8
$\begingroup$

Prove that there exist infinitely many positive integers $n$ such that $\frac{1^2+2^2+\cdots+n^2}{n}$ is a perfect square. Obviously, $1$ is the least integer having this property. Find the next two least integers with this property.

The given condition is equivalent to $(2n+2)(2n+1) = 12p^2$ where $p$ is a positive integer. Then since $\gcd(2n+2,2n+1) = 1$, we have that $2n+2 = 4k_1$ and $2n+1 = k_2$. We must have that $k_1$ is divisible by $3$ or that $k_2$ is divisible by $3$. If $k_1$ is divisible by $3$ and $k_2$ is not, then we must have that $k_1$ is divisible by $9$ and so $2n+2 = 36m$. Then we need $3mk_2$ to be a perfect square where $k_2+1 = 36m$. Thus if $3mk_2 = r^2$, we get $m = \dfrac{1}{72}\left(\sqrt{48r^2+1}+1\right)$.

$\endgroup$
  • 1
    $\begingroup$ So, $$(2n+2)(2n+1)=12p^2$$ where $(2n+2,2n+1)=1=(n+1,2n+1)$ $\endgroup$ – lab bhattacharjee Aug 7 '16 at 17:23
  • $\begingroup$ Pell's equation is the heart of it. Just show that if $a+b\sqrt{48}=(7+\sqrt{48})^k$ , then $a\equiv 3\pmod 4$ when $k$ is odd, and you are done. $\endgroup$ – Thomas Andrews Aug 7 '16 at 17:27
  • $\begingroup$ @labbhattacharjee How does that help? $\endgroup$ – user19405892 Aug 7 '16 at 17:31
  • $\begingroup$ Then next $n=337$. $\endgroup$ – Thomas Andrews Aug 7 '16 at 17:39
  • 2
    $\begingroup$ You haven't proven that $2n+1$ must be a multiple of $3$, you've just asserted it. @user19405892 $\endgroup$ – Thomas Andrews Aug 7 '16 at 17:43
3
$\begingroup$

From the equation $(2n+1)(2n+2) = 12p^2$, we know that there are two possibilities:

  • $2n+2$ is of the form $4x^2$, $2n+1$ is of the form $3y^2$.
  • $2n+2$ is of the form $12x^2$, $2n+1$ is of the form $y^2$.

This follows from $2n+1, 2n+2$ being relatively prime. Thus, solutions correspond to integer solutions of the equations $4x^2-3y^2 = 1$ and $12x^2-y^2 = 1$. We claim that the former equation has infinitely many solutions. These correspond to the solutions of $a^2-3b^2 = 1$ where $a$ is even.

The theory of Pell equations can be used to show that the solutions of $a^2-3b^2 = 1$ are given by the powers $a+b\sqrt{3} = (2+\sqrt{3})^i$ for $i\ge 0$. Even without this theory in hand, we can check directly that these give solutions: $$ a^2-3b^2 = (a+b\sqrt{3})(a-b\sqrt{3}) = (2+\sqrt{3})^i(2-\sqrt{3})^i = \big[(2+\sqrt{3})(2-\sqrt{3})\big]^i = 1 $$

It only remains to check that infinitely many of these solutions have $a$ as even. An easy inductive argument shows that $a$ is even precisely when $i$ is odd, so there are infinitely many solutions.

The first two nontrivial solutions are $(2+\sqrt{3})^3 = 26+15\sqrt{3}$ and $(2+\sqrt{3})^5 = 362+209\sqrt{3}$. These values of $a$ correspond to $n=337$ and $n=65521$. To check that these are indeed the smallest nontrivial solutions to $(2n+1)(2n+2)=12p^2$, there are two approaches. The first is to apply the theory of Pell equations to note that there are no solutions to $a^2-3b^2=1$ other than those given above, and that there are no solutions at all to $12x^2-y^2=1$. The second method is to check by brute force that no other values less than $65521$ yield solutions.

$\endgroup$
5
$\begingroup$

This can be solved by using the quadratic formula on $2n^2 + 3n + (1-6p^2)=0$ to convert to the Pell's equation $r^2 - 48p^2 = 1$. Standard techniques can be used to generate infinitely many solutions, though we need to be careful to keep only those with $r \equiv 3\pmod{4}$.

For example, we can arrive at infinitely many solutions by writing $r+p\sqrt{48} = (7+\sqrt{48})^{2k+1}$, beginning from the primitive solution $r=7, p=1$.

Here is a proof that results from that process:

If $(n,p)$ is a positive solution to $(n+1)(2n+1)=6p^2$, then $(97n+168p+72,56n+97p+42)$ is a larger solution, which we can see by computing directly:

$$((97n+168p+72) + 1)(2(97n+168p+72) + 1) - 6(56n+97p+42)^2$$ $$ = (n+1)(2n+1)-6p^2$$

Beginning from the solution $(n,p)=(1,1)$, this generates infinitely many solutions by induction.

$\endgroup$
  • $\begingroup$ @RossMillikan The solution is $(n,p)=(1,1)$, or equivalently $(r,p)=(7,1)$. I've clarified. $\endgroup$ – Slade Aug 7 '16 at 20:39
1
$\begingroup$

Using this: $$ \frac{\sum_{k=1}^N k^2}{N}= \frac{1}{6} (N+1) (2 N+1)=K^2 $$ the condition can be written as (as you pointed out): $$(N+1) (2 N+1)=6K^2$$ through numerical search, the last part of the problem is solved $$N=1,337,65521$$

Not really a satisfying answer i know.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.