here's a sketch of how you'd do the 2nd question. The idea is to, given an nonidentity matrix A, construct a matrix B that doesn't commute with A.
Given an A, since it is not the scalar multiple of an identity,
1. there is a vector v such that Av is not the scalar multiple of v. In particular Av is linearly independent to v.
2. choose a basis of R^n that contains the 2 vectors {v,Av}
3. let B be a matrix that takes v to v, i.e. Bv=v, and takes Av to 0, B(Av)=0.
4. then verify that ABv \neq BAv and so AB \neq BA.
I haven't justified each step rigorously , but these are just direct applications of the basic theorems in LA. except for step 1, can't think of a simple reason for that.
---
if g is an element of G, then f(g) is an element of Aut(G). that is, f(g) is an automorphism. remember, automorphisms are functions.
the formula f(g)(h)=ghg^-1 is telling you that
f(g) is the automorphism on G that takes a h in G to
if the notation is confusing (as it often does when you have a function that outputs another function)then try writing
("to every
there is an associated automorphism
that acts on G according the following rule:
dont get confused, there are 2 'levels' of morphisms here, the first level is the fact that the map
is a homomorphism
, and the 2nd one is the fact that the map
is an automorphism