A non-hypergeometric \(E\)-function (Q2058615)

In the landmark paper [Abh. Preuß. Akad. Wiss., Phys.-Math. Kl. 1929, No. 1, 70 S. (1929; JFM 56.0180.05)], \textit{C. L. Siegel} defined in 1929 the class of \(E\)-functions. These functions are power series of the form \(\sum_{n=0}^\infty a_n z^n/n! \in \overline{\mathbb Q}[[z]]\) solutions of a linear differential equation with coefficients in \(\overline{\mathbb Q}[z]\) and whose Taylor coefficients \(a_n\) satisfy certain archimedean and non-archimedean growth conditions. The simplest example of an \(E\)-function is \(\exp(z)\). Siegel also showed that the series of the form \({}_pF_q[a_1, \ldots, a_p; b_1, \ldots, b_q; z^{q-p+1}]\) are \(E\)-functions when \(q\ge p\), the \(a_j\)'s and \(b_j\)'s are rational numbers and non of the \(b_j\)'s is a negative integer. They are called hypergeometric \(E\)-functions. Siegel's motivation was to extend the Diophantine properties of the exponential function to \(E\)-functions, and for instance he proved the transcendence of the values of Bessel's function \(J_0\) at any non-zero algebraic point. His program culminated with the celebrated Siegel-Shidlovskii Theorem (1956) concerning the transcendence degree over \(\overline{\mathbb Q}(z)\) of the values of \(E\)-functions solutions of a linear differential system with coefficients in \(\overline{\mathbb Q}(z)\). The exact nature of \(E\)-functions remains largely mysterious and \textit{C. L. Siegel} had asked in [Transcendental numbers. Princeton, NJ: Princeton University Press (1949; Zbl 0039.04402)] if the class of \(E\)-functions coincides with the \(\overline{\mathbb Q}[z]\)-algebra generated by hypergeometric \(E\)-functions as defined above. Very few progress had been made on this question till 2004 when \textit{V. A. Gorelov} proved that the answer to Siegel's question if positive for each \(E\)-function solution of a differential equation of order \(\le 2\) [Math. Notes 75, No. 4, 513--529 (2004; Zbl 1062.34096); translation from Mat. Zametki 75, No. 4, 549--565 (2004)]. However, in 2019, \textit{S. Fischler} and the reviewer [``On Siegel's problem for E-functions'', Rend. Semin. Mat. Univ. Padova (to appear), \url{arXiv:1910.06817}] gave a strong reason to believe that the answer was negative in general: they showed that a positive answer would contradict a generalization to exponential periods of Grothendieck's classical periods conjecture, recently formulated by Fresán and Jossen. In the beautiful paper under review, Fresán and Jossen prove unconditionally that the answer to Siegel's question is indeed negative in general: amongst many possible examples, they show that the \(E\)-function \[ \sum_{n=0}^\infty \Big(\sum_{m=0}^{[2n/3]} \frac{(1/4)_{n-m}}{(2n-3m)!(2m)!}\Big)z^n \] is transcendental over the \(\overline{\mathbb Q}[z]\)-algebra generated by hypergeometric \(E\)-functions. Their method (formulated in terms of \(\mathcal{D}\)-modules and tannakian categories) is based for one part on \textit{Y. André}'s theory of \(E\)-operators [Ann. Math. (2) 151, No. 2, 705--740 (2000; Zbl 1037.11049)] and for the other part on \textit{N. M. Katz}'s classification [Exponential sums and differential equations. Princeton, NJ: Princeton University Press (1990; Zbl 0731.14008)] of the possible differential Galois groups of differential equations satisfied by (polynomials in) hypergeometric \(E\)-functions. Their key observation is that in certain circumstances, the properties of the Galois groups in Katz's classification imply that the singularities of the Fourier-Laplace transform of the \(E\)-operators satisfied by \(E\)-functions should be either collinear or form an equilateral triangle if Siegel's question has a positive answer. They then prove that such a geometric configuration is not always satisfied by \(E\)-functions that fall under the scope of these circumstances, in particular for the explicit \(E\)-function displayed above.