150 Chapter 3 Vector Spaces

Change of Basis for a General Vector Space

Everything we have done so far can easily be generalized to apply to any finite-
dimensional vector space. We begin by defining coordinate vectors for an n-dimen-
sional vector space.

Definition Let V be a vector space and let E = {v1 , v2 , . . . , vn } be an ordered basis for V . If
v is any element of V , then v can be written in the form

v = c1 v1 + c2 v2 + · · · + cn vn

where c1 , c2 , . . . , cn are scalars. Thus, we can associate with each vector v a unique
vector c = (c1 , c2 , . . . , cn )T in Rn . The vector c defined in this way is called the
coordinate vector of v with respect to the ordered basis E and is denoted [v] E . The
ci ’s are called the coordinates of v relative to E.

The examples considered so far have all dealt with changing coordinates in R2 .
Similar techniques could be used for Rn . In the case of Rn , the transition matrices will
be n × n.

EXAMPLE 5 Let
E = [v1 , v2 , v3 ] = [(1, 1, 1)T , (2, 3, 2)T , (1, 5, 4)T ]
F = [u1 , u2 , u3 ] = [(1, 1, 0)T , (1, 2, 0)T , (1, 2, 1)T ]
Find the transition matrix from E to F. If
x = 3v1 + 2v2 − v3 and y = v1 − 3v2 + 2v3
find the coordinates of x and y with respect to the ordered basis F.

Solution
As in Example 4, the transition matrix is given by
⎧ ⎫⎧ ⎫ ⎧ ⎫
⎪
⎪ 2 −1 0⎪⎪ ⎪
⎪ 1 2 1⎪⎪ ⎪
⎪ 1 1 −3 ⎪
⎪
U −1 V = ⎪⎪ −1 1 −1 ⎪⎪⎪ ⎪1 5⎪⎪ =⎪
⎪ −1 −1 0⎪⎪
⎪
⎩ ⎪
⎭⎪ ⎩
3 ⎪
⎭ ⎪⎩ ⎪
⎭
0 0 1 1 2 4 1 2 4
The coordinate vectors of x and y with respect to the ordered basis F are given by
⎧ ⎫⎧ ⎫ ⎧ ⎫
⎪
⎪ 1 1 −3 ⎪ ⎪ ⎪
⎪ 3⎪
⎪ ⎪
⎪ 8⎪⎪
[x] F = ⎪⎪
⎪
⎩
−1 −1 0⎪⎪
⎪
⎭⎩
⎪
⎪
⎪ 2⎪
⎪
⎪ = ⎪
⎪
⎪ −5 ⎪⎪
⎪
⎭ ⎩ ⎭
1 2 4 −1 3
and ⎧ ⎫⎧ ⎫ ⎧ ⎫
⎪
⎪ 1 1 −3 ⎪
⎪ ⎪
⎪ 1⎪⎪ ⎪
⎪ −8 ⎪
⎪
[y] F = ⎪
⎪ −1 −1
⎪
⎩
0⎪⎪
⎪
⎭⎪
⎪ −3 ⎪
⎪
⎩
⎪ ⎪
⎪ 2⎪
⎭=⎪
⎪ ⎩
⎪
⎪
⎭
1 2 4 2 3
The reader may verify that
8u1 − 5u2 + 3u3 = 3v1 + 2v2 − v3
−8u1 + 2u2 + 3u3 = v1 − 3v2 + 2v3
 3.5 Change of Basis 151

If V is any n-dimensional vector space, it is possible to change from one basis to

another by means of an n × n transition matrix. We will show that such a transition
matrix is necessarily nonsingular. To see how this is done, let E = {w1 , . . . , wn } and
F = {v1 , . . . , vn } be two ordered bases for V . The key step is to express each basis
vector w j as a linear combination of the vi ’s:

w1 = s11 v1 + s21 v2 + · · · + sn1 vn

w2 = s12 v1 + s22 v2 + · · · + sn2 vn
..
. (4)
wn = s1n v1 + s2n v2 + · · · + snn vn

Let v ∈ V . If x = [v] E , it follows from (4) that

v = x1 w1 + x2 w2 + · · · + xn wn

n n n
= s1 j x j v1 + s2 j x j v2 + · · · + sn j x j vn
j=1 j=1 j=1

Thus, if y = [v] F , then


n
yi = si j x j i = 1, . . . , n
j=1

and hence
y = Sx
The matrix S defined by (4) is referred to as the transition matrix. Once S has been
determined, it is a simple matter to change coordinate systems. To find the coordinates
of v = x1 w1 + · · · + xn wn with respect to {v1 , . . . , vn }, we need only calculate y = Sx.
The transition matrix S corresponding to the change of basis from {w1 , . . . , wn } to
{v1 , . . . , vn } can be characterized by the condition

Sx = y if and only if x1 w1 + · · · + xn wn = y1 v1 + · · · + yn vn (5)

Taking y = 0 in (5), we see that Sx = 0 implies that

x1 w1 + · · · + xn wn = 0

Since the wi ’s are linearly independent, it follows that x = 0. Thus, the equation
Sx = 0 has only the trivial solution and hence the matrix S is nonsingular. The inverse
matrix is characterized by the condition

S −1 y = x if and only if y1 v1 + · · · + yn vn = x1 w1 + · · · + xn wn

Accordingly, S −1 is the transition matrix used to change basis from {v1 , . . . , vn } to

{w1 , . . . , wn }.

EXAMPLE 6 Suppose that in P3 we want to change from the ordered basis [1, x, x 2 ] to the ordered
basis [1, 2x, 4x 2 − 2]. Since [1, x, x 2 ] is the standard basis for P3 , it is easier to find
152 Chapter 3 Vector Spaces

the transition matrix from [1, 2x, 4x 2 − 2] to [1, x, x 2 ]. Since

1 = 1 · 1 + 0x + 0x 2
2x = 0 · 1 + 2x + 0x 2
4x − 2 = −2 · 1 + 0x + 4x 2
2

the transition matrix is ⎧ ⎫

⎪
⎪1 0 −2 ⎪
⎪
S=⎪
⎪
⎪ 0 2 0⎪⎪
⎪
⎩ ⎭
0 0 4
The inverse of S will be the transition matrix from [1, x, x 2 ] to [1, 2x, 4x 2 − 2]:
⎧ ⎫
⎪
⎪ 1 0 12 ⎪ ⎪
⎪ ⎪
S −1 = ⎪⎪
⎪
⎪ 0 1
0 ⎪
⎪
⎪
⎪
⎩ 2
1⎭
0 0 4

Given any p(x) = a + bx + cx 2 in P3 , to find the coordinates of p(x) with respect to

[1, 2x, 4x 2 − 2], we simply multiply
⎧ ⎫ ⎧ ⎫
⎪ 1 ⎪⎧ ⎫ ⎪
⎪
⎪ 1 0 2⎪⎪ ⎪ a⎪ ⎪ c ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪ ⎪ ⎪
⎪ ⎪
⎪ ⎪ ⎪
⎪ ⎪
⎪ = ⎪ 1 ⎪ ⎪
⎪0 2 0⎪
⎪
1
⎪⎪ b ⎪
⎩ ⎭ ⎪ ⎪
⎪ b
2 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪ ⎪
⎩ 1⎭ c ⎩1 ⎪ ⎭
0 0 4 4
c

Thus,

p(x) = (a + 12 c) · 1 + ( 12 b) · 2x + 14 c · (4x 2 − 2)

We have seen that each transition matrix is nonsingular. Actually, any nonsingular
matrix can be thought of as a transition matrix. If S is an n × n nonsingular matrix and
{v1 , . . . , vn } is an ordered basis for V , then define {w1 , w2 , . . . , wn } by (4). To see that
the w j ’s are linearly independent, suppose that

n
xjwj = 0
j=1

It follows from (4) that


n n
si j x j v j = 0
i=1 j=1

By the linear independence of the vi ’s, it follows that


n
si j x j = 0 i = 1, . . . , n
j=1

or, equivalently,
Sx = 0
 3.5 Change of Basis 153

Since S is nonsingular, x must equal 0. Therefore, w1 , . . . , wn are linearly independent

and hence they form a basis for V . The matrix S is the transition matrix corresponding
to the change from the ordered basis {w1 , . . . , wn } to {v1 , . . . , vn }.
In many applied problems, it is important to use the right type of basis for the par-
ticular application. In Chapter 5, we will see that the key to solving least squares prob-
lems is to switch to a special type of basis called an orthonormal basis. In Chapter 6,
we will consider a number of applications involving the eigenvalues and eigenvectors
associated with an n × n matrix A. The key to solving these types of problems is to
switch to a basis for Rn consisting of eigenvectors of A.

SECTION 3.5 EXERCISES

1. For each of the following, find the transition matrix 7. Given
corresponding to the change of basis from {u1 , u2 } ⎧ ⎫ ⎧ ⎫ ⎧ ⎫
v1 = ⎪
⎩1⎪ v2 = ⎪
⎩2⎪ S=⎪
to {e1 , e2 }: ⎭, ⎭, ⎩3 5⎪⎭
2 3 1 −2
(a) u1 = (1, 1)T , u2 = (−1, 1)T
(b) u1 = (1, 2)T , u2 = (2, 5)T find vectors w1 and w2 so that S will be the transi-
tion matrix from {w1 , w2 } to {v1 , v2 }.
(c) u1 = (0, 1) , T
u2 = (1, 0) T
8. Given
2. For each of the ordered bases {u1 , u2 } in Exer- ⎧ ⎫ ⎧ ⎫ ⎧ ⎫
cise 1, find the transition matrix corresponding to v1 = ⎪
⎩2⎪ ⎭ , v2 = ⎪ ⎩1⎪ ⎭, S = ⎪ ⎩4 1⎪ ⎭
the change of basis from {e1 , e2 } to {u1 , u2 }. 6 4 2 1

3. Let v1 = (3, 2)T and v2 = (4, 3)T . For each or- find vectors u1 and u2 so that S will be the transi-
dered basis {u1 , u2 } given in Exercise 1, find the tion matrix from {v1 , v2 } to {u1 , u2 }.
transition matrix from {v1 , v2 } to {u1 , u2 }. 9. Let [x, 1] and [2x − 1, 2x + 1] be ordered bases for
4. Let E = [(5, 3)T , (3, 2)T ] and let x = (1, 1)T , P2 .
y = (1, −1)T , and z = (10, 7)T . Determine the (a) Find the transition matrix representing the
values of [x] E , [y] E , and [z] E . change in coordinates from [2x − 1, 2x + 1]
5. Let u1 = (1, 1, 1)T , u2 = (1, 2, 2)T , to [x, 1].
u3 = (2, 3, 4)T . (b) Find the transition matrix representing
(a) Find the transition matrix corresponding to the the change in coordinates from [x, 1] to
change of basis from {e1 , e2 , e3 } to {u1 , u2 , u3 }. [2x − 1, 2x + 1].
10. Find the transition matrix representing the change
(b) Find the coordinates of each of the following
of coordinates on P3 from the ordered basis
vectors with respect to {u1 , u2 , u3 }:
[1, x, x 2 ] to the ordered basis
(i) (3, 2, 5)T (ii) (1, 1, 2)T
(iii) (2, 3, 2) T [1, 1 + x, 1 + x + x 2 ]
6. Let v1 = (4, 6, 7)T , v2 = (0, 1, 1)T , 11. Let E = {u1 , . . . , un } and F = {v1 , . . . , vn } be two
v3 = (0, 1, 2)T , and let u1 , u2 , and u3 be the vectors ordered bases for Rn , and set
given in Exercise 5.
(a) Find the transition matrix from {v1 , v2 , v3 } to U = (u1 , . . . , un ), V = (v1 , . . . , vn )
{u1 , u2 , u3 }.
Show that the transition matrix from E to F can be
(b) If x = 2v1 + 3v2 − 4v3 , determine the coordi- determined by calculating the reduced row echelon
nates of x with respect to {u1 , u2 , u3 }. form of (V |U ).