Conjugation asks how an element looks after a reversible change of labels. In a permutation group it relabels cycles, but not every relabeling produces a different permutation. A centralizer records the relabelings that leave an element unchanged; its cosets count the distinct results. We use rightmost-first composition throughout.
Why a centralizer is a subgroup
Definition 8.1.1
(Centralizer and conjugacy class)
For , the centralizer is . The conjugacy class of is .
The centralizer contains . If commute with , then also commutes with , and so does ; hence the subgroup test gives . Its elements need not commute with each other: centralizing is a condition on each element relative to , not a claim that the centralizer is abelian.
Theorem 8.1.2
(Class size equals the index of the centralizer)
Let be a group and . The map from the set of left cosets to is a well-defined bijection. If is finite,
Proof
(A bijection from cosets onto the class)
Here is treated only as a set of cosets. The map is well defined: if with , then . It is injective: equality implies , so and . It is onto by the definition of . For finite , Lagrange's theorem gives the displayed formula.
Proof X-ray
Where each hypothesis enters
Well-definedness uses only , and it works because the correction sits on the right of : a change of representative is a change inside a left coset. Injectivity is the same computation read backwards: the key move is to turn equality of two conjugates into the commutation relation , which says exactly that lies in the centralizer. Surjectivity costs nothing, since the class is defined as the image. The argument closes with Lagrange's counting, so finiteness of is used only in the last line. Normality of is never used, because we count cosets and never multiply them.
Quick check
(Show that if and only if lies in the center .)
Apply the theorem and ask when an index equals 1.
Solution · Answer
By the theorem, , and an index is 1 exactly when . That means for every , that is, . (For the class is .)
Common mistake
(Counting conjugators instead of conjugates)
The map on all of is not injective: each element of the class is produced by exactly different . For in the 24 elements give only 8 distinct conjugates, each obtained 3 times. The class size is , not .
Cycle type and worked examples
Proposition 8.1.3
(Conjugation relabels cycles)
For a cycle and ,
Consequently two permutations in are conjugate if and only if they have the same multiset of cycle lengths, fixed points included.
Proof
(Relabel each cycle by )
Apply the left side to : returns , then advances it to and relabels the result as . Points outside the relabeled cycle are fixed. The same argument applies to every disjoint cycle of a general permutation, so conjugation preserves cycle lengths. Conversely, if two permutations have the same multiset of cycle lengths, choose a bijection sending the entries of each cycle of the first to the corresponding entries of a matching cycle of the second, and match the fixed points; this bijection conjugates the first into the second.
Worked example 8.1.4
(A 3-cycle in )
Take . Any commuting with must send the unique fixed point 4 to a fixed point of , so . The choice of among 1, 2, 3 determines and . These choices give exactly , which do commute with . Thus and the class has elements. Directly, there are four choices of the three moved points and two cyclic orientations on each, again giving 8. The two calculations check different aspects of the result.
Worked example 8.1.5
(A product of two transpositions in )
Let . Its class consists of the permutations of type : , and , so and . The centralizer is . As a check, sends to , so it conjugates to . The centralizer is larger than , which has only 2 elements.
Worked example 8.1.6
(A transposition in (fixed points matter))
Let in . The class is the set of all transpositions, so and . Directly, commutes with exactly when maps the set to itself, so either fixes or swaps 1 and 2 (2 choices) and permutes the fixed points arbitrarily ( choices); this gives . The three fixed points are what make the centralizer large.
Counterexample mode
A centralizer need not be abelian
Claim: since every element of commutes with , the centralizer is abelian. Take in . Then and both lie in , because they are disjoint from . But, composing rightmost first, while , and these differ. So is not abelian. The correct statement is only that lies in the center of its own centralizer.
Quick check
(How many elements does the conjugacy class of a transposition in have, and what is the order of its centralizer?)
Count the transpositions, then use the theorem.
Solution · Answer
Conjugation preserves cycle type and all transpositions are conjugate, so the class is the set of transpositions. The centralizer has order ; it is for .
Conjugation also acts on subgroups
Definition 8.1.7
(Normalizer)
For a subgroup , is again a subgroup. The normalizer of is .
Proposition 8.1.8
(Number of conjugate subgroups)
If is finite and , the number of distinct subgroups is . In particular is normal precisely when this number is 1.
Proof
(The coset argument again)
The map is well defined: if with , then . It is injective: if , then , so . It is onto the set of conjugates by definition, and counting cosets gives the index.
Common mistake
(Normalizing a subgroup is not centralizing its generator)
Do not confuse with the centralizer of one generator: an element can carry a generator to a different generator of the same subgroup without commuting with it. In , is normal because it has index two. The transposition conjugates to its inverse, so it normalizes but does not centralize that 3-cycle. The subgroup and the individual element answer different questions.
Quick check
(In let . Find and the number of conjugates of .)
Test whether conjugates back into .
Solution · Answer
Conjugating by gives , so and its square are not in the normalizer, while and are. Hence , of index 3, and there are three conjugates: , and . Since the number is not 1, is not normal.