DETERMINANTS

1. Cofactor expansion
The determinant of a square matrix A is a real number, denoted by det(A),
which has many interesting fundamental properties. In this section, we will first
explain how to compute this number.
Definition 1.1. When A is a 1 × 1 matrix A = (a), the determinant of A is defined
by:
det(A) = a.
( )
a b
When A is a 2 × 2 matrix A = , the determinant of A is defined by:
c d
det(A) = ad − bc.
Definition 1.2. When A is a 3 × 3 matrix:
 
a b c
A= d e f 
g h i
the determinant of A is defined by:
( ) ( ) ( )
e f b c b c
det(A) = a × det − d × det + g × det .
h i h i e f
Example 1.3.  
−2 1 2
A= 7 0 1 
1 2 −3
det(A) = −2 × (0 − 2) − 7 × (−3 − 4) + 1 × (1 − 0) = 4 + 49 + 1 = 54.
As we can see in Definition 1.2, the determinant of a 3 × 3 matrix A is defined
by using determinants of smaller 2 × 2 matrices which are ”contained inside” A.
This leads us to the following definition.
Definition 1.4. Let A be an n × n matrix and fix two numbers i and j such that
1 ≤ i ≤ n and 1 ≤ j ≤ n.
The submatrix Ai,j is the (n − 1) × (n − 1) matrix obtained from A by removing
row i and column j.
Example 1.5.  
−2 1 2
A= 7 0 1 
1 2 −3
There are 9 submatrices of A:
( ) ( ) ( )
0 1 7 1 −2 1
A1,1 = A1,2 = ... A3,3 = .
2 −3 1 −3 7 0
1
 We are now ready to define the determinant of an n × n matrix A for any
natural number n: we will do it by induction on n, using the determinants of the
submatrices Ai,j .
( )
a1,1 a1,2
Notice that when A is a 2 × 2 matrix A = , the submatrices A1,1
a2,1 a2,2
and A2,1 are:
A1,1 = (a2,2 ) and A2,1 = (a1,2 )
and therefore:
det(A) = a1,1 det(A1,1 ) − a2,1 det(A2,1 ) = a1,1 a2,2 − a2,1 a1,2 .
Similarly, when A is a 3 × 3 matrix with coefficients (ai,j ) with 1 ≤ i ≤ 3 and
1 ≤ j ≤ 3, the formula in Definition 1.2 is equivalent to:
det(A) = a1,1 det(A1,1 ) − a2,1 det(A2,1 ) + a3,1 det(A3,1 ).
and the definition of the determinant of an n × n matrix will be a generalization of
this formula.

Definition 1.6. Let A be an n × n matrix. The determinant of A is defined by

induction on n by the following formula, which is called the cofactor expansion
along the first column of A.
• When n = 1 and A = (a1,1 ),
det(A) = a1,1 .
• When n ≥ 2,
det(A) = a1,1 det(A1,1 ) − a2,1 det(A2,1 ) + ... + (−1)n+1 an,1 det(An,1 ).
Notice the terms ai,1 × det(Ai,1 ) are added up with alternate signs: the first
sign is +, then −, then +, then − ... etc... With this formula, it is clear that we
can compute the determinant of an n × n matrix for any integer n, by reducing it
to smaller determinants, and then smaller determinants ... as necessary until we
get the result. Here is an example with a 4 × 4 matrix.
Example 1.7. Let A be the following 4 × 4 matrix:
 
1 3 −1 1
 0 0 3 4 
A=  0
.
5 0 2 
2 0 5 2
We obtain:
   
0 3 4 3 −1 1
det(A) = 1 × det  5 0 2  − 0 + 0 − 2 × det  0 3 4 
0 5 2 5 0 2
( ) ( ) ( )
3 4 3 4 −1 1
= 0 − 5 × det + 0 − 6 × det + 0 − 10 × det
5 2 0 2 3 4

= −5 × (6 − 20) − 6 × (6 − 0) − 10 × (−4 − 3) = 104.

2
Definition 1.8. Let A be an n × n matrix. For each i and j such that
1 ≤ i ≤ n and 1 ≤ j ≤ n,
the cofactor Ci,j is defined by:
Ci,j = (−1)i+j det(Ai,j ).
In other words,
• when i + j is even: Ci,j = det(Ai,j )
• when i + j is odd: Ci,j = − det(Ai,j ).
Notice that, from this definition, the cofactor Ci,j is the determinant of the
submatrix Ai,j , with a sign + or − depending on the position of the entry ai,j in
the matrix A. The sign of (−1)i+j can be easily remembered by using the following
very convenient diagram:
 
+ − + − ... ...
 − + − ... ... ... 
 
 + − + ... ... ... 
... ... ... ... ... ...
Example 1.9. With ( )
2 −1
A=
3 5
we get
C1,1 = 5, C1,2 = −3, C2,1 = 1, C2,2 = 2.
Example 1.10. When  
−2 1 2
A= 7 0 1 
1 2 −3
we have 9 cofactors:
( ) ( ) ( )
0 1 7 1 −2 1
C1,1 = det = −2, C1,2 = − det = 22 ... C3,3 = det = −7.
2 −3 1 −3 7 0
Using cofactors, the formula giving the determinant of A can be written simply
as:
det(A) = a1,1 C1,1 + a2,1 C2,1 + ... + an,1 Cn,1 .
As we said before, this formula is called the cofactor expansion along the first
column of A. The following result is extremely interesting, important and useful.
Theorem 1.11. Let n ≥ 2 be an integer and A an n × n matrix. The determinant
of A can be computed by cofactor expansion along any column or any row of A.
In other words, the following formulas hold.
• For any fixed number j such that 1 ≤ j ≤ n:
det(A) = a1,j C1,j + a2,j C2,j + ... + an,j Cn,j .
This formula is called the cofactor expansion along column j of A.
• For any fixed number i such that 1 ≤ i ≤ n:
det(A) = ai,1 Ci,1 + ai,2 Ci,2 + ... + ai,n Ci,n .
This formula is called the cofactor expansion along row i of A.
Proof. We will prove the theorem by induction on n.
3
 ( )
a1,1 a1,2
• When n = 2 and A = , the cofactors are
a2,1 a2,2
C1,1 = a2,2 , C1,2 = −a2,1 , C2,1 = −a1,2 and C2,2 = a1,1
and the reader can easily check that all four cofactor expansions give:
a1,1 a2,2 − a2,1 a1,2 = det(A).
• As the proof below involves some quite complicated computations, for any
integer n ≥ 1, we will use the following notation:
∑
n
αk = α1 + α2 + ... + αn
k=1

to denote the sum of n real numbers α1 , α2 , ..., αn .

• Assume that the theorem is true for some integer n ≥ 2, let A be an
(n + 1) × (n + 1) matrix and fix an integer j such that 2 ≤ j ≤ n + 1. Then,
by cofactor expansion along column 1:
∑
n+1
det(A) = (−1)k+1 ak,1 det(Ak,1 )
k=1

and each Ak,1 is a n × n matrix so by induction hypothesis the determinant

det(Ak,1 ) can be computed by cofactor expansion along any column of Ak,1 .
So we can compute each det(Ak ) by cofactor expansion along column j − 1,
which corresponds to column j of A with the entry ak,j removed. For each
i such that 1 ≤ i ≤ n + 1 and i ̸= k, the position of the entry ai,j in column
j − 1 of the matrix Ak,1 is on row i if i < k or on row i − 1 if i > k. It
follows that:
∑
k−1 ∑
n+1
det(Ak,1 ) = (−1)i+j−1 ai,j det(A{i,k},{1,j} ) + (−1)i+j−2 ai,j det(A{i,k},{1,j} )
i=1 i=k+1

where A{i,k},{1,j} denotes the (n − 1) × (n − 1) matrix obtained from A by

removing rows i and k and columns 1 and j. And so by putting together
the two formulas, det(A) is equal to the sum of:
∑ k−1
n+1 ∑
(−1)i+j+k ak,1 ai,j det(A{i,k},{1,j} )
k=2 i=1

and of:
∑
n ∑
n+1
(−1)i+j+k−1 ak,1 ai,j det(A{i,k},{1,j} ).
k=1 i=k+1

Now let us compute the cofactor expansion along column j of A:

∑
n+1
(−1)i+j ai,j det(Ai,j )
i=1

Each det(Ai,j ) can be computed by cofactor expansion along column 1,

which is column 1 of A with the entry ai,1 removed. For each k such that
1 ≤ k ≤ n + 1 and k ̸= i, the position of the entry ak,1 in the matrix Ai,j
4
 is on row k if k < i or on row k − 1 if k > i. It follows that for each i such
that 1 ≤ i ≤ n + 1:
∑
i−1 ∑
n+1
det(Ai,j ) = (−1)k+1 ak,1 det(A{i,k},{1,j} ) + (−1)k ak,1 det(A{i,k},{1,j} )
k=1 k=i+1

where the matrices A{i,k},{1,j} are as above. And so by putting together

the two formulas, the cofactor expansion along column j of A is equal to
the sum of:
∑∑
n+1 i−1
(−1)i+j+k+1 ak,1 ai,j det(A{i,k},{1,j} )
i=2 k=1

and of:
∑
n ∑
n+1
(−1)i+j+k ak,1 ai,j det(A{i,k},{1,j} )
i=1 k=i+1

so one can see that both formulas are equal, and therefore the cofactor
expansion along column j of A is equal to det(A):
∑
n+1
det(A) = (−1)i+j ai,j det(Ai,j )
i=1

and this holds for all j such that 2 ≤ j ≤ n + 1.

• Assume now again that the theorem is true for some n ≥ 2, let A be an
(n + 1) × (n + 1) matrix and fix a number i such that 1 ≤ i ≤ n + 1. Then:
∑
n+1
det(A) = (−1)k+1 ak,1 det(Ak,1 )
k=1

and each Ak,1 is a n × n matrix so by induction hypothesis the determinant

det(Ak,1 ) can be computed by cofactor expansion along any row of Ak,1 .
Three cases occur:
– When k < i, we compute det(Ak,1 ) by cofactor expansion along row
i − 1 of Ak,1 , which is row i of A with the entry ai,1 removed. For each
j such that 2 ≤ j ≤ n + 1, the position of the entry ai,j in the matrix
Ak,1 is on row i − 1 and column j − 1. So we get:
∑
n+1
det(Ak,1 ) = (−1)i+j−2 ai,j det(A{i,k},{1,j} )
j=2

where A{i,k},{1,j} denotes the (n − 1) × (n − 1) matrix obtained from

A by removing rows i and k and columns 1 and j.
– When k = i, we leave the term (−1)i+1 ai,1 det(Ai,1 ) as it is.
– When k > i, we compute det(Ak,1 ) by cofactor expansion along row i
of Ak,1 , which is row i of A with the entry ai,1 removed. For each j
such that 2 ≤ j ≤ n + 1, the position of the entry ai,j in the matrix
Ak,1 is on row i and column j − 1. We get:
∑
n+1
det(Ak,1 ) = (−1)i+j−1 ai,j det(A{i,k},{1,j} )
j=2
5
 and so by putting the two formulas together, det(A) is equal to the sum of
(−1)i+1 ai,1 det(Ai,1 ) and of the two sums:
∑ ∑
i−1 n+1
(−1)i+j+k−1 ak,1 ai,j det(A{i,k},{1,j} )
k=1 j=2

and
∑ n+1
n+1 ∑
(−1)i+j+k ak,1 ai,j det(A{i,k},{1,j} ).
k=i+1 j=2
Now let us look at the cofactor expansion along row i of A:
∑
n+1
(−1)i+j ai,j det(Ai,j )
j=1

and, for each j ≥ 2, let us compute det(Ai,j ) by cofactor expansion along

the first column, which is the first column of A with the entry ai,1 removed.
For each integer k such that 1 ≤ k ≤ n + 1, when k < i, the position of the
entry ak,1 in the matrix Ai,j is on row k and column 1. When k > i, it is
on row k − 1 and column 1. So we get, for each j ≥ 2:
∑
i−1 ∑
n+1
det(Ai,j ) = (−1)k+1 ak,1 det(A{i,k},{1,j} ) + (−1)k ak,1 det(A{i,k},{1,j} )
k=1 k=i+1

where the matrices A{i,k},{1,j} are as above. Finally, the cofactor expansion
along row i is equal to the sum of (−1)i+1 ai,1 det(Ai,1 ) and of the two sums:
∑∑
n+1 i−1
(−1)i+j+k−1 ak,1 ai,j det(A{i,k},{1,j} )
j=2 k=1

and
∑ n+1
n+1 ∑
(−1)i+j+k ak,1 ai,j det(A{i,k},{1,j} ).
j=2 k=i+1
By the previous discussion, this sum is equal to det(A) so we have proved
that the cofactor expansion along row i of A is equal to det(A):
∑
n+1
det(A) = (−1)i+j ai,j det(Ai,j )
j=1

and this holds for all i such that 1 ≤ i ≤ n + 1.

• By induction on n, we have proved the theorem for all integers n ≥ 2 and
for all n × n matrices.
□
Example 1.12 (Example 1.3). Take
 
−2 1 2
A= 7 0 1 
1 2 −3
as in Example 1.3. The cofactor expansion along the 2nd row of A gives us:
( ) ( )
1 2 −2 1
−7 × det + 0 − 1 × det = −7 × (−7) − 1 × (−5) = 54
2 −3 1 2
6
which is the determinant of A as computed in Example 1.3. The reader can easily
convince himself by computation that all nine cofactor expansions will give the
same result.
Example 1.13 (Example 1.7).
 
1 3 −1 1
 0 0 3 4 

A= .
0 5 0 2 
2 0 5 2
The cofactor expansion along the 2nd row of A gives:
   
1 3 1 1 3 −1
det(A) = −3 det  0 5 2  + 4 det  0 5 0 
2 0 2 2 0 5
and we can compute the first determinant by cofactor expansion along the first
column, and the second one by cofactor expansion along the 2nd row to get:
( ) ( ) ( )
5 2 3 1 1 −1
det(A) = −3 det − 6 det + 20 det
0 2 5 2 2 5
and so
det(A) = −30 − 6 + 140 = 104.
By Theorem 1.11, it is clear that we can choose the most convenient row or
column in order to compute the cofactor expansion that gives det(A), and that
”convenient” means here a row or a column with a lot of zeros.
Example 1.14.  
2 3 −5 2
 5 9 0 1 
A=  2 0
.
0 0 
3 3 0 6
Here, the best way is to compute det(A) by cofactor expansion along either the 3rd
row or the 3rd column of A. We do it along the 3rd row:
 
3 −5 2
det(A) = 2 det  9 0 1 
3 0 6
and now we compute this determinant by cofactor expansion along the 2nd column:
( )
9 1
det(A) = 10 det = 10 × 51 = 510.
3 6

2. Some properties of the determinant

The main propertiy of cofactor expansion has many interesting consequences, a
few of which are listed below.
Proposition 2.1. Let A be an n × n matrix. If A has a row or a column which
consists only of zeros, then
det(A) = 0.
Proof. By cofactor expansion along the zero row, or the zero column, it is clear
that all terms are zero so det(A) = 0. □
7
 The following result is extremely useful and will appear many times in further
sections.
Proposition 2.2. If A is an upper triangular matrix, a lower triangular matrix
or a diagonal matrix, then the determinant of A is equal to the product of the
diagonal entries of A:
det(A) = a1,1 a2,2 ...an,n .
Proof. When A is upper triangular, by cofactor expansion along the first column
of A we get:
det(A) = a1,1 det(A1,1 )
since all the entries a2,1 , a3,1 , ...an,1 are zero. But the matrix A1,1 is again upper
triangular, so by induction on n the result is clear. When A is lower triangular, we
proceed similarly by cofactor expansion along the first row. When A is diagonal,
both ways work. □

Example 2.3.  
2 3 0 6
 0 −5 9 −1 
A=
 0
.
0 −1 0 
0 0 0 3
One can easily check that:
det(A) = 2 × (−5) × (−1) × 3 = 30.
The following proposition is also important and concerns the transpose.
Proposition 2.4. For any n × n matrix A, we have:
det(t A) = det(A).
Proof. We will prove the proposition by induction on n.
• When n = 1, the result is clear since t A = A.
• Suppose that the result is true for some n and let A be an (n + 1) × (n + 1)
matrix. Set
B =t A.
B has coefficients bi,j for all integers i and j between 1 and n + 1 and
bi,j = aj,i .
We compute det(B) by cofactor expansion along the first row and get:
det(B) = b1,1 det(B1,1 )−b1,2 det(B1,2 )+b1,3 det(B1,3 )+...+(−1)n+2 b1,n+1 det(B1,n+1 ).
For each j such that 1 ≤ j ≤ n + 1 we obviously have:
b1,j = aj,1
by definition of the transpose and
B1,j =t Aj,1
because B1,j is the matrix t A from which row 1 and column j have been
removed, but row 1 of t A is column 1 of A and column j of t A is row j
of A so t B1,j is the matrix A from which row j and column 1 have been
8
 removed, that is Aj,1 . Now since the matrices Aj,1 are n × n matrices, by
induction hypothesis
det(Aj,1 ) = det(t Aj,1 ) = det(B1,j )
so finally
det(B) = a1,1 det(A1,1 )−a2,1 det(A2,1 )+a3,1 det(A3,1 )+...+(−1)n+2 an+1,1 det(An+1,1 )
which is the cofactor expansion along the first column of A so
det(B) = det(A).
• By induction on n, we have proved the proposition for all integers n and
for all n × n matrices.
□

3. Computing determinants by row reduction

In this section, we want to explain how to compute determinants using the row-
reduction method. To do this, we need first to investigate the effect of elementary
row operations on the determinant of an n × n matrix.
Theorem 3.1. Let n ≥ 2 be an integer and A an n × n matrix. Let B be the
matrix obtained from A by performing one elementary row operation on A.
(1) If B is obtained from A by exchanging two rows, then
det(B) = − det(A).
(2) If B is obtained from A by multiplying one row by a constant k, then
det(B) = k det(A).
(3) If B is obtained from A by adding to one row the product of another row
by a constant, then
det(B) = det(A).
( ) ( )
a b c d
Proof. (1) When A is a 2 × 2 matrix: A = and B = is
c d a b
obtained from A by exchanging the two rows, it is clear that:
det(B) = cb − ad = − det(A).
Now suppose that the result is true for all n × n matrices and let A be
an (n + 1) × (n + 1) matrix. Let B be the matrix obtained from A by
exchanging rows i and j and let us compute det(B) by cofactor expansion
along another row k, which is always possible since A and B have at least
three rows. We obtain:
det(B) = (−1)k+1 bk,1 det(Bk,1 ) + ... + (−1)k+n+1 bk,n+1 det(Bk,n+1 ).
But row k of B is exactly row k of A so
bk,l = ak,l
for all l such that 1 ≤ l ≤ n + 1. Furthermore, each Bk,l is obtained from
Ak,l by exchanging two rows: the rows corresponding to rows i and j of A,
so by induction hypothesis:
det(Bk,l ) = − det(Ak,l )
9
 for all l such that 1 ≤ l ≤ n + 1. It follows that
det(B) = −(−1)k+1 ak,1 det(Ak,1 ) − ... − (−1)k+n+1 ak,n+1 det(Ak,n+1 )
which is the cofactor expansion along row k of A multiplied by (−1), so
finally:
det(B) = − det(A).
By induction on n, the result holds for all integers n and all n × n matrices.
(2) Let A be an n × n matrix and let B be the matrix obtained from A by
multiplying row i by a constant k. Then, by cofactor expansion along row
i, we get:
det(B) = (−1)i+1 bi,1 det(Bi,1 ) + ... + (−1)i+n bi,n det(Bi,n )
but it is clear that, for all j such that 1 ≤ j ≤ n,
bi,j = kai,j
since row i is multiplied by k and that
Bi,j = Ai,j
since the other rows of A are unchanged. So we get:
det(B) = (−1)i+1 kai,1 det(Ai,1 ) + ... + (−1)i+n kai,n det(Ai,n )
and this is just the cofactor expansion along row i of A multiplied by k,
and therefore
det(B) = k det(A).
(3) Let A be an n × n matrix and B be the matrix obtained from A by adding
to row i another row l multiplied by a constant k. Let us compute det(B)
by cofactor expansion along row i.
det(B) = (−1)i+1 bi,1 det(Bi,1 ) + ... + (−1)i+n bi,n det(Bi,n ).
For each j such that 1 ≤ j ≤ n, we have:
bi,j = ai,j + kal,j
and
Bi,j = Ai,j
since the rows of A other than row i are unchanged. So finally, det(B) is
equal to the sum of:
(−1)i+1 ai,1 det(Ai,1 ) + ... + (−1)i+n ai,n det(Ai,n )
and of:
(−1)i+1 kal,1 det(Ai,1 ) + ... + (−1)i+n kal,n det(Bi,n ).
Now the first sum is exactly the cofactor expansion along row i of A, so
it is equal to det(A). The second sum corresponds to k times the cofactor
expansion along row i of a matrix C. The matrix C is obtained from A by
deleting row i and replacing it by row l, with all the other rows unchanged,
which means that C has two equal rows: row i and row l. Therefore, by
exchanging these two rows, we see that necessarily:
det(C) = − det(C) = 0
10
 so that the second sum, which is equal to det(C) is zero. Finally we obtain:
det(B) = det(A)
which completes the proof.
□
Corollary 3.2. Theorem 3.1 remains valid if we replace the word row by the word
column.
Corollary 3.3. Let A be an n × n matrix. If A has two identical rows or two
identical columns, then
det(A) = 0.
Proof. As in the proof of Theorem 3.1, by exchanging the two identical rows (or
columns), the matrix A remains unchanged so we have
det(A) = − det(A) = 0.
□
Theorem 3.1 provides a new method to compute the determinant of an n × n
matrix A: by a sequence of elementary row operations, we transform A into a matrix
B which is in row-echelon form. The determinant of B is then easily computed,
and we deduce from it the determinant of A. More precisely, the following result
holds.
Theorem 3.4. Let n ≥ 2 be an integer and A an n × n matrix. There exist an
upper triangular matrix B, an integer p ≥ 1 and p nonzero constants k1 , k2 , ..., kp
such that:
det(B) = k1 k2 ..kp det(A) = b1,1 b2,2 ...bn,n
where the bi,i are the diagonal entries of B for 1 ≤ i ≤ n.
Proof. By a sequence of p elementary row operations, we transform A into a matrix
B which is in row-echelon form. B is then necessarily upper triangular and by
Proposition 2.2
det(B) = b1,1 b2,2 ...bn,n .
Furthermore, by Theorem 3.1 there exist p nonzero constants k1 , ..., kp such that:
det(B) = k1 k2 ..kp det(A).
More precisely, for each i such that 1 ≤ i ≤ p,
(1) if the i-th operation is an exchange of two rows, then
ki = −1,
(2) if the i-th operation is a dilatation by a constant k ̸= 0 then
ki = k,
(3) if the i-th operation is a transvection, then
ki = 1.
So the result follows and we can compute det(A) by the formula:
b1,1 b2,2 ...bn,n
det(A) =
k1 k2 ...kp
□
11
 We now give a few examples of practical computations.
Example 3.5 (Example 1.3).
 
−2 1 2
A= 7 0 1 
1 2 −3
We proceed by elementary row operations:
     
−2 1 2 1 2 −3 1 2 −3
 7 0 1 ∼ 7 0 1  ∼  0 −14 22 
1 2 −3 −2 1 2 0 5 −4
   
1 2 −3 1 2 −3
∼  0 1 − 11 7
 ∼  0 1 11  .
7
0 5 −4 0 0 27 7
It follows that  
1 2 −3
 
 
B=  0 1 11 
7 
 
0 0 27 7
is a row-echelon form of A and that
27 27
det(B) = b1,1 b2,2 b3,3 = 1 × 1 × =
7 7
because B is upper triangular. Furthermore, the first operation we used was an
exchange so k1 = −1, the second and third operation were transvections so k2 =
k3 = 1. The fourth operation was dividing the second row by −14 so k4 = −1
14 , the
fifth operation was again a transvection so k5 = 1. Finally, we get:
det(B) = k1 k2 k3 k4 k5 det(A)
which gives us the equation:
27 1
= det(A)
7 14
and this implies det(A) = 54.
In the next example, we see what happens when A is not invertible.
Example 3.6.  
2 1 5
A= 1 2 1 
4 5 7
We proceed by elementary row operations:
       
2 1 5 1 2 1 1 2 1 1 2 1
 1 2 1 ∼ 2 1 5  ∼  0 −3 3 ∼ 0 −3 3  .
4 5 7 4 5 7 0 3 −3 0 0 0
It follows that  
1 2 1
B= 0 −3 3 
0 0 0
12
is a row-echelon form of A and that
det(B) = 1 × (−3) × 0 = 0
so finally
det(A) = 0.
Notice that the first operation was an exchange so k1 = −1. The second, third and
fourth operations were transvections so k2 = k3 = k4 = 1. So we get det(B) =
− det(A) but here we needn’t mind about this, because anyway the result is zero.
This method is very effective for larger matrices: when A is a 4 × 4 or a 5 × 5
matrix, in most cases this method is much quicker and easier than cofactor expan-
sion.
Example 3.7 (Example 1.7).
 
1 3 −1 1
 0 0 3 4 

A= .
0 5 0 2 
2 0 5 2
We proceed by using only transvections:
       
1 3 −1 1 1 3 −1 1 1 3 −1 1 1 3 −1 1
 0 0 3 4   3 4   3 4   0 −4 2 
 ∼ 0 0 ∼ 0 0 ∼ 1 
 0 5 0 2   0 5 0 2   0 5 0 2   0 5 0 2 
2 0 5 2 0 −6 7 0 0 −1 7 2 0 −1 7 2
     
1 3 −1 1 1 3 −1 1 1 3 −1 1
 0 1 −4 2   0 1 −4 2   0 1 −4 2 
∼   ∼  ∼ 
0 0 20 −8   0 0 −1 −36   0 0 −1 −36 
0 0 3 4 0 0 3 4 0 0 0 −104
It follows that
det(A) = det(B) = 1 × 1 × (−1) × (−104) = 104.

4. The determinant of an invertible matrix

We are now ready to prove the following extremely important theorem.
Theorem 4.1. A square matrix A is invertible if and only if
det(A) ̸= 0.
Proof. Let A be an n × n matrix. When n = 1, the result is obvious. When n ≥ 2,
let B be the reduced row-echelon form of A. By Theorem 3.1, there exists a finite
number p ≥ 1 of nonzero numbers k1 , k2 , ..., kp such that
det(B) = k1 k2 ...kp det(A).
• If A is not invertible, B is an upper triangular matrix whose bottom row is
zero. By Proposition 2.2, det(B) = 0 because bn,n = 0 so det(A) = 0.
• If A is invertible, then B = In and det(B) = 1. It follows that
1
det(A) =
k1 k2 ...kp
so det(A) ̸= 0.
□
13
 The following proposition is also very important: it concerns the determinant of
a product of matrices.
Proposition 4.2. Let A and B be n × n matrices. We have:
det(AB) = det(A) det(B).
Proof. We proceed in the following steps.
• First we prove that det(EB) = det(E) det(B) when E is an elementary ma-
trix. Notice that the determinant of elementary matrices is easily computed
from Theorem 3.1 and the fact that det(In ) = 1. We get:
(1) If E is a transposition, then det(E) = −1.
(2) If E is a dilatation that multiplies row i by k ̸= 0, then det(E) = k.
(3) If E is a transvection, then det(E) = 1.
The result det(EB) = det(E) det(B) follows from Theorem 3.1.
• When A is invertible, A is a product of p elementary matrices:
A = E1 E2 ...Ep
for some integer p ≥ 1. By induction on p, we prove easily that:
det(A) = det(E1 ) det(E2 )... det(Ep )
and that
det(AB) = det(E1 ) det(E2 )... det(Ep ) det(B)
and it follows that det(AB) = det(A) det(B).
• When A is not invertible, det(A) = 0 but AB is also not invertible (if AB
were invertible, there would exist C such that ABC = In so A would be
invertible with A−1 = BC) and so det(AB) = 0 = det(A) det(B) which
completes the proof.
□

Proposition 4.3. Let A be an n × n invertible matrix. Then:

1
det(A−1 ) =
det(A)
Proof. We have AA−1 = In and det(In ) = 1 so the result
det(A) det(A−1 ) = 1
follows from Proposition 4.2. □

5. The comatrix
In this section, we want to describe another new method to compute the inverse
of an n × n matrix A, by using only determinants. We begin by constructing the
matrix of the cofactors of A.
Definition 5.1. Let n ≥ 2 be an integer and A an n × n matrix. The comatrix
of A is the n × n matrix whose entries are the cofactors Ci,j of A (see Definition
1.8). It is denoted by Com(A).
14
Example 5.2 (Example 1.3).
 
−2 1 2
A= 7 0 1 
1 2 −3
We compute the cofactors one by one and get:
 
−2 22 14
Com(A) =  7 4 5 .
1 16 −7
The following proposition shows all the importance of the comatrix.
Proposition 5.3. Let n ≥ 2 be an integer and A an n × n matrix. The following
formula holds:
t
Com(A)A = At Com(A) = det(A)In .
Proof. Set B = At Com(A) and let bi,j be the entries of B, for 1 ≤ i ≤ n and
1 ≤ j ≤ n. The diagonal entries of B are:
bi,i = ai,1 Ci,1 + ai,2 Ci,2 + ... + ai,n Ci,n
and this is exactly the cofactor expansion along row i of A, so
bi,i = det(A)
for each i. Now let us fix i and j such that i ̸= j and compute:
bi,j = ai,1 Cj,1 + ai,2 Cj,2 + ... + ai,n Cj,n .
By looking closely at this formula, we see that it is the cofactor expansion along
row j of a matrix D. All the rows of D except row j are equal to the corresponding
rows of A, and row j of D is equal to row i of A. It follows that the rows i and
j of the matrix D are identical and thus by Corollary 3.3 we have det(D) = 0. It
follows that
bi,j = 0 when i ̸= j.
Thus B = det(A)In . The other way around is proved similarly. □
The following theorem gives a general formula to compute the inverse of an n × n
matrix.
Theorem 5.4. Let n ≥ 2 be an integer and A an n × n matrix.
• If det(A) = 0 then A is not invertible.
• If det(A) ̸= 0 then A is invertible and its inverse is given by the formula:
1 t
A−1 = Com(A).
det(A)
Proof. The first part of the theorem is proved in Theorem 4.1. The second part
follows directly from Proposition 5.3. □
Example 5.5. When ( )
a b
A=
c d
is a 2 × 2 matrix we get:
( )
d −c
Com(A) =
−b a
15
and det(A) = ad − bc, so when det(A) ̸= 0 we get:
( )
−1 1 d −b
A = .
ad − bc −c a
Example 5.6 (Example 1.3).
 
−2 1 2
A= 7 0 1 
1 2 −3
We have det(A) = 54 and Com(A) has been computed in Example 5.2 above:
 
−2 22 14
Com(A) =  7 4 5 .
1 16 −7
It follows that:
 −2 7 1

  54 54 54
−2 7 1  
1   
A −1
= 22 4 16  =

22
54
4
54
16
54
.

54  
14 5 −7
14 5 −7
54 54 54
Example 5.7.  
1 −2 2
A= 1 −1 3 
2 −4 5
We compute
det(A) = 1 × (−5 + 12) − 1 × (−10 + 8) + 2 × (−6 + 2) = 1
and  
7 1 −2
Com(A) =  2 1 0 
−4 −1 1
and so  
7 2 −4
A−1 = 1 1 −1  .
−2 0 1
6. Cramer’s formulas
Theorem 5.4 provides us with a new method for solving systems of linear equa-
tions, under some conditions.
Definition 6.1. A system of linear equations is called a Cramer system if the
following two conditions are satisfied:
• The number of variables is equal to the number of equations.
• The coefficient matrix of the system is invertible.
More precisely, a Cramer system has the following form:


 a1,1 x1 + a1,2 x2 +... + a1,n xn = b1

a2,1 x1 + a2,2 x2 +... + a2,n xn = b2

 ... ... ... ... ...

an,1 x1 + an,2 x2 +... + an,n xn = bn
16
The coefficient matrix of the system is the square matrix:
 
a1,1 a1,2 ... ... a1,n
 a2,1 a2,2 ... ... a2,n 
A=  ...


... ... ...
an,1 an,2 ... ... an,n
and A is invertible, therefore the system has a unique solution which is:
−
→ −
→
x = A−1 b .
By looking closely at this equation and by using Theorem 5.4, we obtain the fol-
lowing theorem.
Theorem 6.2 (Cramer’s formulas). Let n ≥ 2 be an integer, A an n×n invertible
−
→
matrix and b a column vector in Rn . The system of linear equation
−
→
A− →
x = b
has a unique solution  
x1
−
→  x2 
x = 
 ... 
xn
which can be computed by the following formulas:
−
→→
det(−
→v 1 ...−
→
v i−1 b −v i+1 ...−
→
v n)
xi =
det(A)
for each i such that 1 ≤ i ≤ n. Here
−
→
v 1, −
→
v 2 , ..., −
→
vn
denote the columns of the matrix A and
−
→→
(−
→
v 1 ...−
→
v i−1 b −
v i+1 ...−
→
v n)
denotes the n × n matrix obtained from A by deleting column i and replacing it
−
→
with the column vector b .
Proof. By Theorem 5.4, A being invertible, det(A) ̸= 0 and the following formula
gives the solution:
−
→ 1 t −
→
x = Com(A) b .
det(A)
Therefore, for each i the variable xi is given by the formula:
1
xi = (b1 C1,i + b2 C2,i + ... + bn Cn,i )
det(A)
and the numerator of this expression is exactly the cofactor expansion along column
i of a matrix whose columns are all equal to the columns of A except column i which
−
→
is equal to the column vector b . The result follows. □

We give a few examples of practical computations. Notice that this method only
works for Cramer systems. The other cases need to be solved by more general
methods, as described in previous lessons.
17
Example 6.3.
{
x1 + x2 =3
3x1 − x2 =5
This is a Cramer system, because its coefficient matrix
( )
1 1
A=
3 −1
is clearly invertible with det(A) = −4. We compute the solution with Cramer’s
formulas: ( )
3 1
det
5 −1 −8
x1 = = =2
det(A) −4
( )
1 3
det
3 5 −4
x2 = = = 1.
det(A) −4
Example 6.4.

 x1 + x2 + x3 =0
2x1 − x2 − x3 =6

x1 + 2x2 − x3 =7
Here, the coefficient matrix is:
 
1 1 1
A= 2 −1 −1 
1 2 −1
so first we need to compute det(A). We do it by the row reduction method, using
only transvections:
       
1 1 1 1 1 1 1 1 1 1 1 1
 2 −1 −1  ∼  0 −3 −3  ∼  0 1 −11  ∼  0 1 −11 
1 2 −1 0 1 −2 0 1 −2 0 0 9
It follows that det(A) = 9 and the matrix A is invertible. So the system is a Cramer
system and we can solve it with the formulas:
 
0 1 1
det  6 −1 −1 
7 2 −1 18
x1 = = =2
det(A) 9
 
1 0 1
det  2 6 −1 
1 7 −1 9
x2 = = =1
det(A) 9
 
1 1 0
det  2 −1 6 
1 2 7 −27
x3 = = = −3.
det(A) 9
18
Example 6.5. 

 x1 + x2 =3

3x1 − x2 −4x4 = 5

 −x1 + 2x2 − x3 =0

3x1 + 2x2 − x3 −4x4 = 8
First we compute the determimant of the coefficient matrix
 
1 1 0 0
 3 −1 0 −4 
A=  −1

2 −1 0 
3 2 −1 −4
and we obtain by row reduction, using only transvections:
       
1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0
 3 −1 0 −4   0 −4 0 −4   0 −1 −1 −4   0 −1 −1 −4 
 ∼ ∼ ∼ 
 −1 2 −1 0   0 3 −1 0   0 3 −1 0   0 0 −4 −12 
3 2 −1 −4 0 −1 −1 −4 0 −1 −1 −4 0 0 0 0
which gives
det(A) = 0.
This system is not a Cramer system, and cannot be solved by using Cramer’s
formulas. It has infinitely many solutions.

19