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

how find $a_1,a_2,...,a_n$ ,$ a_i\in\mathbb R$-{0}(all possible $a_i$,i=1,2,..,n) such that $$\sum_{i=1}^na_i^m=\sum_{i=1}^na_i $$ $m=1,2,...,n+1$ Thanks in adnance

share|cite|improve this question
I don't understad your question. Taking $a_i=1$ for all $i$ seems to do what you want... – Mariano Suárez-Alvarez Feb 18 '13 at 18:35
@Mariano Suárez-Alvarez:i mean all possible a_i – Maisam Hedyelloo Feb 18 '13 at 18:38
Then please say so in the body of the question. – Mariano Suárez-Alvarez Feb 18 '13 at 18:39

Let $p_k=a_1^k+\dots+a_n^k$ for each $k$, and let $e_k$ be the elementary symmetric function of degree $k$ on the $a$'s.

Your hypothesis is that $$\text{$p_k=p_1$ for all $k\in\{1,\dots,n+1\}$.}\tag{$\star$}$$ The Newton identities tell us that $ke_k=\sum_{i=1}^k(-1)^{i-1}e_{k-i}p_i$ for all $k\in\{1,\dots,n\}$, and in our case this means that $$ke_k=p_1\sum_{i=1}^k(-1)^{i-1}e_{k-i}.$$

It follows from the hypothesis $(\star)$ that $e_k=\frac{1}{k!}p_1(p_1-1)\dots(p_1-k+1)=\binom{p_1}{i}$ for $0\leq k\leq n+1$, as one can easily check. In particular, $\binom{p_1}{n+1}=0$, and we must have $p_1\in\{0,\dots,n\}$. We also have that the $a$s are then the roots of the polynomial $$\sum_{i=0}^n(-1)^i\binom{p_1}{i}t^{n-i}.$$

Since we are looking for solutions which nn-zero components, one can easily see that this is only possible if $p=n$, and then $a_1=\dots=a_n=1$.

share|cite|improve this answer
In a previous version of this answer, I had forgotten the $(n+1)$th equation. – Mariano Suárez-Alvarez Feb 18 '13 at 20:09

The expression $s_m:=\sum_{i=1}^n a_i^m$ is symmetric in the $a_i$, hence can be expressed as a polynomial in the elementary symmetric polynomials $\sigma_1,\ldots,\sigma_n$, which are the coefficients of $$\tag1X^n-\sigma_1 X^{n-1}+\sigma_2X^{n-2}\pm\ldots+(-1)^n\sigma_n=(X-a_1)\cdots(X-a_n)$$ Clearly, $s_1=\sigma_1$. Next we find $s_2=\sigma_1^2-2\sigma_2$. Since generally $\sigma_1^k=s_k+(\text{expression in other terms})+k!\sigma_k$ , one can show that the $\sigma_k$ are uniquely determined recursively if we are given a constant $s=s_1=\ldots=s_{n+1}$. Thus for given $s$ there is exactly one suitable polynomial $(1)$ and thus (up to permutation) exactly one solution $(a_1,\ldots,a_n)$. An obvious solution for $s=k$, $0\le k\le n$ is given by $a_1=\ldots=a_k=1$, $a_{k+1}=\ldots=a_n=0$. For other values of $s$, the solutions are not that simple and in fact involve complex numbers $a_i$.

Let's consider the simple case $n=2$. As seen above, $\sigma_1=s_1=s$, $\sigma_2=\frac12(\sigma_1^2-s_2)=\frac{s^2-s}2$, hence we need to solve $X^2-sX+\frac{s^2-s}2=0$. Of course the roots are $\frac {s\pm \sqrt{2s-s^2}}2$.

share|cite|improve this answer
For $n=2$ and $s=\dfrac12$ you get $a_1=\dfrac{1+\sqrt{3}}{4}, \ a_2=\dfrac{1-\sqrt{3}}{4}$ right? But $a_1^3+a_2^3\neq\dfrac12$. I think in your solution you are not taking into account that $s_{n+1}=s_1$. – P.. Feb 18 '13 at 20:48
Indeed his solution is essentially equal to what I had written earlier. – Mariano Suárez-Alvarez Feb 18 '13 at 20:52

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.