Extended Field Trip

Having explored the realms of groups and rings, we can now explore fields!

For any group G with identity 0 and set K \subseteq G, define K^* := K \backslash \{0 \}.

Definition 1. A field is a ring \mathbb F with multiplicative identity 1 such that (\mathbb F, +) and (\mathbb F^*, \cdot) form Abelian groups. Given two fields \mathbb F, \mathbb K, we call \mathbb K a field extension of \mathbb F if \mathbb F \subseteq \mathbb K.

Our discussion on rings built up to our discussion on polynomials, because irreducible polynomials help us generate new fields. Let f \in \mathbb F[x] denote an irreducible polynomial.

Example 1. The ring \mathbb F[x]/\langle f \rangle forms a field extension of \mathbb F. In particular, \mathbb C \cong \mathbb R[x]/\langle x^2 + 1 \rangle is a field extension of \mathbb R. Denote i \equiv x + \langle x^2 + 1 \rangle.

Example 2. Slightly more generally, for any prime number p, \mathbb Q[x]/\langle x^2 - p \rangle is a field extension of \mathbb Q. Denote \sqrt{p} \equiv x + \langle x^2 - p \rangle.

Observe that in both examples, the root \alpha := x + \langle f \rangle satisfies f(\alpha) = 0. Therefore, \mathbb F[x]/\langle f \rangle contains both \mathbb F and \alpha. Fix \mathbb F \subseteq \mathbb K as fields.

Definition 2. For any K \subseteq \mathbb K \backslash \mathbb F, denote the smallest sub-field of \mathbb K containing \mathbb F and K by \mathbb F(K), so that \mathbb F \subseteq \mathbb F(K) \subseteq \mathbb K.

  • In the case K = \{\alpha\}, denote \mathbb F(\alpha) := \mathbb F(\{\alpha\}), which we call a simple extension of \mathbb F.

Theorem 1. \mathbb F[x]/\langle f \rangle \cong \mathbb F(\alpha).

Proof. Define the map \phi : \mathbb F[x] \to \mathbb F(\alpha) by \phi(g) = g(\alpha). We remark that \ker(\phi) = \langle f \rangle, and by construction,

\phi(\mathbb F[x]) \subseteq \mathbb F(\alpha).

By the first isomorphism theorem for rings, \phi(\mathbb F[x]) \cong \mathbb F[x]/\langle f \rangle is a field that contains \mathbb F \cup \{\alpha\}. By Definition 2,

\mathbb F(\alpha) \subseteq \phi(\mathbb F[x]).

Therefore, \mathbb F[x]/\langle f \rangle \cong \phi(\mathbb F[x]) = \mathbb F(\alpha).

Since \alpha is constructed as a root of f in the field extension \mathbb K = \mathbb F[x]/\langle f \rangle, we call \mathbb F(\alpha) an algebraic extension of \mathbb F.

Definition 2. An element \alpha \in \mathbb K is algebraic (over \mathbb F) if there exists a polynomial f \in \mathbb F[x] such that f(\alpha) = 0.

  • The element \alpha is transcendental over \mathbb F if it is not algebraic over \mathbb F.

A field extension \mathbb K \supseteq \mathbb F is algebraic if every element in \mathbb K is algebraic over \mathbb F.

Example 3. Using Examples 1 and 2, \mathbb C = \mathbb R(i) \supseteq \mathbb R and \mathbb Q(\sqrt p) \supseteq \mathbb Q are an algebraic extensions.

Theorem 2. For any \alpha \in \mathbb K, exactly one of the following holds.

  • If \alpha is algebraic, then there exists a unique monic irreducible polynomial f_{\mathbb F}^{\alpha} \in \mathbb F[x] such that f_{\mathbb F}^{\alpha}(\alpha) = 0 and \mathbb F[x]/\langle f_{\mathbb F}^{\alpha} \rangle \cong \mathbb F(\alpha).
  • If \alpha is transcendental, then \mathbb F(\alpha) \cong Q(\mathbb F[x]), where the right-hand side denotes the fraction field of \mathbb F[x].

Proof. Define the evaluation map \phi : \mathbb F[x] \to \mathbb F(u) by \phi(g) = g(\alpha), as per Theorem 1, which can be shown to be a ring homomorphism.

Suppose \alpha is algebraic. Then there exists a monic polynomial f \in \mathbb F[x] such that f(\alpha) = 0. Since \mathbb F[x] is a UFD, write f as a product of monic irreducible polynomials. If \alpha is not a root of any of these polynomials, then f(\alpha) \neq 0, a contradiction. Therefore, there exists a monic irreducible polynomial f_{\mathbb F}^{\alpha} \in \mathbb F[x] such that f_{\mathbb F}^{\alpha}(\alpha) = 0. By the argument in Theorem 1, \mathbb F[x] / \langle f_{\mathbb F}^{\alpha} \rangle \cong \mathbb F(\alpha) as fields. Furthermore, if there exists another monic irreducible polynomial g \in \mathbb F[x] such that g(\alpha) = 0, then

\mathbb F[x] / \langle f_{\mathbb F}^{\alpha} \rangle \cong \mathbb F(\alpha) \cong \mathbb F[x] / \langle g \rangle

implies that \langle f_{\mathbb F}^{\alpha} \rangle = \langle g \rangle. Since both polynomials are monic and irreducible, we must have f_{\mathbb F}^{\alpha} = g, establishing uniqueness.

Suppose \alpha is transcendental. Then for any f \in \mathbb F[x], f(\alpha) \neq 0. Therefore, we can extend \phi to act on \phi : Q(\mathbb F[x]) \to \mathbb F(\alpha) by defining

\phi(f/g) := f(\alpha)/g(\alpha).

The observation

\mathbb F \cup \{\alpha \} \subseteq \phi(Q(\mathbb F[x])) \subseteq \mathbb F(\alpha)

implies that \phi(Q(\mathbb F[x])) = \mathbb F(\alpha). However, note that \ker(\phi) = \{ 0 \}, so that \phi is injective. Therefore, \phi is a bijective ring homomorphism, so that

Q(\mathbb F[x])  \cong \phi( Q(\mathbb F[x]) ) = \mathbb F(\alpha).

Definition 3. We can regard field extensions \mathbb K \supseteq \mathbb F as vector spaces over the field \mathbb F. Define the degree of the extension by [\mathbb K : \mathbb F] := \dim_{\mathbb F}(\mathbb K). If this number is finite, we say that \mathbb K is a finite extension of \mathbb F.

Theorem 3. If \alpha \in \mathbb K is algebraic over \mathbb F, then [\mathbb F(\alpha) : \mathbb F] = \deg(f_{\mathbb F}^{\alpha}), so that \mathbb F(\alpha) \supseteq \mathbb F is a finite extension.

Proof. Denote n:= \deg(f_{\mathbb F}^\alpha). We claim that K := \{1,\alpha,\dots, \alpha^{n-1}\} form a basis for \mathbb F(\alpha). For linear independence, the existence of nonzero scalars c_k \in \mathbb F such that

\displaystyle \sum_{k=0}^{n-1} c_k \alpha^k = 0

imply the existence of a polynomial g(x) := \sum_{k=0}^{n-1} c_k x^k \in \mathbb F[x] such that g(\alpha) = 0. By Theorem 2, g \in \langle f_{\mathbb F^{\alpha}} \rangle so that

n-1 = \deg(g) \geq \deg(f_{\mathbb F}^{\alpha}) = n,

a contradiction. Therefore, K must be linearly independent. For the spanning set, write f_{\mathbb F}^{\alpha}(x) = \sum_{k=0}^{n-1} d_kx^k + x^n. Since f_{\mathbb F}^{\alpha}(\alpha) = 0, we must have

\displaystyle \alpha^n = \sum_{k=0}^{n-1} (-d_k)\alpha^k \in \mathrm{span}_{\mathbb F}(K).

Inductively, \alpha^m \in \mathrm{span}_{\mathbb F}(K) for m \geq n, since \alpha^{m-1} \in \mathrm{span}_{\mathbb F}(K) implies

\begin{aligned} \alpha^m &= \alpha \cdot \alpha^{m-1} = \alpha \cdot \sum_{k=0}^{n-1} d_k \alpha^k \\ &= \sum_{k=1}^{n-1} d_{k-1} \alpha^{k}  + d_{n-1} \alpha^n  \in \mathrm{span}_{\mathbb F}(K).\end{aligned}

Then

\mathbb F( \alpha ) \cong \mathbb F[x]/\langle f_{\mathbb F}^{\alpha} \rangle \cong \mathrm{span}_{\mathbb F}(K).

Theorem 4. If \mathbb K \supseteq \mathbb F is a finite extension, then it is an algebraic extension.

Proof. Suppose \mathbb K \supseteq \mathbb F is a finite extension so that [K : F] =: n < \infty. Fix \alpha \in K. The set \{1,\alpha, \dots, \alpha^n\} has n+1 > n elements, and thus must be linearly dependent. Therefore, there exists scalars c_i \in \mathbb F, not all 0, such that

\displaystyle \sum_{i=0}^n c_i \alpha^i = 0.

Defining the polynomial f := \sum_{i=0}^n c_i \mu_i \in \mathbb F[x], we have f(\alpha) = 0. Therefore, \alpha is algebraic.

Can we measure “stacks” or “towers” of field extensions? Of course!

Lemma 1. Every vector space V over \mathbb F has a basis.

Proof. Assume V is not finite-dimensional for the sake of non-triviality. Let \mathcal S denote the collection of linearly independent subsets of V, and partially order its elements via \subseteq. For any chain \mathcal C \subseteq \mathcal S, the linearly independent subset K := \bigcup_{C \in \mathcal C} C is an upper bound for every subset in \mathcal C. By Zorn’s lemma, \mathcal S contains a maximal element B, that is, a linearly independent subset of V. It suffices to check that V = \mathrm{span}(B). To that end, the existence of \mathbf v \in V\backslash \mathrm{span}(B) implies B \subsetneq B \cup \{\mathbf v \}, so that by the maximality of B, B \cup \{\mathbf v \} \subseteq B, a contradiction.

Remark 1. We establish detailed proofs for other domain-specific results out of necessity, rather than curiosity. This is an example of an applied mathematics motivation combined with pure mathematical techniques.

Theorem 5 (Tower Rule). Let \mathbb F \subseteq \mathbb K \subseteq \mathbb L be field extensions. Then \mathbb F \subseteq \mathbb L is finite if and only if \mathbb F \subseteq \mathbb K and \mathbb K \subseteq \mathbb L. In this case, we recover the tower rule:

[\mathbb L : \mathbb F] = [\mathbb L : \mathbb K] \cdot [\mathbb K : \mathbb F].

Proof. Using Lemma 1, let \mathcal U = \{u_i : i \in I\} \subseteq \mathbb L be a basis of \mathbb L over \mathbb K and \mathcal V = \{v_j : j \in J\} \subseteq \mathbb K be a basis of \mathbb K over \mathbb F. We claim that

\mathcal W := \{u_i v_j : (i,j) \in I \times J\}

forms a basis of \mathbb L over \mathbb F. For linear independence,

\displaystyle \sum_{(i,j) \in I \times J} c_{ij} u_i v_j = \sum_{j \in J} \left( \sum_{i \in I} c_{ij} u_i \right) v_j.

Since \mathcal U, \mathcal V are linearly independent,

\begin{aligned}  \sum_{(i,j) \in I \times J} c_{ij} u_i v_j = 0 \quad &\Rightarrow \quad \sum_{j \in J} \left( \sum_{i \in I} c_{ij} u_i \right) v_j = 0 \\ &\Rightarrow \quad  \sum_{i \in I} c_{ij} u_i = 0 \\ &\Rightarrow \quad  c_{ij} = 0.\end{aligned}

Therefore, \mathcal W is linearly independent.

To show that \mathbb L = \mathrm{span}_{\mathbb F}(\mathcal W), use \mathbb L = \mathrm{span}_{\mathbb K}(\mathcal U) and \mathbb K = \mathrm{span}_{\mathbb F}(\mathcal V) to construct c_i, d_{ij} such that for any w \in \mathbb L,

\begin{aligned} w = \sum_{i \in I} c_i u_i &= \sum_{i \in I} \left( \sum_{j \in J} d_{ij} v_j \right) u_i \\ &= \sum_{(i,j) \in I \times J} d_{ij} u_i v_j \in \mathrm{span}_{\mathbb F}(\mathcal W).\end{aligned}

We note, of course, that only finitely many terms here are nonzero, so that the sum is still finite, and hence, well-defined. Here, the finite-ness results hold, and in this case,

\begin{aligned} [\mathbb L : \mathbb F] &= |\mathcal W| = |I \times J| \\ &= |I| \cdot |J| \\ &= |\mathcal U| \cdot |\mathcal V| \\ &= [\mathbb L : \mathbb K] \cdot [\mathbb K : \mathbb F]. \end{aligned}

This “stacking” of field extensions is responsible for several impossibility results in constructions in Euclidean geometry.

For now, we want to recall the seemingly trivial field extension \mathbb R \subseteq \mathbb C. Recall that f_{ \mathbb R}^i(x) = x^2 + 1. By Theorem 3, [\mathbb C : \mathbb R] = \deg(f_{ \mathbb R}^i) = 2. Yet, the fundamental theorem of algebra also tells us that for any f \in \mathbb C[x], there exists at least one \alpha \in \mathbb C such that f(\alpha) = 0. In this case, we call \mathbb C algebraically closed.

It turns out that every field \mathbb F has an (effectively unique) algebraic extension \overline{\mathbb F} that is algebraically closed. By effectively unique, we mean that for any algebraic extension \mathbb K \supseteq \mathbb F that is algebraically closed, \mathbb K \cong \overline{\mathbb F} as fields. Using this uniqueness, we can conclude that \overline{\mathbb R} = \mathbb C.

We will explore this notion of algebraic closure next time.

—Joel Kindiak, 5 Feb 26, 1334H

,

Published by


Leave a comment