Elementary symmetric polynomials proof

530 Views Asked by At

In the book Polynomial Invariants of Finite Groups by Larry Smith, he proved the algebraic independence of elementary symmetric polynomials as follows:

Suppose $g(e_1,…,e_n) =0$ where $ g $ is not the zero polynomial, write $g$ as the sum of $cx_1^{a_1 - a_2} x_2^{a_2 -a_3}... x_n^{a_n}$ where $a_1 \geq ... \geq a_n$. Let the leading term of $g$ be $kx_1^{a_1 - a_2} ... x_n^{a_n}$.

Then the leading term of $g(e_1,…,e_n)$ is $kx_1^{a_1} ... x_n^{a_n}$ and hence k =0.

I don’t really understand the last part, it feels to me like the author just substitute the $e_i$’s into the leading term of g. If so, why does direct substitution of $e_i$’s into the leading term of $g$ work here?

1

There are 1 best solutions below

0
On BEST ANSWER

Define an order of the monomials as follows: $x_1^{a_1} \cdots x_n^{a_n} > x_1^{b_1} \cdots x_n^{b_n}$ if and only $a_j + \cdots + a_n > b_j + \cdots + b_n$ for the first value of $j$ for which $a_j + \cdots + a_n \neq b_j + \cdots + b_n$. Call this the $S$-order. We would need to prove that this gives a well-defined order of course. This is well-defined, and there is a unique maximal element in any polynomial w.r.t the $S$-order.

With this definition if $x_1^{a_1 - a_2} \cdots x_n^{a_n}$ is the leading term of $g(x_1, \dots, x_n)$ w.r.t. the $S$-order, then $x_1^{a_1} \cdots x_n^{a_n}$ is the maximal term of $g(e_1, \dots, e_n)$ w.r.t. the lexicographical order ($L$-order). This is true simply because $a_j = (a_j - a_{j+1}) + \cdots + (a_{n-1} - a_n) + a_n$, and $x_1^{a_1} \cdots x_n^{a_n}$ is the unique maximal term of $e_1^{a_1 - a_2} \cdots e_n^{a_n}$ w.r.t the $L$-order.

More explicitly, let $x_1^{b_1} \cdots x_n^{b_n}$ be maximal w.r.t. the $L$ order. Assume there is a minimal $j$ such that $b_j > a_j$. $x_1^{b_1} \cdots x_n^{b_n}$ must occur in a polynomial $e_1^{b_1 - b_2} \cdots e_n^{b_n}$, where $x_1^{b_1 - b_2} \cdots x_n^{b_n}$ occurs in $g(x_1, \dots, x_n)$. But then $x_1^{b_1 - b_2} \cdots x_n^{b_n} > x_1^{a_1 - a_2} \cdots x_n^{a_n}$ w.r.t. the $S$-order. We conclude that $x_1^{b_1} \cdots x_n^{b_n} = x_1^{a_1} \cdots x_n^{a_n}$.