# Regarding identities with sums of consecutive squares

This comic http://abstrusegoose.com/63 points out an interesting identity with sums of consecutive squares. Let us take positive integers $k, p, q$, with $p < q$, and ask if $k^2 + \ldots + (k + p)^2 = (k + p + 1)^2 + \ldots + (k + q)^2$. One may see this is equivalent to $(p+1)(6k^2 + 6kp + 2p^2 + p) = (q - p)(6k^2 + 6k(p + q + 1) + 2p^2 + p(2q + 3) + 2q^2 + 3q + 1)$. Surprisingly, there is an easy family of triples $(k, p, q)$ of solutions: $(p(2p + 1), p, 2p)$. Searching a bit, I found a few values of $k,p,q$ which solve the equation and are not of the given form:

4 34 44

12 38 51

16 126 163

18 16 24

60 50 75

67 92 131

What can we say about the solutions which are not of that form? Are there infinitely many? Is there another form?

-
The family you mention is the case where there are one more square on the left than on the right. Maybe try for two more, or three more, etc. I also noticed in two of the exceptional cases we have $(p,q)=(2t,3t)$ which may give another family. –  coffeemath Jul 5 '13 at 17:33
Define the pyramidal numbers $$P(n)=1^2+2^2+\cdots+n^2={n(n+1)(2n+1)\over6}$$ Then your left side is $P(k+p)-P(k-1)$, right side is $P(k+q)-P(k+p)$, so you are looking for solutions to $2P(r)=P(s)+P(t)$. Not an answer, but maybe a neater way to pose the question, as each variable is on its own. –  Gerry Myerson Jul 6 '13 at 1:58