Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

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

I had a few questions about the roots of power sum symmetric polynomials:

  1. Given that $x_1^k+x_2^k+x_3^k= 0$ for all $k \not \equiv 0\mod 3$ and is non-zero otherwise, if we assume none of the $x_i$'s equal 0, does this imply that the $\{x_i\|1\leq i \leq 3 \}=\{a\zeta_3^k|1\leq k\leq 3 \}$ where $\zeta_3$ is a primitive third root of unity and $a$ is an arbitrary constant.

  2. If we had 6 variables instead of 3 (i.e. $x_1^k+x_2^k+x_3^k+x_4^k+x_5^k+x_6^k = 0$ for all $k \not \equiv 0\mod 3$ and is non-zero otherwise), under the same assumption that none of the $x_i$'s is $0$, could we still conclude that the $\{x_i\}=\{a\zeta_3^k|1\leq k\leq 3 \}\cup \{b\zeta_3^k|1\leq k\leq 3 \}$ where $\zeta_3$ is a primitive third root of unity and $a$ and $b$ are arbitrary constants?

I appreciate any insight any of you have. Thanks!

share|cite|improve this question
up vote 2 down vote accepted

Complete rewrite to make it more general.

Let $x_1,...,x_n$ be non-zero complex numbers such that $G_k=\sum_{i=1}^n x_i^k = 0$ iff $k\not\equiv 0 \pmod m$.

Define the generating function $$g(z)=\sum_{k=0}^\infty G_k z^k = \sum_{i=1}^n \frac{1}{1-x_iz}$$

Now since $G_k=0$ when $k\not\equiv 0\pmod m$, let $\zeta_m$ be a primitive $m$th root of unity, we see that $g(\zeta_m z) = g(z)$ for all $z$ where the series converges, and hence for all $z$ in the full analytic continuation.

In particular, then $z\to \zeta_m z$ must send the poles of $g(z)$ to the poles of $g(z)$. The poles of $g(z)$ are the numbers of the form $\frac{1}{x_i}$, so for each $i$, there must be a $j$ such that $\zeta_m/x_i = 1/x_j$ , or$\zeta_m x_j = x_i$. That means that $z\to\zeta_m z$ must permute the $x_i$. In particular, we can partition $x_i$ into orbits under this action. (In particular, it must be the case that $m|n$.)

Part 1 is the case $m=n=3$. You have to deal with the case where some or more of the $x_i$ are zero, but that's easy because that reduces to the problem when $n<3, m=3$, which only has zero solutions.

Part 2 is the case $m=3, n=6$. Again, you have to take into account the cases where an $x_i$ is zero, whih reduces to the case where $m=n=3$ or all the $x_i$ are zero.

Note, we didn't actually use that the $G_{mk}\not = 0$. That was a red herring.

(There's a slight wrinkle when the $x_i$ are not distinct, but then you can just remove one instance each of $x_1,\zeta_m x_1,...\zeta_m^{m-1} x_1$ from the set and apply the result for the subset by induction. Essentially, if $x$ occurs $j$ times in $x_1,...,x_n$ the $\zeta_m x$ occurs $j$ times, too.]

share|cite|improve this answer

Hint for 1: Apply Vieta's Formula. What can we say about $\sum x_i$ and $\sum x_i x_j$, and $x_1 x_2 x_3$? Hence the roots are ...

Hint for 2:

The Vandemonde matrix has non-zero determinant (viewed as a polynomial in x_k).

Newton's identities tells us that the values of $\sum x_k, k\geq 6$ are determined by the values for $k=0$ to 5. From the question, the image has 2 degrees of freedom.

Hence, it is sufficient to find a set of 2 linearly independent vectors.

(This works for 1 too, but uses slightly more machinery.)

share|cite|improve this answer
Thanks! I didn't think of Vieta's Formula. That's helpful. – Fortunato Jan 9 '13 at 21:39
I'm confused about this statement: "Newton's identities tells us that the values of "$\sum x_k$, k≥6 are determined by the values for k=0 to 5. From the question, the image has 2 degrees of freedom." Did you mean $\sum x^k$? Also, where do the two degrees of freedom come from? When $k=3$ and when $k \neq 3$? – Fortunato Jan 10 '13 at 15:17
Construct the matrix $A_{i,j} = x_j ^i$ for $i, j = 1$ to $6$. This has det $\prod x_i \prod(x_i-x_j)$. (the initial $\prod x_i$ arises because the true Vandemonde uses all 1's; doesn't make a difference). Then, you're asking for the set of solutions to $A \cdot 1 = (0, 0, a, 0, 0, b)^T$. Since $A$ has full rank, this set of solutions is a deg 2 subspace. (so the 2 degrees of freedom came from the choices of $\sum x_i ^3, \sum x_i ^6$ – Calvin Lin Jan 10 '13 at 15:28
And so I'm then looking for 2 linearly independent vector solutions of the form $(x_1,\ldots , x_6)$ that satisfy both $\sum x_i^3 =a$ and $\sum x_i ^6 =b$. And the only possibilities (up to a permutation of variables) are of the form $(c1 \zeta _3, c1 \zeta _3^2, c1, c2 \zeta _3, c2 \zeta _3^2, c2)$ where $c1$ and $c2$ are constants and $\zeta _3$ is a primitive 3rd root of unity? – Fortunato Jan 10 '13 at 16:02
I'm still having trouble seeing how this leads to the conclusion. Can I have a nudge? – Fortunato Jan 10 '13 at 18:45

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.