The Fibonacci recurrence $F_n=F_{n-1}+F_{n-2}$ allows values for all indices $n\in\mathbb{Z}$. There is an almost endless list of properties of these numbers in all sorts of ways. The below question might even be known. Yet, if true, I like to ask for alternative proofs.

Question. Does the following identity hold? We have $$\frac{\sum_{k=0}^{\infty}\frac{F_{n+k}}{k!}}{\sum_{k=0}^{\infty}\frac{F_{n-k}}{k!}}=e,$$ independent of $n\in\mathbb{Z}$.

  • Perhaps this is also true with F replaced by G, where G is a Fibonacci recurrence but G_0 and G_1 are two arbitrary numbers, excluding 0,0. Gerhard "Can't Build E From Nothing" Paseman, 2017.04.12. – Gerhard Paseman Apr 12 '17 at 21:04
  • 7
    Can't you just do the sums using the Binet formula? – Aaron Bergman Apr 12 '17 at 21:05
  • As I mentioned above, it would be nice to see alternative or novel proofs. – T. Amdeberhan Apr 12 '17 at 21:19
  • @GerhardPaseman I suspect your comment is true, and can be proved from the special cases $n=0,1$ of the OP's assertion by taking linear combinations. – Greg Martin Apr 13 '17 at 17:07

It follows from the identity

$$F_{n+k} = \sum_{j=0}^k {k \choose j} F_{n-j}$$

which is obtained by applying the standard recurrence $k$ times to the left side, each time splitting up each term into two terms.

Indeed, this gives

$$\sum_{k=0}^{\infty} \frac{F_{n+k}}{k!} = \sum_{k=0}^{\infty} \frac{ \sum_{j=0}^k {k \choose j} F_{n-j}}{k!}= \sum_{k=0}^{\infty} \sum_{j=0}^{k} \frac{1}{j! (k-j)!} F_{n-j} = \sum_{j=0}^{\infty} \sum_{k-j=0}^{\infty} \frac{1}{(k-j)!} \frac{F_{n-j}}{j!} = \sum_{j=0}^{\infty} e \frac{F_{n-j}}{j!}$$

  • 3
    OoOooO that is slick! – user78249 Apr 12 '17 at 22:21
  • 2
    This is cool, too. – T. Amdeberhan Apr 12 '17 at 22:38
  • It is not an answer, It's a Golden answer! – Amin235 Apr 18 '17 at 21:44

Mathematica tells me it's a consequence of the two series (distinguished by $\pm$): $$\sum_{k=0}^\infty\frac{F_{n\pm k}}{k!}=\frac{e^{\sqrt{5}} \phi^n-(1-\phi)^n}{\sqrt{5}\, \exp(\phi^{\mp 1})},$$ with $\phi$ the golden ratio (the positive solution of $1+1/\phi=\phi$). So the ratio of the $\pm$ series equals $e^{\phi-1/\phi}=e$, and the entire $n$-dependence drops out of the ratio.

The identity also holds for $n\in\mathbb{R}$, with the usual generalisation $F_z=(\phi^z-\phi^{-z}\cos\pi z)/\sqrt{5}$ of the Fibonacci numbers to real argument $z$.

  • 1
    This is cool, indeed. – T. Amdeberhan Apr 12 '17 at 22:37
  • 1
    As I suspected (and Greg Martin commented), this extends to general nontrivial Fibonacci recurrences with $G_0=a$ and $G_1=b$. Writing the RHS above as $f(n)$, the corresponding value using $G$ is $af(-1) + bf(0)$ for the positive version, and similarly for the negative. Again cancellation occurs leaving a power of $e$. Gerhard "Or Use General Linear Nonsense" Paseman, 2017.04.13. – Gerhard Paseman Apr 13 '17 at 18:04

More generally, $$\sum_{k=0}^\infty F_{n+k} \frac{x^k}{k!} = e^x\sum_{k=0}^\infty F_{n-k}\frac{x^k}{k!},\tag{1}$$ which is equivalent to Will Sawin's identity.

Similarly, $$e^x\sum_{k=0}^\infty F_{n+k}\frac{x^k}{k!}= \sum_{k=0}^\infty F_{n+2k}\frac{x^k}{k!},\tag{2}$$ and more generally, from the formula $$(-1)^q F_{p-q}+F_{\!q}\,\phi^p = F_{\!p}\,\phi^q,$$ where $\phi = (1+\sqrt5)/2$, and the analogous formula with $\phi$ replaced by $(1-\sqrt5)/2$, we get for any integers $p$, $q$, and $n$, $$ e^{(-1)^q F_{p-q}\ x}\sum_{k=0}^\infty F_{q}^k F_{n+pk}\frac{x^k}{k!} = \sum_{k=0}^\infty F_p^k F_{n+qk} \frac{x^k}{k!}. $$ The cases $p=1, q=-1$ and $p=1,q=2$ give $(1)$ and $(2)$.

Related identities (also involving Lucas numbers) can be found in L. Carlitz and H. H. Ferns, Some Fibonacci and Lucas Identities, Fibonacci Quarterly 8 (1970), 61–73.

Your Answer


By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.