## Multiplicities of representations of $\Gamma /{\Gamma }_{1}$in ${M}_{k}\left({\Gamma }_{1}\right)$and ${S}_{k}\left({\Gamma }_{1}\right)$

### 2002.9

Abstract
This paper shows that for an given irreducible representation $\rho$  of $\Gamma /{\Gamma }_{1}$  , the two functions $dim\left({M}_{k}\left({\Gamma }_{1},\rho \right)\right)$  and $dim\left({S}_{k}\left({\Gamma }_{1},\rho \right)\right)$  of $k$  are almost linear functions.

1 Main theorem

Let $\Gamma$  and ${\Gamma }_{1}$  be two Fuchsian subgroups of $S{L}_{2}\left(\mathbb{Q}\right)$  of the first type, and assume that ${\Gamma }_{1}$  is a normal subgroup of $\Gamma$  with finite index.
For ${\Gamma }_{0}=\Gamma$  or ${\Gamma }_{1}$  , let ${M}_{k}\left({\Gamma }_{0}\right)$  ( ${A}_{k}\left({\Gamma }_{0}\right)$  , ${S}_{k}\left({\Gamma }_{0}\right)$  ) be the space of modular forms (meromorphic modular forms, cusp forms) of ${\Gamma }_{0}$  with weight $k$  . Let $Y=\Gamma \{\mathbb{H}}^{*}$  and $X={\Gamma }_{1}\{\mathbb{H}}^{*}$  . Let $\Omega \left(X\right)$  ( $\Omega \left(Y\right)$  ) be the field of meromorphic functions of $X$  ( $Y$  ). Then ${A}_{0}\left(\Gamma \right)\sim =\Omega \left(Y\right)$  and ${A}_{0}\left({\Gamma }_{1}\right)\sim =\Omega \left(X\right)$  , in a natural way.
$\Gamma$  has a representation ${\pi }_{k}$  in ${A}_{k}\left({\Gamma }_{1}\right)$  defined by
 $\begin{array}{c}\left({\pi }_{k}\left(\gamma \right)f\right)\left(z\right)=f{|}_{\left[}{\gamma }^{-1}{\right]}_{k}\left(z\right)=\left(cz+d{\right)}^{-k}f\left({\gamma }^{-1}\left(z\right)\right)\end{array}$ (1),
where ${\gamma }^{-1}=\left(\begin{array}{cc}a& b\\ c& d\end{array}\right)\in \Gamma$  and $f\in {A}_{k}\left({\Gamma }_{1}\right)$  . It is easy to see that ${\pi }_{k}$  is a representation of $\Gamma$  and ${\pi }_{k}\left({\Gamma }_{1}\right)$  is trivial. Therefore we obtain a representation of $\Gamma /{\Gamma }_{1}$  . We wrote ${\pi }_{k}$  for it, too. Notice that ${\pi }_{k}$  stables ${M}_{k}\left({\Gamma }_{1}\right)$  and ${S}_{k}\left({\Gamma }_{1}\right)$  . We write ${\rho }_{k}$  ( ${\sigma }_{k}$  ) for the restriction of ${\pi }_{k}$  to ${M}_{k}\left({\Gamma }_{1}\right)$  ( ${S}_{k}\left({\Gamma }_{1}\right)$  ).
Let $Rep$  ( $Re{p}_{0}$  ) be the set of all representations of $\Gamma /{\Gamma }_{1}$  ( $\overline{\Gamma }/\overline{{\Gamma }_{1}}$  ), where $\overline{\Gamma }=\Gamma /\Gamma \cap \left\{±\mathbf{I}\right\}$  and $\overline{{\Gamma }_{1}}={\Gamma }_{1}/{\Gamma }_{1}\cap \left\{±\mathbf{I}\right\}$  . In this paper, we regard a representations of $\overline{\Gamma }/\overline{{\Gamma }_{1}}$  as a representation of $\Gamma /{\Gamma }_{1}$  naturally without mentioned. Let $R$  ( ${R}_{0}$  ) be the regular representation of $\Gamma /{\Gamma }_{1}$  ( $\overline{\Gamma }/\overline{{\Gamma }_{1}}$  ). If $-\mathbf{I}$  is not in $\Gamma$  , then $Rep=Re{p}_{0}$  , $R={R}_{0}$  , and let $Re{p}_{1}=Rep$  , ${R}_{1}=R$  . If $-\mathbf{I}$  is in ${\Gamma }_{1}$  , then $Rep=Re{p}_{0}$  , $R={R}_{0}$  , and let $Re{p}_{1}$  be empty. If $-\mathbf{I}$  is in $\Gamma$  but not in ${\Gamma }_{1}$  , then $Rep\ne Re{p}_{0}$  , $R\ne {R}_{0}$  and let $Re{p}_{1}=Rep\Re{p}_{0}$  , ${R}_{1}$  be the representation of $\Gamma /{\Gamma }_{1}$  such that $R={R}_{0}\oplus {R}_{1}$  . Then ${R}_{i}={\oplus }_{\rho \in Re{p}_{i}}dim\left(\rho \right)\rho \left(i=0,1\right)$  .
Theory of modular forms tells us that ${\rho }_{k}$  and ${\sigma }_{k}$  have decompositions
 $\begin{array}{c}{\rho }_{k}={\oplus }_{\rho \in Re{p}_{i}}{M}_{k}\left({\Gamma }_{1},\rho \right)\end{array}$ (2),
and
 $\begin{array}{c}{\sigma }_{k}={\oplus }_{\rho \in Re{p}_{i}}{S}_{k}\left({\Gamma }_{1},\rho \right)\end{array}$ (3),
where $i=0$  or 1 and $i\equiv k\left(mod2\right)$  .
The main result of this paper is the following theorem.
Theorem Assume $\rho \in Re{p}_{i}$  ( $i=0,1$  ), then
 $\begin{array}{c}{lim}_{k\to \infty ,k\equiv i\left(mod2\right)}\frac{dim{M}_{k}\left({\Gamma }_{1},\rho \right)}{kdim\left(\rho {\right)}^{2}},{lim}_{k\to \infty ,k\equiv i\left(mod2\right)}\frac{dim{S}_{k}\left({\Gamma }_{1},\rho \right)}{kdim\left(\rho {\right)}^{2}}\end{array}$ (4)
exist and both equal to $\frac{1}{4\pi }{\int }_{\Gamma \\mathbb{H}}\frac{dxdy}{{y}^{2}}$  .
We give an important application of the above theorem.
Let $\Gamma$  be ${\Gamma }_{0}\left(N\right)$  and ${\Gamma }_{1}$  be ${\Gamma }_{1}\left(N\right)$  . Notice that $l\left({\Gamma }_{0}\left(N\right)\right)=N{\prod }_{p:prime,p|N}\left(1+{p}^{-1}\right)$  . If $\chi \left(-I\right)=1$  , we have ${lim}_{k\to \infty ,keven}\frac{dim\left({M}_{k}\left(\chi ,N\right)\right)}{k}={lim}_{k\to \infty ,keven}\frac{dim\left({S}_{k}\left(\chi ,N\right)\right)}{k}=N{\prod }_{p:prime,p|N}\left(1+{p}^{-1}\right).$  If $\chi \left(-I\right)=-1$  , we have ${lim}_{k\to \infty ,kodd}\frac{dim\left({M}_{k}\left(\chi ,N\right)\right)}{k}={lim}_{k\to \infty ,kodd}\frac{dim\left({S}_{k}\left(\chi ,N\right)\right)}{k}=N{\prod }_{p:prime,p|N}\left(1+{p}^{-1}\right).$

2 Proof of the main result

Let $X$  , $Y$  be the same as in section 1. Write $\mu$  for $\left[\overline{\Gamma }:\overline{{\Gamma }_{1}}\right]$  . Let $i=0,1$  and when $i=1$  , we assume $-\mathbf{I}$  is not in ${\Gamma }_{1}$  .
Select out a $p\in \Gamma \\mathbb{H}$  which has $\mu$  different lifts ${p}_{1},...,{p}_{\mu }$  in $X$  . Select out a lift ${\stackrel{~}{p}}_{j}$  in $\mathbb{H}$  for each ${p}_{j}$  .
By Riemann-Roch theorem, we can find an ${f}_{1}\in {A}_{i}\left({\Gamma }_{1}\right)$  such that ${f}_{1}\left({\stackrel{~}{p}}_{1}\right)=1$  and ${f}_{1}\left({\stackrel{~}{p}}_{j}\right)=0$  for $j\in \left\{2,...,\mu \right\}$  . For any $\alpha \in \Gamma /{\Gamma }_{1}$  , we can set ${f}_{\alpha }={f}_{1}|\left[\alpha {\right]}_{i}$  . Let ${W}_{i}$  be the $\mathbb{C}$  -linear subspace of ${A}_{i}\left({\Gamma }_{1}\right)$  spanned by $\left\{{f}_{\alpha }:\alpha \in \Gamma /{\Gamma }_{1}\right\}$  . It is easy to see that ${W}_{i}$  is stable under ${\pi }_{i}\left(\Gamma /{\Gamma }_{1}\right)$  .
We have the following lemma.
Lemma ${A}_{0}\left(\Gamma \right){\otimes }_{\mathbb{C}}{W}_{i}\sim ={A}_{i}\left({\Gamma }_{1}\right)$  . And ${W}_{i}$  and ${R}_{i}$  are two isomorphic representations of $\Gamma /{\Gamma }_{1}$  .
• Proof. We only need to find $\mu$  elements in ${W}_{i}$  which are linear independent over ${A}_{0}\left(\Gamma \right)$  . We can select out a subset $T$  of $\Gamma /{\Gamma }_{1}$  , which is a lift of $\overline{\Gamma }/\overline{{\Gamma }_{1}}$  . We are going to show that $\left\{{f}_{\alpha }:\alpha \in T\right\}$  is linear independent over ${A}_{0}\left(\Gamma \right)$  . Otherwise, there are ${g}_{\alpha }\in {A}_{0}\left(\Gamma \right)\left(\alpha \in T\right)$  , not all zero, such that ${\sum }_{\alpha \in T}{g}_{\alpha }{f}_{\alpha }=0$  . We may assume all ${g}_{\alpha }$  are holomorphic at ${\stackrel{~}{p}}_{1},...,{\stackrel{~}{p}}_{\mu }$  , and moreover ${g}_{\beta }\left({\stackrel{~}{p}}_{1}\right)=1$  for some $\beta \in T$  . Then $0={\sum }_{\alpha \in T}{g}_{\alpha }{f}_{\alpha }{|}_{\left[{\beta }^{-1}{\right]}_{i}}.$  But the right side of the above equality equals 1 at ${\stackrel{~}{p}}_{1}$  . This is a contradiction.
Therefore ${A}_{0}\left(\Gamma \right){\otimes }_{\mathbb{C}}{W}_{i}\sim ={A}_{i}\left({\Gamma }_{1}\right)$  .
Comparing the traces, we find that ${W}_{i}\sim ={R}_{i}$  .
We are going to prove the main theorem.
Find an even ${n}_{0}$  , and a $\phi \in {S}_{{n}_{0}\left(\Gamma \right)}$  such that $\phi \cdot {f}_{\alpha }$  are all cusp forms. Let ${W}_{\phi }$  be the $\mathbb{C}$  -linear subspace of ${S}_{{n}_{0}}\left(\Gamma \right)$  spanned by $\left\{\phi {f}_{\alpha }:\alpha \in \Gamma /{\Gamma }_{1}\right\}$  . Then ${W}_{\phi }{\otimes }_{\mathbb{C}}{S}_{k}\left(\Gamma \right)⟶{S}_{i+{n}_{0}+k}\left({\Gamma }_{1}\right)$  and ${W}_{\phi }{\otimes }_{\mathbb{C}}{M}_{k}\left(\Gamma \right)⟶{M}_{i+{n}_{0}+k}\left({\Gamma }_{1}\right)$  are both injective.
Hence for $\rho \in Re{p}_{i}$  , ${liminf}_{k\to \infty ,2|k}\frac{\left[{\rho }_{i+{n}_{0}+k}:\rho \right]}{i+{n}_{0}+k}\ge {liminf}_{k\to \infty ,2|k}\frac{dim\left({M}_{k}\left(\Gamma \right)\right)\cdot \left[{R}_{i}:\rho \right]}{i+{n}_{0}+k}$
 $\begin{array}{c}={liminf}_{k\to \infty ,k\equiv imod2}\frac{dim\left({M}_{k}\left(\Gamma \right)\right)dim\left(\rho \right)}{k}\end{array}$ (5).
Shimura [2] tells us that $l\left(\Gamma \right)=\left\{\begin{array}{cc}{lim}_{k\to \infty }\frac{{M}_{k}\left(\Gamma \left(N\right)\right)}{k}={lim}_{k\to \infty }\frac{{S}_{k}\left(\Gamma \left(N\right)\right)}{k}& \left(-\mathbf{I}\mathbf{i}\mathbf{s}\mathbf{n}\mathbf{o}\mathbf{t}\mathbf{i}\mathbf{n}\Gamma \right)\\ {lim}_{k\to \infty keven}\frac{{M}_{k}\left(\Gamma \left(N\right)\right)}{k}={lim}_{k\to \infty keven}\frac{{S}_{k}\left(\Gamma \left(N\right)\right)}{k}& \left(-\mathbf{I}\in \Gamma \right)\end{array}$  exists and equals to $\frac{1}{4\pi }{\int }_{\Gamma \\mathbb{H}}\frac{\mathbf{d}x\mathbf{d}y}{{y}^{2}}$  . So $l\left({\Gamma }_{1}\right)=l\left(\Gamma \right)\cdot \mu ,$  and
 $\begin{array}{c}{lim}_{k\to \infty k\equiv i\left(mod2\right)}\frac{dim\left({\rho }_{k}\right)}{k}=l\left(\Gamma \right)\cdot \mu \end{array}$ (6).
Comparing (5) and (6), we find that
 $\begin{array}{c}{lim}_{k\to \infty k\equiv i\left(mod2\right)}\frac{\left[{\rho }_{k}:\rho \right]}{k}=l\left(\Gamma \right)\cdot dim\left(\rho \right)\end{array}$ (7).
Similarly, we have
 $\begin{array}{c}{lim}_{k\to \infty k\equiv i\left(mod2\right)}\frac{\left[{\sigma }_{k}:\rho \right]}{k}=l\left(\Gamma \right)\cdot dim\left(\rho \right)\end{array}$ (8).
We have now proved the main result of this paper.

3 A example

Let $\Gamma$  be ${\Gamma }_{0}\left(N\right)$  and ${\Gamma }_{1}$  be ${\Gamma }_{1}\left(N\right)$  . Let $\chi \left(modN\right)$  be a Dirichlet's character. Let ${M}_{k}\left(N,\chi \right)$  be the space $\left\{f\in {M}_{k}\left({\Gamma }_{0}\left(N\right)\right):f{|}_{\left[\gamma {\right]}_{k}}=\chi \left(d\right)f,\gamma =\left(\begin{array}{cc}a& b\\ c& d\end{array}\right)\in {\Gamma }_{0}\left(N\right)\right\}$  and ${S}_{k}\left(N,\chi \right)$  be the space ${M}_{k}\left(N,\chi \right)\cap {S}_{k}\left(N\right)$  . Notice that $l\left({\Gamma }_{0}\left(N\right)\right)=N{\prod }_{p:prime,p|N}\left(1+{p}^{-1}\right)$  .
If $\chi \left(-I\right)=1$  , we have ${lim}_{k\to \infty ,keven}\frac{dim\left({M}_{k}\left(\chi ,N\right)\right)}{k}={lim}_{k\to \infty ,keven}\frac{dim\left({S}_{k}\left(\chi ,N\right)\right)}{k}=N{\prod }_{p:prime,p|N}\left(1+{p}^{-1}\right).$  If $\chi \left(-I\right)=-1$  , we have ${lim}_{k\to \infty ,kodd}\frac{dim\left({M}_{k}\left(\chi ,N\right)\right)}{k}={lim}_{k\to \infty ,kodd}\frac{dim\left({S}_{k}\left(\chi ,N\right)\right)}{k}=N{\prod }_{p:prime,p|N}\left(1+{p}^{-1}\right).$  References

1. J.P.Serre. Linear Representations of Finite Groups. Spring-Verlag, New York. 1977
2. G.Shimura. Introduction to the Arithmetic Theory of Automorphic Functions. Princeton University Press, Princeton, New Jersey. 1971
3. O.Forster. Lectures on Riemann Surfaces. Springer-Verlag, New York. 1981
4. N.Koblitz. Introduction to Elliptic Curves and Modular Forms. Springer-Verlag, New York. 1984

$\text{0}$  E-mail: byhsie@math.pku.edu.cn