0
$\begingroup$

This is the integral of the reciprocal gamma function,$$F=\int_0^\infty dx/\Gamma(x)= e+\int_0^\infty \frac{e^{-t}\ dt}{\log^2 t+\pi^2}$$ $$=e+\int_{-\infty}^{\infty}\frac{\exp(s-e^s)\ ds}{s^2+\pi^2}$$ $$=e+\frac1{2\pi i}\int_{-\infty}^{+\infty}[\frac1{s-i\pi}-\frac1{s+i\pi}]\exp(s-e^s)\ ds$$Now, defining $f(s):=\exp(s-e^s)$ and observing that $f(s+2\pi i)=f(s)$ and $f(i\pi)=-e$, consider the rectangular closed contour $\mathcal{C}=[-A,A]\cup[A,A+2\pi i]\cup[A+2\pi i,-A+2\pi i]\cup[-A+2\pi i,-A]$. The contour integral$$\frac1{2\pi i}\oint_{\mathcal C}\frac{f(s)\ ds}{s-i\pi}=f(i\pi)=-e$$and also$$=\frac1{2\pi i}\int_{-A}^A [\frac1{s-i\pi}-\frac1{s+i\pi}]f(s)\ ds+\frac1{2\pi i}(\int_{-A+2\pi i}^{-A}+\int^{A+2\pi i}_{A})\frac{f(s)\ ds}{s-i\pi}$$ Taking the $\lim_{A\to\infty}$, and as the "left branch" $\int_{-A+2\pi i}^{-A}\frac{f(s)\ ds}{s-i\pi} \to 0$, we get $$F=-\lim_{A\to\infty}\frac{e^A}{2\pi}\int_0^{2\pi}\frac{\exp(-e^A \cos\theta-ie^A \sin\theta)e^{i\theta}\ d\theta}{A+i(\theta-\pi)}$$

Question : does there exist a method, standard (Laplace, stationary phase, saddle point...) or otherwise, to make this limit explicit ?

$\endgroup$
2
  • $\begingroup$ You have also $$ F=e+{1\over{\pi}}Re(\int_0^{\infty} {{e^{it}}\over{\log t +i\pi/2}}dt) $$ $\endgroup$
    – user75485
    Aug 4, 2015 at 15:38
  • $\begingroup$ Yes, but my (maybe naive) hope, through the use of asymptotic analysis, would be that the limit depends on the behavior of the integrand at "decisive points": stationary phase ($\theta=\pi/2$ and $3\pi/2$), maximum amplitude ($\theta=-\pi$), or some optimal trade-off between amplitude and frequency of oscillations. $\endgroup$ Aug 5, 2015 at 10:21

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.