A product approximation of the gamma function (Q2356072)

A number of approximation formulas for the gamma function are proposed. The obtained approximations generalize the classical results by Stirling, Ramanujan and more recent results by \textit{J. Sándor} and \textit{L. Debnath} [J. Math. Anal. Appl. 249, No. 2, 569--582 (2000; Zbl 0965.26007)]. In particular the following class of approximations for any real \(a\) and \(b\) is introduced: \[ \Gamma(n+1)\sim \mu_n(a,b):= \sqrt{2 \pi n} \left(\frac{n}{e}\right)^n \left(1 + \frac{1}{n + a}\right)^{-b} \eqno{(1)} \] and it was shown that \(\mu_n(-1, 1/2)\) is the best approximation among all approximation given by (1). Generalizing this result the following class of approximations is considered too: \[ \Gamma(n+1)\sim \tau_n(a,b):= \sqrt{2 \pi n} \left(\frac{n}{e}\right)^n \prod\limits_{i=1}^{k} \left(1 + \frac{1}{(n - 1/2)^{i}}\right)^{\lambda_i}, \eqno{(2)} \] where \(\lambda_1, \lambda_2, \ldots, \lambda_k\) are arbitrary real numbers. The values of \(\lambda_1, \lambda_2, \ldots, \lambda_k\) provided the most accurate approximations are found for \(k=1, 2, 3, 4, 5\). The theory of completely monotonic functions is used to derive the necessary estimates.