On p. 263 of Borwein's paper entitled “Computational Strategies for the Riemann zeta function”, the following identity is stated: $$\sum_{n=k+2}^{\infty} \binom{n-1}{k} (\zeta(n) -1) =1 .$$It is valid for all nonnegative integer $k$.
The author points to other formulas that are relevant, including (43) on p. 262. However, I don't see how this identity follows from any previously established series evaluations, nor have I found a reference in the article with regards to this particular identity.
Questions
- How can the aforementioned identity be derived? Can you provide a proof?
- Are there any references to articles or books that describe and prove this identity?
- Can the identity be generalised? Are there variations of it involving binomial sums which are also valid for all nonnegative $k$?
$$ \sum_{n=k+2}^{\infty} \binom{n-1}{k} (\zeta(n) -1) = \sum_{n=k+2}\binom{n-1}{k}\sum_{m=2}^{\infty} \frac{1}{m^{n}} \\ = \sum_{m=2}^{\infty} \sum_{n=k+2}^{\infty} \binom{n-1}{k}\frac{1}{m^{n}} $$ From $$\binom{n-1}{k} = (-1)^{n-1-k}\binom{-1-k}{n-1-k}$$ we have $$ (sum) = \sum_{m=2}^{\infty}\sum_{n=k+2}^{\infty} (-1)^{n-1-k}\binom{-1-k}{n-1-k}\frac{1}{m^{n}} \\ =\sum_{m=2}^{\infty}\frac{1}{m^{k+1}}\sum_{n=1}^{\infty}\binom{-k-1}{n}\left(-\frac{1}{m}\right)^{n} \\ =\sum_{m=2}^{\infty}\frac{1}{m^{k+1}}\left[\left(1 - \frac{1}{m}\right)^{-k-1} - 1\right]\qquad(\text{by Newton's binomial theorem}) \\ =\sum_{m=2}^{\infty}\left(\frac{1}{(m-1)^{k+1}} - \frac{1}{m^{k+1}}\right) = 1. $$