5

Let $(a_n)_{n \in \mathbb{N}}$ be the sequence of integers defined recursively by $a_0 = 0$, $a_1 = 1, a_{n+2} = 4a_{n+1}+a_{n}$ for $n \geq 0$. Find the common divisors of $a_{1986}$ and $a_{6891}$.

I think it is true that $\gcd(a_m,a_n) = a_{\gcd(m,n)}$, but I am not sure how to prove it. From this the answer follows since $\gcd(1986,6891) = 3$ and so $\gcd(a_{1986,6891}) = a_3 = 17$.

user19405892
  • 15,592
  • 1
    Just pointing out that this reminded me of a putnam problem from last year: A2 2015. Anyway, this is a nice question +1 – Andres Mejia Jul 22 '16 at 00:26

2 Answers2

4

Your $\,a_i\,$ satisfy the same addition law as the Fibonacci numbers (with the same easy proof as as in this answer using matrix multiplication). Therefore the short proof I gave in this answer shows that this sequence too is a strong divisibility sequence, i.e. $\, (a_m,a_n) = a_{\large (m,n)}\,$ which immediately yields the sought result.

Remark $\ $ Responding to comments, below are further details. As in the first link we have

$$ \begin{bmatrix}a_2 &a_1\\ a_1 & a_0\end{bmatrix} = \begin{bmatrix}4 &1\\ 1 & 0\end{bmatrix},\quad \begin{bmatrix} a_{n+2} &\!\!\! a_{n+1}\\ a_{n+1} & \!\!\!a_n\end{bmatrix} = \begin{bmatrix} a_{n+1} &\!\! a_{n}\\ a_{n} & \!\!\!\!a_{n-1}\end{bmatrix} \begin{bmatrix}4 &1\\ 1 & 0\end{bmatrix}$$

Thus we infer by induction

$$ A_n := \begin{bmatrix} a_{n+1} &\!\! a_{n}\\ a_{n} &\!\!\! a_{n-1}\end{bmatrix} = \begin{bmatrix}4 &1\\ 1 & 0\end{bmatrix}^n\! =\, A_1^n $$

Therefore we deduce that $\,A_{m+n} = A_1^{m+n} = A_1^m A_1^n = A_m A_n,\ $ i.e.

$$ \begin{align} \begin{bmatrix} a_{m+n+1} &\!\! a_{m+n}\\ a_{m+n} &\!\!\!\! a_{m+n-1}\end{bmatrix} &= \begin{bmatrix} a_{m+1} &\!\! a_{m}\\ \!a_{m} & \!\!\!\!a_{m-1}\end{bmatrix} \begin{bmatrix} a_{n+1} &\!\! a_{n}\\ a_{n} & \!\!\!\!a_{n-1}\end{bmatrix}\\[.5em] &= \begin{bmatrix}a_{m+1}a_{n+1}+a_m a_n &\! a_{m+1}a_n+a_m a_{n-1}\\ a_m a_{n+1}+a_{m-1}a_n &\! a_m a_n + a_{m-1} a_{n-1} \end{bmatrix}\\ \end{align} $$

This yields the addition law $\ a_{m+n} =\, a_{m+1} a_n +a_m a_{n-1}.\ $ For example

$$ \begin{align} \begin{bmatrix}a_8 & a_7\\ a_7 & a_6\end{bmatrix} &= \begin{bmatrix}a_4 & a_3\\ a_3 & a_2\end{bmatrix} \begin{bmatrix}a_5 & a_4\\ a_4 & a_3\end{bmatrix}\\[.4em] &= \begin{bmatrix}72 & \!17\\ 17 &\! 4\end{bmatrix} \begin{bmatrix}305 &\!\!\!\!72\\ 72 &\!\!\!\!\! 17\end{bmatrix} = \begin{bmatrix}23184 &\!\!\! 5473\\ 5473 &\!\!\! 1292\end{bmatrix} \end{align} $$


For $\,f_n = \dfrac{x^n-y^n}{x-y},\,$ satisfying $\,f_{n+2} = \overbrace{(x\!+\!y)}^{f_2} f_{n+1}-xy\, f_n,\,$ a very similar proof shows that it has an analogous addition law: $\, f_{n+m}\ =\ f_m\ f_{n+1}\,\ -\ xy\, f_{m-1} f_n,\ $ i.e.

$$ \dfrac{x^{n+m}\!-y^{n+m}}{x-y}\,=\, \dfrac{x^{m}\!-y^{m}}{x-y}\,\dfrac{x^{n+1}\!-y^{n+1}}{x-y} - xy\,\dfrac{x^{m-1}-y^{m-1}}{x-y}\,\dfrac{x^{n}-y^{n}}{x-y}\qquad$$

To help dispel doubts in the comments, here is an Alpha verification of the prior equation.

Bill Dubuque
  • 272,048
0

Using the characteristic polynomial (you might prove the following by induction) we get that $x^{n+2}=4x^{n+1}+x^n$, implying that $$a_n=\left(2+\sqrt{5}\right)^{n}+\left(2-\sqrt{5}\right)^n$$ so we can denote this with $\alpha^n+\beta^n.$

We first let $k$ be a nonnegative odd integer, and note that $a_n>0$, and that $\frac{a_kn}{a_n}$ is rational. Then by expanding, we obtain that

$$\frac{a_{kn}}{a_n}=\alpha^{(k-1)n}-\alpha^{(k-2)n}\beta^n+...+\beta^{(k-1)n}$$

which means that $\frac{a_{kn}}{a_n}$ is of the form $a+b\sqrt{5}$. Recalling that it is indeed rational, we get that it must be an integer, so $a_n \vert a_{kn}$.

In the case that $k$ is even,

$$\frac{a_{kn}}{a_n}=\frac{\alpha^{kn}+\beta^{kn}}{\alpha^n+\beta^n}=\sum_{i=0}^{k-1}(-1)^i\cdot\alpha^{k-1-i}\cdot\beta^k+2\beta^k$$

which is again of the form $a+b\sqrt{5}$ so we get a similar argument here.

So, we get that if $m \vert n$, then $a_m \vert a_n$.

J. W. Tanner
  • 60,406
Andres Mejia
  • 20,977
  • 1
    I don't see how "if $m \mid n$, then $a_m \mid a_n$" proves the result. – user19405892 Jul 22 '16 at 01:11
  • It doesnt, it is strictly weaker. You would also need the addition property mentioned in the other answer to prove your last statement. Otherwise, im not certain about how to show that there are no other common divisors. – Andres Mejia Jul 22 '16 at 02:12