Algebraic Cycle¶

A dimension-graded formal integer sum of integral closed subvarieties that can be compared by geometric equivalence and assembled into Chow groups.

Version: v2 · 2026-09-06 · History
Domain-specific #: 1261
Origin domain: algebraic geometry
Subdomain: intersection theory

Core Idea¶

An algebraic cycle on an algebraic variety or scheme is a finite, or in the appropriate locally finite setting locally finite, formal integer-weighted sum of its integral closed subvarieties. If the generators have dimension \(k\), a cycle has the form

\[ \alpha=\sum_Z n_Z[Z],\qquad n_Z\in\mathbb Z, \]

where only finitely many coefficients are nonzero in the ordinary finite-type case. The brackets denote the generator associated with the subvariety; they do not mean that its points have been numerically added. Cycles are graded by dimension or, on an equidimensional ambient space, by codimension. This grading keeps divisors, zero-cycles, curves, and higher-dimensional cycles in the degrees where their geometric operations make sense.^[1]

The construction becomes especially powerful when cycles are compared under a specified equivalence relation. Rational equivalence identifies differences generated by divisors of rational functions on one-dimension-larger subvarieties. Quotienting the group of cycles by rational equivalence gives a Chow group. On suitably regular spaces, intersection products organize these groups into a Chow ring. Other equivalence relations, such as algebraic, homological, or numerical equivalence, answer different questions and must not be silently substituted for rational equivalence.^[2]^[3]

The abstraction is therefore not merely “a subvariety” and not merely “a linear combination.” It is a geometric accounting system with four linked commitments: integral geometric generators, integer multiplicities, degree control, and an explicit rule for when formally different sums express the same geometric class. That package lets algebraic geometry turn families of subvarieties into algebraic invariants while retaining enough geometry for pullback, pushforward, intersection, and comparison with cohomology.

Structural Signature¶

The ambient algebraic space — a variety, scheme, or algebraic space on which closed integral subspaces and their dimensions are defined.
The geometric generators — integral closed subvarieties \(Z\), recorded as symbols \([Z]\).
The integer multiplicities — coefficients \(n_Z\in\mathbb Z\); negative coefficients are formal differences, not negative physical subspaces.
The dimension or codimension grade — all generators in a homogeneous cycle share the prescribed degree.
The free abelian cycle group — addition combines coefficients of equal generators and makes formal subtraction possible.
The chosen cycle equivalence — rational equivalence or another explicitly named relation determines which cycles represent one class.
The resulting class group and operations — Chow groups, and where hypotheses permit, intersections, pushforwards, pullbacks, and cycle-class maps.

Recognition test. A case is an algebraic cycle only when the carrier is algebraic-geometric, the terms are integral closed subvarieties, the coefficients and grade are explicit, and any claimed equality of classes names its equivalence relation. A weighted sum of arbitrary geometric quantities does not qualify. Neither does a single subvariety unless it is being used as the corresponding generator with coefficient one.

What It Is Not¶

Not a singular or cellular chain. Those are topological chain constructions generated by simplices or cells with boundary operators. Algebraic cycles are generated by algebraic subvarieties and use algebraic equivalence relations.
Not just a divisor. A divisor is the codimension-one case. Algebraic cycles also include zero-cycles and arbitrary higher-codimension cycles.
Not a homology class. A cycle may determine a cohomology or homology class under a cycle-class map, but distinct cycles can have the same topological class, and an algebraic cycle exists before that map is applied.
Not an effective cycle by default. Effectivity requires every \(n_Z\ge 0\). General cycles allow negative coefficients.
Not a set-theoretic union. Multiplicity and cancellation are essential; \([Z]+[Z]=2[Z]\), whereas a union would forget the repeated contribution.
Not one universal quotient. Rational, algebraic, homological, and numerical equivalence have different strengths and yield different groups.

Scope of Application¶

The concept operates throughout intersection theory. Codimension-one cycles encode Weil divisors; zero-cycles encode formal sums of closed points; cycles of complementary codimension may admit intersection products when transversality or moving arguments make the construction valid. Proper morphisms support pushforward of cycles with degree factors, while flat morphisms support pullback. The precise hypotheses matter: these operations are not unrestricted set maps.

Cycles also connect algebraic geometry to topology and arithmetic. Cycle-class maps compare algebraic classes with singular, étale, or other cohomology theories. The Hodge and Tate conjectures ask, in different settings, which cohomological classes arise from algebraic cycles. Correspondences between varieties are represented by cycles on products and support the organization of motives. These applications reuse the same generator–multiplicity–grading–equivalence structure rather than merely sharing the word “cycle.”

Clarity¶

The formal-sum notation separates geometric support from multiplicity. Writing \(3[C]-2[D]\) says that the curve \(C\) contributes with multiplicity three and \(D\) is formally subtracted twice; it does not create a geometric region with negative area. Specifying \(Z_k(X)\) or \(Z^p(X)\) also prevents a common ambiguity: the first normally grades by dimension \(k\), while the second grades by codimension \(p\).

Equivalence must be carried in the notation or prose. The statement “\(\alpha=0\)” in the free cycle group is stronger than “\(\alpha\sim_{\mathrm{rat}}0\).” In the latter, \(\alpha\) can be a nonzero formal sum that vanishes only after passage to \(CH_k(X)\). Clear writing therefore distinguishes cycles, equivalence classes of cycles, and their images in a cohomology group.

Manages Complexity¶

Varieties contain infinitely many subvarieties, often arranged in complicated families. The cycle group replaces this uncontrolled collection by a linear bookkeeping layer. Grading prevents incompatible dimensions from mixing, while equivalence collapses deformations or principal-divisor changes that should not affect the selected invariant. The result is not that geometry becomes easy, but that difficult geometric questions acquire algebraic targets.

This compression is selective. Rational equivalence deliberately forgets some placement information while retaining information relevant to intersection theory. Numerical equivalence forgets more by comparing only intersection numbers. Choosing the quotient is consequently a modeling decision: too weak an equivalence leaves unwieldy detail, while too strong an equivalence can erase the phenomenon under study.

Abstract Reasoning¶

Algebraic cycles support reasoning at three levels. At the generator level one follows individual subvarieties and multiplicities. At the group level one adds, subtracts, pulls, and pushes formal sums. At the quotient level one reasons with invariant classes and checks that operations respect the equivalence relation. Moving between these levels is disciplined: a statement about representatives must be shown independent of representative before it becomes a statement about classes.

The construction also exposes proof obligations. An intersection product requires more than an intuitive geometric crossing; one needs conditions or a moving lemma that makes the class well defined. A pushforward coefficient includes the degree of the induced function-field extension when dimensions are preserved. Such details are not decorative. They are how the abstraction protects geometric information while permitting algebraic manipulation.

Knowledge Transfer¶

The cycle pattern transfers from divisors on curves to higher-codimension geometry. On a smooth projective curve, divisors are formal sums of closed points and principal divisors generate rational equivalence. The same formal mechanism generalizes: replace points by integral \(k\)-dimensional subvarieties, replace rational functions on the whole curve by rational functions on suitable subvarieties, and retain the quotient logic.^[4]

Transfer also works across cohomology theories, but only through explicit cycle-class maps. The common lesson is to keep three layers separate: the algebraic representative, its equivalence class, and its realization in another invariant. This prevents a topological equality from being mistaken for rational equivalence and makes clear which conclusions survive a change of realization.

Examples¶

Divisors on a curve. For a smooth projective curve \(C\), a divisor \(\sum_P n_P[P]\) is a codimension-one algebraic cycle. If \(f\) is a nonzero rational function, its zeros and poles with orders form the principal divisor \(\operatorname{div}(f)\), which is rationally equivalent to zero.
A principal divisor on the projective line. For the coordinate function \(t\) on \(\mathbb P^1\), \(\operatorname{div}(t)=[0]-[\infty]\). Thus the two points give different generators in the free cycle group but the same class in \(CH_0(\mathbb P^1)\).
An effective plane curve. A closed integral curve \(C\subset\mathbb P^2\) gives the effective codimension-one cycle \([C]\); \(m[C]\) records multiplicity \(m\), not a different underlying point set.
Correspondence. A suitable integral subvariety of \(X\times Y\) can define a correspondence between \(X\) and \(Y\). Composition uses pullback, intersection, and pushforward under appropriate hypotheses.

Structural Tensions¶

Representative versus class: explicit supports make geometry visible, while quotient classes make invariance manageable. Diagnostic: is the claim about a formal cycle, an equivalence class, or a realization of that class?
Effectivity versus group completion: effective cycles match actual positive geometric pieces, but subtraction is needed for groups and relations. Diagnostic: do all coefficients have to be nonnegative, or is formal cancellation being used?
Dimension versus codimension: both are useful, yet conversions require an equidimensional ambient setting and must not be assumed universally. Diagnostic: has the ambient dimension needed to translate the grading been fixed?
Fine versus coarse equivalence: rational equivalence retains more information than numerical equivalence but is usually harder to compute. Diagnostic: which equivalence relation licenses the asserted equality?
Algebraic versus topological visibility: cycle-class maps expose topological consequences, while homologically invisible algebraic cycles may still be nontrivial in Chow groups. Diagnostic: is vanishing claimed in the cycle group, the Chow group, or only a cohomological realization?
Autonomy vs. reduction: Linear Combination, Equivalence Relation, and Quotient expose the portable skeleton, but dimension grading, subvariety generators, geometric equivalences, and Chow-group operations remain autonomous. Diagnostic: Can the case be recognized without varieties or schemes, cycles, and a declared geometric equivalence? If not, retain Algebraic Cycle.

Structural–Framed Character¶

The reusable structure is a typed free-abelian construction followed by a controlled quotient. Its generators are not arbitrary objects: they are integral closed pieces inside a fixed algebraic ambient space. Its types are geometric degrees. Its relations are produced by geometric families or rational functions. This framing explains why ordinary linear-combination language covers only the arithmetic shell. The inferential content comes from which objects may be generators and which geometric motions license identification.

Structural Core vs. Domain Accent¶

The structural core is “form integer combinations of typed generators and quotient by declared relations.” Algebraic geometry supplies the indispensable accent: subvarieties, dimension, rational functions, proper and flat morphisms, intersections, and cycle-class maps. Removing that accent yields a generic presented abelian group, not an algebraic cycle. This is why the concept belongs as a domain-specific abstraction rather than as a new prime.

prime:linear_combination supplies the formal scale-and-add skeleton; algebraic cycles specialize it to integer coefficients and subvariety generators.
prime:equivalence_relation supplies the logic by which cycles are partitioned into rational, algebraic, homological, or numerical classes.
prime:quotient supplies passage from the cycle group to the corresponding group of classes.
prime:grading is reflected in the dimension/codimension decomposition when that catalog surface is available.

Relationships to Other Abstractions¶

Current abstraction Algebraic Cycle Domain-specific

Parents (1) — more general patterns this builds on

Algebraic Cycle is a kind of Linear Combination Prime

prime:linear_combination supplies the formal scale-and-add skeleton; algebraic cycles specialize it to integer coefficients and subvariety generators.

Hierarchy path (1) — routes to 1 parentless root

Algebraic Cycle → Linear Combination → Aggregation → Micro Macro Linkage

Neighborhood in Abstraction Space¶

Algebraic Cycle sits in a sparse region of the domain-specific corpus (83^rd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (1565 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With¶

Do not identify an algebraic cycle with the linear combination prime alone. That prime allows many substrates and coefficient systems; it does not require integral closed subvarieties, geometric grading, or rational equivalence. Do not identify it with a subvariety, because a cycle can combine several supports with signed multiplicity. Do not identify a Chow class with its representative: different cycles may define the same class. Finally, “cycle” in graph theory, dynamical systems, homological algebra, and periodic processes is a homonym unless the algebraic-geometric generator structure is present.

References¶

[1] The Stacks Project Authors. “Chow Groups of Spaces,” especially the definitions of cycles and rational equivalence. https://stacks.math.columbia.edu/download/spaces-chow.pdf registry ↩

[2] The Stacks Project Authors. “Rational Equivalence,” Chow Homology and Chern Classes. https://stacks.math.columbia.edu/tag/0AZG registry ↩

[3] William Fulton. Intersection Theory, second edition. Springer, 1998. https://doi.org/10.1007/978-1-4612-1700-8 registry ↩

[4] Robert Lazarsfeld. Positivity in Algebraic Geometry I. Springer, 2004. https://doi.org/10.1007/978-3-642-18808-4 registry ↩