3
$\begingroup$

I can't seem to use certain methods such as $\varepsilon$-N, L'Hôspital's Rule, Riemann Sums, Integral Test and Divergence Test Contrapositive or Euler's Integral Representation to prove that

$\lim_{n-> \infty} \frac{H_n}{n} = 0$ where $H_n$ is the nth Harmonic number $= \sum_{i=1}^{n} \frac{1}{i} = 1 + \frac{1}{2} + \frac{1}{3} + \frac{1}{4} + ... + \frac{1}{n}$?

I was able to prove it using the Monotone Convergence Theorem, and I think polylogarithms make this easy. I would like to know if any of these are correct/can be modified to be correct/even possible to use.

Here are my attempts:

1 $\varepsilon-\delta$

$\forall \varepsilon > 0, \exists N > 0$ s.t. $|\frac{H_n}{n} - 0| < \varepsilon$ whenever $n > N$

$|\frac{H_n}{n} - 0|$

$=|\frac{H_n}{n}|$

$ \leq |\frac{H_n}{n}|$

$ \leq \frac{|H_n|}{|n|}$

$ \leq \frac{|H_n|}{|n|} < \varepsilon$ if $n > N=\frac{H_n}{\epsilon}$

But I haven't really isolated n, so I don't think that's allowed?

2 L'Hôspital's Rule

$\lim_{n-> \infty} \frac{H_n}{n}$

$= \lim_{n-> \infty} \frac{\frac{\partial}{\partial n} H_n}{1}$

Is there a discrete version of the Fundamental Theorem of Calculus that allows us to evaluate $\frac{\partial}{\partial n} H_n$? Maybe it's this one, but I don't understand it. Does $\frac{\partial}{\partial n} H_n$ even have meaning?

Or is there some way to show that $\lim_{n-> \infty} \frac{H_n}{n} = \lim_{n-> \infty} \frac{1}{n}\int_{1}^{n} \frac{1}{x} dx$?

3 Riemann Sums

$\lim_{n-> \infty} \frac{H_n}{n}$

$= \lim_{n-> \infty} \frac{\sum_{i=1}^{n}\frac{1}{i}}{n}$

$= \lim_{n-> \infty} \frac{1}{n}\sum_{i=1}^{n}\frac{1}{i}$

$= \lim_{n-> \infty} \frac{1}{n}\sum_{i=1}^{n}\frac{1}{i/n}\frac{1}{n}$

$= \lim_{n-> \infty} \frac{1}{n} \lim_{n-> \infty} \sum_{i=1}^{n}\frac{1}{i/n}\frac{1}{n}$

$= \lim_{n-> \infty} \frac{1}{n} \int_{0}^{1} \frac{1}{x} dx$

$= \lim_{n-> \infty} \frac{1}{n} ln|x||_{0}^{1}$

$= \lim_{n-> \infty} \frac{1}{n} (ln|1| - ln|0|)$

$= \lim_{n-> \infty} \frac{1}{n} (- ln|0|)$

$= \lim_{n-> \infty} \frac{1}{n} (- ln|1/n|)$ if that's even allowed

$= \lim_{n-> \infty} \frac{1}{1} (- 1/(1/n)) \lim_{n-> \infty} \frac{-1}{n^{2}}$ by L'Hôspital's Rule

$= \lim_{n-> \infty} \frac{1}{n^{2}} (1/(1/n))$

$= \lim_{n-> \infty} \frac{1}{n^{2}} (n)$

$= \lim_{n-> \infty} \frac{1}{n} = 0$

4 Integral Test and Divergence Test Contrapositive

If the series $\sum_{n=1}^{\infty} \frac{H_n}{n}$ is convergent then $\lim_{n-> \infty} \frac{H_n}{n} = 0$

If the integral $\int_{1}^{\infty} \frac{H_x}{x} dx$ is convergent then the series $\sum_{n=1}^{\infty} \frac{H_n}{n}$ is convergent.

I don't know how to integrate $\int_{1}^{\infty} \frac{H_x}{x} dx$.

How about $\int_{1}^{\infty} \frac{\int_{1}^{x} \frac{1}{y} dy}{x} dx$ ? If that is convergent does that mean the series $\sum_{n=1}^{\infty} \frac{H_n}{n}$ is convergent? It doesn't seem to be convergent in the first place though:

$\int_{1}^{\infty} \frac{\int_{1}^{x} \frac{1}{y} dy}{x} dx$

$= \int_{1}^{\infty} \int_{1}^{x} \frac{1}{xy} dy dx$

$= \int_{1}^{\infty} \frac{ln|y|}{x}|_{y=1}^{y=x} dx$

$= \int_{1}^{\infty} \frac{ln|x|}{x} dx$

$= \int_{1}^{\infty} \frac{lnx}{x} dx$

$= (lnx)^{2}|_{1}^{\infty} = \infty$

5 Euler's Integral Representation

Euler proved that $H_n = \int_{0}^{1} \frac{1-x^{n}}{1-x} dx$

So, $\lim_{n-> \infty} \frac{H_n}{n}$

$= \lim_{n-> \infty} \frac{\int_{0}^{1} \frac{1-x^{n}}{1-x} dx}{n}$

$= \lim_{n-> \infty} \frac{\partial}{\partial n}\int_{0}^{1} \frac{1-x^{n}}{1-x} dx$ by L'Hôspital's Rule

$= \lim_{n-> \infty} \int_{0}^{1} \frac{\partial}{\partial n} \frac{1-x^{n}}{1-x} dx$

$= \lim_{n-> \infty} \int_{0}^{1} \frac{\partial}{\partial n} \frac{1-x^{n}}{1-x} dx$

$= \lim_{n-> \infty} \int_{0}^{1} \frac{\partial}{\partial n} \frac{-x^{n}}{1-x} dx$

$= \lim_{n-> \infty} \int_{0}^{1} \frac{-nx^{n-1}}{1-x} dx$

$= \lim_{n-> \infty} \int_{0}^{1} \frac{nx^{n-1}}{x-1} dx$

$= \lim_{n-> \infty} n \int_{0}^{1} \frac{x^{n-1}}{x-1} dx$

$= \lim_{n-> \infty} n \int_{0}^{1} \frac{x^{n-1} - 1}{x-1} + \frac{1}{x-1}dx$

$= \lim_{n-> \infty} n \int_{0}^{1} \sum_{i=0}^{n-2} x^{i}+ \frac{1}{x-1}dx$

It looks like I will get an ln(0) again...

$\endgroup$
  • $\begingroup$ It seems that this question and the one you voted to close as duplicates are both of high quality, I will later flag for moderator attention and see if they can merge the two. $\endgroup$ – user99914 May 15 '18 at 3:30
3
$\begingroup$

Although this is an old answered post, an unaddressed question is whether we can formulate a proof using the Cauchy definition of limit. The answer is yes. The proof relies on the observation that for every positive integer $n$, $$\sum_{k=2^{n-1}+1}^{2^n} \frac{1}{k} < \sum_{k=2^{n-1}+1}^{2^n} \frac{1}{2^{n-1}} = 1$$ which easily follows because $k > 2^{n-1}$ for every $k \in \{2^{n-1} + 1, \ldots, 2^n\}$. Consequently, $$H_{2^n} = 1 + \sum_{m=1}^n \sum_{k=2^{m-1} +1}^{2^m} \frac{1}{k} < 1 + \sum_{m=1}^n 1 = n+1.$$ Therefore, $$0 < \frac{H_{2^n}}{2^n} < \frac{n+1}{2^n} $$ This suggests that for sufficiently small $\epsilon$, the choice $N = 2^{1/\epsilon}$ will give the desired bound (the details of establishing the resulting inequality is left to the reader as an exercise).

$\endgroup$
6
$\begingroup$

The discrete version of L'Hôpital is Stolz: $$\lim_{n\to\infty}\frac{H_n}{n} = \lim_{n\to\infty}\frac{H_{n+1}-H_n}{(n+1)-n} = \lim_{n\to\infty}\frac{\frac{1}{n+1}}{1} = 0. $$

$\endgroup$
  • $\begingroup$ Thanks! Any idea though if any of my attempts will work? $\endgroup$ – BCLC Apr 13 '14 at 9:04
  • 1
    $\begingroup$ As $H_n/n>0$, only a upper bound is required. And $H_n-1=$lower Riemann sum$\le$integral$\le$ improper integral. $\endgroup$ – Martín-Blas Pérez Pinilla Apr 13 '14 at 17:38
  • $\begingroup$ @Martin-Blaz Perez Pinilla, are you referring to attempt #3? $\endgroup$ – BCLC Apr 25 '14 at 7:22
  • 1
    $\begingroup$ +1. It was nice you pointed out the H$\hat{\rm o}$pital-Stolz correlation. $\endgroup$ – Felix Marin Jul 12 '14 at 23:54
4
$\begingroup$

We have $$\ln(n+1)\le H_n\le\ln(n)+1\tag{1}$$

$$\frac{\ln(n+1)}n\le \frac{H_n}{n}\le\frac{\ln(n)+1}{n}$$

Both the right and left sides tend to zero as $n\to\infty$ using L'hospital. And therefore by the squeeze theorem, our limit is $0$


$$\int_1^{n+1} f(x)\,{\rm d}x \le \sum_{i=1}^nf(i)\le 1+\int_1^n f(x)\,{\rm d}x\text{ for monotonically decreasing $f$}$$

Here $f(x)=\frac{1}{x}$

$\endgroup$
  • $\begingroup$ Thanks! Any idea though if any of my attempts will work? $\endgroup$ – BCLC Apr 13 '14 at 9:05
4
$\begingroup$

This is almost obvious if one remembers the following elementary (but not so easy to prove) result:

If $a_{n} \to L$ as $n \to \infty$ then $\dfrac{a_{1} + a_{2} + \cdots + a_{n}}{n} \to L$ as $n \to \infty$.

Clearly here $a_{n} = 1/n \to 0$ as $n \to \infty$ and $H_{n} = a_{1} + a_{2} + \cdots + a_{n}$ so that $H_{n}/n \to 0$ as $n \to \infty$.

$\endgroup$
  • $\begingroup$ Thanks! Any idea though if any of my attempts will work? $\endgroup$ – BCLC Apr 13 '14 at 9:02

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.