If $ab-1$, $bc-1$, $ca-1$, $ab-a-b+c$, $bc-b-c+a$, $ca-c-a+b$ are perfect squares, then are $ab+a+b-c$, $bc+b+c-a$, $ca+c+a-b$ also perfect squares?

1.2k Views Asked by At

About a month ago, a friend of mine taught me that there exist many sets of three positive integers $(a,b,c)$ where $a\not=b,b\not=c$ and $c\not=a$ such that each of $$ab-1,\ bc-1,\ ca-1,\ ab-a-b+c,\ bc-b-c+a,\ ca-c-a+b$$ is a perfect square.

Then, I found that there exist infinitely many such sets. For example, for $$(a,b,c)=(F_{2n-1},F_{2n+1},F_{2n+3})$$ where $F_k$ is the $k$-th Fibonacci number defined by $F_0=0,F_1=1,F_{k}=F_{k-2}+F_{k-1}\ (k\ge 2)$, we have $$\color{red}{ab-1}=F_{2n-1}F_{2n+1}-1=\color{blue}{{F_{2n}}^2}$$ $$\color{red}{bc-1}=F_{2n+1}F_{2n+3}-1=\color{blue}{{F_{2n+2}}^2}$$ $$\color{red}{ca-1}=F_{2n+3}F_{2n-1}-1=F_{2n}F_{2n+2}+1=\color{blue}{{F_{2n+1}}^2}$$ $$\begin{align}\color{red}{ab-a-b+c}&=F_{2n-1}F_{2n+1}-F_{2n-1}-F_{2n+1}+F_{2n+3}\\&={F_{2n}}^2+1-F_{2n-1}+F_{2n+2}\\&={F_{2n}}^2+1+2F_{2n}\\&=\color{blue}{(F_{2n}+1)^2}\end{align}$$ $$\begin{align}\color{red}{bc-b-c+a}&=F_{2n+1}F_{2n+3}-F_{2n+1}-F_{2n+3}+F_{2n-1}\\&={F_{2n+2}}^2+1-F_{2n+3}-F_{2n}\\&={F_{2n+2}}^2+1-2F_{2n+2}\\&=\color{blue}{(F_{2n+2}-1)^2}\end{align}$$ $$\begin{align}\color{red}{ca-c-a+b}&=F_{2n+3}F_{2n-1}-F_{2n+3}-F_{2n-1}+F_{2n+1}\\&={F_{2n+1}}^2+1-F_{2n-1}-F_{2n+2}\\&={F_{2n+1}}^2+1-2F_{2n+1}\\&=\color{blue}{(F_{2n+1}-1)^2}\end{align}$$

I have also found many other such sets which are not included in this Fibonacci family, and...

Interestingly enough, for every set $(a,b,c)$ I found, each of $$ab+a+b-c,\ bc+b+c-a,\ ca+c+a-b$$ is also a perfect square.

For example, for $(a,b,c)=(F_{2n-1},F_{2n+1},F_{2n+3})$, we have $$\begin{align}\color{red}{ab+a+b-c}&=F_{2n-1}F_{2n+1}+F_{2n-1}+F_{2n+1}-F_{2n+3}\\&={F_{2n}}^2+1+F_{2n-1}-F_{2n+2}\\&={F_{2n}}^2+1-2F_{2n}\\&=\color{blue}{(F_{2n}-1)^2}\end{align}$$ $$\begin{align}\color{red}{bc+b+c-a}&=F_{2n+1}F_{2n+3}+F_{2n+1}+F_{2n+3}-F_{2n-1}\\&={F_{2n+2}}^2+1+F_{2n+3}+F_{2n}\\&={F_{2n+2}}^2+1+2F_{2n+2}\\&=\color{blue}{(F_{2n+2}+1)^2}\end{align}$$ $$\begin{align}\color{red}{ca+c+a-b}&=F_{2n+3}F_{2n-1}+F_{2n+3}+F_{2n-1}-F_{2n+1}\\&={F_{2n+1}}^2+1+F_{2n-1}+F_{2n+2}\\&={F_{2n+1}}^2+1+2F_{2n+1}\\&=\color{blue}{(F_{2n+1}+1)^2}\end{align}$$

So, here is my question.

Question : Are the following propositions true?

Proposition 1 (False) : For positive integers $(a,b,c)$ such that $a\not=b,b\not=c$ and $c\not=a$, if each of $$ab-1,\ bc-1,\ ca-1,\ ab-a-b+c,\ bc-b-c+a,\ ca-c-a+b$$ is a perfect square, then each of $$ab+a+b-c,\ bc+b+c-a,\ ca+c+a-b$$ is a perfect square.

Proposition 2 (True) : For positive integers $(a,b,c)$ such that $\color{red}{1\le} a\lt b\lt c$ and $ab+a+b-c\ge 0$, if each of $$ab-1,\ bc-1,\ ca-1,\ ab-a-b+c,\ bc-b-c+a,\ ca-c-a+b$$ is a perfect square, then each of $$ab+a+b-c,\ bc+b+c-a,\ ca+c+a-b$$ is a perfect square.

Proposition 3 : For positive integers $(a,b,c)$ such that $\color{red}{2\le} a\lt b\lt c$, if each of $$ab-1,\ bc-1,\ ca-1,\ ab-a-b+c,\ bc-b-c+a,\ ca-c-a+b$$ is a perfect square, then each of $$ab+a+b-c,\ bc+b+c-a,\ ca+c+a-b$$ is a perfect square.

Proposition 4 (True) : For positive integers $(a,b,c)$ such that $\color{red}{1\le} a\lt b\lt c$, if each of $$ab-1,\ bc-1,\ ca-1,\ ab\mp (a+ b-c),\ bc\mp (b+ c- a),\ ca\mp (c+ a-b)$$ is a perfect square, then $c=a+b+2\sqrt{ab-1}$ holds.

Added : Proposition 1 is false because of $(a,b,c)=(1,5,65)$ found by Oleg567. Then, I added proposition 2 and 3. Also, I added proposition 4 (see Tito Piezas III's answer).

Can anyone help?

Update : I crossposted to MO.

Added : Proposition 2 and 4 are true (see my answer for the details). However, we still don't know if proposition 3 is true.

2

There are 2 best solutions below

6
On BEST ANSWER

(Update.) A quick search of,

$$\begin{aligned} ab-1\;&=d^2\\ ac-1\;&=e^2\\ bc-1\;&=f^2 \end{aligned}$$

with $1<a<b<c$ gives,

$$2,\;5,\;13\\2,\;13,\;25\\2,\;25,\;41\\5,\;10,\;29\\5,\;13,\;34\\5,\;29,\;58$$ Certain patterns are immediately apparent. One of which is that the first 100 solutions with $c$ below a bound (and even higher) ALL satisfy,

$$(a+b+c)^2 = 2(a^2+b^2+c^2+2)\tag1$$

Equivalently (by solving for $c$),

$$c=a+b\pm2\sqrt{ab-1}=a+b\pm 2d$$

This has implications for the original proposition by the OP. (The first solution that does not satisfy the relation is $2,\,5,\,925.$)

I. Assumption 1.

There is a missing assumption in the Question section. It suffices to find two integers $a,b$ such that,

$$ab-1 = d^2\tag{2a}$$

then define $c$ as,

$$\color{blue}{c = a + b - 2 d}\tag{2b}$$

and these two conditions are sufficient to ensure that the following nine expressions necessarily are squares,

$$\begin{aligned} a b-1\; &= d^2\\ a c-1\; &= (a - d)^2\\ b c-1 \; &= (b - d)^2\\ a b - a - b + c \; &= (1 - d)^2\\ b c + a - b - c \; &= (1 - b + d)^2\\ a c - a + b - c \; &= (1 - a + d)^2\\ a b + a + b - c \; &= (1 + d)^2\\ b c - a + b + c \; &= (1 + b - d)^2\\ a c + a - b + c \; &= (1 + a - d)^2 \end{aligned}$$

There are broad families to the 2 conditions. One is $a,\,b,\,c = x-ny,\;x+ny,\;2(x+my)$ where $x,y$ solve the Pell equation,

$$x^2-(m^2+n^2)y^2 = 1$$

This covers the Fibonacci family by the OP which was just the case $m,n= 2,1.$ (See also this similar post.)

II. Assumption 2.

For the counter-examples found by Oleg567 to the original proposition, if,

$$a=1,\quad \text{so}\quad b,c = p^2+1,\;q^2+1\tag3$$

$$\begin{aligned} a b-1\; &= p^2\\ a c-1\; &= q^2\\ \color{brown}{b c-1} \; &= p^2+q^2+p^2q^2\\ a b - a - b + c \; &= q^2\\ b c + a - b - c \; &= p^2q^2\\ a c - a + b - c \; &= p^2\\ \end{aligned}$$

If $q=2p^2$, then $\color{brown}{b c-1}$ also becomes a square. However, the remaining three numbers will not necessarily be so.

III. Summary

It is asked, "For $2\leq a < b < c$, if the six numbers are perfect squares, then are the three (remaining) numbers perfect squares?".

I assume that the six expressions being squares are not sufficient, by themselves, to ensure that the remaining three expressions will be also squares (like Oleg's counter-examples). The fact that for $a>1$ the remaining three numbers were squares seem to be more due to the additional condition $(1)$.

But it would be interesting to find $a,b,c$ with $a>1$ such that only the first six numbers are squares.

0
On

I'm posting an answer just to inform that we know proposition 2 and 4 are true from a comment by a user duje on MO.

From Lemma 7 of this paper duje showed in his comment, we can say

If each of $$ab-1,\ \ bc-1,\ \ ca-1$$ is a perfect square and $0\lt a\lt b\lt c\lt 3ab\in\mathbb Z$, then $c=a+b+2\sqrt{ab-1}$ holds.

By adding Tito Piezas III's findings to this result, we can say

If each of $$ab-1,\ \ bc-1,\ \ ca-1$$ is a perfect square and $0\lt a\lt b\lt c\lt 3ab\in\mathbb Z$, then each of $$ab\mp (a+ b-c),\ bc\mp (b+ c- a),\ ca\mp (c+ a-b)$$ is a perfect square.

Then, duje pointed that since $1\le a\lt b\lt c,ab+a+b-c\ge 0\Rightarrow c\lt 3ab,$ we can say that proposition 2 and 4 are true.