Covariant (Invariant Theory)¶

A polynomial map between group representations that transforms equivariantly, carrying the symmetry action on its input into the corresponding action on its output.

Version: v3 · 2026-09-19 · History
Domain-specific #: 1573
Domain group: Formal Sciences
Origin domain: Mathematics
Subdomain: Invariant Theory → Mathematics
Aliases: Invariant-theoretic covariant, Classical covariant, Module of covariants

Core Idea¶

In invariant theory, a covariant is a polynomial construction that respects a group action. Given a group \(G\), representations \(V\) and \(W\), and a polynomial map \(F:V\to W\), the map is a covariant when

\[ F(g\cdot v)=g\cdot F(v) \]

for every admissible \(g\) and \(v\). Thus transforming the input form and then applying the construction produces the same result as constructing first and transforming the output. The object combines polynomial dependence with equivariance.^[1]

Classically, \(V\) is a space of binary forms and \(W\) another space of forms. A covariant is polynomial in the input coefficients and in the output variables; its degree records coefficient degree and its order records degree in the output variables. An invariant is the order-zero/scalar-output special case. Hessians, discriminants, Jacobians, and transvectants supply canonical examples and construction operations.^[2]

The recognition invariant is group representations + polynomial map + compatible input/output actions + equivariance identity + declared grading + invariant-theoretic interpretation or generation.

How would you explain it like I'm…

Same Either Order Recipe

Think of a recipe that turns one shape of numbers into another using only adding and multiplying. It is a covariant if it doesn't matter which you do first: turn or stretch the input, then use the recipe, or use the recipe, then turn or stretch the answer the same way. Either order gives the same result.

Formulas That Move with the Change

In a part of math called invariant theory, people study formulas that behave nicely when you change the viewpoint, for example by rotating or stretching. A covariant is a formula that takes an algebraic expression and builds a new expression from it, using only multiplying and adding its numbers. The special property is that it 'moves along' with the change: if you transform the input and then apply the formula, you get the same answer as applying the formula first and then transforming the output. When the output is just a single number that doesn't move at all, it's called an invariant, a special kind of covariant.

Equivariant Polynomial Map

In invariant theory, a group G acts on spaces V and W. A covariant is a polynomial map F from V to W that is compatible with the group actions: F(g . v) = g . F(v) for every g and v. So transforming first and then applying F gives the same result as applying F and then transforming. This property is called equivariance. Classically, V is a space of binary forms (homogeneous polynomials in two variables), and a covariant produces another form whose coefficients are polynomials in the original coefficients. Its degree counts its degree in the input coefficients and its order counts its degree in the output variables. An invariant is the special case of order zero, where the output is a scalar. Hessians, discriminants, Jacobians and transvectants are standard examples and tools for building new covariants.

Given a group G and representations V and W, a covariant is a polynomial map F: V -> W that is G-equivariant, F(g . v) = g . F(v) for all admissible g and v, combining polynomial dependence with equivariance. In the classical theory of binary forms, V and W are spaces of forms, and a covariant is polynomial both in the coefficients of the input form and in the output variables. It is graded by degree (in the input coefficients) and order (in the output variables); invariants are the order-zero, scalar-valued special case. The Hessian, discriminant and Jacobian give canonical examples, and transvectants are an operation for constructing covariants from others. Recognizing a covariant requires the group representations, the polynomial map, compatible input and output actions, the equivariance identity, a declared grading, and an invariant-theoretic interpretation or generation scheme.

Structural Signature¶

Acting group: algebraic or linear group \(G\), often \(SL_2\), \(GL_n\), or a reductive group.
Input representation: vector space or affine representation \(V\) whose points encode forms or tensors.
Output representation: representation \(W\) carrying the result.
Polynomial dependence: coordinate functions of \(F\) lie in the coordinate ring of \(V\).
Equivariance law: \(F\circ \rho_V(g)=\rho_W(g)\circ F\).
Homogeneity/grading: coefficient degree, output order, weight, or multidegree.
Covariant module/algebra: covariants organized over the invariant ring with addition and multiplication where defined.
Generators and relations: finite systems, syzygies, or transvectant constructions.
Vanishing and orbit behavior: evaluations reflect stabilizers, null cones, degeneracy, or orbit structure.
Coordinate–object distinction: formulas depend on chosen coordinates while equivariance certifies the represented construction.

What It Is Not¶

It is not a covariant vector or lower-index tensor merely because the word “covariant” appears. Tensor covariance concerns component transformation under basis change; the invariant-theory object is a polynomial equivariant map between representation spaces.

It is not arbitrary equivariance. A continuous neural-network layer, set map, or differential operator can be equivariant without being a polynomial covariant in invariant theory. It is not necessarily an invariant: scalar-valued covariants with trivial output action are invariants, while positive-order covariants retain output variables. “Covariant” also should not be confused with a coefficient of covariance in probability.

Scope of Application¶

Covariants encode canonical constructions on binary and higher forms, classify orbits, detect singularities and degeneracies, describe moduli, and support explicit calculations in algebraic geometry and representation theory. Classical work seeks finite generating families and relations; computational invariant theory turns those questions into Gröbner-basis, Reynolds-operator, and module calculations.^[3]

The concept extends beyond binary forms to multiple forms, tensors, matrices, quivers, and algebraic representations. Exact finiteness and algorithmic guarantees depend on the base field, characteristic, and group; reductivity assumptions that are harmless over characteristic zero may fail elsewhere.

Clarity¶

The action on both spaces must be stated. The same coordinate formula can be covariant for one pair of actions and fail for another. If \(W\) has the trivial action, equivariance reduces to invariance; if \(W=V\), it says the polynomial self-map commutes with the action.

In classical notation, degree and order answer different questions. Coefficient degree measures how the construction depends on the input form; order measures degree in the auxiliary/output variables. Weight and character conventions can introduce additional gradings, so a draft must declare its convention rather than infer it from terminology.

Manages Complexity¶

Equivariance replaces infinitely many coordinate-change checks with one commuting relation. Grading partitions a large covariant algebra into finite-dimensional pieces. Generators compress every covariant into polynomial combinations of a finite set when a finiteness theorem applies, while syzygies record nonunique presentations.

This organization also exposes hardness. Generator sets can be large, degree bounds severe, and positive-characteristic behavior subtle. A computational list is meaningful only with the group, representation, base field, grading, and completeness claim attached.

Abstract Reasoning¶

Specify the base field, group, and input/output representations.
Write the two actions explicitly and identify any characters or weights.
Propose a polynomial map from coefficients or intrinsic operations.
Verify the equivariance identity symbolically or representation-theoretically.
Determine coefficient degree, output order, multidegree, and parity constraints.
Test evaluations on representative orbits, stabilizers, and degenerate forms.
Generate additional covariants through products, contractions, polarization, or transvection.
Establish generators and relations with theorems or auditable computation.
Separate coordinate normalizations from intrinsic conclusions.

Knowledge Transfer¶

The portable insight is to require a derived object to transform in step with its source. That is precisely Equivariance. What the specialist abstraction adds is polynomial algebra on representation spaces, invariant rings/modules, classical gradings, and generator questions.

The proposed immediate parent is therefore Equivariance; the covariant is a mathematically rich specialization, not an alias for every equivariant map.

Examples¶

Hessian. The Hessian of a binary form is another binary form whose coefficients are polynomial in the original coefficients and whose transformation follows the induced group action. It is a positive-order covariant.^[4]

Discriminant. A binary form’s discriminant is scalar-valued (or relative-invariant-valued under a character). It is the invariant special case rather than evidence that all covariants are scalars.

Non-example. A polynomial feature map chosen without checking the group actions is not a covariant even if it performs well numerically.

Structural Tensions¶

Coordinate formulas versus intrinsic equivariance.
Compact generators versus complicated syzygies.
Classical characteristic-zero theory versus modular invariant theory.
Absolute invariants versus relative/character-valued behavior.
Symbolic completeness versus computational feasibility.
One terminology versus tensor and statistical homonyms.

Structural–Framed Character¶

The group actions, polynomial map, commuting identity, and grading are structural. Choice of form space, field, group, normalization, and computational representation is mathematically framed.

Structural Core vs. Domain Accent¶

The portable core is transformation-compatible construction. Polynomial representations, binary forms, invariant rings, transvectants, degrees/orders, and syzygies are constitutive invariant-theory accent, so the node is domain-specific.

This entry is a kind of Equivariance.

Equivariance is the broader abstraction this entry instantiates. Polynomial Function, Group Action, Representation, Tensor, and Geometric Transformation are related; none covers the complete invariant-theory identity.

Relationships to Other Abstractions¶

Current abstraction Covariant (Invariant Theory) Domain-specific

Parents (1) — outgoing typed relations

Covariant (Invariant Theory) is a kind of Equivariance Prime

Equivariance is the broader abstraction this entry instantiates.

Hierarchy paths (3) — routes to 3 parentless roots

Covariant (Invariant Theory) → is a kind of Equivariance → is a kind of Invariance

Show alternative paths (2)

Neighborhood in Abstraction Space¶

Covariant (Invariant Theory) sits in a sparse region of the domain-specific corpus (88^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Group-Theoretic Subgroup & Cohomology Structures (7 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-25

Not to Be Confused With¶

Covariant and contravariant tensor components.
Covariance of random variables.
A generic equivariant map lacking polynomial invariant-theory structure.
An invariant, except as the scalar-output special case.
Coordinate covariance asserted without specified group actions.
A particular generating set treated as the definition.

References¶

[1] Peter J. Olver, Classical Invariant Theory, Cambridge University Press, 1999. registry ↩a ↩b

[2] Igor V. Dolgachev, Lectures on Invariant Theory, London Mathematical Society Lecture Note Series 296, Cambridge University Press, 2003. registry ↩ Show verification details

[3] Harm Derksen and Gregor Kemper, Computational Invariant Theory, 2^nd ed., Springer, 2015. DOI 10.1007/978-3-662-48422-7. registry ↩ Show verification details

[4] David A. Craven, review discussion of classical invariant theory, Bulletin of the American Mathematical Society 36, 1999, including the equivariant-polynomial definition and Hessian example. withdrawn registry ↩