Motivation
The radical criterion connects a field question to a group question: in characteristic zero, a separable polynomial is solvable by radicals exactly when the Galois group of its splitting field is solvable. The companion chapter proves that criterion, and another constructs a quintic with Galois group . One group-theoretic step was left as background: why is not solvable?
We will supply that step without assuming the simplicity of every alternating group. The small group already contains the obstruction. Its conjugacy classes are few enough to count, yet their sizes leave no room for a proper nontrivial normal subgroup. That simplicity makes equal to its own commutator subgroup, so a derived series cannot reach the identity. The argument then applies to for every because it contains a copy of .
Definitions
For a group , the commutator of is . The commutator subgroup is generated by all commutators. It is normal: conjugating a commutator by gives . The quotient is abelian because each commutator becomes the identity. A group is solvable when the derived sequence
eventually reaches the identity subgroup. This agrees with the normal-series definition used in the radical criterion; we prove the equivalence below.
A group is simple if its only normal subgroups are the identity and the whole group. A nonabelian simple group is not solvable, but simplicity alone does not imply nonsolvability: the cyclic group of prime order is simple and abelian. We will check nonabelianness explicitly.
The alternating group is the subgroup of even permutations in . The sign map is onto because a transposition is odd, so . Cycle notation is composed from right to left. Conjugation relabels the entries of each cycle and preserves cycle type.
Derived series and normal series
Proposition 7.4.1
(Equivalent solvability tests)
A group has a finite normal series with abelian successive quotients if and only if its derived series eventually reaches the identity. Every subgroup of a solvable group is solvable.
Proof
(Commutators fall into the next term)
If the derived series reaches the identity, it is itself a normal series with abelian quotients: each commutator subgroup is normal in the preceding group. Conversely, suppose
has abelian quotients. Since is abelian, every commutator from lies in . Induction gives , so .
For a subgroup , every commutator of two elements of is a commutator of elements of , hence . Inductively for all . If the latter series reaches one, so does the former. The subgroup need not be normal in for this argument.
Worked example 7.4.2
(An abelian group reaches one immediately)
The cyclic group is abelian, so every commutator is the identity and . Its derived sequence terminates in one step. It is also simple because its order is prime. This example separates simple from nonabelian simple.
Counting the conjugacy classes of
Proposition 7.4.3
(The conjugacy-class sizes)
The five conjugacy classes of have sizes .
Proof
(Centralizers in , then in )
The even cycle types in are: the identity; a product of two disjoint transpositions; a 3-cycle; and a 5-cycle. A 4-cycle and a single transposition are odd; a disjoint 3-cycle and transposition are odd. Counting gives respectively
elements. For instance, choose the fixed letter in five ways and one of the three pairings of the remaining four to obtain fifteen double transpositions. Choose three letters and one of their two orientations for twenty 3-cycles. The number of 5-cycles is .
The size of the conjugacy class of in is , by orbit–stabilizer for conjugation. To calculate each centralizer, first work in and then retain its even elements.
A 3-cycle has centralizer of size six in : an element commuting with it is one of its three powers on the moving letters and independently may swap the other two letters. The latter swap is odd and commutes with the cycle, so exactly half of the centralizer is even. Its centralizer in has size three and its conjugacy class has size twenty. Thus all twenty 3-cycles form one class.
A double transposition has one fixed letter, which a commuting permutation must preserve. On the other four letters the commuting permutation may swap within either of the two pairs and may exchange the pairs; these choices give eight elements. One swap within a single pair is odd, so exactly four of the eight are even. The class size is , and all fifteen double transpositions form one class.
The centralizer of a 5-cycle in consists of its five powers. To see this, a commuting permutation is determined by where it sends one letter, and that image may be any of the five letters. Every power of a 5-cycle is even, so the centralizer in still has size five. Each 5-cycle class in therefore has size twelve. There are twenty-four 5-cycles, giving two distinct classes of size twelve. We do not need to name which cycles are in which class.
Proof X-ray
Why the 5-cycles split
The conjugacy class contains all 24 five-cycles, but the centralizer contains no odd permutation. Passing from to its index-two even subgroup leaves a centralizer of size five and halves each conjugacy orbit. Thus the 24 elements split into two classes of 12; silently treating them as one class would spoil the normal-subgroup count.
Why normality is the decisive condition
A subgroup need not be a union of whole conjugacy classes of the ambient group. For example, contains the two nonidentity powers of one 3-cycle but not the other eighteen 3-cycles. It has order three, so it is a proper subgroup of ; it is not normal because the twenty 3-cycles form one ambient conjugacy class. This is consistent with simplicity. Counting class sizes rules out normal subgroups because conjugation by every group element must preserve them; it cannot rule out arbitrary subgroups.
The two classes of 5-cycles also illustrate why centralizers are useful. The full symmetric group can use odd relabelings as well as even ones. An odd relabeling may move a 5-cycle to the other orbit, so being conjugate in does not by itself imply being conjugate in . Computing orbit sizes inside the actual group avoids that mistake.
Concept lensGeometric
The class sizes as rotations of an icosahedron
A standard fact, not proved here, is that is isomorphic to the rotation group of the regular icosahedron. The class sizes then count rotations. The fifteen axes through opposite edge midpoints give fifteen half-turns; the ten axes through opposite face centers give twenty rotations by ; the six axes through opposite vertices give twenty-four rotations by multiples of . Conjugation preserves the angle of a rotation, measured between and , so the rotations by and those by form two classes of twelve: the geometric form of the split computed above.
Simplicity and nonsolvability
Theorem 7.4.4
(The group is nonabelian simple)
has no proper nontrivial normal subgroup and is nonabelian.
Proof
(Normal subgroups are unions of classes)
Let . If it contains an element, normality forces it to contain the element's entire conjugacy class. Thus is one plus a sum of some of . If is nontrivial, its order is at least thirteen. Lagrange's theorem says that divides 60; the only possible orders from thirteen through sixty are .
No sum of the displayed form equals 15 or 20: a single nonidentity class gives 13, 16 or 21, and two or more nonidentity classes already give at least . Nor can a sum equal 30: after subtracting the identity, no selection of totals 29. Hence and . This proves simplicity.
The 3-cycles and do not commute. For example, using right-to-left composition, sends to , while sends to . Both permutations are even, so is nonabelian.
Theorem 7.4.5
(The symmetric group is not solvable)
For every integer , is not solvable.
Proof
( is its own derived subgroup and embeds in )
Because is nonabelian, some commutator is nonidentity and is nontrivial. It is normal in , so simplicity forces . Therefore every term of the derived sequence is and the sequence never reaches one. Thus is not solvable.
For , permute the first five symbols and fix the remaining symbols. This embeds as a subgroup of . If were solvable, the subgroup proposition would make solvable, a contradiction. This proves the claim without requiring a proof that is simple for every larger .
Nonabelianness was used at the first step, to make nontrivial; simplicity alone would not suffice.
Counterexample mode
A simple group can be solvable
Claim: every simple group is nonsolvable. Take the cyclic group . By Lagrange's theorem its only subgroups have orders and , so it is simple. It is abelian, so every commutator is trivial and : the derived series reaches the identity after one step, and is solvable. The repaired statement, used above, is that a nonabelian simple group is nonsolvable, because then is a nontrivial normal subgroup and must equal .
Worked examples
Worked example 7.4.6
(Why a normal subgroup cannot have order thirty)
A normal subgroup containing a 5-cycle contains one entire twelve-element 5-cycle class, as well as the identity. Its order is at least thirteen. To have order thirty, its nonidentity part would have size twenty-nine, but no combination of complete nonidentity class sizes sums to twenty-nine. The argument concerns classes under , not arbitrary individual elements.
Worked example 7.4.7
(The explicit quintic dependency)
The companion cubic-and-quintic unit proves that over has Galois group . The radical-criterion unit proves the characteristic-zero criterion. Our theorem supplies the missing group fact: is not solvable. Taken together, these results show that this particular quintic is not solvable by radicals over . The conclusion uses both companion proofs; this unit alone does not recompute the splitting field or prove the radical criterion.
Worked example 7.4.8
(A solvable quintic is not a contradiction)
The polynomial splits after adjoining a fifth root of two and a primitive fifth root of unity, both radical adjunctions. Its degree is also five. The nonsolvability theorem concerns a group, not every polynomial of degree five. Different quintics can have different Galois groups and different solvability behavior.
Worked example 7.4.9
(A larger symmetric group)
Inside , let the copy of act on and fix . If had a terminating derived series, its subgroup would have one too. Since , this is impossible. We did not need to enumerate the conjugacy classes of or .
Common mistakes
Common mistake
(Simple does not mean without proper subgroups)
Simplicity does not mean that a group has no proper subgroups, only no proper nontrivial normal subgroups. The group contains the proper subgroup of order three, which is not normal because the other eighteen 3-cycles are conjugate to but lie outside it.
Common mistake
(Conjugate in is not conjugate in )
The class count must be done in itself. All twenty-four 5-cycles are conjugate in , but in they form two classes of twelve, because the centralizer of a 5-cycle contains no odd permutation. Treating them as one class of twenty-four would give a wrong list of class sizes, and the class-sum argument for simplicity is valid only with the true classes.
A subgroup of a solvable group is solvable even when it is not normal; the derived-series inclusion proves this directly. Finally, one nonsolvable quintic disproves a proposed radical formula that works for every quintic by specialization, but a formal theorem about a generic polynomial and generic coefficients requires its own precise setting and is not proved here.
Summary
Conjugacy-class sizes force to be simple. Its noncommuting elements make its commutator subgroup nontrivial, hence all of . This perfect subgroup sits inside every for , preventing those symmetric groups from being solvable. Coupled with the separately proved radical criterion and explicit quintic, this gives a concrete nonsolvable equation.
Exercises
Quick check
(What is , and why?)
Solution · Answer
It has order because sign maps onto the two-element sign group.
Quick check
(How many 3-cycles and how many double transpositions are there on five letters?)
Solution · Answer
Choose three of the five letters and one of two cyclic orientations, giving . Choose the fixed letter in five ways and pair the other four in three ways, giving 15.
Quick check
(What are the conjugacy-class sizes?)
Solution · Answer
Their sizes are .
Quick check
(Why must a normal subgroup of be a union of conjugacy classes, and why can it not have order thirty?)
Solution · Answer
Normality says for every . Its nonidentity elements would be a union of classes totaling 29, impossible using sizes .
Quick check
(Why is a normal subgroup, and what is it?)
Solution · Answer
Conjugating by an element sends it to another commutator; the subgroup generated by all such elements is invariant. It is a nontrivial normal subgroup, hence equals all of .
Quick check
(Why does simplicity alone not prove nonsolvability?)
Solution · Answer
A cyclic group of prime order is simple but abelian, so its derived subgroup is trivial.
Quick check
(Why does imply is nonsolvable for ?)
Solution · Answer
Every subgroup of a solvable group is solvable, while the embedded copy of is not.
Quick check
(Does a degree-five polynomial necessarily have Galois group ?)
Solution · Answer
No. For example, is solvable by radicals and its group is solvable as shown in the companion unit.
Exercise 7.4.10
(Which separate results are needed to deduce that is not solvable by radicals?)
Core
Solution(3)
- 1
Identify the Galois group
The cubic-and-quintic unit computes the Galois group of over as .
- 2
Use nonsolvability of
Key step
This unit proves that is not solvable, since it contains the nonabelian simple group .
- 3
Apply the radical criterion
The characteristic-zero radical criterion then shows that is not solvable by radicals over .
Because: In characteristic zero, solvability by radicals is equivalent to solvability of the Galois group.