 6∈ L, which
implies ℘0 (z1 ) 6= 0. As argued above, the roots of f (z) = ℘(z) − ℘(z1 ) in F0 are ±z1 ,
thus z2 ≡ ±z1 mod L. We also have ℘0 (z1 ) = ℘0 (z2 ), and this forces z2 ≡ z1 mod L, since
℘0 (−z1 ) = −℘0 (z1 ) 6= ℘0 (z1 ) because ℘0 (z1 ) 6= 0.
It remains only to show that Φ(z1 + z2 ) = Φ(z1 ) + Φ(z2 ). So let z1 , z2 ∈ F0 ; we may
assume that z1 , z2 , z1 + z2 6∈ L since the case where either z1 or z2 lies in L is immediate,
and if z1 + z2 ∈ L then z1 and z2 are inverses modulo L, a case treated above.
The points P1 = Φ(z1 ) and P2 = Φ(z2 ) are affine points in EL (C), and the line ` between
them cannot be vertical because P1 and P2 are not inverses (since z1 and z2 are not). So
let y = mx + b be an equation for this line, and let P3 be the third point where the line
intersects the curve EL . Then P1 +P2 +P3 = 0, by the definition of the group law on EL (C).
Now consider the function `(z) = −℘0 (z) + m℘(z) + b. It is an elliptic function of order 3
with a triple pole at 0, so it has three zeros in the fundamental region F0 , two of which are
z1 and z2 . Let z3 be the third zero in F0 . The point Φ(z3 ) lies on both the line ` and the
elliptic curve EL (C), hence it must lie in {P1 , P2 , P3 }; moreover, we have a bijection from
{z1 , z2 , z3 } to {Φ(z1 ), Φ(z2 ), Φ(z3 )} = {P1 , P2 , P3 }, and this bijection must send z3 to P3 if
P3 is distinct from P1 and P2 . If P3 coincides with exactly one of P1 or P2 , say P1 , then
`(z) has a double zero at z1 and we must have z3 = z1 ; and if P1 = P2 = P3 then clearly
z1 = z2 = z3 . Thus in every case we must have P3 = Φ(z3 ).
We have P1 + P2 + P3 = 0, so it suffices to show z1 + z2 + z3 ∈ L, since this implies

Φ(z1 + z2 ) = Φ(−z3 ) = −Φ(z3 ) = −P3 = P1 + P2 = Φ(z1 ) + Φ(z2 ).

Let Fα be a fundamental region for L whose boundary does not contain any zeros or
poles of `(z) and replace z1 , z2 , z3 by equivalent points in Fα if necessary.
Applying Theorem 14.17 to g(z) = z and f (z) = `(z) yields

`0 (z)
Z
1 X
z dz = ordw (`)w = z1 + z2 + z3 − 3 · 0 = z1 + z2 + z3 , (1)
2πi ∂Fα `(z)
w∈Fα

where the boundary ∂Fα of Fα is oriented counter-clockwise.

Let us now evaluate the integral in (1); to ease the notation, define f (z) := `0 (z)/`(z),
which we note is an elliptic function (hence periodic with respect to L). We then have
Z Z α+ω1 Z α+ω1 +ω2 Z α+ω2 Z α
zf (z) dz = zf (z)dz + zf (z)dz + zf (z)dz + zf (z)dz
∂Fα α α+ω1 α+ω1 +ω2 α+ω2
Zα+ω1 Z α+ω2 Z α Z α
= zf (z)dz + (z + ω1 )f (z)dz + (z + ω2 )f (z)dz + zf (z)dz
α α α+ω1 α+ω2
Z α+ω2 Z α
= ω1 f (z)dz + ω2 f (z)dz. (2)
α α+ω1

18.783 Fall 2023, Lecture #15, Page 2

Note that we have used the periodicity of f (z) to replace f (z + ωi ) by f (z), and to cancel
integrals in opposite directions along lines that are equivalent modulo L.
For any closed (not necessarily simple) curve C and a point z0 6∈ C, the quantity
Z
1 dz
2πi C z − z0
is the winding number of C about z0 , and it is an integer (it counts the number of times the
curve C “winds around” the point z0 ); see [1, Lem. 4.2.1] or [4, Lem. B.1.3].
The function `(α + tω2 ) parametrizes a closed curve C1 from `(α) to `(α + ω2 ) = `(α),
as t ranges from 0 to 1. The winding number of C1 about the point 0 is the integer
Z Z 1 0 Z α+w2 0 Z α+ω2
1 dz 1 ` (α + tω2 ) 1 ` (z) 1
c1 := = ω2 dt = dz = f (z)dz. (3)
2πi C1 z − 0 2πi 0 `(α + tω2 ) 2πi α `(z) 2πi α
Similarly, the function `(α + tω1 ) parameterizes a closed curve C2 from `(α) to `(α + ω1 ),
and we obtain the integer
Z Z 1 0 Z α+ω1 0 Z α+ω1
1 dz 1 ` (α + tω1 ) 1 ` (z) 1
c2 := = ω1 dt = dz = f (z) dz. (4)
2πi C2 z − 0 2πi 0 `(α + tω1 ) 2πi α `(z) 2πi α
Plugging (3), and (4) into (2), and applying (1), we see that
z1 + z2 + z3 = c1 ω1 − c2 ω2 ∈ L,
as desired.

15.2 The j-invariant of a lattice

Definition 15.2. The j-invariant of a lattice L is defined by
g2 (L)3 g2 (L)3
j(L) = 1728 = 1728 .
∆(L) g2 (L)3 − 27g3 (L)2
Recall that ∆(L) 6= 0, by Lemma 14.33, so j(L) is always defined.
The elliptic curve EL : y 2 = 4x3 − g2 (L)x − g3 (L) is isomorphic to the elliptic curve
y = x3 + Ax + B, where g2 (L) = −4A and g3 (L) = −4B. Thus
2

g2 (L)3 (−4A)3 4A3

j(L) = 1728 = 1728 = 1728 = j(EL ).
g2 (L)3 − 27g3 (L)2 (−4A)3 − 27(−4B)2 4A3 + 27B 2
Thus the j-invariant of a lattice L is the same as the j-invariant of the corresponding elliptic
curve EL . We now define the discriminant of an elliptic curve so that it agrees with the
discriminant of the corresponding lattice.
Definition 15.3. The discriminant of an elliptic curve E : y 2 = x3 + Ax + B is
∆(E) = −16(4A3 + 27B 2 ).
This definition applies to any elliptic curve E/k defined by a short Weierstrass equation,
whether k = C or not, but for the moment we continue to focus on elliptic curves over C.
Recall from Theorem 13.14 that elliptic curves E/k and E 0 /k are isomorphic over k̄ if
and only if j(E) = j(E 0 ). Thus over an algebraically closed field like C, the j-invariant
characterizes elliptic curves up to isomorphism. We now define an analogous notion of
isomorphism for lattices.

18.783 Fall 2023, Lecture #15, Page 3

Definition 15.4. Lattices L and L0 are said to be homothetic if L0 = λL for some λ ∈ C× .
Theorem 15.5. Two lattices L and L0 are homothetic if and only if j(L) = j(L0 ).
Proof. Suppose L and L0 are homothetic, with L0 = λL. Then
X 1 X 1
g2 (L0 ) = 60 4
= 60 4
= λ−4 g2 (L).
0∗
w ∗
(λω)
ω∈L ω∈L

Similarly, g3 (L0 ) = λ−6 g3 L, and we have

(λ−4 g2 (L))3 g2 (L)3

j(L0 ) = 1728 = 1728 = j(L).
(λ−4 g2 (L))3 − 27(λ−6 g3 (L))2 g2 (L)3 − 27g3 (L)2
To show the converse, let us now assume j(L) = j(L0 ). Let EL and EL0 be the corre-
sponding elliptic curves. Then j(EL ) = j(EL0 ). We may write

EL : y 2 = x3 + Ax + B,

with −4A = g2 (L) and −4B = g3 (L), and similarly for EL0 , with −4A0 = g2 (L0 ) and
−4B 0 = g3 (L0 ). By Theorem 13.13, there is a µ ∈ C× such that A0 = µ4 A and B 0 = µ6 B,
and if we let λ = 1/µ, then g2 (L0 ) = λ−4 g2 (L) = g2 (λL) and g3 (L0 ) = λ−6 g3 (L) = g3 (λL),
as above. We now show that this implies L0 = λL.
Recall from Theorem 14.29 that the Weierstrass ℘-function satisfies

℘0 (z)2 = 4℘(z)3 − g2 ℘(z) − g3 .

Differentiating both sides yields

2℘0 (z)℘00 (z) = 12℘(z)2 ℘0 (z) − g2 ℘0 (z)

g2
℘00 (z) = 6℘(z)2 − . (5)
2
By Theorem 14.28, the Laurent series for ℘(z; L) at z = 0 is
∞ ∞
1 X 1 X
℘(z) = 2 + (2n + 1)G2n+2 z 2n = 2 + an z 2n ,
z z
n=1 n=1

where a1 = g2 /20 and a2 = g3 /28.

Comparing coefficients for the z 2n term in (5), we find that for n ≥ 2 we have
n−1
!
X
(2n + 2)(2n + 1)an+1 = 6 ak an−k + 2an+1 ,
k=1

and therefore
n−1
6 X
an+1 = ak an−k .
(2n + 2)(2n + 1) − 12
k=1
This allows us to compute an+1 from a1 , . . . , an−1 , for all n ≥ 2. It follows that g2 (L) and
g3 (L) uniquely determine the function ℘(z) = ℘(z; L) (and therefore the lattice L where
℘(z) has poles), since ℘(z) is uniquely determined by its Laurent series expansion about 0.
Now consider L0 and λL, where we have g2 (L0 ) = g2 (λL) and g3 (L0 ) = g3 (λL). It follows
that ℘(z; L0 ) = ℘(z; λL) and L0 = λL, as desired.

18.783 Fall 2023, Lecture #15, Page 4

Corollary 15.6. Two lattices L and L0 are homothetic if and only if the corresponding
elliptic curves EL and EL0 are isomorphic.
Thus homethety classes of lattices correspond to isomorphism classes of elliptic curves over C,
and both are classified by the j-invariant. Recall from Theorem 13.12 that every complex
number is the j-invariant of an elliptic curve E/C. To prove the Uniformization Theorem
we just need to show that the same is true of lattices.

15.3 The j -function

Every lattice [ω1 , ω2 ] is homothetic to a lattice of the form [1, τ ], with τ in the upper half
plane H := {z ∈ C : im z > 0}; we may take τ = ±ω2 /ω1 with the sign chosen so that
im τ > 0. This leads to the following definition of the j-function.
Definition 15.7. The j-function j : H → C is defined by j(τ ) = j([1, τ ]). We similarly
define g2 (τ ) = g2 ([1, τ ]), g3 (τ ) = g3 ([1, τ ]), and ∆(τ ) = ∆([1, τ ]).
Note that for any τ ∈ H, both −1/τ and τ + 1 lie in H (the maps τ 7→ 1/τ and τ 7→ −τ
both swap the upper and lower half-planes; their composition preserves them).
Theorem 15.8. The j-function is holomorphic on H, and satisfies j(−1/τ ) = j(τ ) and
j(τ + 1) = j(τ ).
Proof. From the definition of j(τ ) = j([1, τ ]) we have
g2 (τ )3 g2 (τ )3
j(τ ) = 1728 = 1728 .
∆(τ ) g2 (τ )3 − 27g3 (τ )2
The series defining
X 1 X 1
g2 (τ ) = 60 and g3 (τ ) = 140
(m + nτ )4 (m + nτ )6
m,n∈Z m,n∈Z
(m,n)6=(0,0 (m,n)6=(0,0)

converge absolutely for any fixed τ ∈ H, by Lemma 14.22, and they converge uniformly over
τ in any compact subset of H. The proof of this last fact is straight-forward but slightly
technical; see [2, Thm. 1.15] for the details. It follows that g2 (τ ) and g3 (τ ) are holomorphic
on H, and therefore ∆(τ ) = g2 (τ )3 − 27g3 (τ )2 is also holomorphic on H. Since ∆(τ ) is
nonzero for all τ ∈ H, by Lemma 14.33, the j-function j(τ ) is holomorphic on H as well.
The lattices [1, τ ] and [1, −1/τ ] = −1/τ [1, τ ] are homothetic, and the lattices [1, τ + 1]
and [1, τ ] are equal; thus j(−1/τ ) = j(τ ) and j(τ + 1) = j(τ ), by Theorem 15.5.

15.4 The modular group

We now consider the modular group
  
a b
Γ = SL2 (Z) = : a, b, c, d ∈ Z, ad − bc = 1 .
c d
As proved in Problem Set 8, the group Γ acts on H via linear fractional transformations
 
a b aτ + b
τ= ,
c d cτ + d
and it is generated by the matrices S = 10 −1


0 and T = ( 10 11 ). This implies that the
j-function is invariant under the action of the modular group; in fact, more is true.

18.783 Fall 2023, Lecture #15, Page 5

Lemma 15.9. For any τ, τ 0 ∈ H we have j(τ ) = j(τ 0 ) if and only if τ 0 = γτ for some
γ ∈ Γ.

Proof. We have j(Sτ ) = j(−1/τ ) = j(τ ) and j(T τ ) = j(τ + 1) = j(τ ), by Theorem 15.8, It
follows that if τ 0 = γτ then j(τ 0 ) = j(τ ), since S and T generate Γ.
To prove the converse, let us suppose that j(τ ) = j(τ 0 ). Then by Theorem 15.5, the
lattices [1, τ ] and [1, τ 0 ] are homothetic So [1, τ 0 ] = λ[1, τ ], for some λ ∈ C× . There thus
exist integers a, b, c, and d such that

τ 0 = aλτ + bλ
1 = cλτ + dλ
1
From the second equation, we see that λ = cτ +d . Substituting this into the first, we have
 
aτ + b a b
τ0 = = γτ, where γ = ∈ Z2×2 .
cτ + d c d

Now [1, τ 0 ] = λ[1, τ ] implies im τ 0 = |λ|2 im τ , since τ, τ 0 ∈ H and fundamental parallelograms

for [1, τ 0 ] = λ[1, τ ] must have the same area. But we also have
 

0 aτ + b im (aτ + b)(cτ̄ + d) (ad − bc) im τ
im τ = im(γτ ) = im = 2
= = (det γ)|λ|2 im τ,
cτ + d |cτ + d| |cτ + d|2

and therefore det γ = 1 and γ ∈ SL2 (Z).

Lemma 15.9 implies that when studying the j-function it suffices to study its behavior on
Γ-equivalence classes of H, that is, the orbits of H under the action of Γ. We thus consider
the quotient of H modulo Γ-equivalence, which we denote by H/Γ.2 The actions of γ and
−γ are identical, so taking the quotient by PSL2 (Z) = SL2 (Z)/{±1} yields the same result,
but for the sake of clarity we will stick with Γ = SL2 (Z).
We now wish to determine a fundamental domain for H/Γ, a set of unique representatives
in H for each Γ-equivalence class. For this purpose we will use the set

F = {τ ∈ H : re(τ ) ∈ [−1/2, 1/2) and |τ | ≥ 1, such that |τ | > 1 if re(τ ) > 0}.

Lemma 15.10. The set F is a fundamental domain for H/Γ.

Proof. We need to show that for every τ ∈ H, there is a unique τ 0 ∈ F such
that τ 0 = γτ ,

for some γ ∈ Γ. We first prove existence. Let us fix τ ∈ H. For any γ = ac db ∈ Γ we have
 

aτ + b im (aτ + b)(cτ̄ + d) (ad − bc) im τ im τ
im(γτ ) = im = 2
= 2
= (6)
cτ + d |cτ + d| |cτ + d| |cτ + d|2

Let cτ + d be a shortest vector in the lattice [1, τ ]. Then c and d must be relatively
prime,
and we can pick integers a and b so that ad − bc = 1. The matrix γ0 = ac db then
maximizes the value of im(γτ ) over γ ∈ Γ. Let us now choose γ = T k γ0 , where k is chosen
so that re(γτ ) ∈ [1/2, 1/2), and note that im(γτ ) = im(γ0 τ ) remains maximal. We must
have |γτ | ≥ 1, since otherwise im(Sγτ ) > im(γτ ), contradicting the maximality of im(γτ ).
2
Some authors write this quotient as Γ\H to indicate that the action is on the left.

18.783 Fall 2023, Lecture #15, Page 6

 i∞

ρ i

-1 -1/2 0 1/2 1

Figure 1: Fundamental domain F for H/Γ, with i = eπ/2 and ρ = e2πi/3 .

Finally, if τ 0 = γτ 6∈ F, then we must have |γτ | = 1 and re(γτ ) > 0, in which case we
replace γ by Sγ so that τ 0 = γτ ∈ F.
It remains to show that τ 0 is unique. This is equivalent to showing that any two Γ-
equivalent
points in F must coincide. So let τ1 and τ2 = γ1 τ1 be two elements of F, with
γ1 = ac db , and assume im τ1 ≤ im τ2 . By (6), we must have |cτ1 + d|2 ≤ 1, thus

1 ≥ |cτ1 + d|2 = (cτ1 + d)(cτ̄1 + d) = c2 |τ1 |2 + d2 + 2cd re τ1 ≥ c2 |τ1 |2 + d2 − |cd| ≥ 1,

where the last inequality follows from |τ1 | ≥ 1 and the fact that c and d cannot both be zero
(since det γ = 1). Thus |cτ1 + d| = 1, which implies im τ2 = im τ1 . We also have |c|, |d| ≤ 1,
and by replacing γ1 by −γ1 if necessary, we may assume that c ≥ 0. This leaves 3 cases:

1. c = 0: then |d| = 1 and a = d. So τ2 = τ1 ± b, but | re τ2 − re τ1 | < 1, so τ2 = τ1 .

2. c = 1, d = 0: then b = −1 and |τ1 | = 1. So τ1 is on the unit circle and τ2 = a − 1/τ1 .
Either a = 0 and τ2 = τ1 = i, or a = −1 and τ2 = τ1 = ρ.
√
3. c = 1, |d| = 1: then |τ1 + d| = 1, so τ1 = ρ, and im τ2 = im τ1 = 3/2 implies τ2 = ρ.

In every case we have τ1 = τ2 as desired.

Theorem 15.11. The restriction of the j-function to F defines a bijection from F to C.

Proof. Injectivity follows immediately from Lemmas 15.9 and 15.10. It remains to prove
surjectivity. We have
 
∞
X 1  X 1 X 1 
g2 (τ ) = 60 4
= 60 2
4
+ 4
.
(m + nτ )  m (m + nτ ) 
n,m∈Z m=1 n,m∈Z
(m,n)6=(0,0) n6=0

The second sum tends to 0 as im τ → ∞. Thus we have

∞
X π4 4π 4
lim g2 (τ ) = 120 m−4 = 120 ζ(4) = 120 = ,
imτ →∞ 90 3
m=1

18.783 Fall 2023, Lecture #15, Page 7

where ζ(s) is the Riemann zeta function. Similarly,

π6 8π 6
lim g3 (τ ) = 280 ζ(6) = 280 = .
imτ →∞ 945 27
Thus  3  2
4 4 8 6
lim ∆(τ ) = π − 27 π = 0.
imτ →∞ 3 27
(this explains the coefficients 60 and 140 in the definitions of g2 and g3 ; they are the smallest
pair of integers that ensure this limit is 0). Since ∆(τ ) is the denominator of j(τ ), the
quantity j(τ ) = g2 (τ )3 /∆(τ ) is unbounded as im τ → ∞.
In particular, the j-function is non-constant, and by Theorem 15.8 it is holomorphic on H.
The open mapping theorem implies that j(H) is an open subset of C; see [4, Thm. 3.4.4].
We claim that j(H) is also a closed subset of C. Let j(τ1 ), j(τ2 ), . . . be an arbitrary
convergent sequence in j(H), converging to w ∈ C. The j-function is Γ-invariant, by
Lemma 15.9, so we may assume the τn all lie in F. The sequence im τ1 , im τ2 , . . . must
be bounded, say be B, since j(τ ) → ∞ as im τ → ∞, but the sequence j(τ) , j(τ2 ), . . .
converges; it follows that the τn all lie in the compact set

Ω = {τ : re τ ∈ [−1/2, 1/2], im τ ∈ [1/2, B]}.

There is thus a subsequence of the τn that converges to some τ ∈ Ω ⊂ H. The j-function is

holomorphic, hence continuous, so j(τ ) = w. It follows that the open set j(H) contains all
its limit points and is therefore closed.
The fact that the non-empty set j(H) ⊆ C is both open and closed implies that j(H) = C,
since C is connected. It follows that j(F) = C, since every element of H is Γ-equivalent to
an element of F (Lemma 15.10) and the j-function is Γ-invariant (Lemma 15.9).

Corollary 15.12 (Uniformization Theorem). For every elliptic curve E/C there exists a
lattice L such that E = EL .

Proof. Given E/C, pick τ ∈ H so that j(τ ) = j(E) and let L0 = [1, τ ]. We have

j(E) = j(τ ) = j(L0 ) = j(EL0 ),

so E is isomorphic to EL0 , by Theorem 13.13, where the isomorphism is given by the map
(x, y) 7→ (µ2 x, µ3 y) for some µ ∈ C× . If now let L = µ1 L0 , then E = EL .

References
[1] Lars V. Ahlfors, Complex analysis, third edition, McGraw Hill, 1979.

[2] Tom M. Apostol, Modular functions and Dirichlet series in number theory, second edi-
tion, Springer, 1990.

[3] Joseph H. Silverman, The arithmetic of elliptic curves, second edition, Springer 2009.

[4] Elias M. Stein and Rami Shakarchi, Complex analysis, Princeton University Press, 2003.

18.783 Fall 2023, Lecture #15, Page 8