On a lark (as a followup to this question), I was playing around with Wolfram alpha, and it seems that

$$\int_{0}^{1} \frac{dx}{1+{}_2F_{1}\left(\frac{1}{n},x;\frac{1}{n};\frac{1}{n}\right)} = \frac{\log\left(\frac{2n}{2n-1}\right)}{\log\left(\frac{n}{n-1}\right)}$$

Could someone take a stab at proving this?


First note that $${}_2F_{1}\left(\frac{1}{n},x;\frac{1}{n};\frac{1}{n}\right) = 1 + \frac{x}{n} + \frac{x(x+1)}{2! n^2} + \frac{x(x+1)(x+2)}{3! n^3} + \frac{x(x+1)(x+2)(x+3)}{4! n^4} + \cdots$$ $$ = 1 + (-x) \frac{-1}{n} + \frac{(-x)(-x-1)}{2!} \left( \frac{-1}{n} \right)^2 + \frac{(-x)(-x-1)(-x-2)}{3!} \left( \frac{-1}{n} \right)^3 + \cdots$$ $$ = \left(1 - \frac1{n} \right)^{-x}$$ Hence, $${}_2F_{1}\left(\frac{1}{n},x;\frac{1}{n};\frac{1}{n}\right) = \left(\frac{n}{n-1} \right)^x$$ Hence, $$\int_{0}^{1} \frac{dx}{1+{}_2F_{1}\left(\frac{1}{n},x;\frac{1}{n};\frac{1}{n}\right)} = \int_{0}^{1} \frac{dx}{1+\left(\frac{n}{n-1} \right)^x} = \int_{0}^{1} \frac{dx}{1+e^{kx}}$$ where $e^k = \frac{n}{n-1}$

Hence, we are interested in evaluating an integral of the form $$\int_0^1 \frac1{1+e^{kx}} dx$$ Let $t = 1 + e^{kx}$. We get $dt = k e^{kx} dx = k (t-1) dx$ $$\int_0^1 \frac1{1+e^{kx}} dx = \int_2^{1+e^k} \frac1{k} \frac{dt}{t(t-1)} = \frac1{k} \int_2^{1+e^k} \left(\frac{dt}{t-1} - \frac{dt}{t} \right) = \frac1{k} \left( k - \log(e^k+1) + \log(2) \right)$$

In our case, $e^k = \frac{n}{n-1}$ and hence the integral becomes $$1 - \frac{\log(2n-1) - \log(n-1)}{\log(n) - \log(n-1)} + \frac{\log(2)}{\log(n) - \log(n-1)} = \frac{\log \left( \frac{2n}{2n-1} \right)}{\log \left(\frac{n}{n-1} \right)}$$


To make GEdgar's solution more explicit: whenever you have equal numerator and denominator parameters in a hypergeometric function, you can cancel them:

$${}_2 F_1\left({{1/n}\atop{}}{{}\atop{1/n}}{{x}\atop{}}\mid \frac1{n}\right) = \sum_{j=0}^\infty \frac{(1/n)_j (x)_j}{(1/n)_j}\frac{(1/n)^j}{j!} = \sum_{j=0}^\infty (x)_j\frac{(1/n)^j}{j!} = {}_1 F_0\left({{x}\atop{—}}\mid \frac1{n}\right)$$

${}_1 F_0$ can subsequently be re-expressed as a binomial series by exploiting identities for converting Pochhammer symbols to binomial coefficients...

  • $\begingroup$ Meta-comment: shouldn't this question be tagged [special-functions]? $\endgroup$ – guren Jun 15 '11 at 4:53
  • $\begingroup$ You're right. I fixed it. $\endgroup$ – t.b. Jun 15 '11 at 5:14

${}_2F_1(1/n,x;1/n,1/n) = {}_1F_0(x;;1/n) = (1-1/n)^{-x}$


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.