Theorem 2.16.2 says that every permutation can be expressed as a product of transpositions.
A permutation is odd if it can be expressed as a product of an odd number of transpositions and even if it can be expressed as a product of an even number of transpositions.
(Sometimes people refer to the parity of a permutation to mean whether it is odd or even. We won’t do this since we want to save the word parity for integers.)
is odd.
is even.
is even.
The expression for an -cycle as a product of transpositions
in Lemma 2.16.1 shows that an cycle is even if is odd and odd if is even.
It seems possible that a permutation could be odd AND even at the same time, but this isn’t the case.
No permutation is both odd and even.
To prove this we need to do a little work.
For any and any distinct numbers we have
This will be a problem set exercise. ∎
We know every permutation can be written as a product of disjoint cycles. We are now going to be interested in how many cycles it takes to express a given permutation, and we will include all 1-cycles in our count. For example, in the identity has four cycles, the transposition has three cycles, the permutation
has two cycles as does , and the permutation has one cycle only.
If has cycles and is a transposition then has or cycles.
Let be a disjoint cycle decomposition for , remembering that we include any 1-cycles in this product so that each number between 1 and appears in exactly one of these cycles.
Let . There are two possibilities: either and belong to the same , or they belong to different s.
In the first case, because disjoint cycles commute we can assume and belong to , which we can write as for some distinct numbers and . Lemma 2.17.2 then shows us that
has cycles.
In the second case, because disjoint cycles commute we can assume belongs to and to . Write the disjoint cycles as and as . Then multiplying Lemma 2.17.2 on the left by gives
(2.4) |
and so
has cycles. ∎
Let’s consider two examples to illustrate the two cases in this proof. Take , , , , , and
so the number of cycles in is equal to 3. We are in the first case of the proof since 1 and 2 both belong to the same cycle from . You can check that
so that
has cycles.
Next take , and , , , and
so the number of cycles in is again 3. We are in the second case of the proof since 1 and 2 belong to and . Rewriting as and using the identity (2.4),
(you should check this by computing the left hand side). It follows
disjoint cycles commute | ||||
and has cycles.
The parity of an integer is whether it is even or odd. Two integers are said to have the same parity if they are either both even or odd, otherwise they are said to have the opposite parity.
Let be a product of transpositions, each of which belongs to . If is even then the number of cycles in has the same parity as , and if is odd then the number of cycles in has the opposite parity to .
For example, let be the odd number 3. In the product of two transpositions is equal to which has two cycles. The parity of the number of cycles is opposite to the parity of , and the same is true of any product of transpositions in for any odd . Now let be the even number 4. In the product of four transpositions is equal to which has two cycles. The parity of the number of cycles is the same as the parity of , and the same is true of any product of transpositions in for any even .
We will do the case when is even, the odd case being similar. We prove by induction on that the number of cycles in has the same parity as . The base case is when the product is just which has cycles (the 2-cycle from and then one-cycles), an odd number of cycles.
For example, if then might be
which has five cycles.
For the inductive step, consider a product . If is even then is odd, so by the inductive hypothesis has an odd number of cycles, and by Lemma 2.17.3 has or cycles, which in either case is an even number. If is odd then is even, so by the inductive hypothesis has an even number of cycles and then by the same Lemma as before has cycles, an odd number. ∎
Finally we can prove the main theorem, that no permutation is both odd and even.
Suppose is even and we can write as a product of transpositions, and also as a product of transpositions. Lemma 2.17.4 shows that both and has the same parity as the number of cycles in , in particular, and have the same parity. The argument for odd is similar. ∎
The sign of a permutation is if it is even and if it is odd.
We write for the sign of the permutation .
So if can be written as a product of transpositions, .
For any two permutations and ,
If can be written as a product of transpositions and can be written as a product of transpositions, then can be written as a product of transpositions. So
∎
This rule about the sign of a product means that
an even permutation times an even permutation is even,
an even permutation times an odd permutation is odd, and
an odd permutation times an odd permutation is even
just like when we multiply odd and even integers.
Even length cycles are odd and odd length cycles are even.
If is any permutation, .
We saw in the proof of Lemma 2.16.1 that if is any -cycle,
so an -cycle can be written as a product of transpositions. The number of transpositions in this expression therefore has the opposite parity to , as required.
If is a product of transpositions, . But the inverse of a transposition is a transposition, so is also the product of transpositions.
∎
Another way to express the first part of this lemma would be to say that .