Motivation
A radical equation suggests a Galois symmetry: once one root is known, multiplying it by an th root of unity gives another. This symmetry produces a cyclic Galois group only under the right hypotheses. The base field must contain the relevant roots of unity, the characteristic must not divide , and the degree of the chosen radical must be checked. With those hypotheses, Hilbert 90 also gives a converse: every cyclic extension of degree is generated by an th root of a base-field element.
Let be a field, let be an integer not divisible by the characteristic of , and suppose contains a primitive th root of unity . Thus has multiplicative order exactly . Since has degree and the powers are distinct, they are all of its roots in an algebraic closure. For take .
From an irreducible radical to a cyclic group
Theorem 6.5.1
(An irreducible radical gives a cyclic extension)
Let , and suppose is irreducible over . If , then is a cyclic Galois extension of degree . Its automorphisms are
Proof
(Every root is a root-of-unity multiple of )
Irreducibility gives . Every root of in an algebraic closure is : the quotient of any root by is an th root of unity. All roots therefore belong to . The derivative has no common root with , since in and . Thus the polynomial is separable, and is its splitting field. It follows that is Galois.
For each , the one-root embedding theorem gives a -embedding of sending to . Its image is , so it is an automorphism. These maps are distinct. The map with has order because its th power sends to . Hence the Galois group is cyclic of order .
Worked example 6.5.2
(A cyclic cubic over a rational-function field)
Let and let . The polynomial is Eisenstein at in , so it is irreducible over . The field contains a primitive cube root . Therefore is Galois of degree three, generated by . The base field is not : including the roots of unity is part of the example's hypothesis.
Common mistake
(The polynomial's exponent is not automatically the extension degree)
An element satisfying may have smaller minimal polynomial. The condition alone does not imply , even when the roots of unity are in . Irreducibility, or an independent degree calculation, must be supplied.
Worked example 6.5.3
(A reducible fourth-power equation)
Take , and . The field contains the primitive fourth root , but
is reducible over . The root satisfies , yet has degree two. Its automorphism group cannot have order four. The polynomial's exponent records an equation, not the degree of every root.
From a cyclic extension to a radical
Theorem 6.5.4
(Cyclic extensions are radical when roots of unity are present)
Suppose is a cyclic Galois extension of degree , generated by , and contains a primitive th root . Then there is with
Consequently the minimal polynomial of is .
Proof
(A Hilbert 90 eigenvector)
The base element has norm , since every automorphism fixes it. Multiplicative Hilbert 90 gives a nonzero such that , or .
The element is fixed by , because . Since generates the Galois group, its fixed field is , so . Its conjugates under the powers of are
and they are distinct because has order and . All are roots of the minimal polynomial of over , so that polynomial has degree at least . The tower law gives , hence equality and . The monic polynomial has degree and vanishes at , so it is the minimal polynomial.
Proof X-ray
How a norm-one root of unity becomes a radical
The proof turns the root of unity into a radical in three moves. First, lies in , so every automorphism fixes it and its norm is ; this is where the root-of-unity hypothesis enters the existence step. Second, Hilbert 90, which needs cyclic with generator , converts norm one into an eigenvector: a nonzero with . Third, the eigenvalue controls everything else. The power is fixed because , and it lies in because generates the whole group. The conjugates are distinct because has order exactly , so the degree count forces . If the eigenvalue had a smaller order , the same element would satisfy and could generate only a proper subfield.
Worked example 6.5.5
(A quadratic extension)
Let have characteristic different from two and let be quadratic Galois. Here . Hilbert 90 supplies with , so and . If , one can take . This is the degree-two case of the general cyclic result, not an assumption that every finite Galois group is cyclic.
The role of the hypotheses
Proposition 6.5.6
(Irreducibility can be replaced by the degree)
Under the same root-of-unity and characteristic hypotheses, if and , then is irreducible and the extension is cyclic Galois of degree .
Proof
(The minimal polynomial has full degree)
The minimal polynomial of divides . It is monic and has degree , so it equals . The first theorem now applies. The degree condition is a sufficient substitute for irreducibility; the relation alone is not.
Counterexample mode
Irreducibility alone does not give a Galois extension
Claim: the first theorem holds without the root-of-unity hypothesis, so irreducibility of alone makes cyclic Galois. Over , the polynomial is irreducible, but its real root generates a nonnormal cubic field: the two nonreal roots are not in that real field. Thus the extension is not Galois. The base does not contain a primitive cube root of unity. After adjoining that root of unity, the splitting behavior changes, but its degree must still be calculated separately. Concretely, has degree but only the identity automorphism, because the only root of it contains is the real one. The repaired statement keeps the hypothesis that contains a primitive th root of unity.
Worked example 6.5.7
(Why the characteristic condition matters)
Let and . The polynomial is irreducible by Eisenstein at , but its derivative is zero and the extension is inseparable. In characteristic , , so there is no primitive th root of unity. This example violates the hypotheses and cannot be used to conclude that the extension is cyclic Galois.
The condition is also useful. For , is reducible and has a repeated root at zero; generates only the base field. The trivial degree-one case remains valid: generates and the Galois group has one element.
A useful prime-degree irreducibility test
For a prime exponent, the root-of-unity hypothesis turns the absence of a base-field root into an irreducibility test. Without that hypothesis, a polynomial of prime degree greater than three can factor without having a linear factor, so the reasoning needs the Galois structure.
Theorem 6.5.8
(Prime radical criterion)
Let be prime, let , and suppose contains a primitive th root of unity. For , the polynomial is irreducible over if and only if .
Proof
(Galois before irreducible)
If for , then is a root and the polynomial is reducible. Conversely, suppose is not an th power and choose a root in an algebraic closure. The field contains all roots , and is separable. Thus is Galois even before we know that the polynomial is irreducible.
Each automorphism sends to . The map sending the automorphism to is an injective homomorphism into the group of th roots of unity: lies in and is fixed, so the exponents add under composition. Its image has order dividing the prime . If the image were trivial, the Galois group would be trivial, so and would be an th power in , contrary to assumption. The image therefore has order . Hence , and the minimal polynomial of has degree and equals .
Proof X-ray
Why a prime exponent makes one missing root enough
The root of unity in does two jobs before irreducibility is known: it makes Galois, and it makes a homomorphism, because the multipliers are fixed and therefore add. The real idea is the target group: the th roots of unity have prime order, so the image is either trivial or everything. The hypothesis excludes only the trivial case. For composite the image can be a proper nontrivial subgroup, which is exactly what happens for below.
Worked example 6.5.9
(A cubic over )
The element has order three, so it is a primitive cube root of unity. The nonzero cubes in are and : the cyclic multiplicative group has order six, and cubing has image of size two. Since is not a cube, is irreducible over . Adjoining a root gives a cyclic Galois extension of degree three. The conclusion uses both the noncube test and the primitive-root hypothesis.
Worked example 6.5.10
(A composite exponent needs more care)
In , is not a fourth power: the exponent of in a nonzero rational function raised to the fourth power is divisible by four, while that of is two. Nevertheless
is reducible. Thus simply replacing the prime by a composite and testing whether is an th power would give a false criterion.
Intermediate fields in the cyclic radical case
Suppose is cyclic of degree with and . For each divisor of , put . Its order is , so its fixed field has degree over by the Galois correspondence.
Proposition 6.5.11
(The radical intermediate fields)
For every divisor ,
Every intermediate field arises this way.
Proof
(Fixed elements and a degree count)
The element is fixed by , because . Thus is contained in . Its conjugates under powers of are . The factor has order , so these give distinct roots of its minimal polynomial. Its degree is at least and at most , hence the fields are equal. A cyclic group has exactly one subgroup of each order dividing , so the Galois correspondence gives every intermediate field on this list.
Worked example 6.5.12
(A degree-six tower)
Over , let . Eisenstein at makes irreducible, and contains primitive sixth roots of unity. The intermediate field of degree two is , generated by a square root of ; the one of degree three is , generated by a cube root of . The degrees one and six give and . These are all four intermediate fields because six has exactly four positive divisors.
Changing the radical generator
If and , then and . Thus multiplying by an th power from the base field does not change the generated extension when the radical is rescaled accordingly. This observation gives a sufficient relation between presentations; it is not a complete classification of all radical generators without further hypotheses.
Worked example 6.5.13
(Two presentations of the same extension)
In , if and , then and . The automorphism sending to also sends to . The polynomials and therefore give different equations for the same embedded field extension.
Proposition 6.5.14
(Conjugates and eigenvalues of the generator)
In the cyclic radical construction of the converse theorem, where , the -basis consists of eigenvectors for as a -linear operator, with eigenvalues .
Proof
(Powers of diagonalize )
The basis follows from . For each ,
The eigenvalues are distinct because has order . This diagonalization uses the roots of unity inside the base field; it is not a general diagonalization theorem for arbitrary Galois automorphisms or for characteristic dividing the group order.
Powers of a radical generator
The same orbit calculation determines the degree of every power of . For any integer , let , with the usual convention . The conjugates of are , and the factor has order . They give exactly distinct values. In a finite Galois extension, the degree of the minimal polynomial equals the number of distinct conjugates, because every embedding extends to an automorphism of the normal extension. Therefore
In particular is another field generator exactly when . If is negative, is defined because and the same argument applies. This is a statement about powers of a chosen generator; it does not say every generator of is a scalar multiple of such a power.
For , the powers and have degrees three and two, while generates the full degree-six extension. This explicit degree formula recovers the intermediate-field calculations above and also makes clear why a nonzero power can fail to generate the original field over the given base.
Practice and solutions
Quick check
(Why are all roots of in under the root-of-unity hypothesis?)
Solution · Answer
Every root is for some , and both factors lie in the field.
Quick check
(What extra fact besides shows is irreducible?)
Solution · Answer
The degree condition forces the degree- polynomial to equal the minimal polynomial.
Quick check
(In the converse proof, why does have norm one?)
Solution · Answer
All Galois automorphisms fix , so its norm is .
Quick check
(Why are the conjugates distinct?)
Solution · Answer
Equality would imply after canceling , so .
Quick check
(Why does irreducibility of over not make its real-root field Galois?)
Solution · Answer
Its other two roots are nonreal and are absent from the real field; the extension is not normal.
Quick check
(What happens to in characteristic ?)
Solution · Answer
Its derivative is zero, so the irreducible polynomial is inseparable and does not give a cyclic Galois extension.
Quick check
(If for , how are and related?)
Solution · Answer
They are equal, since and lie in .
Quick check
(Which eigenvalue does have on in the constructed basis?)
Solution · Answer
The eigenvalue is .