Gauss Composition: Definition, Theory, and Applications

• Gauss composition is the arithmetic operation on equivalence classes of binary quadratic forms with a common discriminant, corresponding to ideal multiplication in quadratic orders, and involves forms f(x,y)=ax^2+bxy+cy^2 with discriminant Δ=b2−4ac.
• The composition maps transformations of quadratic forms induced by PSL2(Z) to multiplication of ideals, preserving discriminants and forming inverse elements when b reflects the change of orientation.
• Modern formulations extend Gauss composition to nonprimitive forms, degenerate cases, and arbitrary base schemes and use techniques from algebraic geometry and modular functions, for advanced topics in algebraic number theory.

Gauss composition is the arithmetic operation on equivalence classes of binary quadratic forms having a common discriminant. In its classical form, it transports multiplication of ideal classes in a quadratic order to forms

$f(x,y)=ax^2+bxy+cy^2,\qquad \Delta(f)=b^2-4ac.$

For primitive forms, proper $\mathrm{SL}_2(\mathbb Z)$-classes of fixed discriminant form an abelian group; the identity is the principal form, inversion changes $b$ to $-b$, and composition corresponds to multiplication of oriented invertible ideals. Modern formulations extend this correspondence to arbitrary base schemes, degenerate and nonprimitive forms, traceable modules, moduli of double covers, congruence-level class groups, and higher composition laws.

1. Classical arithmetic formulation

A binary quadratic form over $\mathbb Z$ is a polynomial

$f(x,y)=ax^2+bxy+cy^2$

with discriminant

$\Delta=b^2-4ac.$

A form is primitive when $\gcd(a,b,c)=1$. Proper equivalence is induced by the right action of $\mathrm{SL}_2(\mathbb Z)$: $$f\sim f\circ \gamma,\qquad \gamma\in\mathrm{SL}_2(\mathbb Z).$$ For negative nonsquare discriminant, positive-definite primitive forms of discriminant $\mathrm{SL}_2(\mathbb Z)$0 correspond to ideal classes of the quadratic order of discriminant $\mathrm{SL}_2(\mathbb Z)$1. For positive nonsquare discriminant, the corresponding group is the narrow ideal class group. The orientation retained by $\mathrm{SL}_2(\mathbb Z)$2-equivalence distinguishes a class from its inverse; passing to ordinary $\mathrm{SL}_2(\mathbb Z)$3-equivalence identifies these orientations.

To a form $\mathrm{SL}_2(\mathbb Z)$4 one associates a quadratic order and a rank-two ideal, for example

$\mathrm{SL}_2(\mathbb Z)$5

The central identity is

$\mathrm{SL}_2(\mathbb Z)$6

The composed form is the form corresponding to the product ideal $\mathrm{SL}_2(\mathbb Z)$7. In this sense, Gauss composition is ideal multiplication transported through the form–ideal correspondence.

When suitable coprimality and congruence conditions hold, the coefficient formula can be written

$\mathrm{SL}_2(\mathbb Z)$8

$\mathrm{SL}_2(\mathbb Z)$9

where

$b$0

and

$b$1

The coefficient formula is a concrete realization of ideal multiplication, but the ideal-theoretic construction is more robust: it applies to nonmaximal orders, arbitrary quadratic orders, and settings in which no convenient common leading coefficient exists.

Gauss’s original notation often writes

$b$2

In that convention, $b$3 is one quarter of the modern discriminant. The distinction between proper and improper equivalence, and between $b$4 and $b$5, is essential for even and imprimitive forms. Generalized composition can allow forms with different determinants, provided their determinants differ from the determinant of the composite by rational square factors and satisfy the required content and integrality conditions (Celis et al., 2021).

2. Forms, quadratic algebras, and traceable modules

The base-scheme formulation replaces integer coefficients by vector bundles. Let $b$6 be a scheme. A linear binary quadratic form consists of a rank-two locally free $b$7-module $b$8, a rank-one locally free module $b$9, and a section

$-b$0

Locally, after choosing bases $-b$1 of $-b$2 and $-b$3 of $-b$4,

$-b$5

The coefficient line $-b$6 permits the value line of the form to vary over the base. The ordinary formulation corresponds to $-b$7, while the twisted formulation relevant to oriented forms takes

$-b$8

A quadratic algebra over $-b$9 is a rank-two locally free $\mathbb Z$0-algebra $\mathbb Z$1. A $\mathbb Z$2-module $\mathbb Z$3 is traceable if it is rank two over $\mathbb Z$4 and the trace of the action of every element of $\mathbb Z$5 on $\mathbb Z$6 agrees with its trace under multiplication on $\mathbb Z$7. Traceability generalizes the condition that a module be a full ideal. It remains meaningful over nonreduced bases, rings with zero divisors, and degenerate quadratic algebras.

The fundamental correspondence is

$\mathbb Z$8

It is functorial in $\mathbb Z$9 and commutes with arbitrary base change. Under the correspondence,

$f(x,y)=ax^2+bxy+cy^2$0

Thus the quotient line of the quadratic algebra records the determinant data of the form.

Locally, put

$f(x,y)=ax^2+bxy+cy^2$1

and let

$f(x,y)=ax^2+bxy+cy^2$2

The algebra and module structures are

$f(x,y)=ax^2+bxy+cy^2$3

$f(x,y)=ax^2+bxy+cy^2$4

The matrix of multiplication by $f(x,y)=ax^2+bxy+cy^2$5 is

$f(x,y)=ax^2+bxy+cy^2$6

whose trace is $f(x,y)=ax^2+bxy+cy^2$7, equal to the trace of $f(x,y)=ax^2+bxy+cy^2$8 in $f(x,y)=ax^2+bxy+cy^2$9. The discriminant of the algebra is

$\Delta=b^2-4ac.$0

and the canonical identification

$\Delta=b^2-4ac.$1

identifies it with the discriminant of the form. Consequently, the correspondence is discriminant preserving (Wood, 2010).

The inverse construction starts with $\Delta=b^2-4ac.$2 and sets

$\Delta=b^2-4ac.$3

The module action defines a map

$\Delta=b^2-4ac.$4

which, after dualization, produces the associated linear binary quadratic form.

3. Geometric and moduli-theoretic realization

The correspondence has a global geometric construction. For the projective bundle

$\Delta=b^2-4ac.$5

the form determines a morphism

$\Delta=b^2-4ac.$6

The quadratic algebra and its module are obtained from the hypercohomology of the complexes

$\Delta=b^2-4ac.$7

and

$\Delta=b^2-4ac.$8

When $\Delta=b^2-4ac.$9 is not a zero divisor, these complexes recover the structure sheaf of the zero locus and its relative $\gcd(a,b,c)=1$0: $\gcd(a,b,c)=1$1 The derived construction is necessary in degenerate cases. If $\gcd(a,b,c)=1$2, its ordinary zero locus is all of $\gcd(a,b,c)=1$3 and its global functions give only $\gcd(a,b,c)=1$4, whereas the hypercohomological construction still produces the correct rank-two algebra.

The correspondence is an equivalence of moduli stacks. After bases are chosen, both sides are parametrized by the coefficient space $\gcd(a,b,c)=1$5 with compatible $\gcd(a,b,c)=1$6 and $\gcd(a,b,c)=1$7 actions. Taking quotient stacks yields an equivalence between the moduli stack of linear binary quadratic forms and the moduli stack of quadratic algebras equipped with traceable modules. This stack-theoretic formulation explains its compatibility with arbitrary base change.

The discriminant has a geometric interpretation. A linear binary quadratic form defines a double cover, and its discriminant is the branch section. For a form on $\gcd(a,b,c)=1$8,

$\gcd(a,b,c)=1$9

the associated cover is reduced precisely when $\mathrm{SL}_2(\mathbb Z)$0 is not identically zero, integral when that discriminant is not a square, and smooth when the ordinary discriminant of the discriminant form is nonzero. The module is a line bundle when $\mathrm{SL}_2(\mathbb Z)$1 is nonzero in every fiber.

For a splitting type

$\mathrm{SL}_2(\mathbb Z)$2

the coefficient degrees are

$\mathrm{SL}_2(\mathbb Z)$3

$\mathrm{SL}_2(\mathbb Z)$4

$\mathrm{SL}_2(\mathbb Z)$5

The associated double covering pair has arithmetic genus and module degree

$\mathrm{SL}_2(\mathbb Z)$6

These numerical relations yield quotient-stack strata describing the universal Jacobian of degree-$\mathrm{SL}_2(\mathbb Z)$7 line bundles on genus-$\mathrm{SL}_2(\mathbb Z)$8 double covers of $\mathrm{SL}_2(\mathbb Z)$9 (Erman et al., 2011).

4. Primitivity, invertibility, and composition

A linear form is primitive when its coefficients generate the unit ideal locally. The base-scheme correspondence identifies primitivity with invertibility: $$f\sim f\circ \gamma,\qquad \gamma\in\mathrm{SL}_2(\mathbb Z).$$0 Here invertible means locally free of rank one over $$f\sim f\circ \gamma,\qquad \gamma\in\mathrm{SL}_2(\mathbb Z).$$1. If $$f\sim f\circ \gamma,\qquad \gamma\in\mathrm{SL}_2(\mathbb Z).$$2 is invertible, one may normalize to $$f\sim f\circ \gamma,\qquad \gamma\in\mathrm{SL}_2(\mathbb Z).$$3, so that $$f\sim f\circ \gamma,\qquad \gamma\in\mathrm{SL}_2(\mathbb Z).$$4 and $$f\sim f\circ \gamma,\qquad \gamma\in\mathrm{SL}_2(\mathbb Z).$$5 is generated by $$f\sim f\circ \gamma,\qquad \gamma\in\mathrm{SL}_2(\mathbb Z).$$6 as a $$f\sim f\circ \gamma,\qquad \gamma\in\mathrm{SL}_2(\mathbb Z).$$7-module. Covering the base by loci on which $$f\sim f\circ \gamma,\qquad \gamma\in\mathrm{SL}_2(\mathbb Z).$$8, $$f\sim f\circ \gamma,\qquad \gamma\in\mathrm{SL}_2(\mathbb Z).$$9, or $\mathrm{SL}_2(\mathbb Z)$00 is invertible gives the global assertion.

For an integral base, a full ideal is a rank-two ideal locally free over the base. Full ideals are traceable modules, and for nondegenerate objects every traceable module becomes, after suitable scaling, a full ideal. Two full ideals represent the same ideal class exactly when they are isomorphic as modules. Thus, in the nondegenerate integral setting,

$\mathrm{SL}_2(\mathbb Z)$01

When the modules are invertible, multiplication is

$\mathrm{SL}_2(\mathbb Z)$02

or equivalently tensor product followed by the natural identification

$\mathrm{SL}_2(\mathbb Z)$03

Transporting this operation to forms gives Gauss composition.

The distinction between invertible modules and arbitrary traceable modules is essential. Primitive forms correspond to invertible modules and recover the group-like composition law. Nonprimitive forms correspond to traceable modules that need not be invertible; the general classification therefore cannot be stated solely in terms of ideal classes.

Degenerate forms are included. When

$\mathrm{SL}_2(\mathbb Z)$04

the associated quadratic algebra may have nilpotents. The zero form corresponds to

$\mathrm{SL}_2(\mathbb Z)$05

with $\mathrm{SL}_2(\mathbb Z)$06 acting trivially on

$\mathrm{SL}_2(\mathbb Z)$07

This traceable module is not an ideal of $\mathrm{SL}_2(\mathbb Z)$08. An ideal-only formulation would therefore omit valid forms, whereas traceable modules provide the complete parametrization. The formulation also remains valid under specialization, including reduction modulo primes where a nondegenerate form can become degenerate (Wood, 2010).

A related orthogonal formulation identifies the quadratic algebra canonically with the even Clifford algebra. For a binary quadratic module $\mathrm{SL}_2(\mathbb Z)$09, the even Clifford algebra $\mathrm{SL}_2(\mathbb Z)$10 is quadratic and satisfies

$\mathrm{SL}_2(\mathbb Z)$11

Its odd Clifford module is pseudoregular, equivalently traceable in the quadratic case. The norm construction from a pseudoregular module is quasi-inverse to the Clifford construction. Primitive forms correspond to invertible modules, and tensor product of invertible modules becomes Gauss composition. This gives a discriminant-preserving categorical explanation of the classical law (Voight et al., 6 Nov 2025).

5. Congruence refinements and class-field theory

Classical composition can be refined by imposing congruence conditions on forms. Let $\mathrm{SL}_2(\mathbb Z)$12 be an imaginary quadratic field with fundamental discriminant $\mathrm{SL}_2(\mathbb Z)$13, and let

$\mathrm{SL}_2(\mathbb Z)$14

Consider primitive positive-definite forms of discriminant $\mathrm{SL}_2(\mathbb Z)$15 whose leading coefficients are relatively prime to $\mathrm{SL}_2(\mathbb Z)$16: $\mathrm{SL}_2(\mathbb Z)$17 The associated ideals are relatively prime to $\mathrm{SL}_2(\mathbb Z)$18.

Instead of ordinary proper equivalence, one uses a congruence subgroup $\mathrm{SL}_2(\mathbb Z)$19 of level $\mathrm{SL}_2(\mathbb Z)$20. For a subgroup

$\mathrm{SL}_2(\mathbb Z)$21

the corresponding subgroup of principal ideals consists of those generated by elements congruent modulo $\mathrm{SL}_2(\mathbb Z)$22 to an element of $\mathrm{SL}_2(\mathbb Z)$23. The resulting form classes are identified with

$\mathrm{SL}_2(\mathbb Z)$24

where $\mathrm{SL}_2(\mathbb Z)$25 is the group of fractional ideals relatively prime to $\mathrm{SL}_2(\mathbb Z)$26 and $\mathrm{SL}_2(\mathbb Z)$27 is the relevant congruence-sensitive subgroup of principal ideals.

The generalized composition law is transported from ideal multiplication: $\mathrm{SL}_2(\mathbb Z)$28 For $\mathrm{SL}_2(\mathbb Z)$29, the congruence subgroup is $\mathrm{SL}_2(\mathbb Z)$30 and the quotient is the ray class group: $\mathrm{SL}_2(\mathbb Z)$31 For $\mathrm{SL}_2(\mathbb Z)$32, the subgroup is $\mathrm{SL}_2(\mathbb Z)$33 and the quotient is the ring class group: $\mathrm{SL}_2(\mathbb Z)$34

The Galois correspondence can be made explicit with modular functions. If $\mathrm{SL}_2(\mathbb Z)$35 denotes the field of meromorphic modular functions for $\mathrm{SL}_2(\mathbb Z)$36 with rational Fourier coefficients, Shimura reciprocity gives

$\mathrm{SL}_2(\mathbb Z)$37

The form represented by $\mathrm{SL}_2(\mathbb Z)$38 determines an ideal class and hence an Artin automorphism characterized by

$\mathrm{SL}_2(\mathbb Z)$39

for suitable modular functions $\mathrm{SL}_2(\mathbb Z)$40. Thus congruence-refined Gauss composition simultaneously describes generalized form classes, ray or ring class groups, and Galois groups of class fields (Jung et al., 2019).

6. Geometric, higher, and computational extensions

Gauss composition has geometric realizations beyond quadratic orders. For double covers of $\mathrm{SL}_2(\mathbb Z)$41, a linear binary quadratic form parametrizes a double cover together with a traceable module. On a smooth cover, traceable modules are line bundles; on a reduced singular cover, they are rank-one torsion-free sheaves. Tensor product of line bundles on a fixed cover gives the geometric composition law and realizes the group structure of the universal Jacobian.

Allowing singular and nonreduced covers produces a stack of generalized line bundles. A bounded open substack defined by

$\mathrm{SL}_2(\mathbb Z)$42

is smooth, irreducible, unirational, and universally closed, contains the universal Jacobian as a dense open substack, and is generally not separated. This compactification preserves the degree-two map to $\mathrm{SL}_2(\mathbb Z)$43 rather than replacing the limiting object by a stable curve (Erman et al., 2011).

The Bhargava cube gives a geometric realization of composition through planes in the space of $\mathrm{SL}_2(\mathbb Z)$44 matrices. To a form

$\mathrm{SL}_2(\mathbb Z)$45

associate the traceless matrix

$\mathrm{SL}_2(\mathbb Z)$46

Given two appropriate matrices $\mathrm{SL}_2(\mathbb Z)$47, form the rank-two plane

$\mathrm{SL}_2(\mathbb Z)$48

If $\mathrm{SL}_2(\mathbb Z)$49 is an oriented basis of this plane, its determinant polynomial

$\mathrm{SL}_2(\mathbb Z)$50

represents the composed class: $\mathrm{SL}_2(\mathbb Z)$51 in the primitive case. Using $\mathrm{SL}_2(\mathbb Z)$52 as the two slices of a $\mathrm{SL}_2(\mathbb Z)$53 cube produces the same composition through Bhargava’s cube law (Aka et al., 2023).

The same mechanism underlies higher composition laws. Binary cubic forms, pairs of binary quadratic forms, pairs of quaternary alternating $\mathrm{SL}_2(\mathbb Z)$54-forms, and senary alternating $\mathrm{SL}_2(\mathbb Z)$55-forms each possess a discriminant, a projectivity condition, and a parametrization by ideal data over a quadratic ring. A companion form $\mathrm{SL}_2(\mathbb Z)$56 combines with $\mathrm{SL}_2(\mathbb Z)$57 to produce an element

$\mathrm{SL}_2(\mathbb Z)$58

of the quadratic ring. Multiplication gives the coefficient operation

$\mathrm{SL}_2(\mathbb Z)$59

where

$\mathrm{SL}_2(\mathbb Z)$60

The resulting polynomial identities are higher analogues of Gauss composition, with bilinear or multilinear substitutions encoding multiplication of the associated ideals (Chinta et al., 6 Feb 2026).

For binary cubic forms, Bhargava’s parametrization uses pairs $\mathrm{SL}_2(\mathbb Z)$61 satisfying

$\mathrm{SL}_2(\mathbb Z)$62

The ideal class of $\mathrm{SL}_2(\mathbb Z)$63 lies in the $\mathrm{SL}_2(\mathbb Z)$64-torsion of the quadratic order’s class group, while units modulo cubes can contribute additional classes. Composition is induced by

$\mathrm{SL}_2(\mathbb Z)$65

and becomes a group law only on $\mathrm{SL}_2(\mathbb Z)$66-equivalence classes (Nativi, 2019).

An algorithmic interpretation identifies Gauss composition with inversion of the wedge map

$\mathrm{SL}_2(\mathbb Z)$67

For $\mathrm{SL}_2(\mathbb Z)$68, decomposability is the Pfaffian relation

$\mathrm{SL}_2(\mathbb Z)$69

Given two forms of common discriminant, their coefficients determine a decomposable bivector. Inverting the wedge map recovers two integral vectors, which become slices of a Bhargava cube; the remaining slicing yields the composed form. The inversion algorithm uses one gcd computation, Bézout coefficients, a congruence solution, and verification through the Plücker relations (Chua, 2024).

Across these formulations, the invariant mechanism is unchanged: $\mathrm{SL}_2(\mathbb Z)$70 Primitivity or projectivity ensures invertibility, orientation preserves the class-group structure, discriminant compatibility permits composition, and the resulting product is transported back to a form. The classical composition of binary quadratic forms is therefore the rank-two instance of a broader algebraic, geometric, categorical, and arithmetic theory.