5

Compute: $$L=\lim_{n\to \infty}\frac{1+\sqrt{2}+\dots+\sqrt{n}}{n\sqrt{n}}$$

I know some calculation of this limit using integrals, the anwser is $L=\frac{2}{3}$. However, I have an idea to compute it simpler by using just simple inequalities. My idea is to find an upper-bound for $1+\sqrt{2}+\dots+\sqrt{n}$ of the form $\frac{2}{3}n\sqrt{n+c}$ for some constant $c>0$ and its lower-bound is just $\frac{2}{3}n\sqrt{n}$. Hence, we got $2/3\leq L\leq \lim_{n\to \infty}\frac{2/3n\sqrt{n+c}}{n\sqrt{n}}=2/3$.

  1. For lower-bound: I can simply prove $1+\sqrt{2}+\dots+\sqrt{n}\geq \frac{2}{3}n\sqrt{n}$ for every $n$ by induction.
  2. For upper-bound: I'm stucking at choosing $c$, I think $c=7$ works.

How can I continue this direction?

Blue
  • 83,939
  • There is another possible approach : use the generalized harmonic numbers and their asymptotics. – Claude Leibovici Oct 23 '24 at 04:19
  • https://math.stackexchange.com/q/2415683/42969, https://math.stackexchange.com/q/1526398/42969, https://math.stackexchange.com/q/1172144/42969 – Martin R Oct 23 '24 at 04:51
  • 1
    I can’t find the answer in the old topics. So what can $c$ be? Does $c=7$ work? – Tung N. Dinh Oct 23 '24 at 05:11
  • 5
    I think it's pretty clear why none of the duplicates answer OP's question. The problem isn't merely to compute this limit, but to explain how to finish the argument that OP started. Skimming through the listed "duplicates", I don't see that anyone has approached this problem using the technique OP came up with, so they won't help OP finish their argument. – Chris Grossack Oct 23 '24 at 05:58
  • 2
    In that vein, though, OP might find these two questions helpful: 1, 2. They aren't directly the inequality you asked about, but they're in the same spirit (and will also finish your proof sketch). – Chris Grossack Oct 23 '24 at 06:19

2 Answers2

3

My claim : $\frac{2}{3}(n+1)\sqrt{n+1} \gt \sum_{r=1}^n \sqrt{r}$
To prove this we shall use induction, for $n=1$ this is true.
Now assume this be true for $n-1$. So we have :
$\frac{2}{3}n\sqrt{n} \gt \sum_{r=1}^{n-1}\sqrt{r}$
$ \Rightarrow \frac{2}{3} n\sqrt{n} +\sqrt{n}\gt \sum_{r=1}^n \sqrt{r}$
So if we show $\frac{2}{3}(n+1)\sqrt{n+1} \gt \frac{2}{3} n\sqrt{n} +\sqrt{n}$ we are done
$\frac{2}{3}(n+1)\sqrt{n+1} - \frac{2}{3} n\sqrt{n} -\sqrt{n}$
$ = \frac{2}{3} \left( (n+1)^{\frac{3}{2}} - (n)^{\frac{3}{2}}\right)-\sqrt{n}$
$ = \frac{2}{3}\left( \sqrt{n+1} - \sqrt{n} \right) \left( n+1 + n + \sqrt{n(n+1)} \right) -\sqrt{n}( \sqrt{n+1} - \sqrt{n} )( \sqrt{n+1} + \sqrt{n} )$
$ = \frac{2}{3}\left( \sqrt{n+1} - \sqrt{n} \right) \left( n+1 + n + \sqrt{n(n+1)} -\frac{3}{2}n -\frac{3}{2}\sqrt{n(n+1)}\right) $
$ = \frac{2}{3}\left( \sqrt{n+1} - \sqrt{n} \right) \left( \frac{1}{2} +\frac{n+1}{2} -\frac{1}{2} \sqrt{n(n+1)} \right)$
$ =\frac{1}{3}\left( \sqrt{n+1} - \sqrt{n} \right) \left(1+ \sqrt{n+1}\left(\sqrt{n+1}-\sqrt{n} \right)\right) \gt 0$
So we have proved for $n$.
Thus we have : $$\frac{2}{3}(n+1)\sqrt{n+1} \gt \sum_{r=1}^n \sqrt{r} \gt \frac{2}{3}n\sqrt{n}$$

Rithvick
  • 124
1

Remark. We can use the method of unknown coefficients (by assuming the form of the upper and lower bounds) and the Mathematical Induction to evaluate the limit.

First of all, we need to guess the appropriate form of the upper and lower bounds. Here we guess that there exist constants $L$ and $c_1, c_2$ such that for sufficiently large $n$, $$L\cdot n\sqrt{n} - c_1 n\le 1 + \sqrt{2} + \cdots + \sqrt{n} \le L \cdot n\sqrt{n} + c_2 n.$$ (Note: We can guess other forms. For example, if we replace $c_1n$ and $c_2n$ with $c_1\sqrt{n}$ and $c_2\sqrt{n}$ respectively, it still works. )

If the Mathematical Induction works, we have \begin{align*} 1 + \sqrt{2} + \cdots + \sqrt{n} + \sqrt{n+1} &\le L\cdot n \sqrt{n} + c_2 n + \sqrt{n+1} \\ &\le L \cdot (n+1)\sqrt{n+1} + c_2(n+1), \tag{1} \end{align*} and \begin{align*} 1 + \sqrt{2} + \cdots + \sqrt{n} + \sqrt{n+1} &\ge L\cdot n\sqrt{n} - c_1 n + \sqrt{n+1}\\ &\ge L\cdot (n+1)\sqrt{n+1} - c_1 (n+1). \tag{2} \end{align*}

From (1), we have $$L[(n+1)^{3/2} - n^{3/2}] + c_2 - \sqrt{n+1} \ge 0,$$ or $$L[(n+1) - n^{3/2}(n+1)^{-1/2}] + \frac{c_2}{\sqrt{n+1}} - 1 \ge 0. \tag{3}$$ Taking limit on (3), using $\lim_{n\to \infty} [(n+1) - n^{3/2}(n+1)^{-1/2}] = \frac32$, we have $\frac32 L - 1 \ge 0$. Thus, $L \ge \frac23$.

From (2), we have $$c_1 + \sqrt{n+1} - L[(n+1)^{3/2} - n^{3/2}] \ge 0,$$ or $$\frac{c_1}{\sqrt{n+1}} + 1 - L[(n+1) - n^{3/2}(n+1)^{-1/2}] \ge 0.\tag{4}$$ Taking limit on (4), using $\lim_{n\to \infty} [(n+1) - n^{3/2}(n+1)^{-1/2}] = \frac32$, we have $1 - \frac32 L \ge 0$. Thus, $L \le \frac23$.

Thus, we have $L = \frac23$.

The rest is smooth. For example, we choose $c_1= c_2 = 1$, and indeed the Mathematical Induction works.

River Li
  • 49,125