My goal is to get an inequality $\forall t>0$ for the following integral

$$ \int_0^t \left(\sum_{n=1}^\infty \exp(-n^2 t_0)\right)^2\,\mathrm{d}t_0 \le f(t). $$

The goal is to at least lose the the square of the series. Initially I thought it was possible to find a reasonable $f(t)$ using the Mean value theorem. However, there is not an easy upper-bound for $\sum_{n=1}^\infty \exp(-n^2 t_0)$ since it explodes around $t_0 \approx 0$. Does anybody have a clue?


Well, your integral is $\infty$ whenever $t>0$. We get for $u>0$: $$ \left(\sum_{n=1}^{\infty}\exp(-n^2u)\right)^2 = \sum_{n=1}^{\infty}\sum_{m=1}^{\infty}\exp(-(n^2+m^2)u)$$ By the Fubini-Tonelli theorem we can exchange the order of integration and summation for a nonnegative function. Fix $t>0$ and define $\mathcal{A}_t=\{(n,m)\in \mathbb{N}^2: (n^2+m^2)t\geq 1\}$. Then:
\begin{align} &\int_0^t \left(\sum_{n=1}^{\infty}\exp(-n^2u)\right)^2 du \\ &= \sum_{n=1}^{\infty}\sum_{m=1}^{\infty}\int_0^t \exp(-(n^2+m^2)u)du \\ &= \sum_{n=1}^{\infty}\sum_{m=1}^{\infty}\frac{1-\exp(-t(n^2+m^2))}{n^2+m^2}\\ &\geq \sum_{(n,m)\in\mathcal{A}_t} \frac{1-\exp(-t(n^2+m^2))}{n^2+m^2}\\ &\geq (1-e^{-1})\sum_{(n,m)\in\mathcal{A}_t} \frac{1}{n^2+m^2}\\ &=\infty \end{align} where the final equality can be shown because $\sum_{n=1}^{\infty}\sum_{m=1}^{\infty} \frac{1}{n^2+m^2}=\infty$, and $\sum_{(n,m)\in\mathcal{A}_t}$ only neglects a finite number of terms in comparison to $\sum_{n=1}^{\infty}\sum_{m=1}^{\infty}$.

  • $\begingroup$ I do not think you did the expansion correct, since the correct Cauchy product expansion should be: $\sum_{n=1}^\infty\sum_{m=1}^{n-1} \exp((-m^2-(n-m)^2)t_0)$. This would lead to the integral result $\int_0^{t} \sum_{n=1}^\infty\sum_{m=1}^{n-1} \exp((-m^2-(n-m)^2)t_0) = \sum_{n=1}^\infty\sum_{m=1}^{n-1} \frac{\exp((-m^2-(n-m)^2)t_0) - 1}{(-m^2-(n-m)^2}$. $\endgroup$ – Amp Mar 13 '16 at 19:17
  • $\begingroup$ @Amp : Why do you sum only from $m=1$ to $n-1$? $\endgroup$ – Michael Mar 13 '16 at 19:23
  • $\begingroup$ Excuse me, the integral should be $\int_0^{t} \sum_{n=1}^\infty\sum_{m=1}^{n-1} \exp((-m^2-(n-m)^2)t_0) dt_0= \sum_{n=1}^\infty\sum_{m=1}^{n-1} \frac{\exp((-m^2-(n-m)^2)t) - 1}{-m^2-(n-m)^2}$ $\endgroup$ – Amp Mar 13 '16 at 19:23
  • $\begingroup$ I have used the Cauchy product rule here, en.wikipedia.org/wiki/Cauchy_product $\endgroup$ – Amp Mar 13 '16 at 19:24
  • $\begingroup$ Before integrating, you need a correct value of the double sum. $\endgroup$ – Michael Mar 13 '16 at 19:24

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.