Iterated integrals and higher order automorphic forms (Q2493391)

The theory of higher-order automorphic forms has come into existence only within the past five years or so, but it already has an interesting history thanks to the work of a number of leading mathematicians. In [Acta Arith. 103, No. 3, 209--223 (2002; Zbl 1020.11025)], \textit{G. Chinta}, \textit{N. Diamantis} and \textit{C. O'Sullivan} introduced holomorphic higher-order forms and the larger space of smooth \((C^\infty)\) higher-order forms, and classified the latter for order \(s= 2\). In both spaces the forms are ``entire'' in the Hecke sense; that is, they have at most polynomial growth at the parabolic cusps of the relevant Fuchsian group. (The interest in the larger space of smooth forms appears to stem from the possibility of a future study of higher-order Maass wave forms.) The purpose of the article under review is to generalize the classification theorem of the 2002 paper to higher-order automorphic forms of all orders \(\geq 2\). Let \(\Gamma\) be a finitely generated Fuchsian group of the first kind (i.e., with finite hyperbolic volume), acting on the upper half-plane \(H\) and having parabolic elements. For their classification result (Theorem 4.4) the authors require \(\Gamma\) to be torsion-free as well, let \(k\in 2\mathbb Z^+\). If \(f\) is a function defined on \(H\), consider the operator \((f|_k\gamma)(z)= (cz+ d)^{-k} f(\gamma z)\), for \(\gamma= (\begin{smallmatrix} * & *\\ c & d\end{smallmatrix})\in \Gamma\). Traditionally, an automorphic form of weight \(k\) satisfies the transformation law \[ f|_k(\gamma- 1)= 0\quad\text{for all }\gamma\in\Gamma.\tag{\(*\)} \] In the definition of higher-order automorphic form of weight \(k\) and order \(s\) \((s\in \mathbb Z^+)\), \((*)\) is replaced by the more general transformation property \[ f|_k(\gamma_1- 1)\cdots(\gamma_s- 1)= 0\quad\text{for }\gamma_j\in \Gamma,\;1\leq j\leq s.\tag{\(**\)} \] Theorem 4.4 may be stated as follows. Suppose \(\Gamma\) has \(2g\) hyperbolic generators and \(m\geq 1\) parabolic generators. (For \(\Gamma\) torsion-free there are no elliptic generators.) Let \(M^s_k\) denote the space of smooth, entire automorphic forms of weight \(k\) and order \(s\geq 2\). Then, \[ M^{s+1}_k\cong M^s_k\oplus M_k',\tag{\(***\)} \] where the indicated isomorphism refers to the vector space structure and the index \(i\) runs from 1 to \((2g+ m-1)^2\). Note that \((***)\) can be iterated to obtain \(M^{s+1}_k\) as (an isomorphic image of) a direct sum of \[ \sum^s_{\ell= 0}(2g+ m-1)^\ell \] copies of \(M_k'\). (The space of order 2 forms characterized in the 2002 paper is somewhat smaller than that which appears here, since in the earlier article the authors imposed the further condition, \(f|_k(\pi- 1)= 0\), for all parabolic \(\pi\in\Gamma\).) \textit{K.-T. Chen's} result on iterated integrals referred to in the title [Trans. Am. Math. Soc. 156, 359--379 (1971; Zbl 0217.47705), Theorem 3.1] is needed here to establish Proposition 4.1, and in the authors' proof of this proposition the application of Chen's Theorem 3.1 requires that \(\pi_1(\Gamma/H^+)= \Gamma\). This fails if \(\Gamma\) has elements of finite order. We note that, in the absence of Proposition 4.1, the result here would be the much weaker statement: \(M^{s+1}_k\cong M^s_k\oplus S\), with \(S\) some subspace of the direct sum of \((2q+ m-1)^2\) copies of \(M_k'\). Since the Acta paper derives the characterization in the case \(s= 2\) without the assumption that \(\Gamma\) is free of torsion, it seems reasonable to suspect that the result of the present work carries over to the more general case as well.