This matrix then operates on a column vector as a
permutation
0 1 0 0 0 a1 a2
0 0 0 1 0 a2 a4
1 0 0 0 0 a3 = a1
0 0 0 0 1 a4 a5
Permutations and determinants 0 0 1 0 0 a5 a3
Math 130 Linear Algebra
D Joyce, Fall 2015 Thus, the permutation matrix permutes the rows
of another matrix. The row 1 is replaced by row 2,
row 2 by row 1, row 3 by row 4, row 4 by row 5,
One way to construct determinants is in terms and row 5 by row 3.
of permutations. That construction depends on a
particular property of permutations, namely, their
parity. Preview of permutations and determinants.
When we construct the determinant of a square n×
n matrix, which we’ll do in a moment, it will be
Permutations. A permutation of the set defined as a sum/difference of n! terms, each term
{1, 2, 3, . . . , n} is a list of its n elements where being a product of n elements, one element chosen
each element appears exactly once in the list. For out of each row and column.
instance, the six permutations of the set {1, 2, 3} Our first question is: why are there n! ways
are 123, 132, 213, 231, 312, and 321. In general, to choose one element out of each row and col-
there are n! permutations of a set of size n. This umn? Each choice is determined by which col-
expression n!, read n factorial, is the product of umn to choose for each row. So, if the element
the integers from 1 through n. a1σ1 is chosen for the first row, a2σ2 is chosen for
the second row, . . ., and anσn is chosen for the nth
A permutation σ is often written as a list, such as
row, that choice is determined by the permutation
σ = 24153, but you can also treat it as a function
σ = σ1 σ2 . . . σn , that is, a permutation of the set
on the set. This permutation σ is a function on the
{1, 2, . . . , n}. So, there are n! choices, and each
set {1, 2, 3, 4, 5}, where σ sends 1 to 2, 2 to 4, 3 to
corresponds to a permutation.
1, 4 to 5, and 5 to 3. We’ll write σ1 = 2, σ2 = 4,
σ3 = 1, σ4 = 5, and σ5 = 3. We’ll form all n! products of n elements, one el-
ement chosen out of each row and column. Half
of these n! we’ll add, the other half we’ll subtract.
Permutation matrices. One way to look at a The result will be the determinant. We’ll add those
permutation is to treat it as a matrix itself. First, that correspond to “even permutations” and sub-
think of the permutation as an operation rather tract those that correspond to “odd permutations”.
than a list. For instance, associate to the permuta- Before we can do that, we’ll have to define what it
tion σ = 24153 the following 5 × 5 matrix means for a permutation to be even or odd.
Parity of a permutation. A transposition is a
0 1 0 0 0
0 0 0 1 0 particularly simple permutation. It exchanges ex-
1
0 0 0 0 changes exactly two elements and leaves all the oth-
0 0 0 0 1 ers fixed. Some transpositions of the set of 5 ele-
0 0 1 0 0 ments, for example, are 21345, 12543, and 52341,
1
but the permutation 23145 isn’t a transposition Theorem 1. A permutation cannot be expressed
since 1, 2, and 3 are all moved. as both the composition of an even number of trans-
Each permutation can be expressed as a product positions and an odd number of transpositions.
of transpositions in many ways. One nice way to
visualize a permutation is by drawing lines connect- Proof. Suppose a permutation σ can be expressed
ing the initial list of numbers from 1 through n to as a composition of transpositions in two ways σ =
their final positions. For example, the permutation τ1 · · · τm = ρ1 · · · ρn where each τi and ρi is a trans-
−1
σ = 24153 can be drawn as position. Then σ −1 = (ρ1 · · · ρn )−1 = ρ−1 n · · · ρ1 ,
and since the inverse of any transposition is itself,
therefore σ −1 = ρn · · · ρ1 . Therefore the identity
1 XXX * 1
permutation is the product of m + n transposi-
XXX
2 H
X X 2
z tions, τ1 · · · τm ρn · · · ρ1 . In the following lemma,
H
HHH we’ll show that that identity permutation can only
3 3
H be expressed as a composition of an even number of
*
H
4 XXX 4 j
H
transpositions. That will imply that m + n is even.
XXX
5
X X 5
z Hence m and n have the same parity. q.e.d.
Lemma 2. The identity permutation can only be
In this figure, there are 4 arrow crossings. Each
expressed as the composition of an even number of
one can be interpreted as a transposition, and you
transpositions.
can use that interpretation as a way to represent
the permutation as a composition of transpositions. Proof. Consider a diagram that expresses the iden-
Here it’s shown as the product of these four trans- tity as a composition of transpositions.
positions: (23), (45), (12), and (34).
1 -1 -1H *1 -1 1 - - - - - - - - - - 1
H @ @
H
2H *2
-2 j2
H -2 2 R
@ - - @
R - - - 2
H
H @ @ @ @ @
3 j3
H -3 -3
HH*3 3 R
@ - - - @
R R
@ R
@ - @
R 3
H @ @ @
4 -4H 4 -4 j4
H 4 - - R
@ - - @
R - @
R - - 4
H
*
@ @
5 - 5 HH
j5 -5 -5 5 - - @
R - - - - - - - @
R - 5
There are many other ways that this permuta- Each strand in the diagram starts at one number
tion can be represented as composition of trans- and ends at the same number.
positions. Just take any drawing where you draw Consider the strands labelled 2 and 3 in the di-
curves going to the right from the numbers 1, 2, agram, drawn in red and green, respectively. The
3, 4, and 5 to their correct targets. (You’ll need red strand starts out above the green. Each time
to make sure that all the crossings are only double the crosses the green strand, it switches from above
crossings and that there are only finitely many of to below, or from below to above. Since it ends up
crossings.) The number of crossings will vary, but above the green where it started, there have to be
we’ll the parity of the number of crossings doesn’t an even number of times that the red strand crosses
change. We expressed this example as a composi- the green strand.
tion of 4 transpositions, so however it’s expressed, Likewise, for any other pair of numbers i and
it will have to involve an even number of transpo- j, the number of times the i strand crosses the j
sitions. strand has to be even.
2
Therefore, the total number of crossings has to dicated with +, odd with −.
be even. That says that the identity transformation
can only be expressed as the composition of an even 1234 + 2134 − 3124 + 4123 −
number of transpositions. q.e.d. 1243 − 2143 + 3142 − 4132 +
1324 − 2314 + 3214 − 4213 +
We can now define the parity of a permutation 1342 + 2341 − 3241 + 4231 −
σ to be either even if its the product of an even 1423 + 2413 − 3412 + 4312 −
number of transpositions or odd if its the product 1432 − 2431 + 3421 − 4321 +
of an odd number of transpositions. The sign of
Each entry in the above table gives one term in
σ, denoted sgn σ, is defined to be 1 if σ is an even
the determinant of A. Thus, reading down the first
permutation, and −1 if σ is an odd permutation.
column, we see that the determinant starts out with
the following six terms:
Construction of the determinant. The de- +a11 a22 a33 a44 − a11 a22 a34 a43 − a11 a23 a32 a44
terminant of a square n × n matrix A is sum of +a11 a23 a34 a42 + a11 a24 a32 a43 − a11 a24 a33 a42
n! terms, one for each permutation σ of the set But besides these, there are 18 more terms.
{1, . . . n}, where each term is
Math 130 Home Page at
sgn σ A1σ1 · · · Anσn http://math.clarku.edu/~ma130/
The entries A1σ1 , . . . , Anσn are one taken one from
each row i and each column σi , and sgn σ is the sign
of the permutation σ. Symbolically,
X
|A| = sgn σA1σ1 · · · Anσn
σ
This definition usually is used to compute deter-
minants when n is small, 2 or 3, and it agrees with
what we did above. But when n is 4 or greater,
there are so many terms that it isn’t practical to
use the definition to compute the value of a deter-
minant. There are much faster ways of computing
determinants.
The determinant of a 4 × 4 matrix. Let’s take
a generic matrix.
a11 a12 a13 a14
a21 a22 a23 a24
A=
a31
a32 a33 a34
a41 a42 a43 a44
Look at all 4! = 24 permutations of the set
{1, 2, 3, 4} and their parities. Even parities are in-
