2

Let $A$ be a central simple algebra over a field $F$.

Let $K$ be a maximal subfield of $A$ with $[K:F]=n$ and assume $K$ is Galois extension of $F$.

Let $\sigma_1,\sigma_2,\cdots,\sigma_n$ be all the Galois automorphisms of $K$ over $F$. Then By Skolen-noether theorem, there exists invertible elements $x_i$ in $A$ such that $\sigma_i$ (on $K$) are simply conjugation (on $K$) by $x_i$.

Claim: $x_1k_1+x_2k_2 + \cdots + x_nk_n=0$ for $k_i\in K$ implies $k_i=0$ for all $i$.

I proved this for a simple case, as follows.

If $x_1k_1+ x_2k_2=0$ for $k_1,k_2\neq 0$, then we see that $x_1$ and $x_2$ differ by an element of $K^*$, i.e. $x_1=x_2(-k_2k_1^{-1})$. Then on $K$, conjugation by $x_1$ and conjugation by $x_2$ coincide, i.e. $\sigma_1=\sigma_2$, contradiction.

I could not proceed in the case $x_1k_1+x_2k_2+x_3k_3=0$ to get a contradiction. How to proceed for this case, and in particular for claim? Any hint is sufficient.

I was thinking to use $F$-independence of $\sigma_i$'s (Dedekind's theorem), but I didn't get direction.

Beginner
  • 10,836
  • 1
    I think you can do this with ideas similar to the proof of $F$-independence of the Galois group. If you multiply the given linear dependence relation from the left by an element $k\in K$, and you move the $k$s to other side of $x_i$s you get a modified relation like $$x_1\sigma_1(k)k_1+x_2\sigma_2(k)k_2+\cdots+x_n\sigma_n(k)k_n=0.$$ Then work the usual way. Assume that the linear dependency relation involves a minimal number of $x_i$, pick a useful $k$ for the above, eliminate one of the variables etc. – Jyrki Lahtonen Jan 22 '19 at 20:09

0 Answers0