1
$\begingroup$

So I was trying to find a series expansion of $\Gamma(x+1)$ (which is the analytic continuation of $x!$ when I bumped into this sum $$\lambda(x)=\sum_{n=1}^\infty\left(\frac xn-\ln\left(1+\frac xn\right)\right)$$According to Desmos, the derivative of this function is the harmonic series $H_x$. The analytic continuation of it is $$\int_0^1\frac{t^x-1}{t-1}dt$$So I want to show that $$\lambda'(x)=H_x$$So I start like this: $$\frac{d}{dx}\sum_{n=1}^\infty\left(\frac xn-\ln\left(1+\frac xn\right)\right)=\lambda'(x)$$Then we switch the derivative and sum symbol: $$\sum_{n=1}^\infty\frac{d}{dx}\left(\frac xn-\ln\left(1+\frac xn\right)\right)$$And then everything is straightforward until we arrive at $$\sum_{n=1}^\infty\left(\frac1n-\frac1{n+x}\right)$$Which could be proven that it telescopes. So did I do everything correctly so far? I am a bit suspicious about switching the summation symbol and derivative symbol at the beginning. But numerically this is correct.

And if you are wondering what the series expansion of $\Gamma(x+1)$ looks like, then here it is:$$\Gamma(x+1)=\sum_{n\ge0}\frac{\left(-\gamma x+\sum_{k=1}^\infty\frac xk-\ln\left(1+\frac xk\right)\right)^n}{n!}$$ Where $\gamma$ is Euler's constant. An interesting note is that $$\lambda(1)=\gamma$$

$\endgroup$
7
  • $\begingroup$ You may be interested in the Weierstrass factorisation of the $\Gamma$ function. $\endgroup$
    – Somos
    Commented Mar 6, 2023 at 3:08
  • $\begingroup$ @Somos That's how I made the series expansion. I started using it and then took the natural logarithm of both sides, then taking the exponential function of both sides and using its series expansion to get what I wrote above. $\endgroup$ Commented Mar 6, 2023 at 4:42
  • 1
    $\begingroup$ dlmf.nist.gov/5.7.E6 $\endgroup$
    – Gary
    Commented Mar 6, 2023 at 6:03
  • 1
    $\begingroup$ Informally, note that $\sum_{n=1}^\infty\left(\frac{1}{n}-\frac{1}{n+x}\right)=\sum_{n=1}^\infty\frac{1}{n}-\sum_{n=x+1}^\infty\frac{1}{n}=\sum_{n=1}^x\frac{1}{n}=H_x$. This is not a viable proof but at least lets you see the connection. $\endgroup$ Commented Mar 6, 2023 at 11:46
  • $\begingroup$ @Aaron Hendrickson The sum $\sum_{n=1}^\infty \frac{1}{n}$ is a divergent series. To avoid this issue , insert the limit as i did in my solution below. $\endgroup$ Commented Mar 6, 2023 at 22:53

2 Answers 2

2
$\begingroup$

$$\frac{d}{dx}\sum_{n=1}^\infty\left(\frac xn-\ln\left(1+\frac xn\right)\right)=\sum_{n=1}^\infty\left(\frac{1}{n}-\frac{1}{n+x}\right)$$

$$=\lim_{N\to \infty}\sum_{n=1}^N\left(\frac{1}{n}-\frac{1}{n+x}\right)$$

$$=\lim_{N\to \infty}\left[\sum_{n=1}^N\frac{1}{n}-\sum_{n=1}^N\frac{1}{n+x}\right]$$

$$=\lim_{N\to \infty}\left[\sum_{n=1}^N\frac{1}{n}-\left(\underbrace{\sum_{n=1}^{N-x}\frac{1}{n+x}}_{n\to n-x}+\underbrace{\sum_{n=N-x+1}^{N}\frac{1}{n+x}}_{n\to N-n}\right)\right]$$

$$=\lim_{N\to \infty}\left[\sum_{n=1}^N\frac{1}{n}-\sum_{n=x+1}^{N}\frac{1}{n}-\sum_{n=0}^{x-1}\frac{1}{N+x-n}\right]$$

$$=\lim_{N\to \infty}\left(\sum_{n=1}^N\frac{1}{n}-\sum_{n=x+1}^{N}\frac{1}{n}\right)-\lim_{N\to \infty}\sum_{n=0}^{x-1}\frac{1}{N+x-n}$$

$$=\lim_{N\to \infty}\left(\sum_{n=1}^x\frac{1}{n}\right)-\lim_{N\to \infty}\sum_{n=0}^{x-1}\frac{1}{N+x-n}$$

$$=\sum_{n=1}^x\frac{1}{n}-0$$

$$=H_x$$

$\endgroup$
1
  • 1
    $\begingroup$ Ali, check out the book "Summability Calculus". There is a preprint in arXiv (arxiv.org/abs/1209.5739). I bet there is an alternative proof in there...and you might find the book to be interesting. $\endgroup$ Commented Mar 6, 2023 at 23:17
2
$\begingroup$

Hoping that it not too advanced for the time being, consider $$A_p(x)=\sum_{n=1}^\infty\left(\frac xn-\log\left(1+\frac xn\right)\right)$$ $$A_p(x)=x H_p+\log (\Gamma (p+1))-\left(\zeta ^{(1,0)}(0,p+x+1)-\zeta ^{(1,0)}(0,x+1)\right)$$ $$\lambda(x)=\underset{p\to \infty }{\text{limit}}A_p(x)=\gamma x+\log (\Gamma (x+1))$$ $$\lambda'(x)=\gamma+\psi ^{(0)}(x+1)=H_x$$

Edit

Otherwise, $$\frac xn-\log\left(1+\frac xn\right)=\sum_{k=2}^\infty (-1)^k\,\frac{ x^k}{k\ n^k}$$ Assuming that we can intergchange the order of summation $$\sum_{n=1}^\infty (-1)^k\,\frac{ x^k}{k\ n^k}=(-1)^k\,\frac{ x^k \zeta (k)}{k}$$ $$\sum_{k=1}^\infty (-1)^k\,\frac{ x^k \zeta (k)}{k}=\gamma x+\log (\Gamma (x+1))$$

If you differentiate deirectly $$\frac d {dx} A_p(x)=\sum_{n=1}^p\left(\frac 1n- \frac{1}{n+x}\right)=H_p+H_x-H_{p+x}$$ Using the asymptotics of harmonic numbers $$H_q=\log (q)+\gamma +\frac{1}{2 q}-\frac{1}{12 q^2}+O\left(\frac{1}{q^4}\right)$$ use it twice and continue with Taylor series to obtain $$\frac d {dx} A_p(x)=H_x-\frac{x}{p}+\frac{x (x+1)}{2 p^2}+O\left(\frac{1}{p^3}\right)$$

$\endgroup$
4
  • $\begingroup$ @KamalSaleh $\zeta^{(1,0)}(0,z)$ means $\frac{\mathrm d}{\mathrm d w}\zeta(w,z)|_{w=0}$ $\endgroup$ Commented Mar 6, 2023 at 19:04
  • $\begingroup$ @AaronHendrickson And $\zeta(w, z)$ is the hurwitz zeta function, correct? $\endgroup$ Commented Mar 6, 2023 at 19:08
  • $\begingroup$ @KamalSaleh Yes. Hurwitz zeta. $\endgroup$ Commented Mar 6, 2023 at 20:11
  • 1
    $\begingroup$ @KamalSaleh $\sum_{n=1}^p(x/n-\log(1+x/n))=-\zeta ^{(1,0)}(0,p+x+1)+\zeta ^{(1,0)}(0,x+1)+x H_p+\log (\Gamma (p+1))-\frac{1}{2} \log (2 \pi )+\frac{\log (\pi )}{2}+\frac{\log (2)}{2}$ $\endgroup$ Commented Mar 6, 2023 at 20:34

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .