08-401/The Fundamental Theorem

From Drorbn
Jump to: navigation, search
Announcements go here

The statement appearing here, which is a weak version of the full fundamental theorem of Galois theory, is taken from Gallian's book and is meant to match our discussion in class. The proof is taken from Hungerford's book, except modified to fit our notations and conventions and simplified as per our weakened requirements.

Here and everywhere below our base field F will be a field of characteristic 0.



Theorem. Let E be a splitting field over F. Then there is a bijective correspondence between the set \{K:E/K/F\} of intermediate field extensions K lying between F and E and the set \{H:H<\operatorname{Gal}(E/F)\} of subgroups H of the Galois group \operatorname{Gal}(E/F) of the original extension E/F:


The bijection is given by mapping every intermediate extension K to the subgroup \operatorname{Gal}(E/K) of elements in \operatorname{Gal}(E/F) that preserve K,

\Phi:\quad K\mapsto\operatorname{Gal}(E/K):=\{\phi:E\to E:\phi|_K=I\},

and reversely, by mapping every subgroup H of \operatorname{Gal}(E/F) to its fixed field E_H:

\Psi:\quad H\mapsto E_H:=\{x\in E:\forall h\in H,\ hx=x\}.

This correspondence has the following further properties:

  1. It is inclusion-reversing: if H_1\subset H_2 then E_{H_1}\supset E_{H_2} and if K_1\subset K_2 then \operatorname{Gal}(E/K_1)>\operatorname{Gal}(E/K_2).
  2. It is degree/index respecting: [E:K]=|\operatorname{Gal}(E/K)| and [K:F]=[\operatorname{Gal}(E/F):\operatorname{Gal}(E/K)].
  3. Splitting fields correspond to normal subgroups: If K in E/K/F is the splitting field of a polynomial in F[x] then \operatorname{Gal}(E/K) is normal in \operatorname{Gal}(E/F) and \operatorname{Gal}(K/F)\cong\operatorname{Gal}(E/F)/\operatorname{Gal}(E/K).
The Fundamental Theorem of Galois Theory, all in one.


The four lemmas below belong to earlier chapters but we skipped them in class (the last one was also skipped by Gallian).

Zeros of Irreducible Polynomials

Lemma 1. An irreducible polynomial over a field of characteristic 0 has no multiple roots.

Proof. See the proof of Theorem 20.6 on page 362 of Gallian's book. \Box

Uniqueness of Splitting Fields

Lemma 2. Let \phi:F_1\to F_2 be an isomorphism of fields, let f_1\in F_1[x] be a polynomial and let f_2=\phi(f_1), and let E_1 and E_2 be splitting fields for f_1 and f_2 over F_1 and F_2, respectively. Then there is an isomorphism \bar\phi:E_1\to E_2 (generally not unique) that extends \phi.

Proof. See the proof of Theorem 20.4 on page 360 of Gallian's book. \Box

The Primitive Element Theorem

The celebrated "Primitive Element Theorem" is just a lemma for us:

Lemma 3. Let a and b be algebraic elements of some extension E of F. Then there exists a single element c of E so that F(a,b)=F(c). (And so by induction, every finite extension of E is "simple", meaning, is generated by a single element, called "a primitive element" for that extension).

Proof. See the proof of Theorem 21.6 on page 375 of Gallian's book. \Box

Splitting Fields are Good at Splitting

Lemma 4. (Compare with Hungerford's Theorem 10.15 on page 355). If E is a splitting field of some polynomial f over F and some irreducible polynomial p\in F[x] has a root v in E, then p splits in E.

Proof. Let L be a splitting field of p over E. We need to show that if w is a root of p in L, then w\in E (so all the roots of p are in E and hence p splits in E). Consider the two extensions

E=E(v)/F(v) and E(w)/F(w).

The "smaller fields" F(v) and F(w) in these two extensions are isomorphic as they both arise by adding a root of the same irreducible polynomial (p) to the base field F. The "larger fields" E=E(v) and E(w) in these two extensions are both the splitting fields of the same polynomial (f) over the respective "small fields", as E/F is a splitting extension for f and we can use the sub-lemma below. Thus by the uniqueness of splitting extensions (lemma 2), the isomorphism between F(v) and F(w) extends to an isomorphism between E=E(v) and E(w), and in particular these two fields are isomorphic and so [E:F]=[E(v):F]=[E(w):F]. Since all the degrees involved are finite it follows from the last equality and from [E(w):F]=[E(w):E][E:F] that [E(w):E]=1 and therefore E(w)=E. Therefore w\in E. \Box

Sub-lemma. If E/F is a splitting extension of some polynomial f\in F[x] and z is an element of some larger extension L of E, then E(z)/F(z) is also a splitting extension of f.

Proof. Let u_1,\ldots,u_n be all the roots of f in E. Then they remain roots of f in E(z), and since f completely splits already in E, these are all the roots of f in E(z). So


and E(z) is obtained by adding all the roots of f to F(z). \Box

Proof of The Fundamental Theorem

The Bijection

Proof of \Psi\circ\Phi=I. More precisely, we need to show that if K is an intermediate field between E and F, then E_{\operatorname{Gal}(E/K)}=K. The inclusion E_{\operatorname{Gal}(E/K)}\supset K is easy, so we turn to prove the other inclusion. Let v\in E-K be an element of E which is not in K. We need to show that there is some automorphism \phi\in\operatorname{Gal}(E/K) for which \phi(v)\neq v; if such a \phi exists it follows that v\not\in E_{\operatorname{Gal}(E/K)} and this implies the other inclusion. So let p be the minimal polynomial of v over K. It is not of degree 1; if it was, we'd have that v\in K contradicting the choice of v. By lemma 4 and using the fact that E is a splitting extension, we know that p splits in E, so E contains all the roots of p. Over a field of characteristic 0 irreducible polynomials cannot have multiple roots (lemma 1) and hence p must have at least one other root; call it w. Since v and w have the same minimal polynomial over K, we know that K(v) and K(w) are isomorphic; furthermore, there is an isomorphism \phi_0:K(v)\to K(w) so that \phi_0|_K=I yet \phi_0(v)=w. But E is a splitting field of some polynomial f over F and hence also over K(v) and over K(w). By the uniqueness of splitting fields (lemma 2), the isomorphism \phi_0 can be extended to an isomorphism \phi:E\to E; i.e., to an automorphism of E. but then \phi|_K=\phi_0|_K=I so \phi\in\operatorname{Gal}(E/K), yet \phi(v)=w\neq v, as required. \Box

Proof of \Phi\circ\Psi=I. More precisely we need to show that if H<\operatorname{Gal}(E/F) is a subgroup of the Galois group of E over F, then H=\operatorname{Gal}(E/E_H). The inclusion H<\operatorname{Gal}(E/E_H) is easy. Note that H is finite since we've proven previously that Galois groups of finite extensions are finite and hence \operatorname{Gal}(E/F) is finite. We will prove the following sequence of inequalities:

|H|\leq|\operatorname{Gal}(E/E_H)|\leq [E:E_H]\leq |H|

This sequence and the finiteness of |H| imply that these quantities are all equal and since H<\operatorname{Gal}(E/E_H) it follows that H=\operatorname{Gal}(E/E_H) as required.

The first inequality above follows immediately from the inclusion H<\operatorname{Gal}(E/E_H).

By the Primitive Element Theorem (Lemma 3) we know that there is some element u\in E so that E=E_H(u). Let p be the minimal polynomial of u over E_H. Distinct elements of \operatorname{Gal}(E/E_H) map u to distinct roots of p, but p has exactly \deg p roots. Hence |\operatorname{Gal}(E/E_H)|\leq\deg p=[E:E_H], proving the second inequality above.

Let \sigma_1,\ldots,\sigma_n be an enumeration of all the elements of H, let u_i:=\sigma_iu (with u as above), and let f be the polynomial


Clearly, f\in E[x]. Furthermore, if \tau\in H, then left multiplication by \tau permutes the \sigma_i's (this is always true in groups), and hence the sequence (\tau u_i=\tau\sigma u_i)_{i=1}^n is a permutation of the sequence (u_i)_{i=1}^n, hence

\tau f=\prod_{i=1}^n(x-\tau u_i)=\prod_{i=1}^n(x-u_i)=f,

and hence f\in E_H[x]. Clearly f(u)=0, so p|f, so [E:E_H]=\deg p\leq \deg f=n=|H|, proving the third inequality above. \Box

The Properties

Property 1. If H_1\subset H_2 then E_{H_1}\supset E_{H_2} and if K_1\subset K_2 then \operatorname{Gal}(E/K_1)>\operatorname{Gal}(E/K_1).

Proof of Property 1. Easy. \Box

Property 2. [E:K]=|\operatorname{Gal}(E/K)| and [K:F]=[\operatorname{Gal}(E/F):\operatorname{Gal}(E/K)].

Proof of Property 2. If K=E_H, then |\operatorname{Gal}(E/K)|=|\operatorname{Gal}(E/E_H)|=[E:E_H]=[E:K] as was shown within the proof of \Phi\circ\Psi=I. But every K is E_H for some H, so |\operatorname{Gal}(E/K)|=[E:K] for every K between E and F. The second equality follows from the first and from the multiplicativity of the degree/order/index in towers of extensions and in towers of groups:

[K:F] = \frac{[E:F]}{[E:K]} = \frac{|\operatorname{Gal}(E/F)|}{|\operatorname{Gal}(E/K)|} = [\operatorname{Gal}(E/F):\operatorname{Gal}(E/K)].\quad\Box

Property 3. If K in E/K/F is the splitting field of a polynomial in F[x] then \operatorname{Gal}(E/K) is normal in \operatorname{Gal}(E/F) and \operatorname{Gal}(K/F)\cong\operatorname{Gal}(E/F)/\operatorname{Gal}(E/K).

Proof of Property 3. We will define a surjective (onto) group homomorphism \rho:\operatorname{Gal}(E/F)\to\operatorname{Gal}(K/F) whose kernel is \operatorname{Gal}(E/K). This shows that \operatorname{Gal}(E/K) is normal in \operatorname{Gal}(E/F) (kernels of homomorphisms are always normal) and then by the first isomorphism theorem for groups, we'll have that \operatorname{Gal}(K/F)\cong\operatorname{Gal}(E/F)/\operatorname{Gal}(E/K).

Let \sigma be in \operatorname{Gal}(E/F) and let u be an element of K. Let p be the minimal polynomial of u in F[x]. Since K is a splitting field, lemma 4 implies that p splits in K[x], and hence all the other roots of p are also in K. As \sigma(u) is a root of p, it follows that \sigma(u)\in K and hence \sigma(K)\subset K. But since \sigma is an isomorphism, [\sigma(K):F]=[K:F] and hence \sigma(K)=K. Hence the restriction \sigma|_K of \sigma to K is an automorphism of K, so we can define \rho(\sigma)=\sigma|_K.

Clearly, \rho is a group homomorphism. The kernel of \rho is those automorphisms of E whose restriction to K is the identity. That is, it is \operatorname{Gal}(E/K). Finally, as E/F is a splitting extension, so is E/K. So every automorphism of K extends to an automorphism of E by the uniqueness statement for splitting extensions (lemma 2). But this means that \rho is onto. \Box