4
$\begingroup$

The Fibonacci and Lucas numbers are defined for all integers $n$ by the recurrence relations $$F_n=F_{n-1}+F_{n-2}\text{ where }F_1=1\text{ and }F_2=1,$$ $$L_n=L_{n-1}+L_{n-2}\text{ where }L_1=1\text{ and }L_2=3.$$ I would like to simplify expressions of the form $$F_n^2-F_m^2.$$ By "simplify" I mean write as some product of Fibonacci or Lucas numbers, or a quotient of said products. When $n-m=2k$ is even this is easy to do using the formulae: \begin{align*} F_{n+k}+F_{n-k}=F_nL_k\text{ where $k$ is even,}\\ F_{n+k}+F_{n-k}=L_nF_k\text{ where $k$ is odd.}\\ \end{align*} Therefore, \begin{align*} F_n^2-F_m^2&=(F_n-F_m)(F_n+F_m)\\ &=(F_{m+2k}-F_m)(F_{m+2k}+F_m)\\ &=(F_{(m+k)-k}-F_{(m+k)-k})(F_{(m+k)+k}+F_{(m+k)-k})\\ &=L_{m+k}F_{m+k}L_kF_k\\ \end{align*} This trick obviously doesn't work when $n-m$ is odd. Does anyone have any suggestions as to how to find a similar expression in this instance?

$\endgroup$
  • $\begingroup$ Are you aware of math.stackexchange.com/q/41102 ? $\endgroup$ – Jean Marie Mar 17 '16 at 11:30
  • $\begingroup$ I had a look at the page but cannot see how it's related, unless you mean to suggest having a look at the book referenced on that page...? $\endgroup$ – Auslander Mar 17 '16 at 12:18
  • $\begingroup$ You are right. I thought to another reference, but I don't find it back. $\endgroup$ – Jean Marie Mar 17 '16 at 12:27
1
$\begingroup$

I want to point out that, in the case $n-m$ is even, then you can also write $$F_n^2-F_m^2 = F_{n+m}\cdot F_{n-m}.$$

Which probably can be obtained easily from Binet expression for Fibonacci numbers.

For the case $n-m$ odd, I have only the following negative result:

In general, the expression $F_n^2-F_m^2$ when $n-m$ is odd cannot be written as a product of $8$ (or less) Fibonacci and/or Lucas numbers.

Since this is a negative result, I established it simply by computing the expression for various values of $n$ and $m$ and begin unable to decompose it as the product of Fibonacci and/or Lucas number.

The only non trivial result I found empirically (I'm sure it can be proved easily) is the following, which holds only for $n=4k+2$, and $m=n/2=2k+1$ (hence $n-m$ is odd): $$ F_n^2-F_{n/2}^2 = F_5\cdot F_{n/2-1}\cdot F_{n/2}\cdot F_{n/2} \cdot F_{n/2+1} $$

$\endgroup$
  • $\begingroup$ I'd also be happy with a quotient of products were it possible. I'll adjust the question to reflect this. Your efforts are appreciated, as is your example. $\endgroup$ – Auslander Mar 17 '16 at 12:32
  • 1
    $\begingroup$ Clearly, ruling out quotients experimentally is much more difficult (maybe impossible). Anyway, it seems that if a general expression involving quotients do exist, then the indices of the Fibonacci or Lucas numbers involved must grow much faster than $n$ and $m$. For example, for $n\le 10^6$ there is no $F_n$ or $L_n$ which is divisible by one of the prime factors of $F_{36}^2-F_{11}^2=14930263\cdot 14930441$. $\endgroup$ – Giovanni Resta Mar 17 '16 at 13:00

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.