Algebraic Number Fields and Rings of Integers

⚡ Key Takeaways

An algebraic number field $K$ is a finite-degree field extension of $\mathbb{Q}$ ; its ring of integers $\mathcal{O}_K$ is the set of elements satisfying monic integer polynomials. Key result: $\mathcal{O}_K$ is a Dedekind domain — a Noetherian, integrally closed domain of Krull dimension 1 — so every nonzero ideal factors uniquely into prime ideals, restoring "unique factorization" at the level of ideals even when elements fail it. Applies to algebraic number theory, Diophantine equations, class field theory, and cryptography. Foundations laid by Kummer, Dedekind, and Hilbert (1840s–1900s).

💡 The Core Insight

Replace numbers by ideals. The multiplicative structure that breaks for elements (non-unique factorization) is repaired by passing to the monoid of nonzero ideals, which becomes a free abelian object on the prime ideals. The two great finiteness theorems — finiteness of the class number and Dirichlet's unit theorem — quantify how far $\mathcal{O}_K$ is from being a PID and how big its group of units is. Everything in classical algebraic number theory orbits these two invariants.

DefinitionNumber Field

A number field is a field $K$ with $\mathbb{Q} \subseteq K$ and $[K : \mathbb{Q}] = n < \infty$ . The integer $n$ is the degree of $K$ . By the primitive element theorem, $K = \mathbb{Q}(\theta)$ for some $\theta$ , and $\theta$ is a root of an irreducible monic polynomial $m_\theta \in \mathbb{Q}[x]$ of degree $n$ .

DefinitionAlgebraic Integer and Ring of Integers

An element $\alpha \in K$ is an algebraic integer if it is a root of a monic polynomial with coefficients in $\mathbb{Z}$ :

1

\quad \alpha^m + c_{m-1}\alpha^{m-1} + \cdots + c_1\alpha + c_0 = 0,\qquad c_i \in \mathbb{Z}.

Derivation hint: equivalently $\alpha$ generates a finitely generated $\mathbb{Z}$ -module $\mathbb{Z}[\alpha]$ ; integrality is closure under such finite generation. The set of all algebraic integers in $K$ is the ring of integers $\mathcal{O}_K$ , equal to the integral closure of $\mathbb{Z}$ in $K$ .

TheoremStructure of

\mathcal{O}_K

Let $K$ be a number field of degree $n$ . Then: 1. $\mathcal{O}_K$ is a free $\mathbb{Z}$ -module of rank $n$ ; it admits an integral basis $\{\omega_1,\dots,\omega_n\}$ with $\mathcal{O}_K = \mathbb{Z}\omega_1 \oplus \cdots \oplus \mathbb{Z}\omega_n$ . 2. $\mathcal{O}_K$ is a Dedekind domain: Noetherian, integrally closed, and of Krull dimension 1 (every nonzero prime ideal is maximal). 3. Every nonzero ideal $\mathfrak{a} \subseteq \mathcal{O}_K$ factors uniquely as a product of prime ideals,

5

\quad \mathfrak{a} = \mathfrak{p}_1^{e_1}\,\mathfrak{p}_2^{e_2}\cdots \mathfrak{p}_g^{e_g},

Derivation hint: Dedekind domains are precisely the domains where the monoid of nonzero ideals is free abelian on the prime ideals; uniqueness follows by localizing at each $\mathfrak{p}_i$ to a DVR.

TheoremClass Number and Unit Theorems

For a number field $K$ of signature $(r_1, r_2)$ :

(Finiteness of class number) The ideal class group $\mathrm{Cl}(K) = I_K / P_K$ , the quotient of fractional ideals by principal fractional ideals, is finite. Its order $h_K$ is the class number.

(Dirichlet's unit theorem) The unit group factors as

6

\quad \mathcal{O}_K^\times \cong \mu_K \times \mathbb{Z}^{\,r_1 + r_2 - 1},

Derivation hint: the logarithmic embedding sends units into a hyperplane of $\mathbb{R}^{r_1+r_2}$ ; its image is a full lattice (covolume = regulator), and the kernel is the finite group of roots of unity $\mu_K$ . where $\mu_K$ is the finite cyclic group of roots of unity in $K$ .

Proof

\mathcal{O}_K

is a free

\mathbb{Z}

-module of rank

n

Step 1 — Integers form a ring. If $\alpha,\beta$ are integral, then $\mathbb{Z}[\alpha,\beta]$ is a finitely generated $\mathbb{Z}$ -module. Any element of a ring that is finitely generated as a $\mathbb{Z}$ -module is integral (Cayley–Hamilton applied to multiplication). Hence $\alpha\pm\beta$ , $\alpha\beta$ are integral, so $\mathcal{O}_K$ is a ring.

Step 2 — $\mathcal{O}_K$ spans $K$ over $\mathbb{Q}$ . For any $\alpha \in K$ , clearing denominators of its minimal polynomial gives $m\alpha \in \mathcal{O}_K$ for some $m \in \mathbb{Z}_{>0}$ . So $K = \mathbb{Q}\otimes_\mathbb{Z} \mathcal{O}_K$ and $\mathcal{O}_K$ contains a $\mathbb{Q}$ -basis of $K$ .

Step 3 — Sandwiching by lattices. Pick a $\mathbb{Q}$ -basis $\{\beta_1,\dots,\beta_n\}\subseteq \mathcal{O}_K$ of $K$ . The trace form is nondegenerate (separability), so there is a dual basis $\{\beta_1^*,\dots,\beta_n^*\}$ with $\mathrm{Tr}(\beta_i\beta_j^*)=\delta_{ij}$ . For $\alpha = \sum a_j \beta_j^* \in \mathcal{O}_K$ , the coefficients are $a_j = \mathrm{Tr}(\alpha\beta_j) \in \mathbb{Z}$ . Hence

7

\quad \bigoplus_{j} \mathbb{Z}\beta_j \;\subseteq\; \mathcal{O}_K \;\subseteq\; \bigoplus_j \mathbb{Z}\beta_j^* .

Derivation hint: integrality forces traces $\mathrm{Tr}(\alpha\beta_j)$ to be rational integers, locking $\mathcal{O}_K$ between two free modules of rank $n$ .

Step 4 — Conclusion. A subgroup of a free abelian group of rank $n$ that contains a free subgroup of rank $n$ is itself free of rank $n$ (structure theorem for f.g. modules over the PID $\mathbb{Z}$ ). Therefore $\mathcal{O}_K \cong \mathbb{Z}^n$ . $\blacksquare$

∎

ProofUnique factorization of ideals — sketch

Step 1 — Noetherian. $\mathcal{O}_K \cong \mathbb{Z}^n$ is a Noetherian $\mathbb{Z}$ -module, hence Noetherian as a ring.

Step 2 — Dimension 1. If $\mathfrak{p}\neq 0$ is prime, then $\mathfrak{p}\cap\mathbb{Z} = (p)$ for a rational prime $p$ , and $\mathcal{O}_K/\mathfrak{p}$ is a finite integral domain, hence a field. So $\mathfrak{p}$ is maximal.

Step 3 — Integrally closed. $\mathcal{O}_K$ is by definition the integral closure of $\mathbb{Z}$ in $K$ , hence integrally closed in its fraction field $K$ .

Step 4 — Containment is divisibility. In a Dedekind domain one shows every ideal contains a product of primes, and that for any nonzero ideal $\mathfrak{a}$ there is a fractional ideal $\mathfrak{a}^{-1}$ with $\mathfrak{a}\mathfrak{a}^{-1} = \mathcal{O}_K$ . Invertibility makes the monoid of nonzero ideals a group modulo cancellation, forcing unique factorization into primes exactly as in $\mathbb{Z}$ . $\blacksquare$

∎

mathematician · Richard Dedekind | 1831–1916

Dedekind, a student of Gauss and Dirichlet at Göttingen, introduced the modern concept of an ideal in his 1871 supplement to Dirichlet's Vorlesungen über Zahlentheorie, generalizing Kummer's "ideal numbers" into actual subsets of a ring. He defined the ring of integers, proved unique factorization of ideals, and effectively created the axiomatic theory of what we now call Dedekind domains. His insistence on set-theoretic, structural definitions — ideals as sets rather than formal symbols — was a decisive step toward abstract algebra. Dedekind also gave the modern definition of infinite sets and constructed the reals via Dedekind cuts.

🧠 Intuition

Think of $\mathcal{O}_K$ as a lattice sitting inside the real vector space $K\otimes_\mathbb{Q}\mathbb{R} \cong \mathbb{R}^{r_1}\times\mathbb{C}^{r_2}$ , via the embedding $\alpha\mapsto(\sigma_1(\alpha),\dots)$ . This is the Minkowski embedding, and it turns arithmetic questions into geometry-of-numbers questions: ideals become sublattices, the norm becomes a covolume ratio, and Minkowski's convex-body theorem produces small elements in any ideal — the engine behind class-number finiteness. For unique factorization, the mental model is: primes of $\mathbb{Z}$ can split, stay inert, or ramify when lifted to $\mathcal{O}_K$ . A prime $p$ becomes $\mathfrak{p}_1^{e_1}\cdots\mathfrak{p}_g^{e_g}$ , and the fundamental identity $\sum e_i f_i = n$ (with $f_i$ the residue degrees) governs the bookkeeping. The class group measures the "twisting" that prevents prime ideals from all being principal: if $h_K = 1$ , every ideal is principal and $\mathcal{O}_K$ is a PID with genuine unique factorization of elements.

Figure

Minkowski lattice of

\mathcal{O}_K

for an imaginary quadratic field, with a fundamental domain and a small ideal element found inside a convex body

ExampleRing of integers of

\mathbb{Q}(\sqrt{d})

and a factorization

Given. Let $d$ be a squarefree integer, $K=\mathbb{Q}(\sqrt{d})$ , $n=2$ .

Step 1 — Integral basis. Standard computation: an element $a+b\sqrt d$ ( $a,b\in\mathbb{Q}$ ) is integral iff its trace $2a$ and norm $a^2-db^2$ are in $\mathbb{Z}$ . Analysis of cases gives:

8

\quad \mathcal{O}_K=\begin{cases}\mathbb{Z}[\sqrt d], & d\equiv 2,3 \pmod 4,\\[4pt]\mathbb{Z}\!\left[\dfrac{1+\sqrt d}{2}\right], & d\equiv 1 \pmod 4.\end{cases}

Derivation hint: solve the integrality conditions $2a\in\mathbb{Z}$ , $a^2-db^2\in\mathbb{Z}$ modulo 4; the half-integer basis appears exactly when $d\equiv1$ .

Step 2 — Discriminant. $d_K = d$ if $d\equiv1\!\pmod4$ , else $d_K=4d$ .

Step 3 — Take $d=-5$ . Then $d\equiv3\!\pmod4$ , so $\mathcal{O}_K=\mathbb{Z}[\sqrt{-5}]$ , $d_K=-20$ . Here $6=2\cdot3=(1+\sqrt{-5})(1-\sqrt{-5})$ — non-unique factorization of elements. As ideals:

9

\quad (6)=\mathfrak{p}_2^2\,\mathfrak{p}_3\,\mathfrak{p}_3',\quad \mathfrak{p}_2=(2,1+\sqrt{-5}),\ \mathfrak{p}_3=(3,1+\sqrt{-5}),\ \mathfrak{p}_3'=(3,1-\sqrt{-5}).

Derivation hint: factor $(2),(3)$ via reduction of $x^2+5$ mod 2 and mod 3, then regroup the principal factorizations.

Boxed answer. $\boxed{\,\mathcal{O}_{\mathbb{Q}(\sqrt{-5})}=\mathbb{Z}[\sqrt{-5}],\ d_K=-20,\ h_K=2\,}$ — the class number 2 is precisely the obstruction making $6$ factor two ways.

ExampleUnit group and regulator of

\mathbb{Q}(\sqrt{2})

Given. $K=\mathbb{Q}(\sqrt2)$ , real quadratic, signature $(r_1,r_2)=(2,0)$ , so unit rank $=r_1+r_2-1=1$ .

Step 1 — Roots of unity. A real field has $\mu_K=\{\pm1\}$ .

Step 2 — Fundamental unit. Solve Pell: $x^2-2y^2=\pm1$ . Smallest nontrivial solution $(x,y)=(1,1)$ gives the fundamental unit

10

\quad \varepsilon = 1+\sqrt2,\qquad N_{K/\mathbb{Q}}(\varepsilon)=(1+\sqrt2)(1-\sqrt2)=-1.

Derivation hint: units have norm $\pm1$ ; the minimal Pell solution generates all units modulo $\pm1$ by Dirichlet's theorem with rank 1.

Step 3 — Structure and regulator. $\mathcal{O}_K^\times=\{\pm(1+\sqrt2)^k:k\in\mathbb{Z}\}\cong\{\pm1\}\times\mathbb{Z}$ . The regulator is

11

\quad R_K=\big|\log|\sigma_1(\varepsilon)|\big|=\log(1+\sqrt2)\approx 0.8814.

Derivation hint: with unit rank 1 the regulator is the single entry $|\log|\varepsilon||$ of the $1\times1$ logarithmic-embedding matrix.

Boxed answer. $\boxed{\,\mathcal{O}_K^\times\cong\{\pm1\}\times\langle 1+\sqrt2\rangle,\ R_K=\log(1+\sqrt2)\approx0.8814\,}$

⚠️ Common mistake

$\mathcal{O}_K \neq \mathbb{Z}[\theta]$ in general. Even when $K=\mathbb{Q}(\theta)$ , the ring of integers can be strictly larger than $\mathbb{Z}[\theta]$ . The index $[\mathcal{O}_K : \mathbb{Z}[\theta]]^2 = \mathrm{disc}(m_\theta)/d_K$ measures the gap. The classic failure: $\theta=\sqrt{-3}$ gives $\mathbb{Z}[\sqrt{-3}] \subsetneq \mathcal{O}_K = \mathbb{Z}[\frac{1+\sqrt{-3}}{2}]$ , index 2. A field is called monogenic when some generator gives $\mathcal{O}_K=\mathbb{Z}[\theta]$ — not all are.

⚠️ Common mistake

Unique factorization of ideals is NOT unique factorization of elements. $\mathcal{O}_K$ always has unique ideal factorization, but is a UFD iff it is a PID iff $h_K=1$ . Conflating these is the single most common error. Likewise, $\mathfrak{a}\mid\mathfrak{b} \iff \mathfrak{a}\supseteq\mathfrak{b}$ — "to divide is to contain," which reverses the naive intuition.

Counterexample

A non-monogenic ring of integers. Dedekind's own example: $K=\mathbb{Q}(\theta)$ with $\theta^3-\theta^2-2\theta-8=0$ . Here $\mathcal{O}_K$ is not of the form $\mathbb{Z}[\alpha]$ for any single $\alpha$ ; the prime $2$ splits completely as $(2)=\mathfrak{p}_1\mathfrak{p}_2\mathfrak{p}_3$ with three distinct primes of residue degree 1, which is impossible for $\mathbb{Z}[\alpha]/(2)$ to realize since a cubic over $\mathbb{F}_2$ cannot have three distinct linear factors (only two monic linear polynomials exist mod 2). This "common index divisor" obstruction shows monogenicity genuinely fails.

connection

The theory threads through much of algebra and number theory. The factorization of primes connects to Galois Theory: for $K/\mathbb{Q}$ Galois, the Galois group acts transitively on primes above each $p$ , and the decomposition/inertia groups encode splitting. Class Field Theory generalizes the class group into a description of all abelian extensions of $K$ via the idele class group. The Minkowski embedding ties the subject to Geometry of Numbers and lattice theory. The analytic side — the Dedekind Zeta Function $\zeta_K(s)$ — packages $h_K$ , $R_K$ , $d_K$ , and $|\mu_K|$ into the class number formula as its residue at $s=1$ . Finally, Dedekind Domains abstract the purely algebraic content, and Valuation Theory / $p$ -adic completions give the local-global perspective via Adeles and Ideles.

🔬 Where this appears

⏳ History

- 1801 — Gauss's Disquisitiones Arithmeticae studies binary quadratic forms and composition, implicitly the class group of quadratic fields.

- 1832 — Gauss introduces the Gaussian integers

\mathbb{Z}[i]

, the first ring of integers studied as such, and proves it is a UFD.

- 1844–1847 — Kummer, attacking Fermat's Last Theorem, discovers that cyclotomic integers

\mathbb{Z}[\zeta_p]

fail unique factorization and invents "ideal numbers" to restore it; proves FLT for regular primes.

- 1871 — Dedekind defines ideals as sets, the ring of integers, and proves unique factorization of ideals in his supplement to Dirichlet's lectures.

- 1882 — Kronecker develops a parallel divisor-theoretic foundation; Dedekind's set-theoretic approach ultimately prevails.

- 1846 / 1871 — Dirichlet's unit theorem established and refined; the regulator emerges as a key invariant.

- 1896 — Minkowski's geometry of numbers gives clean proofs of class-number finiteness and bounds on discriminants (no unramified extensions of

\mathbb{Q}

).

- 1897 — Hilbert's Zahlbericht unifies and systematizes the field, setting the agenda (ramification, Hilbert class field) for the 20th century.

- 1920s–1930s — Artin, Hasse, Chevalley, and Takagi build class field theory; Chevalley introduces ideles, completing the local-global synthesis.

open-problem

Gauss's class number problem (imaginary quadratic). Determining all imaginary quadratic fields with a given class number was hard: the complete list of the nine fields with $h_K=1$ ( $d=-1,-2,-3,-7,-11,-19,-43,-67,-163$ ) was only settled by Heegner (1952, with gaps), Baker, and Stark (1960s). For each fixed $h$ , effective determination came from Goldfeld–Gross–Zagier.

Real quadratic class number 1. Conjecturally infinitely many real quadratic fields have $h_K=1$ (Gauss), but this is open — we cannot even prove infinitely many number fields are PIDs. The Cohen–Lenstra heuristics predict densities of class-group structures, yet remain conjectural.

Monogenicity. Deciding whether a given field is monogenic, and counting power-integral bases, is a deep and largely open problem (Gaál, Győry). Most number fields of degree $\geq 4$ are conjecturally non-monogenic in a density sense (Bhargava–Shankar–Wang).

Vandiver's conjecture. Whether $p \nmid h^+_{\mathbb{Q}(\zeta_p)}$ (the plus-part class number) holds for all primes is open and has consequences in Iwasawa theory and FLT-style arguments.

Algebraic Number Fields and Rings of Integers

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also