Let $F_n$ denote the $n$-th Fibonacci number and $\phi$ be the golden ratio, that $\phi = \frac{1+\sqrt{5}}{2}$. Find a closed form for the sum:

$$\sum_{n=0}^{\infty} \frac{1}{(5\phi)^n(n+2)} \sum_{k=0}^{n} \frac{F_{k+1} F_{n-k+1}}{k+1}$$

I don't know how to tackle the sum. A closed form exists as the book suggests but there is no hind nor answer on how to tackle it. And I have not come up with some idea to kill this sum. Can anyone help me find a closed form?


By setting $\sigma=\frac{-1+\sqrt{5}}{2},\overline{\sigma}=\frac{-1-\sqrt{5}}{2}$ we have: $$ \sum_{k\geq 1}F_k\, x^k = \frac{x}{1-x-x^2} = \frac{-\sigma}{x-\sigma}+\frac{\overline{\sigma}}{x-\overline{\sigma}}\tag{1} $$ hence: $$ \sum_{k\geq 1}\frac{F_k}{k}\,x^k = \frac{1}{\sqrt{5}}\,\log\left(\frac{1+\sigma x}{1+\overline{\sigma}x}\right)\tag{2}$$ and: $$ \sum_{n\geq 0}x^{n+1}\sum_{h=0}^{n}\frac{F_{h+1}F_{n-h+1}}{h+1}=\frac{1}{\sqrt{5}(1-x-x^2)}\,\log\left(\frac{1+\sigma x}{1+\overline{\sigma}x}\right) \tag{3}$$ so: $$ \sum_{n\geq 0}\frac{t^{n+2}}{n+2}\sum_{h=0}^{n}\frac{F_{h+1}F_{n-h+1}}{h+1}=\int_{0}^{t}\frac{dt}{\sqrt{5}(1-x-x^2)}\,\log\left(\frac{1+\sigma x}{1+\overline{\sigma}x}\right) \tag{4}$$ and the problem boils down to finding the right $t$ and computing the integral in the RHS, through partial fraction decomposition and $\int\frac{\log x}{x}\,dx=\log^2 x$.

Can you take it from here? The final answer should be something like $\frac{5}{4} \left(3+\sqrt{5}\right) \log\left[\frac{2}{41} \left(13+\sqrt{5}\right)\right]^2$, but I have to check my computations.

  • $\begingroup$ Very nicely done! Your strategy is particularly appealing to me given my affection for logarithmic integrals. Any chance an integral of that type has already been solved here? $\endgroup$ – David H Aug 3 '15 at 16:54
  • $\begingroup$ @DavidH: probably it was, but I am not so good at fighting with the search system :) $\endgroup$ – Jack D'Aurizio Aug 3 '15 at 16:55
  • $\begingroup$ Wikipedia on 'Fibonacci' Sec. 4.1. $\endgroup$ – BruceET Aug 3 '15 at 16:57
  • 1
    $\begingroup$ @Masacroso: there is no EGF involved, in order to go from $(1)$ to $(2)$ you just need to integrate termwise $(1)$. $\endgroup$ – Jack D'Aurizio Aug 3 '15 at 23:51
  • 1
    $\begingroup$ @Tolaso: you just need to use the partial fraction decomposition appearing in $(1)$ to compute a primitive in terms of a squared logarithm. $\endgroup$ – Jack D'Aurizio Aug 4 '15 at 12:51

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.