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

How do I compute the following limit or show it doesn't exists? $$\lim_{n\rightarrow\infty}\frac{n+\sqrt{n}+\cdots+\sqrt[n]{n}}{n}$$

I've struggled with this problem for a while now so I would appreciate a complete solution.

share|cite|improve this question
In your struggle, what have you come up with, and what did you try to use? – user23211 Jun 27 '12 at 0:48
@ymar Intuitively, the first few terms are significant because they are large and the last bunch of terms contribute because there are many of them. I have tried to split up the sum accordingly to no avail. – ryan Jun 27 '12 at 0:56
@ryan: how have you tried splitting the sum? (Have you tried splitting into the first $r$ terms and the rest without specifying $r$ in advance?) – Qiaochu Yuan Jun 27 '12 at 1:02
Stolz-Cesáro theorem is the way to go in such cases. It works fast and nice. – user 1618033 Jun 27 '12 at 4:51
up vote 12 down vote accepted

The limit is $2$.

Indeed by the Stolz-Cesáro theorem it suffices to show

$$\lim_{n\rightarrow\infty}\frac{\sum_{k=1}^{n+1}(n+1)^{1/k}-\sum_{k=1}^{n}n^{1/k}}{(n+1)-n}=\lim_{n\rightarrow\infty}1+(n+1)^{1/(n+1)}+\sum_{k=2}^{n}\left \{(n+1)^{1/k}-n^{1/k}\right \}=2.$$

Thus we need only show


The concavity of $x\mapsto x^{1/k}$ implies $(n+1)^{1/k}-n^{1/k}\leq n^{1/k-1}/k.$


$$\begin{align*} \sum_{k=2}^{n}(n+1)^{1/k}-n^{1/k}&\leq\sum_{k=2}^n\frac{n^{1/k-1}}{k}\\ &\leq\frac{1}{\sqrt n}\sum_{k=2}^n\frac{1}{k}=\mathcal{O}\left (\frac{\log n}{\sqrt n}\right ). \end{align*}$$

Since the relevant sum is manifestly positive for each $n$, it follows that $\lim_{n\rightarrow\infty}\sum_{k=2}^{n}(n+1)^{1/k}-n^{1/k}=0.$

share|cite|improve this answer

Well, since there is already an answer, let me link to this blog post where I describe how to compute this limit (and estimate the rate of convergence) by splitting the sum.

share|cite|improve this answer

The first term of the sum contributes $1$, so the focus would be on determining the limit of the average of the sum of the remaining terms. Since $n^{1 \over k}$ is approximately $1$ for the vast majority of these terms, it makes sense that this average will converge to $1$. Formally:

Given $\epsilon > 0$, we have that $1 < n^{1 \over k} < 1 + \epsilon$ whenever $k > k_0 = \frac{\ln(n)}{\ln(1 + \epsilon)}$. Thus we have $$\frac{n - k_0 - 1}{n} < {\sum_{k > k_0}n^{1 \over k} \over n} < (1 + \epsilon)\frac{n - k_0}{n}$$ On the other hand, for $2 \leq k \leq k_0$ the first term is largest, so we can bound each term by the first and we get $${\sum_{2 \leq k \leq k_0}n^{1 \over k} \over n} < \frac{k_0}{\sqrt{n}}$$ Adding these together and plugging back in for $k_0$ we get $$\frac{n -\frac{\ln(n)}{\ln(1 + \epsilon)}-1}{n} < {\sum_{k \geq 2}n^{1 \over k} \over n} < (1 + \epsilon)\frac{n -\frac{\ln(n)}{\ln(1 + \epsilon)}}{n} + \frac{\ln(n)}{\sqrt{n}\ln(1 + \epsilon)}$$ The right-hand side of this leads to $$\limsup_{n \rightarrow \infty} {\sum_{k \geq 2}n^{1 \over k} \over n} \leq 1 + \epsilon$$ The left-hand side leads to $$\liminf_{n \rightarrow \infty} {\sum_{k \geq 2}n^{1 \over k} \over n} \geq 1 $$ These hold for any $\epsilon > 0$, so the overall limit is $1$. Adding this to the $k=1$ term, we see the original limit is $2$.

share|cite|improve this answer

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.