i i “main”

2007/2/16
page 258

i i

258 CHAPTER 4 Vector Spaces

23. Show that the set of all solutions to the nonhomoge- and let
neous differential equation
S1 + S2 = {v ∈ V :




y + a1 y + a2 y = F (x), v = x + y for some x ∈ S1 and y ∈ S2 } .

where F (x) is nonzero on an interval I , is not a sub-

space of C 2 (I ). (a) Show that, in general, S1 ∪ S2 is not a subspace
of V .
24. Let S1 and S2 be subspaces of a vector space V . Let
(b) Show that S1 ∩ S2 is a subspace of V .
S1 ∪ S2 = {v ∈ V : v ∈ S1 or v ∈ S2 }, (c) Show that S1 + S2 is a subspace of V .
S1 ∩ S2 = {v ∈ V : v ∈ S1 and v ∈ S2 },

4.4 Spanning Sets

The only algebraic operations that are defined in a vector space V are those of addition
and scalar multiplication. Consequently, the most general way in which we can combine
the vectors v1 , v2 , . . . , vk in V is

c1 v1 + c2 v2 + · · · + ck vk , (4.4.1)

where c1 , c2 , . . . , ck are scalars. An expression of the form (4.4.1) is called a linear

combination of v1 , v2 , . . . , vk . Since V is closed under addition and scalar multiplica-
tion, it follows that the foregoing linear combination is itself a vector in V . One of the
questions we wish to answer is whether every vector in a vector space can be obtained
by taking linear combinations of a finite set of vectors. The following terminology is
used in the case when the answer to this question is affirmative:

DEFINITION 4.4.1
If every vector in a vector space V can be written as a linear combination of v1 ,
v2 , . . . , vk , we say that V is spanned or generated by v1 , v2 , . . . , vk and call the
set of vectors {v1 , v2 , . . . , vk } a spanning set for V . In this case, we also say that
{v1 , v2 , . . . , vk } spans V .

This spanning idea was introduced in the preceding section within the framework
of differential equations. In addition, we are all used to representing geometric vectors
in R3 in terms of their components as (see Section 4.1)

v = ai + bj + ck,

where i, j, and k denote the unit vectors pointing along the positive x-, y-, and z-axes,
respectively, of a rectangular Cartesian coordinate system. Using the above terminology,
we say that v has been expressed as a linear combination of the vectors i, j, and k, and
that the vector space of all geometric vectors is spanned by i, j, and k.
We now consider several examples to illustrate the spanning concept in different
vector spaces.

Example 4.4.2 Show that R2 is spanned by the vectors

v1 = (1, 1) and v2 = (2, −1).

i i

i i
 i i “main”
2007/2/16
page 259

i i

4.4 Spanning Sets 259

Solution: We must establish that for every v = (x1 , x2 ) in R2 , there exist constants
c1 and c2 such that

v = c1 v1 + c2 v2 . (4.4.2)
y
That is, in component form,
(4/3, 4/3)
(1, 1) (x1 , x2 ) = c1 (1, 1) + c2 (2, −1).
(2, 1)
v1
v
Equating corresponding components in this equation yields the following linear system:
x
c1 + 2c2 = x1 ,
(2/3,1/3) c1 − c2 = x2 .
v2 (2,1)
In this system, we view x1 and x2 as fixed, while the variables we must solve for are c1
Figure 4.4.1: The vector and c2 . The determinant of the matrix of coefficients of this system is
v = (2, 1) expressed as a linear
combination of v1 = (1, 1) and 1 2
v2 = (2, −1). = −3.
1 −1

Since this is nonzero regardless of the values of x1 and x2 , the matrix of coefficients is
invertible, and hence for all (x1 , x2 ) ∈ R2 , the system has a (unique) solution according
to Theorem 2.6.4. Thus, Equation (4.4.2) can be satisfied for every vector v ∈ R2 , so the
given vectors do span R2 . Indeed, solving the linear system yields

y c1 = 13 (x1 + 2x2 ), c2 = 13 (x1 − x2 ).

c2v2 Hence,
v2

c2 (x1 , x2 ) = 13 (x1 + 2x2 )v1 + 13 (x1 − x2 )v2 .
v1
v2 c 1
v
For example, if v = (2, 1), then c1 = and c2 = 13 , so that v = 43 v1 + 13 v2 . This is
v1 4
c1v1 3
x illustrated in Figure 4.4.1.

Figure 4.4.2: Any two More generally, any two nonzero and noncolinear vectors v1 and v2 in R2 span R2 ,
noncollinear vectors in R2 span since, as illustrated geometrically in Figure 4.4.2, every vector in R2 can be written as a
R2 . linear combination of v1 and v2 .

Example 4.4.3 Determine whether the vectors v1 = (1, −1, 4), v2 = (−2, 1, 3), and v3 = (4, −3, 5)
span R3 .
Solution: Let v = (x1 , x2 , x3 ) be an arbitrary vector in R3 . We must determine
whether there are real numbers c1 , c2 , c3 such that

v = c1 v1 + c2 v2 + c3 v3 (4.4.3)

or, in component form,

(x1 , x2 , x3 ) = c1 (1, −1, 4) + c2 (−2, 1, 3) + c3 (4, −3, 5).

Equating corresponding components on either side of this vector equation yields

c1 − 2c2 + 4c3 = x1 ,
−c1 + c2 − 3c3 = x2 ,
4c1 + 3c2 + 5c3 = x3 .

i i

i i
 i i “main”
2007/2/16
page 260

i i

260 CHAPTER 4 Vector Spaces

Reducing the augmented matrix of this system to row-echelon form, we obtain

 
1 −2 4 x1
 0 1 −1 −x1 − x2 .
0 0 0 7x1 + 11x2 + x3

It follows that the system is consistent if and only if x1 , x2 , x3 satisfy

7x1 + 11x2 + x3 = 0. (4.4.4)

Consequently, Equation (4.4.3) holds only for those vectors v = (x1 , x2 , x3 ) in R3

whose components satisfy Equation (4.4.4). Hence, v1 , v2 , and v3 do not span R3 .
Geometrically, Equation (4.4.4) is the equation of a plane through the origin in space,
and so by taking linear combinations of the given vectors, we can obtain only those
vectors which lie on this plane. We leave it as an exercise to verify that indeed the three
given vectors lie in the plane with Equation (4.4.4). It is worth noting that this plane
forms a subspace S of R3 , and that while V is not spanned by the vectors v1 , v2 , and v3 ,
S is.

The reason that the vectors in the previous example did not span R3 was because
they were coplanar. In general, any three noncoplanar vectors v1 , v2 , and v3 in R3
span R3 , since, as illustrated in Figure 4.4.3, every vector in R3 can be written as a linear
combination of v1 , v2 , and v3 . In subsequent sections we will make this same observation
from a more algebraic point of view.

z
v
c1v1  c2v2  c3v3

v1
c3v3
c1v1
v3 c2v2 v2
y

Figure 4.4.3: Any three noncoplanar vectors in R3 span R3 .

Notice in the previous example that the linear combination (4.4.3) can be written as
the matrix equation
Ac = v,
where the columns of A are the given vectors v1 , v2 , and v3 : A = [v1 , v2 , v3 ]. Thus,
the question of whether or not the vectors v1 , v2 , and v3 span R3 can be formulated as
follows: Does the system Ac = v have a solution c for every v in R3 ? If so, then the
column vectors of A span R3 , and if not, then the column vectors of A do not span R3 .
This reformulation applies more generally to vectors in Rn , and we state it here for the
record.

Theorem 4.4.4 Let v1 , v2 , . . . , vk be vectors in Rn . Then {v1 , v2 , . . . , vk } spans Rn if and only if, for
the matrix A = [v1 , v2 , . . . , vk ], the linear system Ac = v is consistent for every v in
Rn .

i i

i i
 i i “main”
2007/2/16
page 261

i i

4.4 Spanning Sets 261

Proof Rewriting the system Ac = v as the linear combination

c1 v1 + c2 v2 + · · · + ck vk = v,
we see that the existence of a solution (c1 , c2 , . . . , ck ) to this vector equation for each v
in Rn is equivalent to the statement that {v1 , v2 , . . . , vk } spans Rn .
Next, we consider a couple of examples involving vector spaces other than Rn .

Example 4.4.5 Verify that

       
1 0 1 1 1 1 11
A1 = , A2 = , A3 = , A4 =
0 0 0 0 10 11
span M2 (R).
Solution: An arbitrary vector in M2 (R) is of the form
 
a b
A= .
c d
If we write
c1 A1 + c2 A2 + c3 A3 + c4 A4 = A,
then equating the elements of the matrices on each side of the equation yields the system
c1 + c2 + c3 + c4 = a,
c2 + c3 + c4 = b,
c3 + c4 = c,
c4 = d.
Solving this by back substitution gives
c1 = a − b, c2 = b − c, c3 = c − d, c4 = d.
Hence, we have
A = (a − b)A1 + (b − c)A2 + (c − d)A3 + dA4 .
Consequently every vector in M2 (R) can be written as a linear combination of A1 , A2 ,
A3 , and A4 , and therefore these matrices do indeed span M2 (R).

Remark The most natural spanning set for M2 (R) is

       
1 0 0 1 0 0 0 0
, , , ,
0 0 0 0 1 0 0 1
a fact that we leave to the reader as an exercise.

Example 4.4.6 Determine a spanning set for P2 , the vector space of all polynomials of degree 2 or less.
Solution: The general polynomial in P2 is
p(x) = a0 + a1 x + a2 x 2 .
If we let
p0 (x) = 1, p1 (x) = x, p2 (x) = x 2 ,
then
p(x) = a0 p0 (x) + a1 p1 (x) + a2 p2 (x).
Thus, every vector in P2 is a linear combination of 1, x, and x 2 , and so a spanning set
for P2 is {1, x, x 2 }. For practice, the reader might show that {x 2 , x + x 2 , 1 + x + x 2 } is
another spanning set for P2 , by making the appropriate modifications to the calculations
in this example.

i i

i i
 i i “main”
2007/2/16
page 262

i i

262 CHAPTER 4 Vector Spaces

The Linear Span of a Set of Vectors

Now let v1 , v2 , . . . , vk be vectors in a vector space V . Forming all possible linear combina-
tions of v1 , v2 , . . . , vk generates a subset of V called the linear span of {v1 , v2 , . . . , vk },
denoted span{v1 , v2 , . . . , vk }. We have

span{v1 , v2 , . . . , vk } = {v ∈ V : v = c1 v1 + c2 v2 + · · · + ck vk , c1 , c2 , . . . , ck ∈ F }.
(4.4.5)

For example, suppose V = C 2 (I ), and let y1 (x) = sin x and y2 (x) = cos x. Then

span{y1 , y2 } = {y ∈ C 2 (I ) : y(x) = c1 cos x + c2 sin x, c1 , c2 ∈ R}.

From Example 1.2.16, we recognize y1 and y2 as being nonproportional solutions to

the differential equation y

+ y = 0. Consequently, in this example, the linear span of
the given functions coincides with the set of all solutions to the differential equation
y

+ y = 0 and therefore is a subspace of V . Our next theorem generalizes this to show
that any linear span of vectors in any vector space forms a subspace.

Theorem 4.4.7 Let v1 , v2 , . . . , vk be vectors in a vector space V . Then span{v1 , v2 , . . . , vk } is a subspace

of V .

Proof Let S = span{v1 , v2 , . . . , vk }. Then 0 ∈ S (corresponding to c1 = c2 = · · · =

ck = 0 in (4.4.5)), so S is nonempty. We now verify closure of S under addition and
scalar multiplication. If u and v are in S, then, from Equation (4.4.5),

u = a1 v1 + a2 v2 + · · · + ak vk and v = b1 v1 + b2 v2 + · · · + bk vk ,

for some scalars ai , bi . Thus,

u + v = (a1 v1 + a2 v2 + · · · + ak vk ) + (b1 v1 + b2 v2 + · · · + bk vk )
= (a1 + b1 )v1 + (a2 + b2 )v2 + · · · + (ak + bk )vk
= c1 v1 + c2 v2 + · · · + ck vk ,

where ci = ai + bi for each i = 1, 2, . . . , k. Consequently, u + v has the proper form

for membership in S according to (4.4.5), so S is closed under addition. Further, if r is
any scalar, then

ru = r(a1 v1 + a2 v2 + · · · + ak vk )
= (ra1 )v1 + (ra2 )v2 + · · · + (rak )vk
= d1 v1 + d2 v2 + · · · + dk vk ,

where di = rai for each i = 1, 2, . . . , k. Consequently, ru ∈ S, and so S is also closed

under scalar multiplication. Hence, S = span{v1 , v2 , . . . , vk } is a subspace of V .

Remarks

1. We will also refer to span{v1 , v2 , . . . , vk } as the subspace of V spanned by

v1 , v2 , . . . , vk .
2. As a special case, we will declare that span(∅) = {0}.

i i

i i
 i i “main”
2007/2/16
page 263

i i

4.4 Spanning Sets 263

Example 4.4.8 If V = R2 and v1 = (−1, 1), determine span{v1 }.

Solution: We have
span{v1 } = {v ∈ R2 : v = c1 v1 , c1 ∈ R}
= {v ∈ R2 : v = c1 (−1, 1), c1 ∈ R}
= {v ∈ R2 : v = (−c1 , c1 ), c1 ∈ R}.
Geometrically, this is the line through the origin with parametric equations x = −c1 ,
y = c1 , so that the Cartesian equation of the line is y = −x. (See Figure 4.4.4.)

y
(—c1, c1)
c1v1
The subspace of ⺢2 spanned
by the vector v1
(1, 1)
(1, 1)
v1
x

Figure 4.4.4: The subspace of R2 spanned by v1 = (−1, 1).

Example 4.4.9 If V = R3 , v1 = (1, 0, 1), and v2 = (0, 1, 1), determine the subspace of R3 spanned by
v1 and v2 . Does w = (1, 1, −1) lie in this subspace?
Solution: We have
span{v1 , v2 } = {v ∈ R3 : v = c1 v1 + c2 v2 , c1 , c2 ∈ R}
= {v ∈ R3 : v = c1 (1, 0, 1) + c2 (0, 1, 1), c1 , c2 ∈ R}
= {v ∈ R3 : v = (c1 , c2 , c1 + c2 ), c1 , c2 ∈ R}.
Since the vector w = (1, 1, −1) is not of the form (c1 , c2 , c1 + c2 ), it does not lie in
span{v1 , v2 }. Geometrically, span{v1 , v2 } is the plane through the origin determined by
the two given vectors v1 and v2 . It has parametric equations x = c1 , y = c2 , z = c1 + c2 ,
which implies that its Cartesian equation is z = x + y. Thus, the fact that w is not
in span{v1 , v2 } means that w does not lie in this plane. The subspace is depicted in
Figure 4.4.5.

z
The subspace of ⺢3 spanned
by v1
(1, 0, 1), v2
(0, 1, 1)

v2
v1

x w
(1, 1, 1) does not

lie in span{v1, v2}

Figure 4.4.5: The subspace of R3 spanned by v1 = (1, 0, 1) and v2 = (0, 1, 1) is the plane
with Cartesian equation z = x + y.

i i

i i
 i i “main”
2007/2/16
page 264

i i

264 CHAPTER 4 Vector Spaces

Example 4.4.10 Let

     
1 0 0 1 0 0
A1 = , A2 = , A3 =
0 0 1 0 0 1
in M2 (R). Determine span{A1 , A2 , A3 }.
Solution: By definition we have

span{A1 , A2 , A3 } = {A ∈ M2 (R) : A = c1 A1 + c2 A2 + c3 A3 , c1 , c2 , c3 ∈ R}
      
1 0 0 1 0 0
= A ∈ M2 (R) : A = c1 + c2 + c3
0 0 1 0 0 1
   
c1 c2
= A ∈ M2 (R) : A = , c1 , c2 , c3 ∈ R .
c2 c3

This is the set of all real 2 × 2 symmetric matrices.

Example 4.4.11 Determine the subspace of P2 spanned by

p1 (x) = 1 + 3x, p2 (x) = x + x 2 ,

and decide whether {p1 , p2 } is a spanning set for P2 .

Solution: We have

span{p1 , p2 } = {p ∈ P2 : p(x) = c1 p1 (x) + c2 p2 (x), c1 , c2 ∈ R}

= {p ∈ P2 : p(x) = c1 (1 + 3x) + c2 (x + x 2 ), c1 , c2 ∈ R}
= {p ∈ P2 : p(x) = c1 + (3c1 + c2 )x + c2 x 2 , c1 , c2 ∈ R}.

Next, we will show that {p1 , p2 } is not a spanning set for P2 . To establish this, we need
give only one example of a polynomial in P2 that is not in span{p1 , p2 }. There are many
such choices here, but suppose we consider p(x) = 1 + x. If this polynomial were in
span{p1 , p2 }, then we would have to be able to find values of c1 and c2 such that

1 + x = c1 + (3c1 + c2 )x + c2 x 2 . (4.4.6)

Since there is no x 2 term on the left-hand side of this expression, we must set c2 = 0.
But then (4.4.6) would reduce to

1 + x = c1 (1 + 3x).

Equating the constant terms on each side of this forces c1 = 1, but then the coefficients of
x do not match. Hence, such an equality is impossible. Consequently, there are no values
of c1 and c2 such that the Equation (4.4.6) holds, and therefore, span{p1 , p2 } = P2 .

Remark In the previous example, the reader may well wonder why we knew from
the beginning to select p(x) = 1 + x as a vector that would be outside of span{p1 , p2 }.
In truth, we only need to find a polynomial that does not have the form p(x) = c1 +
(3c1 + c2 )x + c2 x 2 and in fact, “most” of the polynomials in P2 would have achieved
the desired result here.

i i

i i
 i i “main”
2007/2/16
page 265

i i

4.4 Spanning Sets 265

Exercises for 4.4

Key Terms 8. If S is a spanning set for a vector space V , then any
Linear combination, Linear span, Spanning set. proper subset S
of S is not a spanning set for V .

9. The vector space of 3 × 3 upper triangular matrices is

Skills spanned by the matrices Eij where 1 ≤ i ≤ j ≤ 3.

• Be able to determine whether a given set of vectors 10. A spanning set for the vector space P2 must contain a
S spans a vector space V , and be able to prove your polynomial of each degree 0, 1, and 2.
answer mathematically.
11. If m < n, then any spanning set for Rn must contain
• Be able to determine the linear span of a set of vec- more vectors than any spanning set for Rm .
tors. For vectors in Rn , be able to give a geometric
description of the linear span. 12. The vector space P of all polynomials with real coef-
ficients cannot be spanned by a finite set S.
• If S is a spanning set for a vector space V , be able to
write any vector in V as a linear combination of the Problems
elements of S.
For Problems 1–3, determine whether the given set of vectors
• Be able to construct a spanning set for a vector space spans R2 .
V . As a special case, be able to determine a spanning
set for the null space of an m × n matrix. 1. {(1, −1), (2, −2), (2, 3)}.

• Be able to determine whether a particular vector v in 2. {(2, 5), (0, 0)}.

a vector space V lies in the linear span of a set S of
vectors in V . 3. {(6, −2), (−2, 2/3), (3, −1)}.

Recall that three vectors v1 , v2 , v3 in R3 are coplanar if and

True-False Review only if
For Questions 1–12, decide if the given statement is true or det([v1 , v2 , v3 ]) = 0.
false, and give a brief justification for your answer. If true, For Problems 4–6, use this result to determine whether the
you can quote a relevant definition or theorem from the text. given set of vectors spans R3 .
If false, provide an example, illustration, or brief explanation
of why the statement is false. 4. {(1, −1, 1), (2, 5, 3), (4, −2, 1)}.
1. The linear span of a set of vectors in a vector space V 5. {(1, −2, 1), (2, 3, 1), (0, 0, 0), (4, −1, 2)}.
forms a subspace of V .
6. {(2, −1, 4), (3, −3, 5), (1, 1, 3)}.
2. If some vector v in a vector space V is a linear com-
bination of vectors in a set S, then S spans V . 7. Show that the set of vectors

3. If S is a spanning set for a vector space V and W is a {(1, 2, 3), (3, 4, 5), (4, 5, 6)}
subspace of V , then S is a spanning set for W .
does not span R3 , but that it does span the subspace
4. If S is a spanning set for a vector space V , then every
of R3 consisting of all vectors lying in the plane with
vector v in V must be uniquely expressible as a linear
equation x − 2y + z = 0.
combination of the vectors in S.
8. Show that v1 = (2, −1), v2 = (3, 2) span R2 , and ex-
5. A set S of vectors in a vector space V spans V if and
press the vector v = (5, −7) as a linear combination
only if the linear span of S is V .
of v1 , v2 .
6. The linear span of two vectors in R3 is a plane through
the origin. 9. Show that v1 = (−1, 3, 2), v2 = (1, −2, 1), v3 =
(2, 1, 1) span R3 , and express v = (x, y, z) as a linear
7. Every vector space V has a finite spanning set. combination of v1 , v2 , v3 .

i i

i i
 i i “main”
2007/2/16
page 266

i i

266 CHAPTER 4 Vector Spaces

10. Show that v1 = (1, 1), v2 = (−1, 2), v3 = (1, 4) span For Problems 22–24, determine whether the given vector v
R2 . Do v1 , v2 alone span R2 also? lies in span{v1 , v2 }.
11. Let S be the subspace of R3 consisting of all vectors 22. v = (3, 3, 4), v1 = (1, −1, 2), v2 = (2, 1, 3) in R3 .
of the form v = (c1 , c2 , c2 − 2c1 ). Show that S is
spanned by v1 = (1, 0, −2), v2 = (0, 1, 1). 23. v = (5, 3, −6), v1 = (−1, 1, 2), v2 = (3, 1, −4) in
R3 .
12. Let S be the subspace of R4 consisting of all vectors
of the form v = (c1 , c2 , c2 − c1 , c1 − 2c2 ). Determine 24. v = (1, 1, −2), v1 = (3, 1, 2), v2 = (−2, −1, 1) in
a set of vectors that spans S. R3 .
13. Let S be the subspace of R3 consisting of all solutions
to the linear system 25. If p1 (x) = x − 4 and p2 (x) = x 2 − x + 3, determine
whether p(x) = 2x 2 − x + 2 lies in span{p1 , p2 }.
x − 2y − z = 0.
26. Consider the vectors
Determine a set of vectors that spans S.      
1 −1 0 1 30
For Problems 14–15, determine a spanning set for the null A1 = , A2 = , A3 =
2 0 −2 1 1 2
space of the given matrix A.
  in M2 (R). Determine span{A1 , A2 , A3 }.
1 2 3
14. A =  3 4 5 .
27. Consider the vectors
5 6 7
     
12 −2 1
1 2 3 5 A1 = , A2 =
−1 3 1 −1
15. A =  1 3 4 2 .
2 4 6 −1
in M2 (R). Find span{A1 , A2 }, and determine whether
16. Let S be the subspace of M2 (R) consisting of all sym- or not  
metric 2 × 2 matrices with real elements. Show that S B=
3 1
is spanned by the matrices −2 4
     
1 0 0 0 0 1 lies in this subspace.
A1 = , A2 = , A3 = .
0 0 0 1 1 0
28. Let V = C ∞ (I ) and let S be the subspace of V
17. Let S be the subspace of M2 (R) consisting of all skew- spanned by the functions
symmetric 2 × 2 matrices with real elements. Deter-
mine a matrix that spans S. f (x) = cosh x, g(x) = sinh x.

18. Let S be the subset of M2 (R) consisting of all upper (a) Give an expression for a general vector in S.
triangular 2 × 2 matrices.
(b) Show that S is also spanned by the functions
(a) Verify that S is a subspace of M2 (R).
(b) Determine a set of 2 × 2 matrices that spans S. h(x) = ex , j (x) = e−x .

For Problems 19–20, determine span{v1 , v2 } for the given For Problems 29–32, give a geometric description of the sub-
vectors in R3 , and describe it geometrically. space of R3 spanned by the given set of vectors.
19. v1 = (1, −1, 2), v2 = (2, −1, 3).
29. {0}.
20. v1 = (1, 2, −1), v2 = (−2, −4, 2).
30. {v1 }, where v1 is any nonzero vector in R3 .
21. Let S be the subspace of R3
spanned by the vectors
v1 = (1, 1, −1), v2 = (2, 1, 3), v3 = (−2, −2, 2). 31. {v1 , v2 }, where v1 , v2 are nonzero and noncollinear
Show that S also is spanned by v1 and v2 only. vectors in R3 .

i i

i i
 i i “main”
2007/2/16
page 267

i i

4.5 Linear Dependence and Linear Independence 267

32. {v1 , v2 }, where v1 , v2 are collinear vectors in R3 . 34. Prove that

33. Prove that if S and S
are subsets of a vector space V span{v1 , v2 , v3 } = span{v1 , v2 }
such that S is a subset of S
, then span(S) is a subset
of span(S
). if and only if v3 can be written as a linear combination
of v1 and v2 .

4.5 Linear Dependence and Linear Independence

As indicated in the previous section, in analyzing a vector space we will be interested in
determining a spanning set. The reader has perhaps already noticed that a vector space
V can have many such spanning sets.

Example 4.5.1 Observe that {(1, 0), (0, 1)}, {(1, 0), (1, 1)}, and {(1, 0), (0, 1), (1, 2)} are all spanning
sets for R2 .

As another illustration, two different spanning sets for V = M2 (R) were given in Exam-
ple 4.4.5 and the remark that followed. Given the abundance of spanning sets available
for a given vector space V , we are faced with a natural question: Is there a “best class
of” spanning sets to use? The answer, to a large degree, is “yes”. For instance, in Exam-
ple 4.5.1, the spanning set {(1, 0), (0, 1), (1, 2)} contains an “extra” vector, (1, 2), which
seems to be unnecessary for spanning R2 , since {(1, 0), (0, 1)} is already a spanning set.
In some sense, {(1, 0), (0, 1)} is a more efficient spanning set. It is what we call a mini-
mal spanning set, since it contains the minimum number of vectors needed to span the
vector space.3
But how will we know if we have found a minimal spanning set (assuming one
exists)? Returning to the example above, we have seen that

span{(1, 0), (0, 1)} = span{(1, 0), (0, 1), (1, 2)} = R2 .

Observe that the vector (1, 2) is already a linear combination of (1, 0) and (0, 1), and
therefore it does not add any new vectors to the linear span of {(1, 0), (0, 1)}.
As a second example, consider the vectors v1 = (1, 1, 1), v2 = (3, −2, 1), and
v3 = 4v1 + v2 = (7, 2, 5). It is easily verified that det([v1 , v2 , v3 ]) = 0. Consequently,
the three vectors lie in a plane (see Figure 4.5.1) and therefore, since they are not collinear,
the linear span of these three vectors is the whole of this plane. Furthermore, the same
plane is generated if we consider the linear span of v1 and v2 alone. As in the previous
example, the reason that v3 does not add any new vectors to the linear span of {v1 , v2 }
is that it is already a linear combination of v1 and v2 . It is not possible, however, to
generate all vectors in the plane by taking linear combinations of just one vector, as we
could generate only a line lying in the plane in that case. Consequently, {v1 , v2 } is a
minimal spanning set for the subspace of R3 consisting of all points lying on the plane.
As a final example, recall from Example 1.2.16 that the solution space to the differ-
ential equation
y

+ y = 0

3 Since a single (nonzero) vector in R2 spans only the line through the origin along which it points, it cannot
span all of R2 ; hence, the minimum number of vectors required to span R2 is 2.

i i

i i