Given the system: $$ \begin{cases} x^2y^2+x^2z^2=axyz & \\ y^2z^2+y^2x^2=bxyz &\\ z^2x^2+z^2y^2=cxyz \end{cases} $$ The solution could be gotten in a very tedious way. Is it possible to solve it considering some symmetry property of the system if any? Thanks in advance.

  • $\begingroup$ Factor, factor, factor. $\endgroup$ – John Jul 18 '14 at 22:24

Divide all three equations by $x y z$, and observe that the resulting system of equations has in effect only three variables. This is readily solved, and one can then map these back to $\{x,y,z\}$ rather easily.

Edit: As noted by Yeldarbskich, the division by $x y z$ is only valid if all are nonzero. Consequently there will also be three families of solutions corresponding to (at least) two of $\{x,y,z\}$ being zero.

  • $\begingroup$ From the start, the system has three variables, right? Can you please elaborate a bit ? $\endgroup$ – lab bhattacharjee Jul 18 '14 at 16:45
  • $\begingroup$ I was being deliberately a bit glib. What I meant more specifically is that after dividing through by $x y z$, the three equations are functions of $\{xy/z,y z/x,z x/y\}.$ $\endgroup$ – Semiclassical Jul 18 '14 at 16:46

If is clear we cannot have some $a,b,c$ positive while the rest negative. Otherwise, $xyz = 0$ and we only have trivial solution.

For simplicity, I will assume $a, b, c > 0$.

If any one of $x,y,z$ vanishes, say $x = 0$, it is clear the equations implies $y z = 0$. This means if one of $x,y,z$ vanish, at least two of $x,y,z$ vanish. If two of $x,y,z$ vanish, then the third one is arbitrary.

To eliminate this sort of trivial solutions, we will first assume $x, y, z > 0$. If we divide everything by $x^2 y^2 z^2$, we have $$\begin{cases} \frac{1}{z^2} + \frac{1}{y^2} &= \frac{a}{xyz}\\ \frac{1}{x^2} + \frac{1}{z^2} &= \frac{b}{xyz}\\ \frac{1}{y^2} + \frac{1}{x^2} &= \frac{c}{xyz} \end{cases} \quad\iff\quad \begin{cases} \frac{1}{x^2} &= \frac{b+c-a}{2xyz}\\ \frac{1}{y^2} &= \frac{a+c-b}{2xyz}\\ \frac{1}{z^2} &= \frac{a+b-c}{2xyz} \end{cases} \tag{*1} $$ This implies $$\begin{align} \frac{1}{(xyz)^2} &= \left(\frac{b+c-a}{2}\right)\left(\frac{a+c-b}{2}\right)\left(\frac{a+b-c}{2}\right)\frac{1}{(xyz)^3}\\ \iff\quad\quad xyz &= \left(\frac{b+c-a}{2}\right)\left(\frac{a+c-b}{2}\right)\left(\frac{a+b-c}{2}\right) \end{align} $$ Substitute this back into $(*1)$ immediately give us

$$\begin{cases} x &= x_0 \stackrel{def}{=} \sqrt{\left(\frac{a+c-b}{2}\right)\left(\frac{a+b-c}{2}\right)}\\ y &= y_0 \stackrel{def}{=} \sqrt{\left(\frac{b+c-a}{2}\right)\left(\frac{a+b-c}{2}\right)}\\ z &= z_0 \stackrel{def}{=} \sqrt{\left(\frac{b+c-a}{2}\right)\left(\frac{a+c-b}{2}\right)} \end{cases}\tag{*2}$$

What happens if some or all of the $x,y,z$ are negative? If one look back the original set of equations, the LHS is positive. Under the assume that $a, b, c > 0$, we find $xyz$ is positive. This means in general, the set of non-trivial solutions of the original equation is given by that in $(*2)$ with an even number of $x,y,z$ flipped sign. i.e. $$(x,y,z) = (x_0,y_0,z_0), (x_0,-y_0,-z_0), (-x_0,y_0,-z_0) \text{ or } (-x_0,-y_0,z_0)$$


I just want to add that Semiclassical's answer above has given you the right idea, but one should be careful to note that the possibility of any of x,y, and z could be zero, which would not work for that method. You should go back and check for solutions for combinations of x,y, and z being zero.

  • 1
    $\begingroup$ In particular, Semiclassical's method won't give you the trivial solution x=y=z=0. $\endgroup$ – Yeldarbskich Jul 18 '14 at 16:53
  • $\begingroup$ Good point. Mea culpa. $\endgroup$ – Semiclassical Jul 18 '14 at 16:55
  • $\begingroup$ It's not that bad actually. I just looked and if any are zero then you must have the trivial solution, but a solution nonetheless. $\endgroup$ – Yeldarbskich Jul 18 '14 at 16:56
  • $\begingroup$ Actually, if any two of $x$, $y$ and $z$ are zero, then the equations are trivially satisfied regardless of the value of third variable (or of the coefficients $a$, $b$ and $c$). So there are three families of trivial solutions, which correspond to the three coordinate axes in $(x,y,z)$ space, and which intersect at the point $x=y=z=0$. $\endgroup$ – Ilmari Karonen Jul 18 '14 at 18:44
  • $\begingroup$ "Good point. Mea culpa." $\endgroup$ – Yeldarbskich Jul 18 '14 at 20:09

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.