7
$\begingroup$

In essence, I want to prove that the product of any two distinct elements in the set $\{n^2, n^2+1, ... , (n+1)^2-1\}$ is never a perfect square for a positive integers $n$. I have no idea on how to prove it, but I've also yet to find a counterexample to this statement. Can anyone help?

$\endgroup$
  • $\begingroup$ just inequalities. All positive integers, can we have $n^2 < 2 w^2 < 2 (w+1)^2 < (n+1)^2 \; \; ? \; \;$ How about $n^2 < 3 w^2 < 3 (w+1)^2 < (n+1)^2 \; \; ? \; \;$ $\endgroup$ – Will Jagy Feb 17 at 20:17
  • 1
    $\begingroup$ is $n^2$ included in the set? If so, why is $(n+1)^2$ not included? $\endgroup$ – Dr. Mathva Feb 17 at 20:17
  • 2
    $\begingroup$ @Dr.Mathva Because then the statement is obviously false, with (n(n+1))^2 $\endgroup$ – gnasher729 Feb 17 at 20:20
7
$\begingroup$

First, note that $n^2$ can't be one of the elements as the other element would also need to be a perfect square for the product to be a perfect square. As gnasher729 commented to the answer, this is why $\left(n + 1\right)^2$ is not included.

Assume there are $2$ such elements, $n^2 + a$ and $n^2 + b$, with $a \neq b$, $1 \le a, b \le 2n$ and, WLOG, $a \lt b$. Thus, consider their product to be a perfect square of $n^2 + c$, i.e.,

$$\left(n^2 + a\right)\left(n^2 + b\right) = \left(n^2 + c\right)^2 \tag{1}\label{eq1}$$

for some integer $a \lt c \lt b$. As such, for some positive integers $d$ and $e$, we have that

$$a = c - d \tag{2}\label{eq2}$$ $$b = c + e \tag{3}\label{eq3}$$

Substitute \eqref{eq2} and \eqref{eq3} into \eqref{eq1} to get

$$\left(n^2 + \left(c - d\right)\right)\left(n^2 + \left(c + e\right)\right) = \left(n^2 + c\right)^2 \tag{4}\label{eq4}$$

Expanding both sides gives

$$n^4 + 2cn^2 + \left(e - d\right)n^2 + c^2 + c\left(e - d\right) - ed = n^4 + 2cn^2 + c^2 \tag{5}\label{eq5}$$

Removing the common terms on both sides and moving the remaining terms, apart from the $n^2$ one, to the right gives

$$\left(e - d\right)n^2 = -c\left(e - d\right) + ed \tag{6}\label{eq6}$$

Note that $e \le d$ doesn't work because the LHS becomes non-positive but the RHS becomes positive. Thus, consider $e \gt d$, i.e., let

$$e = d + m, \text{ where } m \ge 1 \tag{7}\label{eq7}$$

Using \eqref{eq3} - \eqref{eq2}, this gives

$$b - a = e + d \lt 2n \Rightarrow 2d + m \lt 2n \Rightarrow d \lt n - \frac{m}{2} \tag{8}\label{eq8}$$

Also,

$$ed = \left(d + m\right)d \lt \left(n + \frac{m}{2}\right)\left(n - \frac{m}{2}\right) = n^2 - \frac{m^2}{4} \lt n^2 \tag{9}\label{eq9}$$

Since $e \gt d$ means that $-c\left(e - d\right) \lt 0$, the RHS of \eqref{eq6} cannot be a positive integral multiple of $n^2$, so it can't be equal to the LHS.

$\endgroup$
  • $\begingroup$ Or, by (7), and by $c>0,$ since $e-d$ and $c$ are positive we have, in (6) that $n^2\le (e-d)n^2<ed$ which contradicts (9)....+1. $\endgroup$ – DanielWainfleet Feb 18 at 21:49
4
$\begingroup$

For any two numbers $n^2+a,\ n^2+b;\ 0<a<b<(2n+1)$, their product will satisfy $n^4<n^4+(a+b)n^2+ab<(n^2+2n+1)^2$.

All of the squares between $n^4$ and $(n^2+2n+1)^2$ will have the form $(n^2+m)^2=n^4+2mn^2+m^2;\ 1\le m\le 2n$

If $n^4+(a+b)n^2+ab$ is a perfect square, it will be one of the squares between $n^4$ and $(n^2+2n+1)^2$ and hence equal to $n^4+2mn^2+m^2$ for some $m$.

Thus $(a+b)=2m;\ ab=m^2$ for some $m$

Rearranging, we get $m^2=\frac{a^2+2ab+b^2}{4}=ab=\frac{4ab}{4}$, or $a^2+b^2=2ab$.

This implies both $a\mid b$ and $b\mid a$, meaning $b=a$ and the numbers being multiplied to obtain a perfect square are not distinct.

Added by edit: John Omielan comments (for the specific case $k=1$) that my original answer fails to consider possible solutions of the form $a+b=2m-k;\ ab=m^2+kn^2$. He separately provides a more complete answer that addresses those cases. Marty Cohen comments that I can only properly conclude $n^4+(a+b)n^2+ab=n^4+2mn^2+m^2 \Rightarrow n^2(a+b-2m)=(m^2-ab)$. Let me address those shortcomings.

If $(n^2+a)(n^2+b)=(n^2+m)(n^2+m)$ then either $a=b=m$ (addressed in my original answer) or $a<m<b$ (which this edit will address). If $n^2(a+b-2m)=(m^2-ab)$, then $(m^2-ab)$ is divisible by $n^2$, so $(m^2-ab)=kn^2=n^2(a+b-2m)$, or $(a+b-2m)=k\Rightarrow a+b=2m+k$.

$m,a,b$ have limits on their sizes. $m<b\le 2n\Rightarrow m^2< 4n^2$. Also $a<b\le2n \Rightarrow ab<4n^2$ Hence, $|(m^2-ab)|<4n^2 \Rightarrow |k|=(a+b-2m)<4$. For $(a+b)=2m+k,\ |k|=0,1,2,3$, where $k=0$ corresponds to the case where $a=b=m$.

The midpoint $t$ between $a$ and $b$ is the average $t=\frac{a+b}{2}$. Let $b-t=r,\ t-a=r$. Note that if $a$ and $b$ have different parity, $t$ and $r$ may have half integral values. Finally, $b-a=2r$, but since $b\le2n,\ a\ge 1$, then $2r=b-a<2n\Rightarrow r<n$.

$(n^2+a)(n^2+b)=((n^2+t)-r)((n^2+t)+r)=(n^2+t)^2-r^2$. Therefore, unless $m>t$, $(n^2+a)(n^2+b)<(n^2+m)^2$. But $t=\frac{a+b}{2}=m-\frac{k}{2}$. $m-t=\frac{1}{2},1,\frac{3}{2}$. Within the constraints of the original question, $m$ must be larger than the average of $a$ and $b$, but it must be very close to that average.

Substituting $t-\frac{k}{2}$ for $m$ in $(n^2+m)^2$, we get $((n^2+t)-\frac{k}{2})^2$. Can that equal $(n^2+a)(n^2+b)=((n^2+t)-r)((n^2+t)+r)=(n^2+t)^2-r^2$? Letting $s=(n^2+t)$ for simplicity in keeping track during expansion, we ask whether $(s-\frac{k}{2})^2=(s-r)(s+r)=s^2-r^2\Rightarrow ks-\frac{k^2}{4}=r^2\Rightarrow k(n^2+t)-\frac{k^2}{4}=r^2$?

$r<n\Rightarrow r^2<n^2$. It is obvious for $3\ge k>1$ that $k(n^2+t)-\frac{k^2}{4}>n^2>r^2$. For $k=1$, $n^2+t-\frac{1}{4}>n^2 \iff t>\frac{1}{4}$. For $a<m<b$, $\min(a+b)=4 \Rightarrow \min(t)=2; 2>\frac{1}{4}$.

There are no solutions to the question for $a<m<b$.

$\endgroup$
  • 1
    $\begingroup$ Simpler: if $a+2+b^2 = 2ab$ then $(a-b)^2 = 0$ so $a = b$. Nice answer. $\endgroup$ – marty cohen Feb 17 at 21:48
  • 3
    $\begingroup$ Note that $n$ is a fixed value. How do you know that, for example, $a + b = 2m - 1$ and $ab = n^2 + m^2$ can't be true instead? It might be intuitively obvious to you, & others, but it's not to me. This is why I try to explicitly show this doesn't occur in my answer. $\endgroup$ – John Omielan Feb 17 at 21:49
  • $\begingroup$ I think this answer is wrong. All you can conclude is that $(a+b)n^2+ab = 2mn^2+m^2$ or $n^2(a+b-2m) = m^2-ab$. This is not a polynomial identity in $n$, so you can not conclude that $a+b = 2m$ and ab = n^2$. $\endgroup$ – marty cohen Feb 18 at 17:59

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.