Let $c>1$ be a real number, let $\alpha$ be a complex number with $0<\mathrm{Re}(\alpha)\leq 1$ and $T$ a large positive real number. Now consider the integral $$\frac{1}{2\pi i}\int_{c-iT}^{c+iT}\frac{\Gamma\left(w\right)}{\Gamma\left(w+\alpha\right)}dw.$$ Recalling the definition of Beta function $$B\left(a,b\right)=\int_{0}^{1}\left(1-t\right)^{a-1}t^{b-1}dt,\,\textrm{Re}\left(a\right)>0,\,\textrm{Re}\left(b\right)>0$$ from the inverse Mellin transform we get $$f\left(t\right)=\frac{\Gamma\left(a\right)}{2\pi i}\int_{c-i\infty}^{c+i\infty}t^{-b}\frac{\Gamma\left(b\right)}{\Gamma\left(a+b\right)}db,\,c>0,\,\textrm{Re}\left(a\right)>0$$ where $$f\left(t\right)=\begin{cases} \left(1-t\right)^{a-1} & 0<t<1\\ 0, & t\geq1. \end{cases}$$ Then $$\frac{1}{2\pi i}\int_{c-i\infty}^{c+i\infty}\frac{\Gamma\left(w\right)}{\Gamma\left(w+\alpha\right)}dw=0$$ so it means that $$\frac{1}{2\pi i}\int_{c-iT}^{c+iT}\frac{\Gamma\left(w\right)}{\Gamma\left(w+\alpha\right)}dw\rightarrow 0$$ as $T\rightarrow \infty.$

Question: Is it possible to find a function $f(\alpha,T)$ such that $$\frac{1}{2\pi i}\int_{c-iT}^{c+iT}\frac{\Gamma\left(w\right)}{\Gamma\left(w+\alpha\right)}dw=O(f(\alpha,T))$$ and $$\lim_{T\rightarrow\infty}f\left(\alpha,T\right)=0? $$

I tried to use the residue theorem but the limitation $0<\mathrm{Re}(\alpha)\leq 1$ is annoying. Thank you.

  • $\begingroup$ Plugging $t = 1$ in $f(t) = (1-t)^{a-1} 1_{t \in (0,1)}$ is allowed only if $\Re(a) > 1$. For $\Re(a) < 1$, $f(1) = \infty$ so you'll hardly find the bound you want. $\endgroup$
    – reuns
    Sep 16 '17 at 21:27
  • $\begingroup$ @reuns So the inversion is not valid if $0<\mathrm{Re}\left(\alpha\right)<1$? $\endgroup$ Sep 16 '17 at 21:29
  • $\begingroup$ It is valid in the sense of distributions or $L^2$. And it is valid pointwise away from $t=1$. $\endgroup$
    – reuns
    Sep 16 '17 at 21:31

About the Fourier/Laplace/Mellin inversion theorem.

Let $\phi_a(x) = \int_{-a}^a e^{ikx}dk = \frac{e^{iax}-e^{-iax}}{ix} = \frac{2 \sin(ax)}{x} = a \, \Phi'(ax)$ where $\Phi(x) = \int_{-\infty}^x \frac{2\sin(y)}{y} dy, \Phi(-\infty) = 0, \Phi(+\infty) = 2\pi$. If $F,F' \in L^1$ then integrating by parts $$\int_{-\infty}^\infty \hat{F}(k) dk = \lim_{a \to \infty} \int_{-a}^a \int_{-\infty}^\infty F(x)e^{-ikx}dx dk = \lim_{a \to\infty}\int_{-\infty}^\infty \phi_a(x) F(x)dx \\= \lim_{a \to\infty}-\int_{-\infty}^\infty \Phi(ax) F'(x)dx = -\int_{-\infty}^\infty 2 \pi 1_{x > 0} F'(x)dx =2\pi F(0)$$

Here $F(x) = f(e^x)e^{\Re(b) x}, f(t) = (1-t)^{a-1}$ so $F' \not \in L^1$ for $\Re(a) \in (0,1)$ but you can look at $F \ast n e^{-\pi n^2 x^2}$, obtaining that

$$\lim_{a \to \infty}\int_{-a}^a \hat{F}(k) dk = \lim_{n \to \infty}\int_{-\infty}^\infty \hat{F}(k) e^{-\pi k^2/n^2} dk= \lim_{n \to \infty} F \ast n e^{-\pi n^2 x^2}|_{x=0} = \infty$$


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.