Quadratic Fields and the Gaussian Integers

Quadratic fields are degree-2 extensions

Q (d)

of the rationals; the Gaussian integers

Z [i]

are the integers of the simplest imaginary one,

Q (i)

. Key result:

Z [i]

is a Euclidean domain, hence a unique factorization domain, and its primes classify exactly which integers are sums of two squares. Applies across algebraic number theory, Diophantine equations, and cryptography. Foundational work by Gauss (1801, 1832), Dirichlet, and Dedekind (19th century).

A quadratic field is any field $K = Q (d)$ with $d$ a squarefree integer $\neq = 0, 1$ ; it is a two-dimensional vector space over $Q$ . The ring of integers $O_{K}$ inside it is the central object of study, and for $K = Q (i)$ it equals the Gaussian integers $Z [i] = {a + bi : a, b \in Z}$ . These rings matter because they are the first nontrivial laboratory in which the failure—or survival—of unique factorization is visible, motivating Dedekind's theory of ideals. A surprising fact: although $Z [i]$ has unique factorization, the closely related $Z [- 5]$ does not, since $6 = 2 \cdot 3 = (1 + - 5) (1 - - 5)$ . The arithmetic of $Z [i]$ encodes a classical theorem of Fermat about sums of two squares with striking transparency.

The Core Insight

Number-theoretic questions about $Z$ become transparent when lifted into a larger ring where rational primes can split. The prime $5 = (2 + i) (2 - i)$ factors in $Z [i]$ precisely because $5 \equiv 1 (mod 4)$ and is a sum of two squares, $5 = 2^{2} + 1^{2}$ . The behavior of a rational prime $p$ in $O_{K}$ —split, inert, or ramified—is the engine driving the entire theory, governed by quadratic reciprocity through the Legendre/Kronecker symbol.

Formal Statement

Quadratic Field

For a squarefree integer $d \in / {0, 1}$ , the quadratic field is

K = Q (d) = {x + y d : x, y \in Q} .

⟳adjoin a root of the irreducible

t^{2} - d

Q

; the quotient

Q [t] / (t^{2} - d)

is a field of degree 2.

K

is real quadratic if

d > 0

and imaginary quadratic if

d < 0

. The nontrivial Galois automorphism is conjugation

σ (x + y d) = x - y d

Norm and Trace

For $α = x + y d \in K$ , define the trace and norm by

Tr (α) = α + σ (α) = 2 x, N (α) = α σ (α) = x^{2} - d y^{2} .

⟳these are the coefficients of the characteristic polynomial

t^{2} - Tr (α) t + N (α)

of multiplication-by-

α

on the

Q

-vector space

K

The norm is multiplicative: $N (α β) = N (α) N (β)$ . For the Gaussian integers, $N (a + bi) = a^{2} + b^{2}$ .

Ring of Integers

The ring of integers $O_{K}$ is the integral closure of $Z$ in $K$ —the set of $α \in K$ satisfying a monic polynomial with $Z$ coefficients. Explicitly,

O_{K} = ⎩ ⎨ ⎧ Z [d], Z [\frac{1 + d}{2}], d \equiv 2, 3 (mod 4), d \equiv 1 (mod 4) .

⟳

α = x + y d

is integral iff

Tr (α) = 2 x \in Z

and

N (α) = x^{2} - d y^{2} \in Z

; analyzing parities of

2 x, 2 y

modulo 4 yields the two cases.

Discriminant

The discriminant of $K = Q (d)$ is

D_{K} = {d, 4 d, d \equiv 1 (mod 4), d \equiv 2, 3 (mod 4) .

It satisfies

D_{K} \equiv 0

1 (mod 4)

and equals

det (Tr (ω_{i} ω_{j}))

for any integral basis

{ω_{1}, ω_{2}}

⟳compute the Gram matrix of traces

Tr (ω_{i} ω_{j})

for the integral basis from [3]; its determinant gives [4].

Gaussian Integers are Euclidean

The ring $Z [i] = O_{Q (i)}$ is a Euclidean domain with respect to the norm $N (a + bi) = a^{2} + b^{2}$ . Consequently it is a principal ideal domain and a unique factorization domain. Its units are exactly ${1, - 1, i, - i}$ , the elements of norm 1.

:::proof[Euclidean property of $Z [i]$ ] Step 1 — Goal. Given $α, β \in Z [i]$ with $β \neq = 0$ , find $q, r \in Z [i]$ with $α = q β + r$ and $N (r) < N (β)$ .

Step 2 — Exact quotient. Compute in the field

Q (i)

\frac{α}{β} = u + v i, u, v \in Q .

⟳rationalize

α \overset{ˉ}{β} / (β \overset{ˉ}{β}) = α \overset{ˉ}{β} / N (β)

, splitting real and imaginary parts.

Step 3 — Round. Choose integers

m, n

with

∣ u - m ∣ \leq \frac{1}{2}

and

∣ v - n ∣ \leq \frac{1}{2}

, and set

q = m + ni

r = α - q β

. Step 4 — Bound the remainder. Since

r / β = (u - m) + (v - n) i

N (r) = N (β) [(u - m)^{2} + (v - n)^{2}] \leq N (β) (\frac{1}{4} + \frac{1}{4}) = \frac{1}{2} N (β) < N (β) .

⟳apply multiplicativity

N (r) = N (β) N (r / β)

and bound each rounded coordinate by

\frac{1}{2}

Step 5 — Conclude. The division algorithm holds, so

Z [i]

is Euclidean; the standard chain Euclidean

\Rightarrow

PID

\Rightarrow

UFD finishes the proof.

■

Classification of Gaussian Primes

Up to units, the primes of $Z [i]$ are exactly:

$1 + i$ , the unique ramified prime above $2$ , with $2 = - i (1 + i)^{2}$ ;
For rational primes $p \equiv 1 (mod 4)$ : two conjugate split primes $π, \overset{π}{ˉ}$ with $p = π \overset{π}{ˉ} = a^{2} + b^{2}$ ;
For rational primes $p \equiv 3 (mod 4)$ : $p$ itself remains inert (prime in $Z [i]$ ).

⟳a rational prime

p

splits iff

- 1

is a quadratic residue mod

p

, i.e.

x^{2} \equiv - 1

is solvable; by Euler's criterion this happens iff

p \equiv 1 (mod 4)

:::svg[Splitting of rational primes in the Gaussian integer lattice $Z [i]$ ]

Why It Holds / Derivation

Sums of Two Squares from Gaussian Arithmetic

Step 1. A rational prime

p

is a sum of two squares iff

p = N (a + bi)

for some

a + bi \in Z [i]

, iff

p

is reducible (splits) in

Z [i]

. Step 2.

p

splits iff the equation

x^{2} \equiv - 1 (mod p)

has a solution, since then

p ∣ x^{2} + 1 = (x + i) (x - i)

while

p ∤ (x \pm i)

, so

p

is not prime. Step 3. By Euler's criterion,

(\frac{- 1}{p}) = (- 1)^{(p - 1) /2} = {+ 1, - 1, p \equiv 1 (mod 4), p \equiv 3 (mod 4) .

⟳

(- 1)^{(p - 1) /2}

records whether the order-2 element

- 1

lies in the squares of the cyclic group

(Z / p)^{\times}

Step 4 — Full count (Jacobi/Gauss). The number

r_{2} (n)

of representations of

n

as an ordered sum of two squares is

r_{2} (n) = 4 d ∣ n \sum χ (d), χ (d) = ⎩ ⎨ ⎧ 1 - 1 0 d \equiv 1 (4) d \equiv 3 (4) d \equiv 0 (2) .

⟳count Gaussian integers of norm

n

by factoring

n

into Gaussian primes and tracking the four units; the divisor sum tracks split-prime exponents.

■

Carl Friedrich Gauss | 1777–1855

Gauss introduced the integers $a + bi$ in his 1832 memoir on biquadratic reciprocity, proving their unique factorization and using them to formulate the biquadratic reciprocity law—a problem he could not state cleanly within $Z$ alone. His 1801 Disquisitiones Arithmeticae had already developed the genus theory of binary quadratic forms, the form-theoretic shadow of quadratic field arithmetic. The "Gaussian integers" bear his name precisely because he recognized that enlarging the number system was the natural setting for reciprocity. Dedekind and Dirichlet later recast this form-language into the modern ideal-and-field framework.

Building Intuition

Think of $Z [i]$ as the square lattice in the complex plane, with the norm measuring squared distance to the origin. Euclidean division "rounds to the nearest lattice point," and the bound $N (r) \leq \frac{1}{2} N (β)$ holds geometrically because any point in a unit cell is within distance $\frac{1}{2}$ of a corner. A rational prime $p$ either fragments into two lattice points (split), sits at a doubled lattice point (ramified, only $p = 2$ ), or remains atomic because no lattice point has norm $p$ (inert). The whole story of which primes are sums of two squares is just asking: does the circle of radius $p$ pass through a lattice point?

Worked Examples

:::example[Factoring 65 in $Z [i]$ ] Given: $65 = 5 \cdot 13$ , with $5 \equiv 1$ and $13 \equiv 1 (mod 4)$ .

Step 1.

5 = (2 + i) (2 - i)

, since

2^{2} + 1^{2} = 5

. Step 2.

13 = (3 + 2 i) (3 - 2 i)

, since

3^{2} + 2^{2} = 13

. Step 3. Combine:

65 = (2 + i) (3 + 2 i) \cdot (2 - i) (3 - 2 i)

. Compute

(2 + i) (3 + 2 i) = 6 + 4 i + 3 i + 2 i^{2} = 4 + 7 i

, and

(2 + i) (3 - 2 i) = 6 - 4 i + 3 i + 2 = 8 - i

. Step 4. This yields the two essentially different representations

65 = N (4 + 7 i) = 4^{2} + 7^{2}, 65 = N (8 - i) = 8^{2} + 1^{2} .

⟳distinct pairings of conjugate split primes give distinct norm-form representations.

65 = 4^{2} + 7^{2} = 8^{2} + 1^{2}

(two representations, confirming

r_{2}

-style multiplicity).

Computing $D_K$ and $\mathcal{O}_K$ for $\mathbb{Q}(\sqrt{5})$

Given:

d = 5

, which is squarefree and

5 \equiv 1 (mod 4)

. Step 1. By [3],

O_{K} = Z [\frac{1 + 5}{2}]

. The generator

ω = \frac{1 + 5}{2}

(the golden ratio) is integral, satisfying

ω^{2} - ω - 1 = 0

. Step 2. By [4], since

d \equiv 1 (mod 4)

, the discriminant is

D_{K} = 5

. Step 3. Check via the trace form on basis

{1, ω}

Tr (1) = 2

Tr (ω) = 1

Tr (ω^{2}) = Tr (ω + 1) = 3

D_{K} = det (2113) = 6 - 1 = 5.

⟳the discriminant equals

det (Tr (ω_{i} ω_{j}))

; entries are

Tr (1 \cdot 1), Tr (1 \cdot ω), Tr (ω \cdot ω)

O_{K} = Z [\frac{1 + 5}{2}], D_{K} = 5

Common Misconceptions

The ring of integers is NOT always $Z [d]$ . For

d \equiv 1 (mod 4)

you must use the half-integer basis

\frac{1 + d}{2}

. Using

Z [5]

instead of

Z [\frac{1 + 5}{2}]

gives an order of index 2, with wrong discriminant (

20

5

) and a non-maximal, badly-behaved ideal theory. Norm being prime is sufficient but not necessary for an element to be prime. If

N (α)

is a rational prime then

α

is a Gaussian prime; but

α = 3

is a Gaussian prime with

N (3) = 9

composite. Inert primes have composite (square) norm.

Unique factorization can fail

In $Z [- 5] = O_{Q (- 5)}$ (here $- 5 \equiv 3 (mod 4)$ so this is the full ring of integers),

6 = 2 \cdot 3 = (1 + - 5) (1 - - 5),

and all four factors are irreducible and non-associate, with norms

4, 9, 6, 6

⟳check no element has norm 2 or 3 (i.e.

a^{2} + 5 b^{2} \in {2, 3}

is insoluble), so the factors cannot be split further or matched up.

The class number is $h (Q (- 5)) = 2 > 1$ , measuring exactly this failure. Unlike $Z [i]$ , this ring is a Dedekind domain but not a PID—ideals still factor uniquely even though elements do not.

Connections

Quadratic fields are the degree-2 case of Algebraic Number Fields, and their integers are the simplest Dedekind Domains where Unique Factorization of Ideals replaces element factorization. The split/inert/ramified trichotomy is governed by Quadratic Reciprocity through the Kronecker symbol $(\frac{D _{K}}{p})$ . The failure of unique factorization is measured by the Ideal Class Group, whose order is the class number $h$ , and the structure of units in real quadratic fields is the content of Dirichlet's Unit Theorem and the Pell Equation. Imaginary quadratic fields connect to Complex Multiplication and the theory of elliptic curves with CM.

Real-World Applications

Cryptography (Gaussian integer RSA variants): factorization difficulty in $Z [i]$ underlies certain public-key proposals and lattice-based schemes.
Coding theory: Gaussian-integer and Eisenstein-integer lattices yield efficient signal constellations (QAM) and error-correcting codes over $Z [i] / π$ .
Sums of two squares algorithms: Cornacchia's algorithm uses Gaussian arithmetic to represent primes $p \equiv 1 (mod 4)$ as $a^{2} + b^{2}$ , used in primality proving (ECPP).
Diophantine equation solving: the equation $y^{2} = x^{3} - 1$ and many Mordell equations are solved by factoring over $Z [i]$ or $Z [ω]$ .
Crystallography and lattice physics: the Gaussian and Eisenstein lattices model 2D periodic structures and quasi-crystal substructures.
Fast multiplication / FFT over finite fields: primes $p \equiv 1 (mod 4)$ admit square roots of $- 1$ , enabling complex-style number-theoretic transforms.
Pell-equation-based continued fraction algorithms: real quadratic units drive efficient solutions used in computational number theory.

Historical Development

1640 — Fermat states (in a letter to Mersenne) that every prime $p \equiv 1 (mod 4)$ is a sum of two squares, without published proof.
1749–1770 — Euler proves Fermat's two-square theorem and develops infinite descent and quadratic residue tools.
1798 — Legendre formulates quadratic reciprocity (with an incomplete proof) and the Legendre symbol.
1801 — Gauss publishes Disquisitiones Arithmeticae, giving the first complete proofs of quadratic reciprocity and the genus theory of binary quadratic forms.
1832 — Gauss introduces $Z [i]$ to state biquadratic reciprocity, proving its unique factorization.
1840s — Dirichlet proves the unit theorem and the class number formula for quadratic forms; Eisenstein develops $Z [ω]$ for cubic reciprocity.
1846–1871 — Kummer's ideal numbers (for cyclotomic fields) inspire Dedekind's general theory.
1871 — Dedekind defines ideals and integral closure, recasting quadratic-field arithmetic as the theory of Dedekind domains in his supplements to Dirichlet's lectures.
1900s — Hilbert's Zahlbericht and class field theory systematize quadratic fields as the genesis of all abelian extension theory.

Subtleties & Limits

Class number problems remain partly open. Gauss conjectured (proved by Heilbronn 1934, effectivized by Baker–Stark 1966) that there are exactly nine imaginary quadratic fields with class number 1:

d = - 1, - 2, - 3, - 7, - 11, - 19, - 43, - 67, - 163

. By contrast, it is still unknown whether infinitely many real quadratic fields have class number 1—the Gauss class number 1 problem for real fields is open. The Cohen–Lenstra heuristics predict that roughly

75.4%

of real quadratic fields are principal, but no proof exists. Effective bounds on class numbers and the precise distribution of split versus inert primes (linked to the Generalized Riemann Hypothesis through

L (1, χ)

) remain active frontiers.