0% found this document useful (0 votes)
124 views34 pages

Chapter 5 - Determinants

This document provides an overview of determinants and their properties. It begins by stating that determinants will be defined in multiple ways and properties proven, with a focus on using properties to calculate determinants rather than examinable proofs. It then provides: - A brief history of determinants and their relation to matrices. - Examples of calculating 2x2 and 3x3 determinants. - A definition of the sign of a permutation that is used to define the general nxn determinant. - Properties of permutations that are used to prove properties of determinants. - The main properties of determinants, including how row/column operations affect the determinant. The document concludes by stating that determinants can

Uploaded by

Christopher
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
124 views34 pages

Chapter 5 - Determinants

This document provides an overview of determinants and their properties. It begins by stating that determinants will be defined in multiple ways and properties proven, with a focus on using properties to calculate determinants rather than examinable proofs. It then provides: - A brief history of determinants and their relation to matrices. - Examples of calculating 2x2 and 3x3 determinants. - A definition of the sign of a permutation that is used to define the general nxn determinant. - Properties of permutations that are used to prove properties of determinants. - The main properties of determinants, including how row/column operations affect the determinant. The document concludes by stating that determinants can

Uploaded by

Christopher
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 34

Chapter 5

Determinants

MATH2601 Slides, 2021 – p. 1


5 Determinants
In this chapter we shall look at a several approaches to
defining determinants and proving their properties.
Mainly, the idea is to give you the background and to show you
how the topic can be developed: proofs from the definition will
not be examinable.
However, you will be expected to know how to use their
properties to calculate determinants quickly, and to be able to
use these properties in linear algebra proofs and
computations in subsequent chapters.
As a historical aside, note that determinants predate matrices.
They were invented because they determine when a square
system of linear equations has a solution; J.J. Sylvester
coined the term matrix (latin for “womb”) for the array of
numbers that gave birth to determinants.

MATH2601 Slides, 2021 – p. 2



a a

11 12
Recall det(a11 ) = |(a11 )| = a11 and = a11 a22 − a12 a21 .
a21 a22
Also

a a a
11 12 13 a a
a a
a a

22 23 21 23 21 22
a a a = a − a + a

21 22 23 11 12 13
a32 a33 a31 a33 a31 a32
a31 a32 a33
= a11 a22 a33 − a11 a23 a32 − a12 a21 a33 + a12 a23 a31
+ a13 a21 a32 − a13 a22 a31
X
= sign(σ) a1σ(1) a2σ(2) a3σ(3)
σ∈S3

where S3 is the group of permutations of 3 objects and


sign(σ) = ±1 depending on the permutation σ .

MATH2601 Slides, 2021 – p. 3


Notice that we could also have written the 2 × 2 determinant as
X
sign(σ) a1σ(1) a2σ(2)
σ∈S2

From this it seems clear how we could generalise — the


problem is defining sign(σ).
To do that let us first recall our notation for permutations on
Ωn = {1, 2, 3, . . . , n} from chapter 1 e.g.
!
1 2 3
1 3 2

on Ω3 swaps 2 and 3 and leaves 1 unchanged.

MATH2601 Slides, 2021 – p. 4


We next note that we could easily omit the first row of this
matrix, and so just write our permutation as a re-ordering of
the elements of Ωn : [p1 , p2 , · · · , pn ] where 1 7→ p1 , 2 7→ p2 etc.
So [2, 1, 4, 3] is the permutation on Ω4 that swaps 1 and 2 and
also swaps 3 and 4.
In a permutation [p1 , p2 , · · · , pn ], an inversion occurs
whenever a larger number precedes a smaller.
For example, on Ω6
[3, 6, 2, 1, 4, 5]
contains seven inversions:
• 3 precedes 1 and 2 (two inversions)
• 6 precedes 1, 2, 4 and 5 (four inversions)
• 2 precedes 1 (one inversion)

MATH2601 Slides, 2021 – p. 5


Definition 5.1 If a permutation σ = [p1 , p2 , · · · , pn ] of Ωn
contains k inversions then its sign, sign(σ) = (−1)k .
A permutation is called even or odd depending on whether
the number of inversions is even or odd.
Our example [3, 6, 2, 1, 4, 5] on Ω6 is odd; the case [2, 1, 4, 3] on
Ω4 has 2 inversions and so is even .
If we now look back at the definition of the 3 × 3 determinant,
we see that the sign in front of each term does match up with
the sign of the permutations:

sign([1, 2, 3]) = 1 sign([1, 3, 2]) = −1 sign([2, 1, 3]) = −1


sign([3, 2, 1]) = −1 sign([2, 3, 1]) = 1 sign([3, 1, 2]) = 1

This prompts us to make our formal definition of the


determinant for arbitrary square matrices. . .

MATH2601 Slides, 2021 – p. 6


Definition 5.2 The determinant of an n × n matrix
A = (aij )i,j=1,...,n is
X
det(A) = sign(σ)a1σ(1) a2σ(2) · · · anσ(n) .
σ∈Sn

In order to prove properties of the determinant, we need a few


properties of the sign of a permutation. To do that we
introduce a useful object:
Definition 5.3 Define the nth difference product ∆n by
n
Y
∆n = (i − j)
i=1,j>i
= (1 − 2)(1 − 3) . . . (1 − n)(2 − 3)(2 − 4) . . . ((n − 1) − n)

MATH2601 Slides, 2021 – p. 7


For example
∆3 = (1 − 2)(1 − 3)(2 − 3)
We now define the action of a permutation σ ∈ Sn on ∆n by
n
Y
σ(∆n ) = (σ(i) − σ(j))
i=1,j>i

So, for example the permutation σ = [2, 3, 1] on ∆3 gives

σ(∆3 ) = (2 − 3)(2 − 1)(3 − 1) = ∆3

We see from this example that all a permutation will do is


rearrange the factors of ∆n , with some of the terms having
their signs changed.

MATH2601 Slides, 2021 – p. 8


In fact, it is clear that the number of terms whose signs are
changed must be the same as the number of inversions in the
permutation. So
Proposition 5.1 For any σ ∈ Sn , σ(∆n ) = sign(σ)∆n .

Lemma 5.2 If α, β ∈ Sn then sign(α ◦ β) = sign(α) sign(β), and


sign(α−1 ) = sign(α).
Proof: Using the previous result

sign(α ◦ β)∆n = α ◦ β(∆n ) = α(β(∆n ))


= α(sign(β)∆n ) = sign(α) sign(β)∆n

and the first part follows.


The second part is now immediate, since clearly the identity
permutation has no inversions and so is even. 

MATH2601 Slides, 2021 – p. 9


This lemma tells us that sign : Sn → {−1, 1} is a group
homomorphism, where {−1, 1} is a group under multiplication.

One last result on permutations:


Proposition 5.3 Every transposition (i.e. swap of two
elements) in Sn is odd.
Proof: Suppose σ ∈ Sn swaps i and i + j , with j ≥ 1 and does
nothing else.
Then there will be 2j − 1 inversions in σ :
i + j will precede the j numbers i + 1, i + 2,. . . , i + j − 1 and i;
each of the j − 1 numbers i + 1, i + 2,. . . , i + j − 1 precede i.
There are no other inversions, so sign(σ) = (−1)2j−1 = −1 

And now we can prove all the properties of the determinant


we need. . .

MATH2601 Slides, 2021 – p. 10


Theorem 5.4 Let A ∈ Mp,p . Then
X
a) det(A) = sign(σ)aσ(1)1 aσ(2)2 · · · aσ(n)n .
σ∈Sn

b) det(A) = det(AT ) and det(A∗ ) = det(A).


c) If any row or column of A is zero, det(A) = 0.
d) If a permutation is applied to the rows or columns of A,
then the determinant is multiplied by the sign of the
permutation.
e) If A has two rows or two colums the same, then
det(A) = 0.
f) If any row or column of A has a multiple of another row or
column (respectively) added to it, the determinant is
unchanged.

MATH2601 Slides, 2021 – p. 11


Before we go through the proof of this theorem, we note one
important consequence, whose proof we will cover after
proving theorem 5.4:
Theorem 5.5 If det is considered as a map on the rows or
columns of a matrix (that is, from p copies of Fp to F), then it is
multilinear and alternating.
We note a useful special case of this last result, namely if any
row or column of A is multiplied by a constant α, then the
determinant is multiplied by α.

MATH2601 Slides, 2021 – p. 12


Proof (theorem 5.4): For part (a), note that each term
sign(σ)a1σ(1) a2σ(2) · · · anσ(n) in the sum for the determinant
could be re-ordered by the columns instead of the rows as

sign(σ −1 )aσ−1 (1)1 aσ−1 (2)2 · · · aσ−1 (n)n

since sign(σ) = sign(σ −1 ).


But adding all these term over all σ −1 ∈ Sn gives the same
sum as adding them over all σ ∈ Sn (why?), proving (a).

The first part of (b) now follows since the (i, j)th entry of AT is
the (j, i)th entry of A.
The second part of (b) follows from the first and the properties
of the complex conjugate.
It follows that in the rest of the proof, the results for columns
follow from the results for rows, so we just prove those.
MATH2601 Slides, 2021 – p. 13
For (c), suppose that the ith row is zero.
Each term a1σ(1) · · · anσ(n) contains an entry aiσ(i) = 0.
So each product in the sum defining the determinant is zero
and det(A) = 0.

For (d), let B be A with rows permuted by α−1 , so that

bij = aα(i)j

All permutations σ ∈ Sn can be written as σ = τ ◦ α (why?), so


X
det(B) = sign(τ ◦ α)b1,τ ◦α(1) · · · bn,τ ◦α(n) ,
τ ∈Sn

where sign(τ ◦ α) = sign(τ ) sign(α) from lemma 5.2.

MATH2601 Slides, 2021 – p. 14


Now

b1,τ ◦α(1) · · · bn,τ ◦α(n) = aα(1),τ ◦α(1) · · · aα(n),τ ◦α(n)


= a1,τ (1) · · · an,τ (n)

by commutativity of multiplication.
So det(B) = sign(α) det(A) = sign(α−1 ) det(A).
(Note the special case of a row or column swap, which
changes the sign of the determinant.)
For (e), note that if A has two rows the same, then when we
swap them we do not change the matrix, but from the special
case of (d) we change the sign of the determinant.
The only way this can be so is if det(A) = 0.

MATH2601 Slides, 2021 – p. 15


Finally for (f), suppose that β times row s is added to row t,
and call the resulting matrix B .
Then btj = atj + βasj (sic) for all j and so
X
det(B) = sign(σ)a1σ(1) · · · asσ(s) · · · (atσ(t) + βasσ(t) ) · · · anσ(n)
σ∈Sn
X
= sign(σ)a1σ(1) · · · asσ(s) · · · atσ(t) · · · anσ(n)
σ∈Sn
X
+β sign(σ)a1σ(1) · · · asσ(s) · · · asσ(t) · · · anσ(n)
σ∈Sn

The first sum on the right hand side is det(A).


The second term is zero from part (e). 

MATH2601 Slides, 2021 – p. 16


And now we prove theorem 5.5:
Theorem 5.5 If det is considered as a map on the rows or
columns of a matrix (that is, from p copies of Fp to F), then it is
multilinear and alternating.
Proof: Part (d) of theorem 5.4 gives us the alternating
property, and (along with part (b)) implies we need only check
linearity in the first row to prove the result.
So let A = (aij ) and suppose a1j = λa′1j + a′′1j : the first row of A
is the sum of the row vectors λ(a′11 a′12 . . . ) and (a′′11 a′′12 . . . ).
Let A′ be the matrix whose first row is the a′1j but all the other
rows are the same as those of A and similarly A′′ whose first
row is the a′′1j instead.

MATH2601 Slides, 2021 – p. 17


Then
X
det(A) = sign(σ)a1σ(1) a2σ(2) · · · anσ(n)
σ∈Sn
X  
= sign(σ) λa′1σ(1) + a′′1σ(1) a2σ(2) · · · anσ(n)
σ∈Sn
X
=λ sign(σ)a′1σ(1) a2σ(2) · · · anσ(n)
σ∈Sn
X
+ sign(σ)a′′1σ(1) a2σ(2) · · · anσ(n)
σ∈Sn
= λ det(A′ ) + det(A′′ )

proving the result. 

MATH2601 Slides, 2021 – p. 18


We need one more definition and result to allow us to
efficiently calculate determinants:
Definition 5.4 For A ∈ Mp,p (F) let Aij be the matrix obtained
by deleting row i and column j . We call Aij the (i, j)-minor of
A.
Lemma 5.6 Suppose that row i of matrix A ∈ Mp,p (F) is zero
except for entry aij .
Then det(A) = (−1)i+j aij det(Aij ).
The same result would apply if column j of A were zero except
for aij , of course.
So this lemma implies that, e.g.

1 3 2
1 2
2+2
0 2 0 = (−1) ×2×

1 −1


1 5 −1
MATH2601 Slides, 2021 – p. 19
Proof: Apply the permutation [1, 2, . . . , i − 1, p, i, i + 1, . . . , p − 1]
to the rows of A, and then apply
[1, 2, . . . , j − 1, p, j, j + 1, . . . , p − 1] to the columns of the
resulting matrix.
We then have a matrix B whose last row is (0 0 . . . 0 aij ) and
whose (p, p)-minor is the minor Aij of A.
For example,

1 3 2 1 3 2 1 2 3

0 2 0 −→ 1 5 −1 −→ 1 −1 5



1 5 −1 0 2 0 0 0 2

The permutation of the rows has p − i inversions, and the


permutation of the columns p − j inversions, so

det(B) = (−1)2p−i−j det(A) i.e. det(A) = (−1)i+j det(B) .


MATH2601 Slides, 2021 – p. 20
Now bpj = 0 except for j = p, so any permutation in Sp that
does not fix p will give a zero term in the sum for the
determinant.
The set of all permutations that do fix p is just Sp−1 , and so
X
det(B) = bpp sign(σ)b1σ(1) b2σ(2) · · · bp−1 σ(p−1) .
σ∈Sp−1

But the sum given here is just the determinant of the


(p, p)-minor of B , which we constructed to be the same as the
Aij , and bpp = aij .
Putting these results together gives us the required formula. 

MATH2601 Slides, 2021 – p. 21


Lemma 5.6 and the last part of theorem 5.4 allows us to
efficiently calculate determinants as follows:

1) Add appropriate multiples of one row (or column) to all


the other rows or columns until one row (or column) has
only one non-zero entry. This does not change the
determinant.
2) Use lemma 5.6 to reduce the determinant to one that is
one row and column smaller.
3) Repeat steps 1 and 2 until you have a 2 × 2 determinant,
for which you use the formula we learnt in first year.

−1 0 0 0

1 3 4 1
Example 5.1 Find = 42.

2 2 1 5

0 1 2 7
MATH2601 Slides, 2021 – p. 22
An alternative way of using row operations to calculate
determinants is to row reduce the matrix to upper triangular
form, keeping careful track of any row swaps.
This works because of
Proposition 5.7 The determinant of an upper or lower
triangular matrix is the product of its diagonal elements.
Proof: If A is upper triangular, then aij = 0 if i > j .
It follows that a product a1σ(1) a2σ(2) · · · anσ(n) is non-zero only if

σ(1) ≥ 1 , σ(2) ≥ 2 , . . . , σ(n) ≥ n .

Since σ is a permutation, this means that σ(k) = k for all k ,


that is, σ = e, the identity permutation.
So det(A) = sign(e)a1e(1) a2e(2) · · · an e(n) = a11 a22 · · · ann .
The lower triangular result follows similarly, or from
theorem 5.4 b). 
MATH2601 Slides, 2021 – p. 23
We now turn to proving the multiplicativity of the determinant:

det(AB) = det(A) det(B)

and the results that flow from it.


We need another idea to prove that:
Definition 5.5 Recall that an elementary row operation on a
matrix is one of the following:
1) swapping two rows
2) multiplying one row by a non-zero scalar
3) adding a multiple of one row to another
An elementary matrix is a matrix obtained from an identity
matrix after an elementary row operation.
Doing an elementary row operation on a matrix is the same as
multiplying the matrix on the left by the equivalent elementary
matrix, as is easy to see.
MATH2601 Slides, 2021 – p. 24
Now a simple corollary of proposition 5.7 is that
det(I) = 1 for all identity matrices.
If follows from theorem 5.4 that for elementary matrices E ,
1) if E swaps two rows, det(E) = −1;
2) if E multiplies one row by a non-zero scalar λ, det(E) = λ;
3) if E adds a multiple of one row to another, det(E) = 1.
But now we can see that
if E is an elementary matrix, then det(EA) = det(E) det(A) for
any matrix A.
Because multiplying by E applies the row operation E codes
up, and so changes the determinant of A according to
theorem 5.4 in a way precisely accounted for by multiplying
det(A) by det(E).

MATH2601 Slides, 2021 – p. 25


We next need to put a result from first year into our new
language:
Lemma 5.8 If A is invertible there is a sequence of
elementary matrices E1 , E2 , . . . Ek such that
k
Y
A = E1 E2 · · · Ek and det(A) = det(Ei ) .
i=1

Proof: An invertible matrix A can be row reduced to the


identity, which is the same as saying A can be multiplied by a
sequence of elementary matrices to give the identity, i.e.
Fk Fk−1 · · · F1 A = I for some elementary matrices Fi .
But the inverse of an elementary matrix is an elementary
matrix and the product follows with Ei = Fi−1 .
The determinant formula follows from multiplicativity for
determinant on elementary matrices. 
MATH2601 Slides, 2021 – p. 26
Lemma 5.9 A matrix A is invertible if and only if det(A) 6= 0.
Proof: If A is invertible it is the product of elementary matrices
from lemma 5.8: A = E1 E2 · · · Ek
Q
But then det(A) = i det(Ei ), and all the Ei have non-zero
determinant.
So if A is invertible then det(A) 6= 0.
Conversely, if A is not invertible, it does not have full rank, and
so the columns are linearly dependent.
So we can add multiples of some of the columns to another
one to give a matrix with a zero column: doing this has not
changed the determinant by theorem 5.4 (f), but the resulting
matrix must have determinant zero by theorem 5.4 (c).
Hence if A is not invertible then det(A) = 0. 

MATH2601 Slides, 2021 – p. 27


Theorem 5.10 For any two matrices A and B ,

det(AB) = det(A) det(B) .

Proof: Suppose B is invertible and B = E1 E2 . . . Ek as a


product of elementary matrices. Then

det(AB) = det (AB) = det(B T AT )


T


= det(EkT . . . E1T A)
= det(Ek ) . . . det(E1 ) det(A)
= det(B) det(A) = det(A) det(B) .

On the other hand if B is not invertible then there is a non-zero


x with Bx = 0.
But then ABx = 0 and so AB is not invertible.
So both det(AB) and det(A) det(B) are zero. 
MATH2601 Slides, 2021 – p. 28
We have already used this result to show that determinant is a
similarity invariant; one other easy corollary is:
−1 1
Corollory 5.11 If A is invertible, then det(A ) = .
det(A)

Before we turn to the cofactor expansion, I want to highlight


something from the proof of lemma 5.9 that may be a little
buried in all that we have done, but is useful in the next
chapter:
Lemma 5.12 A matrix A has determinant zero if and only if it
has linearly dependent columns and hence if and only if it has
linearly dependent rows.

MATH2601 Slides, 2021 – p. 29


The other standard formula for a determinant is known as the
cofactor expansion, and we finish this chapter with that.
Definition 5.6 In matrix A the number cij = (−1)i+j det(Aij ) is
called the cofactor of element aij .
The formula in lemma 5.6 can thus be rewritten as
det(A) = aij cij for a matrix with only one non-zero entry on row
i (or column j ).
Theorem 5.13 For any A ∈ Mp,p (F) and any fixed j ,
p
X
det(A) = aij cij
i=1

That is to say, pick a column (fixed j ), multiply each element


by its cofactor and add up to get the determinant.
You can do this with rows as well of course.
MATH2601 Slides, 2021 – p. 30
Proof: With j fixed, for each i define matrix Ai to be the same
matrix as A except that every entry in column j other than the
(i, j) one is zero.
 
1 2 3
So if A = 4 5 6 then with j = 2 we have
 
7 8 9
     
1 2 3 1 0 3 1 0 3
A1 = 4 0 6 , A2 = 4 5 6 A 3 = 4 0 6  .
     
7 0 9 7 0 9 7 8 9

Lemma 5.6 then tells us that det(Ai ) = aij cij . In the above,
det(A1 ) = 2 × [(−1) × (36 − 42)] = 12,
det(A2 ) = 5 × [(1) × (9 − 21)] = −60 and
det(A3 ) = 8 × [(−1) × (6 − 12)] = 48
MATH2601 Slides, 2021 – p. 31
But the j th column of A is the sum of the j th columns of the
Ai , and each other column of Ai is the same as the
corresponding one of A.
Our formula now follows since the determinant is linear in the
columns of A from theorem 5.5. 
For our example we get det(A) = 12 − 60 + 48 = 0, which we
know is the case as the rows are dependent: twice the second
row minus the third row gives the first row.
Note: you should avoid using the cofactor expansion to
calculate determinants except for the 2 × 2 case.
Always try to do some row (or column) operations to find the
determinant.

MATH2601 Slides, 2021 – p. 32


One last use of the cofactors is an alternative way of
calculating inverse:
Definition 5.6 For A ∈ Mp,p (F) the adjugate (also called the
classical adjoint and adjunct) of A, adj(A) is the transpose
of the matrix of cofactors:

adj(A)ij = cji

adj(A)
Theorem 5.14 For an invertible matrix A, A−1 = .
det(A)
Proof: Suppose that Aj is the same matrix as A but with
column j replaced by column i, for i 6= j .
Then det(Aj ) = 0 by theorem 5.4 (c), but expanding det(Aj ) by
Xp
column i in theorem 5.13 gives det(Aj ) = aki ckj = 0 .
k=1

MATH2601 Slides, 2021 – p. 33


Putting this together with theorem 5.13 we find
p
X
aki ckj = δij det(A) .
k=1

p
X
So that the (j, i) entry of adj(A)A is ckj aki = δij det(A).
k=1
So that adj(A) A = (det(A))I . 
In the 2 × 2 case this gives the familiar formula
! !
a b 1 d −b
= .
c d ad − bc −c a

This is the only case where we use the adjugate-inverse


formula.
MATH2601 Slides, 2021 – p. 34

You might also like