I'm having troubles in understanding a rapidly solved exercise, which is actually a sort of chain of consequences. It says: be $A; B$ two matrices $n\times n$ over a field $K$ and be $F_A, F_B$ their associated linear transformations. Then:
- $AB$ is invertible, then $A$ and $B$ are invertible. $\color{blue}{\text{This is ok}}$.
- $AB$ is invertible iff $F_{AB}: K^n \to K^n$ is an isomorphism. $\color{blue}{\text{This is ok}}$ (where $F_{AB} = F_A \circ F_B$).
- Having $F_{AB}$ then $F_B$ is injective and $F_A$ is surjective. $\color{red}{\text{This is not ok}}$.
Here I have problems in understanding. What does ensure me that if $F_{AB}$ is invertible, then $F_B$ is injective and $F_A$ is surjective? Why can't I have other combinations like the opposite one?
The only possible explanation I managed to give is this: in order for $F_{AB}$ to exist, we need something like $F_A: K^n \to K^p$ and $F_B: K^q \to K^n$, in this way $F_{AB}: K^p \to K^n$. This actually requires (for our purposes) $p = q = n$ $\color{red}{\text{Is this right?}}$.
In this way if we work on $K^n$ then $F_B$ is surely surjective because the codomain and the image are the same. I don't understand why $F_A$ is injective though. I'm losing myself in a glass of water perhaps...