T1A/T1B, hmcheungae Math 2121 Tutorial 4 October 4, 2018

It is my experience that proofs involving matrices can be shortened by 50% if one throws
the matrices out.——————————————————————- Emil Artin

Matrix Operations:

• Denote Mm×n (R) the set of all m × n matrices with real number entries.
• (Sum of two matrix ). If A = [aij ] ∈ Mm×n (R) and B = [bij ] ∈ Mm×n (R), then

A + B = [aij + bij ].

• (Scalar multiplication). If A = [aij ] ∈ Mm×n (R) and k ∈ R, then kA = [kaij ].

• If A ∈ Mk×m (R) is the standard matricx of T and B ∈ Mm×n (R) is the standard matrix of U , then
we define the matrix product AB = C, where (T ◦ U )(v) = Cv for v ∈ Rn , where U : Rn → Rm
and T : Rm → Rk are linear transformations.

Write B = [b1 b2 · · · bn ], then the product AB = [Ab1 Ab2 · · · Abn ].

Entrywisely, the (i, j)-entry of AB is given by

m
X
cij = aik bkj = ai1 b1j + ai2 b2j + · · · + aim bmj .
k=1

It is the dot product between row i of A and column j of B.

   
u1 v1
 u2   v2 
• (Dot product). Given vectors u =  .  and v =  .  in Rn , define
   
 ..   .. 
un vn

(u, v) = uT v = u1 v1 + u2 v2 + · · · + un vn = kukkvk cos θ,

p p
where θ is the angle between u and v, kuk = u21 + u22 + · · · + u2n and kvk = v12 + v22 + · · · + vn2 .

Property of Inverse:
• Let A be an n × n matrix and define T : Rn → Rn by T (x) = Ax. The matrix A is invertible if the
function T is invertible. Its inverse is the unique matrix A−1 such that T −1 (x) = A−1 x.
– To show an n × n matrice A is invertible, suffices to find an n × n matrix B such that

AB = BA = In .

• If A is invertible, then the equation Ax = b has a unique solution x = A−1 b.

• Suppose A, B, and A1 , A2 , · · · , Ak are invertible,
– (AB)−1 = B −1 A−1
– (A1 A2 · · · Ak )−1 = A−1 −1 −1
k · · · A2 A1 .
– (Ak )−1 = (A−1 )k .
1 −1
– k ∈ R − {0} implies (kA)−1 = A .
k
T1A/T1B, hmcheungae Math 2121 Tutorial 4 Page 2 of 6

– (A−1 )−1 = A
– (AT )−1 = (A−1 )T .

Property of Transpose:
• Suppose L : Rn → Rm is a linear transformation. There exists a unique linear transformation
LT : Rm → Rn , called the transpose of L, such that (L(u), v) = (u, LT (v)) for all u ∈ Rn and
v ∈ Rm .

The standard matrix of LT is the transpose of the standard matrix of L.

• The transpose of A = [aij ] ∈ Mm×n is AT = [aji ] ∈ Mn×m .
• (A + B)T = AT + B T .
• (AB)T = B T AT .
• (AT )T = A.
• (kA)T = kAT , where k ∈ R.
• If AT = A, then A is said to be a symmetric matrix. e.g.,
 
1 7 3
A = 7 4 −5 .
3 −5 6

Algorithm to find inverse:

Suppose A is an n × n matrix.
• A is invertible if and only if A is row equivalent to I.
• If [A | I] is transformed to [I | B] by applying a sequence of elementary row operations, then A−1 = B.

The Invertible Matrix Theorem:

Suppose A is an n × n matrix, then the following statements are equivalent:
• A is invertible.
• rref(A) = I.
• A has n pivot positions.
• The columns of A forms a linearly independent set.
• The columns of A span Rn .
• The equation Ax = 0 has only the trivial solution.
• For any b ∈ Rn , the equation Ax = b has at least one solution.
• For any b ∈ Rn , the equation Ax = b has a unique solution.
• For any b ∈ Rm , the equation Ax = b has at most one solution.
• The linear transformation x 7→ Ax is one-to-one.
• The linear transformation x 7→ Ax is onto.
• There is an n × n matrix C such that CA = In .
• There is an n × n matrix D such that AD = In .
• The transpose matrix AT is invertible.

Cont.
T1A/T1B, hmcheungae Math 2121 Tutorial 4 Page 3 of 6

Example 1
Given A ∈ Mm×n (R) and B ∈ Mn×p (R), prove that

(AB)T = B T AT .

Answer: It is easy to see that both (AB)T and B T AT have the same size (p × m).
It suffices to show the corresponding elements of both matrices are equal.
Denote Xij the (i, j)-th entry of X. we have

((AB)T )ij = (AB)ji

Xn
= Ajk Bki
k=1
Xn
= Bki Ajk
k=1
Xn
= (B T )ik (AT )kj
k=1
= (B AT )ij
T

The claim follows.

Example 2
Prove that if A and B are n × n symmetric matrices, then

AB is symmetric if and only if A and B commute.

Note that C is symmetric means C T = C, while C and D commute means CD = DC.

Answer:
If AB is symmetric, then AB = (AB)T = B T AT = BA, which means that A and B commute.
Conversely, if A and B commutes, then AB = BA = B T AT = (AB)T , so that AB is symmetric.

Example 3
Let A be an n × n matrix. Prove that
(a) AAT is a symmetric matrix, and
(b) If A is a square matrix, then A + AT is symmetric and A − AT is skew-symmetric.
(c) Show that each square matrix A can be decomposed into the sum of a symmetric matrix and a skew-
symmetric matrix.
Remark:
If AT = A, then A is said to be a symmetric matrix.
If AT = −A, then A is said to be a skew-symmetric matrix.
Answer:

(a) AAT is a symmetric matrix:

(AAT ) = (AT )T AT = AAT .

Cont.
T1A/T1B, hmcheungae Math 2121 Tutorial 4 Page 4 of 6

(b) A + AT is symmetric:
(A + AT )T = AT + (AT )T = AT + A = A + AT .
A − AT is skew-symmetric:
(A − AT )T = AT − (AT )T = AT − A = −(A − AT )

(c) A can be decomposed into the sum of a symmetric matrix and a skew-symmetric matrix:
1
1
A + AT + A − AT .


A=
2 2

Example 4
A square matrix A is said to be nilpotent if Ak = 0 for some positive integer k.
Let N be an n × n nilpotent matrix.
(a) Show that I − N is invertible.
(b) Hence show that every n × n matrix can be written in the form B − N B for some n × n matrix B.
Answer:
Idea:
In order to prove I − N to be invertible, one need to find (I − N )(something) = I = (something)(I − N ).
One only have two matrices I and N in hand. And one only know N k = 0. It is a good indication that
(something) should involve combinations of I and N .
Meanwhile, the identity 1 − xn = (1 − x)(1 + x + x2 + · · · + xn−1 ) you have learnt in high school can be
generalized to the matrix version:
I − X n = (I − X) 1 + X + X 2 + · · · + X n−1 .

(a) Suppose N k = 0 for some k ∈ N, we get

I = I − Nk
= (I − N )(I + N + N 2 + · · · + N k−1 )
= (I + N + N 2 + · · · + N k−1 )(I − N )
Hence N is invertible.
(b) For any given n × n matrix C, there is a n × n matrix B = (I + N + N 2 + · · · + N k−1 )C such that
C = (I − N )(I + N + N 2 + · · · + N k−1 )C = (I − N )B = B − N B.
 
0 a b
Remark: Let us have some example of nilpotent matrix: N = 0 0 c .
0 0 0
One can check
    
0 a b 0 a b 0 0 ac
N 2 = 0 0 c  0 0 c  = 0 0 0  .
0 0 0 0 0 0 0 0 0
And
    
0 0 ac 0 a b 0 0 0
N 3 = N 2 N = 0 0 0  0 0 c  = 0 0 0 .
0 0 0 0 0 0 0 0 0
 
1 −a −b
Then I − N = 0 1 −c is an element in the Heisenberg group.
0 0 1

Cont.
T1A/T1B, hmcheungae Math 2121 Tutorial 4 Page 5 of 6

Example 5
Apply matrix inverse to solve


x+z−w =1

x + y + 2z = 2
.


x + 2y + 2z + 2w = −1
2y + z + 4w = −2

     
1 x 1 0 1 −1
2 y  1 1 2 0 
−1, where x =  z  and A = 1 2 2 2 .
Answer: This is equivalent to Ax =      

−2 w 0 2 1 4
We first find A−1 .
   
1 0 1 −1 1 0 0 0 1 0 1 −1 1 0 0 0
 −  0 1 1 1 −1 1 0 0 
 1 1 2 0 0 1 0 0   
 1 2 2 2 0 0 1 0 →

 0 2 1 3 −1 0 1 0 
0 2 1 4 0 0 0 1 0 2 1 4 0 0 0 1
 
1 0 1 −1 1 0 0 0
 0 1 1 1 −1 1 0 0 
→
−  0 0 −1 1

1 −2 1 0 
0 0 −1 2 2 −2 0 1
 
1 0 1 −1 1 0 0 0
 0 1 1 1 −1 1 0 0 
→
−  0 0 −1 1

1 −2 1 0 
0 0 0 1 1 0 −1 1
 
1 0 1 0 2 0 −1 1
 0 1 1 0 −2 1 1 −1 
→
−  0 0 −1 0 0 −2 2 −1 


0 0 0 1 1 0 −1 1
 
1 0 0 0 2 −2 1 0
 0 1 0 0 −2 −1 3 −2 
→
−  0 0 −1 0 0 −2 2 −1 


0 0 0 1 1 0 −1 1
 
1 0 0 0 2 −2 1 0
 0 1 0 0 −2 −1 3 −2 
→
−  
 0 0 1 0 0 2 −2 1 
0 0 0 1 1 0 −1 1

 
2 −2 1 0
−2 −1 3 −2
Hence A−1 =  .
0 2 −2 1 
1 0 −1 1
This implies
        
x 1 2 −2 1 0 1 −3
y  2
  = x = A−1   = 
  −2 −1 3 −2  2  −3
  =  .
z  −1  0 2 −2 1  −1  4 
w −2 1 0 −1 1 −2 0
Therefore, the solution is
(x, y, z, w) = (−3, −3, 4, 0).

Cont.
T1A/T1B, hmcheungae Math 2121 Tutorial 4 Page 6 of 6

Example 6
Let A be an n × n matrix with the property that the homogeneous system Ax = 0 has only the trivial
solution. Then A is invertible (nonsingular).
Equivalently, if A is singular (noninvertible), then the homogeneous system Ax = 0 has a nontrivial
solution.
Answer:
Note that Ax = 0 has only the trivial solution is equivalent to LA is one-to-one, where LA (x) = Ax.
Also, Ax = 0 has only the trivial solution is equivalent to every column of A has a pivot position.
Since A is an n × n matrix, this implies every row of A has a pivot position.
This is equivalent to LA is onto. Therefore, LA is invertible, and hence A is invertible.
By taking contrapositive, we get the last if-then statement.

Example 7
Let B be an n × n skew-symmetric matrix. Then A = In − B is (non-singular) invertible.
Note that C is skew-symmetric iff C T = −C.
[Hint: Use the fact A is invertible iff Ax = 0 has a unique trivial solution;
And consider xT x, x = 0 ⇐⇒ xT x = 0.]
Answer:
Let A = In − B, where B T = −B.
It suffices to show Ax = 0 =⇒ x = 0.
Then (In − B)x = 0 and so x = Bx.
This implies xT x = xT Bx.
Taking transpose of both sides, we get

(xT x)T = (xT Bx)T

xT x = (x)T B T ((x)T )T
xT x = xT (−B)x = −xT x


x1
 x2 
This implies xT x = 0. Let x =  .  and then x21 + x22 + · · · + x2n = 0.
 
 .. 
xn
We get x1 = x2 = · · · = xn = 0, i.e., x = 0.

The End.