7
$\begingroup$

I am wondering if any knows how to compute a closed form for the following two series.

  1. $$\sum_{m=1}^{n}\frac{(-1)^m}{m^2}\binom{2n}{n+m}$$
  2. $$\sum_{m=1}^{n}\frac{(-1)^m}{m^4}\binom{2n}{n+m}$$

Mathematica gave a result in terms of the Hypergeometric function, which I don't understand. I was wondering if anyone can illustrate how to express these two series in a nice closed form, without using the Hypergeomeyric function. I was thinking along the line using a combinatorial identity or method, but I having problems figuring it out.

$\endgroup$
  • 2
    $\begingroup$ Result in terms of the hypergeometric function is given when there are no nice closed forms , i.e., when there are no closed forms in terms of standard mathematical functions. $\endgroup$ – Prasun Biswas Jul 4 '15 at 3:19
  • 3
    $\begingroup$ Somewhat nicer expression for (1) is $$2^{n-1}\left(\Psi^{(1)}(n+1)-\frac{\pi^2}{6}\right)\frac{(2n-1)!!}{n!}$$ $\endgroup$ – d.k.o. Jul 4 '15 at 3:51
  • $\begingroup$ Thanks for the reply. This is an interesting formula. How did you go about deriving it? Do you have any insight on how to compute a nicer expression for the second identity? $\endgroup$ – Hatchet Jul 4 '15 at 6:29
8
$\begingroup$

Let $$ g(n,m)=\frac1{\binom{2n}{n}}\sum_{j=1}^n\frac{(-1)^j}{j^m}\binom{2n}{n+j}\tag{1} $$ then it can be shown that $$ g(n,m)=g(n-1,m)+\frac{g(n,m-2)}{n^2}\tag{2} $$ The proof of $(2)$ is below.

Since $$ \sum_{j=0}^k(-1)^j\binom{n}{j}=(-1)^k\binom{n-1}{k}\tag{3} $$ we get $$ g(n,0)=-\frac12\tag{4} $$ then, using $(2)$ and $(4)$, we get $$ g(n,2)=-\frac12H_n^{(2)}\tag{5} $$ Therefore, $$ \bbox[5px,border:2px solid #C0A000]{\sum_{j=1}^n\frac{(-1)^j}{j^2}\binom{2n}{n+j}=-\frac12H_n^{(2)}\binom{2n}{n}}\tag{6} $$ where $$ H_n^{(k)}=\sum_{j=1}^n\frac1{j^k}\tag{7} $$


Using $(2)$ and $(5)$, we get that $$ g(n,4)=-\frac14\left(\left(H_n^{(2)}\right)^2+H_n^{(4)}\right)\tag{8} $$ therefore, $$ \bbox[5px,border:2px solid #C0A000]{\sum_{j=1}^n\frac{(-1)^j}{j^4}\binom{2n}{n+j}=-\frac14\left(\left(H_n^{(2)}\right)^2+H_n^{(4)}\right)\binom{2n}{n}}\tag{9} $$


Proof of $\boldsymbol{(2)}$:

Note that $$ \begin{align} g(n,m) &=\sum_{j=1}^n\frac{(-1)^j}{j^m}\frac{n!}{(n-j)!}\frac{n!}{(n+j)!}\\ &=\frac{(-1)^n}{n^m\binom{2n}{n}}+\sum_{j=1}^{n-1}\frac{(-1)^j}{j^m}\frac{(n-1)!}{(n-j-1)!}\frac{(n-1)!}{(n+j-1)!}\frac{n^2}{n^2-j^2}\\ g(n-1,m) &=\hphantom{\frac{(-1)^n}{n^m\binom{2n}{n}}+}\sum_{j=1}^{n-1}\frac{(-1)^j}{j^m}\frac{(n-1)!}{(n-j-1)!}\frac{(n-1)!}{(n+j-1)!}\\ g(n,m)-g(n-1,m) &=\frac{(-1)^n}{n^m\binom{2n}{n}}+\sum_{j=1}^{n-1}\frac{(-1)^j}{j^m}\frac{(n-1)!}{(n-j-1)!}\frac{(n-1)!}{(n+j-1)!}\frac{j^2}{n^2-j^2}\\ &=\frac{(-1)^n}{n^m\binom{2n}{n}}+\frac1{n^2}\sum_{j=1}^{n-1}\frac{(-1)^j}{j^{m-2}}\frac{n!}{(n-j)!}\frac{n!}{(n+j)!}\\ &=\frac1{n^2}\sum_{j=1}^n\frac{(-1)^j}{j^{m-2}}\frac{n!}{(n-j)!}\frac{n!}{(n+j)!}\\ &=\frac{g(n,m-2)}{n^2} \end{align} $$

$\endgroup$
  • 1
    $\begingroup$ Brilliant! Using a recursive relation was the ticket. Thanks for the help. $\endgroup$ – Hatchet Jul 5 '15 at 8:05
5
$\begingroup$

The first expression:

$\begin{align}\sum\limits_{m=1}^{n}\frac{(-1)^m}{m^2}\binom{2n}{n+m} &= \binom{2n}{n}\sum\limits_{m=1}^{n}\frac{(-1)^m}{m^2}\binom{n}{m}\frac{m!n!}{(m+n)!}\\ &= \binom{2n}{n}\sum\limits_{m=1}^{n}\frac{(-1)^m}{m}\binom{n}{m}\frac{(m-1)!n!}{(m+n)!}\\ &= \binom{2n}{n}\sum\limits_{m=1}^{n}\frac{(-1)^m}{m}\binom{n}{m}\int_0^1 x^{m-1}(1-x)^n\,dx\\ &= \binom{2n}{n}\int_0^1\int_0^{x}\sum\limits_{m=1}^{n}\binom{n}{m}(-1)^my^{m-1} \frac{(1-x)^n}{x}\,dy\,dx\\ &= \binom{2n}{n}\int_0^1\int_0^{x}\frac{(1-y)^n -1 }{y}. \frac{(1-x)^n}{x}\,dy\,dx\\ &= \binom{2n}{n}\int_0^1 \frac{(1-x)^n}{x}\int_0^{x}\frac{(1-y)^n -1 }{y} \,dy\,dx\\ &= \binom{2n}{n}\int_0^1 \frac{x^n}{1-x}\int_0^{1-x}\frac{(1-y)^n -1 }{y} \,dy\,dx\\ &= -\binom{2n}{n}\int_0^1 \frac{x^n}{1-x}\int_0^{1-x}\frac{1-(x+y)^n}{1-(x+y)} \,dy\,dx\\ &= -\binom{2n}{n}\int_0^1 \frac{x^n}{1-x}\left[\sum\limits_{m=1}^{n}\frac{(x+y)^m}{m}\right]_0^{1-x}\,dx\\ &= -\binom{2n}{n}\int_0^1 \frac{x^n}{1-x}\sum\limits_{m=1}^{n}\frac{1-x^m}{m}\,dx\\ &= -\binom{2n}{n}\sum\limits_{m=1}^{n}\frac{1}{m}\sum\limits_{k=1}^{m}\frac{1}{n+k}= -\binom{2n}{n}\left(\sum\limits_{m=1}^{n}\frac{H_{m+n}}{m} - H_n^2\right)\end{align}$

The second one looks like a troublesome calculation to make.

$$\sum\limits_{m=1}^{n}\frac{(-1)^m}{m}\binom{2n}{n+m} = -\binom{2n}{n}(H_{2n} - H_n)$$

With @Chris's sis's help in chat I figured out how to simplify the expression further:

$$\begin{align}\sum\limits_{m=1}^{n}\frac{H_{m+n}}{m} &= \sum\limits_{m=1}^{n}\frac{H_m}{m}+\sum\limits_{m=1}^{n}\sum\limits_{k=1}^{n} \frac{1}{m(k+m)}\\&= \frac{1}{2}(H_n^{(2)}+H_n^2) + \frac{1}{2}\sum\limits_{m=1}^{n}\sum\limits_{k=1}^{n} \left(\frac{1}{m(k+m)}+\frac{1}{k(k+m)}\right)\\&= \frac{1}{2}(H_n^{(2)}+H_n^2) + \frac{1}{2}\sum\limits_{m=1}^{n}\sum\limits_{k=1}^{n} \frac{1}{mk}\\&= H_n^2 + \frac{1}{2}H_n^{(2)} \tag{1} \end{align}$$

Hence, $\displaystyle \sum_{m=1}^{n}\frac{(-1)^m}{m^2}\binom{2n}{n+m} = -\frac{1}{2}\binom{2n}{n}H_n^{(2)}$

Edit: For the second one, a short computation similar to one above shows:

$$\sum\limits_{m=1}^{n}\frac{(-1)^m}{m^4}\binom{2n}{n+m} = -\sum\limits_{m=1}^{n}\sum\limits_{j=1}^{m}\sum\limits_{k=1}^{j} \frac{H_{k+n} - H_n}{mjk}$$

Couldn't figure out a more direct line of computation, used differencing to complete the proof:

Call the triple sum $\displaystyle F(n) = \sum\limits_{m=1}^{n}\sum\limits_{j=1}^{m}\sum\limits_{k=1}^{j} \frac{H_{k+n} - H_n}{mjk}$

Then,

$\begin{align}F(n+1) - F(n) &= \sum\limits_{m=1}^{n+1}\sum\limits_{j=1}^{m}\sum\limits_{k=1}^{j} \frac{H_{k+n+1} - H_{n+1}}{mjk} - \sum\limits_{m=1}^{n}\sum\limits_{j=1}^{m}\sum\limits_{k=1}^{j} \frac{H_{k+n} - H_n}{mjk}\\ &= \sum\limits_{m=1}^{n+1}\sum\limits_{j=1}^{m}\sum\limits_{k=1}^{j} \frac{H_{k+n} - H_n - \frac{k}{(n+1)(n+k+1)} }{mjk} - \sum\limits_{m=1}^{n}\sum\limits_{j=1}^{m}\sum\limits_{k=1}^{j} \frac{H_{k+n} - H_n}{mjk}\\ &= \frac{1}{n+1}\sum\limits_{j=1}^{n+1}\sum\limits_{k=1}^{j} \frac{H_{k+n+1} - H_{n+1}}{jk} - \sum\limits_{m=1}^{n}\sum\limits_{j=1}^{m}\sum\limits_{k=1}^{j} \frac{k}{mjk(n+1)(n+k+1)}\\ &= \frac{1}{n+1}\left(\sum\limits_{j=1}^{n+1}\sum\limits_{k=1}^{j} \frac{H_{k+n+1} - H_{n+1}}{jk} - \sum\limits_{m=1}^{n}\sum\limits_{j=1}^{m} \frac{H_{n+j+1} - H_{n+1}}{mj}\right)\\&= \frac{1}{(n+1)^2}\sum\limits_{j=1}^{n+1} \frac{H_{n+j+1} - H_{n+1}}{j} = \frac{H_{n+1}^{(2)}}{2(n+1)^2}\end{align}$

In the last line I used the previous result $(1)$.

Hence, $\displaystyle F(n) = \frac{1}{2}\sum\limits_{k=1}^{n} \frac{H_{k}^{(2)}}{k^2} = \frac{\left(H_{n}^{(2)}\right)^2 + H_{n}^{(4)}}{4}$

$\endgroup$
  • $\begingroup$ Looks nice. Thanks. Do you know of any identities or ways to bound $(H_n^{(2)})^2$ $\endgroup$ – Hatchet Jul 5 '15 at 8:08
  • $\begingroup$ @Hatchet I suppose we could use $\frac{3}{2} - \frac{1}{n+1} = 1+\sum\limits_{k=2}^{n} \frac{1}{k(k+1)} \le H_n^{(2)} = 1+\sum\limits_{k=2}^{n} \frac{1}{k^2} \le 1+\sum\limits_{k=2}^{n} \frac{1}{k(k-1)} = 2 - \frac{1}{n}$ as a weak bound. $\endgroup$ – r9m Jul 5 '15 at 8:36
3
$\begingroup$

This is not an answer but it is too long for a comment.

For the first expression, d.k.o. gave a very elegant expression involving the first derivative of the digamma function and I do not think that I could provide anything simpler. $$S_n=\sum_{m=1}^{n}\frac{(-1)^m}{m^2}\binom{2n}{n+m}=2^{n-1}\left(\Psi^{(1)}(n+1)-\frac{\pi^2}{6}\right)\frac{(2n-1)!!}{n!}$$ For large values of $n$, we can replace $\Psi^{(1)}(n+1)$ by the expansion $$\Psi^{(1)}(n+1)=\frac{1}{n}-\frac{1}{2 n^2}+\frac{1}{6 n^3}-\frac{1}{30 n^5}+O\left(\frac{1}{n^6}\right)$$ which leads to quite accurate results.

For example

  • $S_{10}=-\frac{90915148181}{635040}\approx -143164.443$ while the above expansion gives $\approx -143164.413$.
  • $S_{100}=\approx -7.40226818241923\times 10^{58}$ while the above expansion gives $\approx -7.40226818240414\times 10^{58}$.

For both expressions, we can notice that the logarithm of the negative value of the sum varies almost linearly with $n$.

$\endgroup$
  • $\begingroup$ Thanks for the reply. Do you believe that both of these sums run parallel to a logarithm, i.e. can they be expressed as a logarithm? How would you find an elegant expression for the second series? $\endgroup$ – Hatchet Jul 4 '15 at 6:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.