Consider the equation $$\frac{a^2}{a^2-1} \cdot \frac{b^2}{b^2-1} = \frac{c^2}{c^2-1}.$$ Of course, there are solutions to this like $(a,b,c) = (9,8,6)$. Is there any known approximation for the number of solutions $(a,b,c)$, when $2 \leq a,b,c \leq k$ for some $k \geq 2.$

More generally, consider the equation $$\frac{a_1^2}{a_1^2-1} \cdot \frac{a_2^2}{a_2^2-1} \cdot \ldots \cdot \frac{a_n^2}{a_n^2-1} = \frac{b_1^2}{b_1^2-1} \cdot \frac{b_2^2}{b_2^2-1}\cdot \ldots \cdot \frac{b_m^2}{b_m^2-1}$$ for some natural numbers $n,m \geq 1$. Similarly to the above question, I ask myself if there is any known approximation to the number of solutions $(a_1,\ldots,a_n,b_1,\ldots,b_m)$, with natural numbers $2 \leq a_1, \ldots, a_n, b_1, \ldots, b_m \leq k$ for some $k \geq 2$. Of course, for $n = m$, all $2n$-tuples are solutions, where $(a_1,\ldots,a_n)$ is just a permutation of $(b_1,\ldots,b_n)$.

  • 6
    $\begingroup$ "I wonder" means "I haven't put any work into this myself"? $\endgroup$ – Gerry Myerson May 5 at 23:49
  • 3
    $\begingroup$ What does like mean in "solutions ... like $(a,b,c)=(9,8,6)$"? $\endgroup$ – Gerry Myerson May 6 at 12:29
  • 4
    $\begingroup$ Let $a^2\to u+v$, $b^2\to u-v$ and $c$ as parameter. Then solving difference of squares $\Bigl(2 (u - c^2)\Bigr)^2 - \Bigl(2 v\Bigr)^2 = (2 c^2 - 1)^2 - 1$. (a,b,c)=(9,8,6),(26,15,13),(55,24,22),(99,35,33),(50,49,35),(161,48,46),(120,55,50),(244,63,61),(351,80,78),(485,99,97),... $\endgroup$ – Dmitry Ezhov May 6 at 14:28
  • 1
    $\begingroup$ @Dmitry do all $c$ of the form $n^2-3$ give solutions? Most of the solutions you give have $c$ of the form. $\endgroup$ – Gerry Myerson May 7 at 1:45
  • 1
    $\begingroup$ @Gerry yes, for $c$ in 2..1000 give all $n$ in 3..31 $\endgroup$ – Dmitry Ezhov May 7 at 5:03

It seems worth noting that the equation in the title does have infinitely many solutions in positive integers, as for all $n$ it is satisfied by $$a={n(n^2-3)\over2},\ b=n^2-1,\ c=n^2-3.$$ The number of solutions of this form with $a\le k$ will be on the order of $\root3\of{2k}$, but Dmitry has found solutions not of this form.

| cite | improve this answer | |

Here's another infinite family. Let $x,y$ be positive integers such that $x^2-2y^2=\pm1$ – there are infinitely many such pairs. Let $a=x^2$, $b=2y^2$, $c=xy$, then a little algebra will show that $(a,b,c)$ satisfy the equation in the title.

E.g., $x=3$, $y=2$ leads to $(9,8,6)$, and $x=7$, $y=5$ yields $(49,50,35)$, two triples already found by Dmitry, while $x=17$, $y=12$ gets us $(289,288,204)$.

This infinite family is much thinner than the one in the other answer.

[I seem to have become disconnected from the account under which I posted the other answer.]

EDIT: A third infinite family. $$a=4n(n+1)(n^2+n-1),\ b=(2n+1)(2n^2+2n-1),\ c=2(2n+1)(n^2+n-1)$$

| cite | improve this answer | |

Above equation shown below, has solution:

$\frac{a^2}{a^2-1} \cdot \frac{b^2}{b^2-1} = \frac{c^2}{c^2-1}$




Where, w=[1/(36p^2-7)]

For, $p=0$ we get:


| cite | improve this answer | |
  • 1
    $\begingroup$ For which values of $p$ does this lead to a solution in integers? Very few, I think. Your formula gives $b=8-28(7p-2)/(36p^2-7)$, and a linear over a quadratic can only be an integer for finitely many integers $p$. $\endgroup$ – Gerry Myerson Jun 8 at 23:36
  • $\begingroup$ There are infinitely many integer solutions with $3b=4c$. Let $n$ be odd, and $(8+3\sqrt7)^n=x+y\sqrt7$, $x,y$ integers. Then $a=3y$, $b=x$, $c=3x/4$ is a solution in integers with $3b=4x$. $n=1$ gives $(9,8,6)$; $n=3$ gives $(2295,2024,1518)$. $\endgroup$ – Gerry Myerson Jun 9 at 3:17

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.