|
The following problem was communicated to us by Ronald L. Graham. The sequence of numbers $\left\{a_{i}\right\}$ defined recursively in the abstract begins 1,1,1,1,2,3,7,23,59,314,1529,8209,83313,620297,7869898,126742987,1687054711,$\ldots$, and grows on the order of $\exp \left(n^{2}\right)$[1].
It had been observed that more than a hundred values of the sequence were integers and a proof that the entire sequence is integral was sought. We show in Theorem 1 that the sequence is indeed integral and discuss some generalizations.
Theorem 1. The sequence $\left\{a_{i}\right\}$ defined by the recurrence
$$\tag1
a_{n}=\frac{a_{n-3} a_{n-1}+a_{n-2}^{2}}{a_{n-4}} \text { for } n>3
$$
with initial values $a_{0}=a_{1}=a_{2}=a_{3}=1$ is integral.
Proof. To show the sequence consists only of integers, we assume that the sequence is integral up to and including eight consecutive members $B(0), B(1), B(2), \ldots, B(7)$. To show that $B(8)$ and hence the rest of the sequence is integral, we need to show that
$$
B(5) B(7)+B(6)^{2} \equiv 0\pmod{B(4)}
$$
We first show in Lemma 1 that $B(4)$ is relatively prime to $B(1) B(2)$, and then complete the proof by showing in Lemma 2 that $B(1) B(2)\left[B(5) B(7)+B(6)^{2}\right] \equiv 0\pmod{B(4)}$.
Lemma 1. The numbers $B(4)$ and $B(1) B(2)$ are relatively prime.
Proof. We first observe that as long as the sequence is integral, consecutive members are relatively prime. For if a prime $p \mid a_{n}$ and $p \mid a_{n-1}$, then from the recurrence (1), we also have $p \mid a_{n-2}$. Then from $p \mid \operatorname{gcd}\left(a_{n-1}, a_{n-2}\right)$, we have $p\left|a_{n-3}, \ldots, p\right| a_{2}=1$. Now let $p$ be a prime divisor of $B(4)$. Then $p$ cannot divide $B(1)$, else $B(1) B(5)=B(2) B(4)+B(3)^{2}$ implies $p \mid B(3)$. Also, since $B(0) B(4)=B(1) B(3)+B(2)^{2}, p$ cannot divide $B(2)$, for otherwise we have $p \mid B(1) B(3)$.
Lemma 2. The sum $B(5) B(7)+B(6)^{2} \equiv 0\pmod{B(4)}$
Proof. We use the recursion formula (1) for $B(n)$, with $n=4,5,6,7$ to get the identities
\begin{align}B(0) B(4)=B(1) B(3)+B(2)^{2} \tag i\\ B(1) B(5)=B(2) B(4)+B(3)^{2}\tag{ii} \\ B(2) B(6)=B(3) B(5)+B(4)^{2} \tag{iii}\\ B(3) B(7)=B(4) B(6)+B(5)^{2}\tag{iv
}\end{align}So that, applying (ii), (iv), (iii), again (iii), and (i),
$$
\begin{aligned}
B(5) B(7)+B(6)^{2} & \equiv B(1) B(2) B(5) B(7)+B(1) B(2) B(6) B(6) & & \text { from Lemma } 1 \\
& \equiv B(2) B(3) B(3) B(7)+B(1) B(2) B(6) B(6) & & \text { by (ii) } \\
& \equiv B(2) B(3) B(5) B(5)+B(1) B(2) B(6) B(6) & & \text { by (iv) } \\
& \equiv B(2) B(3) B(5) B(5)+B(1) B(3) B(5) B(6) & & \text { by (iii) } \\
& \equiv B(2) B(2) B(5) B(6)+B(1) B(3) B(5) B(6) & & \text { by (iii) } \\
&=B(5) B(6)[B(1) B(3)+B(2) B(2)] & \\
& \equiv 0\pmod{B(4)} \quad \text { by (i). }
\end{aligned}
$$
|
|