Let $x>0$. Calculate:

$$\lim_{R\rightarrow\infty} \int_{\frac{1}{2}-iR}^{\frac{1}{2}+iR}\frac{x^s}{s}ds$$

for $x=1$ it is quite simple. Otherwise, I used the following contours:

enter image description here

The left one when $x>1$ and the right one when $0<x<1$. Let us discuss the case $x>1$.

The integral on $\Gamma_1$ is what we want. The integral on $\Gamma_3$ vanishes since:

$$\bigg| \int_{\Gamma_3}f(s)\bigg|\leq \int_{\Gamma_3}|f(s)| \leq 2R \bigg|\frac{x^s}{s}\bigg| \leq 2R \frac{e^{\ln x \Re s}}{\sqrt{2}R}=\sqrt{2}e^ {-R\ln x} \rightarrow 0$$

But I have no idea what to do with $\Gamma_2$ and $\Gamma_4$.

  • $\begingroup$ The Maple command $$int((exp(1))^{k*(1/2+I*y)}/(1/2+I*y), y = -infinity .. infinity, CauchyPrincipalValue) assuming\,k > 0 $$ produces $2\pi$ and $$int((exp(1))^{k*(1/2+I*y)}/(1/2+I*y), y = -infinity .. infinity, CauchyPrincipalValue) assuming\,k <0 $$ produces $-2\pi$. The command $$int(1/(1/2+I*y), y = -infinity .. infinity, CauchyPrincipalValue) $$ gives $\pi .$ $\endgroup$
    – user64494
    Jul 31 '13 at 19:08
  • 1
    $\begingroup$ @user64494: Thanks, but I'm not intersted in the final solution - just how to get to it. $\endgroup$
    – catch22
    Jul 31 '13 at 19:52

For the parts $\Gamma_2$ and $\Gamma_4$, you can estimate the integral by estimating the integrand

$$\left\lvert \frac{\exp \bigl((\sigma \pm iR)\log x\bigr)}{\sigma \pm iR} \right\rvert \leqslant \frac{1}{R} e^{\sigma\log x}$$

and then integrating

$$\left\lvert\int_{\Gamma_{2,\,4}} \frac{x^s}{s}\, ds\right\rvert \leqslant \frac{1}{R} \int_{-R}^{1/2} e^{\sigma\log x}\, d\sigma = \frac{1}{R} \frac{x^{1/2} - x^{-R}}{\log x} \leqslant \frac{C(x)}{R}.$$

For the $x < 1$ case, the estimate is similar.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.