-2

I’d like to calculate $$ \lim_{n\to\infty}\frac{\ln n}{n}$$ by using the squeeze theorem.

So I’m looking for a sequence with members that are bigger than $\frac{\ln n}{n}$, but which also converges to zero.

How is it possible to evaluate this sequence without the exponent?

amWhy
  • 209,954
  • i would consider the function $$f(x)=\frac{\ln(x)}{x}$$ for $x>0$ – Dr. Sonnhard Graubner Nov 10 '17 at 16:56
  • What do you mean by evaluating without the exponent? – Sil Nov 10 '17 at 17:12
  • 1
    What facts about the logarithm can be used? – Daniel Fischer Nov 10 '17 at 17:12
  • @Sil, I meant that I wouldn’t like to use the fact that I can substitute $\frac{\ln n}{n}$ by $\frac{n}{e^n}$ – John Doe Nov 10 '17 at 17:20
  • @Daniel Fischer, I think that if I can prove that $\ln n < \sqrt(n)$, it will be enough. So there are mostly any ways to find a sequence that grows faster than $\ln n$ and slower than $n$. – John Doe Nov 10 '17 at 17:24
  • I think @AlexM's answer here Why is $\frac{\ln\infty}{\infty}$ equal to $\frac\infty\infty$? is what you want. You can prove that inequality for integers by induction. See this: Proof by induction: $\log n < n$ for $n ≥ 1$. – Sil Nov 10 '17 at 17:25
  • @Sil, no, they use exponential representation. Okay, I got the exact case. I’d like to prove that $\ln n < \sqrt n$, but I can’t prove it. – John Doe Nov 10 '17 at 17:30
  • Where they use the exponential representation? And why are you afraid of exponential function anyway? Doesn't make much sense talking about $\ln n$ without knowing exponential function first :) – Sil Nov 10 '17 at 17:32
  • 2
    Which definition of $\ln$ are you using? – Daniel Fischer Nov 10 '17 at 17:36
  • @Sil, I don’t afraid of it, I just want to make it in a different way.:) I’ve already come to an idea how to do it. So if you can help me to prove that $\ln n < \sqrt n$ by induction, I got what I wanted.:) – John Doe Nov 10 '17 at 17:43
  • $\ln\sqrt{x}<\sqrt{x}\iff \frac{1}{2}\ln x<\sqrt{x}$ – Stu Nov 10 '17 at 17:43
  • @Daniel Fischer I’d like to evaluate $ln n$ with a polynomial function and not to use $\ln n$ definition. – John Doe Nov 10 '17 at 17:45
  • @Jack I’ve finally understood what I want to show. I want to prove by induction that $\ln n < \sqrt n$ for all positive integers. So I need to prove that if $\ln n < \sqrt n$ then it follows that $\ln (n+1) < \sqrt {n+1}$. And I need your help with this. – John Doe Nov 10 '17 at 21:06
  • 1
    @JohnDoe To avoid confusion, I suggest to create new question asking just for $\ln n < \sqrt{n}$ using induction. However notice there are many proofs of this on this site using calculus, so either use one of them, or mention in the question that you dont want calculus to be used... If you post such question, don't forget to mention what you have tried. – Sil Nov 11 '17 at 07:44
  • Why do you need $\log n <\sqrt{n} $ when the job is easily done by $\log n<2\sqrt{n}$ as mentioned by @Stu? – Paramanand Singh Nov 11 '17 at 08:06
  • Also you should specify the specific definition of $\log n$? The solution to the question depends on the definition of $\log$ function. – Paramanand Singh Nov 11 '17 at 08:17
  • @Paramanand Singh $ \ln \sqrt n < \sqrt n$ implies that I know that $ \ln n$ (or $\ln \sqrt n$) has a slower growth than $n$ (or $\sqrt n$) and I want to show it by induction. So I want to show that if $\ln n < \sqrt n$, then it follows that $\ln (n+1) < \sqrt {n+1}$. – John Doe Nov 11 '17 at 09:33
  • 1
    Showing $\log n<2\sqrt{n}$ is easier and it shows that $\log$ has a slower growth rate than square root. So I am wondering why you are not going the easy way. – Paramanand Singh Nov 11 '17 at 11:10
  • @Paramanand Singh I have nothing against it, but how can I prove this fact by induction method. That’s what I need now. To show that logarithm is less than square root of any positive integer.:) It can be $ 2 \sqrt n$, it doesn’t matter. – John Doe Nov 11 '17 at 14:42
  • 2
    JohnDoe #3091 If you have another question, or want to do what is suggested in comments, (and your most recent edit), ask it in a separate question, please. – amWhy Nov 11 '17 at 15:22

3 Answers3

1

The two properties of $\log$ that we will use are:

  • $\log a>0$ when $a>1$.
  • $\log(ab)=\log(a)+\log(b)$ when $a,b>0$.

From these, you get:

  1. If $0<a<b$ then $\log a<\log b.$
  2. When $0<a$ and $x$ is rational, we have $\log\left(a^x\right)=x\log a$.

We will show that $n^{1/n}$ is decreasing for $n\geq 3$ and (obviously) bounded below by $1$.

From this (and our assumptions about $\log$) we see that $a_n=\log(n^{1/n})=\frac{\log n}{n}$ is decreasing and bounded below, and hence must converge.

But then the sequence $a_{n^2}$ must converge to the same value, and $a_{n^2}=\frac{2}{n}a_n$. So $a_{n^2}\to 0$, and hence $a_n\to 0$.

Now, we just need to prove that $(n+1)^{1/(n+1)}\leq n^{1/n}$ when $n\geq 3$.

This is equivalent to:

$$n^{n+1}\geq (n+1)^n$$

or:

$$n\geq \left(1+\frac{1}{n}\right)^n=\sum_{k=0}^{n}\binom{n}{k}n^{-k}$$

Now, each of the terms in the sum is know to be $\leq 1$. But you also have that, when $n\geq 3$, that the last two terms $(k=n-1,n)$ add to:

$$\frac{n^2+1}{n^n}$$

which is less than $1$ for $n\geq 3$. So we have:

$$\left(1+\frac{1}{n}\right)^n\leq 1+\sum_{k=0}^{n-2}\binom{n}{k}n^{-k}\leq n,$$

which proves what we needed.


Aside: There is a stronger inequality which comes from $\binom{n}{k}\leq \frac{n^k}{2^{k-1}}$ for $k\geq 1$, giving that $\left(1+\frac1n\right)^n\leq 3.$


An alternative approach is to show that there is a constant $C$ such that for all $n$: $C\leq \sqrt{n}-\log(n)$. From this, we deduce that: $$\frac{\log n}{\sqrt{n}}\leq 1-\frac{C}{\sqrt{n}}$$ and therefore $$0\leq \frac{\log n}{n}\leq \frac{1}{\sqrt{n}}-\frac{C}{n}$$ and thus, by the squeeze theorem, $\frac{\log n}{n}\to 0$.

To prove this, you need the inequality $\log(1+x)\leq x$. Then you can prove by induction that for $n\geq 5$ that $\sqrt{n}-\log(n)<\sqrt{n+1}-\log(n+1)$. Thus, $C$ is the minimum of the values for $n=1,2,3,4,5$.

Thomas Andrews
  • 177,126
0

With $\varphi(x)=\dfrac{\ln x}{x^{1/2}}$ for $x>0$, $\varphi'(x)<0$ for all $x>e^{2}$, so $\varphi(x)\leq\varphi(e^{2})=\text{constant}$, so $\dfrac{\ln x}{x}\leq\varphi(e^{2})\dfrac{1}{x^{1/2}}$, now take $x\rightarrow\infty$.

user284331
  • 55,591
0

$y_n: =n^{1/n}$,

Find $\lim_{n \rightarrow \infty } log y_n$.

Step 1:

$(1+x)^n >(n^2/4)x^2,$ $x\ge 0$, $n\ge 2,$

where $x$ real, and $n$ a positive integer.

Proof:

Binomial expansion :

$(1+x)^n > 1 + nx + (n(n-1)/2)x^2+...>$

$(n(n-1)/2)x^2 \ge (n^2/4)x^2 ,$

for $n\ge 2.$

(Since $2(n-1) \ge n$).

Step 2:

$n^{1/n} \le 1+ 2/√n $, $ n \ge 2$

Proof:

Set $x= 2/√n.$

$(1+2/√n)^n \gt (n^2/4 )(2/√n)^2;$

$(1+2/√n) \gt n^{1/n} \gt 1.$

(Note: $n > 1$, then $n^{1/n} \gt 1$)

Step 3:

$1\le \lim_{n \rightarrow \infty} n^{1/n} \le$

$\lim_{n \rightarrow \infty} (1 + 1/√n) =1$.

Step 4:

$y_n =n^{1/n}.$

$\lim_{n \rightarrow \infty} \log(y_n) =0,$

since log is continuos.

Peter Szilas
  • 20,344
  • 2
  • 17
  • 28