Closed form for $\sum_{k=1}^{n}\binom{n}{k}\frac{\left(-1\right)^k }{k}$ and $\sum_{k=1}^{n}\binom{n}{k}\frac{1 }{k+m}$

163 Views Asked by At

Consider the following binomial summations:

$$\sum_{k=0}^{n}\binom{n}{k}\frac{\left(-1\right)^k }{k+m}\tag{I}$$

Where $m \in \mathbb N_{\ge1}$

$$\sum_{k=1}^{n}\binom{n}{k}\frac{1 }{k+m}\tag{II}$$

Where $m \in \mathbb N_{\ge1}$


$\text{(I)}$

I use the following identity:

$$\binom{n}{k}\frac{1}{k+m}=\frac{1}{m}\binom{n+m}{m}^{-1}\binom{n}{k}\binom{k+m-1}{k}$$

So the sum changes to:

$$\sum_{k=0}^{n}\binom{n}{k}\frac{\left(-1\right)^k }{k+m}=\frac{1}{m}\binom{n+m}{m}^{-1}\sum_{k=0}^{n}\binom{n}{k}\color{red}{\binom{k+m-1}{k}\left(-1\right)^k}$$ $$=\frac{1}{m}\binom{n+m}{m}^{-1}\sum_{k=0}^{n}\binom{n}{n-k}\color{red}{\binom{-m}{k}}$$$$=\frac{1}{m}\binom{n+m}{m}^{-1}\binom{n-m}{n}$$

But is there any closed form when $m=0$?, in other words does there exist any closed form for the following sum:

$$\sum_{k=1}^{n}\binom{n}{k}\frac{\left(-1\right)^k }{k}$$


$\text{(II)}$

$$\sum_{k=1}^{n}\binom{n}{k}\frac{1 }{k+m}=\frac{1}{m}\binom{n+m}{m}^{-1}\sum_{k=1}^{n}\binom{n}{k}\color{red}{\binom{k+m-1}{k}} $$ $$=\frac{1}{m}\binom{n+m}{m}^{-1}\sum_{k=1}^{n}\binom{n}{n-k}\color{red}{\binom{-m}{k}\left(-1\right)^k}$$

The final answer depends on the closed form of alternating sign Vandermonde's convolution, which apparently there does not exist such closed form.

My question is that : what is the closed form of $\text{(II)}$?

2

There are 2 best solutions below

0
On BEST ANSWER

For $\text{(II)}$, there is a closed form $$S_n=\sum_{k=1}^{n}\binom{n}{k}\frac{1 }{k+m}=\frac 1m \left(2^n-1+(-1)^{-m} n B_{-1}(m+1,n) \right)\tag 1$$ where appears the incomplete beta function.

For $\text{(I)}$ $$T_n=\sum_{k=1}^{n}\binom{n}{k}\frac{(-1)^k }{k+m}=\frac{\Gamma (m)\, \Gamma (n+1)}{\Gamma (m+n+1)}-\frac{1}{m}\tag 2$$ which, for $m=0$ reduces to $-H_n$.

To make a series expansion of $(2)$ around $m=0$, recall first $$\Gamma(m)=\frac{1}{m}-\gamma +\frac{1}{12} \left(6 \gamma ^2+\pi ^2\right) m+O\left(m^2\right)$$ and $$\Gamma(m+n+1)=\Gamma (n+1)+m \Gamma (n+1) \psi (n+1)+O\left(m^2\right)$$ So $$T_n=-\psi (n+1)-\gamma +O\left(m\right)=-H_n+O\left(m\right)$$

0
On

From $(1+x)^n=\sum_{k=0}^n \binom{n}{k}x^k$, we have $$x^{m-1}(1+x)^n=\sum_{k=0}^n \binom{n}{k}x^{k+m-1}$$ integrating from $0$ to $1$ yields $$\int_0^1 x^{m-1}(1+x)^ndx=\int_{0}^1 x^{m-1} dx+\sum_{k=1}^n \binom{n}{k}\frac{1}{m+k}$$

But in the case $m=0$, the integral $\int_0^1 x^{-1}dx$ diverges.