Here I have a limit to which I arrived while working on a seperate integral through Mellin Transforms. $$\lim\limits_{s\to -1^{-}}\Big[\psi_{(0)}(s)-\frac{\pi}{2}\tan\left(\frac{\pi s}{2}\right)\Big]$$ Here, we have $\psi_{(0)}(s)$ which represents the digamma function. I graphed the whole thing on Desmos to see what it looked like approaching $-1$, and it seems very likely that the limit approaches $$1-\gamma$$ Here, $\gamma$ is the Euler-Mascheroni constant. I would like to know if there is a concrete way of evaluating this limit. I tried doing some work with it: $$=\lim\limits_{x\to 0}\Big[\psi_{(0)}(x-1)-\frac{\pi}{2}\tan\left(\frac{\pi}{2}(x-1)\right)\Big]$$ $$=\lim\limits_{x\to 0}\Big[\frac{1}{1-x}+\psi_{(0)}(x)-\frac{\pi}{2}\tan\left(\frac{\pi}{2}(x-1)\right)\Big]$$ I don't really know where to go from here. I figure maybe a Taylor expansion could do the trick. However, the expansions for both the digamma and tangent functions are largely unrelated, it seems. I'm curious to see a solution to this problem, and wish you all good luck!

A natural extension of this question would be to find: $$\lim\limits_{s\to (-1-2k)^{-}}\Big[\psi_{(0)}(s)-\frac{\pi}{2}\tan\left(\frac{\pi s}{2}\right)\Big]\,\,\forall\,\,k\in Z^{+}$$ This generalized limit might be the bane of my existance.


You're correct about the limiting value and correct that a series expansion is the way to go.

Starting with the series expansion at $s=0$, which is $$ \Gamma(s) = \frac1s + \gamma + O(|s|), $$ and repeatedly using the functional equation $\Gamma(s+1)=s\Gamma(s)$, one can prove by induction that the series expansion at $s=-k$ ($k$ a nonnegative integer) is $$ \Gamma(s) = \frac{(-1)^k}{k!} \bigg( \frac1{s+k} + H_k - \gamma + O(|s+k|) \bigg), $$ where $H_k = \sum_{j=1}^k \frac1j$ is the $k$th harmonic number. Therefore $$ \Gamma'(s) = \frac{(-1)^k}{k!} \bigg( {-}\frac1{(s+k)^2} + O(1) \bigg), $$ from which we calculate (by long division) that $$ \psi_{(0)}(s) = \frac{\Gamma'(s)}{\Gamma(s)} = -\frac1{s+k} + H_n - \gamma + O(|s+k|). $$ Since $\tan$ is an odd function (and is itself a logarithmic derivative of $\sin$), its series expansion at odd negative integers $-k$ is going to be simply $-1/(s+k) + O(|s+k|)$. Therefore the series expansion of their difference is $$ \psi_{(0)}(s) - \frac\pi2 \tan\bigg( \frac{\pi s}2 \bigg) = H_n - \gamma + O(|s+k|), $$ which gives you the desired limits.

  • $\begingroup$ Thanks for the detailed explanation Mr. Martin, I could follow along perfectly. I haven't worked a lot with the digamma function before so I was at a loss for an expansion with negative arguments. This is awesome...again, thanks for sharing! $\endgroup$ – Brian Constantinescu Jul 20 at 8:30

Let $t=s+1\to 0^-$, then $$\psi_{(0)}(s)-\frac{\pi}{2}\tan\left(\frac{\pi s}{2}\right)=\psi_{(0)}(t-1)+\frac{\pi/2}{\tan\left(\frac{\pi t}{2}\right)}=\psi_{(0)}(t-1)+\frac{1}{t}+o(1).$$ Now recall that $\psi_0(1+x)-\psi_0(x)=\frac{1}{x}$ which implies $$\psi_{(0)}(t-1)=\psi_{(0)}(t)-\frac{1}{t-1}=\psi_{(0)}(t+1)-\frac{1}{t}-\frac{1}{t-1}.$$ Therefore, as $s\to 1^-$, we have that $t\to 0^-$, and
$$\psi_{(0)}(s)-\frac{\pi}{2}\tan\left(\frac{\pi s}{2}\right)=\psi_{(0)}(t+1)-\frac{1}{t-1}+o(1)\to \psi_{(0)}(1)+1=1-\gamma.$$

  • $\begingroup$ Thanks Robert! Funny enough, right as you made this post I edited my question to show I had thought of the same manipulation that you provided in your hint. Alas, I didn't know where to go from there (and it appears I arrived to something wrong too). I appreciate the quick response! $\endgroup$ – Brian Constantinescu Jul 20 at 8:23
  • $\begingroup$ @BrianConstantinescu I edited my answer. $\endgroup$ – Robert Z Jul 20 at 9:01
  • $\begingroup$ Cheers, it's an elegant and concise solution! Thanks for taking the time to help me out. $\endgroup$ – Brian Constantinescu Jul 20 at 9:49

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.