1

The question is to find this infinite series:

$$ \sum_{r = 1}^{\infty} \frac{1}{r(r + \frac{1}{3})} = \; ? $$


I have tried applying the Riemann sum to the Integral technique as follows:

Assuming the lower and upper boundaries (a and b) are 0 and 1. Now, we know that

$$ \int_{0}^{1} {f(x)} \; dx = \lim_{n \to \infty} \frac{1}{n} \sum_{r = 1}^{n} f(\frac{r}{n}) $$

Therefore,

$$ \sum_{r = 1}^{\infty} \frac{1}{r(r + \frac{1}{3})} = \lim_{n \to \infty} \sum_{r = 1}^{n} \frac{1}{r(r + \frac{1}{3})} $$

$$ = \lim_{n \to \infty} \frac{1}{n} \sum_{r = 1}^{n} \frac{n}{r(r + \frac{1}{3})} = \lim_{n \to \infty} \frac{1}{n} \sum_{r = 1}^{n} \frac{n}{r} \frac{1}{({r+\frac{1}{3}})} $$

I'm stuck here.

Is there a different approach to this problem? Or, could this problem be solved this way?


Edit: Given in this comment below, does not answer my question. The answer is completely incomprehensible to me.

  • 1
    This sum doesn't really have anything to do with Riemann sums. I'm not sure what level of math you are comfortable with, but with complex analysis you can find the answer is $9-\frac{\sqrt{3}}{2}\pi-\frac{9}{2}\ln (3)$. https://www.wolframalpha.com/input?i=sum%281%2F%28r%28r%2B1%2F3%29%29%2Cr%2C1%2Cinfty%29 – QC_QAOA Jul 05 '22 at 02:31
  • @QC_QAOA I specifically used the Riemann sums because it is an infinite series, and after matching it to the $\frac{r}{n}$ format, we could reduce it to a definite integral, and then evaluate. – Rohan Bari Jul 05 '22 at 02:33
  • Riemann sums $\leftrightarrow$ require sums of functions of the form $f(r/n)$. Your sum is not of this form so it doesn't turn into a Riemann sum – QC_QAOA Jul 05 '22 at 02:35
  • @QC_QAOA Yes, that is where I'm stuck. It's not doable, seemingly. – Rohan Bari Jul 05 '22 at 02:36
  • @QC_QAOA I'm curious of how you apply directly Complex Analysis for this question. – Paresseux Nguyen Jul 05 '22 at 02:36
  • @zwim Does not seem so. The final answer approximates ~1.33. – Rohan Bari Jul 05 '22 at 02:37
  • 2
    @Parasseux_Nguyen The basic idea is to use the formula in the link below. You can transform sums over rational functions into residues of a related function of (that includes $\cosh(\pi z)$). With a little finagling you can solve any sum of rational functions this way. https://en.wikipedia.org/wiki/Residue_theorem#An_infinite_sum – QC_QAOA Jul 05 '22 at 02:39
  • @QC_QAOA : Ah ok. Got it. Nice approach. – Paresseux Nguyen Jul 05 '22 at 02:49

3 Answers3

8

Using partial fractions, we have $$\dfrac{1}{r(r+\tfrac{1}{3})} = \dfrac{3}{r}-\dfrac{3}{r+\tfrac{1}{3}}.$$ Then, by using the identity $\dfrac{1}{n} = \displaystyle\int_{0}^{1}x^{n-1}\,dx$, we have $$S:= \sum_{r = 1}^{\infty}\dfrac{1}{r(r+\tfrac{1}{3})} = \sum_{r = 1}^{\infty}\left[\dfrac{3}{r}-\dfrac{3}{r+\tfrac{1}{3}}\right] = 3\sum_{r = 1}^{\infty}\int_{0}^{1}\left(x^{r-1}-x^{r-2/3}\right)\,dx.$$ Because every term is positive, we can then interchange the order of the summation and integral. $$S = 3\int_{0}^{1}\sum_{r = 1}^{\infty}\left(x^{r-1}-x^{r-2/3}\right)\,dx.$$ The summation is of a geometric series which converges for $|x| < 1$ (i.e. almost everywhere on $[0,1]$), so we have $$S = 3\int_{0}^{1}\dfrac{1-x^{1/3}}{1-x}\,dx.$$ Now, just substitute $x = u^3$, $dx = 3u^2\,du$ and use a trigonometric substitution to finish.

JimmyK4542
  • 54,331
2

The logarithmic derivative of the Gamma function, a famous special function generalizing the factorial operations to complex inputs, is called the digamma function, $\psi$. For complex $z$, it can be shown (from the Weierstrass factorization of the gamma function) that $$\psi(z+1) = - \gamma + \sum_{n=1}^{\infty} \frac{z}{n(n+z)}$$ So note $$\sum_{n=1}^{\infty} \frac{z}{n(n+z)} = \psi(z+1)+\gamma$$ and so $$\sum_{n=1}^{\infty} \frac{\frac{1}{3}}{n(n+\frac{1}{3})} = \psi(\frac{4}{3})+\gamma$$ Thus $$\sum_{n=1}^{\infty} \frac{1}{n(n+\frac{1}{3})} = 3(\psi(\frac{4}{3})+\gamma)$$

  • Thank you very much for the answer. However, I'm not sure what those unique functions do right now. I will have a look into them sometime later. – Rohan Bari Jul 05 '22 at 03:18
0

Making the problem more general, for $a>0$, we can compute the partial sum $$S_n=\sum_{r=1}^n \frac 1{r(r+a)}$$ Using partial fraction decomposition $$S_n=\frac 1 a (-\psi (a+n+1)+\psi (a+1)+\psi (n+1)+\gamma )$$ and using the asymptotics $$S_n=\frac{\psi(a+1)+\gamma }a-\frac{1}{n}+\frac{ (a+1)}{2 n^2}-\frac{ (a+1) (2 a+1)}{6 n^3}+O\left(\frac{1}{n^4}\right)$$