Past exam of the mathematics course of the University of Cambridge / 2003 / iii / Paper 24 (Modular Forms)

Let $\mathbb{H}$ be the upper half-plane and put $q=e^{2\pi i\tau }$. A weight-$k$ modular form for the modular group is a holomorphic function on $\mathbb{H}$ satisfyingwhose Fourier expansion at the modular cusp at infinity has only nonnegative powers of $q$. These functions form the vector space $M_k$. Its subspace $S_k$ consists of the cusp forms, whose constant Fourier coefficient is zero. For the full modular group every cusp is equivalent to infinity, so this is the complete cusp condition. The matrix $-I$ gives $M_k=0$ for odd $k$, and the valence formula for the modular group excludes nonzero forms of negative weight.

The normalized Eisenstein series start as $E_4=1+240q+O(q^2)$ and $E_6=1-504q+O(q^2)$. Consequently the modular discriminantis a nonzero cusp form of weight twelve, with order one at infinity. Its contribution there already exhausts the valence formula, so it has no zero in $\mathbb{H}$. Multiplication by $\Delta$ takes $M_k$ injectively into $S_{k+12}$. Conversely, dividing any member of $S_{k+12}$ by $\Delta$ is holomorphic in $\mathbb{H}$, has weight $k$, and has no negative power at infinity because its numerator vanishes there. Thus multiplication by the modular discriminant gives

The valence formula for the modular group shows that a weight-zero form minus its constant term must vanish: otherwise it would be a cusp form with positive order at infinity and total zero order zero. Hence $\dim M_0=1$. A weight-two form satisfies $f(i)=i^{2}f(i)=-f(i)$, so any nonzero one would contribute at least $1/2$ to the valence formula, exceeding $2/12$. Therefore $M_2=0$.

Every even integer $k\ge 4$ has a representation $k=4a+6b$ with $a,b\ge 0$. Indeed, if $k/2$ is even, take $b=0$; if it is odd, take $b=1$ and $a=(k-6)/4$. The form $R_k=E_4^a{E}_6^b$ has constant term one. Taking the constant Fourier coefficient therefore maps $M_k$ onto $\mathbb{C}$, with kernel $S_k=\Delta M_{k-12}$. Starting with the negative-weight vanishing and the two initial dimensions givesand hence the dimension of level-one modular forms isThis also proves finite dimensionality, rather than presupposing it in the recurrence.

For the generation assertion, if $f\in M_k$ has constant coefficient $a_0$, subtract $a_0{R}_k$. The difference is a cusp form, and so equals $\Delta h$ with $h\in M_{k-12}$. Induction on weight expresses $h$ as a polynomial in $E_4,E_6$; substituting the polynomial expression for $\Delta$ does the same for $f$. Constants start the induction, while weight two contributes nothing. The result is a weighted homogeneous polynomial, assigning weights four and six to its variables. In fact the monomials $E_4^a{E}_6^b$ with $4a+6b=k$ number exactly the displayed dimension: $b$ runs from zero to $\lfloor k/6\rfloor$ with parity $k/2$. Their spanning property therefore makes them a basis in each weight. This identifies the polynomial ring of level-one modular forms:

The preceding dimension of level-one modular forms gives $\dim M_{16}=2$ and $\dim S_{16}=\dim M_4=1$. Thus $E_4\Delta$, which begins with $q$, spans $S_{16}$. A normalized weight-sixteen Eisenstein series and $E_4^4$ both have constant term one, so their difference is a multiple of $E_4\Delta$.

There is a normalization error in the printed auxiliary expansion: for the usual normalized $E_{16}$, the coefficient multiplier is $16320/3617$, rather than $255/(8\cdot 3617)$. Indeed, the Fourier expansion of a normalized Eisenstein series gives $-32/B_{16}$, and the Bernoulli number $B_{16}=-3617/510$ gives $16320/3617$. The printed multiplier is smaller by a factor of $512$. With the correct normalization, the first Fourier coefficients yieldComparing all positive-index coefficients therefore gives the exact identityThe Fourier coefficients of $E_4^4$ are integers, as are those of $E_4\Delta$: $E_4$ has its integral divisor-sum expansion and the modular discriminant has the product $q{\prod }_{r\ge 1}(1-q^r{)}^{24}$. Also $3456000+16320=960\cdot 3617$, and $\gcd (16320,3617)=1$. Reducing the identity modulo $3617$ and cancelling the invertible multiplier proves the weight-sixteen Eisenstein congruenceThe desired congruence is valid; it is the auxiliary normalization in the PDF that requires correction.

Fix an even weight $k$ and write $\Gamma =S{L}_2(\mathbb{Z})$. Extend the slash operator for modular forms to positive-determinant rational matrices by the determinant-normalized slash operatorThe automorphy factor identity makes this a right action. Let ${\mathcal{D}}_n$ be the integral matrices of determinant $n>0$. Their left $\Gamma$-cosets have unique determinant-n matrix representatives for Hecke operatorsAn integral row operation obtained from Bézout's identity makes the first column $(a,0{)}^T$, and a row shear reduces the other top entry modulo $d$, proving this description. The usual Hecke operator isThis normalization has $T_1=1$. Right multiplication by $\Gamma$ permutes the left cosets in ${\mathcal{D}}_n$, so $T_{n}f$ has the same modular transformation law as $f$. Each summand is holomorphic in $\mathbb{H}$. Holomorphy at the cusp and preservation of cusp forms follow directly from the coefficient formula below; the full modular group has only one cusp orbit. Thus the operators act on both $M_k$ and $S_k$.

If $f={\sum }_{r\ge 0}{a}_r{q}^r$, summing over $b$ kills terms unless $d\mid r$. Writing $r=d\ell$ gives output exponent $a\ell$, so a term of exponent $m$ requires $a\mid m$ and has original index $r=dm/a=nm/a^2$. Its multiplier is $n^{k-1}{d}^{1-k}=a^{k-1}$. The Fourier coefficients of a composite-index Hecke operator are consequentlyFor $m=0$, this means $a_0(T_{n}f)={\sigma }_{k-1}(n)a_0(f)$. In particular $a_1(T_{n}f)=a_n(f)$. For a prime number $p$,with a nonexistent fractional-index coefficient interpreted as zero.

The Hecke multiplication relations follow from these coefficients. For coprime indices, the two divisor choices combine uniquely and give $T_m{T}_n=T_{mn}$. For prime powers put $A=p^{k-1}$, and, on formal Fourier series, let $U_p(\sum a_m{q}^m)=\sum a_{pm}{q}^m$ and $V_p(\sum a_m{q}^m)=\sum a_m{q}^{pm}$. These individual maps need not preserve level one, but $U_p{V}_p=1$, and the coefficient formula givesMultiplying on the left by $T_p=U_p+A{V}_p$ and cancelling $U_p{V}_p$ yieldsInduction gives $T_{p^u}{T}_{p^v}={\sum }_{j=0}^{\min (u,v)}{p}^{j(k-1)}{T}_{p^{u+v-2j}}$; combining distinct prime factors provesThe modular Hecke algebra is therefore commutative and generated by its prime-index operators. The identity $U_p{V}_p=1$, rather than the generally false reversed identity, is what proves the prime-power recurrence.

On $S_k$ use the positive definite Petersson inner productExponential cusp decay makes this integral finite. The Hecke operators are self-adjoint for the Petersson inner product. Here is the correspondence argument. For a positive-determinant matrix $\alpha$, set $H=\Gamma \cap {\alpha }^{-1}\Gamma \alpha$. The finite sum over the double coset $\Gamma \alpha \Gamma$ unfolds its pairing intoChanging variables $z=\alpha \tau$ preserves hyperbolic area, and $\mathrm{Im}(\alpha \tau )=\det (\alpha )y/|c\tau +d{|}^2$ shows that the determinant-normalized slash factors transfer $\alpha$ from $f$ to ${\alpha }^{-1}$ on $g$. The new quotient is by $\alpha H{\alpha }^{-1}=\Gamma \cap \alpha \Gamma {\alpha }^{-1}$; folding it back gives the operator for the reversed double coset. For $\alpha \in {\mathcal{D}}_n$, its adjugate $n{\alpha }^{-1}$ again lies in ${\mathcal{D}}_n$. Positive scalar matrices act trivially in the determinant normalization, so ${\alpha }^{-1}$ and its adjugate give the same action. The adjugate involution preserves the entire determinant-$n$ correspondence, and the external normalization $n^{k/2-1}$ is real. Thus

Finite-dimensional spectral theory now gives a simultaneous orthonormal eigenbasis on $S_k$, since the operators commute and are self-adjoint. For a common Hecke eigenform $f$ with eigenvalues ${\lambda }_n$, the first-coefficient formula gives $a_n(f)={\lambda }_n{a}_1(f)$. If $a_1(f)=0$, every coefficient vanishes and $f=0$, a contradiction. Rescale each eigenform to have $a_1=1$. The resulting basis is orthogonal, andAll these eigenvalues and coefficients are real. A fixed simultaneous eigenspace is one-dimensional, because its normalized Fourier coefficients are uniquely determined by the eigenvalues.

Applying the multiplication relations to an eigenform givesThe recurrence yields ${\sum }_{r\ge 0}{a}_{p^r}{X}^r=(1-a_{p}X+p^{k-1}{X}^2{)}^{-1}$. Multiplicativity then gives the Euler product of a Hecke eigenforminitially in a right half-plane, for example $\mathrm{Re}s>k/2+1$ using the Fourier coefficient bound for a cusp form. Modular inversion and the Mellin transform of a cusp-form L-function continue $\Lambda (f,s)=(2\pi {)}^{-s}\Gamma (s)L(f,s)$ to an entire function and give $\Lambda (f,s)=i^k\Lambda (f,k-s)$.

There is also an integral structure. If $d=\dim S_k$, choose $F_j={\Delta }^j{R}_{k-12j}$, $1\le j\le d$, with $R_w$ a monomial in $E_4,E_6$ of weight $w$ and constant term one. These forms have integral Fourier coefficients and start with $q^j$; their first $d$ coefficients form a unitriangular matrix. They are a basis, and every cusp form with integral coefficients is an integral combination of them, by successive elimination of its first $d$ coefficients. The coefficient formula shows that $T_n$ preserves this lattice. Its characteristic polynomial is consequently monic with integer coefficients, proving that the Hecke eigenvalues are algebraic integers. Together with self-adjointness, this gives real algebraic integers without requiring any unproved bound for their sizes.

For even $k\ge 4$, the full space decomposes as $M_k=\mathbb{C}{E}_k\oplus S_k$. The normalized Eisenstein series is a simultaneous eigenform withIts coefficients are $-2k{\sigma }_{k-1}(m)/B_k$, and the elementary divisor identity ${\sigma }_{k-1}(m){\sigma }_{k-1}(n)={\sum }_{d\mid \gcd (m,n)}{d}^{k-1}{\sigma }_{k-1}(mn/d^2)$ verifies the eigenvalue formula, including the constant coefficient. Thus the cusp eigenbasis plus $E_k$ diagonalizes the operators on $M_k$. Weight two and odd weights have no forms; in weight zero the constant function has eigenvalue ${\sigma }_{-1}(n)$. As a concrete cuspidal example, $S_{12}=\mathbb{C}\Delta$, so $T_n\Delta =\tau (n)\Delta$ for its Fourier coefficients, and the recurrence gives $\tau (p^2)=\tau (p{)}^2-p^{11}$.

The analytic input for these Mellin transforms is elementary cusp decay. For a level-one weight-$k$ cusp form, the invariant norm of a modular form $y^{k/2}|f(x+iy)|$ is invariant under the modular group. It is bounded on the standard fundamental region, because the cusp expansion decays exponentially, and hence bounded on all of $\mathbb{H}$. Fourier inversion givesTaking $y=1/n$ proves the Fourier coefficient bound for a cusp form $c_n=O(n^{k/2})$. In particular rational translates have polynomially bounded coefficients and exponential decay as $y\to \infty$, uniformly in their real parts. These facts justify the initial Mellin integrations and the later entire continuation.

Use $q=e^{2\pi i\tau }$, so $g(iy)={\sum }_{n\ge 1}{b}_n{e}^{-2\pi ny}$. Put $u=\mathrm{Re}s$. For $u>\max \{0,\sigma +1\}$,The Gamma integral and absolute convergence therefore justify exchanging summation and integration. For every positive integer $n$, substitution $t=2\pi ny$ givesSumming these identities proves the Mellin transform formula for the Dirichlet series:The power of the positive real variable $y$ uses the real logarithm; the initial half-plane stated above is a sufficient one even when $\sigma$ is negative.

Choose $b=(ad-1)/N\in \mathbb{Z}$, and take the matrixAt $w=\tau -d/N$ its denominator is $Nw+d=N\tau$, while its numerator is $aw+b=a\tau -1/N$. ConsequentlyApplying the weight-$k$ transformation law of the cusp form at $w\in \mathbb{H}$ proves the rational-cusp inversion of a level-one cusp form:

The rational translate $g(\tau )=f(a/N+\tau )$ has Fourier coefficients $e^{2\pi ian/N}{c}_n$, with the same polynomial bound as $c_n$. Thus the preceding Mellin transform gives, initially for $\mathrm{Re}s$ large, the completion of the additive twist of a cusp-form L-functionWrite $\epsilon =i^k=(-1{)}^{k/2}$. Nonzero level-one forms have even weight, so ${\epsilon }^2=1$. Split the integral at $1/N$. In its lower part substitute $y=1/(N^{2}v)$, and use rational-cusp inversion of a level-one cusp form with $\tau =iv$:The bounds reverse from $(0,1/N)$ to $(\infty ,1/N)$, so

Both cusp values in this integral are $O(e^{-2\pi y})$ as $y\to \infty$. On any compact set of $s$-values, the remaining powers of $Ny$ are bounded by fixed positive powers of $y$; differentiation in $s$ merely introduces powers of $\log (Ny)$. The integral and its differentiated integrals therefore converge uniformly on compact sets. It defines an entire function of $s$, agreeing with the original completion in its half-plane of convergence, and hence supplies the required analytic continuation to $\mathbb{C}$.

For the opposite rational cusp, $a^{\prime }=-d$ has inverse $d^{\prime }=-a$ modulo $N$. Denote the two individual upper integrals by $I_a(s)$ and $I_{-d}(s)$. The representation givesReplacing $s$ by $k-s$ in the first equality and using ${\epsilon }^2=1$ proves the functional equation of an additive cusp-form twist:For $N=1$ this specializes to the entire completion and functional equation of the ordinary L-function of a cusp form.

Use the Fourier transform convention $\hat{h}(\xi )={\int }_{\mathbb{R}}h(x)e^{-2\pi ix\xi }dx$. For a Schwartz function $h$, the Poisson summation formula isTo prove it, periodize $h$ as $P(x)={\sum }_{n\in \mathbb{Z}}h(x+n)$. Rapid decay of all derivatives makes this a smooth periodic function, with uniform termwise differentiability. Its $m$th Fourier coefficient iswhere absolute convergence justifies splitting the real-line integral into unit intervals. Integration by parts gives rapid decay of the coefficients of $P$, so its Fourier series converges absolutely and uniformly to $P$. Evaluate it at zero to obtain the formula.

For $\mathrm{Im}\tau >0$ and $z\in \mathbb{C}$, take the complex Gaussian function $h(x)=e^{\pi i\tau x^2+2\pi izx}$. It is a Schwartz function of the real variable $x$. Its Fourier transform isThe real Gaussian integral proves this first for $\tau =iy$, $y>0$, and real $z$. Normal convergence of the defining integrals makes both sides entire in $z$ and holomorphic in $\tau \in \mathbb{H}$, so the identity theorem extends it to the stated domain. Choose the square-root branch holomorphic for $-i\tau$ in the right half-plane and positive when $\tau =iy$.

Applying Poisson summation gives the transformation of the Jacobi theta function:The series on the right is ${\vartheta }_{00}(z/\tau ,-1/\tau )$. Rearranging therefore givesThe branch choice is part of the formula; at $\tau =i$ its square-root factor equals one.

Fix $\tau \in \mathbb{H}$ and let $\Lambda =\mathbb{Z}+\mathbb{Z}\tau$. We use the binary convention for the theta function with characteristics in the PDF. Its nonzero exponential prefactor does not change the zeros of the translated ${\vartheta }_{00}$. The four characteristic functions are entire in $z$, by normal convergence of their Gaussian series.

We first prove their zero information, rather than assume that the quotient has no poles. Pair the terms with indices $n$ and $-n-1$ to getThe logarithmic derivative $D(z)={\vartheta }_{00}^{\prime }(z)/{\vartheta }_{00}(z)$ is periodic under $z\mapsto z+1$ and satisfies $D(z+\tau )=D(z)-2\pi i$, by the spatial transformation formulas. Choose a fundamental parallelogram with zero-free boundary. Its vertical-side integrals cancel, and subtracting the top horizontal integral from the bottom one gives $2\pi i$. The argument principle therefore counts exactly one zero inside, with multiplicity. Thus the displayed zero is simple and is the unique one modulo $\Lambda$. The zeros of theta functions with characteristics are consequentlyAll are simple. In particular the three even theta constants are nonzero and ${\vartheta }_{11}^{\prime }(0)\ne 0$.

Let $R(z)$ be the product of these four functions divided by ${\vartheta }_{11}(2z)$. The denominator's simple zeros occur exactly at the four half-periods modulo $\Lambda$, and each is cancelled by the corresponding simple numerator zero. Hence $R$ extends to an entire function. Under $z\mapsto z+1$, the four numerator signs multiply to one, as do the two successive shifts of the denominator. Under $z\mapsto z+\tau$, the numerator has multiplier $q^{-2}{t}^{-4}$, where $t=e^{2\pi iz}$; the denominator, whose argument shifts by $2\tau$, has the same multiplier. Thus $R$ is an entire elliptic function with periods $1,\tau$. It is bounded on a fundamental parallelogram and hence on the plane, so Liouville's theorem makes it constant.

At $z=0$, the simple odd zero gives ${\vartheta }_{11}(z)/{\vartheta }_{11}(2z)\to 1/2$. Therefore the Jacobi theta duplication identity isAt a zero of the displayed denominator, this is interpreted by removable continuation. Equivalently, multiply through by the denominator to obtain an entire product identity valid for every $z$.

Here being modular of weight $k$ means obeying the weight transformation law, or equivalently being invariant under the slash operator for modular forms:This transformation property alone does not impose holomorphy; a holomorphic modular form additionally satisfies interior and cusp holomorphy. That distinction is needed for the nonholomorphic completion in the next part.

Put $F(\tau )=f(N\tau )$. For $\gamma \in {\Gamma }_0(N)$, conjugation by $\mathrm{diag}(N,1)$ gives the integral determinant-one matrixSince the denominator in the latter action is $(c/N)(N\tau )+d=c\tau +d$, the full-group modularity of $f$ givesThus argument dilation lowers the group to the Gamma 0 congruence subgroup while retaining the weight. If $f$ is a holomorphic modular form, cusp holomorphy is preserved too: an integral matrix $\mathrm{diag}(N,1)\gamma$ of determinant $N$ factors as an integral determinant-one matrix times an upper triangular positive-determinant matrix. The corresponding slash translate is a constant times $f((a\tau +b)/d)$ with positive $a/d$, whose expansion at infinity has no negative exponents. This checks every rational cusp; for a cusp form its constant term remains zero. It is the level-one instance of the degeneracy map for cusp forms.

Logarithmically differentiate the convergent product of the modular discriminant. On compact subsets of $\mathbb{H}$, absolute convergence permits termwise differentiation and rearrangement of the resulting divisor sums:The division is valid because $\Delta$ has no interior zeros. Its weight-twelve inversion law is $\Delta (-1/\tau )={\tau }^{12}\Delta (\tau )$. Taking a logarithmic derivative, including the derivative $1/{\tau }^2$ of $-1/\tau$, yieldsThus the Eisenstein series of weight two has the anomalous transformation

Let $y=\mathrm{Im}\tau$. Under inversion, $y$ becomes $y/|\tau {|}^2$. Using ${\tau }^2-|\tau {|}^2=2iy\tau$, the anomalous term cancels:The Fourier expansion gives $E_2(\tau +1)=E_2(\tau )$, and translation leaves $y$ unchanged. Since translation and inversion generate the projective modular group and weight two is unchanged by $-I$, the almost holomorphic weight-two Eisenstein series satisfiesIt is modular in the transformation sense, but is not a holomorphic modular form. Substituting $\mathrm{Im}(\gamma \tau )=y/|c\tau +d{|}^2$ into this law also gives the useful full-group anomaly

Write $F(\tau )={\sum }_{M\mid N}c(M)E_2(M\tau )$ and $C={\sum }_{M\mid N}c(M)/M$. For $\gamma =\left(\begin{array}{cc}a& b\\ c& d\end{array}\right)\in {\Gamma }_0(N)$, each divisor $M$ of $N$ also divides the lower-left entry $c$. ThusThe full transformation of the Eisenstein series of weight two consequently givesSumming with the prescribed coefficients provesIf $C=0$, this is exactly the desired weight-two law. Conversely, test it at $\gamma =\left(\begin{array}{cc}1& 0\\ N& 1\end{array}\right)\in {\Gamma }_0(N)$. Its anomaly factor $6N(N\tau +1)/(\pi i)$ is nonzero throughout the upper half-plane; if $F$ is modular it forces $C=0$. This proves cancellation of weight-two Eisenstein anomalies:Equivalently the inverse-height terms cancel in $\sum c(M)E_2^{\ast }(M\tau )$, leaving precisely the holomorphic function $F$. Eliminating $c(1)$ shows that all such combinations are spanned by $E_2(\tau )-M{E}_2(M\tau )$ for $M>1$ dividing $N$.