T1A/T1B, hmcheungae Math 2121 Tutorial 7 October 25, 2018

Mathematics is no more computation than typing is literature.—— John Allen Paulos

Determinant:
• Given a matrix X ∈ Rn and an arbitrary n × n matrix A, define

Π(X, A) = the product of the entries of A in the nonzero positions of X.

For example,
   
Y 0 0 1 a b c
1 0 0 , d e f  = cdh.
0 1 0 g h i

Suppose A is an n × n matrix. Then

X
det A = Π(X, A)(−1)inv(X)
X∈Sn
P
where the notation X∈Sn means “compute Π(X, A)(−1)inv(X) for each n × n permutation matrix
X and then take the sum of all of the resulting numbers.”

This general formula simplifies to the following expressions for n = 1, 2, 3 (which you should remember):
– det[a] = a.
 
a b
– det = ad − bc.
c d
 
a b c
– det d e f  = a(ei − f h) − b(di − f g) + c(dh − ef ).
g h i
Summary of properties of determinant.
1. The determinant is defined only for square matrices!
2. Determinant is linear in each row (column) when the other rows (columns) are fixed.
3. If one interchanges two rows (columns) of a matrix A, the determinant changes sign.
4. For a triangular (in particular, for a diagonal) matrix its determinant is the product of the diagonal
entries. In particular, det I = 1.
5. If a matrix A has a zero row (or column), det A = 0.
6. If a matrix A has two equal rows (columns), det A = 0.
7. If one of the rows (columns) of A is a linear combination of the other rows (columns), i.e. if the matrix
is not invertible, then det A = 0;
More generally,
8. det A = 0 if and only if A is not invertible, or equivalently
9. det A 6= 0 if and only if A is invertible.
10. det A does not change if we add to a row (column) a linear combination of the other rows
(columns). In particular, the determinant is preserved under the row (column) replacement.
T1A/T1B, hmcheungae Math 2121 Tutorial 7 Page 2 of 10

11. det AT = det A.

12. det(AB) = (det A)(det B).
13. If A is an n × n matrix, then det(aA) = an det A.
Remark:
• Since det A = det(AT ), the statements that we knew about columns are true for rows too.
• The property 13 follows from the linearity of the determinant, recall that to multiply a matrix A by a
we have to multiply each row by a, and that each multiplication multiplies the determinant by a.
• If A is a permutation matrix then det A = (−1)inv(A) .
• If X ∈ Sn , then X T ∈ Sn and inv(X) = inv(X T ). This implies property 11.
• If A is an n × n matrix, then det A is a nonzero multiple of det(RREF (A)).

Evaluate det A by cofactor expansion along a row or column

Given an n × n matrix A, define A(i,j) as the (n − 1) × (n − 1) submatrix formed by removing row i and
column j from A.
If A is the n × n matrix
 
a11 a12 a13 ··· a1n
 a21 a22 a23
 ··· a2n 

A =  a31 a32 a33
 ··· a2n 

 .. .. .. .. .. 
 . . . . . 
an1 an2 an3 ··· ann
then
det A = a11 det A(1,1) − a12 det A(1,2) + a13 det A(1,3) − · · · − (−1)n a1n det A(1,n) , (along first row.)
and also
det A = a11 det A(1,1) − a21 det A(2,1) + a31 det A(3,1) − · · · − (−1)n an1 det A(n,1) , (along first column.)
The geometrical interpretation of determinants.
• If u, v ∈ R2 are two vectors and A = [u v], then the area of the parallelogram with sides u and v is
| det A|.
• Suppose T : R2 → R2 is a linear transformation with standard matrix A.
If S is a parallelogram in the R2 plane, then the area of T (S) is the area of S times | det A|.
Finding Matrix Inverse by using determinants
• The adjoint of an n × n matrix A is adj A = [Cij ]T , where Cij = (−1)i+j det A(i,j) ,
which is called (i, j)-cofactor of A.
• For any n × n matrix A, A(adj A) = (adj A)A = (det A)I.
1
• If det A 6= 0, then A−1 = adj A.
det A
Cramer’s Rule (Optional)
Consider the equation Ax = b, where A is an n × n invertible matrix.
• Let Ai be the matrix obtained from A by replacing column i of A by b.
det Ai
• Each unknown in the equation Ax = b is unique determined by xi = .
det A

Cont.
T1A/T1B, hmcheungae Math 2121 Tutorial 7 Page 3 of 10

Example 1
(True/False about determinant.)
1. If A is an n × n matrix with nonzero determinant and AB = AC, then B = C.
2. A square matrix with zero diagonal entries is never invertible.
3. Row operations do not affect determinant.

4. If column of a n × n matrix A forms a basis of Rn , then det(A) = 0.

5. For any lower triangular matrix A, det A is the sum of the entries on its diagonal.
6. If det A = 0, then A = 0n×n .

7. For any n × n matrix A, det(−A) = − det A.

8. If A is 3 × 3 matrix with det A = 0, then one column of A must be multiple of the other.
9. If AT A = AAT = I, then det A = ±1.
10. Let A ∈ M2×3 (R) and B ∈ M3×2 (R). Then det(BA) = 0.

11. If AT = −A and n is odd, then det A = 0.

12. If AT = −A and n is even, then det A = 0.
13. If Ak = 0 for some positive integer k, then det A = 0.

14. If A and B are similar, than det A = det B.

Answer:

1. True: det A 6= 0 is equivalent to A is invertible. Then

AB = AC =⇒ B = A−1 (AB) = A−1 (AC) = C.

 
0 1
2. False: Let A = which is of zero diagonal entries, we get det A = −1, which is the same as A
1 0
is invertible.

3. False: Theorem 0.3 c, Chapter 5 Section 2,

A row replacement operation on A does not change the determinant.
A row interchange changes the sign of the determinant.
A row scaling also scales the determinant by the same scalar factor.

4. False:
Column of a n × n matrix A forms a basis of Rn
iff column of a n × n matrix A is linearly independent and span Rn
iff every column has a pivot position and every row has a pivot position
iff LA is one-to-one and onto
iff LA is invertible
iff A is invertible
iff det A 6= 0.

Cont.
T1A/T1B, hmcheungae Math 2121 Tutorial 7 Page 4 of 10

 
1 0
5. False: Let A = , we get det A = 1 · 2 = 2 6= 3 = 1 + 2
a 2
The correct statement should be “For any lower triangular matrix A,
det A is the product of the entries on its diagonal.”
6. False: Let Eij be the matrix with 1 in (i, j)-entry,
all other entries is zero.
The det En1 = 0 but En1 6= 0n×n .
7. False: Note that det(cA) = cn det(A).
In our case, det(−A) = (−1)n det A
Consider a matrix A with det A 6= 0, and then if det(−A) = − det A
implies (−1)n det A = − det A
implies (−1)n = −1
implies n is odd, which may not be true.
 
0 0 0
8. False: Let A = 1 2 7, then det A = 0 but No pair of columns is multiple of each other.
1 3 4
9. True:
Since det A = det AT , we get
AAT = I =⇒ det(AAT ) = det I = 1 =⇒ det A det AT = 1 =⇒ (det A)2 = 1 =⇒ det A = ±1.

10. True:
Note that x 7→ BAx is onto implies x 7→ Bx is onto.
Taking contrapositive, we get x 7→ Bx is not onto implies x 7→ BAx is not onto.
By example 2 in Tut 5/ example 6 in Tut 6,
there is No onto maps from R2 to R3 . Hence x 7→ Bx is not onto.
In particular, x 7→ BAx is not onto and thus not invertible.
By Invertible Matrix Theorem, we get det(BA) = 0.
11. True:
AT = −A =⇒ det(AT ) = det(−A) = (−1)n det A =⇒ det A = (−1)n det A. If n is odd, we get
det A = −1 det A =⇒ 2 det A = 0 =⇒ det A = 0.
12. False:  
0 −1
If n is even, it may not be true in general. For example, det = 1 6= 0.
1 0
13. True:
Ak = On×n =⇒ det(Ak ) = det On×n =⇒ (det A)k = 0 =⇒ det A = 0.
14. True:
A = P BP −1 for some invertible P . This implies
det(A) = det(P BP −1 )
= det(P ) det(B) det(P −1 )
= det(P ) det(P −1 ) det(B)
= det(P P −1 ) det(B)
= det(I) det(B)
= det(B)

Cont.
T1A/T1B, hmcheungae Math 2121 Tutorial 7 Page 5 of 10

Example 2
Show that
(a)
 
a + b + 2c a b
det  c b + c + 2a b  = 2(a + b + c)3 .
c a c + a + 2b

(b)

a2 c2 + ca
 
bc
2
det a + ab
 b2 ca  = 4a2 b2 c2 .
ab b2 + bc c2

(c)

x2
 
1 x
det 1 y y 2  = (z − x)(x − y)(y − z).
1 z z2

This is a particular case of the so-called Vandermonde determinant.

Answer:

(a)
   
a + b + 2c a b a+b+c 0 −a − b − c
det  c b + c + 2a b  = det  0 a + b + c −a − b − c
c a c + a + 2b c a c + a + 2b
(−R3 + R1 → R1 , −R3 + R2 → R2 )
 
a+b+c 0 0
= det  0 a+b+c 0 
c a 2(a + b + c)
(C3 + C1 → C3 , C2 + C3 → C3 )
= 2(a + b + c)3 .

(b)

a2 c2 + ca
 2
c2 + ca
  
bc a bc
2
det a + ab b2 ca  = det ab b2 − bc −c2  (−R1 + R2 → R2 )
2 2 2
ab b + bc c ab b + bc c2
 2
c2 + ca

a bc
= 2 det ab b2 − bc −c2  (−R2 + R3 → R3 )
0 bc c2
 2 
a 0 ca
= 2 det ab b2 0  (−R3 + R1 → R1 , R3 + R2 → R2 )
0 bc c2
= 2 a2 b2 c2 + (ca)(ab)(bc)

= 4a2 b2 c2 .

Cont.
T1A/T1B, hmcheungae Math 2121 Tutorial 7 Page 6 of 10

(c)
x2 x2
   
1 x 1 x
det 1 y y 2  = det 0 y−x y 2 − x2 
1 z z2 0 z−x z 2 − x2
x2
 
1 x
= det 0 y−x (y − x)(y + x)
0 z−x (z − x)(z + x)
x2
 
1 x
= (y − x)(z − x) det 0 1 y + x
0 1 z+x
x2
 
1 x
= (y − x)(z − x) det 0 1 y + x
0 0 z−y
= (y − x)(z − x)(z − y)
= (z − x)(x − y)(y − z)

Example 3
(a) Let v1 , v2 be vectors in R2 and let A be the 2 × 2 matrix with columns v1 , v2 . Prove that | det A| is
the area of the parallelogram with two sides given by the vectors v1 , v2 .
Consider first the case when v1 = (x1 , 0)T .
To treat general case v1 = (x1 , y1 )T left multiply A by a rotation matrix that transforms vector v1 into
(x˜1 , 0)T . [Hint: What is the determinant of a rotation matrix?]
(b) Let points A, B and C in the plane R2 have coordinates (x1 , y1 ), (x2 , y2 ) and (x3 , y3 ) respectively.
Show that the area of triangle ABC is the absolute value of
 
1 x1 y1
1
det 1 x2 y2  .
2
1 x3 y3
[ Hint: use row operations and geometric interpretation of 2 × 2 determinants (area). ]

Answer:
   
x1 x
(a) Consider v1 = , v2 = 2 and let A = [v1 v2 ], Rθ be the rotation anticlockwise by some angle
y1 y2
   
x˜1 x˜
θ such that Rθ v1 = and Rθ v2 = 2 .
0 y˜2
 
cos θ − sin θ
Note det(Rθ ) = det = cos2 θ + sin2 θ = 1. Then
sin θ − sin θ
  
x˜ x˜2 
| det A| = | det Rθ det A| = | det(Rθ A)| = | det[Rθ v1 Rθ v2 ]| = det 1

= |x˜1 ||y˜2 |
0 y˜2 

Note that rotation does not change the area. Thus the area of parallelogram spanned by Rθ v1 and
Rθ v2 is the same as the area of parallelogram spanned by v1 and v2 .
   
x˜ x˜
Now it suffices to show |x˜1 ||y˜2 | = the area spanned by Rθ v1 = 1 and Rθ v2 = 2 .
0 y˜2
There are four cases:

Cont.
T1A/T1B, hmcheungae Math 2121 Tutorial 7 Page 7 of 10

(a) x˜1 > 0, y˜1 > 0.

(b) x˜1 > 0, y˜1 < 0.
(c) x˜1 < 0, y˜1 > 0.
(d) x˜1 < 0, y˜1 < 0.

Case ii is reduced to case i by reflecting along x-axis.

Case iii is reduced to case i by reflecting along y-axis.
Case iv is reduced to case i by reflecting along x-axis and reflecting along y-axis.
Case i is clearly true since |x˜1 | is the base of the triangle and |y˜2 | is the height of the triangle.

Thus we get

| det A| = the area of parallelogram spanned by v1 and v2 .

(b)
     

1 1 x1 y1   1 x1 y1 
1  
 det 1
2 x2 y2  = det 0 x2 − x1 y2 − y1 
    
2
 1 x3 y3   0 x3 − x1 y3 − y1 
  
1 x − x1 y2 − y1 
= det 2
2 x3 − x1 y3 − y1 
  
1 x − x1 x3 − x1 
= det 2
2 y2 − y1 y3 − y1 
   
1 x2 − x1 x − x1
= area of parallelogram spanned by , 3
2 y2 − y1 y3 − y1
   
x2 − x1 x − x1
= area of triangle spanned by , 3
y2 − y1 y3 − y1
~
= area of triangle spanned by AB, AC~
= area of triangle ∆ABC

Example 4
 
7 2 1
Use adjugate method to find the inverse of A =  0 3 −1.
−3 4 −2
Answer:

       T
3 −1 0 −1 0 3
det − det det
 4 −2   −3 −2 −3 4 
  T  
  −2 3 9 −2 8 −5
 2 1 7 1 7 2 
adj(A) = − det det − det = 8 −11 −34 =  3 −11 7  .

 4 −2 −3 −2  −3 4 

   −5 7 21 9 −34 21
 2 1 7 1 7 2 
det − det det
3 −1 −3 −2 −3 −2

Note that
 
7 2 1
det A = det  0 3 −1 = (7)(3)(−2) + (2)(−1)(−3) + (1)(0)(4) − (−3)(3)(1) − (7)(4)(−1) − (0)(2)(−2)
−3 4 −2

Cont.
T1A/T1B, hmcheungae Math 2121 Tutorial 7 Page 8 of 10

= −42 + 6 + 9 + 28 = 1.
Hence
   
−2 8 −5 −2 8 −5
1 1
A−1 = adj(A) =  3 −11 7 = 3 −11 7  .
det A 1
9 −34 21 9 −34 21

Example 5
(a) Let A be an n × n matrix and Ax = b and det A 6= 0.
Prove that the linear system has the unique solution
det A1 det A2 det An
x1 = , x2 = , · · · , xn = ,
det A det A det A
where Ai is the matrix obtained from A replacing the i-th column of A by b.

[Hint: det(CD) = det C det D.]

(b) Use Cramers rule to solve the system

x1 − 2x2 + x3 = 3

2x1 + x2 − x3 = 5 .

3x1 − x2 + 2x3 = 12


Answer:

(a) Since det(A) 6= 0, by properties of the determinant, we know that A is invertible.

Claim: Ax = b has a unique solution.
Note that A−1 b is a solution since A(A−1 b) = (AA−1 )b = b, so we know that a solution exists.

Let s be an arbitrary solution to the equation, so As = b.

But then

s = (A−1 A)s = A−1 (As) = A−1 b,

we see that A−1 b is the only solution.

For each integer i, 1 ≤ i ≤ n, let ai denote the i-th column of A.

Let ei denote the i-th column of the identity matrix In , let Xi denote the matrix obtained from In by
replacing column i with the column vector x.

Note that

AXi = A[e1 · · · ei−1 x ei+1 · · · en ] = [Ae1 · · · Aei−1 Ax Aei+1 · · · Aen ] = [a1 · · · ai−1 b ai+1 · · · an ] = Ai

Since det Xi = (−1)i+i xi det In−1 = xi , and then

det Ai = det A det Xi = xi det A.

det Ai
This implies xi = for each i = 1, · · · , n.
det A

Cont.
T1A/T1B, hmcheungae Math 2121 Tutorial 7 Page 9 of 10

 
1 −2 1
(b) ∆ = det 2 1 −1.
3 −1 2
Expanding along the top row,
     
1 −1 2 −1 2 1
∆ = 1× −(−2)× +1× = 1×(2−1)+2×(4+3)+1×(−2−3) = 1+14−5 = 10
−1 2 3 2 3 −1

Then
 
3 −2 1    
det  5 1 −1   5 −1 5 1
3 × 1 − 1 − 12 − (−2) × +1×
12 −1 2 12 2 12 −1
x1 = =
det ∆ 10
3 × 1 + 2 × 22 + 1 × (−17) 30
= = = 3.
10 10

 
1 3 1
det 2 5 −1  
3 12 2 1 3 1       
1  1 5 −1 2 −1 2 5
x2 = = 2 5 −1 = 1× −3× +1×
det ∆ 10 10 12 2 3 2 3 12
3 12 2

1
= (22 − 3 × 7 + 9) = 1.
10

 
1 −2 3
det 2 1 5
3 −1 12
      
1 1 5 2 5 2 1
x3 = = 1× − (−2) × +3×
det ∆ 10 −1 12 3 12 3 −1

1
= (17 + 2 × 9 + 3 × (−5)) = 2.
10

Midterm: True or False

Indicate which of the following is TRUE or FALSE.

1. If a system of linear equations has two different solutions, then it must have infinitely many solutions.
2. If A is an m × n matrix and the equation Ax = b is consistent for every b in Rm , then A has m pivot
columns.
3. A linear system with no free variables has a unique solution.

4. If a linear system Ax = b has more than one solution, then so does the linear system Ax = 0.
5. If u, v, w ∈ R2 are all nonzero, then w is a linear combination of u and v.
6. If A, B, and C are matrices with AB = AC, then B = C.

7. If A and B are m × n matrices, then both AB T and AT B are defined.

Cont.
T1A/T1B, hmcheungae Math 2121 Tutorial 7 Page 10 of 10

8. If A and B are n × n matrices with AB = BA, and if A is invertible, then A−1 B = BA−1
9. If two matrices are row equivalent, then they have the same column space.
10. If two matrices are row equivalent, then they have the same null space.
Answer:

1. True: If x1 , x2 are different solutions of Ax = 0. Then A (tx1 ) = 0 and A ((1 − t)x2 ) = 0. This
implies A (tx1 + (1 − t)x2 ) = 0 for any t ∈ R. Hence the claim follows.

2. True: If A is an m × n matrix and the equation Ax = b is consistent for every b in Rm , then every
row of A has a pivot position. Since there are m rows, A has m pivot columns.
 
1 0 0
3. False: Consider  0 1 0 .
0 0 1
4. True: If x1 , x2 are different solutions of Ax = b. Then 0, x1 − x2 are different solutions of Ax = 0.
   
1 0
5. False: Consider u = v = and w = .
0 1
       
1 0 0 0 0 0 0 0
6. False: Consider A = ,B= and C = . Then AB = = AC but B 6= C.
0 0 0 1 1 0 0 0

7. True: If A, B are m × n matrices, then B T is n × m matrix and hence AB T is a m × m matrix. Also

AT is a n × m matrix and hence AT B is a n × n matrix.
8. True: A is invertible, then A−1 (BA)A−1 = A−1 (AB)A−1 and hence A−1 B = BA−1 .
           
1 0 1 0 1 0 1 1 0 1
9. False: is row equivalent to . But Col =R and Col =R .
1 0 0 0 1 0 1 0 0 0
10. True: If A and B are two row equivalent matrices then they share the same null space. This fact,
which is in fact a little theorem, can be proved as follows:
Suppose x ∈Nul(A). Then Ax = 0.
Also since A is row equivalent to B so B = E1 E2 · · · Ek A where each Ei is an elementary matrix.
(Recall that an elementary matrix is the matrix obtained from I from performing any elementary row
operation.)
Now, Bx = E1 E2 · · · Ek Ax = 0 and so x ∈Nul(B) as well. So Nul(A) ⊆Nul(B). Similarly Nul(B) ⊆Nul(A).
It is now clear that Nul(A) =Nul(B).

The End.