Lecture Notes: Determinant of a Square Matrix

Yufei Tao
Department of Computer Science and Engineering
Chinese University of Hong Kong
taoyf@cse.cuhk.edu.hk

1 Determinant Definition
Let A = [aij ] be an n × n matrix (i.e., A is a square matrix). Given a pair of (i, j), we define M ij
to be the (n − 1) × (n − 1) matrix obtained by removing the i-th row and j-th column of A. For
example, suppose that
 
1 2 1
A =  3 0 -2 
-1 -1 2

Then:
     
2 1 1 1 1 1
M 21 = , M 22 = , M 32 =
-1 2 -1 2 3 -2

We are now ready to define determinants:

Definition 1. Let A = [aij ] be an n × n matrix. If n = 1, its determinant, denoted as det(A),
equals a11 . If n > 1, we first choose an arbitrary i∗ ∈ [1, n], and then define the determinant of A
recursively as:
n
X ∗
det(A) = (−1)i +j · ai∗ j · det(M i∗ j ). (1)
j=1

Besides det(A), we may also denote the determinant of A as |A|. Henceforth, if we apply (1)
to compute det(A), we say that we expand A by row i∗ . It is important to note that the value of
det(A) does not depend on the choice of i∗ . We omit a proof of this fact, but illustrate it in the
following examples.

Example 1 (Second-Order Determinants). In general, if A = [aij ] is a 2 × 2 matrix, then

det (A) = a11 a22 − a12 a21 .

For instance:
 
 2 1 

 -1 = 2 × 2 − 1 × (−1) = 5.
2 

We may verify the above by definition as follows. Choosing i∗ = 1, we get:

 
 2 1  1+1
 -1 2  = (−1)
  · 2 · det(M 11 ) + (−1)1+2 · 1 · det(M 12 )

= 2 × 2 + (−1) × (−1) = 5.

1
Alternatively, choosing i∗ = 2, we get:
 
 2 1  2+1
 -1 2  = (−1)
  · (−1) · det(M 21 ) + (−1)2+2 · 2 · det (M 22 )

= 1 × 1 + 2 × 2 = 5.

Example 2 (Third-Order Determinants). Suppose that

 
1 2 1
A =  3 0 -2 
-1 -1 2

Choosing i∗ = 1, we get:
     
 0 −2   3 −2   3 0 
det(A) = 1 
  − 2   + 1 
−1 2   −1 2   −1 −1 
= 1(0 − 2) − 2(6 − 2) + 1(−3 − 0) = −13.

Alternatively, choosing i∗ = 2, we get:

     
 2 1   1 1   1 2 
det(A) = −3    + 0
  − (−2)  
−1 2  −1 2  −1 −1 
= (−3)(4 + 1) + 0(2 + 1) + 2(−1 + 2) = −13.

2 Properties of Determinants

Expansion by a Column. Definition 1 allows us to compute the determinant of a matrix by row

expansion. We may also achieve the same purpose by column expansion.

Lemma 1. Let A = [aij ] be an n × n matrix with n > 1. Choose an arbitrary j ∗ ∈ [1, n]. The
determinant of A equals:
n
X ∗
det(A) = (−1)i+j · aij ∗ · det(M ij ∗ ).
i=1

The value of the above equation does not depend on the choice of j ∗ .

We omit a proof but illustrate the lemma with an example below. Henceforth, if we compute
det(A) by the above lemma, we say that we expand A by column j ∗ .

Example 3. Suppose that

 
1 2 1
A =  3 0 -2 
-1 -1 2

2
Choosing j ∗ = 1, we get:
     
 0 −2   2 1   2 1 
det(A) = 1 
  − 3   + (−1) 
−1 2   −1 2   0 −2 
= 1(0 − 2) − 3(4 + 1) − 1(−4 − 0) = −13.

Corollary 1. Let A be a square matrix. Then, det(A) = det(AT ).

Proof. Note that expanding A by column k is equivalent to expanding AT by row k.

Corollary 2. If A has a zero row or a zero column, then det(A) = 0.

Proof. If A has a zero row, then det(A) = 0 follows from expanding A by that row. The case
where A has a zero column is similar.

Determinant of a Row-Echelon Matrix. The next lemma shows that the determinant of a
matrix in row-echelon form is simply the product of the values on the main diagonal.

Lemma 2. Let A = [aij ] be an n × n matrix in row-echelon form. Then, det(A) = Πni=1 aii .

Proof. We can prove the lemma by induction. First, correctness is obvious for n = 1. Assuming
correctness for n ≤ t − 1 (for t ≥ 2), consider n = t. Expanding A by the first row gives:
n
X
det(A) = (−1)1+j · a1j · det(M 1j ). (2)
j=1

From induction we know that det(M 11 ) = Πni=2 aii . Furthermore, for j > 1, det(M 1j ) = 0 because
the first column of M 1j contains all 0’s. It thus follows that (2) equals Πni=1 aii .

Determinants under Elementary Row Operations. Let A = [aij ] be an n × n matrix. Define

the elementary row operations on A:

1. Switch two rows of A.

2. Multiply all numbers of a row of A by the same non-zero value c.

3. Let r i and r j be two row vectors of A. Update row r i to r i + c · r j , where c can be any real
value.

Next, we refer to the above as Operation 1, 2, and 3, respectively.

Lemma 3. The determinant of A

1. should be multiplied by −1 after Operation 1;

2. should be multiplied by c after Operation 2;

3
 3. has no change after Operation 3.

Again, we omit a proof, but illustrate the lemma with an example.

Example 4.
       

 1 2 1   1
  2 1   1
  2 1   1
  2 1 


 3 0 -2 = 3
  0 -2 = 0
  -6 -5 = 0
  -6 -5  = −13.

 -1 -1 2   0 1 3   0 1 3   0 0 13/6 

Here is another derivation giving the same result:

       
 1 2 1   1 2 1   1 2 1   1 2 1 
      
 3 0 -2  = −  -1 -1 2  = −  0
 1 3  = − 0 1 3  = −13.
     
 -1 -1 2   3 0 -2   0 -6 -5   0 0 13 

Corollary 3. If A has two identical rows or columns, then det(A) = 0.

Proof. We prove only the row case. Switching the two rows gets back the same matrix. However, by
Lemma 3, the determinant of the matrix should be multiplied by −1. Therefore, we get det(A) =
−det(A), meaning det(A) = 0.

Determinant under Matrix Multiplication. We state the following result without proof:

Lemma 4. Let A, B be n × n matrices. It holds that det(AB) = det(A) · det(B).

Example 5.
 
 1 2 1 
 
 3 0 -2  = −13
 
 -1 -1 2 
 
 -2 1 0 
 
 0 1 1  = −3
 
 1 -2 0 
    
 1 2 1 -2 1 0   -1 1 2 
  
 3 0 -2   0 1 1  =  -8
 7 0  = 39.
 
 -1 -1 2 1 -2 0   4 -6 -1 

Relationships with Ranks. We close the lecture notes by giving two lemmas that relate deter-
minants to ranks:

Lemma 5. Let A be an n × n matrix. A has rank n if and only if det(A) 6= 0.

4
Proof. We can first apply elementary row operations to convert A into row-echelon form A∗ . Thus,
A has rank n if and only if A∗ has rank n. Since A∗ is a square matrix, that it has rank n is
equivalent to saying that all the numbers on its main diagonal are non-zero. Thus, by Lemma 2,
we know that A∗ has rank n if and only if det(A∗ ) 6= 0. Finally, by Lemma 3, det(A) 6= 0 if and
only if det(A∗ ) 6= 0. We thus complete the proof.

Let A be an m × n matrix. We say that B is a submatrix of A if we can obtain B is by

discarding some rows and columns of A. For instance, suppose that
 
1 2 3 4
A =  5 6 7 8  (3)
3 2 1 0

Then all the following are submatrices of A:

 
1 2 3  
 5 6 6 7
7 ,
 , [0].
2 1
3 2 1

We state the next lemma without proof:

Lemma 6. Let A be an m × n matrix. A has rank r if and only if both the following hold
simultaneously:

• A has an r × r submatrix B such that det (B) 6= 0.

• all (r + 1) × (r + 1) submatrices of A have determinant 0.

Example 6. Let A be as shown in (3). It can be verified that rank A = 2. Indeed, you can easily
find a 2 × 2 submatrix of A with a non-zero determinant. On the other hand, any 3 × 3 submatrix
of A always has determinant 0.