Throughout this question $n$ is a positive integer greater than 1. Consider the following well-known identity by Euler, $$\sum_{k=1}^{n-1} \binom{2n}{2k}B_{2k}B_{2n-2k}=-(2n+1)B_{2n}.$$ Rather unconventionally, we can also write the above identity as $$\sum_{k=1}^{n} \binom{2n}{2k}B_{2k}B_{2n-2k}=-2n B_{2n}.$$

I was wondering if the following identity is also known. $$\sum_{k=1}^{n} \binom{2n}{2k}B_{2k}B_{2n-2k}=\sum_{k=1}^{n} 2^{2k} \binom{2n}{2k}B_{2k}B_{2n-2k}.$$

This would give the following nice identity $$\sum_{k=1}^{n}(2^{2k}-1) \binom{2n}{2k}B_{2k}B_{2n-2k}=0.$$

My proof involves Newton’s Identities and simple manipulations of the equation derived.


This is well known. Equation (11) in is this Wolfram article implies that


If $n>2$ is even, then $E_{n-2}(0)=0$, and in the sum all the $B_k$ are zero for odd $k$. This is your identity.

I think that the rule for Bernoulli numbers identities is: if it is a simple algebraic manipulation, it is surely known.


Here's a simple proof of the identity. Let $$B(x) = \frac{x}{e^x-1} = \sum_{n} B_n \frac{x^n}{n!}.$$ Then $$\sum_n (2^n-1) B_n \frac{x^n}{n!} = B(2x) - B(x) = -\frac{x}{e^x+1}$$ and $$\bigl(B(2x) -B(x)\bigr) B(x)=-\frac{x^2}{e^{2x}-1}=-\frac{x}{2}B(2x).$$ So $$\sum_{k=0}^n \binom nk (2^k-1)B_k B_{n-k}=-2^{n-2}n B_{n-1}.$$ For $n$ even this reduces to the OP's identity since $B_n=0$ for $n$ odd and greater than 1.


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.