A subgroup can be normal in one place but not another
Let be a subgroup of . Conjugation by an element of sends to the subgroup .
Definition 6.1.1
(Normalizer)
The normalizer of a subgroup is
This set records exactly which elements preserve the entire subgroup under conjugation. It is not necessary that such an element fix every member of separately.
Why the normalizer is a subgroup
Theorem 6.1.2
(The normalizer is the largest subgroup normalizing )
For a subgroup , the set is a subgroup of that contains , and is normal in . If is a subgroup with , then . In particular, is normal in exactly when .
Proof
(Closure, containment and maximality)
The identity normalizes . If , then
Also, from , conjugating both sides by gives . Thus inverses remain in the set, and is a subgroup. Every satisfies by closure of , so ; and is normal in by the definition of the latter. Finally, if and , then every satisfies , hence . Taking gives the last statement.
Proof X-ray
Where the definition by equality is used
The key move in closure is that is an equality of sets, so it may be substituted inside the outer conjugation by . The definition by equality also makes inverses free: conjugating by returns , whereas a definition by the inclusion would need a separate argument for . Maximality needs no computation at all: " is normal in " means exactly for all , which is the membership test for applied to each . No finiteness is used anywhere.
Common mistake
(Preserving as a set is not fixing its elements)
The equality allows conjugation to permute the elements of . Requiring for every would instead describe the centralizer of , generally a smaller subgroup. Normality is a setwise condition.
Quick check
(Does every element of belong to ?)
Use closure of under products and inverses.
Solution · Answer
Yes; closure and inverses inside give for .
Quick check
(If , what is ?)
Compare the definition of normality with that of the normalizer.
Solution · Answer
All of , by the definition of normality.
A cyclic subgroup inside
Worked example 6.1.3
(A cyclic subgroup of order 6 in )
Let and . The two disjoint cycles commute, so the generator has order 6. Its cube is and its fourth power is ; contains the products of powers of these two cycles. Its only element of order 2 is , and its two elements of order 3 are the two nonidentity powers of .
If normalizes , conjugation preserves orders, so it sends the unique order-2 element to itself. Therefore it preserves , and hence its complement . Conversely, any permutation preserving both blocks fixes under conjugation and sends to one of its two nonidentity powers. It consequently sends the generators of into and, by applying the same argument to its inverse, normalizes . Thus
For example, normalizes but conjugates to its inverse, so it does not centralize every element of . The normalizer is proper: sends to , which is not in . Thus is normal in its normalizer, but not in all of .
Counterexample mode
One preserving element does not prove normality in the whole group
The transposition above preserves setwise. That single successful conjugation does not make normal in : fails to preserve it. Normality in the ambient group requires preservation by every ambient element.
Worked example 6.1.4
(A subgroup of order 2: normalizer equals centralizer)
In let . Conjugation fixes and permutes the nonidentity elements of , and there is only one, so exactly when commutes with . Hence is the centralizer, of order , because the class of has 3 elements. The number of conjugates of is therefore : namely , and . Since the number is not 1, is not normal in , although it lies in the normal Klein four-subgroup.
Worked example 6.1.5
(A 3-cycle subgroup in )
Let in . Conjugating by gives , which lies in only if it equals or ; that happens exactly when preserves , and then . So consists of the six permutations of . It is larger than , which has order 3, and is normal in it; there are conjugates of .
Quick check
(In the example, does belong to the normalizer?)
Check that it preserves the blocks and .
Solution · Answer
Yes; it preserves both blocks, and conjugation reverses the 3-cycle: .
Quick check
(Must an element of the normalizer fix every element of individually?)
Look at the effect of on .
Solution · Answer
No; sends the 3-cycle to its distinct inverse , yet it normalizes .