0

The following problem appeared on a problem set I'm working on: "Let $A$ be a finite-dimensional associative algebra with $1$ over a field $k$. Show that an element $a$ in $A$ has a multiplicative inverse if $a$ is not a zero divisor" (emphasis mine). I'd appreciate if y'all could check my solution to see that it's correct.

Let $a \in A$ not be a zero divisor. Then the map $T : A \to A$ given by $T : x \mapsto ax$ is an injective linear transformation from a finite-dimensional vector space $A$ to itself, as $\ker(T) = \{ x \in A : ax = 0 \} = \{ 0 \}$ by the assumption that $a$ isn't a zero divisor. Thus $T$ must also be onto, so in particular there exists a unique $b \in A$ such that $ab = 1$. Let us fix this $b$.

We know that $b$ is a right-inverse to $A$, so must show that it's also a left-inverse, i.e. that $ba = 1$. First, we wanna show that a right-inverse to $b$ exists, for which it'll suffice to show that $b$ is not a zero-divisor. If it were the case that $bx = 0$, then we'd have $a (bx) = (ab) x = 0$. But $(ab) x = 1x = x$, so $x = 0$. Thus we know that $b$ is not a zero-divisor, so there exists a unique $c \in A$ such that $bc = 1$.

It now remains to show that $a = c$. This follows as \begin{align*} a(bc) & = (ab) c \\ = a 1 & = 1 c \\ = a & = c . \end{align*} Thus $ab = ba = 1$, so $b$ can rightly be called a multiplicative inverse to $a$.

Is my solution correct? Thanks!

AJY
  • 9,086
  • Granted that $T: A \to A$ is surjective and injective, it has a right inverse $S$, which is a linear endomorphism; but how do we know $Sx = bx$ for some $b \in A$? – Robert Lewis Jul 13 '18 at 00:39
  • Congratulations, your proof is equivalent to several on the site. You did a pretty good job writing it up, but next time you should do things differently. You should search the site for your question first. You may, quite often, discover your question already exists. To achieve your aim you can do one of two things: 1) compare your solution to existing solutions, and be satisfied that in fact others have done the same thing as you, so you are correct; 2) post your own version as a solution to the question, so you get feedback. But please do not needlessly duplicate content. – rschwieb Jul 13 '18 at 13:35
  • Also be aware that what I'm describing above isn't written in stone. You may, in fact, have a very specific question about a step in a proof, and that may distinguish your question from duplicates. Please use your best judgement and ask yourself whether or not a new post would actually contribute something. – rschwieb Jul 13 '18 at 13:37

0 Answers0