Tell me more ×
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.
up vote 2 down vote favorite

Let $p \in \mathbb{R}$. Series of the form $$S_p = \sum_{k \in \mathbb{N^*}} \frac{1}{k^p}$$ converge if and only if $p > 1$. Let us define $$S_{p, n} \triangleq \sum_{k = 1}^{n} \frac{1}{k^p},$$ the truncation of the $p$-series at its $n$-th term. (It is evident that $S_{p, n} \space \xrightarrow{n\to\infty} \space S_p$.) Is there a closed expression for $S_{p, n}$, at least for some values of $p$ (integers must be easier)?

share|improve this question
\stackrel{n \to \infty}{\longrightarrow} should do the trick. – Josh May 14 '11 at 0:41
$\overset{n\rightarrow \infty }{\longrightarrow }$ can be obtained with \overset{n\rightarrow \infty }{\longrightarrow } – Américo Tavares May 14 '11 at 0:54
I deleted an answer because I replied to a different question (closed form for $S_p$, with $p$ even or negative) – Américo Tavares May 14 '11 at 0:58
It depends on what you mean by "closed form expression." It is surprisingly difficult to give a precise definition of this. – Qiaochu Yuan May 14 '11 at 3:32
1  
Here's a plausibility argument on why you're supposed to expect the polygamma functions (as shown in Robert's and Eric's answers) to pop up: recall the usual gamma function relation $\Gamma(n+1)=n\Gamma(n)$ and then derive from it $\log\Gamma(n+1)=\log\,n+\log\Gamma(n)$. As noted, the polygamma functions are derivatives of $\log\Gamma$, and through the identity $$\frac{\mathrm d^n}{\mathrm dz^n}\log\,z=\frac{(-1)^{n-1}(n-1)!}{z^n}$$, you'll find that a connection formula for $\psi^{(n)}(z+k)$ and $\psi^{(n)}(z)$ would necessarily involve a series of reciprocal powers. – J. M. May 14 '11 at 4:34
show 1 more comment

3 Answers

up vote 3 down vote accepted

In Maple notation, for $p \ge 2$ integer, $$S_{p,n} = \zeta(p) + \frac{(-1)^{p-1}}{(p-1)!} \Psi(p-1,n+1),$$ where $\psi(m,\space·)$ is the $m$-th polygamma function.

share|improve this answer
1  
For completeness: if $p=1$, $$\sum_{k=1}^n k^{-1}=\gamma+\psi^{(0)}(n+1)$$ where $\gamma$ is the Euler-Mascheroni constant. (The zeta function diverges for $p=1$.) For noninteger $p$, there has been research into analytically continuing the polygamma function such that Robert's expression still makes sense. See this for instance. – J. M. May 14 '11 at 3:45
I could be wrong, but I doubt this is what Luke had in mind when he asked for a closed form. – Gerry Myerson May 14 '11 at 11:31
up vote 3 down vote

Yes, there is. These are called the generalized harmonic numbers.

With polygamma function: I suggest looking at this paper, they have many interesting things. In particular, if we define $$ H_n(z;r)=\sum_{k=1}^n\frac{1}{(z+k-1)^r} $$ (even more generalized) then we have for integers $p$ $$ H_n(z;p)=\frac{(-1)^{p-1}}{(p-1)!}(\psi^{(p-1)}(z+n)-\psi^{(p-1)}(z)) $$ where $$ \psi^{(k-1)}(z)=\frac{d^k}{dx^k}\left(\log\Gamma(x)\right) \biggr|_{x=z}. $$ This is equation (1.14) of the paper, but I think they forgot a factor of $(p-1)!$ on the denominator. Taking the case $z=1$ gives what you are looking for.

With zeta function: we can write $$H_n(r)=\sum_{k=1}^n\frac{1}{k^r}=\zeta(r)-\zeta(r,n+1),$$ where $\zeta(s,a)$ is the Hurwitz zeta function. This $\space$ really clear, but by using identities regarding $\zeta(s,a)$ we can change the identity into $\space$ previous one, and other things.

Hope that helps.

share|improve this answer
2  
In short, the (generalized) harmonic numbers and the polygamma function are essentially the same thing (differing only by a constant); and are tightly related to Hurwitz zeta. – J. M. May 14 '11 at 2:59
@J.M. That is a nice way to put it! – Eric May 14 '11 at 3:06
A tiny tip, BTW: for linking to SD, it usually makes better sense to use the DOI instead of the actual SD link; for some reason they tend to go stale quickly. In your case, http://dx.doi.org/10.1016/S0096-3003(01)00172-2 would be a better link to the paper in your answer. – J. M. May 14 '11 at 3:36
I could be wrong, but I doubt this is what Luke had in mind when he asked for a closed form. – Gerry Myerson May 14 '11 at 11:31
On the other hand, Luke did not seem to complain about Eric's and Robert's answers... – J. M. May 14 '11 at 11:38
show 1 more comment
up vote 0 down vote

To the best of my knowledge, there is no closed form expression for $S_{p,n}$.

EDIT: Let me amend my answer to read, to the best of my knowledge, there is no closed form expression in terms of functions familiar to, say, students of 1st-year Calculus.

share|improve this answer

Your Answer

 
discard

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.