On \(q\)-summation and confluence (Q1010977)

One justification (among many) for the interest of \(q\)-special functions is that they usually are deformations of classical special functions. For instance, the \textit{basic} hypergeometric series \({}_{2} \Phi_{1}(a,b;c;q,z)\) introduced by Heine degenerates into the \textit{classical} hypergeometric series \({}_{2} F_{1}(\alpha,\beta;\gamma;z)\) when \(q \to 1\) provided, for instance, \(a = q^{\alpha}\), \(b = q^{\beta}\) and \(c = q^{\gamma}\). In the simplest sense, this means that each coefficient of the series \({}_{2} \Phi_{1}(a,b;c;q,z)\) tends to the corresponding coefficient of the series \({}_{2} F_{1}(\alpha,\beta;\gamma;z)\). In a more elaborate sense, one finds that \({}_{2} \Phi_{1}\) satisfies a Fuchsian linear \(q\)-difference equation of order \(2\) that degenerates into a a Fuchsian linear differential equation of order \(2\) (the so-called hypergeometric equation) according to the following rule: setting \(d_{q} f(z) := \frac{f(q z) - f(z)}{(q - 1) z}\), one has \(d_{q} \to d/dz\). In a still more elaborate sense, it was shown by the reviewer in [Ann. Inst. Fourier 50, No.~4, 1021--1071 (2000; Zbl 0957.05012) and Ann. Sci. Éc. Norm. Supér. (4) 36, No.~6, 925--968 (2003; Zbl 1053.39033)] that ``galoisian'' objects linked to \({}_{2} \Phi_{1}\) (\textit{e.g.} its Birkhoff connection matrix) degenerate into corresponding objects linked to \({}_{2} F_{1}\) (\textit{e.g.} its monodromy matrix). This is part of a theory limited to \textit{fuchsian} \(q\)-difference and differential equations. The degeneracy process alluded to was nicknamed \textit{confluence} by Ramis, because it bears some resemblance to the degeneracy of the usual hypergeometric series to the irregular ones by confluence of their singularities. It is a natural goal to extend the theory to irregular \(q\)-difference and differential equations, in particular to the Stokes phenomenon; and the paper under review is the first breakthrough in that direction. The first obvious case to consider is the Euler series \(\sum_{n \geq 0} (-1)^{n} n! x^{n+1}\), which is solution of the differential equation \(x^{2} y' + y = x\). (Indeed, this equation is a touchstone for any theory of the Stokes phenomenon.) A natural \(q\)-deformation to consider is the \(q\)-series: \[ \hat{E}_{q}(x) := \sum_{n \geq 0} (-1)^{n} [n!]_{q} x^{n+1}, \text{~where~} [n!]_{q} := \prod_{k=1}^{n} \frac{q^{k} - 1}{q - 1}, \] which is solution of the \(q\)-deformed equation \(x^{2} d_{q} y + y = x\). (This is a non homogeneous first order equation, but it could be turned into a homogeneous second order equation.) Now, a curious dichotomy arises: if \(|q| > 1\), the series diverges and, as well for the \(q\)-difference as for the differential equation, we are in the realm of irregular equations. If \(|q| < 1\) however the series \(\hat{E}_{q}\) converges with radius of convergence \(|1 - q|\), which is pinched to zero when \(q \to 1\); although the \(q\)-difference equation is Fuchsian, the above mentioned theory of confluence does not apply. (The case \(|q| = 1\) is not considered here, as it involves completely different methods; actually, the first important analytic study of that case is recent and due to the first author [J. Noncommut. Geom. 3, No. 1, 125--149 (2009; Zbl 1168.39002)].) As shown by the authors, this dichotomy is rather general and accordingly the paper is divided into two parts. The tools to study the Euler equations (and other irregular equations) have been known for a while: these are the summation processes along authorized directions at \(0\), the asymptotic properties of the sums thereby obtained and the Stokes phenomenon following from the diversity of the sums. Both parts of the paper refer to these tools, and one part uses the corresponding tools for \(q\)-difference equations. For simplicity, the authors assume \(q > 0\). The case \(q > 1\) (second part of the paper). In the last decade, the second author, Changgui Zhang (alone and with J.-P. Ramis) has developped a theory of asymptotics and summation for irregular \(q\)-difference equations. As always, when dealing when extensions of the world of ODEs to the \(q\)-world, the choice of \(q\)-analogues is far for unique. For instance, there are (at least) four possible processes of summation, according to the type of integration (discrete \textit{vs} continuous) and to the type of kernel (theta \textit{vs} \(q\)-exponential). By the way, this is related to the possibility of two \(q\)-Borel transforms: \[ \sum a_n x^n \mapsto \sum \frac{a_n}{[n!]_q} \xi^n \quad \text{or} \quad \sum a_n x^n \mapsto \sum \frac{a_n}{q^{n(n-1)/2}} \xi^n. \] In the second part of the paper, various sums for \(\hat{E}_{q}\) are compared and it is shown how they all degenerate to the classical summation of the Euler series when \(q \to 1\). Then, these results are extended to a rather general class of irregular \(q\)-deformations of irregular differential equations. The case \(0 < q < 1\) (first part of the paper). Here, the situation is rather more surprising. The authors show (as well for \(\hat{E}_{q}\) as for a more general class of Fuchsian \(q\)-deformations of irregular differential equations) how the canonical analytic solutions of the \(q\)-difference equations degenerate (in given sectors) to the summations of the irregular differential equation at stake. This is for instance applied to the degeneracy of \textit{plain} (Fuchsian) basic hypergeometric series into \textit{confluent} (irregular) hypergeometric series. This paper is a major breakthrough in the direction of understanding the \(q\)-deformation of the Stokes phenomemon.