1

I have the following double sum that I am trying to simplify into a single sum: \begin{align} \sum_{s=0}^\infty \sum_{m=0}^\infty b_{2s,2m}x^{2s}(1-x^2)^{m} \end{align} where $b_{s,m}$' are coefficients that don't have any specific structure.

But I am having difficulty finding the coefficients of the new sum $\sum_{k} a_k x^k$.

Here is what I tried to do. First, it is clear that only even coefficients should be considered.

Second, we can use the Binomial theorem to write \begin{align} (1-x^2)^{m}= \sum_{i=0}^{m} { m \choose i} (-1)^i x^{2i} \end{align} so we can combine everything to \begin{align} \sum_{s=0}^\infty \sum_{m=0}^\infty b_{2s,2m}x^{2s} \sum_{i=0}^{m} { m \choose i} (-1)^i x^{2i} \end{align}

But at this point, I start getting lost.

Second approach I was thinking we can define \begin{align} f(x)=\sum_{s=0}^\infty \sum_{m=0}^\infty b_{2s,2m}x^{2s}(1-x^2)^{m} \end{align} and then find the power series of $f(x)$ around $x=0$. The first coefficients are given by \begin{align} a_0&=\frac{f(0)}{1}=b_{0,0}\\ a_1&=\frac{f'(0)}{1}=0\\ a_2&=\frac{f''(0)}{2}=- \sum_{m=0}^\infty m b_{0,m}+ \sum_{m=0}^\infty b_{2,m} \end{align} but again I am stuck with finding other coefficients.

Lisa
  • 2,941

1 Answers1

2

The sum is $\sum_{s=0}^\infty \sum_{m=0}^\infty b_{2s,2m}x^{2s} \sum_{i=0}^{m} { m \choose i} (-1)^i x^{2i}$

We find the coefficient of $x^{2k}$:

We may have $0 \leq s \leq k$, for a fixed $s$ we require a $x^{2(k-s)}$ term from the $(1-x^2)^m$ factor, in particular we require that $m \geq k-s$. The coefficient of the $x^{2(k-s)}$ term in $(1-x^2)^m$ is ${m \choose (k-s)} (-1)^{k-s}$, so the coefficient of $x^{2k}$ is $\sum_{s=0}^k \sum_{m \geq (k-s)} b_{2s,2m} {m \choose (k-s)} (-1)^{k-s} = c_k$, the final sum is $\sum_{k=0}^{\infty} c_kx^{2k}$

  • Can one establish a recursion between $c_k$? There seem to be some terms that repeat. – Lisa Jun 25 '21 at 20:13
  • Not sure what you mean, could you specify what sort of recurrence you're looking for/what terms repeat? – porridgemathematics Jun 25 '21 at 20:18
  • I mean. It look like we can write $c_{k+1}$ in terms of $c_{k}$. Maybe I am wrong. – Lisa Jun 25 '21 at 20:20
  • You might be interested in the following question that is related to this one: https://math.stackexchange.com/questions/4183166/if-sum-s-0-infty-sum-m-0-infty-b-s-mx2s1-x2m-0-for-all-x-i – Lisa Jun 25 '21 at 21:14