On a problem of Erdős and Graham (Q5892373)

For positive integers \(k\), \(d\) let \(f(x,k,d)=x(x+d)\cdots (x+(k-1)d)\), the product of \(k\) terms in the arithmetic progression of difference \(d\) and first term \(x\). Several authors have studied equations of the form \(f(x,k,d)=y^{\ell}\), with the most spectacular result being due to Erdős and Selfridge who showed in 1975 that the above equation has no positive integer solutions \((x,y,\ell)\) with \(\ell\geq 2\) when \(k\geq 2\) and \(d=1\). Erdős and Graham asked if the Diophantine equation \[ \prod_{i=1}^r f(x_i,k_i,1)=y^2 \] has, for fixed \(r\geq 1\), and \(\{k_1,\ldots,k_r\}\) with \(k_i\geq 4\) for \(i=1,\ldots,r\) at most finitely many solutions in positive integers \((x_1,x_2,\ldots,x_r,y)\) with \(x_i+k_i\leq x_{i+1}\) for \(i=1,\ldots,r-1\). After surveying the literature on this problem, the author moves on and proves that the above equation with \(r=2\), \(k_1=k_2=4\), and \(m:=x_2-x_1\) has the property that all its positive integer solutions \((x_1,y)\) satisfy \(1\leq x_1\leq 1.08m\). The proof uses Runge's method and is elementary. Using this result, and a numerical sieve, the author finds all the solutions when \(m\leq 10^6\), the result being that the only solution is \((x_1,x_2)=(33,1680)\) (corresponding to \(m=1647\)), for which \[ 33\times 34\times 35\times 36\times 1680\times 1681\times 1682\times 1683=3361826160^2, \] a solution which has already appeared in the literature in a different context.