Linear Algebra II

Akiyoshi Tsuchiya ([email protected])

1 Vector Spaces
Recall properties of Rn .

• Two operations on Rn are defined.

– (addition) u + v ∈ Rn for u, v ∈ Rn .
– (scalar multiplication) cu ∈ Rn for u ∈ Rn and c ∈ R.

• The following properties hold for any u, v, w ∈ Rn and c, d ∈ R:

1. u + v = v + u;
2. (u + v) + w = u + (v + w);
3. u + 0 = u, where 0 is the zero vector in Rn ;
4. u + (−u) = 0;
5. c(u + v) = cu + cv;
6. (c + d)u = cu + du;
7. c(du) = (cd)u;
8. 1u = u.

Basically, other properties of Rn can be proven by these properties. This means that these
properties are essential for Rn . If operations of a set satisfy these essential properties, then
the set and Rn should have same properties. We call the set a vector space:

Definition 1.1. Let V be a set of objects. V is called a vector space if

• Two operations on V are defined.

– (addition) u + v ∈ V for u, v ∈ V .
– (scalar multiplication) cu ∈ V for u ∈ V and c ∈ R.

• The two operations satisfy the following properties (axioms) hold for any u, v, w ∈ V
and c, d ∈ R:

1. u + v = v + u;
2. (u + v) + w = u + (v + w);
3. There exists x ∈ V such that for any u ∈ V , u + x = u. We denote this vector by
0, which is called the zero vector of V ;

1
 4. For any u ∈ V , there exists u′ ∈ V such that u + u′ = 0. We denote this vector
by −u, which is called the additive inverse of u;
5. c(u + v) = cu + cv;
6. (c + d)u = cu + du;
7. c(du) = (cd)u;
8. 1u = u.

Hence Rn is a vector space under the usual addition and scalar multiplication.

Remark 1.2. An object of a vector space is called a vector.

Example 1.3. Let Mn be the set of n × n matrices. We consider the usual matrix addition
and matrix scalar multiplication. For example, for M2 ,
󰀗 󰀘 󰀗 󰀘 󰀗 󰀘
a11 a12 b11 b12 a11 + b11 a12 + b12
• (addition) + = .
a21 a22 b21 b22 a21 + b21 a22 + b22
󰀗 󰀘 󰀗 󰀘
a11 a12 ca11 ca12
• (scalar multiplication) c = .
a21 a22 ca21 ca22

Then the axioms (1), (2), (5), (6), (7), (8) follow by definition. The zero matrix
󰀵 󰀶
0 ··· 0
󰀹 󰀺
0 = 󰀷 ... . . . ... 󰀸
0 ··· 0

satisfies the axiom (3). Moreovre, for A ∈ Mn , (−1)A satisfies the axiom (4). Hence Mn is
a vector space. In general, the set of all n × m matrices is a vector space under the usual
matrix addition and matrix scalar multiplication.

Example 1.4. Let Pn be the set of all polynomials of degree at most n. For f, g ∈ Pn with
f (t) = a0 + a1 t + · · · an tn and g(t) = b0 + b1 t + · · · + bn tn , we define the addition f + g on Pn
by
(f + g)(t) = (a0 + b0 ) + (a1 + b1 )t + · · · + (an + bn )tn ∈ Pn .
Moreover, for f ∈ Pn and c ∈ R, we define the scalar multiplication cf on Pn by

(cf )(t) = (ca0 ) + (ca1 )t + · · · + (can )tn ∈ Pn .

Then it is easy to show that the axioms (1), (2), (5), (6), (7), (8) hold. The zero polynomial
0 defined by
0(t) = 0 + 0t + · · · + 0tn
satisfies the axiom (3). Moreover, for f ∈ Pn , (−1)f satisfies the axiom (4). Hence Pn is a
vector space under these operations.

2
Example 1.5. For R2 , we define the scalar multiplication c ⊙ (a, b) = (ca, b). Then R2 is not
a vector space under the usual addition and this scalar multiplication. Indeed, the axiom (6)
does not always hold. For example,
(r + s) ⊙ (a, b) = ((r + s)a, b),
but
r ⊙ (a, b) + s ⊙ (a, b) = (ra, b) + (sa, b) = ((r + s)a, 2b).
Proposition 1.6. Let V be a vector space. Then
(a) (The uniqueness of 0) If there exists 0′ ∈ V such that u + 0′ = u for any u ∈ V , then
0′ = 0.
(b) (The uniqueness of −u) Let u ∈ V . If there exist u′ , u′′ ∈ V such that u+u′ = u+u′′ =
0, then u′ = u′′ .
Proof. (a) By applying u + 0′ = u for u = 0, one has 0 + 0′ = 0. On the other hand, from
the axiom (3) we obtain 0′ + 0 = 0′ . Moreover, from the axiom (1) one has 0 + 0′ = 0′ + 0.
Hence 0 = 0 + 0′ = 0′ + 0 = 0′ .
(b) By adding u′′ to the both side of u + u′ = 0, one has
(u + u′ ) + u′′ = 0 + u′′
(u′ + u) + u′′ = 0 + u′′ (axiom (1))
u′ + (u + u′′ ) = 0 + u′′ (axiom (2))
u′ + 0 = 0 + u′′ (assumption)
u′ = u′′ (axiom (3))

Proposition 1.7. Let V be a vector space. For any u ∈ V and c ∈ R, one has
(a) 0u = 0,
(b) c0 = 0,
(c) −u = (−1)u.
Proof. (a)
0u = (0 + 0)u
= 0u + 0u (axiom (6)).
By adding −0u to the both sides, one has
0u + (−0u) = (0u + 0u) + (−0u)
0 = (0u + 0u) + (−0u) (axiom (4))
= 0u + (0u + (−0u)) (axiom (1))
= 0u + 0 (axiom (4))
= 0u (axiom (3)).

3
 (b)

c0 = c(0 + 0) (axiom (3))

= c0 + c0 (axiom (5)).

By adding −c0 to the both sides, one has

c0 + (−c0) = (c0 + c0) + (−c0)

0 = (c0 + c0) + (−c0) (axiom (4))
= c0 + (c0 + (−c0)) (axiom (1))
= c0 + 0 (axiom (4))
= c0 (axiom (3)).

(c) We want to show that u + (−1)u = 0. This implies that −u = (−1)u from axiom (4).

u + (−1)u = 1u + (−1)u (axiom (8))

= (1 + (−1))u (axiom (6))
= 0u
= 0 (a).

Exercise 1.8. For R3 , we define the addition ⊕ by

(a1 , a2 , a3 ) ⊕ (b1 , b2 , b3 ) = (a1 + b1 + 5, a2 + b2 − 7, a3 + b3 + 1)

and the scalar multiplication ⊙ by

c ⊙ (a1 , a2 , a3 ) = (ca1 + 5(c − 1), ca2 − 7(c − 1), ca3 + c − 1)

Then R3 is a vector space with these operations.

1. What is the zero vector of this vector space?

2. Check the axioms of this vector space.

3. Check Proposition 1.7 holds for this vector spaces.

2 Subspaces
Definition 2.1. Let V be a vector space. A subspace of V is a subset W of V that satisfies
the following three properties:
(a) The zero vector of V is in W ;

(b) For any u, v ∈ W , u + v ∈ W ;

4
 (c) For any u ∈ W and c ∈ R, cu ∈ W .

Proposition 2.2. Let V be a vector space. Then any subspace of V is a vector space under
the operations on V .

Proof. Properties (b) and (c)define two operations on W . Moreover, the axioms (1), (2),
(5)–(8) automatically hold in W . Property (a) guarantees axiom (3) in W . If u ∈ W , then
(−1)u = −u ∈ W from property (c). This implies axiom (4) in W . Hence W is a vector
space.

Example 2.3. For any nonzero vector space V has at least two subspaces. One is itself.
One is the zero space. Indeed, let W = {0}. Then

(a) 0 ∈ W .

(b) 0 + 0 = 0 ∈ W (axiom(3)).

(c) c0 = 0 ∈ W (Proposition 1.7 (b)).

Hence {0} is a subspace of V .

Example 2.4. Set

󰀝󰀗 󰀘 󰀞
c 1
A = : c ∈ R ⊂ M2 ,
1 c
󰀝󰀗 󰀘 󰀞
c 0
B = : c ∈ R ⊂ M2 ,
0 c
󰀝󰀗 󰀘 󰀞
a 0
C = : a, b ∈ R ⊂ M2 .
0 b

Then since the zero matrix is not in A, A is not subspace of M2 . Since

󰀗 󰀘
0 0
(a) ∈ B.
0 0
󰀗 󰀘 󰀗 ′ 󰀘 󰀗 󰀘
c 0 c 0 (c + c′ ) 0
(b) + = ∈ B.
0 c 0 c′ 0 (c + c′ )
󰀗 󰀘 󰀗 󰀘
c 0 ac 0
(c) a = ∈ B.
0 c 0 ac
Hence B is a subspace. Similarly, we can show that C is a subspace.

In Example 2.4, since B and C are subspaces of A, B and C are vector spaces from
Proposition 2.2. On the other hand, C is a subset of B. In this case, we can C as a subspace
of B. In fact, we can show the following easily.

Proposition 2.5. Let V be a vector space and W, W ′ subspaces of V . If W ⊂ W ′ , then W

is a subspace of W ′ .

5
Definition 2.6. Let V be a vector space and u1 , . . . , up ∈ V . Then a linear combina-
tion of u1 , . . . , up if a vector in the form c1 u1 + · · · + cp up with c1 , . . . , cp ∈ R. we denote
span{u1 , . . . , up } the set of all linear combinations of u1 , . . . , up . Namely,

span{u1 , . . . , up } := {c1 u1 + · · · + cp up : c1 , . . . , cp ∈ R} ⊂ V.

Theorem 2.7. Let V be a vector space and u1 , . . . , up ∈ V . Then span{u1 , . . . , up } is a

subspace of V .
Proof. (a) 0u1 + · · · + 0up = 0 + · · · + 0 = 0 ∈ span{u1 , . . . , up }.

(b) For a1 u1 + · · · + ap up , b1 u1 + · · · + bp up ∈ span{u1 , . . . , up }, one has

(a1 u1 + · · · + ap up ) + (b1 u1 + · · · + bp up )
= (a1 + b1 )u1 + · · · + (ap + bp )up ∈ span{u1 , . . . , up }.

(c) For a1 u1 + · · · + ap up ∈ span{u1 , . . . , up } and c ∈ R, one has

c(a1 u1 + · · · + ap up ) = (ca1 )u1 + · · · + (cap )up ∈ span{u1 , . . . , up }.

Therefore, span{u1 , . . . , up } is a subspace of Rn .

Definition 2.8. Let W be a subspace of a vector space V . A subset H of W is called a
spanning set for W if W = spanH.
Exercise 2.9. Let Q be the set of all polynomials in Pn which is the divisible by x − 2. Show
that Q is a subspace of Pn . Hint: every polynomial in Q is of the form (x − 2)f (x) where f
is a polynomial of degree at most n − 1.
Exercise 2.10. Let H be the set of all vectors of the form (a − 3b, b − a, a), where a, b are
arbitrary scalars. Namely,

H = {(a − 3b, b − a, a) ∈ R3 : a, b ∈ R}.

Find a spanning set of H and show that H is a subspace of R3 .

3 Linear transformations
Let V and W be vector spaces. Now, we consider a map T from V to T , denoted by
T : V → W , namely, we assign each vector u in V to a vector T (u) in W . We say that T (u)
is the image of u under T . Denote Im(T ) the set of all images T (u) with u ∈ V . We say
that Im(T ) is the image of V under T . We remark that Im(T ) is a subset of W .
Definition 3.1. Let V and W be vector spaces and T : V → W . We say that T is a linear
transformation if
1. T (u + v) = T (u) + T (v) for any u, v ∈ V .

2. T (cu) = cT (u) for any u ∈ V and c ∈ R.

6
Example 3.2. Let A be an m × n matrix and we define a map T : Rn → Rm by x 󰀁→ Ax.
Then T is a linear transformation. In this case, Im(T ) = ColA.
󰀗 󰀘
a b
Example 3.3. We define a map det : M2 → R defined by A = 󰀁→ ad − bc. Then for
󰀗 󰀘 c d
1 0
I2 = , one has det(I2 ) = 1 and det(2I2 ) = 4. Hence 4 = det(2I2 ) ∕= 2 det(I2 ) = 2. This
0 1
implies that det is not a linear transformation.
󰀗 󰀘
a b
Example 3.4. We define a map T : M2 → R defined by 󰀁→ a + d. Then
c d
󰀗 󰀘 󰀗 󰀘 󰀗 󰀘
a b e f a+e b+f
1. For A = ,B = ∈ M2 , A + B = . Hence T (A + B) =
c d g h c+g d+h
(a + e) + (d + h) = (a + d) + (e + h) = T (A) + T (B).
󰀗 󰀘 󰀗 󰀘
a b xa xb
2. For A = and x ∈ R, xA = . Hence T (xA) = xa + xd = x(a + d) =
c d xc xd
xT (A).
Therefore,
󰀅 󰀆 T is a linear transformation. In general, a map Trn : Mn → R defined by
aij 1≤i,j≤n 󰀁→ a11 + · · · + ann is a linear transformation. Trn is called the trace of Mn . Then
Im(Trn ) = R. In particular, Im(Trn ) is a subspace of R.
Proposition 3.5. Let V and W be vector spaces and T : V → W a linear transformation.
Then Im(T ) is a subspace of W .
Proof. (a) T (0V ) = T (00V ) = 0T (0V ) = 0W ∈ Im(T ).

(b) For T (x), T (y) ∈ Im(T ) with x, y ∈ V , one has T (x) + T (y) = T (x + y) ∈ Im(T ).

(c) For T (x) ∈ Im(T ) with x ∈ V and c ∈ R, one has cT (x) = T (cx) ∈ Im(T ).
Therefore, Im(T ) is a subspace of W .
Definition 3.6. Let V and W be vector spaces and T : V → W a linear transformation.
The kernel Ker(T ) of T is the set of all vectors x ∈ V such that T (x) = 0W , where 0W is
the zero vector of W .
Example 3.7. Let A be an m×n matrix and we define a linear transformation T : Rn → Rm
by x 󰀁→ Ax. Then Ker(T ) = NulA.
󰀗 󰀘
a b
Example 3.8. We consider the trace Tr2 of M2 . Then A = ∈ Ker(Tr2 ) if and only
c d
if Tr2 (A) = a + d = 0. Hence
󰀝󰀗 󰀘 󰀞
a b
Ker(Tr2 ) = : a, b, c ∈ R .
c −a

Proposition 3.9. Let V and W be vector spaces and T : V → W a linear transformation.

Then Ker(T ) is a subspace of V .

7
Proof. (a) In the proof of Proposition 3.5, we have already seen that T (0V ) = 0W . Hence
0V ∈ Ker(T ).

(b) For x, y ∈ Ker(T ), T (x + y) = T (x) + T (y) = 0W + 0W = 0W . Hence x + y ∈ Ker(T ).

(c) For x ∈ Ker(T ) and c ∈ R, T (cx) = cT (x) = c0W = 0W . Hence cx ∈ Ker(T ).

Therefore, Ker(T ) is a subspace of V .

4 Linearly independent sets

Definition 4.1. Let V be a vector space. A set of vectors {x1 , . . . , xp } in V is said to be
linearly independent if c1 x1 + · · · cp xp = 0 with c1 , . . . , cp ∈ R implies that c1 = · · · =
cp = 0. Otherwise, we say that {x1 , . . . , xp } is linearly dependent.

Theorem 4.2. Let V be a vector space and S = {v1 , . . . , vp } ⊂ V with p ≥ 2 and v1 ∕= 0.

Then S is linearly dependent if and only if some vj (with j > 1) is a linear combination of
v1 , . . . , vj−1 .

Proof. (⇒) Suppose that c1 v1 + · · · + cp vp = 0 with c1 , . . . , cp ∈ R with ck ∕= 0. Let j be the

maximal index i such that ci ∕= 0. Then one has c1 v1 + · · · + cj vj = 0 with cj ∕= 0. We claim
that j > 1. Indeed, if j = 1, then c1 v1 = 0. Since c1 ∕= 0, v1 = 0, contradiction. On the
c
other hand, since cj ∕= 0, we obtain vj = − cc1j v1 − · · · − j−1
cj
vj−1 .
(⇐) Suppose that vj = c1 v1 + · · · + cj−1 vj−1 with j > 1 and c1 , . . . , cj−1 ∈ R. Then one
has c1 v1 + · · · + cj−1 vj−1 − 1vj + 0vj+1 + · · · + 0vp = 0. This implies that {v1 , . . . , vp } is
linearly dependent.

Example 4.3. Let f1 (t) = 1, f2 (t) = t, f3 (t) = 4 − t be polynomials in Pn .

• {f1 } is linearly independent. In fact, for any vector space V and a vector x ∈ V , {x}
is linearly independent if and only if x ∕= 0.

• {f1 , f2 } is linearly independent. In fact, for any vector space V and x, y ∈ V with
x, y ∕= 0, {x, y} is linearly dependent if and only if there exists c ∈ R such that y = cx.

• {f1 , f2 , f3 } is linearly dependent because f3 = 4f1 − f2 .

Proposition 4.4. Let V and W be vector spaces and T : V → W a linear transformation,

and let x1 , . . . , xp ∈ V . If {T (x1 ), . . . , T (xp )} is linearly independent, then {x1 , . . . , xp } is
linearly independent.

Proof. Suppose that c1 x1 + · · · + cp xp = 0V with c1 , . . . , cp . Then

T (c1 x1 + · · · + cp xp ) = T (0V )
c1 T (x1 ) + · · · + cp T (xp ) = 0W

Since {T (x1 ), . . . , T (xp )} is linearly independent, one has c1 = · · · = cp = 0. Hence

{x1 , . . . , xp } is linearly independent.

8
Example 4.5. We consider the trace Tr2 of M2 . Let
󰀗 󰀘 󰀗 󰀘
1 0 0 0
A= ,B = .
0 0 0 −1

Then {A, B} is linearly independent. However, since Tr2 (A) + Tr2 (B) = 0, {Tr2 (A), Tr2 (B)}
is not linearly independent.
Definition 4.6. Let V and W be linear spaces. A linear transformation T : V → W is said
to be one-to-one (injective) if T (x) = T (y) with x, y ∈ V implies that x = y.
Proposition 4.7. Let V and W be linear spaces and T : V → W a linear transformation.
Then
(a) T is one-to-one if and only if Ker(T ) = {0V }.

(b) Assume that T is one-to-one. If {x1 , . . . , xp } is linearly independent, then {T (x1 ), . . . , T (xp )}
is linearly independent.
Proof. (a) (⇒) Take x ∈ Ker(T ). Then T (x) = 0W = T (0V ). Since T is one-to-one, x = 0V .
Hence Ker(T ) = {0V }.
(⇐) Take x, y ∈ V such that T (x) = T (y). Then

T (x) = T (y)
T (x) − T (y) = 0W
T (x − y) = 0W .

Hence x − y ∈ Ker(T ). Since Ker(T ) = {0}, x − y = 0V . Therefore, x = y. This implies

that T is one-to-one.
(b) Suppose that c1 T (x1 ) + · · · + cp T (xp ) = 0W with c1 , . . . , cp . Then T (c1 x1 + · · · +
cp xp ) = 0W . Hence c1 x1 + · · · + cp xp ∈ Ker(T ). It then follows from (a) that c1 x1 + · · · +
cp xp = 0V . Since {x1 , . . . , xp } is linearly independent, one has c1 = · · · = cp = 0. Hence
{T (x1 ), . . . , T (xp )} is linearly independent.

5 Basis
Definition 5.1. Let W be a subspace of a vector space V . A set of vectors B = {b1 , . . . , bp }
in V is a basis for W if
1. B is linearly independent.

2. W = spanB.
Example 5.2. Set S = {1, t, t2 , . . . , tn } ⊂ Pn . Then S is linearly independent. Suppose that
c0 · 1 + c1 t + · · · + cn tn = 0(t) with c0 , . . . , cn ∈ R. Then c0 = · · · = cn = 0. For example,
when n = 2, c0 = 0 (t = 0), c0 + c1 + c2 = 0 (t = 1) and c0 + 2c1 + 4c2 = 0 (t = 2) imply
c0 = c1 = c2 = 0. Hence S is linearly independent. Since clearly, Pn = spanS, S is a basis
for Pn . In this case, S is called the standard basis for Pn .

9
Example 5.3. Let f1 (t) = 2t − t2 , f2 (t) = 2 + 2t, f3 (t) = 6 + 16t − 5t2 and W =
span{f1 , f2 , f3 } ⊂ P2 . Then since f3 = 5f1 + 3f2 , f3 ∈ span{f1 , f2 }. In particular,
W = span{f1 , f2 }. Since {f1 , f2 } is linearly independent, {f1 , f2 } is a basis for H.

Theorem 5.4 (The Spanning Set Theorem). Let V be a vector space and S = {v1 , . . . , vp } ⊂
V and W = spanS. Then

(a) If one of the vectors in S, say vk , is a linear combination of the remaining vectors in
S, then S \ {vk } still spans W .

(b) If W ∕= {0}, then some subset of S is a basis for W .

Proof. (a) We may assume that vp is a linear combination of v1 , . . . , vp−1 . Namely, vp =

c1 v1 + · · · + cp−1 vp−1 with c1 , . . . , cp−1 ∈ R. Given x ∈ W , we can write x = d1 v1 + · · · + dp vp
with d1 , . . . , dp ∈ R. Then

x = d1 v1 + · · · + dp (c1 v1 + · · · + cp−1 vp−1 ) = (d1 + dp c1 )v1 + · · · + (dp−1 + dp cp−1 )vp−1 .

Hence W ⊂ span{v1 , . . . , vp−1 }. This implies that W = span{v1 , . . . , vp−1 }.

(b) If p = 1, then S is a basis for W because W ∕= {0}. If p ≥ 2, by using (a) repeatedly,
we can get a spanning set S ′ ⊂ S of W such that S ′ is linearly independent. This implies
that S ′ is a basis for W .

Theorem 5.5 (The unique representation theorem). Let V be a vector space and B =
{b1 , . . . , bn } a basis for V . Then for each x ∈ V , there uniquely exist c1 , . . . , cn ∈ V such
that
x = c 1 b1 + · · · + c n bn . (1)

Proof. Since spanB = V , there exist c1 , . . . , cn ∈ V satisfying (1). Suppose that x also has
the representation x = d1 b1 + · · · + dn bn with d1 , . . . , dn ∈ R. Then

0 = x − x = (c1 − d1 )b1 + · · · + (cn − dn )bn .

Since B is linearly independent, c1 − d1 = · · · = cn − dn = 0. Hence ci = di for any

1 ≤ i ≤ n.

Definition 5.6. Let V be a vector space and B = {b1 , . . . , bn } a basis for V . If x ∈ V

has the 󰀵 󰀶 representation x = c1 b1 + · · · + cn bn with c1 , . . . , cn ∈ R, then we write
unique
c1
󰀅 󰀆 󰀹 .. 󰀺 󰀅 󰀆
x B = 󰀷 . 󰀸 ∈ Rn . The vector x B is called the B-coordinate vector of x. The map
cn
󰀅 󰀆
V → R defined by x 󰀁→ x B is called the coordinate mapping of B.
n

n
Example 5.7. We consider the standard 󰀵 󰀶 basis S = {1, t, . . . , t } for Pn . Then for f (t) =
a0
󰀅 󰀆 󰀹 .. 󰀺
a0 + a1 t + · · · + an t ∈ Pn , f S = 󰀷 . 󰀸 ∈ Rn+1 .
n

an

10
Definition 5.8. Let V and W be vector spaces. We say that a linear transformation T :
V → W is onto (surjective) if for any w ∈ W , there exists v ∈ V such that T (v) = w.
Namely, Im(V ) = W .
Definition 5.9. Let V and W be vector spaces. We say that a linear transformation T :
V → W is an isomorphism from V onto W if it is one-to-one and onto. In this case, we
say that V and W are isomorphic, denoted by V ∼
= W.
Theorem 5.10. Let V be a vector space and B = {b1 , . . . , bn } a basis for V . Then the
coordinate mapping of B is an isomorphism from V onto Rn .
Proof. It is an exercise to show that the coordinate mapping is a linear transformation. We
show that the map is one-to-one and󰀅 onto.
󰀆 Let T be the coordinate mapping of B.
Assume that for some x ∈ V , x B = 0Rn . Namely, x ∈ ker(T ). This implies that
x = 0b1 + · · · 0bn = 0V . Hence ker(T ) = {0V }. It then follows from Proposition 4.7 (a) that
T is one-to-one.󰀵 󰀶 󰀵 󰀶
a1 a1
󰀹 .. 󰀺 󰀅 󰀆 󰀹 󰀺
Take any 󰀷 . 󰀸 ∈ Rn . Set x = a1 b1 + · · · + an bn ∈ V . Then T (x) = x B = 󰀷 ... 󰀸.
an an
Hence, T is onto.
From Propositions 4.4 and 4.7 (b) we have the following.
Corollary 5.11. Let V be a vector space and B = {b1󰀅, . . 󰀆. , bn } a󰀅 basis 󰀆 for V . Then
{x1 , . . . , xp } ⊂ V is linearly independent if and only if { x1 B , . . . , xp B } is linearly in-
dependent.
Example 5.12. Let f1 (t) = 1, f2 (t) = 󰀗 t,󰀘 f3 (t) = 4 − t ∈ 󰀗P1󰀘. We recall that 󰀗S =󰀘{1, t} is the
󰀅 󰀆 1 󰀅 󰀆 0 󰀅 󰀆 4
standard basis for P1 . Then f1 S = =: v1 , f2 S = =: v2 , f3 S = =: v3 . We
0 1 −1
recall that {f1 } and {f1 , f2 } are linearly independent but {f1 , f2 , f3 } is linearly dependent.
Hence it follows that {v1 } and {v1 , v2 } are linearly independent but {v1 , v2 , v3 } is linearly
dependent.

6 Dimension
Theorem 6.1. Let V be a vector space and B = {b1 , . . . , bn } a basis for V . Then any set
in V containing more than n vectors must be linearly dependent.
󰀅 󰀆 󰀅 󰀆
Proof. Let {u1 , . . . , up } be a set in V with p > n. Then { u1 B , . . . , u1 B } ⊂ Rn is linearly
dependent. It then follows from Corollary 5.11 that {u1 , . . . , up } is also linearly dependent.

Theorem 6.2. Let V be a vector space and B = {b1 , . . . , bn } a basis for V . Then any basis
for V must consist of exactly n vectors.
Proof. Let A = {a1 , . . . , am } be an arbitrary basis for V . Since B is a basis for V and A is
linearly independent, one has m ≤ n from Theorem 6.1. Since A is a basis for V and B is
linearly independent, one has n ≤ m. Hence we obtain n = m.

11
Definition 6.3. Let V be a vector space. We say that V is finite-dimensional. if V
is spanned by a finite set. In this case, it follows from Theorem 5.4 that V has a basis.
The dimension of V , denoted by dim V , is the number of vectors in a basis for V . For
convenience, we define dim{0} = 0. If V is not spanned by a finite set, then we say that V
is infinite-dimensional.

Example 6.4. We consider the vector space M2 . Then

󰀝󰀗 󰀘 󰀗 󰀘 󰀗 󰀘 󰀗 󰀘󰀞
1 0 0 1 0 0 0 0
, , ,
0 0 0 0 1 0 0 1

is a basis for M2 . Hence one has dim M2 = 4. In general, it follows that dim Mn = n2 .

Example 6.5. We consider the vector space P2 . Then {1, t, t2 } is a basis for P2 . Hence one
has dim P2 = 3. In general, it follows that dim Pn = n + 1.

Example 6.6. Let P is the set of all polynomials. Then we can easily show that P is a
vector space. Moreover, P is infinite-dimensional. Indeed, if P is spanned by a finite set A.
Then we let n be the maximal degree of the polynomials in A. Hence one has tn+1 ∈ / spanA.
n+1
This contradicts that t ∈ P.

Corollary 6.7. Let V be an n-dimensional vector space. Then V ∼

= Rn .
Theorem 6.8. Let V be a finite-dimensional vector space and H a subspace of V . Then any
linearly independent set in H can be expanded to a basis for H. In particular, dim H ≤ dim V .

Proof. If H = {0}, then dim H = 0 ≤ dim V . Otherwise, let S = {u1 , . . . , uk } be any

linearly independent set in H. If S spans H, then S is a basis for H. Otherwise, there
is some uk+1 ∈ H that is not in spanS. But {u1 , . . . , uk+1 } is linearly independent from
Theorem 4.2. So long as the new set does not span H, we can continue this process. But the
number of vectors in a linearly independent set must be at most dim V from Theorem 6.1.
So eventually the expansion of S will span H, hence will be a basis for H.

Theorem 6.9 (The Basis Theorem). Let V be a p-dimensional vector space with p ≥ 1 and
S = {v1 , . . . , vp } ⊂ V . Then the following arguments are equivalent:

(a) S is a basis for V ;

(b) S is linearly independent;

(c) V = spanS.

Proof. (a) ⇒ (b) and (a) ⇒ (c) follow from the definition of a basis.
((b) ⇒ (a)) By Theorem 6.8, S can be expanded to a basis for V . But that basis must
contain exactly p vectors. Hence S must be a basis for V .
((c) ⇒ (a)) Since V is a nonzero space, by Theorem 5.4, a subset S ′ of S is a basis for
V . Since dim V = p, S ′ must contain p vectors. This implies that S = S ′ .

12
7 Inner product spaces
Recall that for u, v ∈ Rn , the standard inner product of u and v is
u ◦ v = ut v = u 1 v 1 + · · · + u n v n .
Proposition 7.1. Let u, v, w ∈ Rn and c ∈ R. Then
1. u ◦ v = v ◦ u,
2. (u + v) ◦ w = u ◦ w + v ◦ w,
3. (cu) ◦ v = c(u ◦ v),
4. u ◦ u ≥ 0, and u ◦ u = 0 ⇐⇒ u = 0.
Definition 7.2. Let V be a vector space. An inner product on V is a function that, to
each pair u, v ∈ V , associates a real number 〈u, v〉 and satisfies the following four axioms
for u, v, w ∈ V and c ∈ R.
1. 〈u, v〉 = 〈v, u〉,
2. 〈u + v, w〉 = 〈u, w〉 + 〈v, w〉,
3. 〈cu, v〉 = c〈u, v〉,
4. 〈u, u〉 ≥ 0, and 〈u, u〉 = 0 ⇐⇒ u = 0.
A vector space with an inner product is called an inner product space.
Example 7.3. For u, v ∈ R2 , define
〈u, v〉 = 4u1 v1 + 5u2 v2 .
Then for any u, v, w ∈ R and c ∈ R
1. 〈u, v〉 = 4u1 v1 + 5u2 v2 = 4v1 u1 + 5v2 u2 = 〈v, u〉.
2. 〈u + v, w〉 = 4(u1 + v1 )w1 + 5(u2 + v2 )w2 = (4u1 w1 + 5u2 w2 ) + (4v1 w1 + 5v2 w2 ) =
〈u, w〉 + 〈v, w〉.
3. 〈cu, v〉 = 4(cu1 )v1 + 5(cu2 )v2 = c(4u1 v1 + 5u2 v2 ) = c〈u, v〉.
4. 〈u, u〉 = 4u21 + 5u22 ≥ 0 and 〈u, u〉 = 0 ⇐⇒ u = 0.
Therefore, 〈u, v〉 defines an inner product on R2 .
Example 7.4. For f, g ∈ Pn , define
󰁝 1
〈f, g〉 = f (t)g(t)dt.
−1

󰁕 1 that 〈f,
It is easy to show g〉 satisfies axioms (1), (2) and (3).
4. 〈f, f 〉 = −1 (f (t)) dt. This is equal to the area of the region between (f (t))2 and the
2

x-axis from t = −1 to t = 1. Hence 〈f, f 〉 ≥ 0 and 〈f, f 〉 = 0 if and only if f = 0.

Therefore, 〈f, g〉 defines an inner product on Pn .

13
Proposition 7.5. Let V be an inner product space. Then for any vector u ∈ V , one has
〈0, u〉 = 0.

Proof. 〈0, u〉 = 〈0 + 0, u〉 = 〈0, u〉 + 〈0, u〉. Hence 〈0, u〉 = 0.

Definition 7.6. Let V an inner product space. For u ∈ V , the norm of u is

󰁳
󰀂u󰀂 = 〈u, u〉.

For u, v ∈ V , u, v are orthogonal if 〈u, v〉 = 0. A unit vector is a vector of norm 1.

Similarly, we can define an orthogonal set, an orthogonal basis, an orthonormal set,
orthonormal basis.

Proposition 7.7. Let V be an inner product space. Then the zero vector 0 is orthogonal to
any vector in V .

Proof. This follows from Proposition 7.5.

Theorem 7.8. Let V be an inner product space. Then

1. 󰀂cu󰀂 = |c|󰀂u󰀂 for u ∈ V and c ∈ R.

2. (The Cauchy–Schwarz inequality) |〈u, v〉| ≤ 󰀂u󰀂 · 󰀂v󰀂 for u, v ∈ V .

3. (The triangle inequality) 󰀂u + v󰀂 ≤ 󰀂u󰀂 + 󰀂v󰀂 for u, v ∈ V .

Proof. (1) follows by the same proof for the usual inner product on Rn .
(2) If u = 0, then from Proposition 7.5 the inequality holds. Assume that u ∕= 0. Let
f (t) = 󰀂tu + v󰀂2 with a scalar t. Then one has f (t) = t2 󰀂u󰀂2 + 2t〈u, v〉 + 󰀂v󰀂2 . Since
󰀂u󰀂2 > 0, f (t) ≥ 0 if and only if the discriminant of f (t) is less than or equal to 0. Hence
〈u, v〉2 − 󰀂u󰀂2 · 󰀂v󰀂2 ≤ 0. Namely, 〈u, v〉2 ≤ 󰀂u󰀂2 · 󰀂v󰀂2 . This implies |〈u, v〉| ≤ 󰀂u󰀂 · 󰀂v󰀂.
(3) One has

󰀂u + v󰀂2 = 󰀂u󰀂2 + 2〈u, v〉 + 󰀂v󰀂2 ≤ 󰀂u󰀂2 + 2|〈u, v〉| + 󰀂v󰀂2

≤ 󰀂u󰀂2 + 2󰀂u󰀂 · 󰀂v󰀂 + 󰀂v󰀂2 (by (2))
= (󰀂u󰀂 + 󰀂v󰀂)2 .

Hence we obtain 󰀂u + v󰀂 ≤ 󰀂u󰀂 + 󰀂v󰀂.

We can prove the Gram–Schmidts process for an inner product space. Moreover, any
finite-dimensional inner product space has an orthogonal (orthonormal) basis.

Theorem 7.9. Let V be an n-dimensional inner product space and B = {b1 , . . . , bn } an

orthonormal basis for V . For u, v ∈ V with u = c1 b1 + · · · + cn bn and v = d1 b1 + · · · + dn bn ,
one has
〈u, v〉 = c1 d1 + · · · + cn dn = [u]B ◦ [v]B .

14
8 Complex vector spaces
We regard “scalars” as real numbers and studied real vector spaces. From now we regard
“scalars” as complex numbers. Namely, we consider complex vector spaces.
Remark 8.1. More general, we can consider a vector space over a field. The basic intuition
is that a field is a set where one can define addition, subtraction, multiplication, and division
with the usual properties. Standard examples are the real numbers R, and the complex
numbers C, the rational numbers Q, and F2 = {0, 1} and so on.
Example 8.2. Cn is a complex vector space, which is called the n-dimensional complex
space.
As is the case in real vector spaces, we can define subspaces, linear transformations,
linear independence, basis and dimensions and we can several propositions about such
properties.
Definition 8.3. Let V be a complex vector space. An Hermitian inner product on V is
a function that, to each pair u, v ∈ V , associates a complex number 〈u, v〉 and satisfies the
following four axioms for u, v, w ∈ V and c ∈ C.
1. 〈u, v〉 = 〈v, u〉,

2. 〈λu + v, w〉 = λ〈u, w〉 + 〈v, w〉,

3. 〈u, λv + w〉 = λ〈u, w〉 + 〈v, w〉,

4. 〈u, u〉 ≥ 0, and 〈u, u〉 = 0 ⇐⇒ u = 0,

√ √ √
where for a + b −1 with a, b ∈ R, a + b −1 denotes the complex conjugate a − b −1.
A vector space with an inner product is called an Hermitian space.
Example 8.4. For x, y ∈ Cn , define

〈x, y〉 = x1 y1 + · · · + xn yn .

Then 〈x, y〉 is not an Hermitian inner product. Indeed, when n = 1,

󰀍󰀅 √ 󰀆 󰀅 √ 󰀆󰀎 √ √ √
1 + −1 , 1 + −1 = (1 + −1)2 = 1 + 2 −1 − 1 = 2 −1 ∈ / R.

Example 8.5. For x, y ∈ Cn , define

〈x, y〉 = x1 y1 + · · · + xn yn .

Then the axioms (1), (2) and (3) are clearly satisfied.
√ √ √ √
(4) Since (a + b −1) · (a + b −1) = (a + b −1) · (a − b −1) = a2 + b2 ≥ 0, 〈x, x〉 ≥ 0.
Moreover, 〈x, x〉 = 0 ⇐⇒ x = 0. Therefore, 〈x, y〉 is an Hermitian inner product on V .
For an Hermitian space, we can define norm, orthogonality, orthogonal set, an or-
thogonal basis, an orthonormal set, orthonormal basis, and we can prove the Gram–
Schmidts process

15