15
$\begingroup$

I'm trying to find the value of the following sum

$$\sum_{n=1}^{\infty} \frac1{n}\left(\frac{np}{p+n}\right)^{n+1}$$

where $0. Any ideas? Thanks.

  • 2
    My guess is that there is no nice closed form solution.2011-06-02
  • 2
    Certainly the $n^{n+1}$ factor will likely preclude an expression in terms of currently known functions...2011-06-02
  • 3
    Guessing you can't guess is a funny guess... the only thing I can say is that I am sure it is absolutely convergent (which I believe you noticed since you worked on it... but I was suspicious at first so I'll just make the worried people stop wondering ) :$$ \sum_{n=1}^{\infty} \frac 1n \left( \frac {np}{p+n} \right)^{n+1} = \sum_{n=1}^{\infty} \frac 1n \left( \frac {p}{\frac pn + 1} \right)^{n+1} \le \sum_{n=1}^{\infty} \frac 1n p^{n+1} \le \sum_{n=1}^{\infty} p^{n+1} = \frac 1{1-p} $$ because $p/n+1 \ge 1$, so $1 \ge \frac 1{p/n+1}$. Now I can sleep. Good luck with it though.2011-06-02
  • 0
    If you define this as function $f(p)$ then the terms of the powerseries begin with $ p^2 - 3/2*p^3 + 31/12*p^4 - 31/9*p^5 + 10007/2160*p^6 - ...$ Let's write *x* for *p* now... Roughly this goes to $f(x) = x*\sum_{k=1}^{\infty} (-1)^k (k-\delta_k) x^k $ where the sequence $delta_k$ looks similar to the harmonic series and doesn't sum up nicely to some estimate of bound. (just another hint, that it should be convergent for $|x|< 1$ )2011-06-02
  • 1
    Reordering summation/collecting like powers of *p* I get the following expression $$f(p) = p \sum_{j=1}^{\infty} (-1)^j c_j p^j $$ where $$c_j = \sum_{k=1}^j (-1)^k \binom {j}{k} k^{k-j-1} $$. The $c_j$ approximate $-j$ so the $\delta$ in my previous comment is now $\delta_j = c_j +j$ (I hope, all indexes and signs are correctly taken from my sketchpad...)2011-06-02
  • 0
    Thank you all for your insights. Something like this is what I expected. I'm not sure if a closed form can be found, so brainstorm is just the way to go now.2011-06-03
  • 1
    Not to be pedantic, but this is not a "sum of a series", but just "a series" (=' sum of the terms of a sequence').2011-07-10
  • 0
    You should somehow use that $$\Bigl({1\over 1+{p\over n}}\Bigr)^n\doteq e^{-p}\ .$$2011-07-10
  • 0
    OP, can you tell us _why_ you are trying to sum the series and _where_ you found the series?2011-07-24
  • 1
    @Srivatsan: It's a far-fetched expression that shows up in a very particular problem that I was working on. The problem would not be very relevant for the series. In the end I was able to follow a completely different path and this series was no longer necessary. Since then I have done no further work on it.2011-07-25
  • 0
    @Patrick, your argument shows that the sum is $\le p^2/(1-p)$. By a similar one, it is also $\ge p^2/(1+p)$.2011-09-24

2 Answers 2