Take the 2-minute tour ×
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.

Is it possible to prove that for the regularized Harmonic series

$$ \tag 1\sum_{n=1}^{\infty} \frac{e^{-n\epsilon}}{n}=\gamma + 1/\epsilon $$

if epsilon is very small $ \epsilon \to 0 $ i can use $ e^{-n\epsilon}=1-n\epsilon $ , to the Harmonic series appear

however if i use the Polylogarithm i get

$$ \sum_{n=1}^{\infty} \frac{e^{-n\epsilon}}{n}=log(1-e^{-\epsilon})$$

so no constant appear, how can I prove $(1)$?

share|improve this question

2 Answers 2

up vote 1 down vote accepted

You cannot prove (1) because it is indeed not true. In fact $$\sum_{n=1}^\infty \frac{e^{-n\varepsilon}}{n}\sim -\log(\varepsilon)$$ as $\varepsilon\to 0^+$, so certainly this cannot be equivalent to $1/\varepsilon$, as would follow from (1).

The above equivalence is proved as follows: as you mentioned, $\sum_{1}^\infty \frac{e^{-n\varepsilon}}{n}=-\log(1-e^{-\varepsilon})$; but $1-e^{-\varepsilon}=\varepsilon+O(\varepsilon^2)$ because $e^{-\varepsilon}=1-\varepsilon +O(\varepsilon^2)$, so $$\log(1-e^{-\varepsilon} )=\log(\varepsilon+O(\varepsilon^2))=\log(\varepsilon)+\log(1+O(\varepsilon))\sim\log(\varepsilon)\, ,$$ which gives the result.

share|improve this answer

In general, you can consider the power series $$f(x) = \sum_{n=1}^\infty \frac{x^n}{n}$$ which is defined (and $\mathcal{C}^\infty$) on the open ball $(-1,1)$. Using this, and standard theorems about power series, you have $$f^\prime(x) = \sum_{n=1}^\infty \frac{d}{dx}\frac{x^n}{n} = \sum_{n=1}^\infty x^{n-1} = \sum_{n=0}^\infty x^{n} = \frac{1}{1-x}$$ and since $f(0)= 0$ (using the very first expression as an infinite series), you get $$f(x) = \int_{0}^x \frac{dt}{1-t} = -\ln(1-x)$$ Now, plugging $x=e^{-\epsilon}\in(0,1)$ (for $\epsilon > 0$), you get $$\sum_{n=1}^\infty \frac{e^{-n\epsilon}}{n} = f(e^{-\epsilon}) = -\ln(1-e^{-\epsilon}) \operatorname*{=}_{\epsilon\to 0^+} \log \frac{1}{\epsilon}+\frac{\epsilon}{2}+o(\epsilon)$$

Therefore, if I have understood correctly your question, the answer is no: you cannot prove (1) — as it is not true.

share|improve this answer
$f(e^{-\varepsilon})$ is not $-{\rm ln}(1+\varepsilon)$. –  Etienne Jul 29 '13 at 19:46
no, you are wrong is $ log(1-e^{-\epsilon}) 4 not $ log(1+\epsilon) $ –  Jose Garcia Jul 29 '13 at 19:46
Oops, my mistake. I'll edit accordingly (thanks! Indeed, a quantity going to $\infty$ makes much more sense). –  Clement C. Jul 29 '13 at 19:47

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.