Fourier expansion of a modular form ($f(\tau )={\sum }_{n\ge 0}{a}_n{q}^n$)

Translation invariance at a cusp gives a Fourier expansion in its local parameter $q$. Holomorphy at the cusp excludes negative powers; vanishing of the constant term defines a cusp form.

A cusp form is a modular form that vanishes at every cusp. At the cusp at infinity for the full modular group it has an expansion ${\sum }_{n\ge 1}{a}_n{q}^n$.

The old subspace is spanned by $h(tz)$ for cusp forms $h$ of level $M<N$, where $M\mid N$ and $t\mid N/M$. It is stable under the Hecke operators, but need not have a basis of eigenvectors for every bad-prime operator. Its Petersson inner product orthogonal complement is the new subspace of cusp forms.

If $M\mid N$ and $D\mid N/M$, conjugating ${\Gamma }_0(N)$ by $\mathrm{diag}(D,1)$ gives matrices $\left(\begin{array}{cc}a& Db\\ c/D& d\end{array}\right)$ in ${\Gamma }_0(M)$. This proves the weight transformation of $f(D\tau )$. At any rational modular cusp, factor the rational change of variable into an integral cusp representative followed by an upper-triangular rational matrix of positive slope. The original cusp expansion then stays holomorphic and has zero constant term. Thus this map takes cusp forms of level $M$ to those of level $N$.

For prime $p$, the two cusps of ${\Gamma }_0(p)$ have widths one and $p$. At infinity the modular discriminant product gives the first pair of orders. At zero use $S\tau =-1/\tau$ and the parameter $q_0=e^{2\pi i\tau /p}$. The weight-twelve transforms are $\Delta {|}_{12}S=\Delta (\tau )$ and $\Delta (p\tau ){|}_{12}S=p^{-12}\Delta (\tau /p)$, giving orders $p$ and one in $q_0$.

If $p$ divides the primitive level $M$ of a trivial-character newform and $N=M{p}^r$, put $e_j=p^{jk/2}f(p^{j}z)$. Then $U_p{e}_0=a_p{e}_0$ and $U_p{e}_j=p^{k/2}{e}_{j-1}$ for $j\ge 1$. The normalized Fricke involution acts by $e_j\mapsto \epsilon e_{r-j}$, where $W_{M}f=\epsilon f$. These two operators act irreducibly: a nonzero invariant subspace contains an eigenvector of $U_p$; its eigenvalue is $a_p$ or zero. Either it contains $e_0$ immediately or applying the reversal and a power of $U_p$ produces $e_0$, using $a_p^2\ne p^k$. Reversal then produces $e_r$, and its successive images produce every basis vector.

The new subspace is the orthogonal complement of the old subspace of cusp forms for the Petersson inner product. A form belongs to it precisely when the adjoints of all lower-level oldform by argument dilation maps annihilate it.

A newform is a normalized Hecke eigenform in the new subspace of cusp forms. Its weight, level and Dirichlet character are part of its specification. For a newform, its Fourier coefficients equal its eigenvalues of Hecke operators.

For a newform of weight $k\ge 2$, level $M$ and trivial Dirichlet character, the bad-prime Hecke operator $U_p$ has the displayed possibilities. The local trace relation on the new subspace of cusp forms gives $U_p=-p^{k/2-1}{W}_p$ when $p\parallel M$, where the normalized local Atkin-Lehner operator squares to one; if $p^2\mid M$, the new-vector relation gives $U_p=0$. In particular $|a_p|<p^{k/2}$, a strict inequality useful for proving irreducibility of old blocks.

On the new subspace of cusp forms, a common eigenspace for the Hecke operators away from the level has dimension one. Distinct newforms, even at different primitive levels, have different systems of these eigenvalues. Consequently the commuting bad-prime operators also act by scalars on these lines. This is the multiplicity-one part of Atkin-Lehner-Li theory, not a consequence of normal operator diagonalization alone. See Stein's statement of the theorem.

For $f={\sum }_{n\ge 1}{a}_n{q}^n$, define $L(f,s)={\sum }_{n\ge 1}{a}_n{n}^{-s}$ in a right half-plane. Its completion $(2\pi {)}^{-s}\Gamma (s)L(f,s)$ is the Mellin transform of $f(iy)$. The modular inversion and exponential cusp decay continue this completion to an entire function and give its reflection $s\mapsto k-s$. A normalized Hecke eigenform also has an Euler product of a Hecke eigenform.

For a cusp form $f=\sum c_n{q}^n$ and coprime integers $a,N$, the additive twist inserts the additive character $e^{2\pi ian/N}$. It is the Dirichlet series associated with the rational translate $f(a/N+\tau )$. Its completed Mellin transform uses the scale $N^s$ and the Gamma function.

For even weight $k$, complete the additive twist of a cusp-form L-function as $M(f,a/N,s)=(N/(2\pi ){)}^s\Gamma (s)L(f,a/N,s)$. Split its Mellin transform at $y=1/N$ and use rational-cusp inversion of a level-one cusp form in the lower integral. The resulting two integrals over $[1/N,\infty )$ converge locally uniformly for every complex $s$, giving an entire continuation and the displayed functional equation. The relation $ad\equiv 1(\mathrm{mod}N)$ identifies the opposite cusp.

If $ad\equiv 1(\mathrm{mod}N)$, the determinant-one matrix with rows $(a,(ad-1)/N)$ and $(N,d)$ sends $\tau -d/N$ to $a/N-1/(N^2\tau )$. The transformation of the cusp form gives the displayed identity. It turns the small-height portion of an additive-twist Mellin transform into exponentially decaying values at the inverse rational cusp.

For a level-one cusp form and any Dirichlet character modulo $N>1$, this translation sum is a cusp form on ${\Gamma }_1(N)\cap {\Gamma }_0(N^2)$. Conjugating that subgroup by $\left(\begin{array}{cc}1& j/N\\ 0& 1\end{array}\right)$ yields integral determinant-one matrices; cusp holomorphy under rational slash operators supplies all cusp conditions. Its exact Fourier coefficients at positive indices are $a_n(f){\sum }_{j\in (\mathbb{Z}/N\mathbb{Z}{)}^{\times }}\chi (j{)}^{-1}{e}^{2\pi inj/N}$. For a primitive Dirichlet character they equal $g(\overline{\chi })\chi (n)a_n(f)$, by the finite Fourier transform of a primitive Dirichlet character. For imprimitive characters the sum can be nonzero at nonunits, so the simplified twist formula need not hold.

For a primitive Dirichlet character $\chi$ modulo $D$, its Gauss sum of a Dirichlet character gives $f_{\chi }=\tau (\overline{\chi }{)}^{-1}{\sum }_u\overline{\chi }(u)f|U_u$, with $U_u=\left(\begin{array}{cc}1& u/D\\ 0& 1\end{array}\right)$. If $\gamma =\left(\begin{array}{cc}a& b\\ c& d\end{array}\right)\in {\Gamma }_0(N{D}^2)$, take $v\equiv d^{2}u(\mathrm{mod}D)$. Then $U_u\gamma U_v^{-1}\in {\Gamma }_0(N)$ and its lower-right entry is congruent to $d$ modulo $N$. Reindexing the sum yields the character $\psi (d)\chi (d{)}^2$. Rational translations preserve vanishing at modular cusps, proving the stated cuspidality as well as the transformation law.

Use the determinant-normalized slash operator and $W_M=\left(\begin{array}{cc}0& -1\\ M& 0\end{array}\right)$. For units $u,v$ with $Nuv\equiv -1(\mathrm{mod}D)$, the matrix identity $U_u{W}_{N{D}^2}=D{\gamma }_u{W}_N{U}_v$ has ${\gamma }_u=\left(\begin{array}{cc}(1+Nuv)/D& u\\ Nv& D\end{array}\right)\in {\Gamma }_0(N)$. A positive scalar matrix acts trivially. The nebentypus supplies $\psi (D)$; reindexing gives $\overline{\chi }(u)=\chi (-N)\chi (v)$. Thus the factor is initially $\psi (D)\chi (-N)\tau (\chi )/\tau (\overline{\chi })$. The primitive Gauss sum of a Dirichlet character identity $\tau (\chi )\tau (\overline{\chi })=\chi (-1)D$ gives the displayed equivalent factor. No assumption that $f$ itself is a Fricke eigenform is needed.

For a weight-$k$ modular form on the full modular group, the quantityis modular invariant. For a cusp form it tends to zero at the cusp and is therefore bounded on a fundamental domain.

For $k>0$, a meromorphic function satisfying the modular transformation law is a cusp form exactly when its invariant norm is bounded. The bound first excludes interior poles. At each cusp a genuine periodic slash translate is bounded by $C{y}^{-k/2}$, which tends to zero. Its punctured-disk expression therefore has a removable singularity with zero value. Conversely cusp Fourier expansions decay exponentially, and finitely many compact core regions control the rest. Positive weight is essential: a nonzero constant in weight zero satisfies the bound but is not a cusp form.

If $f(\tau )={\sum }_{n\ge 1}{a}_n{q}^n$ is a weight-$k$ level-one cusp form, boundedness of its invariant norm givesIntegrate $f(x+iy)e^{-2\pi inx}$ over one period and choose $y=1/n$.

Let even $k\ge 4$ and let $g(\tau )={\sum }_{n\ge 0}{b}_n{q}^n$ be a level-one modular form of weight $k$. If $b_n=O(n^{k/2})$, then $b_0=0$ and $g$ is a cusp form. Otherwise subtract the matching Eisenstein series; its coefficients grow like ${\sigma }_{k-1}(n)$, contradicting the bound at prime indices.

For even $k\ge 12$,The dimension is zero for odd $k$ and for $k<12$.

For $f(\tau )={\sum }_{n\ge 1}{a}_n{q}^n$, termwise integration initially givesA Fricke transformation converts the behavior near zero into cusp decay at infinity, giving analytic continuation and a functional equation for the completed L-function.

With the determinant-normalized slash operator, the map $Jf=i^{k}f{|}_k{W}_N$ satisfies $J^2=1$. The scaled imaginary-axis functions satisfy $F(y)=y^{-k}G(1/y)$, and their Mellin transforms give the displayed entire functional equation. The phase must be specified; omitting it changes the sign in odd or twice-odd weights.

The modular discriminant is the normalized weight-twelve level-one cusp form

The modular discriminant has a simple zero at infinity and no zero in the upper half-plane, by the valence formula for the modular group. Multiplication by it identifies weight-$k$ modular forms with weight-$(k+12)$ cusp forms. Conversely dividing a cusp form by it preserves interior holomorphy and removes one order of vanishing at infinity.

The effective group ${\Gamma }_0(11)$ is torsion-free, has index twelve and has two regular cusps. If its weight-two cusp form space is one-dimensional, any nonzero member has order one at each cusp and no interior zeros by the regular-cusp valence formula on a torsion-free modular curve. Its twelfth power and $\Delta (\tau )\Delta (11\tau )$ have the same weight and divisor. Their ratio is a holomorphic function without poles on the compactified modular curve, hence a nonzero constant. Normalizing the leading Fourier coefficient to one proves the displayed formula, with the holomorphic root chosen to start with $q$. It equals $q{\prod }_{n\ge 1}(1-q^n{)}^2(1-q^{11n}{)}^2$.

Put $S_j={\sum }_{n\ge 1}{\sigma }_j(n)q^n$. Expanding $E_4=1+240S_3$ and $E_6=1-504S_5$ gives $\Delta =(5S_3+7S_5)/12+100S_3^2+8000S_3^3-147S_5^2$. For every integer $d$, $5d^3+7d^5$ is divisible by twelve, by separate congruences modulo three and four. Consequently every coefficient of the modular discriminant is integral, without using a product expansion.

The Ramanujan tau function gives the Fourier coefficients at positive indices of the modular discriminant: $\Delta (z)={\sum }_{n\ge 1}\tau (n)q^n$, with $\tau (1)=1$. The Serre derivative and vanishing of weight-fourteen level-one cusp forms give $qd\Delta /dq=E_2\Delta$ and hence the convolution recurrence

The modular discriminant satisfies $1728\Delta =E_4^3-E_6^2$, while the Serre derivative gives $2\Theta E_6=E_2{E}_6-E_4^2$. Their integral Fourier expansions have $E_4\equiv 1$ and $E_2\equiv E_6(\mathrm{mod}5)$, because $d^5\equiv d(\mathrm{mod}5)$. Reducing the two identities gives $\Delta \equiv \Theta E_6(\mathrm{mod}5)$. The coefficient of $q^n$ in $\Theta E_6$ is $-504n{\sigma }_5(n)\equiv n{\sigma }_5(n)$.

The Fourier expansion of a normalized Eisenstein series and the one-dimensional weight-twelve cusp form space give $691E_{12}=691E_4^3-432000\Delta$. The coefficient of $q^n$, for $n\ge 1$, reduces modulo $691$ to $566{\sigma }_{11}(n)=566\tau (n)$. Since $566$ is invertible modulo $691$, this proves the congruence. Integrality follows from $E_4=1+240\sum {\sigma }_3(n)q^n$ and the product $\Delta =q{\prod }_{n\ge 1}(1-q^n{)}^{24}$.

For every real $s$, the rapidly convergent integralis the analytic continuation of $(2\pi {)}^{-s}\Gamma (s)L(\Delta ,s)$. The product for $\Delta$ makes the integrand positive, so $L(\Delta ,s)>0$ for real $s>0$.

For weight-$k$ cusp forms on $\Gamma$, the Petersson inner product isCusp decay makes the integral convergent.

On level-one cusp forms, determinant-normalized Hecke operators satisfy $⟨T_{n}f,g⟩=⟨f,T_{n}g⟩$. Unfolding their finite correspondences transfers a matrix to its inverse; its adjugate has the same determinant and reverses the correspondence. Since these operators commute, the cusp space has a common orthonormal basis of Hecke eigenforms, with real eigenvalues.

The Rankin–Selberg method represents Dirichlet series built from automorphic forms as integrals against Eisenstein series and studies them by unfolding those integrals.

For two weight-$k$ level-one cusp forms, use the primitive real-analytic Eisenstein series and $d\mu =dxdy/y^2$. The Rankin–Selberg unfolding identity isInitially this follows by absolute convergence, unfolding to $0\le x<1$, and integrating the Fourier series in $x$. Exponential cusp decay and polynomial growth of the primitive real-analytic Eisenstein series make the integral holomorphic for $\mathrm{Re}w>1$, continuing the Dirichlet series to $\mathrm{Re}s>k$.

For real $w>1$, the finite Rankin–Selberg integral $I(f,f,w)$ has a nonnegative integrand. Tonelli theorem unfolds it without an a priori bound on the Dirichlet series. Integration over $x$ gives ${\int }_0^1|f(x+iy){|}^{2}dx={\sum }_{n\ge 1}|a_n{|}^2{e}^{-4\pi ny}$. A second application of Tonelli theorem gives the finite sum in the displayed identity with $s=w+k-1$. Positivity, rather than analytic continuation alone, proves absolute convergence.

By positive unfolding of a cusp-form square, $\sum |a_n{|}^2{n}^{-\beta }<\infty$ for $\beta >k$. Choose $k<\beta <2\sigma -1$. The Cauchy-Schwarz inequality bounds the absolute sum for the L-function of a cusp form by

For cusp forms $f=\sum a_n{q}^n$ and $g=\sum b_n{q}^n$, their Rankin–Selberg convolution in the elementary normalization is $L(f,g,s)={\sum }_{n\ge 1}{a}_n\overline{b_n}{n}^{-s}$.

For normalized level-one Hecke eigenforms, let ${\alpha }_p+{\beta }_p=a_p$, ${\alpha }_p{\beta }_p=p^{k-1}$, and similarly ${\gamma }_p+{\delta }_p=b_p$, ${\gamma }_p{\delta }_p=p^{k-1}$. The Hecke multiplication relations give $a_{p^r}=({\alpha }_p^{r+1}-{\beta }_p^{r+1})/({\alpha }_p-{\beta }_p)$, with the repeated-root case obtained by continuity. Multiplying the two expressions and summing four geometric series proves the displayed local factor. Multiplicativity gives an Euler product; multiplying by $\zeta (2s-2k+2)$ cancels its local numerator and produces the degree-four convolution factors.

For two level-one weight-$k$ cusp forms, $F(s)=\sum a_n{\overline{b}}_n{n}^{-s}$. Unfolding against the nonholomorphic lattice sum normalized by ${\pi }^{-w}\Gamma (w)$ gives the displayed completion as the integral of $f\overline{g}{y}^k$ against that Eisenstein series. Its functional equation is $\mathcal{R}(s)=\mathcal{R}(2k-1-s)$, and its only possible simple poles are at $k-1,k$. The raw series is obtained by dividing out the gamma and zeta factors, so it does not inherit the same unqualified pole statement; zeros of the denominator zeta factor can yield additional possible poles. The raw series is regular at $k-1$ and can have a simple pole at $k$.

If $f$ and $g$ have respective weights $k$ and $l$, and $r=k-l>2$, unfolding the weight-$r$ Eisenstein series gives