I came across this crazy sum, and I have no idea how to tackle it


I tried to approximate it using wolfram alpha widget,

and got something like $1.13134$.

However, I'd like to get an exact solution.

EDIT: i was thinking that maybe we could play around with the ceiling function a little bit. Consider that:


is actually the area under the curve


from $0$ to $\infty$. More formally, this can be expressed as: $$I=\int_0^\infty \frac{1}{\left\lceil{k^k}\right\rceil}\frac{1}{\left\lceil{k!}\right\rceil} dk=\sum_{k=1}^{\infty}\frac{1}{k!k^k}.$$

If we multiply $I$ by: $$L=\int_0^\infty{\left\lceil{k^k}\right\rceil}{\left\lceil{k!}\right\rceil} e^{-k}dk.$$

We get that:

$$I L=\int_0^\infty e^{-k} dk=1.$$


$$\sum_{k=1}^{\infty}\frac{1}{k!k^k}\int_0^\infty{\left\lceil{k^k}\right\rceil}{\left\lceil{k!}\right\rceil} e^{-k}dk=1,$$

so if we could somehow evaluate $L$, then it might be possible to calculate the sum.

Unfortunately, I'm not sure if this reasoning works.

  • 1
    $\begingroup$ Exact solution, seems impossible with Earth math. $\endgroup$ – Mariusz Iwaniuk Apr 13 '19 at 18:42
  • $\begingroup$ $\displaystyle{1 \over k!\, k^{k}} = \left[z^{k}\right]\mathrm{e}^{z/k}$. $\endgroup$ – Felix Marin Apr 13 '19 at 19:38
  • $\begingroup$ I'd like to point out in general $\left(\int_0^\infty f(x) dx\right)\left(\int_0^\infty g(x) dx\right) \ne \int_0^\infty f(x) g(x) dx$ so your $IL=1$ identity is false. (In fact, the integral $L$ diverges). $\endgroup$ – Dark Malthorp May 5 '20 at 0:58

I'm afraid I have to concur with @MariuszIwaniuk. I'll summarise what I found.

In a variant on the famous sophomore's dream calculation, note first that, since in terms of modified Bessel functions $\sum_{n\ge 0}\frac{y^n}{n!(n+1)!}=\frac{I_1(2\sqrt{y})}{\sqrt{y}}$, the substitution $u=-\ln x$ obtains $$\int_0^1\frac{I_1(2\sqrt{-x\ln x})}{\sqrt{-x\ln x}}dx=\sum_{n\ge 0}\frac{(-1)^n}{n!(n+1)!}\int_0^1x^n\ln xdx=\sum_n\frac{\int_0^1u^ne^{-(n+1)u}du}{n!(n+1)!}=\sum_{k\ge 1}\frac{1}{k!k^k}.$$Now we just need to evaluate that integral... which seems equally impossible. I also had no look identifying the value with the inverse symbolic calculator.


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.