Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I would like to estimate the scale of the following series,

$$S(m,t)=e^{-(m+1) t} \sum _{k=0}^{\infty } \frac{t^k}{k!}\left(\sum _{r=0}^k \frac{t^r}{r!}\right)^{m},$$

where $e$ is the base of natural logarithm, $m$ is an nonnegative integer and $t$ is a positive real number.

I turned the equation into another form,

$$S(m,t) = e^{-t} \sum _{k=0}^{\infty } \frac{t^k }{k!}\left(\frac{\Gamma (k+1,t)}{k!}\right)^m.$$ I noticed that $S(0,t)$ = 1, $S(1,t)$ is highly related to the existing question Cauchy product on exponential-looking power series, and $$S(m_2,t)<S(m_1,t)<1\,\text{ if }\,( 1<m_1<m_2).$$ But I cannot go further in giving an appropriate scale of $S(m,t)$.

Is it possible that $S(m, t)$ can be estimated as a polynomial of $t$? I am interested in giving an estimate of $S(m,t)$ as $S(m,t) = O (t^{-\alpha m}),(\alpha >0 )$ or something like that.
Would anyone give me any clue on this? Thanks in advance!

share|cite|improve this question
up vote 2 down vote accepted

It turns out that $S(m,t)=P(A)$ where $A=[X_{m+1}\geqslant\max\{X_i;1\leqslant i\leqslant m\}]$ and $(X_i)_{1\leqslant i\leqslant m+1}$ is i.i.d. with Poisson distribution with parameter $t$. To see this, note that, for every $k\geqslant0$, $$ P(A\mid X_{m+1}=k)=P(X_1\leqslant k)^m=\left(\mathrm e^{-t}\sum_{r=0}^k\frac{t^r}{r!}\right)^m, $$ and that $$ P(A)=E(P(A\mid X_{m+1}))=\sum_{k=0}^{+\infty}\mathrm e^{-t}\frac{t^k}{k!}P(A\mid X_{m+1}=k). $$ Let $\mathbf I=\{1,2,\ldots,m+1\}$, $M=\max\{X_i;i\in\mathbf I\}$, $A_J=[\forall i\in J, X_i=M]$ for every $J\subseteq\mathbf I$, and $B_i=A_{\{1,2,\ldots,i\}}$ for every $i$ in $\mathbf I$. Then $P(A_J)=P(B_i)$ for every $J\subseteq\mathbf I$ of size $i$ and $$ S(m,t)=P(B_1). $$ The union of the sets $A_{\{i\}}$ for $i$ in $\mathbf I$ is the universe hence $(m+1)P(B_1)\geqslant1$, that is, $$ S(m,t)\geqslant\frac1{m+1}. $$ In the other direction, the exact inclusion-exclusion formula $$ 1=\sum_{i=1}^{m+1}(-1)^{i+1}{m+1\choose i}P(B_i), $$ yields the inequality $1\geqslant (m+1)P(B_1)-\frac12m(m+1)P(B_2)$ hence $$ S(m,t)\leqslant\frac1{m+1}+\frac12mP(B_2). $$ At this point, one can write down $P(B_2)$ as the sum of the obvious series but I fail to see how to estimate it accurately... Maybe you have an idea?

share|cite|improve this answer
Very impressive answer, but still I have no idea how to estimate this series…Therefore, I raised another question hoping you could help me out. What is the probability of the number 1 and number 2 employees getting the bonus at a call center? – Erdos Yi Jun 19 '13 at 0:22

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.