a matrix A is a rectangular array of field elements (m rows, n columns).
a linear operator (transformation) is a linear function from one vector space to another.
but...since we can write vectors as an array (1 row, or 1 column, take your pick), once we choose a basis (so the entries in the array are the coordinates of our vector in our basis), it seems that since matrices turn (via matrix multiplication, if we multiply the vectors on the right side) nx1 arrays into mx1 arrays (that is, n-vectors into m-vectors), perhaps we can turn a linear operator into a matrix.
that is, we want to turn a linear function into a collection of numbers. HOW we do so, depends on our choice of bases for the vector spaces involved (different bases give different matrices).
what you haven't told us, is what spaces T operates on, and what field you are using. there's actually lots of possible fields, although most beginning courses stick to the real numbers, because they are very familiar. my guess is that:
T:R2→R2, which is a very important piece of information that you have not told us.
what size matrix takes a 2-vector to a 2-vector? well, it must be 2x2. let's do some detective work.
since [T]β is a 2x2 real matrix, [T]β =
\begin{bmatrix}a&b\\c&d\end{bmatrix}
what we need to do is figure out a,b,c and d. but we need to use the basis β somehow. and this is where the problem is sort of sneaky. for example (x1,x2) is already "coordinatized" by a basis already, but this basis is "invisible", because it's the standard basis for R2. the upshot of this, is that x1 and x2 are "the wrong numbers" (coordinates), because we need to work with the basis β.
we need to find the β-coordinates of (x1,x2).
let's do a somewhat easier problem first. we know that if e1 = (1,0), and e2 = (0,1), then (x1,x2) = x1e1 + x2e2. this is how linear combinations are turned into coordinate arrays. so what we want to do is write:
e1 = (1,0) = a1(2,1) + a2(1,0).
since β is a basis for R2, there's only ONE way to do this. and it's fairly obvious that a1 = 0, a2 = 1. that is:
e1 = [0,1]β. next we want to write:
e2 = (0,1) = b1(2,1) + b2(1,0) = (2b1+b2,b1).
again, it's fairly obvious that b1 = 1, so b2 must be -2. so now we know that e2 = [1,-2]β.
now we are in a position to write (x1,x2) in β-coordinates:
(x1,x2) = x1e1 + x2e2 = x1[0,1]β + x2[1,-2]β = [x2,x1-2x2]β.
now [1,0]β = 1(2,1) + 0(0,1) = (2,1). and by definition (or just by doing the matrix multiplication, writing [1,0]β as a column vector), the first column of [T]β is
[T]β([1,0]β)T.
but if [T]β is really supposed to represent the linear transformation T, then this ought to be [T(2,1)]β. can you continue?