Quantum Cohomology¶

A cohomology algebra whose cup product is deformed by curve-class-weighted genus-zero Gromov–Witten invariants.

Version: v1 · 2026-10-03 · History
Domain-specific #: 13544
Domain group: Formal Sciences
Origin domain: Mathematics
Subdomains: Algebraic Geometry, Symplectic Geometry → Mathematics
Aliases: Quantum Cohomology Ring

Core Idea¶

Quantum cohomology equips the cohomology of a suitable smooth projective or symplectic target with a product that retains the classical cup product at curve class zero and adds genus-zero Gromov–Witten (GW) contributions from nonzero curve classes. A coefficient variable or ring records the classes contributing to each correction. The result is not “quantum” because it is a cohomology theory of microscopic particles; the name arose from the mathematical structure of topological sigma models and curve counting.^[1]^[2]^[3]

The small product is determined by genus-zero three-point invariants. The big product depends on a cohomology parameter: its structure constants are third derivatives of a genus-zero potential and therefore include additional primary insertions. It is false that big quantum cohomology simply contains all GW invariants of all genera or all descendant insertions. Kontsevich and Manin's original formalism gives associative, supercommutative fiberwise algebras with identity when its GW axioms and formal/convergence conditions apply.^[1]

Structural Signature¶

Sig role-phrases: target/cohomology carrier — degree-zero cup product — genus-zero curve-class data — coefficient convention — deformed internal multiplication — WDVV/splitting coherence.

Target and carrier. Fix a suitable target $X$ and its graded cohomology $H^*(X)$. Its classical cup multiplication is the zero-curve-class baseline, not the quantum product's whole identity.^[1]
Curve data. Genus-zero, three-marked-point GW invariants of effective classes $\beta$ provide small-product coefficients. Extra primary marked-point insertions enter the big potential. The invariants are attached to a target and incidence classes; they are not arbitrary numeric corrections.^[1]^[2]
Coefficient convention. A weight $q^\beta$ distinguishes curve classes. In one-generator examples it may be a single $q$; Grassmannians in Bertram's small theory use $\mathbb C[q]$. General sums may instead require a Novikov/formal completion or a convergence domain. One must declare the convention before manipulating products.^[1]^[2]
Multiplication. For a homogeneous basis $\{T_i\}$ and its Poincaré-dual basis $\{T^i\}$, schematically,

$$a *_0 b=\sum_{\beta,i}\langle a,b,T_i\rangle_{0,3,\beta}\,T^i q^\beta.$$

Here the pairing, degree signs and coefficient completion use the selected theory; the $\beta=0$ contribution is $a\smile b$. The formula defines an internal, coefficient-bilinear product only when the GW data and sums are well-defined.^[1]^[2] - Coherence. GW splitting identities appear algebraically as WDVV equations, securing associativity of the product under the theory's hypotheses. WDVV may constrain coefficients strongly, but it is not a universal algorithm that supplies all enumerative numbers without target-specific initial information.^[1]

What It Is Not¶

It is not the ordinary cohomology ring renamed. The classical ring is recovered by discarding positive curve classes, yet the multiplication can change: for projective space the small quantum relation is $h^{*(n+1)}=q$, whereas the ordinary cup power $h^{n+1}$ vanishes. The carrier can be recognizably related while the ring law is not identical.^[1]

It is not GW theory in full. GW invariants can include positive genus, descendants and other structures not determined by a generic small quantum multiplication table. Big multiplication is a genus-zero primary deformation family, not an assertion that every invariant is encoded by one ring. It is also not a generic formal deformation: the corrections are constrained by target geometry, curve classes, the pairing and splitting axioms.^[1]

Nor is a curve-count prediction by mirror symmetry itself quantum cohomology. Such a prediction may be compared with quantum invariants, but a specified cohomology carrier and GW-defined product are needed before there is this algebraic object. A coefficient need not literally count distinct embedded curves without multiplicity: virtual and intersection-theoretic definitions can require more careful interpretation.^[1]^[2]

Scope of Application¶

The original projective-algebraic construction treats smooth projective targets with genus-zero GW data satisfying its axioms; symplectic constructions use pseudoholomorphic curves under stated technical conditions. Seidel's original paper, for example, explicitly assumes (W+) for its quantum cup-product construction rather than asserting all-target validity. Its useful questions include deformed multiplication of cohomology classes, relations among rational-curve incidence invariants, and comparisons of target-dependent structures. The target and coefficient convention are part of the scope, not metadata that can be silently omitted.^[1]^[2]^[3]

For projective spaces, Kontsevich and Manin give both the small relation $h^{*(n+1)}=q$ and a big potential for $\mathbb P^2$ whose associativity equations constrain specified rational-curve numbers. For a Grassmannian, Bertram constructs a small ring over $\mathbb C[q]$ and quantum corrections to Schubert multiplication. These are literal instantiations of one signature but have different bases, incidence problems and presentations.^[1]^[2]

The theory has boundaries. In the original Kontsevich–Manin account, formal versus convergent potentials are distinguished, and the existence and geometric reading of some historical constructions are explicitly qualified. We therefore do not infer that every target has the same polynomial coefficient ring, every count is an integer of literal distinct curves, or every mirror-symmetry number follows directly from a ring presentation.^[1]

Clarity¶

Three levels must stay separate: cohomology classes are elements being multiplied; GW invariants furnish target- and class-indexed structure constants; the quantum product packages those constants into an associative algebra. Setting the curve variables to zero tests whether the claimed formula recovers cup multiplication. That is a structural check, not a claim that positive-degree contributions are small numerically.^[1]^[2]

“Small” and “big” name how much genus-zero insertion data enters the product, not how large the target is. At the small base point, the three-point invariants determine the multiplication. At a general parameter $t$, derivatives of the potential use higher-point genus-zero primary invariants with additional $t$ insertions. The resulting big family can reveal variations invisible in a single fixed multiplication table.^[1]

Manages Complexity¶

Raw rational-curve incidence information is indexed by degree or class and by insertions. Quantum cohomology compresses selected three-point information into structure constants and ring identities; one can multiply basis elements instead of returning to each moduli space for every formal relation. Bertram's Schubert basis and $q$-corrected Pieri rule show this compression in a concrete combinatorial setting.^[2]

Associativity makes the compression useful: comparing $(a*b)*c$ with $a*(b*c)$ forces nontrivial identities among corrections. In the projective-plane potential, Kontsevich and Manin derive a recursion for numbers $N(d)$ with stated initial data. This does not remove the geometric work of defining the GW invariants or proving that $N(d)$ has the intended enumerative meaning.^[1]

Abstract Reasoning¶

The operation is a deformation with a controlled classical limit. If all positive curve-class weights are formally suppressed, the zeroth term is the cup product. The positive terms must satisfy graded algebra laws, not merely fit observed counts. In a one-parameter case this resembles $a*b=a\smile b+qC_1(a,b)+q^2C_2(a,b)+\cdots$, but the $C_d$ are constrained GW data rather than freely chosen bilinear maps.^[1]^[2]

Associativity can be read as a consistency condition across two ways to split a four-point genus-zero problem. Kontsevich and Manin express this through WDVV and show its target-specific enumerative strength for $\mathbb P^2$. They also note that for Calabi–Yau targets the same equations can be much less constraining. Hence “associative” is necessary coherence, not a promise of complete reconstruction everywhere.^[1]

Knowledge Transfer¶

The literal signature transfers from projective-space cohomology to Grassmannian Schubert calculus: choose a target and basis, index genus-zero data by curve class, form the coefficient-weighted internal product, and check the classical limit and associativity. The actual curve counts and ring presentations do not transfer unchanged. In $\mathbb P^n$ a single hyperplane generator gives $h^{*(n+1)}=q$ under the stated convention; in a Grassmannian, partitions/Schubert classes and quantum Pieri/Giambelli relations organize the answer.^[1]^[2]

The broader idea of correcting an operation while retaining an old limit may travel outside geometry, but quantum cohomology itself does not: remove cohomology, genus-zero curve data or the pairing and one has an analogy, not another instance. Its proposed live DAG genus is Algebra over a Ring because the completed/polynomial cohomology carrier has bilinear internal multiplication over the selected coefficient ring. Cohomology Ring is a closely related undeformed limit, not a strict parent under the altered multiplication.

Examples¶

Projective plane and rational-curve recursion. In $\mathbb P^2$, the classical hyperplane class satisfies $h^3=0$, while Kontsevich and Manin's small quantum presentation has $h^{*3}=q$ under their convention. Their big potential adds coordinates for insertions and contains coefficients $N(d)$ for degree-$d$ rational curves through $3d-1$ suitably general points. WDVV yields a recursion from $N(1)=1$; the paper distinguishes its formal derivation from the additional construction/interpretation needed to call the coefficients geometric curve counts.^[1] Mapped back: target/carrier = $H^*(\mathbb P^2)$; classical baseline = $h^3=0$; curve data = class $d[\mathrm{line}]$ three-point corrections and higher-point incidence insertions; coefficient = degree weight $q^d$ or equivalent divisor coordinate; product = $h^{*3}=q$ at the small point and a parameter-varying big product; coherence = WDVV recursion for the specified $N(d)$.

Grassmannian Schubert multiplication. Bertram takes a Grassmannian $G(n-k,n)$ with Schubert classes $\sigma_\lambda$. Degree-$d$ maps $\mathbb P^1\to G$ meeting general Schubert conditions furnish the $q^d$ correction terms. The resulting small product on $H^*(G,\mathbb C)[q]$ is associative, reduces to ordinary intersection product at $q=0$, and yields a quantum Pieri rule with a positive-degree term; the quantum Giambelli determinant is subtler and does not simply receive the same kind of added term.^[2] Mapped back: target/carrier = Grassmannian Schubert cohomology; classical baseline = ordinary Schubert cup multiplication; curve data = degree-$d$ rational-map incidences; coefficient = $\mathbb C[q]$ and $q^d$; product = corrected Schubert multiplication; coherence = associative small quantum ring and compatible Pieri/Giambelli rules.

Structural Tensions¶

Finite calculability versus deformation completeness. Small multiplication uses three-point information and can admit compact presentations such as the projective-space relation or Grassmannian quantum Pieri rule. That makes computation tractable but omits the extra insertion dependence of big multiplication. Retaining the big potential preserves those genus-zero variations but raises additional formal/convergence and data demands.^[1]^[2] Diagnostic: Does the question ask for one fixed multiplication law or how products change with inserted cohomology parameters, and what relevant information disappears when small replaces big?

Enumerative immediacy versus formal validity. Calling a coefficient “the number of curves” gives a strong geometric interpretation, but only under suitable incidence, compactification, multiplicity and GW-construction conditions. Treating it as a formal/virtual invariant can make the algebraic structure well-defined more broadly while weakening the naive distinct-curve reading. Kontsevich and Manin explicitly distinguish formal calculations from additional geometric interpretation, while Bertram supplies a rigorous Grassmannian construction for his incidence numbers.^[1]^[2] Diagnostic: For this coefficient and target, what construction and hypotheses justify a literal enumerative reading rather than only an intersection-theoretic or virtual one?

Structural–Framed Character¶

Quantum cohomology lies toward the structural end of a domain-specific mathematical spectrum: its target/cohomology/GW/product roles and axioms are formal. Vocabulary travel: “quantum” and “cohomology” travel into symplectic and algebraic geometry, but the literal identity requires the GW-derived cohomological multiplication rather than any corrected ring. Evaluative weight: whether a computation is elegant or useful is an evaluation, not part of the identity; algebraic validity rests on defined invariants and associativity. Human-practice dependence: mathematicians choose bases and coefficient conventions, yet after those are stated the product's formal claims are not contingent on community preference. Institutional origin: the historical mathematical-physics origin explains the name, but an institution or research school is not part of a ring's defining data. Import versus recognition: a new target can be recognized as an instance only by verifying its GW construction and product; importing the $\mathbb P^n$ equation into an arbitrary target is invalid.^[1]^[2]

The node remains domain-specific. The operation-deformation skeleton might merit a future-prime review, but cohomology, Poincaré pairing and genus-zero curve invariants are constitutive here, not removable accents. Its character: a formally structural, target-dependent mathematical algebra whose very identity retains its geometric domain.

Structural Core vs. Domain Accent¶

At a neutral level, the core is an operation whose old product is recovered at zero correction, whose indexed corrections preserve an algebraic coherence law. That skeleton is portable and could be investigated as a future prime; no already verified live prime is asserted as its complete parent. The live Algebra over a Ring instead provides a concrete algebraic genus: one coefficient base, module and bilinear internal product.

The indispensable domain accent is the cohomology of a target, degree/class-indexed genus-zero GW coefficients, Poincaré duality pairing, coefficient convention and WDVV-compatible quantum multiplication. Without these, a generic formal deformation remains but quantum cohomology does not. The identity is thus a recurring domain-specific abstraction rather than an alias of Cohomology Ring or a prime.

This entry is a kind of Algebra over a Ring. Quantum cohomology is an algebra over its chosen commutative coefficient ring, with a bilinear internal quantum product.

Relationships to Other Abstractions¶

Current abstraction Quantum Cohomology Domain-specific

Parents (1) — more general patterns this builds on

Quantum Cohomology is a kind of Algebra over a Ring Domain-specific

Quantum cohomology is an algebra over its chosen commutative coefficient ring, with a bilinear internal quantum product.

Hierarchy path (1) — routes to 1 parentless root

Quantum Cohomology → Algebra over a Ring → Linearity

Neighborhood in Abstraction Space¶

Quantum Cohomology sits in a sparse region of the domain-specific corpus (66^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Algebraic Structures, Groups & Operators (42 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With¶

Ordinary Cohomology Ring: same underlying cohomology can support the classical product, but quantum positive-degree terms change its multiplication.^[1]
GW invariants as a list: the ring packages a constrained portion of those data into a product; higher-genus and descendant data are not generically recovered from the small ring.^[1]
Big equals all invariants: big product varies through extra genus-zero primary insertions, not every genus or descendant theory.^[1]
One universal $q$ ring: a single polynomial variable suffices in the cited homogeneous examples, while other targets may need multi-class weights and formal/Novikov conventions.^[1]^[2]
Automatic naive curve counts: GW coefficients may need virtual or intersection-theoretic interpretation and geometry-specific incidence conditions.^[1]^[2]
Mirror prediction itself: comparison with a mirror calculation does not replace construction of the target's quantum product.

References¶

[1] Maxim Kontsevich and Yuri Manin, “Gromov–Witten Classes, Quantum Cohomology, and Enumerative Geometry”, original author/institute-hosted paper (1994), introduction; §§2.1–2.2, 4.3–4.5 and 5.2–5.3, particularly equations (5.16)–(5.17) and (5.24). Historical axiomatic and enumerative-interpretation caveats are retained. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t ↩u ↩v ↩w ↩x ↩y ↩z ↩27 ↩28

[2] Aaron Bertram, “Quantum Schubert Calculus”, original arXiv paper, version 2 (1997), §1 definition and associativity theorem; §§2–4 for Quot-scheme interpretation and quantum Pieri/Giambelli. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q

[3] Paul Seidel, “π1 of Symplectic Automorphism Groups and Invertibles in Quantum Homology Rings”, original author preprint revised 1997, introduction pp.1–3, especially Assumption (W+) and the quantum cup/intersection-product discussion. registry ↩a ↩b