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

What is the limit of $H_n - H_{n/2}$ for $n\to\infty$? $H_n$ is the nth Harmonic number.

And in general, what is the limit of $H_n - H_{a*n}$ for $n\to\infty$? $0 < a < 1$.

Summing these with Python gives the following results.

..............a=1/2...........a=1/3...........a=1/4...........a=1/5 n=1,000.......0.69364743056...1.10061495867...1.38779561112...1.61143991243 n=10,000......0.69319718306...1.09881231534...1.38644437362...1.60963793243 n=100,000.....0.693152180585..1.09863228893...1.38630936124...1.60945791263 n=1,000,000...0.69314768056...1.09861428867...1.38629586112...1.60943991244 n=10,000,000..0.69314723056...1.09861248867...1.38629451112...1.60943811243

These sums are getting smaller, and they can't go to zero, so they have limits. What would be their exact mathematical expressions?

Thank you.

share|cite|improve this question
For example, if n=1,000, a=1/2, then 1/500 + 1/501 + ... + 1/1,000. It would be H_1000 - H_499, where H_n is the nth Harmonic number. – Phillip Jan 25 '12 at 2:38
I'm not sure I understand your formula, but those sequences of numbers clearly converge to $\ln(2)$, $\ln(3)$, $\ln(4)$, $\ln(5)$, etc. – Byron Schmuland Jan 25 '12 at 2:42
Thank you very much! That answers my question. – Phillip Jan 25 '12 at 2:52
On page 82 of the January Monthly, try to find the limit as $n \rightarrow \infty$ of $$ 2 H_n - H_{n^2 + n} = 1 + \frac{1}{2} + \cdots + \frac{1}{n} - \frac{1}{n + 1 } - \cdots - \frac{1}{n + n^2 } \; ? $$ Evidently submitted by Juozas Juvencijus Macys. I changed it slightly. – Will Jagy Jan 25 '12 at 6:43
Is it not just $\gamma$ ? – Ragib Zaman Jan 25 '12 at 7:23
up vote 4 down vote accepted

For easiness, say $b=a^{-1}$. Then, the limit the OP is interested in is $ \lim{}_{n\rightarrow \infty }\sum_{k=n/b}^n \frac{1}{k}$ (assuming $b | n$ of course). We can change the sum to $$\sum_{k=0}^{n-n/b}\frac{b}{bk+n}=\sum_{k=0}^{n-n/b}\frac{1}{\frac{bk}{n}+1}\cdot\frac{b}{n}$$

This is a riemann sum where $dx = b/n$, and by taking limits we see that $$\lim_{n\rightarrow \infty}\sum_{k=0}^{n-n/b}\frac{1}{\frac{bk}{n}+1}\cdot\frac{b}{n} = \int_0^{b-1} \frac{1}{x+1}dx = \log(b)$$

there are a few math GRE subj problems like this ;)

share|cite|improve this answer
Thank you very much. – Phillip Jan 25 '12 at 3:59

Worth remembering is that $H_n=\log(n)+\gamma+o(1)$ where $\gamma=0.577\ldots$ is Euler-Mascheroni constant. Hence, for every $a\gt0$ and $b\gt0$, $$ H_{\lfloor an\rfloor}-H_{\lfloor bn\rfloor}=\log(an)+\gamma-\log(bn)-\gamma+o(1)=\log(a)-\log(b)+o(1), $$ which means exactly that $H_{\lfloor an\rfloor}-H_{\lfloor bn\rfloor}\to\log(a/b)$. In your case, $H_n-H_{\lfloor n/2\rfloor}\to\log(2)$.

A proof of the existence of $\gamma$ uses the expression $$H_n-\log(n)=\sum\limits_{k=1}^{n}u_k\quad\text{with}\quad u_k=\frac1k-\log\left(1+\frac1k\right). $$ Then the series expansion of the logarithm shows that $x-\frac12x^2\lt\log(1+x)\lt x$ for every $x$ in $(0,1)$, hence $0\lt u_k\lt\frac1{2k^2}$ for every $k\geqslant1$ and the series $\sum\limits_{k}u_k$ converges.

share|cite|improve this answer
Thanks a lot, I learnt something here. – Phillip Jan 28 '12 at 2:56
You are welcome. Unrelated: next time, you might wish to wait more than 97 minutes before accepting an answer. – Did Jan 28 '12 at 8:30

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.