≻≻≻≻—≻≻≻≻—Lecture Eleven —≺≺≺≺—≺≺≺≺

Chapter Six: Subspace and Linear Independence

Recall. A (real) vector space (over R) is a set V together with two operations vector
addition ⊕ : V × V −→ V and scalar multiplication ⊙ : R × V −→ V satisfying
(Vα) (closure of addition) V is closed under the operation of addition. Namely, u ⊕ v ∈ V
for all u, v ∈ V .
(Va) (commutativity of addition) u ⊕ v = v ⊕ u for all u, v ∈ V .
(Vb) (associativity of addition) (u ⊕ v) ⊕ w = u ⊕ (v ⊕ w) for all u, v, w ∈ V .
(Vc) (identity element of addition) there is an element in V , denoted by 0, such that
u ⊕ 0 = u for all u ∈ V .
(Vd) (inverse elements of addition) for each u in V , there is an element in V , denoted by
−u, such that u ⊕ −u = 0.

Linear Algebra lecture 11−1

(Vβ) (closure of scalar multiplication) V is closed under the operation of scalar multiplication.
Namely, c ⊙ u ∈ V for all c ∈ R and u ∈ V .
(Ve) (distributivity of scalar multiplication with respect to vector addition) for all u, v ∈ V
and c ∈ R, we have c ⊙ (u ⊕ v) = (c ⊙ u) ⊕ (c ⊙ v).
(Vf) (distributivity of scalar multiplication with respect to field addition) for all u ∈ V
and c, d ∈ R, we have (c + d) ⊙ u = (c ⊙ u) ⊕ (d ⊙ u).
(Vg) for all u ∈ V and c, d ∈ R, we have (c · d) ⊙ u = c ⊙ (d ⊙ u).
(Vh) (identity element of scalar multiplication) for all u ∈ V , we have 1 ⊙ u = u.

Remark. With no ambiguity, we may refer to a real vector space simply as a vector space
and also write u ⊕ v and c ⊙ u simply as u + v and cu respectively.

Linear Algebra lecture 11−2

Recall. Last lecture we showed that R3 together with the addition defined by u ⊕ v =
   
u1 + v 1 cu1
u2 + v2 and scalar multiplication defined by c ⊙ u = cu2 is a real vector space.
   
u3 + v 3   cu3 
 x
 

Furthermore, it is verified that the subset W = x : x ∈ R of R3 is also a vector
 

 x 

space with respect to the same addition and scalar multiplication operations. We say W is
a subspace of R3 as defined below.

Definition. Let V be a vector space and W a nonempty subset of V . If W is a vector

space with respect to the same operations as those in V , then W is called a subspace of V .

Example 6.2.6. Let Pn, n = 0, 1, 2, . . ., denote the vector space consisting all polynomials
of degree smaller than or equal to n and the zero polynomial and P denote the vector space
of all polynomials. We have shown in last lecture that Pn, n = 0, 1, 2, . . ., and P are all
S
vector spaces. Therefore, Pr is a subspace of P for all r ∈ N {0} and Pr is a subspace of
S
Ps for any r ≤ s and r, s ∈ N {0}.
Linear Algebra lecture 11−3
Example 6.2.1. Any vector space V 6= {0} has at least (two) subspaces itself V and the
zero subspace {0}.

Fact. Let V be any vector space. Then V is the largest subspace of V and {0} is the
smallest subspace of V . Namely, if W is a subspace of V , then {0} ⊆ W ⊆ V .

Corollary. If a subset W of a vector space V does not contain the zero vector 0, then
W cannot be a subspace of V .

Example 6.2.7. Let W be the set of all polynomials of degree exactly equal to 2. Then
W is a subset of P2 but not a subspace of P2.

Linear Algebra lecture 11−4

 Theorem 6.2.2. Let V be a vector space with operations ⊕ and ⊙ and let W be a
nonempty subset of V . Then W is a subspace of V if and only if the following conditions
hold:
(Vα) W is closed under the operation of addition. Namely, u ⊕ v ∈ W for all u, v ∈ W .
(Vβ) W is closed under the operation of scalar multiplication. Namely, c ⊙ u ∈ W for all
c ∈ R and u ∈ W .

Proof. Skip.

Linear Algebra lecture 11−5

 a b 0
 
Example 6.2.3&12. Consider W consisting of all 2 × 3 of the form , where
0 c d
a, b, c, and d are arbitrary real numbers. Then W is a subspace of the vector space M23
defined in last lecture.

Example 6.2.7. Let W be the set of all polynomials of degree exactly equal to 2. Then
W is a subset of P2 but not a subspace of P2.

Linear Algebra lecture 11−6

Example 6.2.9. Consider the homogeneous system Ax = 0, where A is an m × n matrix.
Let W be the subset of Rn consisting of all solutions to the homogeneous system. Namely,
W is the solution space of Ax = 0 or we may write W = {x ∈ Rn : Ax = 0}. Then W is
a subspace of Rn. This subspace W is called the null space of A

Remark. Note the solution space of a linear system Ax = b, where A is m × n, is not a

subspace of Rn if b 6= 0.

Linear Algebra lecture 11−7

Definition. Given S = {v1, v2, . . . , vk } a set of vectors in a vector space V , the (linear)
span of S is the subset of V consisting of all linear combinations of v1, v2, . . . , vk . Namely
Span S = Span{v1, v2, . . . , vk} = {c1v1 + c2v2 + · · · + ckvk : c1, c2, . . . , ck ∈ R}. Further-
more, if W = Span S, we say that W is spanned by S or that S spans W .

(" # " # " #)

1 0 0 1 0 0
Example 6.3.2. Consider S = , , . Then the span of S con-
0 0 1 0 0 1
" # " # " # " #
1 0 0 1 0 0 a b
sists all 2 × 2 matrix of the form a +b +c = , a, b, c ∈ R.
0 0 1 0 0 1 b c
Namely, S spans the subset of all symmetric matrices in M22.

Linear Algebra lecture 11−8

 Theorem 6.2.3. Let S = {v1, v2, . . . , vk } be a set of vectors in a vector space V .
Then the subset W of V consisting of all linear combinations of v1, v2, . . . , vk , namely
W = Span S = Span{v1, v2, . . . , vk} = {c1v1 + c2v2 + · · · + ckvk : c1, c2, . . . , ck ∈ R}, is
a subspace of V .
Proof. 1◦ Note S ⊆ W and hence W is not empty.
Furthermore, since V is a vector space, c1v1 + c2v2 + · · · + ck vk ∈ V .
Hence W is a subset of V .
2◦ ”(Vα)”
Given any u, v ∈ W , we have, for some scalars c1, . . . , ck , d1, . . . , dk ,
u = c1v1 + c2v2 + · · · + ck vk and v = d1v1 + d2v2 + · · · + dk vk .
Then u + v = (c1v1 + c2v2 + · · · + ck vk ) + (d1v1 + d2v2 + · · · + dk vk )
= (c1 + d1)v1 + · · · + (ck + dk )vk
and we obtain u + v ∈ W .
3◦ ”(Vβ)”
Given any c ∈ R and u ∈ W ,
one has u = c1v1 + c2v2 + · · · + ck vk for some scalars c1, . . . , ck .
It is easy to see cu = (cc1)v1 + · · · + (cck )vk ∈ W .
4◦ We verify that W ⊆ V satisfies (Vα) and (Vβ) and hence is a subspace of V .
Linear Algebra lecture 11−9
     
 1
 0  
Example. Let S =  0  ,  1  in R3. Then the subspace spanned by S is the x-y
   
 
0 0
 
        

 1 0  a
 
 

plane as span S = a  0  + b  1  : a, b ∈ R =  b  : a, b ∈ R .
     
  
 
0 0 0
  

Example 6.2.3&12. Consider the set

(" # " # " # " # )
1 0 0 0 1 0 0 0 0 0 0 0
S= , , , , .
0 0 0 0 0 0 0 1 0 0 0 1
" # " #
1 0 0 0 1 0
Then span S, consisting of all 2 × 3 matrices of the form a +b +
0 0 0 0 0 0
" # " # " #
0 0 0 0 0 0 a b 0
c +d = for a, b, c, d ∈ R, is a subspace of M23.
0 1 0 0 0 1 0 c d

Example. In P2, let S = {1, t, t2}. Then span S = {c0 + c1t + c2t2 : c0, c1, c2 ∈ R} is a
subspace of P2. Note indeed span S = P2 (S spans P2).

Linear Algebra lecture 11−10

Theorem 6.2.T5. Let S = {v1, . . . , vk } be a set of vectors in a vector space V . Then
span S is the smallest subspace of V containing S. In other words, let W be any sub-
space of V containing S. Then span S ⊆ W .

Proof. 1◦ If W is a subspace containing S,

then any linear combinations of vectors in S, and hence in W ,
must be in W as well.
Consequently, span S ⊆ W .
Corollary. Let S = {v1, . . . , vk } be a set of vectors in a vector space V . Then span S
is the intersection of all subspaces containing S.
Proof. 1◦ Denote U = span S and U ∗ the intersection of all subspaces containing S.
because U is a subspace containing S by Theorem 6.2.3,
the intersection U ∗ ⊆ U .
2◦ On the other hand, Theorem 6.2.T5 says that
any subspace of V containing S contains U .
Hence U ⊆ U ∗. The proof is complete.
Linear Algebra lecture 11−11
Definition. Vectors v1, v2, . . . , vk in a vector space V are said to be linear dependent if
there exists real numbers c1, c2, . . . , ck , not all zero, such that c1v1 + c2v2 + · · · + ck vk = 0.
Otherwise, v1, v2, . . . , vk are called linear independent. That is, we say v1, v2, . . . , vk are
linear independent provided that whenever c1v1 + c2v2 + · · · + ck vk = 0, we must have
c1 = c2 = · · · = ck = 0.
     
1 0 0
Example. The three vectors e1 = 0, e2 = 1, and e3 = 0 in R3 are linear
     
0 0 1
independent.        
1 0 0 1
Example. The four vectors e1 = 0, e2 = 1, e3 = 0, and u = 1 in R3 are
       
0 0 1 1
linear dependent.        
1 0 0 1
Example. The four vectors e1 = 0, e2 = 1, e3 = 0, and v = 1 in R3 are
       
0 0 1 0
linear dependent.
Linear Algebra lecture 11−12
Example 6.3.12. Every set of vectors containing the zero vector is linear dependent.
More precisely, if v1, v2, . . . , vk are vectors in any vector space such that vj = 0 for some
1 ≤ j ≤ k. Then S = {v1, v2, . . . , vk } is linear dependent.

Example 6.3.11. Let p1(t) = t2 + t + 2, p2(t) = 2t2 + t, and p3(t) = 3t2 + 2t + 2. Then
S = {p1(t), p2(t), p3(t)} is linear dependent.

Example. Let p1(t) = t2 + t + 2, p2(t) = 2t2 + t, and p3(t) = 3t2 + 2. Then S =

{p1(t), p2(t), p3(t)} is linear independent.

Linear Algebra lecture 11−13

Theorem 6.3.T3. Let v1, v2, . . . , vk be vectors in a vector space V . Then v1, v2, . . . , vk
are linear dependent if and only if there exists some subscript j such that vj is a linear
combination of the other vectors v1, . . . , vj−1, vj+1, . . . , vk .
Note. This theorem does not say that every vector vj is a linear combination of the other
vectors.        
1 0 0 1
Example. Let v1 = 0, v2 = 1, v3 = 0, and v4 = 1. Then v1, v2, v3 and v4
       
0 0 1 0
are linear dependent. Note v1 = v4 − v2, a linear combination of v2 and v4, v2 = v4 − v1,
a linear combination of v1 and v4, and v4 = v1 + v2, a linear combination of v1 and v2.
However, we cannot write v3 as a linear combination of v1, v2, v4.

Linear Algebra lecture 11−14

 Proof. 1◦ ”⇒”
Since v1, v2, . . . , vk are linear dependent,
there exists real numbers c1, c2, . . . , ck , not all zero,
such that c1v1 + c2v2 + · · · + ck vk = 0.
Let j be any subscript for which cj 6= 0.
c cj+1 ck
Then vj = (− cc1j )v1 + · · · + (− j−1cj )vj−1 + (− cj )v j+1 + · · · + (− cj )vk .
Namely, vj is a linear combination of v1, v2, . . . , vj−1, vj+1 . . . , vk .
2◦ ”⇐”
Suppose vj is a linear combination of the other vectors
v1, v2, . . . , vj−1, vj+1 . . . , vk .
Then there exist real numbers c1, . . . , cj−1, cj+1, . . . , ck
such that vj = c1v1 + · · · + cj−1vj−1 + cj+1vj+1 + · · · + ck vk .
Take cj = −1. Then the above equation becomes
c1v1 + · · · + cj−1vj−1 + cj vj + cj+1vj+1 + · · · + ck vk = 0.
Note cj = −1 6= 0. Hence v1, v2, . . . , vk are linear dependent.

Linear Algebra lecture 11−15

Corollary. If S = {v1, v2, . . . , vk } is linear dependent, then there exists some j such
that the subspace spanned by S \ {vj }, consisting of S with vj deleted, is the same as
the subspace spanned by S.
Proof. 1◦ Since S is linear dependent, by Theorem 6.3.T3, there exists some j
such that vj is a linear combination of v1, v2, . . . , vj−1, vj+1, . . . , vk ,
say, vj = c1v1 + · · · + cj−1vj−1 + cj+1vj+1 + · · · + ck vk
for some real numbers c1, . . . , cj−1, cj+1, . . . , ck .
Denote S ∗ = S \ {vj }, W = span S and W ∗ = span S ∗.
2◦ It is obvious that W ∗ ⊆ W because S ∗ ⊆ S.
3◦ Now given any w ∈ W , we can write
w = d1v1 + · · · + dj−1vj−1 + dj vj + dj+1vj+1 + · · · + dk vk
for some scalars d1, . . . , dk . Then 1◦ gives
w = d1v1 + · · · + dj−1vj−1 + dj (c1v1 + · · · + cj−1vj−1 + cj+1vj+1 + · · · + ck vk )
+dj+1vj+1 + · · · + dk vk
= (d1 + dj c1)v1 + · · · + (dj−1 + dj cj−1)vj−1 + (dj+1 + dj cj+1)vj+1 + · · ·
+(dk + dj ck )vk , a linear combination of vectors in S ∗.
Therefore W ⊆ W ∗ . We complete the proof.
Linear Algebra lecture 11−16
Theorem 6.3.4. The nonzero vectors v1, v2, . . . , vk in a vector space V are linear
dependent if and only if one of the vectors vj , j ≥ 2, is a linear combination of the
preceding vectors v1, v2, . . . , vj−1.

Corollary Let v1, v2, . . . , vk are nonzero vectors in a vector space V . Then the fol-
lowing statements are equivalent.
(a) v1, v2, . . . , vk are linear dependent.
(b) There exists j ≥ 2 such that vj is a linear combination of the preceding vectors
v1, v2, . . . , vj−1.
(c) There exists j such that vj is a linear combination of the other vectors v1, v2, . . . , vj−1,
vj+1, . . . , vk .

Linear Algebra lecture 11−17

Proof. 1◦ ”⇒” Since v1, v2, . . . , vk are linear dependent,
there exists real numbers c1, c2, . . . , ck , not all zero,
such that c1v1 + c2v2 + · · · + ck vk = 0.
Let j be the largest subscript for which cj 6= 0.
That is cj 6= 0 and cj+1 = · · · = ck = 0.
Then c1v1 + c2v2 + · · · + cj vj = 0.
c
It following vj = (− cc1j )v1 + · · · + (− j−1
cj
)vj−1.
Thus vj is a linear combination of v1, v2, . . . , vj−1.
Also note j ≥ 2 because if j = 1,
we get c1v1 = 0, along with c1 6= 0, which implies v1 = 0.
2◦ ”⇐” Suppose, for some j ≥ 2 and real numbers c1, c2, . . . , cj−1,
vj = c1v1 + c2v2 + · · · + cj−1vj−1.
Take cj = −1, cj+1 = 0, . . . , cn = 0.
Then the above equation becomes
c1v1 + · · · + cj−1vj−1 + cj vj + cj+1vj+1 + · · · + ck vk = 0.
Note cj = −1 6= 0. Hence v1, v2, . . . , vk are linear dependent.

Linear Algebra lecture 11−18

Theorem 6.3.T2. Let S1 and S2 be finite subsets of a vector space V and S1 ⊆ S2.
(a) If S1 is linear dependent, then S2 is linear dependent as well.
(b) If S2 is linear independent, then S1 is linear independent as well.

Proof. Skip.
Linear Algebra lecture 11−19