Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

After applying the Lagrange multiplier method, I got the following system of equations, which is quite symmetric:

$(x+y)^2 + (x+z)^2 = \frac{2}{3} \lambda x$

$(y+x)^2 + (y+z)^2 = \frac{2}{3} \lambda y $

$(z+x)^2 + (z+y)^2 = \frac{2}{3} \lambda z$

It is claimed that this system of equations is satisfied only when $x=y=z$. Could anyone explain why?

share|cite|improve this question
up vote 1 down vote accepted

Obviously, $x=y=z$ leads to the solutions $x = y = z = 0$ and $x = y = z = \frac {\lambda}{12}$. Now, let's assume $x \ne y \ne z$. Subtract second equation from first, third from first and third from second, so you'll get equivalent system $$ \begin{align} (x-y)(x+y+2z) &= \frac 23 \lambda (x-y)\\ (x-z)(x+2y+z) &= \frac 23 \lambda (x-z)\\ (y-z)(2x+y+z) &= \frac 23 \lambda (y-z) \end{align} $$ or $$ \begin{align} x+y+2z &= \frac 23 \lambda\\ x+2y+z &= \frac 23 \lambda\\ 2x+y+z &= \frac 23 \lambda \end{align} $$ since we assumed that $x \ne y \ne z$.

Comparing first and second, one may conclude that $$ x+y+2z = x+2y+z $$ or equivalently $$ y = z $$ Considering other two pairs one may conclude $$ x = y = z $$ which is in contradiction with initial assumptions. So the only solutions are those when $x = y = z$.


As it was pointed out, one also may check case when $x = y$ but $z \ne x,y$. In this case consider first and third equations $$ \begin{align} 4x^2 + (x+z)^2 &= \frac 23 \lambda x \\ 2(z+x)^2 &= \frac 23 \lambda z \end{align} $$ Subtract first from second $$ (z-x)(z+3x) = \frac 23 \lambda (z-x) $$ Since we're assuming $z \ne x$, then $$ z+3x = \frac 23 \lambda $$ or explicitly $$ z = \frac 23 \lambda - 3x $$ Substitute it to the any of previous two, let's say second equation $$ \left(\frac 23 \lambda - 2x\right)^2 = \frac 13 \lambda \left( \frac 23 \lambda -3x\right) $$ and if you solve it, you'll get complex values for $x$. So, contradiction.

In the same manner you can assume $z=y$, $x \ne y,z$ and $z=x$, $y \ne x,z$.

share|cite|improve this answer
This does not cover the case that only two out of three are equal (e.g. $x=y$, $z\neq x$, $z \neq y$). – WimC May 2 '13 at 7:34
@WimC you're right, it doesn't. I'll fix it in a minute. – Kaster May 2 '13 at 7:42

Interpreted as $3$-vectors, the left hand side and the right hand side (without the $\lambda$-factor) must be dependent. That means that all $2 \times 2$ minors of the matrix

$$\begin{pmatrix} (x+y)^2+(x+z)^2 & x\\ (y+x)^2+(y+z)^2 & y\\ (z+x)^2+(z+y)^2 & z \end{pmatrix} $$

must vanish (or in other words the cross product of the colums must be zero). Let's consider the first one:

$$ 0 = y \left( (x+y)^2 + (x+z)^2 \right) - x \left((y+x)^2 + (y+z)^2 \right) = (y-x)(x^2 + xy + y^2 + z^2). $$

So either $x=y$ or $x^2 + xy + y^2 + z^2 = 0$. The latter implies that $x=y=z=0$ since the quadratic form is positive definite over $\mathbb{R}$. In either case $x=y$. Other equalities follow similarly from the other minors.

One way to show that the quadratic form is positive definite is to write it as a positive combination of squares:

$$x^2 + xy + y^2 + z^2 = \frac{1}{2}\left(x + \frac{1-\sqrt{3}}{2} y\right)^2 + \frac{1}{2} \left(x + \frac{1+\sqrt{3}}{2} y\right)^2 + z^2$$

share|cite|improve this answer
What do you mean by "all 2×2 minors must vanish"? What matrix are you considering? – MichaelNgelo May 2 '13 at 18:19
@MichaelNgelo Added the matrix explicitly. – WimC May 2 '13 at 18:35

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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