Let $A$ and $B$ be unital Banach algebras with $\phi:A \to B$ a linear mapping.
I am looking specifically at linear mappings which preserves invertiblity and noticed that different papers use different definitions to define preserves invertibility. The one that comes up most, is the following:
Sp$\phi(x) \subset $ Sp$x$ for every $x \in A$.
I am trying to see why this is true though, i.e. why this is equivalent to $x$ invertible $\implies \phi(x)$ invertible.
Suppose that $x \in A$ is invertible and that $\phi$ preserves invertibility with $\phi(x^{-1})=\phi(x)^{-1}$.
Let $\lambda \in $ Sp$\phi(x)$, then $\phi(x) - \lambda\textbf{1}_B$ is not invertible in $B$.
I need to show that $\lambda \in $ Sp$x$. But I am not sure how to do so. I have thought of the following:
Since $\phi(x) - \lambda\textbf{1}_B$ is not invertible in $B$ we have, in particular, that $\phi(x)\neq \lambda$. But I am not sure if this will take me anywhere.
Can anyone please point me in the right direction?
In proofs such as this, it is easier to work with the contrapositive.
If $x-\lambda$ is invertible, then $\phi(x-\lambda)=\phi(x)-\lambda$ is invertible. Thus $\mathbb C\setminus\sigma(x)\subset\mathbb C\setminus\sigma(\phi(x))$, and therefore $\sigma(\phi(x))\subset\sigma(x)$.
On the other hand, if $\sigma(\phi(x))\subset\sigma(x)$ for all $x\in A$, then supposing $x\in A$ is invertible means $0\notin\sigma(x)$. Thus $0\notin\sigma(\phi(x))$, and so $\phi(x)$ is invertible.