The function φβ : V → F n which sends v to [v]β is

an isomorphism.
Although an isomorphism doesn’t mean that V
is identical to F n , it does mean that coordinates
work exactly the same.
One place where this can be confusing is when
Coordinates V is F n , but the basis β is not the standard ba-
Math 130 Linear Algebra sis. Having two bases for the same vector space is
D Joyce, Fall 2015 confusing, but sometimes useful.

Throughout this discussion, F refers to a fixed

field. In application, F will usually be R. Matrices for transformations V → W . So far,
We’ve seen how each linear transformation T : we can represent a transformation T : F n → F m
F n → F m corresponds to an m × n matrix A. The by an m × n matrix A. The entries for A were
matrix A is the standard matrix representing T and determined by what T did to the standard basis of
its j th column consists of the coordinates of the vec- F n . The coordinates of T (ej ) were placed in the
tor T (ej ) where ej is the j th vector in the standard j th column of A. It followed that for any vector v
basis  of F n . in F n , the ith coordinate of it’s image w = T (v)
was
We’ll use this connection as we derive algebraic
operations on linear transformations to translate n
X
those operations to algebraic operations on matri- wi = Ai1 v1 + Ai2 v2 + · · · + Aij vn = Aij vj
ces. j=1
Before we do that, however, we should see that
we can use use matrices to represent linear trans- We can represent a transformation T : V → W
formations between other vector spaces so long as by a matrix as well, but it will have to be relative
we’ve got coordinates for them. to one basis β for V and another basis γ for W .
Suppose that β = (b1 , b2 , . . . , bn ) is an ordered
basis for V and γ = (c1 , c2 , . . . , cm ) is an ordered
Extending the standard matrix to transfor-
basis for W . We’ll put the γ-coordinates of the
mations T : V → W . Coordinates on vector
T (bj ) in the j th column of A, just like we did before.
spaces are determined by bases for those spaces.
If we need to indicate the bases that were used to
Recall that a basis β = (b1 , b2 , . . . , bn ) of a vec-
create the matrix A, we’ll write A = [T ]γβ .
tor space V sets up a coordinate system on V . Each
What we’ve actually done here was that we used
vector v in V can be expressed uniquely as a linear
the isomorphisms φβ : V → Rn and φγ : V → Rm
combination of the basis vectors
to transfer the linear transformation T : V → W
to a linear transformation T 0 : Rn → Rm . Define
v = v1 b1 + v2 b2 + · · · + vn bn .
T 0 as φγ ◦ T ◦ φ−1 β .

The coefficients in that linear combination form the

coordinate vector [v]β relative to β T -
V W
 
v1 φβ φγ
 v2 
[v]β =  .. 
 
.
? T 0 - ?m
Rn R
vn

1
Whereas T describes the linear transformation where a linear transformation T corresponds to the
V → W without mentioning the bases or coor- matrix A = [T ]γβ . The entries Aij of A describe how
dinates, T 0 describes the linear transformation in to express T evaluated at the β-basis vector bj as
terms of coordinates. We used those coordinates to a linear combination of the γ-basis vectors.
describe the entries of the matrix A = [T ]γβ . X
(bj ) = Aij ci

The vector space of linear transformations We’ll use this bijection φγβ to define addition and
HomF (V, W ). Let’s use the notation HomF (V, W ) scalar multiplication on Mm×n , and when we do
for the set of linear transformations V → W , or that, φγβ will be an isomorphism of vector spaces.
more simply Hom(V, W ) when there’s only one field First consider addition. Given T : V → W and
under consideration. S : V → W . Let A = [T ]γβ be the standard matrix
This is actually a vector space itself. If you have for T , and let B = [S]γβ be the standard matrix for
two transformations, S : V → W and T : V → W , S. Then
X
then define their sum by T (bj ) = Aij ci , and
i
(S + T )(v) = S(v) + T (v). X
S(bj ) = Bij ci . Therefore
Check that this sum is a linear transformation by i
X
verifying that (T + S)(bj ) = (Aij + Bij )ci .
(1) for two vectors u and v in V , i
Thus, the entries in the standard matrix for T + S
(S + T )(u + v) = S(u + v) + S(u + v), and are sums of the corresponding entries of the stan-
dard matrices for T and S. We now know how we
(2) for a vector v and a scalar c,
should define addition of matrices.
(S + T )(cv) = c(S + T )(v). Definition 1. We define addition of two m × n
matrices coordinatewise: (A + B)ij = Aij + Bij .
Next, define scalar multiplication on Hom(V, W ) by
Next, consider scalar multiplication. Given a
(cS)(v) = c(S)(v) scalar c and a linear transformation T : V → W ,
let A be the standard matrix for T . Then
and show that’s a linear transformation.
X
T (bj ) = Aij ci . Therefore
There are still eight axioms for a vector space i
that need to be checked to show that with these two X
(cT )(bj ) = cAij ci .
operations Hom(V, W ) is, indeed, a vector space.
i
But note how the operations of addition and scalar
Thus, the entries in the standard matrix for cT are
multiplication were both defined by those opera-
each c times the corresponding entries of the stan-
tions on W . Since the axioms hold in W , they’ll
dard matrice for T . We now know how we should
also hold in Hom(V, W ).
define scalar multiplication of matrices.
The vector space of matrices Mm×n (F ). Definition 2. We define scalar multiplication of a
Mm×n (F ) is the set of m × n matrices with entries scalar c and an m × n matrix A coordinatewise:
in the field F . We’ll drop the subscript F when the (cA)ij = cAij .
field is understood. With these definitions of addition and scalar mul-
When V and W are endowed with bases β and tiplication, Mm×n (F ) becomes a vector space over
'
γ, we have a bijection φγβ : Hom(V, W ) → Mm×n F , and it is isomorphic to HomF (V, W ).

2
Invariants. Note that the isomorphism

Hom(V, W ) ∼
= Mm×n

depends on the ordered bases you choose. With

different β and γ you’ll get different matrices for
the same transformation.
In some sense, the transformation holds intrin-
sic information, while the matrix is a description
which varies depending on the bases you happen to
choose.
When you’re looking for properties of the trans-
formation, they shouldn’t be artifacts of the chosen
bases but properties that hold for all bases.
When we get to looking for properties of trans-
formations, we’ll make sure that they’re invariant
under change of basis.

Math 130 Home Page at

[Link]