Proof of the Normal Basis Theorem in Lang's *Algebra*

791 Views Asked by At

Theorem 13.1 in Chapter 6 of Lang's Algebra (Revised Third Edition) states the following:

Theorem. Let $K:k$ be a finite Galois extension of degree $n$. Let $\sigma_1, \ldots, \sigma_n$ be all the elements of the Galois group of $G$ of $K:k$. Then there is an element $w\in K$ such that $\sigma_1(w), \ldots, \sigma_n(w)$ forms a $k$-basis for $K$.

The proof given is as follows:

Assume that $k$ is infinite. For each $\sigma\in G$, let $X_\sigma$ be a variable, and let $t_{\sigma, \tau}=X_{\sigma^{-1}\tau}$. Let $X_i=X_{\sigma_i}$. Define a polynomial $f(X_1, \ldots, X_n)\in K[X_1, \ldots, X_n]$ as $$f(X_1, \ldots, X_n) = \det(t_{\sigma_i, \sigma_j})$$ Then $f$ is not identically zero, since substituting $1$ in place if $X_{\text{id}}$ and $0$ in place of all other variables yields a non-zero value.

Now here is what I am unable to understand:

Hence $f(\sigma_1(x), \ldots, \sigma_n(x))$ will not vanish for all $x\in K$. (That is, there exists $x\in K$ for which this does not vanish).

Can somebody please explain how the last line is inferred? Thanks.

1

There are 1 best solutions below

2
On BEST ANSWER

If the 3rd edition is like the 1st, then you will see that the preceding theorem establishes that, over an infinite field $K$, any finite group $\{\sigma_1(x),\ldots,\sigma_n(x)\}$ of distinct automorphisms of $K$ consists of functions that are algebraically independent over $K$. Thus if $f(x_1,\ldots,x_n)\in K[x_1,\ldots,x_n]$ is not the zero polynomial, then $f(\sigma_1(x),\ldots,\sigma_n(x))$ will not be the zero function.