\(G\)-functions and multisum versus holonomic sequences (Q1011412): Difference between revisions

A multisum sequence \((a_n)\) is a sequence of complex numbers of the form \[ a_n = \sum_{k \in \mathbb Z^r\mid A_j(n,k) \geq 0,\;j=1, \dots J} C_0^n \prod_{i=1}^r C_i^{k_i} \prod_{j=1}^J A_j(n,k)!^{\varepsilon_j} \] where \(C_i \in \overline{\mathbb Z}\), \(\varepsilon_j=\pm 1\), and \(A_j\) are integral linear forms in \((n,k)\) such that for every \(n\in \mathbb N\) the set \(\{k\in \mathbb Z^r\mid A_j(n,k) \geq 0\), \(j=1,\dots J\}\) is finite. The multisum is balanced if \[ \sum_{j=1}^J \varepsilon_j A_j=0. \] A formal power series \(\sum_{n=0}^\infty b_n T^n \in \mathbb C[[T]]\) is Gevrey type-1 if \(\sum_{n=0}^\infty \frac{b_n}{n!} T^n\) has a nonzero radius of convergence, i.e., there is a positive constant \(C\) such that \(|b_n|\leq C^n n!\) (see [\textit{Y. André}, ``Arithmetic Gevrey series and transcendence. A survey.'' J. Théor. Nombres Bordx. 15, No.~1, 1--10 (2003; Zbl 1136.11315)]). The author proves that every sequence which is a multidimensional sum of a balanced hypergeometric term has an asymptotic expansion of Gevrey type-1 with rational exponents. A series \(\sum_{n=0}^\infty b_n T^n \in \mathbb C[[T]]\) is a \(G\)-function if it is holonomic (the solution of a linear differential equation with polynomial coefficients), and the \(b_n\) are algebraic numbers such that the absolute value of each conjugate of \(b_n\) is exponentially bounded and so is the common denominator of \(b_0, \dots, b_n\). The proof is achieved by first showing that the generating series of a balanced multisum sequence is a \(G\)-function. A consequence is that the generating series of an integer-valued, exponentially bounded holonomic sequence is a \(G\)-function. While \textit{H. Wilf} and \textit{D. Zeilberger} [``An algorithmic proof theory for hypergeometric (ordinary and ``\(q\)'') multisum/integral identities'', Invent. Math. 103, No.~3, 575--634 (1992; Zbl 0739.05007)] proved that every multisum sequence is holonomic, the author shows a holonomic sequence that is not a balanced multisum, namely \[ (2 n + 1)a_{n + 2} -(7 n + 11)a_{ n + 1} + (2 n + 1)a_{n} = 0, \] with initial conditions \(a_0 = 0, a_1 = 1\).