Chapter 18
Rigorous

Field Extensions

00 · Symbol Glossary

$K/F$K is an extension of F

Read "KK over FF." Means FF is a subfield of KK. The slash is notation, not division.

$[K:F]$Degree of the extension

The dimension of KK as a vector space over FF (using FF's scalars, KK's vectors, and KK's own addition and multiplication-by-FF-elements). Finite or infinite.

$F(\alpha)$Simple extension generated by α

The smallest field containing both FF and α\alpha. Read "FF adjoin alpha."

$m_\alpha(x)$Minimal polynomial of α over F

The monic polynomial of least degree in F[x]F[x] having α\alpha as a root, when α\alpha is algebraic over FF.

$F[\alpha]$Subring generated by α

The smallest subring of KK containing FF and α\alpha — polynomial expressions in α\alpha with coefficients in FF, before knowing whether inverses are needed.


01 · Extensions and Degree

Definition — Field Extension and Degree

If FF is a subfield of KK, KK is an extension of FF, written K/FK/F. Since FF's scalars act on KK's elements via KK's own multiplication, KK is naturally a vector space over FF; its dimension is the degree [K:F][K:F].

Example — [C:R] = 2

Every complex number is uniquely a+bia+bi with a,bRa,b\in\mathbb{R} — exactly the statement that {1,i}\{1,i\} is a basis for C\mathbb{C} as an R\mathbb{R}-vector space. So [C:R]=2[\mathbb{C}:\mathbb{R}]=2.

Example — [R:Q] = ∞

R\mathbb{R} contains uncountably many elements but Q\mathbb{Q}-linear combinations of any finite list of reals form only a countable set — so no finite basis can span R\mathbb{R} over Q\mathbb{Q}. [R:Q]=[\mathbb{R}:\mathbb{Q}]=\infty.


02 · The Tower Law

Theorem 18.1 — Tower Law

If FKLF\subseteq K\subseteq L are fields with [K:F][K:F] and [L:K][L:K] both finite, then [L:F][L:F] is finite and:

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

Proof

Let {a1,,am}\{a_1,\ldots,a_m\} be a basis for KK over FF (m=[K:F]m=[K:F]) and {b1,,bn}\{b_1,\ldots,b_n\} a basis for LL over KK (n=[L:K]n=[L:K]). Claim: {aibj:1im,1jn}\{a_ib_j : 1\le i\le m,\,1\le j\le n\} (mnmn elements) is a basis for LL over FF.

Spanning. Any L\ell\in L is a KK-linear combination =jcjbj\ell = \sum_j c_jb_j (cjKc_j\in K). Each cjc_j, being in KK, is an FF-linear combination cj=idijaic_j=\sum_i d_{ij}a_i (dijFd_{ij}\in F). Substituting: =j(idijai)bj=i,jdij(aibj)\ell = \sum_j\big(\sum_i d_{ij}a_i\big)b_j = \sum_{i,j}d_{ij}(a_ib_j) — an FF-linear combination of the aibja_ib_j.

Independence. Suppose i,jdijaibj=0\sum_{i,j}d_{ij}a_ib_j=0 with dijFd_{ij}\in F. Regroup as j(idijai)bj=0\sum_j\big(\sum_i d_{ij}a_i\big)b_j=0. Each coefficient idijai\sum_i d_{ij}a_i lies in KK; since {bj}\{b_j\} is independent over KK, every one of these KK-coefficients must be 00: idijai=0\sum_i d_{ij}a_i=0 for each jj. Since {ai}\{a_i\} is independent over FF, this forces dij=0d_{ij}=0 for every ii (for each fixed jj). So all dij=0d_{ij}=0.

The mnmn products aibja_ib_j span LL and are FF-linearly independent: a basis. So [L:F]=mn=[L:K][K:F][L:F]=mn=[L:K][K:F].

Example — Applying the Tower Law

QQ(2)Q(2,3)\mathbb{Q}\subseteq\mathbb{Q}(\sqrt2)\subseteq\mathbb{Q}(\sqrt2,\sqrt3). If [Q(2):Q]=2[\mathbb{Q}(\sqrt2):\mathbb{Q}]=2 and [Q(2,3):Q(2)]=2[\mathbb{Q}(\sqrt2,\sqrt3):\mathbb{Q}(\sqrt2)]=2 (verified in Section 05), Theorem 18.1 gives [Q(2,3):Q]=2×2=4[\mathbb{Q}(\sqrt2,\sqrt3):\mathbb{Q}] = 2\times2=4 — instantly, without constructing a basis of the full four-dimensional space by hand.


03 · Algebraic Elements and Minimal Polynomials

Definition — Algebraic and Transcendental

Let K/FK/F be an extension. αK\alpha\in K is algebraic over FF if f(α)=0f(\alpha)=0 for some nonzero fF[x]f\in F[x]; otherwise α\alpha is transcendental over FF.

Definition — Minimal Polynomial

If α\alpha is algebraic over FF, its minimal polynomial mα(x)m_\alpha(x) is the monic polynomial of least degree in F[x]F[x] with mα(α)=0m_\alpha(\alpha)=0.

Theorem 18.2 — The Minimal Polynomial Is Irreducible and Unique

mα(x)m_\alpha(x) is irreducible over FF, and it is the unique monic polynomial of its degree vanishing at α\alpha.

Proof

Irreducible. Suppose mα=fgm_\alpha = fg with degf,degg<degmα\deg f,\deg g < \deg m_\alpha. Then f(α)g(α)=mα(α)=0f(\alpha)g(\alpha) = m_\alpha(\alpha)=0; since KK is a field (no zero divisors), f(α)=0f(\alpha)=0 or g(α)=0g(\alpha)=0. Either way, a polynomial of degree strictly less than degmα\deg m_\alpha vanishes at α\alpha — contradicting minimality (after scaling to make it monic, using that FF is a field so any nonzero leading coefficient is invertible). So no such factorization exists (beyond trivial unit factors): mαm_\alpha is irreducible.

Unique. Suppose gg is also monic of degree degmα\deg m_\alpha with g(α)=0g(\alpha)=0. Then (mαg)(α)=00=0(m_\alpha-g)(\alpha) = 0-0=0, and deg(mαg)<degmα\deg(m_\alpha-g) < \deg m_\alpha (the leading terms, both monic of the same degree, cancel). If mαg0m_\alpha-g\neq0, scaling to monic gives a smaller-degree polynomial vanishing at α\alpha, contradicting minimality. So mαg=0m_\alpha-g=0: g=mαg=m_\alpha.


04 · Simple Extensions

The minimal polynomial completely determines the structure of the smallest field containing α\alpha.

Lemma 18.3 — Nonzero Prime Ideals in a PID Are Maximal

In a PID RR, every nonzero prime ideal (p)(p) is maximal.

Proof

Suppose (p)(d)R(p)\subseteq(d)\subseteq R for some dRd\in R. Then dpd\mid p, so p=dcp=dc for some cRc\in R. Since (p)(p) is prime, pp is a prime element (Chapter 15's convention), hence irreducible (Theorem 15.3). So in p=dcp=dc, either dd or cc is a unit. If dd is a unit, (d)=R(d)=R. If cc is a unit, d=pc1d=pc^{-1} is an associate of pp, so (d)=(p)(d)=(p). Either way, no ideal sits strictly between (p)(p) and RR: (p)(p) is maximal.

Theorem 18.4 — Structure of F(α) for Algebraic α

Let αK\alpha\in K be algebraic over FF with minimal polynomial mαm_\alpha of degree nn. Then:

F(α)    F[x]/(mα),[F(α):F]=nF(\alpha) \;\cong\; F[x]/(m_\alpha), \qquad [F(\alpha):F] = n

with {1,α,α2,,αn1}\{1,\alpha,\alpha^2,\ldots,\alpha^{n-1}\} a basis for F(α)F(\alpha) over FF.

Proof

Consider the evaluation homomorphism evα:F[x]K\mathrm{ev}_\alpha: F[x]\to K, ff(α)f\mapsto f(\alpha) (a ring homomorphism, Chapter 13's evaluation example specialized to polynomials). ker(evα)={fF[x]:f(α)=0}\ker(\mathrm{ev}_\alpha) = \{f\in F[x] : f(\alpha)=0\} is an ideal (Theorem 13.3); since F[x]F[x] is Euclidean (Chapter 15), hence a PID, this ideal is (g)(g) for some gg, and gg has the minimal degree among nonzero elements of the ideal (the proof of Theorem 15.1 identifies the generator exactly this way) — so gg is (up to a unit scalar) the minimal polynomial: ker(evα)=(mα)\ker(\mathrm{ev}_\alpha) = (m_\alpha).

(mα)(m_\alpha) is a nonzero prime ideal: mαm_\alpha is irreducible (Theorem 18.2), hence prime in the PID F[x]F[x] (Theorem 15.4). By Lemma 18.3, (mα)(m_\alpha) is maximal. By Theorem 12.6, F[x]/(mα)F[x]/(m_\alpha) is a field.

By the First Isomorphism Theorem (Theorem 13.6): F[x]/ker(evα)im(evα)F[x]/\ker(\mathrm{ev}_\alpha) \cong \mathrm{im}(\mathrm{ev}_\alpha), i.e. F[x]/(mα)F[α]F[x]/(m_\alpha) \cong F[\alpha] (the image is exactly the subring generated by α\alpha over FF, since every polynomial expression in α\alpha is hit). Since the left side is a field, F[α]F[\alpha] is already a field — meaning F[α]=F(α)F[\alpha]=F(\alpha) (a field containing FF and α\alpha that's already closed under inverses needs nothing more adjoined).

Finally, by the division algorithm (Theorem 14.5), every coset in F[x]/(mα)F[x]/(m_\alpha) has a unique representative of degree <n<n (divide by mαm_\alpha and keep the remainder), so {1,x,,xn1}+(mα)\{1,x,\ldots,x^{n-1}\}+(m_\alpha) is a basis for F[x]/(mα)F[x]/(m_\alpha) over FF, corresponding to {1,α,,αn1}\{1,\alpha,\ldots,\alpha^{n-1}\} under the isomorphism. So [F(α):F]=n[F(\alpha):F]=n.

Example — Q(√2)

2\sqrt2 is a root of x22x^2-2, irreducible over Q\mathbb{Q} (no rational root, by the classical proof that 2\sqrt2 is irrational, together with Corollary 14.7 — a degree-2 polynomial with no root can't factor into two degree-1 pieces). So m2(x)=x22m_{\sqrt2}(x)=x^2-2, and by Theorem 18.4:

Q(2)Q[x]/(x22),[Q(2):Q]=2\mathbb{Q}(\sqrt2) \cong \mathbb{Q}[x]/(x^2-2), \qquad [\mathbb{Q}(\sqrt2):\mathbb{Q}]=2

with basis {1,2}\{1,\sqrt2\} — matching the familiar description Q(2)={a+b2:a,bQ}\mathbb{Q}(\sqrt2) = \{a+b\sqrt2 : a,b\in\mathbb{Q}\}.

Example — Constructing C purely algebraically

x2+1x^2+1 is irreducible over R\mathbb{R} (Chapter 15, Section 06's example: no real root). By Theorem 18.4, R[x]/(x2+1)\mathbb{R}[x]/(x^2+1) is a field of degree 2 over R\mathbb{R}, with basis {1,xˉ}\{1,\bar x\} (writing xˉ\bar x for the coset x+(x2+1)x+(x^2+1)) satisfying xˉ2=1\bar x^2 = -1 (since x21x^2\equiv-1 modulo x2+1x^2+1). This field is exactly C\mathbb{C}, with xˉ\bar x playing the role of ii — the complex numbers, built with no appeal to square roots of negative numbers, purely as a quotient of a polynomial ring.


05 · A Degree-4 Extension

Example — [Q(√2,√3):Q] = 4

[Q(2):Q]=2[\mathbb{Q}(\sqrt2):\mathbb{Q}]=2 (Section 04). Working inside K=Q(2,3)K=\mathbb{Q}(\sqrt2,\sqrt3), consider 3\sqrt3 over the base field Q(2)\mathbb{Q}(\sqrt2): 3\sqrt3 satisfies x23x^2-3, and this remains irreducible over Q(2)\mathbb{Q}(\sqrt2) (if 3=a+b2\sqrt3=a+b\sqrt2 for a,bQa,b\in\mathbb{Q}, squaring gives 3=a2+2b2+2ab23=a^2+2b^2+2ab\sqrt2; since 2Q\sqrt2\notin\mathbb{Q}, this forces ab=0ab=0, and checking b=0b=0 gives 3=a23=a^2 with no rational solution, while a=0a=0 gives 3=2b23=2b^2, also no rational solution — so no such a,ba,b exist, and 3Q(2)\sqrt3\notin\mathbb{Q}(\sqrt2)). So [Q(2,3):Q(2)]=2[\mathbb{Q}(\sqrt2,\sqrt3):\mathbb{Q}(\sqrt2)]=2 by Theorem 18.4 applied over the base field Q(2)\mathbb{Q}(\sqrt2).

By the Tower Law (Theorem 18.1): [Q(2,3):Q]=2×2=4[\mathbb{Q}(\sqrt2,\sqrt3):\mathbb{Q}] = 2\times2=4, with basis {1,2,3,6}\{1,\sqrt2,\sqrt3,\sqrt6\} (products of the two individual bases, as in Theorem 18.1's proof).


06 · Finite Extensions Are Algebraic

Theorem 18.5 — Finite Extensions Are Algebraic

If [K:F][K:F] is finite, every αK\alpha\in K is algebraic over FF.

Proof

Let n=[K:F]n=[K:F]. The n+1n+1 elements 1,α,α2,,αn1,\alpha,\alpha^2,\ldots,\alpha^n all live in the nn-dimensional FF-vector space KK, so they are linearly dependent over FF (more vectors than the dimension): there exist c0,,cnFc_0,\ldots,c_n\in F, not all zero, with c0+c1α++cnαn=0c_0+c_1\alpha+\cdots+c_n\alpha^n=0. This is exactly the statement that α\alpha is a root of the nonzero polynomial c0+c1x++cnxnF[x]c_0+c_1x+\cdots+c_nx^n\in F[x]: α\alpha is algebraic over FF.

Extensions built from algebraic elements are automatically finite too

The converse direction — combining Theorem 18.4's degree formula with the Tower Law — shows that adjoining finitely many algebraic elements always produces a finite extension, exactly as Section 05's example demonstrated for 2,3\sqrt2,\sqrt3. Finiteness and "built from algebraic elements" turn out to be two sides of the same coin, a theme Chapter 20's Galois correspondence relies on constantly.


07 · Exercises

EXERCISE 18.1

Apply the Tower Law directly, using the given intermediate degrees.

By Theorem 18.1: [L:F]=[L:K][K:F]=3×5=15[L:F] = [L:K][K:F] = 3\times5=15.

If [K:F]=5[K:F]=5 and [L:K]=3[L:K]=3, find [L:F][L:F] using the Tower Law.

EXERCISE 18.2

Find a polynomial with rational coefficients having 23\sqrt[3]2 as a root, and check irreducibility (no rational root, and it can't factor as degree 1 times degree 2 without a rational root).

23\sqrt[3]2 is a root of x32x^3-2. Rational root test: any rational root would be ±1,±2\pm1,\pm2 (divisors of the constant term), and none satisfy x3=2x^3=2. Since a degree-3 polynomial with no rational root cannot factor into a linear times a quadratic (the linear factor would supply a rational root, by the Factor Theorem, Theorem 14.6), x32x^3-2 is irreducible over Q\mathbb{Q}.

By Theorem 18.4, m23(x)=x32m_{\sqrt[3]2}(x)=x^3-2 and [Q(23):Q]=3[\mathbb{Q}(\sqrt[3]2):\mathbb{Q}]=3.

Find the minimal polynomial of 23\sqrt[3]{2} over Q\mathbb{Q}, and state [Q(23):Q][\mathbb{Q}(\sqrt[3]{2}):\mathbb{Q}].

EXERCISE 18.3

Apply Theorem 18.5 directly — you don't need to find the polynomial explicitly.

Since [Q(2,3):Q]=4[\mathbb{Q}(\sqrt2,\sqrt3):\mathbb{Q}]=4 (Section 05) is finite, Theorem 18.5 guarantees 2+3\sqrt2+\sqrt3 (an element of this field) is algebraic over Q\mathbb{Q}, without needing to exhibit the polynomial. (It happens to satisfy x410x2+1=0x^4-10x^2+1=0, found by repeatedly squaring x=2+3x=\sqrt2+\sqrt3 and eliminating radicals — but Theorem 18.5 guarantees algebraicity before any such computation.)

Using Theorem 18.5 alone (without finding an explicit polynomial), explain why 2+3\sqrt2+\sqrt3 must be algebraic over Q\mathbb{Q}.

EXERCISE 18.4

Apply Theorem 18.4 with the given irreducible polynomial directly.

x2+x+1x^2+x+1 is irreducible over Z2\mathbb{Z}_2 (checking both elements of Z2\mathbb{Z}_2: 02+0+1=100^2+0+1=1\neq0; 12+1+1=101^2+1+1=1\neq0 — no root, and a degree-2 polynomial with no root over a field is irreducible). By Theorem 18.4, Z2[x]/(x2+x+1)\mathbb{Z}_2[x]/(x^2+x+1) is a field with [:Z2]=2[\,\cdot\,:\mathbb{Z}_2]=2, hence 22=42^2=4 elements: basis {1,xˉ}\{1,\bar x\} over Z2\mathbb{Z}_2, giving elements {0,1,xˉ,1+xˉ}\{0,1,\bar x,1+\bar x\}. This is a field with exactly 4 elements — resolving Exercise 17.2's question by explicit construction.

Using Theorem 18.4 and the irreducible polynomial x2+x+1x^2+x+1 over Z2\mathbb{Z}_2, construct a field with exactly 4 elements.


08 · Chapter Summary

ConceptStatement
Extension, degreeK/FK/F; [K:F]=dimFK[K:F] = \dim_F K
Tower Law[L:F]=[L:K][K:F][L:F]=[L:K][K:F] (Thm 18.1)
Algebraic / transcendentalRoot of some / no nonzero fF[x]f\in F[x]
Minimal polynomialLeast-degree monic vanishing at α\alpha; irreducible and unique (Thm 18.2)
Nonzero prime ideals in a PIDAlways maximal (Lemma 18.3)
F(α)F(\alpha) structureF(α)F[x]/(mα)F(\alpha)\cong F[x]/(m_\alpha); [F(α):F]=degmα[F(\alpha):F]=\deg m_\alpha (Thm 18.4)
C\mathbb{C} from R\mathbb{R}R[x]/(x2+1)C\mathbb{R}[x]/(x^2+1) \cong \mathbb{C}
Finite     \implies algebraic[K:F][K:F] finite     \implies every element of KK algebraic over FF (Thm 18.5)

Next: Chapter 19 — Splitting Fields & Algebraic Closures builds the smallest extension in which a given polynomial factors completely into linear pieces, the essential setting for Galois theory.