# product is twice a square

For every positive integer $n$, there exists a set $S\subset \{n^2+1,n^2+2,\dotsc,(n+1)^2-1\}$, such that

$$\prod_{k\in S}k=2m^2$$

for some positive integer $m$

I have no clue about it. Could anyone help me? Thanks a lot.

p.s. Whether or not the problem was a conclusion of an paper is unknown.

-
I think you should write $\prod_{k\in S} k=2m^2$. – Luiz Cordeiro Jan 1 '14 at 20:07
Where did you find this? (Given the proof in Gerry's link, this problem seems to be really hard) – chubakueno Jan 1 '14 at 23:13
I'm still trying to puzzle out that P.S. – Byron Schmuland Jan 1 '14 at 23:32

Consider (as Hugh Denoncourt did in his deleted answer) the vector space $V$ over ${\mathbb F}_2$ with basis corresponding to the prime factors $p_1, \ldots, p_N$ of $n^2 + 1, \ldots, (n+1)^2 - 1$. The number $N$ of such prime factors is OEIS sequence A143346. It appears that $N$ is approximately $1.52 n$ for large $n$, and is less than $2n$ for $6 < n \le 10000$ (the maximum $n$ in the data file). Presumably this is true for all $n > 6$. Let $v(k) = [\alpha_1, \ldots, \alpha_N] \mod 2$ where $k = p_1^{\alpha_1} \ldots p_N^{\alpha_N}$. Now the question is whether $[1,0,\ldots,0]$ is in the linear span of $v(n^2+1), \ldots, v((n+1)^2-1)$. In fact, it appears that this linear span is the whole vector space $V$. That is true for at least all $n \le 200$.
This is nice. I gave a bad reason for why there should be fewer than $2n$ prime factors amongst the integers in the interval. The requirement that $[1,0,\ldots,0]$ be in the span unfortunately makes it so that knowing (or estimating) the dimension is not enough. – Hugh Denoncourt Jan 2 '14 at 7:11