Large values of modular forms (Q2510909)

Here are some of the results in the paper a) Theorem 1.1 says : There exist arbitrarily large values \(N\) such that the set \(S_2^*(N, \chi)\) of primitive forms of weight \(2\), level \(N\) and nebentypus \(\chi\) (i.e., consisting of forms \(f(z) = y F(z)\), \(F\) being a holomorphic, cuspidal primitive and hence eigenform of weight 2, level \(N\), nebentypus \(\chi\)) contains an element \(f\) with \(\| f\|_{\infty} \gg N^{1/4}\), disproving the folklore conjecture that \(\| f\|_{\infty}\ll_{\varepsilon} N^{\varepsilon}\). The method of proof is specific to non-square free levels. b) Theorem 1.2 shows that for any primitive character \(\chi\) modulo \(N\), and any primitive form \(f \in S_2^*(N, \chi)\), \[ \| f\|_{\infty} \gg \prod_{\substack{ c \in \mathbb N,\, p\operatorname{prime}\\ p^c | N \\ p^{c + 1} \nmid N}} p^{(1/2) \lfloor c/2 \rfloor}, \] (which is greater than \(N^{1/4}\) if \(N\) is a perfect square). In other words, there are plenty of \(f \in S_2^*(N, \chi)\) with \(\| f\|_{\infty}\) large. c) Theorem 1.3 is a more general statement for Hilbert-Maass cusp forms, giving a lower bound for the sup norm in terms of the level, weight and the \(L^2\)-norm, which we do not reproduce in this review. d) The statement of Theorem 1.1 is refined in Theorem 2.2 by showing that for an even primitive Dirichlet character \(\chi\) modulo \(p^2\), \(p\) a variable prime, there exists (an explicitly defined) \(z_{\chi}\) in the upper half plane \(\mathfrak h\) such that for all \(f \in S_2^*(p^2, \chi)\), \(|f(z_{\chi})| \gg p^{1/2}\). e) A consequence of Theorem 2.2 is Proposition 2.1, which says that the average \[ M_{\chi}(z) := \sum _{\substack{ g \mathrm{\;in\;an} \\ \mathrm{o.b.\;of\;}S_2(p^2, \chi)}} |g(z)|^2, \, z \in \mathfrak h \] satisfies, for all even primitive Dirichlet characters \(\chi\) modulo \(p^2\), that \(\| M_{\chi}\|_{\infty} \gg_{\varepsilon} (p^2)^{(1/2) - \varepsilon}\). In contrast, by results of \textit{J. Jorgenson} and \textit{J. Kramer} [Geom. Funct. Anal. 14, No. 6, 1267--1277 (2004; Zbl 1078.11027); Ann. Math. (2) 170, No. 1, 1--43 (2009; Zbl 1169.14020)], \(M(z) := M_{\mathbf{1}}(z)\) and the average \(M_{\Gamma_1(N)}(z)\) of the \(M_{\chi}(z)\) over all (not necessarily primitive) even Dirichlet characters, satisfy \(\| M\|_{\infty} \ll 1, \| M_{\Gamma_1(N)}\|_{\infty} \ll 1\). The author explains why this could be interpreted as saying that for certain \(z \in \mathfrak h\), \(|f(z)|\) fluctuates a lot with \(f\). f) In Proposition 2.3 the author establishes bounds involving \(L^r\)-norms of elements of \(S_2^*(p^2, \chi)\), namely, for \(r > 4\), he shows that \(\| f\|_r \rightarrow \infty\) as \(p \rightarrow \infty\) uniformly for all \(L^2\)-normalized \(f \in S_2^*(p^2, \chi)\), \(\chi\) a primitive character modulo \(p^2\). g) In Proposition 2.4, a uniform version of the classical Wilton estimate, the author gives an upper bound for Wilton partial sums, that involves \(\|f\|_{\infty}\) on the right side. { Idea of proofs of Theorems 1.1 and 1.2}. If \(\varphi\) is the automorphic form on \(\mathrm{GL}_2(\mathbb Q) \backslash \mathrm{GL}_2(\mathbb A)\) that corresponds to \(f\), the author notes in Section 3 that \(z \mapsto |f(z)|\) and \(z \mapsto |\varphi(g)|\) have identical images in \(\mathbb R^+\). Since the corresponding Whittaker function \(W\) on \(\mathrm{GL}_2(\mathbb A)\) is obtained by averaging \(f\) against a character on a suitable unipotent subgroup, a lower bound for \(|f|\) can be obtained from one for \(|W|\), or equivalently, for each \(|W_p|\) and \(|W_{\infty}|\) (and one hopes that this gives a good lower bound). In other words, one is integrating \(f\) against a character on a horocycle -- something that yields good results only when \(N\) is not square free, because it is then that the number of cusps exceeds the number of Atkin-Lehner involutions (the Atkin-Lehner involutions force Fourier expansions at cusps to be related, and the cusp \(i \infty\) is not ``good enough''). Continuing in Section 3, the author determines restrictions imposed by primitivity (of \(\chi\) and \(f\)) on the local components of the automorphic representation associated to \(f\), namely these have to be `twist minimal principal series'. This reduces the problem to computing the supremum value of local Whittaker new vectors \(W_0\) for such representations (normalized by \(W_0 = 1\)), which is done in Section 4. The remaining results (i.e., other than Theorems 1.1 and Theorem 1.2) are proved in Section 5. This is a very well written paper, and the author is quite generous with explaining the motivation and strategy behind his proofs, as well as in giving historical context, informal intuition, comments on significance of results and comments on what a true bound for \(\| f\|_{\infty}\) might be, etc.