9

I encountered this fact yesterday: $1$ and $4900$ are the only squares as the sum of $1+4+9+\ldots +n^2$. I was trying to solve this problem using my knowledge of elementary number theory. I reduce it to the point:

Show that $(a,b,c)=(2,5,7)$ is the only positive integer solution to $$ \left\lbrace \begin{array}{} 6\times a^2+1=b^2 \\ 12\times a^2+1=c^2 \end{array} \right. $$ (then let $n=6\times a^2$, you get $4900= 24\times25\times49/6$)

I recognize these as Pell's equations, but I don't know how to proceed.

Gary
  • 36,640
Chao H
  • 141

2 Answers2

4

You are exploring what has been called the Cannonball problem. Édouard Lucas formulated the cannonball problem as a Diophantine equation $$\sum_{n = 1}^N n^2 = M^2$$ or $$\frac{1}{6} N(N + 1)(2N + 1) = \frac{2N^3 + 3N^2 + N}{6} = M^2.$$

Lucas conjectured that the only solutions are $N=1, M=1$, and $N=24, M=70$, using either $1$ or $4900$ cannon balls (as it seems you have observed). It was not until 1918 that G. N. Watson proved it, using a method that is far beyond on the scope of elementary number theory. However it was proved in $1990$ in just over $4$ pages, if you are interested in seeing a simpler proof.

  • The last link is broken, I think you may be logged in to an university account when copying the link. – Andrew Feb 08 '23 at 03:21
  • 1
    Yes, the only place to access the paper is on JSTOR, so you need a login. – Clyde Kertzer Feb 08 '23 at 03:27
  • It’s not that it’s on JSTOR that’s the issue. Click on the link while in incognito mode. – Andrew Feb 08 '23 at 04:25
  • If you don't have access to JSTOR, I don't know any other way to access the paper. Let me know if you find another link that doesn't require login. – Clyde Kertzer Feb 08 '23 at 04:32
  • The issue that the paper is found on JSTOR is NOT the issue. https://www-jstor-org.colorado.idm.oclc.org/stable/2323911?origin=crossref You linked to your universities access page. Try accessing the link on incognito, off your university wifi. It will just bring you to a ucolorado login page. I guess changing the link like this actually works: https://www.jstor.org/stable/2323911?origin=crossref – Andrew Feb 08 '23 at 04:57
  • 1
    Ah, gotchya. Good catch, I just changed the link. – Clyde Kertzer Feb 08 '23 at 05:06
  • 3
    This is why, instead of "a paper", the link text should be a proper citation with title, author's name, year, and journal name. That way, people can find the paper even if something is wrong with the link. – MJD Feb 08 '23 at 05:36
  • @ClydeKertzer I just saw this question. The same Pell equation $p^2-6q^2=1$ appears if we relax the requirement that the first of 24 consecutive squares be $x=1$. Kindly see 2nd answer. – Tito Piezas III Jan 27 '25 at 04:17
1

(More of a long comment.) This is an old question but a Pell equation in it caught my eye. The post is about,

$$1^2+2^2+3^2+\dots+n^2=y^2$$

with stated solution $n=24$ for $n>1$. The OP arrived at a pair of Pell equations, one of which is,

$$6\times a^2+1=b^2$$

Or more accurately since $a=2q$ is even,

$$\; \color{blue}{24}\,q^2+1 =p^2$$

If we relax the requirement that the initial square is $1$, but retain the condition the LHS contains $\color{blue}{24}$ squares, then the same Pell equation appears. Define,

$$F(x) = x^2+(x+1)^2+(x+2)^2+\dots+(x+23)^2=y^2$$

Equivalently,

$$F(x) = 24(x-1)(x+24)+70^2=y^2$$

From the structure of $F(x)$, one can clearly that $F(1) = F(-24) = 70^2$. But since we relax the conditions on $x$, then there are infinitely many sets of $24$ consecutive squares such that their sum is a square, given by,

$$x = (p+24q)^2+4\times23\,pq\quad\\[4pt] \quad y = (p+24q)^2+3\times23\,(p+4q)^2$$

where $(p,q)$ solve the Pell equation,

$$p^2-24q^2=1$$

$p_n = 1,5,49,485,\dots$ (A001079)

$q_n =\, 0,1,10,99,\,\dots\,$ (A004189)

The three smallest cases yield,

$$\sum_{k=0}^{23} (1+k)^2 =70^2$$

$$\sum_{k=0}^{23} (1301+k)^2=6430^2$$

$$\sum_{k=0}^{23} (128601+k)^2=630070^2$$

and so on.

P.S. It seems the initial square conveniently has an ending digit of $1$, and is $x \equiv 1\, (\text{mod}\, 100)$. I've tested it for the first 20 $(p,q)$, but I'm not sure it is true for all.

  • 1
    $x_{n+1}=99x_n-99x_{n-1}+x_{n-2}$ so it continues to end in $01$ – Empy2 Feb 01 '25 at 16:42
  • @Empy2 Thanks for the confirmation. By the way, I finally completed that Heron triangle post. It needed a Somos-5 sequence to extended the OP's observations. – Tito Piezas III Feb 01 '25 at 17:30