Motivation
An irreducible quartic has four conjugate roots, but that fact alone does not tell us how its splitting-field automorphisms permute them. A direct attack might require writing four complicated radicals. A more useful question is: what happens to the three ways of splitting four roots into two unordered pairs? That smaller permutation action is encoded by a cubic polynomial whose coefficients are computed from the quartic.
The resolvent is not magic: it deliberately forgets some information. In particular, a cyclic quartic and a dihedral quartic can produce the same factorization pattern for their resolvents. Our goal is to prove what the resolvent detects, see exactly what it loses, and calculate concrete examples. This section extends the earlier cubic-discriminant test and the explicit splitting field.
Definitions and hypotheses
Let be a field of characteristic different from two. Let
be irreducible and separable, with distinct roots in its splitting field . Irreducibility makes the Galois group transitive on the four roots. Separability makes finite Galois and allows us to identify faithfully with a subgroup of . The hypotheses are separate: in positive characteristic an irreducible polynomial need not be separable.
There are three partitions into unordered pairs:
For them define, respectively,
The cubic resolvent is . A permutation of the roots permutes the three , so its coefficients are fixed by and lie in . This fixed-field argument is useful even before we expand the coefficients.
The resolvent and its discriminant
Theorem 6.3.1
(Resolvent formula and distinct roots)
For the quartic above,
The three are distinct, and .
Proof
(Symmetric functions and root differences)
To check the formula, use the elementary symmetric functions , , , . Expanding the three sums gives
and . These are polynomial identities, so no division or special choice of roots is involved.
The difference of two resolvent roots factors into two differences of quartic roots; for example,
The analogous factorizations for the other two differences involve the remaining four root differences, each once up to sign. No factor vanishes because is separable, proving that is separable. Squaring and multiplying the three difference identities gives precisely the product of the squares of all six quartic root differences. This is the stated equality of discriminants.
Concept lensStructural
What the cubic sees
The three values record the three pair partitions. Their difference identities ensure that distinct partitions really give distinct values, so the cubic cannot accidentally identify two partitions. The splitting field of the cubic is the fixed field of the kernel of the partition action; a nontrivial kernel is exactly the information that the cubic forgets.
The permutation action
Let be the action on the three partitions. The kernel is
Indeed, each listed permutation fixes every partition. Conversely, a permutation fixing all three partitions is determined by the image of , and the four displayed permutations realize the four possible images. Thus the kernel has exactly four elements. Since , the image is all of .
The kernel lies in . Also, a transposition of two root labels induces a transposition of two partition labels. As transpositions generate , the sign of equals the sign of . Consequently the discriminant sign test applied to the quartic agrees with the sign test for the resolvent. Characteristic different from two is essential for a sign change of a nonzero square root of the discriminant to be visible.
Theorem 6.3.2
(Reading the transitive quartic group)
Under the stated hypotheses:
- If is irreducible, then when is a square in , and otherwise.
- If has one root in and an irreducible quadratic factor, then is either a cyclic group or a dihedral group of order eight .
- If splits completely over , then .
The second case requires additional information to distinguish the two groups.
Proof
(Kernel, image and transitivity)
First, the factor degrees of separable are the orbit lengths for the induced action on its roots. This follows from the extension-of-embeddings theorem for a finite Galois splitting field: conjugates of a root are precisely its orbit. Thus irreducibility means the image is transitive on three letters, a subgroup of order three or six. In either case its order is divisible by three.
The intersection is normal in . Conjugation by an element whose image is a 3-cycle permutes the three nonidentity elements of cyclically. Hence is either trivial or all of : an intermediate subgroup of order two would not be invariant. Transitivity of on four roots makes divide , while . If were trivial, would be three or six, a contradiction. Therefore . The image has order three or six, so is twelve or twenty-four. In the former case the image is , and parity shows ; equal orders give . In the latter case . The discriminant square test chooses between them.
If has one linear and one irreducible quadratic factor, its image has order two. The subgroup of preserving the corresponding pair partition has order eight: we may exchange the two pairs and independently swap their elements. This subgroup is the symmetry group of a square, . Thus lies in and, by four-root transitivity, has order four or eight. If its order is eight it is all of . If its order is four, the action is regular. A regular subgroup of order four is or the Klein group; in the latter case all three nonidentity elements act without fixed points, hence are the three double transpositions, giving . But has trivial partition action, contrary to the image of order two. Therefore the order-four case is .
Finally, if splits, the image is trivial, so . A transitive action on four roots requires divisible by four; since , equality follows. The factorization pattern and discriminant therefore give an efficient logical classification, provided irreducibility, separability and the relevant field calculations have actually been established.
Proof X-ray
How the kernel V does the work
All three cases run on one equation and one divisibility. The equation is , from the restriction of to , and the factor pattern of fixes as or , , or . The divisibility is , which is where irreducibility of the quartic enters, through transitivity on four roots. In the irreducible case a -cycle in the image permutes the three nonidentity elements of , so the normal subgroup is or , and excludes . In the case lies in the stabilizer of the fixed partition; order four forces a regular action, and the Klein possibility is excluded because it would be , whose image under is trivial. In the split case and give equality. The discriminant is used only at the very end, to choose between and .
Worked examples
Worked example 6.3.3
(A full symmetric group)
Take over . Its reduction modulo two is . It has no root in , and the only monic irreducible quadratic over is , whose square is . Thus the reduction, and hence , is irreducible. Characteristic zero gives separability. The resolvent is . Its only possible rational roots are , neither of which works, so it is irreducible. Its discriminant is , not a rational square. Hence .
Worked example 6.3.4
(A biquadratic splitting field)
Let . Its four roots are . They are distinct and belong to , whose four sign-change automorphisms act transitively on them. Therefore is irreducible and its splitting field has group . The resolvent calculation confirms the classification:
All three pair partitions are fixed even though individual roots are moved by nontrivial automorphisms.
Worked example 6.3.5
(The dihedral ambiguity)
For , Eisenstein at two gives irreducibility and the resolvent is . The quadratic is irreducible over , so the theorem gives or . The companion quartic unit explicitly proves that the splitting field has degree eight and group . The cubic factorization alone did not supply that last step.
Worked example 6.3.6
(The same resolvent pattern with a cyclic group)
Let . The substitution gives , which is Eisenstein at five. Its roots are the primitive fifth roots of unity. If is one root, each for is another, so the irreducible degree-four extension already splits . Its automorphisms are and their group is , cyclic of order four because has order four modulo five. Here
again one linear factor and an irreducible quadratic. This concrete comparison proves that the resolvent factorization pattern cannot distinguish from .
How to use the test in a new problem
Start with the quartic, not the resolvent. Prove irreducibility over the stated base field and check separability; for instance, compute the gcd of and its formal derivative. Next form from its coefficients and factor it over the same field. Since is a separable cubic, its possible factor-degree patterns are exactly , , and . For the first pattern, determine whether its discriminant is a square. The equality of discriminants lets us use whichever polynomial makes that calculation shorter. For the last pattern, no square test is needed: the kernel and transitivity already force .
When the factorization is , keep the unresolved choice visible. One possible next step is to find the degree of the quartic splitting field: degree four gives and degree eight gives . That instruction is a consequence of the group calculation, not a promise that the degree is easy to compute. In particular, adjoining one quartic root always gives a degree-four field, but that field need not contain the other three roots. The example shows the distinction: the real field generated by its positive root is not its full splitting field, which also contains .
Common mistakes
Common mistake
(A square discriminant gives containment, not equality)
A square discriminant alone implies containment in , not necessarily equality: the biquadratic example has group . The factorization of the resolvent must be read first.
Common mistake
(A rational resolvent root is not a rational quartic root)
A rational resolvent root means that a pair partition is fixed; it does not mean that any individual quartic root is rational. For the resolvent has the root , yet is irreducible over and has no rational root.
In characteristic two the root-difference formula still makes sense, but the sign argument used here does not distinguish even from odd permutations. This chapter makes no characteristic-two classification claim. Irreducibility of the quartic is needed to infer transitivity and eliminate smaller subgroups, as the following example shows.
Counterexample mode
A reducible quartic breaks the 1 + 2 case
Claim: whenever factors as a linear factor times an irreducible quadratic, is or . Take over , which is separable but reducible. Here , , , so , and is irreducible because is not a rational square. With , , , , the roots are and . The splitting field is , of degree , and its group is a Klein four group that is not transitive on the roots, neither nor . The repair is the standing hypothesis that is irreducible.
Summary
The cubic resolvent turns the action on four roots into an action on three pair partitions, with kernel . Its factorization records the image, while the discriminant records parity. Together with transitivity they identify , and , and narrow the remaining case to or . The residual ambiguity is a structural consequence of passing to a quotient, not a computational accident.
Exercises
Quick check
(List the three pair partitions of four labels.)
Solution · Answer
They are , and .
Quick check
(Why are the resolvent roots distinct when the quartic roots are distinct?)
Solution · Answer
Each difference of two resolvent roots is, up to sign, a product of two nonzero differences of quartic roots.
Quick check
(What is the kernel of the partition action of , and why is the action onto ?)
Solution · Answer
It is . The kernel has four elements, so the image has elements, all of .
Quick check
(Why is a quartic Galois group transitive when the quartic is irreducible?)
Solution · Answer
Every root is a conjugate of any other; the finite Galois extension-of-embeddings theorem supplies an automorphism carrying one to the other.
Quick check
(An irreducible quartic has irreducible resolvent. What is its group if the discriminant is a square, and what if it is not?)
Solution · Answer
If the discriminant is a square the group is ; otherwise it is .
Quick check
(What does a completely split resolvent imply?)
Solution · Answer
Its group is , because it lies in the order-four kernel and must act transitively on four roots.
Quick check
(Explain why a single rational resolvent root does not force a rational quartic root.)
Solution · Answer
It fixes a partition as an unordered pair of unordered pairs; it can still permute every individual root. The quartic is an example.
Quick check
(Why is the sign test restricted to characteristic different from two?)
Solution · Answer
In characteristic two, , so the discriminant square root cannot record the sign of a permutation.
Exercise 6.3.7
(Calculate the resolvent of and state what its factorization alone proves.)
Core
Solution(3)
- 1
Read off the coefficients
Here and , so .
- 2
Factor over
The quadratic has no rational root, so factors as .
Because: is not a rational square.
- 3
Apply the classification
The factorization alone proves only that is or . A separate splitting-field computation, , gives .
Exercise 6.3.8
(What extra fact distinguishes the and fifth-cyclotomic examples?)
Core
Solution(2)
- 1
Compare the resolvent patterns
Both resolvents factor as , so in each case the group is or ; the factorization cannot separate them.
- 2
Use the splitting-field degrees
Their splitting fields have degrees eight and four respectively, giving group orders eight and four, hence and . The cyclic example also explicitly exhibits a generator of order four.