I've not looked at finding the limits of sums before, and i'm interested to know how this particular example would be evaluated. The answer I believe should give $\frac{2}{3}$.
3 Answers
The usual way: $\sqrt{x}$ is a continuous function on $[0,1]$, hence it is Riemann-integrable and $$\int_{0}^{1}\sqrt{x}\,dx = \lim_{n\to+\infty}\frac{1}{n}\sum_{k=1}^{n}\sqrt{\frac{k}{n}}\tag{1}$$ The RHS of $(1)$ is your limit and by the fundamental theorem of Calculus the LHS of $(1)$ is $\frac{1}{1+\frac{1}{2}}=\color{red}{\large\frac{2}{3}}$ as wanted.
The unusual way: it is enough to understand the behaviour of $\sum_{k=1}^{n}\sqrt{k}$ with decent accuracy.
For such a task, we may employ creative telescoping, since
$$ \left(k+\frac{1}{2}\right)\sqrt{k+\frac{1}{2}}-\left(k-\frac{1}{2}\right)\sqrt{k-\frac{1}{2}} = \frac{3k^2+\frac{1}{4}}{\left(k+\frac{1}{2}\right)\sqrt{k+\frac{1}{2}}+\left(k-\frac{1}{2}\right)\sqrt{k-\frac{1}{2}}}$$ differs from $\frac{3}{2}\sqrt{k}$ by less than $\frac{1}{60 k^{3/2}}$ for any $k\geq 1$. In particular, $$ \left|\sum_{k=1}^{n}\sqrt{k}-\frac{2}{3}\left(n+\frac{1}{2}\right)^{3/2}\right|\leq \frac{1}{3\sqrt{2}}+\frac{1}{90}\sum_{k\geq 1}\frac{1}{k^{3/2}} = C $$ and the wanted limit is $\color{red}{\large\frac{2}{3}}$.

- 353,855
-
2Jack, I've provided an asymptotic expansion which shows that $$\lim_{n\to \infty}\left( \sum_{k=1}^n \sqrt{k}-\frac23(n+1/2)^{3/2} \right)=\zeta(-1/2)\approx 0.207$$Interestingly, your $C \approx 0.26$ which isn't a bad uniform approximation. – Mark Viola Mar 15 '17 at 05:00
-
@Dr.MV: that is nice but not really surprising, since CT and the EMC formula are ultimately two sides of the same approach. That is revealed by my approach to prove the Stirling's inequality, and by Mathieu's inequality too. – Jack D'Aurizio Mar 15 '17 at 16:54
-
Jack, are you suggesting than the EMSF is based on creative telescoping? I suppose in the abstract, one can view it that way. I am interested in a reference to your proof of Stirling's formula. -Mark – Mark Viola Mar 15 '17 at 18:36
-
@Dr.MV: I was referring to (http://math.stackexchange.com/questions/94722/stirlings-formula-proof/1409131#1409131). The EMC formula applied to $\sum_{n\geq N}\frac{1}{n^k}$ and creative telescoping lead to the same decomposition due to Faulhaber's formula, hence EMC is equivalent to CT for practical purposes. – Jack D'Aurizio Mar 15 '17 at 18:42
METHODOLOGY $1$:
Note that since $\sqrt x$ is monotonically increasing we can bound the sum of interest $S_n=\sum_{i=2}^{n+1}i^{1/2}$ by
$$\int_1^{n+1}\sqrt{x}\,dx\le S_n=\sum_{i=2}^{n+1}i^{1/2}\le \int_2^{n+2}\sqrt x\,dx$$
whereupon evaluating the integrals gives
$$\frac23((n+1)^{3/2}-1)\le \sum_{i=2}^{n+1}i^{1/2}\le \frac23((n+2)^{3/2}-2^{3/2})\tag 1$$
Dividing $(1)$ by $n^{3/2}$ and applying the squeeze theorem yields the coveted result
$$\lim_{n\to \infty}\frac{S_n}{n^{3/2}}=\frac23$$
METHODOLOGY $2$:
This is overkill for the purpose herein, but I thought it might be instructive to present an approach that provides an expansion of $S_n=\sum_{i=2}^{n+1}i^{1/2}$. To that end we proceed.
Using the Euler-Maclaurin Summation Formula, we can write
$$\begin{align} S_n&=\sum_{i=2}^{n+1}i^{1/2}\\\\ &=\int_1^{n+1}\sqrt{x}\,dx+\frac12(\sqrt{n+1}-\sqrt{1})+\frac{B_2}{2!}\frac12\left((n+1)^{-1/2}-^{-1/2}\right)\\\\ &+\frac{B_4}{4!}\frac{3}{8}\left((n+1)^{-5/2}-(1)^{-5/2}\right)+C+O\left((n+1)^{-9/2}\right)\\\\ &=\frac23(n+1)^{3/2}+\frac{\sqrt{n+1}}{2}+C'+\frac{1}{24}(n+1)^{-1/2}-\frac{1}{1920}(n+1)^{-5/2}+O\left((n+1)^{-9/2}\right) \end{align}$$
The constant $C'$ is given by the limit
$$\begin{align} C'&=\lim_{n\to \infty}\left(\sum_{i=2}^{n+1}i^{1/2}-\frac23(n+1)^{3/2}-\frac{\sqrt{n+1}}{2}\right)\\\\ &=\zeta\left(-\frac12\right)\\\\ &\approx 0.207886225 \end{align}$$
Hence, we have
$$\begin{align} S_n=\frac23(n+1)^{3/2}+\frac{\sqrt{n+1}}{2}+\zeta\left(-\frac12\right)+\frac{(n+1)^{-1/2}}{24}-\frac{(n+1)^{-5/2}}{1920}+O\left((n+1)^{-9/2}\right) \tag 2 \end{align}$$
Now, we see from $(2)$ that $\lim_{n\to \infty}\frac{S_n}{n^{3/2}}=\frac23$.

- 179,405
Hint: $$\ \frac1{n^{\frac32}}\sum_{i=\color{red}1}^{\color{red}n}=\frac1n \sum_{i=1}^{n}\Bigl(\frac in\Bigr)^{\!\frac12}$$ is an upper Riemann sum for the integral of a function on the interval $[0,1]$.

- 175,478
-
-
Without $\frac1{n^{3/2}}$, and for the bounds I mentioned, not for the exact bounds of the exercise, but it's easy to deduce the limit for the exercise from the integral. – Bernard Mar 15 '17 at 01:45