Let $a_1>1$ and $$a_{n+1}=a_n^{n+1}+a_n+1, \: \forall n \geq 1$$ Edit: Prove that there is a real number $x \neq 0$ such that $$\lim_{n \to \infty}\frac{a_n}{x^{n!}}=1$$

I think that $x=a_1$ works, and I tried to use the Squeeze Theorem to prove that. From $a_{n+1}>a_n^{n+1}>a_{n-1}^{n(n+1)}>\dots>a_1^{(n+1)!}$ we get that $1<\frac{a_{n+1}}{a_1^{(n+1)!}}$ and then I tried to find another bound for $a_{n+1}$ such that $\frac{a_{n+1}}{a_1^{(n+1)!}}<b_{n+1} \to 1$, but nothing seems to work.

  • $\begingroup$ No, that doesn't work, and the formulation is misleading. You can show the existence of such an $x$, but you can't find it without computing the whole sequence. $\endgroup$ – Professor Vector Jan 12 '18 at 11:54
  • $\begingroup$ Okay, I'll edit the statement! $\endgroup$ – Shroud Jan 12 '18 at 11:55


$$x = \lim_{n\to\infty} a_n^{1/n!}.$$

We know from @frame95's answer that $x$ exists. Now set $\delta_n = \log a_n - n! \log x$ and notice that

$$ \frac{\delta_{n+1}}{(n+1)!} = \frac{\delta_n}{n!} + \frac{1}{(n+1)!}\log\left( 1 + \frac{1}{a_n^n} + \frac{1}{a_n^{n+1}} \right). $$

Since we know that $\delta_n/n! \to 0$, recursively applying the above relation and taking limit shows that

$$ \frac{\delta_n}{n!} = -\sum_{k = 0}^{\infty} \frac{1}{(n+k+1)!}\log\left( 1 + \frac{1}{a_{n+k}^{n+k}} + \frac{1}{a_{n+k}^{n+k+1}} \right) $$

Now by utilizing simple estimations $a_{n+k} \geq a_n^{k!} \geq a_1^{k!}$ and $\log(1+t) \leq t$ for $t \geq 0$, we obtain the following very crude bound

$$ \left| \delta_n \right| \leq \sum_{k=0}^{\infty} \frac{n!}{(n+k+1)!} \cdot 2 a_1^{-k!}. $$

Since each summand is bounded by $2a_1^{-k!}$ and $\sum_{k=0}^{\infty} 2a_1^{-k!} < \infty$, dominated convergence theorem tells that

$$ \limsup_{n\to\infty} \left| \delta_n \right| \leq \sum_{k=0}^{\infty} \lim_{n\to\infty} \frac{n!}{(n+k+1)!} \cdot 2 a_1^{-k!} = 0. $$

Therefore $\delta_n \to 0$ and hence $a_n!/x^{n!} = e^{\delta_n} \to 1$ as required.

  • 1
    $\begingroup$ Wow! Good (team) job guy :) very neat proof! $\endgroup$ – Andrea Marino Jan 13 '18 at 10:36

The guess is wrong, but if you go a step further... Let's stop the chain of inequalities a bit before: $$ a_n> a_{n-1}^n > \ldots > a_2 ^{n!/2} = (\sqrt{a_2} )^{n!}$$ Is this a better estimate? Yes, because $ \sqrt{a_2 }=\sqrt{ a_1^2 + a_1+1} > a_1$. Let's call $d_r = a_r^{1/r!}$. Generalizing this argument, we get $$ a_{r+k} > a_{r+k-1}^{r+k} > \ldots > a_r^{ (r+k)!/r!} = d_r^{(r+k)!}$$ Let's analyze $\{d_r\}$. We will show that it is an increasing and bounded sequence, thus convergent to a number $\ell$. Infact $$ (1) \ \ increasing \ \ \ d_{r+1} = a_{r+1}^{1/(r+1)!} = ( a_r^{r+1} +a_r+1)^{1/(r+1)!} > d_r $$ By induction, one can easily show that $a_n +1 < (a_1+1) a_n^{n+1}$. Thus, $a_{n+1} < (a_1+2)a_n^{n+1}$. This yields $$ (2) \ \ bounded \ \ \ d_{n+1} < (a_1+2)^{1/(n+1)!} d_n < 2^{\sum^{n+1} (1/k!)} a_1/2 < 2^{e-1} a_1 $$ If $x$ exists, then $x=\ell$. Infact $$ \lim_{n \to \infty} \frac{a_n}{x^{n!}} = \lim_{n \to \infty} \left ( \frac{d_n}{x} \right )^{n!} = 1 \ \Rightarrow \ \frac{\ell}{x} = \lim_{n \to \infty} \frac{d_n}{x} = 1 $$ Here my results stop: I can't show the refined result that this $x$ exist!


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.