Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am often working with divergent series all around being this the bread and butter for a theoretical physicist. Thanks to the excellent work of Hardy these have lost their mystical Aurea and so, they have found some applications in concrete computations. In these days I have been involved with the following divergent series


that is clearly divergent. Wolfram Alpha provides the following


being $\Phi(z,s,a)$ the Lerch function (see also Wikipedia). Is there any summation technique in this case?

share|cite|improve this question
$S=-\mathrm i\pi$, obviously... – Did Feb 26 '12 at 14:29
@DidierPiau: Great! Please, put this as an answer and I will accept it. Thanks! – Jon Feb 26 '12 at 14:33
Since logarithm has no natural analytic continuation as a meromorphic function on $\mathbb{C}$, we cannot say $-i\pi$ as a natural choice. Rather, if we understand the summation as the Cauchy principal value of the integral $$\int_{0}^{2} \frac{dx}{1-x},$$ it should be zero. – Sangchul Lee Feb 26 '12 at 14:51

Well $\displaystyle S(x)=\sum_{n=1}^\infty \frac{x^n}n$ so that

$S'(x)=\sum_{n=1}^\infty x^{n-1}=\frac1{1-x}$

and $S(x)=-\log(1-x)+C$

But $S(\frac 12)=\log(2)$ so that $C=0$

and $S(2)$ 'could be' $-\log(1-2)= -\log(e^{\pi i})= -\pi i$ or $-\log(e^{-\pi i})= \pi i\cdots$.

Both choices seem ok by analytic expansion (or none of them since it is at the border! :-)).

Here is a picture of $\mathrm{Im}(-\log\left(1-(x+iy))\right)$ with two continuous paths possible from $z=\frac 12$ to $2$.

share|cite|improve this answer
Raymond, I think Didier is right as $\sum_{n=1}^\infty\frac{2^n}{n}=\Phi(2,1,0)=-i\pi$ as you can check with Mathematica where $\Phi(z,s,a)$ is LerchPhi[z,s,a]. – Jon Feb 26 '12 at 15:17
@Jon: Well Mathematica had to make a choice and $\log(-1)= i \pi\ $ by 'definition' so that $-i \pi\ $ won. This choice seems indeed acceptable. As a physicist you could too consider the symmetry of the situation and accept the mean result $0\ $ (the principal value point of view...) ! – Raymond Manzoni Feb 26 '12 at 15:25
Indeed, I know that this cannot be zero. The reason is that a colleague of mine asked to me the power of a quaternion $\hat i^{\hat i}$, as the case $i^i$ is indeed simple, and, in the effort to give a meaning to this, I met the above sum. – Jon Feb 26 '12 at 15:38
@Jon: Well $i^i= e^{i\log(i)}= e^{i(i\pi/2+2 k\pi i)}= e^{-\pi/2 -2 k\pi}\ \ \ $ so that it admits an infinity of real answers (depending on the branch chosen)! (for the quaternionic 'i' or the classical one). That could be the next move of your colleague... – Raymond Manzoni Feb 26 '12 at 15:45
@Jon: my derivation would be (if that's the question) : $\displaystyle \log\left(\hat{\imath}^{\hat{\imath}}\right)= \hat{\imath}\log(\hat{\imath})= \hat{\imath}\log\left(e^{\hat{\imath}\frac{\pi}2+2k\pi\hat{\imath}}\right)= \hat{\imath}\left(\hat{\imath}\frac{\pi}2+2k\pi\hat{\imath}\right)=-\frac{\pi}2-‌​2k\pi\ $ so that $\displaystyle\hat{\imath}^{\hat{\imath}}=e^{-\frac{\pi}2-2k\pi}\ $ It we consider as usual the principal branch (that $\hat{\imath}=e^{\hat{\imath}\frac{\pi}2}\ $ i.e. $k=0\ $) then the answer is simply $e^{-\frac{\pi}2}\ $ – Raymond Manzoni Feb 26 '12 at 19:41

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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