I have a power series nestled inside a power series, and the inside power series is taken to a power. I need to isolate the term that would go like $t^{2mn}$ so that I could integrate it.

I'm not looking to sum up the series, since that's where I'm coming from.

$$ \sum_{n=0}^{\infty} (-1)^n \left(\sum_{m=0}^{\infty} \; \frac{(-1)^m}{(2m)!} \; \beta^{2m} t^{2m} \right)^n $$

Any ideas are appreciated.

  • 2
    $\begingroup$ The inner sum is just the cosine, no? $\endgroup$ – J. M. is a poor mathematician Mar 28 at 5:09
  • $\begingroup$ @J.M.isnotamathematician. Using your comment, the outer sum is simple too. Cheers $\endgroup$ – Claude Leibovici Mar 28 at 5:31

Here are two approaches to extract the coefficient of $t^{2MN}$. One is based upon multinomial expansion, the other is based upon the representation $\cos t=\frac{1}{2}\left(e^{it}+e^{-it}\right)$. Regrettably, none of them seems really promising. Nevertheless, here we go!

It is convenient to use the coefficient of operator $[t^n]$ to denote the coefficient of $t^n$.

First approach:

We obtain \begin{align*} \color{blue}{[t^{2MN}]}&\color{blue}{\sum_{n=0}^\infty(-1)^n\left(\sum_{m=0}^\infty\frac{(-1)^m}{(2m)!}\beta^{2m}t^{2m}\right)^n}\\ &\sum_{n=0}^\infty(-1)^n[t^{2MN}]\left(\sum_{m=0}^{MN}\frac{(-1)^m}{(2m)!}\beta^{2m}t^{2m}\right)^n\\ &=\sum_{n=0}^\infty(-1)^n[t^{2MN}]\sum_{{k_0+\cdots+k_{MN}=n}\atop{k_j\geq 0,\ ,0\leq j\leq MN}} \binom{n}{k_0,\ldots,k_{MN}}\prod_{j=0}^{MN}\left(\frac{\left(-\beta^2\right)^j}{(2j)!}t^{2j}\right)^{k_j}\\ &\,\,\color{blue}{=\sum_{n=0}^\infty(-1)^n[t^{2MN}]\sum_{{k_0+\cdots+k_{MN}=n}\atop{k_j\geq 0,\ ,0\leq j\leq MN}}\binom{n}{k_0,\ldots,k_{MN}}}\\ &\qquad\quad\color{blue}{\cdot\prod_{j=0}^{MN}\left((2j)!\right)^{-k_j}\left(-\beta^2\right)^{\sum_{j=1}^{N}jk_j}t^{2\sum_{j=1}^{N}jk_j}}\tag{1}\\ \end{align*}

We observe in order to find the coefficient of $t^{2MN}$ the expression (1) boils down to find the solutions of \begin{align*} k_0+k_1+\cdots+k_{MN}&=n\\ k_1+2k_2+\cdots+MNk_{MN}&=MN\qquad\qquad k_j\geq 0, \,\, 0\leq j\leq MN \end{align*} which indicates connections with Bell polynomials.

The second approach is somewhat more promising, since we can completely extract the coefficient of $t^{2MN}$ . But, we will see the final expression is far from being simple.

Second approach:

\begin{align*} \color{blue}{[t^{2MN}]}&\color{blue}{\sum_{n=0}^\infty(-1)^n\left(\sum_{m=0}^\infty\frac{(-1)^m}{(2m)!}\beta^{2m}t^{2m}\right)^n}\\ &=[t^{2MN}]\sum_{n=0}^\infty(-1)^n\cos^n(\beta t)\tag{2}\\ &=[t^{2MN}]\sum_{n=0}^\infty(-1)^n\frac{1}{2^n}\left(e^{i\beta t}+e^{-i\beta t}\right)^n\tag{3}\\ &=[t^{2MN}]\sum_{n=0}^\infty\left(-\frac{1}{2}\right)^ne^{-in\beta t}\left(1+e^{2i\beta t}\right)^n\tag{4}\\ &=[t^{2MN}]\sum_{n=0}^\infty\left(-\frac{1}{2}\right)^n\sum_{j=0}^\infty\frac{\left(-in\beta\right)^j}{j!}t^j\left(1+e^{2i\beta t}\right)^n\tag{5}\\ &=\sum_{n=0}^\infty\left(-\frac{1}{2}\right)^n\sum_{j=0}^{2MN}\frac{\left(-in\beta\right)^j}{j!}[t^{2MN-j}]\left(1+e^{2i\beta t}\right)^n\tag{6}\\ &=\sum_{n=0}^\infty\left(-\frac{1}{2}\right)^n\sum_{j=0}^{2MN}\frac{\left(-in\beta\right)^{2MN-j}}{(2MN-j)!}[t^{j}]\sum_{k=0}^n\binom{n}{k}e^{2ik\beta t}\tag{7}\\ &=\sum_{n=0}^\infty\left(-\frac{1}{2}\right)^n\sum_{j=0}^{2MN}\frac{\left(-in\beta\right)^{2MN-j}}{(2MN-j)!} [t^{j}]\sum_{k=0}^n\binom{n}{k}\sum_{l=0}^\infty \frac{\left(2ik\beta\right)^l}{l!}t^l\tag{8}\\ &=\sum_{n=0}^\infty\left(-\frac{1}{2}\right)^n\sum_{j=0}^{2MN}\frac{\left(-in\beta\right)^{2MN-j}}{(2MN-j)!}\sum_{k=0}^n\binom{n}{k}\frac{(2ik\beta)^j}{j!}\tag{9}\\ &\,\,\color{blue}{=\frac{\left(-\beta^2\right)^{MN}}{(MN)!}\sum_{n=0}^\infty\left(-\frac{1}{2}\right)^nn^{2MN} \sum_{j=0}^{2MN}\binom{2MN}{j}\left(-\frac{2}{n}\right)^j\sum_{k=0}^n\binom{n}{k}k^j}\tag{10} \end{align*}


  • In (2) we use the series expansion $\cos t=\sum_{n=0}^\infty \frac{(-1)^n}{(2n)!}t^{2n}$.

  • In (3) we use the representation $\cos t=\frac{1}{2i}\left(e^{it}+e^{-it}\right)$.

  • In (4) we factor out $e^{-in\beta t}$.

  • In (5) we expand $e^{-in\beta t}$.

  • In (6) we apply the rule $[t^{p-q}]A(t)=[t^p]t^qA(t)$ and we set the upper limit of the sum with index $j$ to $2MN$, since other terms do not contribute.

  • In (7) we change the order of the inner sum $j\to 2MN-j$.

  • In (8) we expand $e^{2ik\beta t}$.

  • In (9) select the coefficient of $t^j$.

  • In (10) we do some final rearrangements.


Seems like a strange approach but the only thing I think would help would be this: https://en.m.wikipedia.org/wiki/Cauchy_product Although you'd need to establish a general form for the coefficients when multiplying more than 2 series, which would include several nested finite sums, making the result really ugly


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.