Motivation
Splitting fields make the permutations of roots visible, while fixed fields turn those permutations into concrete intermediate extensions. We begin with explicit constructions and a coordinate calculation, then use a non-abelian cubic to see why normality matters. Finally we compare quintics and prove two further tests. The preceding units supply splitting-field degree bounds, the finite Galois correspondence and the radical-solvability criterion; the latter is proved in the radical-criterion unit.
Splitting fields in practice
The following computations show how the degree bound behaves, and how to certify a splitting field in concrete cases.
Worked example 6.2.1
(The splitting field of over the rationals has degree 6)
Let and let , so that and . The three roots of are , , , since each cubes to and while are distinct. The polynomial is irreducible over by Eisenstein's criterion at the prime , so . The splitting field is : it contains all three roots, and conversely any field containing all three roots contains and the quotient .
Next we find . The number is a root of , so its minimal polynomial over has degree at most . It is not , because while . Hence and, by the tower law,
So the bound of the degree proposition is attained. By the proposition on finite normal extensions, is normal, and it is separable by characteristic zero. The extension sitting inside it is not normal, as shown in the normality unit.
Worked example 6.2.2
(A degree-four polynomial whose splitting field has degree two)
Let over . We claim . Expanding,
and reducing coefficients modulo gives . Each quadratic factor has no root in : for the values at are , none zero, and the second factor is obtained from the first by , so its values at are the values of the first at , again nonzero. A quadratic with no root in the field is irreducible.
Let be a root of , so . The roots of are and (their sum is ), and the roots of are the negatives and . All four lie in , and is generated by , so it is a splitting field of of degree . Here , and the actual degree is only . Finally is separable: and because . The four roots are therefore distinct.
Computing a fixed field in coordinates
The degree-count shortcut is efficient, but it is reassuring to see the same answer arise from raw linear algebra, without invoking the fundamental theorem.
Worked example 6.2.3
(Fixed field of a single sign change)
Let , which has degree over by the biquadratic theorem in the correspondence unit. The four products span over , because is spanned by the monomials with . Four spanning vectors in a four-dimensional space form a basis. Let be the automorphism with and . Since is multiplicative, . For rational ,
This equals the original element exactly when , and by linear independence of the basis this happens exactly when . The fixed set is therefore , which is . This agrees with the dictionary: the subgroup has order , and its fixed field has degree over .
The same calculation with fixing and negating gives the fixed set , and with negating both roots it gives , since negates and but fixes . The three coordinate computations confirm the three quadratic subfields independently of any general theorem.
A non-abelian example: the splitting field of
The Klein four group is abelian, so every subgroup is normal and every intermediate field is Galois. To see the correspondence distinguish normal from non-normal subgroups, we need a group with non-normal subgroups. The smallest example is .
Worked example 6.2.4
(The splitting field of over )
Let , so that , and let be the real cube root of . The roots of are , , and , so the splitting field is , because the quotient of two roots gives . The polynomial is irreducible over by Eisenstein's criterion at the prime , so . The field lies in while is not real, so and its minimal polynomial over is of degree . By the tower law .
The extension is Galois in characteristic zero, so has order . By the root-permutation theorem in the correspondence unit, an automorphism is determined by its values on and , and since , it is determined by how it permutes the three roots of . Thus embeds in , and by order it is all of . In particular every one of the combinations of an image for and an image for occurs. Define by , (of order ), and by , (complex conjugation, of order ).
The subgroups of are the trivial group, three subgroups of order , the subgroup of order , and itself, so there are exactly six intermediate fields. The group fixes , and , so its fixed field is . Each subgroup of order is the stabilizer of one of the three roots; for example fixes , and since , its fixed field is . The other two order-two subgroups have fixed fields and , which are distinct from because the subgroups are distinct. The lattice of intermediate fields is therefore , , the three cubic fields , , , and .
The only nontrivial proper normal subgroup of is . Hence is the only proper nontrivial Galois intermediate extension, with group , consistent with the fact that is the splitting field of . The real field contains neither nonreal root. Its conjugate cubic fields likewise cannot contain all three roots, since an automorphism carries the real field and its omitted roots to each of them. Thus none is normal over .
This example explains why the correspondence cannot be strengthened to "every intermediate field is normal." It also shows how the degree formulas make the bookkeeping rigid: the three cubic fields have degree and the quadratic field has degree , as forced by the orders and of the corresponding subgroups.
The radical units prove that a separable polynomial in characteristic zero is solvable by radicals exactly when its splitting-field Galois group is solvable (the radical-criterion unit), and that is not solvable (the unit on symmetric groups). We use both results below. Cauchy's theorem and extension of embeddings are additional background for the displayed group calculation.
A solvable quintic and the limits of the criterion
Worked example 6.2.5
(The group of over the rationals)
Let be the real fifth root and a primitive fifth root of unity. The roots of are for , and since , the splitting field is .
First, by Eisenstein at . Second, , because is irreducible: substituting gives , which is Eisenstein at . Both and divide , so divides it. Conversely is a root of over , so and . Hence .
Let and , which has order and is therefore cyclic. Since is the splitting field of , it is Galois, so is normal in , and has order and is abelian, because each automorphism is determined by and these compose by multiplication of exponents. The chain has abelian quotients, so is solvable, in agreement with the explicit tower .
Concept lensStructural
The quintic group as affine maps modulo five
Label the root by . An automorphism with and sends to . Thus acts on the labels by the affine maps with and ; there are twenty such maps, matching . The normal subgroup consists of the translations (), and is the multiplier group .
By contrast, the polynomial is a quintic over with Galois group , as the following argument shows. Combined with the two dependencies above it gives a quintic that is not solvable by radicals. The two quintics together illustrate the point of the criterion: the degree does not decide solvability; the symmetry group of the roots does.
Proposition 6.2.6
(A quintic with group )
The polynomial is irreducible over , and the Galois group of its splitting field over is .
Proof
(Irreducibility, two nonreal roots, and generation)
Let . Eisenstein at makes it irreducible. Put . The derivative has exactly the two real zeros and , and while . Monotonicity and the end behavior therefore give exactly three real roots and two nonreal roots. Complex conjugation swaps the latter pair and fixes the others, giving a transposition in the Galois group.
Irreducibility makes the action on the five roots transitive, by the extension-of-embeddings theorem. Thus its group order is divisible by , and Cauchy's theorem, used here without proof, gives a -cycle. Conjugating the transposition by powers of that cycle gives transpositions along a connected five-vertex cycle: the step between the two exchanged vertices is nonzero modulo . Transpositions along the edges of a connected graph generate the full symmetric group. Hence the Galois group is .
Proof X-ray
Three independent inputs to the argument for the symmetric group of degree five
The proof combines three facts of different kinds. Algebra: Eisenstein irreducibility gives transitivity on the five roots, so divides the group order, and Cauchy's theorem converts that divisibility into a -cycle. Analysis: the derivative count shows that exactly two roots are nonreal, so complex conjugation, which always lies in the group, is a transposition rather than a product of two transpositions. Combinatorics: the conjugates of one transposition by the powers of a -cycle form a connected graph on the five roots, because the step between the two exchanged roots is a unit modulo the prime . A different count of real roots would break the second input: with only one real root, conjugation would be a product of two disjoint transpositions.
The primality of five is used in the last step, and it cannot be dropped from that step.
Counterexample mode
A cycle and a transposition need not generate the symmetric group
Claim: an -cycle and any transposition generate . Take , and . Conjugating by gives , and conjugating again returns , so the conjugates of form a disconnected graph with edges and . Indeed every element of preserves the partition : exchanges the two blocks and fixes each. The transposition sends to , so it is not in , which is the symmetry group of a square, of order rather than . For a prime , every transposition has a step that is a unit modulo , and the generation argument goes through.
A discriminant test for irreducible cubics
The cubic example above used a degree calculation to identify its group. There is another useful route: a square root of the discriminant records whether a permutation is even or odd. The test has hypotheses. Irreducibility supplies transitivity; separability supplies three distinct roots; characteristic different from two makes a sign change visible.
Theorem 6.2.7
(Discriminant and the cubic Galois group)
Let be an irreducible separable cubic over a field of characteristic different from , and let be its splitting field. Write its distinct roots in as , and put . Then . The Galois group is cyclic of order three if is a square in , and is isomorphic to otherwise.
Proof
(Signs of the root-difference product)
Every automorphism fixing permutes the roots, and an odd permutation reverses the sign of while an even permutation preserves it. This sign rule follows by checking a transposition; transpositions generate the symmetric group. Thus is fixed by the Galois group. The finite Galois correspondence identifies the fixed field of the whole group with , so .
The group action on the three roots is faithful because those roots generate the splitting field. It is transitive because irreducibility and extension of embeddings carry any one root to any other. Hence its order is divisible by three and divides six. It has order three or six; in the first case it is the alternating subgroup , and in the second it is all of .
All automorphisms fix exactly when the group is contained in . By the fixed-field statement, this occurs exactly when . Finally, exactly when is a square in : if satisfies , then in the field , so or . This proves both cases. Separability ensures ; in characteristic two the sign argument would not distinguish odd permutations.
For a depressed cubic , direct expansion of the squared product gives . This is a polynomial identity in the coefficients and can be checked by expanding the product of root differences and substituting the elementary symmetric functions. We use the identity in the examples; the group argument itself does not depend on a particular formula for the discriminant.
Worked example 6.2.8
(Two irreducible cubics with different groups)
For , Eisenstein at proves irreducibility and its discriminant is , not a square in . The test gives , matching the six-element calculation above. For , the only possible rational roots are and , and neither is a root; a cubic without a rational root is irreducible over . Its discriminant is , a rational square. Its group is therefore , cyclic of order three. The same degree of the polynomial can lead to different splitting-field degrees.
Common mistake
(The discriminant test needs an irreducible cubic)
For a reducible cubic the test gives wrong answers. Take over , with roots . Then and , which is not a square in ; yet the splitting field is and the Galois group has order , not . Without irreducibility the action is not transitive, and the group can be a proper subgroup of not containing .
A fifth root after adjoining roots of unity
The following result isolates a step used in the radical example. The presence of a primitive root of unity is essential to the stated splitting-field conclusion.
Proposition 6.2.9
(An irreducible fifth-root extension over a field containing the roots of unity)
Let have characteristic zero and contain a primitive fifth root of unity . If is irreducible over and , then is Galois with cyclic group of order five.
Proof
(Five roots, one generator)
Irreducibility gives . All roots are , so they lie in and generate it; they are distinct because characteristic is zero and . The extension is therefore the splitting field of a separable polynomial. The finite-Galois automorphism-count theorem gives five automorphisms. Each is determined by the image of , which must be one of those five roots; because is fixed, the multiplier indices add under composition. Thus the group embeds in the cyclic group of fifth roots of unity and has the same order, so it is cyclic of order five.
The irreducibility condition cannot be omitted. If , all fifth roots already lie in and adjoining one adds no extension at all. Nor does the proposition prove the full radical solvability criterion: it describes one well-controlled radical step after roots of unity have been supplied, not an arbitrary chain of extensions.
Practice and solutions
Quick check
(For , explain why the splitting-field degree is three rather than six.)
Solution · Answer
The polynomial is irreducible by the rational-root test and has discriminant , a square. The theorem gives Galois group of order three. Since the splitting extension is finite Galois, its degree equals the group order.
Quick check
(In the fifth-root proposition, what happens if and the base field already contains a primitive fifth root of unity?)
Solution · Answer
The polynomial is reducible, since is a root, so the proposition's irreducibility hypothesis fails. Every root is a power of the primitive fifth root already in the base; its splitting field is the base field itself, of degree one.
Quick check
(Explain why knowing only that a cubic has three distinct roots does not establish transitivity of its Galois action.)
Solution · Answer
Distinctness is separability, not irreducibility. For example, over has three distinct rational roots. Its splitting field is , and the trivial Galois group fixes each root separately rather than acting transitively.
Quick check
(In the splitting field of , let be the automorphism with and . Show that has order , that it fixes , and that the fixed field of is .)
Solution · Answer
Compute and . Hence , and because , so has order . Next, . So , giving . The element is a root of the irreducible polynomial , so . The fixed field has degree over . Two fields of equal degree with one contained in the other coincide.