10
$\begingroup$

I want to determine the limit given below: $$\lim_{n\to\infty}\frac{1^n+2^n+3^n+\ldots+n^n}{n^{n+1}}$$ I have tried to solve thise several times ,but with no results.I have tried using lema stolz cezaro and managed to find a general formula for the summation, but couldnt prove it $$\frac{1}{n}\sum_{k=0}^{n-1}\left(1-\frac{k}{n}\right)^n$$ Edit: Thank you for your responses,but i want to ask you if it is possible to solve the problem by not using integrals :D. NOTEIm in the eleventh grade and i had been given this problem by my math tutor ,so, as a consequence i dont know how to interpretate the great majority of your answers.Thanks for your time ,gents!

$\endgroup$
11
$\begingroup$

You only need to exploit the identity (proved here): $$\sum_{k=0}^{K}\binom{n+k}{n}=\binom{n+K+1}{n+1}$$ and the trivial bound: $$ m!\binom{n}{m} \leq n^m \leq m!\binom{n+m}{m}$$ to have: $$ \forall m\in\mathbb{N},\quad \lim_{n\to +\infty}\frac{1^m+2^m+\ldots+n^m}{n^{m+1}}=\frac{1}{m+1}$$ from which it follows that your limit is zero.

As an alternative approach, since over $[0,1]$ we have $1-x\leq e^{-x}$:

$$ \frac{1}{n}\sum_{k=0}^{n}\left(1-\frac{k}{n}\right)^n \leq \frac{1}{n}\sum_{k=0}^{n}e^{-k}\leq\frac{1}{n}\sum_{k=0}^{+\infty}e^{-k}=\frac{e}{n(e-1)}.$$

$\endgroup$
  • $\begingroup$ I wanted to ask you soemthing.For the last relation presented when k=n-1 it is 0^infinity.I have solved it the same way and i was told it has some problems. $\endgroup$ – Razvan Paraschiv Apr 21 '15 at 15:23
  • $\begingroup$ @RazvanParaschiv: my last line does not involve any limit, it is just an inequality holding for every $n\geq 1$. The trick is to consider limits just after that and state that the limit is zero by squeezing, since $$\lim_{n\to +\infty}\frac{e}{n(e-1)}=0,$$ no issues. $\endgroup$ – Jack D'Aurizio Apr 21 '15 at 15:28
  • 1
    $\begingroup$ Thank you! I apologize for my miscomprehension $\endgroup$ – Razvan Paraschiv Apr 21 '15 at 15:29
  • $\begingroup$ Jack: does 'from which it follows that your limit is zero' actually follow? After all, $\lim_{n\to\infty}(1+\frac1n)^m=1$ for all $m$, but that doesn't imply the same result when $m$ is replaced by the limiting variable. Isn't this the exact same situation? $\endgroup$ – Steven Stadnicki Apr 21 '15 at 15:34
  • $\begingroup$ @StevenStadnicki: well, in that case one may prove that $$ f(m,n) = \frac{1^m+\ldots+n^m}{n^{m+1}} $$ is monotonic with respect to both $m$ and $n$ and exploit this fact to be allowed to exchange limits. $\endgroup$ – Jack D'Aurizio Apr 21 '15 at 15:37
9
$\begingroup$

Note that $\sum_{j=1}^n j^n \le \int_{1}^{n+1} x^n$. (And recall that $((n+1)/n)^{n}$ has a finite limit.)

$\endgroup$
  • $\begingroup$ Thanks. Yours has the advanateg of remaining more discrete, which I generally like. $\endgroup$ – quid Apr 21 '15 at 15:06
  • 1
    $\begingroup$ I like it too. I think I have managed to find another proof, similar to yours, and without integrals at all. Just see below. $\endgroup$ – Jack D'Aurizio Apr 21 '15 at 15:11
  • $\begingroup$ @JackD'Aurizio I see in my browser above :D $\endgroup$ – RE60K Apr 25 '15 at 15:37
3
$\begingroup$

Here's another point of view. For each $n\in\mathbb N$ choose an integer $k(n)\in\{0,1,\ldots,n\}$ such that the following two properties are satisfied: $$\lim_{n\to\infty}\frac{k(n)}{n}=1\qquad\text{and}\qquad\lim_{n\to\infty}\left(\frac{k(n)}{n}\right)^n=0.\tag{$\ast$}$$ Some possible choices are: $$k(n)=\lfloor n-\log{n}\rfloor\qquad\text{or}\qquad k(n)=\lfloor n-\sqrt n\rfloor\qquad\text{or}\qquad k(n)=\lfloor n^{\frac{n}{n+1}}\rfloor,$$ where $\lfloor x\rfloor$ is the largest integer not exceeding $x$, i.e. the floor function. (That $(\ast)$ holds follows e.g. from $\lim_{n\to\infty}(1+\frac1n)^n=e$ in the first two cases and from $\lim_{n\to\infty}n^{\frac1n}=1$ in the third case.)

Given such a sequence $k(n)$, we have the following bounds: $$\begin{align}0&\leq\frac{1^n+2^n+\ldots+k(n)^n+(k(n)+1)^n+\ldots+n^n}{n^{n+1}}\\&\leq\frac{k(n)k(n)^n+(n-k(n))n^n}{n^{n+1}}\\&=\left(\frac{k(n)}{n}\right)^{n+1}+1-\frac{k(n)}{n}.\end{align}$$ This last sequence converges to zero by $(*)$, so our sequence is bounded between two sequences converging to zero. We may conclude that $$\lim_{n\to\infty}\frac{1^n+2^n+3^n+\ldots+n^n}{n^{n+1}}=0.$$

$\endgroup$
  • $\begingroup$ that is the "razor" function? $\endgroup$ – Razvan Paraschiv Apr 25 '15 at 14:19
  • $\begingroup$ @RazvanParaschiv: Yes, I guess you could call it that. I should probably remark that the main part of this proof (requiring most of the work) consists of finding such a function and proving its properties, i.e. $(\ast)$. If you have any further questions about this, ask and I'll clarify. $\endgroup$ – Dejan Govc Apr 25 '15 at 14:43
  • $\begingroup$ the square brakets signify the full part of their content? $\endgroup$ – Razvan Paraschiv Apr 25 '15 at 14:47
  • $\begingroup$ @RazvanParaschiv: you mean the whole part? Yes, $\lfloor x\rfloor$ is defined as the unique integer $k$ such that $k\leq x<k+1$. See here and here. $\endgroup$ – Dejan Govc Apr 25 '15 at 15:00
  • $\begingroup$ thank you sir indeed ! in my country the notation si the following x<=[x]<=x+1 $\endgroup$ – Razvan Paraschiv Apr 25 '15 at 15:57

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.