Professional Documents
Culture Documents
(1)
T (u + v) = T (u) + (v)
and (2)
T (αu) = αT (u).
These two conditions can be rolled into the single requirement that
Note that we are writing T (x, y) instead of the more formal (and awkward)
T (< x, y >). It is understood that T (x, y) is the image of < x, y > under the
mapping T .
It is routine to verify that T is a linear transformation.
53
Every linear transformation T : Rn → Rm must map the zero vector On of
R to the zero vector Om of Rm . To see why this is true, use the linearity of T
n
to write
T (On ) = T (On + On ) = T (On ) + T (On ),
implying that
T (On ) = Om .
However, a linear transformation may also take nonzero vectors to the zero
vector. For example, the linear transformation defined by
T (x, y) = (x − y, 0),
Example 2 Let
T (x, y) =< x − y, 0, 0 > .
Then T is a linear transformation from R2 to R3 . T is not one-to-one. For
example,
T (1, 1) = T (2, 2) =< 0, 0, 0 > .
In fact, T (x, x) =< 0, 0, 0 > for every number x. T maps infinitely many vectors
to the origin in R3 .
T is also not onto R3 , because no vector in R3 with a nonzero second or
third component is the image of any vector in R2 under T .
Example 3 Let S : R3 → R2 be defined by
54
But S is not one-to-one, because we have just seen that many different 2 -
vectors can map to the same 3 - vector.
There is an efficient test to tell whether a linear transformation is one-to-
one. We know that every linear transformation maps the zero vector to the zero
vector. The transformation is one-to-one exactly when this is the only vector
mapping to the zero vector.
T (u) = Om
T (u) = T (v).
T (u) − T (v) = T (u − v) = Om .
u − v = On .
Then u = v, so T is one-to-one.
Example 4 Let
T (x, y) =< x, y, 0, 0 > .
This mapping is one-to-one. If T (x, y) = T (a, b), then
55
Then
< x+y+2z+8w, y−z, x−w, x−y+z−w, 5x+5y−z, 0, 0 >=< 0, 0, 0, 0, 0, 0, 0 > .
Then
x + y + 2z + 8w = 0,
y−z = 0,
x−w = 0,
x−y+z−w = 0,
5x + 5y − z = 0.
From the second and third equations,
y = z and x = w.
Then, from the first equation,
3z + 9w = 0,
so z = −3w, and then y = z = −3w also.
Now the fourth equation gives us w = 0. Then x = y = z = w = 0, and
< x, y, z, w >= O4 , so by Theorem 1, T is one-to-one.
56
In this way we can compute T (V) as the matrix product of AT with the column
matrix of the components of V.
so
1 −1
AT = 0 0 .
0 0
Now observe that
x−y
x
AT = 0 ,
y
0
giving the coordinates of T (x, y) with respect to the standard basis for R3 . We
can therefore compute the coordinates of T (x, y) as a matrix product.
Example 7 In Example 5 we had
Then
1 −1 2 8
0 1 −1 0
1 0
0 −1
0 1
AT = 0 4.
5 5 −1 0
0 0 0 0
0 0 0 0
We obtain T (x, y, z, w) as the matrix product
x
y
AT z .
AT depends on the basis chosen, because the matrix elements are coordinates
of the images of the standard basis vectors for Rn with respect to the standard
basis vectors for Rm . If we use different bases, we obtain a different matrix,
which would still represent the linear transformation, as long as all vectors are
57
written in terms of the bases being used Different matrix representations formed
in this way for the same linear transformation are related in a special way, but
we will not pursue issue this here.
The matrix representation AT of T enables us to pose questions about T in
terms of linear systems of equations, about which we know a good deal.
First, T : Rn → Rm is one-to-one exactly when T (X) =< 0, 0, · · · , 0 > in
R implies that X =< 0, 0, · · · , 0 > in Rn . This is equivalent to the assertion
m
AT X = O.
The null space of T is exactly the solution space of the homogeneous linear
system AT (X) = O. This solution space has dimension
58
bodies of information. However, matrices are often better suited to computation,
particularly if a software package is being used.
Problems
In each of Problems 1 through 10, determine whether or not the given func-
tion is a linear transformation. If it is, write the matrix representation of T
(using the standard bases) and determine if T is onto and if T is one-to-one.
Also determine the null space of T and its dimension.
59