Dimension sequences of modular forms
Erick Ross · 2025
Read Dimension sequences of modular forms by Erick Ross (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-12340.txt.
Introduction
Erick Ross studies dimension sequences of modular forms in a 2025 math paper. Modular forms are highly symmetric functions that sit at the center of number theory. For each level, one can look at the space of new cusp forms of weight two and record its dimension, forming a sequence as the level grows.
In 2004, Greg Martin conjectured that this sequence takes every possible natural number as a value. Ross investigates several generalizations and variations of that type of problem. In each case, he provides a complete characterization of when such a property holds. The work serves readers in number theory who study how the size of these form spaces behaves across levels.
Catalog record
| CREATOR | Erick Ross |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-12340
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Dimension sequences of modular forms Erick Ross Abstract . For N ≥ 1 N\geq 1 , let S 2 new ( N ) S_{2}^{\text{new}}(N) denote the newspace of cuspidal modular forms of weight 2 2 and level N N . In 2004, Greg Martin conjectured that as a sequence in N N , dim S 2 new ( N ) \dim S_{2}^{\text{new}}(N) takes on all possible natural numbers. In this paper, we investigate several generalizations and variations of this type of problem. In each case, we provide a complete characterization of when such a property holds. 2020 Mathematics Subject Classification Primary 11F11. 1. Introduction For N,k\geq 1 , we let S_{2k}(N) denote the space of cuspidal modular forms of weight 2k and congruence subgroup \Gamma_{0}(N) , and we let S_{2k}^{\textnormal{new}}(N) denote its new subspace. These spaces are of great importance in Number Theory, and we would like to answer certain questions about how large they can be. In particular, we would like to determine which of the natural numbers can be realized as dimensions of these spaces. (Here, 0 is a natural number.) In 2000, Csirik-Wetherell-Zieve proved that the sequence \{\dim S_{2}(N)\}_{N\geq 1} does not take on all natural numbers, interpreting these values as the genera of modular curves [ 7 , Section 4] . In 2004, Martin considered the possible dimensions of S_{2}^{\textnormal{new}}(N) , conjectured that the sequence \{\dim S_{2}^{\textnormal{new}}(N)\}_{N\geq 1} takes on all possible natural numbers [ 8 , Conjecture 27] . As a corollary of certain explicit bounds on \dim S_{2}^{\textnormal{new}}(N) , we recently disproved this conjecture by showing that 67846 can not realized as a value of \{\dim S_{2}^{\textnormal{new}}(N)\}_{N\geq 1} [ 11 , Section 7] . In this paper, we investigate the question of when precisely results of this nature can occur, considering several generalizations and variations on the above problems. First, we consider general weight 2k and show that, in fact, the generalized version of Martin’s conjecture can never hold. Theorem 1.1 . For each fixed k\geq 1 , the sequences \{\dim S_{2k}(N)\}_{N\geq 1} and \{\dim S_{2k}^{\textnormal{new}}(N)\}_{N\geq 1} both omit some natural number. For each k\geq 1 , we also determine the first omitted dimension for these sequences. The above result for the full space S_{2k}(N) is perhaps not too surprising since its dimension has a growth rate oscillating between N and N\log\log N . However, it is much more surprising for the newspace S_{2k}^{\textnormal{new}}(N) , because the growth rate in N of \dim S_{2k}^{\textnormal{new}}(N) oscillates between N/\log\log N and N . We also consider the corresponding question for fixed level and varying weight. We give a complete classification of when these weight-indexed sequences take on every natural number. Theorem 1.2 . For fixed N\geq 1 , the sequence \{\dim S_{2k}(N)\}_{k\geq 1} takes on every natural number iff N\in\{1,2,3,4\} . Similarly, the sequence \{\dim S_{2k}^{\textnormal{new}}(N)\}_{k\geq 1} takes on every natural number iff N\in\{1,2,3,4,8,12,16,18\} . Finally, we consider the corresponding question on the sign pattern spaces S_{2k}^{\sigma}(N) and S_{2k}^{{\textnormal{new}},\sigma}(N) . Recall that for squarefree N , certain operators are defined on S_{2k}(N) ; the Atkin-Lehner involutions W_{p} for p\mid N . Since the W_{p} are involutions, they can only possibly have eigenvalues of \pm 1 . Additionally, the \{W_{p}\}_{p\mid N} turn out to be a commuting family of self-adjoint operators. Hence, they simultaneously diagonalize S_{2k}(N) , and so S_{2k}(N) can be written as a direct sum of the simultaneous eigenspaces of the W_{p} . Define the sign patterns \sigma for N to be the nonzero multiplicative functions on the divisors of N such that \sigma(p)\in\{\pm 1\} for each p\mid N . Then S_{2k}(N)=\bigoplus_{\sigma}S_{2k}^{\sigma}(N),\quad\text{where}\ \ S_{2k}^{\sigma}(N)=\{f\in S_{2k}(N)\ \colon\ W_{p}f=\sigma(p)f\ \ \text{for }p\mid N\}. In other words, the sign pattern spaces S_{2k}^{\sigma}(N) are defined by specifying the sign of the W_{p} eigenvalues for each p\mid N . One can also define the sign pattern newspaces S_{2k}^{{\textnormal{new}},\sigma}(N) in the same way [ 9 , Section 3] . We will study the weight-indexed dimension sequences \{\dim S_{2k}^{\sigma}(N)\}_{k\geq 1} for these sign pattern spaces. (Note that level-indexed sequences would not make sense here since the sign pattern depends on the level.) Of course there are many different decompositions of S_{2k}(N) and S_{2k}^{\textnormal{new}}(N) we could have considered. We chose to study the sign pattern decomposition in particular because this is, in some sense, the finest possible way to decompose S_{2k}^{\textnormal{new}}(N) in a natural way (i.e. that respects the Hecke operators and Galois conjugation) [ 4 , Section 5] . This gives the dimension sequences for these spaces the “best chance” to take on every natural number. For N and \sigma fixed, we classify precisely when the dimension sequences for the sign pattern spaces take on all natural numbers. For shorthand, we will write a sign pattern \sigma for N=p_{1}\cdots p_{t} (with p_{1}<\ldots<p_{t} ) as the sequence of signs \sigma(p_{1}),\ldots,\sigma(p_{t}) . For example, the sign pattern \sigma for N=30 mapping 2\mapsto-1 ; 3\mapsto+1 ; 5\mapsto+1 would be written just as \sigma=-++ . Theorem 1.3 . For fixed squarefree N\geq 1 with sign pattern \sigma , the sequence \{\dim S_{2k}^{\sigma}(N)\}_{k\geq 1} takes on every natural number iff N\in\{1,2,3,5,6,7,10,11,14,15\} and (N,\sigma)\neq(10,+-) . Similarly, the sequence \{\dim S_{2k}^{{\textnormal{new}},\sigma}(N)\}_{k\geq 1} takes on every natural number iff N\in\{1,\allowbreak 2,\allowbreak 3,\allowbreak 5,\allowbreak 6,\allowbreak 7,\allowbreak 10,\allowbreak 14,\allowbreak 15,\allowbreak 21,\allowbreak 26,\allowbreak 30,\allowbreak 42,\allowbreak 66,\allowbreak 70,\allowbreak 78,\allowbreak 210\} or (N,\sigma) is in the list given in Table 5.2 . Finally, we point out an interesting phenomenon that we observed regarding a closely related question. One could ask about when a dimension sequence takes on every natural number exactly once. It turns out that for the full sign pattern spaces, whether or not a dimension sequence satisfies this property depends only on its multiset density. Given a multiset A of natural numbers, the multiset density of A is defined to be \rho(A):=\lim_{x\to\infty}\frac{\#\{a\in A\colon a\leq x\}}{x}, assuming this limit exists. We note that this definition only differs from the typical notion of asymptotic density in that we are considering A here to be a multiset instead of a set (i.e., we count the elements of A with multiplicity). For our purposes, this will be more useful to consider than the typical asymptotic density since we are interested in sequences which may have repeated values. This also means that for sequences, the multiset density can be thought of describing a more general notion of the linear growth rate of a sequence. In the simplest case, a sequence \{a_{k}\}_{k\geq 1} of the form a_{k}=\alpha k+o(k) will have multiset density 1/\alpha . Then we have the following result, which follows from Theorem 1.3 . Corollary 1.4 . For fixed squarefree N\geq 1 with sign pattern \sigma , the sequence \{\dim S_{2k}^{\sigma}(N)\}_{k\geq 1} takes on every natural number exactly once iff it has multiset density \rho=1 . Because of the classification given by Theorem 1.3 , this result is not all that difficult to prove. However, it is quite surprising. Nothing close to this result holds for the other spaces we investigated; this structure emerges only for the full sign pattern dimension sequences \{\dim S_{2k}^{\sigma}(N)\}_{k\geq 1} . This pattern is also somewhat surprising given the manner in which the dimension sequences in question satisfy the desired property. For (N,\sigma)=(14,-+) , the dimension sequence takes on every