25.03.2013 Views

Shimura lifts of half-integral weight modular forms - Department of ...

Shimura lifts of half-integral weight modular forms - Department of ...

Shimura lifts of half-integral weight modular forms - Department of ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

SHIMURA LIFTS OF HALF-INTEGRAL WEIGHT MODULAR FORMS<br />

ARISING FROM THETA FUNCTIONS<br />

DAVID HANSEN AND YUSRA NAQVI<br />

Abstract. In 1973, <strong>Shimura</strong> [8] introduced a family <strong>of</strong> correspondences between <strong>modular</strong><br />

<strong>forms</strong> <strong>of</strong> <strong>half</strong>-<strong>integral</strong> <strong>weight</strong> and <strong>modular</strong> <strong>forms</strong> <strong>of</strong> even <strong>integral</strong> <strong>weight</strong>. Earlier, in unpublished<br />

work, Selberg explicitly computed a simple case <strong>of</strong> this correspondence pertaining<br />

to those <strong>half</strong>-<strong>integral</strong> <strong>weight</strong> <strong>forms</strong> which are products <strong>of</strong> Jacobi’s theta function and level<br />

one Hecke eigen<strong>forms</strong>. Cipra [1] generalized Selberg’s work to cover the <strong>Shimura</strong> <strong>lifts</strong><br />

where the Jacobi theta function may be replaced by theta functions attached to Dirichlet<br />

characters <strong>of</strong> prime power modulus, and where the level one Hecke eigen<strong>forms</strong> are replaced<br />

by more generic new<strong>forms</strong>. Here we generalize Cipra’s results further to cover theta functions<br />

<strong>of</strong> arbitrary Dirichlet characters multiplied by more general Hecke eigen<strong>forms</strong>, and<br />

we use these explicit formulas to compute optimal levels for these <strong>lifts</strong> without appealing<br />

to <strong>Shimura</strong>’s deeper arguments.<br />

1. Introduction and statement <strong>of</strong> results<br />

Let SL2(Z) denote the set <strong>of</strong> all 2-by-2 matrices with integer entries and determinant<br />

1, and let k be a positive integer. We say that f(z) is a <strong>modular</strong> form <strong>of</strong> <strong>weight</strong> k on the<br />

congruence subgroup Γ0(N) with multiplier ψ if f(z) is a holomorphic function on the upper<br />

<strong>half</strong> <strong>of</strong> the complex plane which satisfies f(γz) = (cz + d) kψ(d)f(z) for all γ = <br />

a b<br />

c d ∈<br />

SL2(Z) with c ≡ 0 (mod N). Let Mk(N, ψ) denote the finite-dimensional vector space <strong>of</strong><br />

<strong>modular</strong> <strong>forms</strong> <strong>of</strong> <strong>weight</strong> k on Γ0(N) with multiplier ψ, where ψ is a Dirichlet character <strong>of</strong><br />

modulus N. A <strong>modular</strong> form is called a cusp form if it vanishes at all rational points and<br />

at infinity. We let Sk(N, ψ) denote the subspace <strong>of</strong> Mk(N, ψ) consisting only <strong>of</strong> cusp <strong>forms</strong>.<br />

For k ≥ 2 a positive even integer, we define the Eisenstein series <strong>of</strong> <strong>weight</strong> k by<br />

(1.1) Ek(z) := 1 − 2k<br />

Bk<br />

n=1<br />

∞<br />

σk−1(n)q n ,<br />

where Bn in the nth Bernoulli number and q := e 2πinz . These functions represent the<br />

simplest <strong>modular</strong> <strong>forms</strong> <strong>of</strong> <strong>weight</strong> k, and they lie in Mk(1, 1). More general Eisenstein<br />

series can be defined as follows. Let χ1 (mod a1) and χ2 (mod a2) be primitive Dirichlet<br />

characters <strong>of</strong> conductors a1, a2, not both trivial, and set a = a1a2. The character χ = χ1χ2<br />

has modulus a. If k is positive integer with χ(−1) = (−1) k , then set<br />

(1.2) Ek(χ1, χ2; z) := C(k, χ1, χ2) +<br />

∞ <br />

n=1<br />

1<br />

d|n<br />

χ1(n/d)χ2(d)d k−1<br />

q n ,


2 DAVID HANSEN AND YUSRA NAQVI<br />

where C(k, χ1, χ2) is zero unless a1 = 1, in which case C(k, χ1, χ2) = 1<br />

2L(1 − k, χ),<br />

where L(s, χ) = ∞ n=1 χ(n)n−s is the Dirichlet L-function <strong>of</strong> the character χ. We have<br />

Ek(χ1, χ2; z) ∈ Mk(a, χ); see Chapter 4 <strong>of</strong> [4].<br />

In a classic paper [8], <strong>Shimura</strong> invented the modern theory <strong>of</strong> <strong>modular</strong> <strong>forms</strong> <strong>of</strong> <strong>half</strong><strong>integral</strong><br />

<strong>weight</strong>. Briefly, let N, k be positive integers with ψ a Dirichlet character <strong>of</strong> modulus<br />

4N. We say that f is a <strong>modular</strong> form <strong>of</strong> <strong>weight</strong> k + 1/2 with multiplier ψ if<br />

<br />

c<br />

2k+1 (1.3) f(γz) = ψ(d) ɛ<br />

d<br />

−2k−1<br />

d (cz + d) k+1/2 f(z)<br />

for all γ ∈ Γ0(4N), where ɛd is 1 or i for odd d according to whether d ≡ 1 (mod 4) or<br />

d ≡ 3 (mod 4), respectively, and <br />

c<br />

d is <strong>Shimura</strong>’s extension <strong>of</strong> the Jacobi symbol. As above,<br />

Mk+1/2(N, ψ) denotes the finite-dimensional vector space <strong>of</strong> <strong>weight</strong> k + 1/2 <strong>modular</strong> <strong>forms</strong>,<br />

and Sk+1/2(N, ψ) denotes its subspace <strong>of</strong> cusp <strong>forms</strong>.<br />

Define theta functions<br />

(1.4) θ(χ; z) := <br />

χ(n)n ν q n2<br />

∈ M1/2+ν(4r 2 , χχ ν 4)<br />

n∈Z<br />

for χ a Dirichlet character <strong>of</strong> modulus r, where ν = 0, 1 is chosen such that χ(−1) = (−1) ν .<br />

These functions are the simplest examples <strong>of</strong> <strong>modular</strong> <strong>forms</strong> <strong>of</strong> <strong>half</strong>-<strong>integral</strong> <strong>weight</strong>, and for<br />

k = 1/2, the space is spanned by them (c.f. [7]). For a good introduction to this material,<br />

see [6].<br />

<strong>Shimura</strong> also established a family <strong>of</strong> nontrivial maps between <strong>modular</strong> <strong>forms</strong> <strong>of</strong> <strong>half</strong><strong>integral</strong><br />

<strong>weight</strong> and <strong>modular</strong> <strong>forms</strong> <strong>of</strong> even integer <strong>weight</strong>. These maps, known as the<br />

<strong>Shimura</strong> <strong>lifts</strong>, can be stated as follows.<br />

<br />

Theorem (<strong>Shimura</strong>). Let t be a positive squarefree integer, and suppose that f(z) =<br />

∞<br />

n=1 b(n)qn ∈ Sk+1/2(4N, ψ), where k is a positive integer. If numbers A(n) are defined by<br />

∞<br />

(1.5)<br />

A(n)n −s := L(s − k + 1, ψχ k ∞<br />

4χt) b(tn 2 )n −s ,<br />

n=1<br />

where χt = <br />

t<br />

• is the usual Kronecker character modulo t, then St(f)(z) := ∞ n=1 A(n)qn ∈<br />

M2k(2N, ψ2 ). Moreover, if k > 1, then St(f)(z) is a cusp form.<br />

<strong>Shimura</strong> <strong>lifts</strong> play an important role in several areas <strong>of</strong> modern number theory, including<br />

Tunnell’s famous work [9] on the ancient congruent number problem, and recent work by<br />

Ono [5] on congruences for the partition function. Moreover, in these particular applications,<br />

the relevant <strong>half</strong>-<strong>integral</strong> <strong>weight</strong> <strong>forms</strong> can be written as products <strong>of</strong> integer <strong>weight</strong> <strong>forms</strong><br />

and theta functions. In light <strong>of</strong> these facts, it is desirable to have explicit formulas for the<br />

<strong>Shimura</strong> <strong>lifts</strong> in these cases.<br />

It turns out that much earlier, in unpublished work, Selberg worked out such an explicit<br />

formula. Briefly, for certain <strong>modular</strong> <strong>forms</strong> f(z) ∈ Mk(1, 1), Selberg found that<br />

f(4z)θ(1; z) ∈ M k+1/2(4, 1) <strong>lifts</strong> to f(z) 2 − 2 k−1 f(2z) 2 ∈ M2k(2, 1). Later on, Cipra [1]<br />

generalized Selberg’s work by proving the following result.<br />

n=1


SHIMURA LIFTS OF MODULAR FORMS WITH THETA FUNCTIONS 3<br />

Theorem (Cipra). If f(z) ∈ Sk(N, ψ) is a newform, and θ(χr; z) is the theta function<br />

<strong>of</strong> an even Dirichlet character <strong>of</strong> prime power modulus r = p m , then if we define g(z) :=<br />

f(z)f(p µ z), the <strong>Shimura</strong> lift S1 <strong>of</strong> f(4p µ z)θ(χr; z) is<br />

(1.6) gχr(z) − 2 k−1 χr(2)ψ(2)gχr(2z),<br />

where µ is any integer with µ ≥ m.<br />

Cipra also proves a similar statement for theta functions with odd characters. However,<br />

Cipra’s class <strong>of</strong> eligible <strong>forms</strong> f(z) is limited to new<strong>forms</strong>, and his use <strong>of</strong> theta functions with<br />

characters to prime power moduli is a highly restrictive condition. We prove the following<br />

two theorems, generalizing these results.<br />

Theorem 1.1. Let χr be an even Dirichlet character modulo r, and write χr = χ p α 1<br />

1 χ p α 2<br />

2 ...χ p α j<br />

j<br />

as the factorization <strong>of</strong> χr into Dirichlet characters modulo prime powers p α1<br />

1<br />

, pα2<br />

2<br />

, ..., pαj<br />

j<br />

with p α1<br />

1 pα2<br />

2 ...pαj<br />

j = r. Let f(z) ∈ Mk(N, ψ) be a Hecke eigenform, and set F (z) :=<br />

θ(χr; z)f(4rz) ∈ M k+1/2(4N ′ r 2 , ψχrχ k 4 ) with N ′ = N/ gcd(N, r). If<br />

(1.7) g(z) := <br />

where χd = <br />

p αj j ||d χ p αj j<br />

, then we have<br />

d|r<br />

gcd(d,r/d)=1<br />

f(dz)f(rz/d)χd(−1),<br />

(1.8) S1(F )(z) = gχr(z) − 2 k−1 χr(2)ψ(2)gχr(2z) ∈ M2k(2N ′ r 2 , ψ 2 χ 2 r).<br />

Here gχ is the χ-twist <strong>of</strong> g.<br />

For the case <strong>of</strong> odd characters, the theorem is slightly different, due to the fact that the<br />

relevant theta functions now have <strong>weight</strong> 3/2.<br />

Theorem 1.2. Let χr be an odd Dirichlet character modulo r, and write χr = χ p α 1<br />

1 χ p α 2<br />

2 ...χ p α j<br />

j<br />

as the factorization <strong>of</strong> χr into Dirichlet characters modulo prime powers p α1<br />

1<br />

with p α1<br />

1 pα2<br />

2<br />

, pα2<br />

2<br />

, · · · , pαj<br />

j<br />

· · · pαj<br />

j = r. If F (z) := θ(χr; z)f(4rz) ∈ M k+3/2(4N ′ r 2 , ψχrχ k+1<br />

4 ), where<br />

f(z) ∈ Mk(N, ψ) is a Hecke eigenform, and<br />

(1.9) g(z) := 1<br />

πi<br />

<br />

where χd = <br />

p αj j ||d χ p αj j<br />

, then we have<br />

d|r<br />

gcd(d,r/d)=1<br />

df ′ (dz)f(rz/d)χd(−1),<br />

(1.10) S1(F )(z) = gχr(z) − 2 k χr(2)ψ(2)gχr(2z) ∈ M2k+2(2N ′ r 2 , ψ 2 χ 2 r),<br />

where gχ is the χ-twist <strong>of</strong> g.<br />

The pro<strong>of</strong>s <strong>of</strong> our theorems, like those <strong>of</strong> Selberg and Cipra, are entirely combinatorial,<br />

using only elementary properties <strong>of</strong> Dirichlet series and a multiplicativity relation for the<br />

coefficients <strong>of</strong> our starting form f(z). This multiplicativity is conditional on f(z) being<br />

a Hecke eigenform. However, since any given <strong>modular</strong> form can be written as a linear<br />

combination <strong>of</strong> eigen<strong>forms</strong>, our theorems can be applied to more general products <strong>of</strong> <strong>modular</strong>


4 DAVID HANSEN AND YUSRA NAQVI<br />

<strong>forms</strong> and theta functions by the linearity <strong>of</strong> the <strong>Shimura</strong> lift. Furthermore, we compute the<br />

levels <strong>of</strong> the <strong>lifts</strong> in Theorems 1.1 and 1.2 directly, without appealing to any <strong>of</strong> <strong>Shimura</strong>’s<br />

results. In fact, our theorems are completely independent <strong>of</strong> <strong>Shimura</strong>’s work.<br />

In Section 2, we define and explain the notion <strong>of</strong> a Hecke eigenform and the associated<br />

multiplicativity relations for its coefficients. In Section 3, we present pro<strong>of</strong>s <strong>of</strong> Theorems<br />

1.1 and 1.2, and we discuss a method <strong>of</strong> determining the cuspidality <strong>of</strong> the <strong>lifts</strong> given by<br />

these theorems. We also show how to obtain the optimal level for the lifted <strong>forms</strong>. Section<br />

4 contains a discussion <strong>of</strong> examples and applications.<br />

2. Multiplicativity Properties <strong>of</strong> Modular Form Coefficients<br />

Let f(z) = ∞<br />

n=0 a(n)qn ∈ Mk(N, ψ). The action <strong>of</strong> the nth Hecke operator T ψ n <strong>of</strong> <strong>weight</strong><br />

k on f(z) is given by<br />

(2.1) f(z) | T ψ n =<br />

∞ <br />

m=0<br />

d|(m,n)<br />

ψ(d)d k−1 a(mn/d 2 <br />

) q m .<br />

It is known, by work <strong>of</strong> Hecke, that these operators are linear endomorphisms on Mk(N, ψ).<br />

They also map cusp <strong>forms</strong> to cusp <strong>forms</strong>. Furthermore, there exist <strong>modular</strong> <strong>forms</strong> f(z) ∈<br />

Mk(N, ψ) which are simultaneous eigenfunctions <strong>of</strong> all the Hecke operators; in other words,<br />

they satisfy<br />

(2.2) f(z) | T ψ n = λ(n)f(z)<br />

for all positive integers n, where the λ(n) are complex numbers. If f(z) satisfies these<br />

conditions, we generally refer to it as a Hecke eigenform. In this case, by combining (2.1)<br />

and (2.2), we easily get<br />

(2.3) λ(n)a(m) = <br />

d|(m,n)<br />

ψ(d)d k−1 a(mn/d 2 ).<br />

If a(1) = 1, then this reveals that in fact λ(n) = a(n), and we can then reformulate (2.3)<br />

as follows.<br />

Proposition 2.1. If f(z) = ∞<br />

n=0 a(n)qn ∈ Mk(N, χ) is a simultaneous eigenfunction <strong>of</strong><br />

all the Hecke operators T ψ n with a(1) = 1, then for any positive integers m, n, we have<br />

a(m)a(n) = <br />

d|(m,n)<br />

ψ(d)d k−1 a(mn/d 2 ).


SHIMURA LIFTS OF MODULAR FORMS WITH THETA FUNCTIONS 5<br />

For the Eisenstein series defined in the introduction (see 1.2), we can prove this directly<br />

as follows:<br />

<br />

d|(m,n)<br />

= <br />

d|(m,n)<br />

= <br />

d|(m,n)<br />

d 2 δ|mn<br />

= <br />

d|(m,n)<br />

d 2 δ|mn<br />

χ(d)d k−1 a(mn/d 2 )<br />

<br />

k−1<br />

χ(d)d<br />

δ|mn/d 2<br />

χ1(mn/(d 2 δ))χ2(δ)δ k−1<br />

χ1(d)χ2(d)χ1(mn/(d 2 δ))χ2(δ)d k−1 δ k−1<br />

χ1(mn/(dδ))χ2(dδ)(dδ) k−1<br />

= <br />

χ1(mn/(dδ))χ2(dδ)(dδ) k−1<br />

d|m<br />

δ|n<br />

= a(m)a(n).<br />

Furthermore, we have an “inverse” <strong>of</strong> Proposition 2.2, which we shall refer to as Selberg<br />

inversion.<br />

Proposition 2.2. If f(z) = ∞ n=0 a(n)qn ∈ Mk(N, χ) is a Hecke eigenform with a(1) = 1,<br />

then we have<br />

a(mn) = <br />

µ(d)χ(d)d k−1 a(m/d)a(n/d),<br />

for any positive integers m, n.<br />

Pro<strong>of</strong>. We have that<br />

<br />

d|(m,n)<br />

= <br />

d|(m,n)<br />

= <br />

dδ|(m,n)<br />

= <br />

D|(m,n)<br />

= a(mn).<br />

d|(m,n)<br />

µ(d)χ(d)d k−1 a(m/d)a(n/d)<br />

µ(d)χ(d)d k−1<br />

<br />

δ|(m/d,n/d)<br />

µ(d)χ(dδ)(dδ) k−1 a(mn/(dδ) 2 )<br />

<br />

d|D<br />

<br />

µ(d) χ(D)D k−1 a(mn/D 2 )<br />

χ(δ)δ k−1 a(mn/(dδ) 2 )


6 DAVID HANSEN AND YUSRA NAQVI<br />

3. Pro<strong>of</strong>s <strong>of</strong> Theorems 1.1 and 1.2<br />

We begin by presenting the pro<strong>of</strong> <strong>of</strong> the formula for the lift in Theorem 1.1. From the<br />

definition <strong>of</strong> F (z), we have F (z) = ∞ n=0 b(n)qn with<br />

(3.1) b(n) = <br />

n − m2 <br />

χr(m)a ,<br />

4r<br />

m∈Z<br />

where f(z) = ∞ n=0 a(n)qn is as in the statement <strong>of</strong> Theorem 1.1. As above, the <strong>Shimura</strong><br />

lift is given by<br />

∞<br />

(3.2) S1(F ) = A(n)q n<br />

with the coefficients A(n) defined by<br />

(3.3)<br />

∞<br />

n=1<br />

n=1<br />

A(n)n −s = L(s − k + 1, χrψχ 2k<br />

4 )<br />

We also need the coefficients defined by<br />

∞<br />

(3.4)<br />

cd(n)q n := f(dz)f(rz/d) =<br />

n=0<br />

n=0 m∈Z<br />

∞<br />

b(n 2 )n −s .<br />

n=1<br />

∞ <br />

n − dm<br />

<br />

a(m)a q<br />

r/d<br />

n .<br />

Throughout the pro<strong>of</strong>, we use the convention that a <strong>modular</strong> form coefficient is zero if its<br />

argument is negative or not <strong>integral</strong>. From (3.1) it is easy to see that<br />

(3.5) b(n 2 ) = <br />

(n − m)(n + m)<br />

<br />

χr(m)a<br />

.<br />

4r<br />

m∈Z<br />

This is a finite sum with non-zero coefficients whenever (n − m)(n + m)/(4r) ∈ N. Note<br />

that n + m and n − m must both be even for n and m to be integers with 4|(n 2 − m 2 ). Let<br />

gcd((n − m)/2, r) = d. Thus, m ≡ n (mod 2d) and m ≡ −n (mod 2r/d). Now suppose<br />

gcd(d, r/d) = d ′ > 1. This implies that m ≡ n ≡ −n ≡ 0 (mod 2d ′ ), so d| gcd(m, r) and<br />

so χr(m) = 0. Therefore, we only consider the cases in which gcd(d, r/d) = 1. We have<br />

m = n + 2dm ′ for some m ′ ∈ Z, so n − m = −2dm ′ and n + m = 2n + 2dm ′ . Thus,<br />

(3.6)<br />

(n − m)(n + m)<br />

4r<br />

= −m′ (n + dm ′ )<br />

.<br />

r/d<br />

Also, since m ≡ n (mod d) and m ≡ −n (mod r/d), we have that<br />

= χd(m)χ r/d(m)<br />

= χ r/d(−1)χd(n)χ r/d(n)<br />

χr(m)<br />

= χd(n)χ r/d(−n)<br />

= χ r/d(−1)χr(n),


SHIMURA LIFTS OF MODULAR FORMS WITH THETA FUNCTIONS 7<br />

where the characters are as defined in the statement <strong>of</strong> Theorem 1.1. Since χr(−1) =<br />

χ r/d(−1)χd(−1) = 1, we have χd(−1) = χ r/d(−1) = ±1. Thus, by changing the variable<br />

m ′ to −m, (3.5) becomes<br />

(3.7) b(n 2 ) = χr(n) <br />

d|r<br />

gcd(d,r/d)=1<br />

We now apply Proposition 2.2 to get<br />

b(n 2 ) = χr(n)<br />

<br />

χd(−1) <br />

= χr(n)<br />

= χr(n)<br />

d|r<br />

gcd(d,r/d)=1<br />

<br />

d|r<br />

gcd(d,r/d)=1<br />

<br />

d|r<br />

gcd(d,r/d)=1<br />

χd(−1) <br />

χd(−1) <br />

m(n − dm)<br />

<br />

a<br />

.<br />

r/d<br />

<br />

m∈Z δ|(m,n)<br />

δ|n<br />

m∈Z<br />

µ(δ)ψ(δ)δ k−1 <br />

m<br />

<br />

n − dm<br />

<br />

a a<br />

δr/d δr/d<br />

<br />

k−1<br />

µ(δ)ψ(δ)δ<br />

m∈Z<br />

χd(−1) <br />

µ(δ)ψ(δ)δ k−1 cd(n/δ).<br />

Rewriting these formulas as Dirichlet series immediately gives<br />

(3.8)<br />

∞<br />

b(n 2 )n −s =<br />

<br />

∞ <br />

χd(−1)<br />

n=1<br />

n=1<br />

d|r<br />

gcd(d,r/d)=1<br />

δ|n<br />

n=1 δ|n<br />

<br />

m<br />

<br />

a a<br />

r/d<br />

<br />

n/δ − dm<br />

<br />

r/d<br />

µ(δ)ψ(δ)χr(δ)δ k−1 χr(n/δ)cd(n/δ)n −s ,<br />

and we can easily pull out the reciprocal <strong>of</strong> a Dirichlet L-function to produce<br />

(3.9)<br />

∞<br />

b(n 2 )n −s 1<br />

=<br />

L(s − k + 1, ψχr)<br />

<br />

∞<br />

χd(−1) χr(n)cd(n)n −s .<br />

n=1<br />

d|r<br />

gcd(d,r/d)=1<br />

Multiplying by L(s − k + 1, χrψχ2k 4 ) = L(s − k + 1, χrψχ2 4 ), as in the definition <strong>of</strong> the<br />

<strong>Shimura</strong> lift, we have<br />

(3.10)<br />

∞<br />

A(n)n −s = L(s − k + 1, χrψχ2 4 )<br />

L(s − k + 1, χrψ)<br />

<br />

∞<br />

χd(−1) χr(n)cd(n)n −s .<br />

d|r<br />

gcd(d,r/d)=1<br />

By an easy consideration <strong>of</strong> Euler products, the quotient <strong>of</strong> the L-functions simplifies to<br />

1 − 2 k−1−s χr(2)ψ(2), and rewriting into q-series completes the pro<strong>of</strong> <strong>of</strong> the identity for the<br />

lift.<br />

The pro<strong>of</strong> <strong>of</strong> the equation for the lift in Theorem 1.2 follows along the same lines, with<br />

appropriate changes due to the slightly different expression for the theta function. The<br />

congruence condition reasoning following (3.5) does not change, and (3.7) becomes<br />

(3.11) b(n 2 ) = χr(n) <br />

χr/d(−1) <br />

m(n − dm)<br />

<br />

a<br />

(n − 2dm).<br />

r/d<br />

d|r<br />

gcd(d,r/d)=1<br />

m∈Z<br />

n=1<br />

n=1


8 DAVID HANSEN AND YUSRA NAQVI<br />

Recall that χr is odd here, so χr(−1) = χ r/d(−1)χd(−1) = −1, and so we have that<br />

χd(−1) = −χ r/d(−1) = ±1. Selberg inversion applies again, and the derivatives <strong>of</strong> <strong>modular</strong><br />

<strong>forms</strong> appearing in the definition <strong>of</strong> g(z) arise naturally from the linear form in m and n<br />

appearing in (3.11).<br />

In the odd case, it is not immediately clear that g(z) is in fact a <strong>modular</strong> form, since it<br />

contains derivatives <strong>of</strong> <strong>modular</strong> <strong>forms</strong>. However, it is in fact easy to prove <strong>modular</strong>ity by<br />

employing the following useful fact.<br />

Proposition 3.1. Let f(z) be a <strong>modular</strong> form <strong>of</strong> <strong>weight</strong> k on some subgroup <strong>of</strong> SL2(Z).<br />

Then 1 d<br />

2πi dz f(z) = ( ˜ f(z) + kE2(z)f(z))/12, where E2(z) is the Eisenstein series defined in<br />

(1.1) and ˜ f(z) is a <strong>modular</strong> form <strong>of</strong> <strong>weight</strong> k + 2.<br />

Note that E2 is not a <strong>modular</strong> form (see [6]). Using this proposition, we easily obtain<br />

g(z) = 1<br />

πi<br />

<br />

χd(−1)df ′ (dz)f(rz/d)<br />

= 1<br />

2π 2 i 2<br />

= 1<br />

12πi<br />

= 1<br />

12πi<br />

d|r<br />

gcd(d,r/d)=1<br />

<br />

d|r<br />

gcd(d,r/d)=1<br />

<br />

d|r<br />

gcd(d,r/d)=1<br />

<br />

d|r<br />

gcd(d,r/d)=1<br />

χd(−1)f(rz/d) ∂<br />

∂z f(dz)<br />

χd(−1)f(rz/d)( ˜ fd(z) + kE2(z)f(dz))<br />

χd(−1)f(rz/d) ˜ fd(z),<br />

where ˜ fd(z) is a <strong>modular</strong> form <strong>of</strong> <strong>weight</strong> k+2 and level dN. The sum involving E2’s vanishes<br />

due to cancellation in characters, namely χd(−1) = −χ r/d(−1).<br />

To complete the pro<strong>of</strong>s <strong>of</strong> Theorems 1.1 and 1.2, it suffies to compute the levels <strong>of</strong> the<br />

relevant <strong>Shimura</strong> <strong>lifts</strong>. Because g(z) lies in the space M2k(N ′ r, ψ 2 ), it is easy to see by the<br />

general theory <strong>of</strong> twists (see [6], Sec. 2.2) that gχr(z) ∈ M2k(N ′ r 3 , χ 2 rψ 2 ). However, we can<br />

in fact show that gχr(z) lies in the space M2k(N ′ r 2 , χ 2 rψ 2 ). To do this, we demonstrate the<br />

invariance <strong>of</strong> gχr(z) under a complete set <strong>of</strong> representatives <strong>of</strong> right cosets <strong>of</strong> Γ0(N ′ r 3 ) in<br />

Γ0(N ′ r 2 ). By Proposition 2.5 <strong>of</strong> [2], we have that [Γ0(N ′ r 2 ) : Γ0(N ′ r 3 )] = r, so such a set<br />

<strong>of</strong> representatives is given by<br />

(3.12) αj :=<br />

<br />

1 0<br />

jN ′ r2 <br />

1<br />

for j = 0, 1, 2, ..., r − 1. For convenience, we define the slash operator for γ ∈ GL + 2 (Q) by<br />

(3.13) f(z) |k γ := f(γz)(cz + d) −k (det γ) k/2 .<br />

With this notation, we need to show gχr(z) |k αj = gχr(z) for j = 0, 1, 2, ..., r − 1. Using<br />

Proposition 17 in Sec. 3.3 <strong>of</strong> [3] and defining τ(χr) := r−1<br />

m=0 χr(m)e 2πim/r , we first write


SHIMURA LIFTS OF MODULAR FORMS WITH THETA FUNCTIONS 9<br />

gχr(z) as a sum <strong>of</strong> linear trans<strong>forms</strong>,<br />

where we have set<br />

gχr(z) = r −1 r−1<br />

τ(χr)<br />

v=0<br />

= r −1 r−1<br />

τ(χr)<br />

(3.14) γv :=<br />

It then follows that<br />

gχr(z) | αj = r −1 r−1<br />

τ(χr)<br />

= r −1 r−1<br />

τ(χr)<br />

v=0<br />

= r −1 r−1<br />

τ(χr)<br />

v=0<br />

= r −1 r−1<br />

τ(χr)<br />

= gχr(z).<br />

v=0<br />

v=0<br />

<br />

<br />

¯χr(v)g(z)<br />

v=0<br />

<br />

1 −v/r<br />

.<br />

0 1<br />

<br />

<br />

¯χr(v)g(z)<br />

k<br />

<br />

<br />

¯χr(v)g(z)<br />

k<br />

¯χr(v)g(z) k γv<br />

1 −v/r<br />

0 1<br />

¯χr(v)g(z − v/r)<br />

¯χr(v)g(z) | γv,<br />

k γvαj<br />

1 0<br />

jN ′ r 2 1<br />

1 − jvN ′ r −jN ′ v 2<br />

<br />

jN ′ r 2 jvN ′ r + 1<br />

<br />

1 −v/r<br />

0 1<br />

Note that the first matrix in the fourth line is in Γ0(N ′ r) with d ≡ 1 (mod N ′ ), and so it<br />

has an invariant action on g(z).<br />

Having an explicit form for the lift allows us to check its cuspidality directly, without<br />

using the analytic machinery <strong>of</strong> <strong>Shimura</strong>’s theorem. If f(z) is a cusp form, then it is easy to<br />

see that S1(F )(z) must also be a cusp form, since a sum <strong>of</strong> cusp <strong>forms</strong> is itself cuspidal. We<br />

now consider, as a simple example, the case in which f(z) ∈ Mk(1, 1) is a Hecke eigenform<br />

that is not a cusp form. Let F (z) = θ(χr; z)f(4rz), and recall that<br />

(3.15) g(z) = <br />

χδ(−1)f(δz)f(rz/δ).<br />

δ|r<br />

gcd(δ,r/δ)=1<br />

Also recall that if 2|r, then we have that S1(F )(z) = gχ(z). If r is odd, then we define<br />

h(z) := g(z) − 2k−1g(2z), noting that in this case, the <strong>Shimura</strong> lift is hχ(z). We shall<br />

proceed by computing the Fourier expansions <strong>of</strong> g(z) and h(z) around a complete set <strong>of</strong><br />

cusps.<br />

Let gγ(z) denote g(z) |2k γ. For any γ = <br />

a b<br />

c d ∈ SL2(Z), we have<br />

<br />

a b<br />

(3.16) f(δz) = δ<br />

c d<br />

−k/2 <br />

δ 0 a b<br />

f(z)<br />

.<br />

0 1 c d<br />

k<br />

k


10 DAVID HANSEN AND YUSRA NAQVI<br />

Let δ ′ = gcd(c, δ). We have that there exists an integer y such that (δ/δ ′ )|(cy + d), and so<br />

we get<br />

<br />

<br />

f(δz)<br />

k<br />

<br />

a b<br />

c d<br />

= δ −k/2 <br />

<br />

f(z)<br />

Inserting this into the definition <strong>of</strong> g(z) gives<br />

<br />

<br />

g(z)<br />

2k<br />

<br />

a b<br />

=<br />

c d<br />

=<br />

<br />

δ|r<br />

gcd(δ,r/δ)=1<br />

<br />

δ|r<br />

gcd(δ,r/δ)=1<br />

<br />

aδ/δ ′ δ<br />

<br />

k<br />

′ (ay + b)<br />

c/δ ′ δ ′ (cy + d)/δ<br />

= (δ/δ ′ ) −k <br />

δ ′2z − δ ′ y<br />

f<br />

<br />

<br />

χδ(−1)f(δz)f(rz/δ)<br />

δ<br />

2k<br />

<br />

.<br />

<br />

a b<br />

c d<br />

<br />

δ<br />

−k r/δ<br />

−k χδ(−1)<br />

f<br />

(δ, c) (r/δ, c)<br />

<br />

δ ′ −y<br />

0 δ/δ ′<br />

<br />

δ ′2 z − y ′<br />

δ<br />

<br />

f<br />

δ ′′2 z − y ′′<br />

where δ ′ is as before, δ ′′ = (r/δ, c) and y ′ and y ′′ are integers that depend on δ. This<br />

trans<strong>forms</strong> into<br />

(3.17) gγ(z) = <br />

δ|r<br />

gcd(δ,r/δ)=1<br />

<br />

r<br />

−k χδ(−1) f<br />

(r, c)<br />

δ ′2 z − y ′<br />

δ<br />

<br />

f<br />

δ ′′2 z − y ′′<br />

We now consider<br />

(3.18)<br />

<br />

a<br />

g(2z) <br />

2k c<br />

<br />

b<br />

= 2<br />

d<br />

−k <br />

2<br />

g(z) <br />

2k 0<br />

<br />

0 a<br />

1 c<br />

<br />

b<br />

,<br />

d<br />

<br />

which gives us that g(2z) |2k<br />

a b<br />

c d = gγ(2z) if c is even or gγ((z − x)/2) if c is odd, where<br />

x is some integer that depends on d. This yields<br />

<br />

<br />

h(z)<br />

2k<br />

<br />

a b<br />

= gγ(z) − 2<br />

c d<br />

k−1 gγ(2z) or<br />

= gγ(z) − 2 −k−1 gγ((z − x)/2).<br />

Thus, in all cases, the constant term <strong>of</strong> the Fourier expansion is a constant multiple <strong>of</strong><br />

( r<br />

(r,c) )−k a(0) 2 χδ(−1), and hence this term vanishes if and only if f is a cusp form or<br />

(3.19)<br />

<br />

δ|r<br />

gcd(δ,r/δ)=1<br />

χδ(−1) = 0.<br />

In particular, this sum vanishes if and only if χr decomposes into a product <strong>of</strong> Dirichlet<br />

characters to prime power moduli which are not all even. Note that by [7], this is equivalent<br />

to θ(χr; z) being a cusp form. This same method can be applied to <strong>modular</strong> <strong>forms</strong> <strong>of</strong> higher<br />

level; however, the computations are more complicated.<br />

δ<br />

<br />

.<br />

δ<br />

<br />

,


SHIMURA LIFTS OF MODULAR FORMS WITH THETA FUNCTIONS 11<br />

4. Examples and Applications<br />

In this section, we present some examples illustrating Theorems 1.1 and 1.2 in which we<br />

calculate the explicit form <strong>of</strong> the <strong>Shimura</strong> lift for certain <strong>half</strong>-<strong>integral</strong> <strong>weight</strong> <strong>forms</strong>.<br />

Example 1. We begin with Theorem 1.1, by defining f(z) := η(z) 5 /η(5z), where<br />

η(z) := q<br />

1/24 <br />

n>0<br />

(1 − q n )<br />

is the Dedekind eta function. This function is in the space M2(5, χ5), and in fact we have<br />

f(z) = −5E2(1, χ5; z). We compute the <strong>Shimura</strong> lift <strong>of</strong> f(48z)η(24z) = 1<br />

2f(48z)θ(χ12; z).<br />

To utilize Theorem 1.1, we factor χ12 = χ−4χ−3 and 12 = 22 · 3 to obtain<br />

(4.1) g(z) = f(z)f(12z) − f(3z)f(4z).<br />

Because χ12(2) = 0, the second term in (1.7) vanishes and we have<br />

(4.2)<br />

<br />

η(48z) 5η(24z) <br />

η(z) 5η(12z) 5<br />

S1<br />

=<br />

η(240z) η(5z)η(60z) − η(3z)5η(4z) 5 <br />

= 25q<br />

η(15z)η(20z) χ12<br />

7 +50q 11 +100q 13 +150q 17 +...<br />

Example 2. We now illustrate Theorem 1.2 by computing the lift <strong>of</strong> ∆(60z) 2 θ(χ−15; z),<br />

where ∆(z) = η(z) 24 is the standard discriminant function. To apply our theorem, we must<br />

write ∆(z) 2 as a linear combination <strong>of</strong> Hecke eigen<strong>forms</strong>. The two Hecke eigen<strong>forms</strong>, say<br />

f1(z) and f2(z), <strong>of</strong> <strong>weight</strong> 24 and level 1 have Fourier expansions<br />

(4.3) fi(z) = q + aiq 2 + (195660 − 48ai)q 3 + ...<br />

with a1 = 540 + 12 √ 144169 and a2 = 540 − 12 √ 144169. Hence,<br />

(4.4) ∆(z) 2 = f1(z) − f2(z)<br />

24 √ 144169 .<br />

To apply Theorem 1.2, we factor 15 = 3 · 5 and χ−15 = χ−3χ5 to obtain<br />

(4.5) g(z) = 1<br />

πi<br />

<br />

f ′ 1(z)f1(15z) − 3f ′ 1(3z)f1(5z) + 5f ′ 1(5z)f1(3z) − 15f ′ 1(15z)f1(z)<br />

and hence S1(θ(χ−15; z)f1(60z))(z) = gχ−15 (z) + 224gχ−15 (2z). A similar formula holds for<br />

the lift <strong>of</strong> f2(z).<br />

References<br />

[1] B. A. Cipra. On the <strong>Shimura</strong> lift, après Selberg. J. Number Theory, 32(1):58–64, 1989.<br />

[2] H. Iwaniec. Topics in classical automorphic <strong>forms</strong>, volume 17 <strong>of</strong> Graduate Studies in Mathematics. American<br />

Mathematical Society, Providence, RI, 1997.<br />

[3] N. Koblitz. Introduction to elliptic curves and <strong>modular</strong> <strong>forms</strong>, volume 97 <strong>of</strong> Graduate Texts in Mathematics.<br />

Springer-Verlag, New York, second edition, 1993.<br />

[4] T. Miyake. Modular <strong>forms</strong>. Springer Monographs in Mathematics. Springer-Verlag, Berlin, english edition,<br />

2006. Translated from the 1976 Japanese original by Yoshitaka Maeda.<br />

[5] K. Ono. Distribution <strong>of</strong> the partition function modulo m. Ann. <strong>of</strong> Math. (2), 151(1):293–307, 2000.<br />

[6] K. Ono. The web <strong>of</strong> <strong>modular</strong>ity: arithmetic <strong>of</strong> the coefficients <strong>of</strong> <strong>modular</strong> <strong>forms</strong> and q-series, volume<br />

102 <strong>of</strong> CBMS Regional Conference Series in Mathematics. Published for the Conference Board <strong>of</strong> the<br />

Mathematical Sciences, Washington, DC, 2004.


12 DAVID HANSEN AND YUSRA NAQVI<br />

[7] J.-P. Serre and H. M. Stark. Modular <strong>forms</strong> <strong>of</strong> <strong>weight</strong> 1/2. In Modular functions <strong>of</strong> one variable, VI<br />

(Proc. Second Internat. Conf., Univ. Bonn, Bonn, 1976), pages 27–67. Lecture Notes in Math., Vol. 627.<br />

Springer, Berlin, 1977.<br />

[8] G. <strong>Shimura</strong>. On <strong>modular</strong> <strong>forms</strong> <strong>of</strong> <strong>half</strong> <strong>integral</strong> <strong>weight</strong>. Ann. <strong>of</strong> Math. (2), 97:440–481, 1973.<br />

[9] J. B. Tunnell. A classical Diophantine problem and <strong>modular</strong> <strong>forms</strong> <strong>of</strong> <strong>weight</strong> 3/2. Invent. Math.,<br />

72(2):323–334, 1983.<br />

<strong>Department</strong> <strong>of</strong> Mathematics, Brown University, Providence, RI 02912<br />

E-mail address: david hansen@brown.edu<br />

<strong>Department</strong> <strong>of</strong> Mathematics and Statistics, Swarthmore College, Swarthmore, PA 19081<br />

E-mail address: yusra.naqvi@gmail.com

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!