4
$\begingroup$

The Euler-Mascheroni constant gamma is defined as:

$$\gamma=\lim\limits_{n \rightarrow \infty}\left(\sum\limits_{m=1}^{n} \frac{1}{m} - \log(n)\right)$$

From this previous question Do these series converge to logarithms? $\log(n)$ can be written:

$$\log(n)=\sum\limits_{k=1}^\infty \sum\limits_{a=1}^{n-1}\frac{1}{kn-a}-\sum\limits_{k=1}^\infty\frac{n-1}{kn}$$

$$\gamma=\lim\limits_{n \rightarrow \infty}\left(\sum\limits_{m=1}^{n} \frac{1}{m} - \left(\sum\limits_{k=1}^\infty \sum\limits_{a=1}^{n-1}\frac{1}{kn-a}-\sum\limits_{k=1}^\infty\frac{n-1}{kn}\right)\right)$$

$$\gamma=\lim\limits_{n \rightarrow \infty}\left(\sum\limits_{m=1}^{n} \frac{1}{m} - \sum\limits_{k=1}^\infty \sum\limits_{a=1}^{n-1}\frac{1}{kn-a} + \sum\limits_{k=1}^\infty\frac{n-1}{kn}\right)$$

Can this last expression be further simplified?

$\endgroup$
  • 1
    $\begingroup$ Perhaps it is not good to have a divergent series like $\sum\limits_{k=1}^\infty\frac{n-1}{kn}$ inside your formula. $\endgroup$ – GEdgar Jun 21 '11 at 15:17
  • $\begingroup$ That is, you want $$\log(n)=\sum\limits_{k=1}^\infty \left[\sum\limits_{a=1}^{n-1}\frac{1}{kn-a}-\frac{n-1}{kn}\right]$$ not two separate sums. $\endgroup$ – GEdgar Jun 21 '11 at 15:32
  • $\begingroup$ @GEdgar: That is probably right. $\endgroup$ – Mats Granvik Jun 21 '11 at 16:28
9
$\begingroup$

Careful, we don't actually have

$$\log(n)=\sum\limits _{k=1}^{\infty}\sum\limits _{a=1}^{n-1}\frac{1}{kn-a}-\sum\limits _{k=1}^{\infty}\frac{n-1}{kn},$$

since neither series converges. Instead we have

$$\log(n)=\lim_{M\rightarrow\infty}\sum\limits _{k=1}^{M}\sum\limits _{a=1}^{n-1}\frac{1}{kn-a}-\sum\limits _{k=1}^{M}\frac{n-1}{kn}.$$

Now, since

$$\sum\limits _{k=1}^{M}\frac{n-1}{kn}=\sum\limits _{k=1}^{M}\frac{1}{k}-\sum\limits _{k=1}^{M}\frac{1}{kn}$$

we can rewrite

$$\log(n)=\lim_{M\rightarrow\infty}\sum\limits _{k=1}^{M}\sum\limits _{a=0}^{n-1}\frac{1}{kn-a}-\sum\limits _{k=1}^{M}\frac{1}{k}=\lim_{M\rightarrow\infty}\sum_{k=1}^{nM}\frac{1}{k}-\sum_{k=1}^{M}\frac{1}{k}.$$

Thus we have

$$\gamma=\lim_{n\rightarrow\infty}\lim_{M\rightarrow\infty}\left(\sum_{k=1}^{n}\frac{1}{k}-\sum_{k=1}^{nM}\frac{1}{k}+\sum_{k=1}^{M}\frac{1}{k}\right).$$

Personally, I quite like this limit since it has a nice symmetry when we switch the order of the limits. Also, it generalizes to give limits over l variables which are invariant under permutation. Let $H_{k}$ be the $k^{th}$ harmonic number. Then the above was

$$\gamma=\lim_{n\rightarrow\infty}\lim_{M\rightarrow\infty}\left(H_{n}-H_{nm}+H_{m}\right).$$

Here is $l=3$:

$$\gamma=\lim_{n_{1}\rightarrow\infty}\lim_{n_{2}\rightarrow\infty}\lim_{n_{3}\rightarrow\infty}\left(\left(H_{n_{1}}+H_{n_{2}}+H_{n_{3}}\right)-\left(H_{n_{1}n_{2}}+H_{n_{2}n_{3}}+H_{n_{3}n_{1}}\right)+\left(H_{n_{1}n_{2}n_{3}}\right)\right).$$

In general

$$\gamma=\lim_{n_{1}\rightarrow\infty}\cdots\lim_{n_{l}\rightarrow\infty}\left(\sum_{i_{1}=1}^{l}H_{n_{i_{1}}}-\sum_{i_{1}<i_{2}\leq l}H_{n_{i_{1}}n_{i_{2}}}+\cdots+(-1)^{l}H_{n_{1}\cdots n_{l}}\right).$$

So I guess it depends on what you mean by “simplify further.” I find the form

$$\gamma=\lim_{n\rightarrow\infty}\lim_{m\rightarrow\infty}\left(H_{n}-H_{nm}+H_{m}\right)$$

to be quite simple.

Hope that helps,

$\endgroup$
  • $\begingroup$ Wow thanks, that is a really nice looking expression in the middle of your answer after "Thus we have". If I understood it right, the use of a double limit is the way to make it computable. My choice of words was not so good. I should have left out the word "further". $\endgroup$ – Mats Granvik Jun 21 '11 at 16:20
  • $\begingroup$ Does:$$\lim_{n_1\to\infty}\lim_{n_2\to\infty}\lim_{n_3\to\infty}\frac12(H_{n_1}+H_{n_2}+H_{n_3}-H_{n_1n_2n_3})$$also equal $\gamma$? $\endgroup$ – Akiva Weinberger Jan 26 '16 at 20:21
  • $\begingroup$ Yeah, it does, because that's equal to:$$\lim_{n_1\to\infty}\lim_{n_2\to\infty}\lim_{n_3\to\infty}\\ \frac12\big((H_{n_1}-\ln n_1)+(H_{n_2}-\ln n_2)+(H_{n_3}-\ln n_3)-(H_{n_1n_2n_3}-\ln n_1n_2n_3)\big)$$which goes to $\frac12(\gamma+\gamma+\gamma-\gamma)=\gamma$. Same strategy proves the similar limits in the answer. $\endgroup$ – Akiva Weinberger Jan 26 '16 at 20:29
1
$\begingroup$

@Eric can you give the proof that this is equal (for every possible value of n): $$\ln(n)=\lim_{M\rightarrow\infty}\sum\limits _{k=1}^{M}\sum\limits _{a=1}^{n-1}\frac{1}{kn-a}-\sum\limits _{k=1}^{M}\frac{n-1}{kn}.$$

$\endgroup$
1
$\begingroup$

Setting $m=n$ in Naslund's equation we get

$$ \gamma= \sum_{n=1}^\infty \left(\frac{2}{n}-\sum_{j=n(n-1)+1}^{n(n+1)} \frac{1}{j}\right) $$

This may be used to write a closed form for the digamma function, see Re-Expressing the Digamma.

$\endgroup$

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.