Linear Algebra

Department of Mathematics
Indian Institute of Technology Guwahati

January – May 2019

MA 102 (RA, RKS, MGPP, KVK)

1 / 24
Basis and dimension

Topics:
Linear span
Subspaces
Linear independence
Basis, Dimension & Rank

2 / 24
Linear combination

Definition: A vector v in Rn is a linear combination of the vectors

v1 , v2 , . . . , vk in Rn if there exist real numbers c1 , c2 , . . . , ck such
that
v = c1 v1 + c2 v2 + . . . + ck vk .

3 / 24
Linear combination

Definition: A vector v in Rn is a linear combination of the vectors

v1 , v2 , . . . , vk in Rn if there exist real numbers c1 , c2 , . . . , ck such
that
v = c1 v1 + c2 v2 + . . . + ck vk .

The numbers c1 , c2 , . . . , ck are called the coefficients of the

linear combination.

3 / 24
Linear combination

Definition: A vector v in Rn is a linear combination of the vectors

v1 , v2 , . . . , vk in Rn if there exist real numbers c1 , c2 , . . . , ck such
that
v = c1 v1 + c2 v2 + . . . + ck vk .

The numbers c1 , c2 , . . . , ck are called the coefficients of the

linear combination.

Question: Is the vector [1, 2, 3]> a linear combination of [1, 0, 3]>

and [−1, 1, −3]> ?

3 / 24
Linear combination

Definition: A vector v in Rn is a linear combination of the vectors

v1 , v2 , . . . , vk in Rn if there exist real numbers c1 , c2 , . . . , ck such
that
v = c1 v1 + c2 v2 + . . . + ck vk .

The numbers c1 , c2 , . . . , ck are called the coefficients of the

linear combination.

Question: Is the vector [1, 2, 3]> a linear combination of [1, 0, 3]>

and [−1, 1, −3]> ?

Theorem: A system of linear equations with augmented matrix

[A | b] is consistent if and only if b is a linear combination of the
columns of A.

3 / 24
Span of vectors
Definition: Let S = {v1 , . . . , vk } ⊆ Rn . Then the collection of all
linear combinations of the vectors v1 , . . . , vk is called the span of S
(or span of the vectors v1 , . . . , vk ), and is denoted by span(S) (or
span(v1 , . . . , vk )).

4 / 24
Span of vectors
Definition: Let S = {v1 , . . . , vk } ⊆ Rn . Then the collection of all
linear combinations of the vectors v1 , . . . , vk is called the span of S
(or span of the vectors v1 , . . . , vk ), and is denoted by span(S) (or
span(v1 , . . . , vk )).

Thus

span(S) = {v ∈ Rn | v = c1 v1 +. . .+ck vk for some c1 , . . . , ck ∈ R}.

4 / 24
Span of vectors
Definition: Let S = {v1 , . . . , vk } ⊆ Rn . Then the collection of all
linear combinations of the vectors v1 , . . . , vk is called the span of S
(or span of the vectors v1 , . . . , vk ), and is denoted by span(S) (or
span(v1 , . . . , vk )).

Thus

span(S) = {v ∈ Rn | v = c1 v1 +. . .+ck vk for some c1 , . . . , ck ∈ R}.

Convention: span(∅) = {0}.

4 / 24
Span of vectors
Definition: Let S = {v1 , . . . , vk } ⊆ Rn . Then the collection of all
linear combinations of the vectors v1 , . . . , vk is called the span of S
(or span of the vectors v1 , . . . , vk ), and is denoted by span(S) (or
span(v1 , . . . , vk )).

Thus

span(S) = {v ∈ Rn | v = c1 v1 +. . .+ck vk for some c1 , . . . , ck ∈ R}.

Convention: span(∅) = {0}.

If span(S) = Rn , then S is called a spanning set for Rn .

4 / 24
Span of vectors
Definition: Let S = {v1 , . . . , vk } ⊆ Rn . Then the collection of all
linear combinations of the vectors v1 , . . . , vk is called the span of S
(or span of the vectors v1 , . . . , vk ), and is denoted by span(S) (or
span(v1 , . . . , vk )).

Thus

span(S) = {v ∈ Rn | v = c1 v1 +. . .+ck vk for some c1 , . . . , ck ∈ R}.

Convention: span(∅) = {0}.

If span(S) = Rn , then S is called a spanning set for Rn .

R2 = span(e1 , e2 ), where e1 = [1, 0]> and e2 = [0, 1]> .

4 / 24
Span of vectors
Definition: Let S = {v1 , . . . , vk } ⊆ Rn . Then the collection of all
linear combinations of the vectors v1 , . . . , vk is called the span of S
(or span of the vectors v1 , . . . , vk ), and is denoted by span(S) (or
span(v1 , . . . , vk )).

Thus

span(S) = {v ∈ Rn | v = c1 v1 +. . .+ck vk for some c1 , . . . , ck ∈ R}.

Convention: span(∅) = {0}.

If span(S) = Rn , then S is called a spanning set for Rn .

R2 = span(e1 , e2 ), where e1 = [1, 0]> and e2 = [0, 1]> .

Exercise: Let u = [1, 2, 3]> and v = [−1, 1, −3]> . Describe

span(u, v) geometrically.
4 / 24
Subspaces of Rn

Definition: A set U (6= ∅) ⊆ Rn is called a subspace of Rn if

au + bv ∈ U for every u, v ∈ U and for every a, b ∈ R.

5 / 24
Subspaces of Rn

Definition: A set U (6= ∅) ⊆ Rn is called a subspace of Rn if

au + bv ∈ U for every u, v ∈ U and for every a, b ∈ R.

U = {0} and U = Rn are subspaces of Rn , called trivial

subspaces of Rn .
Any subspace contains 0.

5 / 24
Subspaces of Rn

Definition: A set U (6= ∅) ⊆ Rn is called a subspace of Rn if

au + bv ∈ U for every u, v ∈ U and for every a, b ∈ R.

U = {0} and U = Rn are subspaces of Rn , called trivial

subspaces of Rn .
Any subspace contains 0.
U is a subspace iff U is closed under addition and scalar
multiplication.

5 / 24
Subspaces of Rn

Definition: A set U (6= ∅) ⊆ Rn is called a subspace of Rn if

au + bv ∈ U for every u, v ∈ U and for every a, b ∈ R.

U = {0} and U = Rn are subspaces of Rn , called trivial

subspaces of Rn .
Any subspace contains 0.
U is a subspace iff U is closed under addition and scalar
multiplication.

For any finite subset S of Rn , span(S) is a subspace of Rn .

5 / 24
Subspaces of Rn

Definition: A set U (6= ∅) ⊆ Rn is called a subspace of Rn if

au + bv ∈ U for every u, v ∈ U and for every a, b ∈ R.

U = {0} and U = Rn are subspaces of Rn , called trivial

subspaces of Rn .
Any subspace contains 0.
U is a subspace iff U is closed under addition and scalar
multiplication.

For any finite subset S of Rn , span(S) is a subspace of Rn .

Exercise: Examine whether the sets

S = {[x, y , z]> ∈ R3 : x = y + 1}, V = {[x, y , z]t ∈ R3 : x = 5y }
and U = {[x, y , z]t ∈ R3 : x = z 2 } are subspaces of R3 .

5 / 24
Direct sum of subspaces
Fact: Let A be an m × n matrix. Then U := {x ∈ Rn : Ax = 0} is
a subspace of Rn , called the nullspace of A.

6 / 24
Direct sum of subspaces
Fact: Let A be an m × n matrix. Then U := {x ∈ Rn : Ax = 0} is
a subspace of Rn , called the nullspace of A.

Definition: Let U and V be two subspaces of Rn . Then

U + V := {u + v : u ∈ U, v ∈ V }

is called the sum of the subspaces U and V .

6 / 24
Direct sum of subspaces
Fact: Let A be an m × n matrix. Then U := {x ∈ Rn : Ax = 0} is
a subspace of Rn , called the nullspace of A.

Definition: Let U and V be two subspaces of Rn . Then

U + V := {u + v : u ∈ U, v ∈ V }

is called the sum of the subspaces U and V .

Definition: Let U and V be two subspaces of Rn . If U ∩ V = {0}

then the sum U + V is called the direct sum of U and V and is
denoted by U ⊕ V . Thus

U ⊕ V = U + V and U ∩ V = {0}.

6 / 24
Direct sum of subspaces
Fact: Let A be an m × n matrix. Then U := {x ∈ Rn : Ax = 0} is
a subspace of Rn , called the nullspace of A.

Definition: Let U and V be two subspaces of Rn . Then

U + V := {u + v : u ∈ U, v ∈ V }

is called the sum of the subspaces U and V .

Definition: Let U and V be two subspaces of Rn . If U ∩ V = {0}

then the sum U + V is called the direct sum of U and V and is
denoted by U ⊕ V . Thus

U ⊕ V = U + V and U ∩ V = {0}.

Fact: Let U and V be subspaces of Rn . Then U + V and U ⊕ V

are subspaces of Rn . If z ∈ U ⊕ V then there exist unique u ∈ U
and v ∈ V such that z = u + v.
6 / 24
Linear dependence
Definition: A set {v1 , v2 , . . . , vk } of vectors in Rn is said to be
linearly dependent if one of the vectors vi is a linear combination
of the rest,

7 / 24
Linear dependence
Definition: A set {v1 , v2 , . . . , vk } of vectors in Rn is said to be
linearly dependent if one of the vectors vi is a linear combination
of the rest, i.e., if there are real numbers c1 , c2 , . . . , ck of which at
least one is nonzero such that

c1 v1 + c2 v2 + . . . + ck vk = 0.

7 / 24
Linear dependence
Definition: A set {v1 , v2 , . . . , vk } of vectors in Rn is said to be
linearly dependent if one of the vectors vi is a linear combination
of the rest, i.e., if there are real numbers c1 , c2 , . . . , ck of which at
least one is nonzero such that

c1 v1 + c2 v2 + . . . + ck vk = 0.

We say that the vectors v1 , v2 , . . . , vk are linearly dependent,

to mean that the set {v1 , v2 , . . . , vk } is linearly dependent.

7 / 24
Linear dependence
Definition: A set {v1 , v2 , . . . , vk } of vectors in Rn is said to be
linearly dependent if one of the vectors vi is a linear combination
of the rest, i.e., if there are real numbers c1 , c2 , . . . , ck of which at
least one is nonzero such that

c1 v1 + c2 v2 + . . . + ck vk = 0.

We say that the vectors v1 , v2 , . . . , vk are linearly dependent,

to mean that the set {v1 , v2 , . . . , vk } is linearly dependent.

Any set of vectors containing the 0 is linearly dependent.

7 / 24
Linear dependence
Definition: A set {v1 , v2 , . . . , vk } of vectors in Rn is said to be
linearly dependent if one of the vectors vi is a linear combination
of the rest, i.e., if there are real numbers c1 , c2 , . . . , ck of which at
least one is nonzero such that

c1 v1 + c2 v2 + . . . + ck vk = 0.

We say that the vectors v1 , v2 , . . . , vk are linearly dependent,

to mean that the set {v1 , v2 , . . . , vk } is linearly dependent.

Any set of vectors containing the 0 is linearly dependent.

Exercise: Examine whether the sets

U := {[1, 2, 0]> , [1, 1, −1]> , [1, 4, 2]> } and S := {[1, 4]> , [−1, 2]> }
are linearly dependent.

7 / 24
Linear independence
Definition: A set S = {v1 , v2 , . . . , vk } of vectors in Rn is said to be
linearly independent if S is NOT linearly dependent.

8 / 24
Linear independence
Definition: A set S = {v1 , v2 , . . . , vk } of vectors in Rn is said to be
linearly independent if S is NOT linearly dependent.

S is linearly independent iff

c1 v1 + c2 v2 + . . . + ck vk = 0 ⇒

8 / 24
Linear independence
Definition: A set S = {v1 , v2 , . . . , vk } of vectors in Rn is said to be
linearly independent if S is NOT linearly dependent.

S is linearly independent iff

c1 v1 + c2 v2 + . . . + ck vk = 0 ⇒c1 = c2 = . . . = ck = 0.

8 / 24
Linear independence
Definition: A set S = {v1 , v2 , . . . , vk } of vectors in Rn is said to be
linearly independent if S is NOT linearly dependent.

S is linearly independent iff

c1 v1 + c2 v2 + . . . + ck vk = 0 ⇒c1 = c2 = . . . = ck = 0.

We say that the vectors v1 , v2 , . . . , vk are linearly independent,

to mean that the set {v1 , v2 , . . . , vk } is linearly independent.

8 / 24
Linear independence
Definition: A set S = {v1 , v2 , . . . , vk } of vectors in Rn is said to be
linearly independent if S is NOT linearly dependent.

S is linearly independent iff

c1 v1 + c2 v2 + . . . + ck vk = 0 ⇒c1 = c2 = . . . = ck = 0.

We say that the vectors v1 , v2 , . . . , vk are linearly independent,

to mean that the set {v1 , v2 , . . . , vk } is linearly independent.

Question: Let ei ∈ Rn be the i-th column of the identity matrix In .

8 / 24
Linear independence
Definition: A set S = {v1 , v2 , . . . , vk } of vectors in Rn is said to be
linearly independent if S is NOT linearly dependent.

S is linearly independent iff

c1 v1 + c2 v2 + . . . + ck vk = 0 ⇒c1 = c2 = . . . = ck = 0.

We say that the vectors v1 , v2 , . . . , vk are linearly independent,

to mean that the set {v1 , v2 , . . . , vk } is linearly independent.

Question: Let ei ∈ Rn be the i-th column of the identity matrix In .

Is {e1 , e2 , . . . , en } linearly independent?

8 / 24
Linear independence
Definition: A set S = {v1 , v2 , . . . , vk } of vectors in Rn is said to be
linearly independent if S is NOT linearly dependent.

S is linearly independent iff

c1 v1 + c2 v2 + . . . + ck vk = 0 ⇒c1 = c2 = . . . = ck = 0.

We say that the vectors v1 , v2 , . . . , vk are linearly independent,

to mean that the set {v1 , v2 , . . . , vk } is linearly independent.

Question: Let ei ∈ Rn be the i-th column of the identity matrix In .

Is {e1 , e2 , . . . , en } linearly independent?

Fact: Let S := {v1 , v2 , . . . , vm } ⊆ Rn . Consider the n × m matrix

A := [v1 v2 . . . vm ]. Then S is linearly dependent iff the system
Ax = 0 has a non-trivial solution.

8 / 24
Linear combinations of rows
 
A1
Let A :=  ...  be an m × n matrix. Then
 

Am

9 / 24
Linear combinations of rows
 
A1
Let A :=  ...  be an m × n matrix. Then
 

Am
For ci ∈ R, a := c1 A1 + . . . cm Am is a linear combination of
the rows of A.

9 / 24
Linear combinations of rows
 
A1
Let A :=  ...  be an m × n matrix. Then
 

Am
For ci ∈ R, a := c1 A1 + . . . cm Am is a linear combination of
the rows of A. Note that a is an 1 × n matrix and a> ∈ Rn .

9 / 24
Linear combinations of rows
 
A1
Let A :=  ...  be an m × n matrix. Then
 

Am
For ci ∈ R, a := c1 A1 + . . . cm Am is a linear combination of
the rows of A. Note that a is an 1 × n matrix and a> ∈ Rn .
Note: c1 A1 + . . . cm Am = [c1 , . . . , cm ]A.

9 / 24
Linear combinations of rows
 
A1
Let A :=  ...  be an m × n matrix. Then
 

Am
For ci ∈ R, a := c1 A1 + . . . cm Am is a linear combination of
the rows of A. Note that a is an 1 × n matrix and a> ∈ Rn .
Note: c1 A1 + . . . cm Am = [c1 , . . . , cm ]A. Thus, for any
c ∈ Rm , c> A is a linear combination of rows of A.

9 / 24
Linear combinations of rows
 
A1
Let A :=  ...  be an m × n matrix. Then
 

Am
For ci ∈ R, a := c1 A1 + . . . cm Am is a linear combination of
the rows of A. Note that a is an 1 × n matrix and a> ∈ Rn .
Note: c1 A1 + . . . cm Am = [c1 , . . . , cm ]A. Thus, for any
c ∈ Rm , c> A is a linear combination of rows of A.
The rows of A are linearly dependent iff
c> A = c1 A1 + . . . cm Am = 0 (zero row) for some nonzero
c ∈ Rm .

9 / 24
Linear combinations of rows
 
A1
Let A :=  ...  be an m × n matrix. Then
 

Am
For ci ∈ R, a := c1 A1 + . . . cm Am is a linear combination of
the rows of A. Note that a is an 1 × n matrix and a> ∈ Rn .
Note: c1 A1 + . . . cm Am = [c1 , . . . , cm ]A. Thus, for any
c ∈ Rm , c> A is a linear combination of rows of A.
The rows of A are linearly dependent iff
c> A = c1 A1 + . . . cm Am = 0 (zero row) for some nonzero
c ∈ Rm .

9 / 24
Linear combinations of rows
 
A1
Let A :=  ...  be an m × n matrix. Then
 

Am
For ci ∈ R, a := c1 A1 + . . . cm Am is a linear combination of
the rows of A. Note that a is an 1 × n matrix and a> ∈ Rn .
Note: c1 A1 + . . . cm Am = [c1 , . . . , cm ]A. Thus, for any
c ∈ Rm , c> A is a linear combination of rows of A.
The rows of A are linearly dependent iff
c> A = c1 A1 + . . . cm Am = 0 (zero row) for some nonzero
c ∈ Rm .
The rows of A are linearly dependent iff A> >
1 , . . . , Am are
n
linearly dependent in R ,

9 / 24
Linear combinations of rows
 
A1
Let A :=  ...  be an m × n matrix. Then
 

Am
For ci ∈ R, a := c1 A1 + . . . cm Am is a linear combination of
the rows of A. Note that a is an 1 × n matrix and a> ∈ Rn .
Note: c1 A1 + . . . cm Am = [c1 , . . . , cm ]A. Thus, for any
c ∈ Rm , c> A is a linear combination of rows of A.
The rows of A are linearly dependent iff
c> A = c1 A1 + . . . cm Am = 0 (zero row) for some nonzero
c ∈ Rm .
The rows of A are linearly dependent iff A> >
1 , . . . , Am are
n >
linearly dependent in R , i.e., the columns of A are linearly
dependent.

9 / 24
Linearly dependent rows
Theorem: Let S := {v1 , v2 , . . . , vm } ⊆ Rn and A := [v1 · · · vm ].
Then the following are equivalent.
1 S is linearly dependent.

10 / 24
Linearly dependent rows
Theorem: Let S := {v1 , v2 , . . . , vm } ⊆ Rn and A := [v1 · · · vm ].
Then the following are equivalent.
1 S is linearly dependent.
2 Columns of A are linearly dependent.

10 / 24
Linearly dependent rows
Theorem: Let S := {v1 , v2 , . . . , vm } ⊆ Rn and A := [v1 · · · vm ].
Then the following are equivalent.
1 S is linearly dependent.
2 Columns of A are linearly dependent.
3 Ax = 0 has a nontrivial solution.

10 / 24
Linearly dependent rows
Theorem: Let S := {v1 , v2 , . . . , vm } ⊆ Rn and A := [v1 · · · vm ].
Then the following are equivalent.
1 S is linearly dependent.
2 Columns of A are linearly dependent.
3 Ax = 0 has a nontrivial solution.
4 Rows of A> are linearly dependent.

10 / 24
Linearly dependent rows
Theorem: Let S := {v1 , v2 , . . . , vm } ⊆ Rn and A := [v1 · · · vm ].
Then the following are equivalent.
1 S is linearly dependent.
2 Columns of A are linearly dependent.
3 Ax = 0 has a nontrivial solution.
4 Rows of A> are linearly dependent.
5 rank(A> ) < m.

10 / 24
Linearly dependent rows
Theorem: Let S := {v1 , v2 , . . . , vm } ⊆ Rn and A := [v1 · · · vm ].
Then the following are equivalent.
1 S is linearly dependent.
2 Columns of A are linearly dependent.
3 Ax = 0 has a nontrivial solution.
4 Rows of A> are linearly dependent.
5 rank(A> ) < m.
6 rref(A> ) has a zero row.
Proof: (1) ⇒ (2) ⇒ (3) trivial. Suppose (3) holds. Then

10 / 24
Linearly dependent rows
Theorem: Let S := {v1 , v2 , . . . , vm } ⊆ Rn and A := [v1 · · · vm ].
Then the following are equivalent.
1 S is linearly dependent.
2 Columns of A are linearly dependent.
3 Ax = 0 has a nontrivial solution.
4 Rows of A> are linearly dependent.
5 rank(A> ) < m.
6 rref(A> ) has a zero row.
Proof: (1) ⇒ (2) ⇒ (3) trivial. Suppose (3) holds. Then
x> A> = 0 ⇒ x1 A1 + · · · + xm Am = 0 ⇒ (4) holds.

10 / 24
Linearly dependent rows
Theorem: Let S := {v1 , v2 , . . . , vm } ⊆ Rn and A := [v1 · · · vm ].
Then the following are equivalent.
1 S is linearly dependent.
2 Columns of A are linearly dependent.
3 Ax = 0 has a nontrivial solution.
4 Rows of A> are linearly dependent.
5 rank(A> ) < m.
6 rref(A> ) has a zero row.
Proof: (1) ⇒ (2) ⇒ (3) trivial. Suppose (3) holds. Then
x> A> = 0 ⇒ x1 A1 + · · · + xm Am = 0 ⇒ (4) holds.
Suppose (4) holds. Then rref(A> ) has a zero row ⇒ (5) holds.
Now (5) ⇒ (6) is immediate.

10 / 24
Linearly dependent rows
Theorem: Let S := {v1 , v2 , . . . , vm } ⊆ Rn and A := [v1 · · · vm ].
Then the following are equivalent.
1 S is linearly dependent.
2 Columns of A are linearly dependent.
3 Ax = 0 has a nontrivial solution.
4 Rows of A> are linearly dependent.
5 rank(A> ) < m.
6 rref(A> ) has a zero row.
Proof: (1) ⇒ (2) ⇒ (3) trivial. Suppose (3) holds. Then
x> A> = 0 ⇒ x1 A1 + · · · + xm Am = 0 ⇒ (4) holds.
Suppose (4) holds. Then rref(A> ) has a zero row ⇒ (5) holds.
Now (5) ⇒ (6) is immediate.
Suppose (6) holds. Then EA> = rref(A> ) for some invertible
matrix E . Now e> > >
m rref(A ) = 0 ⇒ Ay = 0, where y := E em .
10 / 24
Basis
Corollary: If m > n then any set of m vectors in Rn is linearly
dependent.

11 / 24
Basis
Corollary: If m > n then any set of m vectors in Rn is linearly
dependent.
Definition: Let S be a subspace of Rn and B ⊆ S. Then B is said
to be a basis for S iff B is linearly independent and span(B) = S.

11 / 24
Basis
Corollary: If m > n then any set of m vectors in Rn is linearly
dependent.
Definition: Let S be a subspace of Rn and B ⊆ S. Then B is said
to be a basis for S iff B is linearly independent and span(B) = S.

The set {1} is a basis for R1 (= R).

11 / 24
Basis
Corollary: If m > n then any set of m vectors in Rn is linearly
dependent.
Definition: Let S be a subspace of Rn and B ⊆ S. Then B is said
to be a basis for S iff B is linearly independent and span(B) = S.

The set {1} is a basis for R1 (= R).

The standard unit vector ei ∈ Rn is the i-th column of the

identity matrix In . The set {e1 , e2 , . . . , en } is a basis for Rn
and is called the standard basis.

11 / 24
Basis
Corollary: If m > n then any set of m vectors in Rn is linearly
dependent.
Definition: Let S be a subspace of Rn and B ⊆ S. Then B is said
to be a basis for S iff B is linearly independent and span(B) = S.

The set {1} is a basis for R1 (= R).

The standard unit vector ei ∈ Rn is the i-th column of the

identity matrix In . The set {e1 , e2 , . . . , en } is a basis for Rn
and is called the standard basis.

Exercise: Find a basis for the subspace S := {x ∈ R4 : Ax = 0},

where  
1 −1 −1 2
A =  2 −2 −1 3 .
−1 1 −1 0

11 / 24
Basis
Theorem: Let S := {v1 , . . . , vr } ∈ Rn and U ⊆ span(S) such that
m := #(U) > r . Then U is linearly dependent.

12 / 24
Basis
Theorem: Let S := {v1 , . . . , vr } ∈ Rn and U ⊆ span(S) such that
m := #(U) > r . Then U is linearly dependent.
Proof. Let U := {u1 , . . . , um }.

12 / 24
Basis
Theorem: Let S := {v1 , . . . , vr } ∈ Rn and U ⊆ span(S) such that
m := #(U) > r . Then U is linearly dependent.
Proof. Let U := {u1 , . . . , um }. Then

ui = ai1 v1 + ai2 v2 + · · · + air vr , 1 ≤ i ≤ m.

12 / 24
Basis
Theorem: Let S := {v1 , . . . , vr } ∈ Rn and U ⊆ span(S) such that
m := #(U) > r . Then U is linearly dependent.
Proof. Let U := {u1 , . . . , um }. Then

ui = ai1 v1 + ai2 v2 + · · · + air vr , 1 ≤ i ≤ m.

   
a11 · · · a1r A1
Let A := 
 .. . . ..  =  .. .
. . .   . 
am1 · · · amr Am

12 / 24
Basis
Theorem: Let S := {v1 , . . . , vr } ∈ Rn and U ⊆ span(S) such that
m := #(U) > r . Then U is linearly dependent.
Proof. Let U := {u1 , . . . , um }. Then

ui = ai1 v1 + ai2 v2 + · · · + air vr , 1 ≤ i ≤ m.

     
a11 · · · a1r A1 v1
Let A := 
 .. . . ..  =  .. . Then u = A  .. .
. . .   .  i i . 
am1 · · · amr Am vr

12 / 24
Basis
Theorem: Let S := {v1 , . . . , vr } ∈ Rn and U ⊆ span(S) such that
m := #(U) > r . Then U is linearly dependent.
Proof. Let U := {u1 , . . . , um }. Then

ui = ai1 v1 + ai2 v2 + · · · + air vr , 1 ≤ i ≤ m.

     
a11 · · · a1r A1 v1
Let A := 
 .. . . ..  =  .. . Then u = A  .. .
. . .   .  i i . 
am1 · · · amr Am vr
Since m > r , the rows of A are linearly dependent.

12 / 24
Basis
Theorem: Let S := {v1 , . . . , vr } ∈ Rn and U ⊆ span(S) such that
m := #(U) > r . Then U is linearly dependent.
Proof. Let U := {u1 , . . . , um }. Then

ui = ai1 v1 + ai2 v2 + · · · + air vr , 1 ≤ i ≤ m.

     
a11 · · · a1r A1 v1
Let A := 
 .. . . ..  =  .. . Then u = A  .. .
. . .   .  i i . 
am1 · · · amr Am vr
Since m > r , the rows of A are linearly dependent. Then there exist
αj ∈ R, j = 1 : m, such that α1 A1 + · · · + αm Am = 01×r .

12 / 24
Basis
Theorem: Let S := {v1 , . . . , vr } ∈ Rn and U ⊆ span(S) such that
m := #(U) > r . Then U is linearly dependent.
Proof. Let U := {u1 , . . . , um }. Then

ui = ai1 v1 + ai2 v2 + · · · + air vr , 1 ≤ i ≤ m.

     
a11 · · · a1r A1 v1
Let A := 
 .. . . ..  =  .. . Then u = A  .. .
. . .   .  i i . 
am1 · · · amr Am vr
Since m > r , the rows of A are linearly dependent. Then there exist
αj ∈ R, j = 1 : m, such that α1 A1 + · · · + αm Am = 01×r . Hence

m
X
αi ui =
i=1

12 / 24
Basis
Theorem: Let S := {v1 , . . . , vr } ∈ Rn and U ⊆ span(S) such that
m := #(U) > r . Then U is linearly dependent.
Proof. Let U := {u1 , . . . , um }. Then

ui = ai1 v1 + ai2 v2 + · · · + air vr , 1 ≤ i ≤ m.

     
a11 · · · a1r A1 v1
Let A := 
 .. . . ..  =  .. . Then u = A  .. .
. . .   .  i i . 
am1 · · · amr Am vr
Since m > r , the rows of A are linearly dependent. Then there exist
αj ∈ R, j = 1 : m, such that α1 A1 + · · · + αm Am = 01×r . Hence
 
m m v1
αi Ai  ...  =
X X
αi ui =
 
i=1 i=1 vr

12 / 24
Basis
Theorem: Let S := {v1 , . . . , vr } ∈ Rn and U ⊆ span(S) such that
m := #(U) > r . Then U is linearly dependent.
Proof. Let U := {u1 , . . . , um }. Then

ui = ai1 v1 + ai2 v2 + · · · + air vr , 1 ≤ i ≤ m.

     
a11 · · · a1r A1 v1
Let A := 
 .. . . ..  =  .. . Then u = A  .. .
. . .   .  i i . 
am1 · · · amr Am vr
Since m > r , the rows of A are linearly dependent. Then there exist
αj ∈ R, j = 1 : m, such that α1 A1 + · · · + αm Am = 01×r . Hence
   
m m v1 v1
αi Ai  ...  = 01×r  ...  =
X X
αi ui =
   
i=1 i=1 vr vr

12 / 24
Basis
Theorem: Let S := {v1 , . . . , vr } ∈ Rn and U ⊆ span(S) such that
m := #(U) > r . Then U is linearly dependent.
Proof. Let U := {u1 , . . . , um }. Then

ui = ai1 v1 + ai2 v2 + · · · + air vr , 1 ≤ i ≤ m.

     
a11 · · · a1r A1 v1
Let A := 
 .. . . ..  =  .. . Then u = A  .. .
. . .   .  i i . 
am1 · · · amr Am vr
Since m > r , the rows of A are linearly dependent. Then there exist
αj ∈ R, j = 1 : m, such that α1 A1 + · · · + αm Am = 01×r . Hence
   
m m v1 v1
αi Ai  ...  = 01×r  ...  = 0.
X X
αi ui =
   
i=1 i=1 vr vr

12 / 24
Basis

Theorem: Let U be a subspace of Rn . Then U has a basis and any

two bases of U have the same number of elements.

13 / 24
Basis

Theorem: Let U be a subspace of Rn . Then U has a basis and any

two bases of U have the same number of elements.
Dimension: The number of elements in a basis of a subspace U of
Rn is called the dimension of U and is denoted by dim(U).

13 / 24
Basis

Theorem: Let U be a subspace of Rn . Then U has a basis and any

two bases of U have the same number of elements.
Dimension: The number of elements in a basis of a subspace U of
Rn is called the dimension of U and is denoted by dim(U).

dim(Rn ) =

13 / 24
Basis

Theorem: Let U be a subspace of Rn . Then U has a basis and any

two bases of U have the same number of elements.
Dimension: The number of elements in a basis of a subspace U of
Rn is called the dimension of U and is denoted by dim(U).

dim(Rn ) = n.

dim({0}) =

13 / 24
Basis

Theorem: Let U be a subspace of Rn . Then U has a basis and any

two bases of U have the same number of elements.
Dimension: The number of elements in a basis of a subspace U of
Rn is called the dimension of U and is denoted by dim(U).

dim(Rn ) = n.

dim({0}) = 0, since span({}) = {0}).

13 / 24
Basis

Theorem: Let U be a subspace of Rn . Then U has a basis and any

two bases of U have the same number of elements.
Dimension: The number of elements in a basis of a subspace U of
Rn is called the dimension of U and is denoted by dim(U).

dim(Rn ) = n.

dim({0}) = 0, since span({}) = {0}).

If v1 , . . . , vm are linearly independent, then

dim(span(v1 , . . . , vm )) =

13 / 24
Basis

Theorem: Let U be a subspace of Rn . Then U has a basis and any

two bases of U have the same number of elements.
Dimension: The number of elements in a basis of a subspace U of
Rn is called the dimension of U and is denoted by dim(U).

dim(Rn ) = n.

dim({0}) = 0, since span({}) = {0}).

If v1 , . . . , vm are linearly independent, then

dim(span(v1 , . . . , vm )) = m.

13 / 24
Basis

Theorem: Let U be a subspace of Rn . Then U has a basis and any

two bases of U have the same number of elements.
Dimension: The number of elements in a basis of a subspace U of
Rn is called the dimension of U and is denoted by dim(U).

dim(Rn ) = n.

dim({0}) = 0, since span({}) = {0}).

If v1 , . . . , vm are linearly independent, then

dim(span(v1 , . . . , vm )) = m.

A set S := {v1 , . . . , vn } ⊆ Rn is a basis of Rn ⇐⇒ S is

linearly independent

13 / 24
Basis

Theorem: Let U be a subspace of Rn . Then U has a basis and any

two bases of U have the same number of elements.
Dimension: The number of elements in a basis of a subspace U of
Rn is called the dimension of U and is denoted by dim(U).

dim(Rn ) = n.

dim({0}) = 0, since span({}) = {0}).

If v1 , . . . , vm are linearly independent, then

dim(span(v1 , . . . , vm )) = m.

A set S := {v1 , . . . , vn } ⊆ Rn is a basis of Rn ⇐⇒ S is

linearly independent ⇐⇒ span(S) = Rn .

13 / 24
Fundamental subspaces associated to a matrix
Definition: Let A be an m × n matrix.
1 The column space / range space of A, denoted col(A), is the
subspace of Rm spanned by the columns of A.

14 / 24
Fundamental subspaces associated to a matrix
Definition: Let A be an m × n matrix.
1 The column space / range space of A, denoted col(A), is the
subspace of Rm spanned by the columns of A.
In other words, col(A) := {Ax | x ∈ Rn }.

14 / 24
Fundamental subspaces associated to a matrix
Definition: Let A be an m × n matrix.
1 The column space / range space of A, denoted col(A), is the
subspace of Rm spanned by the columns of A.
In other words, col(A) := {Ax | x ∈ Rn }.
2 The row space of A, denoted row(A), is the subspace of Rn
spanned by the rows of A.

14 / 24
Fundamental subspaces associated to a matrix
Definition: Let A be an m × n matrix.
1 The column space / range space of A, denoted col(A), is the
subspace of Rm spanned by the columns of A.
In other words, col(A) := {Ax | x ∈ Rn }.
2 The row space of A, denoted row(A), is the subspace of Rn
spanned by the rows of A. In other words,
row(A) := {x> A | x ∈ Rm }
[Here, elements of row(A) are row vectors. How can they be
elements of Rn .

14 / 24
Fundamental subspaces associated to a matrix
Definition: Let A be an m × n matrix.
1 The column space / range space of A, denoted col(A), is the
subspace of Rm spanned by the columns of A.
In other words, col(A) := {Ax | x ∈ Rn }.
2 The row space of A, denoted row(A), is the subspace of Rn
spanned by the rows of A. In other words,
row(A) := {x> A | x ∈ Rm }
[Here, elements of row(A) are row vectors. How can they be
elements of Rn . In strict sense, row(A) := col(A> ).]

14 / 24
Fundamental subspaces associated to a matrix
Definition: Let A be an m × n matrix.
1 The column space / range space of A, denoted col(A), is the
subspace of Rm spanned by the columns of A.
In other words, col(A) := {Ax | x ∈ Rn }.
2 The row space of A, denoted row(A), is the subspace of Rn
spanned by the rows of A. In other words,
row(A) := {x> A | x ∈ Rm }
[Here, elements of row(A) are row vectors. How can they be
elements of Rn . In strict sense, row(A) := col(A> ).]
3 The null space of A, denoted null(A), is the subspace of Rn
consisting of the solutions of the homogeneous linear system
Ax = 0.

14 / 24
Fundamental subspaces associated to a matrix
Definition: Let A be an m × n matrix.
1 The column space / range space of A, denoted col(A), is the
subspace of Rm spanned by the columns of A.
In other words, col(A) := {Ax | x ∈ Rn }.
2 The row space of A, denoted row(A), is the subspace of Rn
spanned by the rows of A. In other words,
row(A) := {x> A | x ∈ Rm }
[Here, elements of row(A) are row vectors. How can they be
elements of Rn . In strict sense, row(A) := col(A> ).]
3 The null space of A, denoted null(A), is the subspace of Rn
consisting of the solutions of the homogeneous linear system
Ax = 0. In other words, null(A) := {x ∈ Rn | Ax = 0}

14 / 24
Fundamental subspaces associated to a matrix
Definition: Let A be an m × n matrix.
1 The column space / range space of A, denoted col(A), is the
subspace of Rm spanned by the columns of A.
In other words, col(A) := {Ax | x ∈ Rn }.
2 The row space of A, denoted row(A), is the subspace of Rn
spanned by the rows of A. In other words,
row(A) := {x> A | x ∈ Rm }
[Here, elements of row(A) are row vectors. How can they be
elements of Rn . In strict sense, row(A) := col(A> ).]
3 The null space of A, denoted null(A), is the subspace of Rn
consisting of the solutions of the homogeneous linear system
Ax = 0. In other words, null(A) := {x ∈ Rn | Ax = 0}
4 The null space of A> : null(A> ) = {x ∈ Rm | A> x = 0}.

14 / 24
Bases of row spaces

Theorem: If two matrices A and B are row equivalent, then

row(B) = row(A)

15 / 24
Bases of row spaces

Theorem: If two matrices A and B are row equivalent, then

row(B) = row(A)

. Proof. A and B are row equivalent ⇒ B = PA, for some

invertible P.

15 / 24
Bases of row spaces

Theorem: If two matrices A and B are row equivalent, then

row(B) = row(A)

. Proof. A and B are row equivalent ⇒ B = PA, for some

invertible P. Thus,

row(B) = {x> B | x ∈ Rn } = {(x> P)A | x ∈ Rn } ⊆ row(A).

15 / 24
Bases of row spaces

Theorem: If two matrices A and B are row equivalent, then

row(B) = row(A)

. Proof. A and B are row equivalent ⇒ B = PA, for some

invertible P. Thus,

row(B) = {x> B | x ∈ Rn } = {(x> P)A | x ∈ Rn } ⊆ row(A).

Similarly, row(A) ⊆ row(B), since A = P −1 B. 

15 / 24
Bases of row spaces

Theorem: If two matrices A and B are row equivalent, then

row(B) = row(A)

. Proof. A and B are row equivalent ⇒ B = PA, for some

invertible P. Thus,

row(B) = {x> B | x ∈ Rn } = {(x> P)A | x ∈ Rn } ⊆ row(A).

Similarly, row(A) ⊆ row(B), since A = P −1 B. 

Corollary: For any A, row(A) = row(rref(A)).

15 / 24
Bases of row spaces

Theorem: If two matrices A and B are row equivalent, then

row(B) = row(A)

. Proof. A and B are row equivalent ⇒ B = PA, for some

invertible P. Thus,

row(B) = {x> B | x ∈ Rn } = {(x> P)A | x ∈ Rn } ⊆ row(A).

Similarly, row(A) ⊆ row(B), since A = P −1 B. 

Corollary: For any A, row(A) = row(rref(A)).

Corollary: For any matrix A, the non-zero rows of rref(A) forms a

basis of row(A).

15 / 24
Bases of column spaces

Question: (a) Suppose A and B are row-equivalent. Are col(A)

and col(B) equal?

16 / 24
Bases of column spaces

Question: (a) Suppose A and B are row-equivalent. Are col(A)

and col(B) equal? No.

16 / 24
Bases of column spaces

Question: (a) Suppose A and B are

 row-equivalent.
  Are col(A)

1 1 1 1
and col(B) equal? No. Take A = and B = .
1 1 0 0

16 / 24
Bases of column spaces

Question: (a) Suppose A and B are

 row-equivalent.
  Are col(A)

1 1 1 1
and col(B) equal? No. Take A = and B = .
1 1 0 0
(b) Suppose A and B are row-equivalent. Do col(A) and col(B)
have same dimension?

16 / 24
Bases of column spaces

Question: (a) Suppose A and B are

 row-equivalent.
  Are col(A)

1 1 1 1
and col(B) equal? No. Take A = and B = .
1 1 0 0
(b) Suppose A and B are row-equivalent. Do col(A) and col(B)
have same dimension? Yes. We will see soon.

16 / 24
Bases of column spaces

Question: (a) Suppose A and B are

 row-equivalent.
  Are col(A)

1 1 1 1
and col(B) equal? No. Take A = and B = .
1 1 0 0
(b) Suppose A and B are row-equivalent. Do col(A) and col(B)
have same dimension? Yes. We will see soon.

Theorem: Let P be an invertible matrix. Then a set

{v1 , v2 , . . . , vm } in Rn is linearly independent iff the set
{Pv1 , Pv2 , . . . , Pvm } is linearly independent.

16 / 24
Bases of column spaces

Question: (a) Suppose A and B are

 row-equivalent.
  Are col(A)

1 1 1 1
and col(B) equal? No. Take A = and B = .
1 1 0 0
(b) Suppose A and B are row-equivalent. Do col(A) and col(B)
have same dimension? Yes. We will see soon.

Theorem: Let P be an invertible matrix. Then a set

{v1 , v2 , . . . , vm } in Rn is linearly independent iff the set
{Pv1 , Pv2 , . . . , Pvm } is linearly independent.

Corollary: Let A := [a1 a2 · · · an ] and rref(A) = [b1 b2 · · · bn ]. If

are bj1 , bj2 , . . . , bjr are pivot columns of rref(A), then
{aj1 , aj2 , . . . , ajr } is a basis of col(A).

16 / 24
Algorithm for computing bases of null spaces

Input: An m × n matrix A.
Output: A matrix X whose columns form a basis of the null
space of A.

17 / 24
Algorithm for computing bases of null spaces

Input: An m × n matrix A.
Output: A matrix X whose columns form a basis of the null
space of A.
1. Compute R = rref(A).

17 / 24
Algorithm for computing bases of null spaces

Input: An m × n matrix A.
Output: A matrix X whose columns form a basis of the null
space of A.
1. Compute R = rref(A).

2. Suppose that R has p-nonzero rows. So it has p-pivot

columns. Interchange columns of R (i.e., choose a
permutation matrix P) so that
 
Ip F
RP = = column interchanged form of R,
0 0

where Ip is the identity matrix of size p.

17 / 24
Bases of null spaces

 
−F
3. Set Y := , where In−p is the identity matrix of size
In−p
n − p.

18 / 24
Bases of null spaces

 
−F
3. Set Y := , where In−p is the identity matrix of size
In−p
n − p.

4. Now interchange rows of Y according to the permutation P.

This means compute
X := PY .

18 / 24
Bases of null spaces

 
−F
3. Set Y := , where In−p is the identity matrix of size
In−p
n − p.

4. Now interchange rows of Y according to the permutation P.

This means compute
X := PY .
Then rank(X ) = n − p and RX = RPY = 0. Thus columns of
X span the null space of R and hence the null space of A.

18 / 24
Example
Compute bases of the null space, row space and the column space
of the matrix  
1 3 3 2
A :=  2 6 9 7 .
−1 −3 3 4

19 / 24
Example
Compute bases of the null space, row space and the column space
of the matrix  
1 3 3 2
A :=  2 6 9 7 .
−1 −3 3 4
 
1 3 0 −1
We have R = rref(A) =  0 0 1 1 .
0 0 0 0

19 / 24
Example
Compute bases of the null space, row space and the column space
of the matrix  
1 3 3 2
A :=  2 6 9 7 .
−1 −3 3 4
 
1 3 0 −1
We have R = rref(A) =  0 0 1 1 . Therefore
0 0 0 0

{[1, 3, 0, −1], [0, 0, 1, 1]} is a basis for the row space of A.

19 / 24
Example
Compute bases of the null space, row space and the column space
of the matrix  
1 3 3 2
A :=  2 6 9 7 .
−1 −3 3 4
 
1 3 0 −1
We have R = rref(A) =  0 0 1 1 . Therefore
0 0 0 0

{[1, 3, 0, −1], [0, 0, 1, 1]} is a basis for the row space of A.

   
 1 3 
 2  ,  9  is a basis for the column space of A.
−1 3
 

19 / 24
Example
Compute bases of the null space, row space and the column space
of the matrix  
1 3 3 2
A :=  2 6 9 7 .
−1 −3 3 4
 
1 3 0 −1
We have R = rref(A) =  0 0 1 1 . Therefore
0 0 0 0

{[1, 3, 0, −1], [0, 0, 1, 1]} is a basis for the row space of A.

   
 1 3 
 2  ,  9  is a basis for the column space of A.
−1 3
 

Solve Rx = 0 to find a basis of null(R),

19 / 24
Example
Compute bases of the null space, row space and the column space
of the matrix  
1 3 3 2
A :=  2 6 9 7 .
−1 −3 3 4
 
1 3 0 −1
We have R = rref(A) =  0 0 1 1 . Therefore
0 0 0 0

{[1, 3, 0, −1], [0, 0, 1, 1]} is a basis for the row space of A.

   
 1 3 
 2  ,  9  is a basis for the column space of A.
−1 3
 

Solve Rx = 0 to find a basis of null(R), or use the previous

algorithm.
19 / 24
Example (cont.)
Interchanging 2nd and 3rd columns of R, we have
 
1 0 3 −1  
I2 F
RP =  0 1 0 1  = .
0 0
0 0 0 0
Now define  
  −3 1
−F  0 −1 
Y := = ,
In−p  1 0 
0 1
where p = 2 and n = 4.

20 / 24
Example (cont.)
Interchanging 2nd and 3rd columns of R, we have
 
1 0 3 −1  
I2 F
RP =  0 1 0 1  = .
0 0
0 0 0 0
Now define  
  −3 1
−F  0 −1 
Y := = ,
In−p  1 0 
0 1
where p = 2 and n = 4.
Finally, interchange 2nd and 3rd row of Y to obtain X , that is,
 
−3 1
 1 0 
X = PY =   0 −1  ,


0 1
which gives a basis of the null space of A.
20 / 24
Rank of a matrix

Theorem: The row space and the column space of a matrix A have
the same dimension, and dim(row(A)) = dim(col(A)) = rank(A).

21 / 24
Rank of a matrix

Theorem: The row space and the column space of a matrix A have
the same dimension, and dim(row(A)) = dim(col(A)) = rank(A).
Proof: Let R := rref(A). Then dim(row(A)) = dim(row(R)) =
number of nonzero rows of R = rank(A).

21 / 24
Rank of a matrix

Theorem: The row space and the column space of a matrix A have
the same dimension, and dim(row(A)) = dim(col(A)) = rank(A).
Proof: Let R := rref(A). Then dim(row(A)) = dim(row(R)) =
number of nonzero rows of R = rank(A).
Also A = ER for some m × m invertible matrix E . Hence
dim(col(A)) = dim(col(R)) = number of pivot columns of R
= rank(A). 

21 / 24
Rank of a matrix

Theorem: The row space and the column space of a matrix A have
the same dimension, and dim(row(A)) = dim(col(A)) = rank(A).
Proof: Let R := rref(A). Then dim(row(A)) = dim(row(R)) =
number of nonzero rows of R = rank(A).
Also A = ER for some m × m invertible matrix E . Hence
dim(col(A)) = dim(col(R)) = number of pivot columns of R
= rank(A). 

Theorem: For any matrix A, we have rank(A> ) = rank(A).

21 / 24
Rank of a matrix

Theorem: The row space and the column space of a matrix A have
the same dimension, and dim(row(A)) = dim(col(A)) = rank(A).
Proof: Let R := rref(A). Then dim(row(A)) = dim(row(R)) =
number of nonzero rows of R = rank(A).
Also A = ER for some m × m invertible matrix E . Hence
dim(col(A)) = dim(col(R)) = number of pivot columns of R
= rank(A). 

Theorem: For any matrix A, we have rank(A> ) = rank(A).

Definition: The nullity of a matrix A is the dimension of its null

space and is denoted by nullity(A).

21 / 24
Rank-nullity theorem

22 / 24
Rank-nullity theorem

Theorem: (Rank-nullity theorem) Let A be an m × n matrix. Then

rank(A) + nullity(A) = n.

22 / 24
Rank-nullity theorem

Theorem: (Rank-nullity theorem) Let A be an m × n matrix. Then

rank(A) + nullity(A) = n.

Proof: Suppose that rank(A) = r . Claim: nullity(A) = n − r .

22 / 24
Rank-nullity theorem

Theorem: (Rank-nullity theorem) Let A be an m × n matrix. Then

rank(A) + nullity(A) = n.

Proof: Suppose that rank(A) = r . Claim: nullity(A) = n − r .

Let R := rref(A). Then R has r nonzero rows. Equivalently, A has

r pivot columns and n − r non-pivot columns.

22 / 24
Rank-nullity theorem

Theorem: (Rank-nullity theorem) Let A be an m × n matrix. Then

rank(A) + nullity(A) = n.

Proof: Suppose that rank(A) = r . Claim: nullity(A) = n − r .

Let R := rref(A). Then R has r nonzero rows. Equivalently, A has

r pivot columns and n − r non-pivot columns.

Hence there are n − r free variables in the solution to the system

Ax = 0.

22 / 24
Rank-nullity theorem

Theorem: (Rank-nullity theorem) Let A be an m × n matrix. Then

rank(A) + nullity(A) = n.

Proof: Suppose that rank(A) = r . Claim: nullity(A) = n − r .

Let R := rref(A). Then R has r nonzero rows. Equivalently, A has

r pivot columns and n − r non-pivot columns.

Hence there are n − r free variables in the solution to the system

Ax = 0. Thus nullity(A) = n − r . (WHY?)

22 / 24
Rank-nullity theorem

Theorem: (Rank-nullity theorem) Let A be an m × n matrix. Then

rank(A) + nullity(A) = n.

Proof: Suppose that rank(A) = r . Claim: nullity(A) = n − r .

Let R := rref(A). Then R has r nonzero rows. Equivalently, A has

r pivot columns and n − r non-pivot columns.

Hence there are n − r free variables in the solution to the system

Ax = 0. Thus nullity(A) = n − r . (WHY?)

If x is a solution with n − r free parameters, then setting all but

one parameter to zero at a time results in n − r linearly
independent solutions. 

22 / 24
The fundamental theorem of invertible matrices

Theorem: Let A be an n × n matrix. Then the following

statements are equivalent.
1. A is invertible.

2. Ax = b has a unique solution for every b in Rn .

3. Ax = 0 has only the trivial solution.

4. The reduced row echelen form of A is In .

5. A is a product of elementary matrices.

23 / 24
The fundamental theorem of invertible matrices

Theorem: Let A be an n × n matrix. Then the following

statements are equivalent.
1. A is invertible.

2. Ax = b has a unique solution for every b in Rn .

3. Ax = 0 has only the trivial solution.

4. The reduced row echelen form of A is In .

5. A is a product of elementary matrices.

6. rank(A) = n.

23 / 24
The fundamental theorem of invertible matrices

7. nullity(A) = 0.

24 / 24
The fundamental theorem of invertible matrices

7. nullity(A) = 0.

8. The column vectors of A are linearly independent.

9. The column vectors of A span Rn .

10. The column vectors of A form a basis for Rn .

24 / 24
The fundamental theorem of invertible matrices

7. nullity(A) = 0.

8. The column vectors of A are linearly independent.

9. The column vectors of A span Rn .

10. The column vectors of A form a basis for Rn .

11. The row vectors of A are linearly independent.

12. The row vectors of A span Rn .

13. The row vectors of A form a basis for Rn .

***

24 / 24