Refined Gromov-Witten Invariants

Updated 31 January 2026

Refined Gromov-Witten invariants are enhanced curve-counting numbers that integrate additional structures like q-refinements and λ-class insertions to capture subtle geometric and arithmetic data.
They employ combinatorial, tropical, and orbifold techniques to relate algebro-geometric moduli spaces with deformation formulas and integrable hierarchies.
This framework underpins modern mirror symmetry and degeneration studies by linking modularity properties and holomorphic anomaly equations to refined enumerative invariants.

Refined Gromov-Witten invariants are enhancements of classical Gromov-Witten theory that insert new structure, such as cohomological, equivariant, tropical, or combinatorial data, into enumerative curve counting. These refinements often encode subtle geometric, arithmetic, or representation-theoretic information, especially in contexts such as mirror symmetry, moduli spaces of sheaves, and log/tropical geometry. This article surveys the principal approaches to refined Gromov-Witten invariants, highlighting structural frameworks, calculation techniques, integrable hierarchies, modularity properties, and deformation principles.

1. Refined Invariants: Definitions and Geometric Frameworks

Refined Gromov-Witten invariants are constructed by inserting additional data—most notably $q$ -refinements, $\lambda$ -class insertions, orbifold structure, or K-theoretic parameters—at the level of stable map moduli spaces or in the combinatorial models for curve counting. For toric surfaces specified by a balanced lattice polygon $\Delta^\circ$ , the logarithmic Gromov-Witten moduli space $\Mbar_{g,\Delta}^{\log}(X_\Delta|\partial X_\Delta,\beta_\Delta)$ carries tautological classes: Hodge class $\lambda_g$ , cotangent-line classes $\psi_i$ , and point-class insertions via evaluation maps. One considers descendant- $\lambda_g$ invariants:

$N_{g,\Delta}^k = \int_{[\Mbar_{g,\Delta}]^{\rm vir}} (-1)^g \lambda_g \prod_{i=1}^n \psi_i^{k_i} \ev_i^*([\pt])$

and packages these into genus-generating functions $F_\Delta^k(u) = \sum_{g\ge 0} N_{g,\Delta}^k u^{2g-2+r-\sum_i k_i}$ , interpreting $q = e^{iu}$ as the quantum parameter (Kennedy-Hunt et al., 2023).

Refinement may also involve constructing IP-counts—counts derived from evaluation maps lifted to abelian covers of symplectic divisors to record additional topological monodromy (“rim tori”) (Tehrani et al., 2014), or further dividing the virtual fundamental class according to correlation data in degeneration formulas for $\mathbb{P}^1$ -bundles (Blomme et al., 2024).

2. $q$ -Refined Tropical Correspondence and Vertex Multiplicities

The tropical approach to refined invariants establishes a combinatorial correspondence between generating series of descendant logarithmic Gromov-Witten invariants and $q$ -refined counts of rational tropical curves. For generic point configurations $p=(p_1,...,p_n)$ in $\RR^2$, rational tropical curves $h:\Gamma\to\RR^2$ are required to pass through $p_i$ at specified ends of valency $\geq k_i+2$ . Each vertex $V$ in the tropical graph $\Gamma$ is assigned a local multiplicity $m_V(q)$ : trivalent (unpointed) vertices obtain $m_V(q)=(-i)(q^{|a\wedge b|/2} - q^{- |a\wedge b|/2})$ for primitive directions $(a, b, -a-b)$ , while pointed higher-valency vertices have

$m_V(q) = \frac{1}{(m-1)!} \sum_{\omega \in \Omega_m} q^{k(\omega)/2}$

with $k(\omega)$ a sum over wedge-products of directions under cyclic orderings.

The refined tropical count is then

$N_{\mathrm{trop}}(q) = \sum_{h \in T^k_{\Delta,p}} \prod_{V \in \mathcal{V}(\Gamma)} m_V(q)$

where $T^k_{\Delta,p}$ is the finite set of rigid tropical curves fulfilling the combinatorial conditions. The primary $q$ -refined correspondence theorem asserts

$F_\Delta^k(u) = N_{\mathrm{trop}}(q)$

after setting $q = e^{iu}$ (Kennedy-Hunt et al., 2023). This identifies the algebro-geometric (logarithmic, descendant, Hodge-class) theory with combinatorial $q$ -refined curve counts, yielding deformation invariance of the refined counts.

3. Degeneration, Gluing, Integrable Hierarchies

Refined GW theory exploits toric degeneration and polyhedral decompositions to relate global invariants to sums over local models. The degeneration formula enables expressing $N_{g,\Delta}^k$ as a sum over genus-spread tropical enhancements $\tilde{h}: \tilde{\Gamma}\to\RR^2$, each contribution factoring by gluing vertex invariants and edge weights. Using Ranganathan–Wise gluing and universal rubber moduli, local vertex contributions reduce to integrals against double ramification cycles associated to tangency matrices $\Delta_V$ . These integrals connect to the quadratic double-ramification hierarchy, essentially encoding an integrable hierarchy for GW theory (Kennedy-Hunt et al., 2023).

For pointed vertices,

$N_{g_V,V} = \int_{[\Mbar_{g_V,n_V}]^{\rm vir}} (-1)^{g_V} \lambda_{g_V} \psi^d \mathrm{DR}_{g_V}(\Delta_V^x)\mathrm{DR}_{g_V}(\Delta_V^y)$

where $d$ is the valency deficit.

The generating functions arising in this context match the Moyal product and dispersionless noncommutative KdV hierarchy, so that refined invariants admit closed-form expressions involving trigonometric sums and match the tropical vertex multiplicities. Reassembling these data yields the total correspondence and supports modular/analytic structure in the generating series.

4. Quasi-Modularity, Holomorphic Anomaly, and Higher-Genus Structure

Refined GW generating series frequently exhibit quasi-modular behavior. Floor diagram algorithms and their $q$ -refined counterparts assemble generating functions for surfaces and threefolds fibered over elliptic curves, with modular properties determined by congruence subgroups $\Gamma_1(n)$ . In the context of Calabi–Yau 5-folds, the refined GW generating series are shown to be quasi-modular functions of $\Gamma_1(3)$ , satisfy extended holomorphic anomaly equations, and possess leading conifold asymptotics governed by the Barnes double-Gamma function (Brini et al., 2024).

For bielliptic surfaces, the introduction of $\lambda$ -class refinements in the GW theory leads to generating series that decompose as sums over quasi-modular forms in fiber and section directions (Blomme, 2024). Similar modularity arises for local mirrors and crepant resolutions.

Orbifold refined invariants expand the theory to stacks and root constructions. The effective one-leg orbifold refined vertex for $P^1_a$ (the $a$ -th root stack at 0) is formulated via disconnected relative moduli spaces with prescribed ramification and orbifold monodromy, and localization techniques yield generating functions expressible as Schur function expansions under explicit framing conditions (Yu et al., 24 Jan 2026). In the smooth $a=1$ case, formulas recover the classical one-leg refined topological vertex. Gluing such vertices enables the computation of partition functions for orbifold local Calabi–Yau threefolds ("football" geometry) as double infinite products, capturing refined combinatorics and orbifold parameters.

Further types of refined GW invariants are constructed by lifting evaluation maps to abelian covers determined by rim tori modules of symplectic divisors, yielding IP-counts and sharp refinements of the symplectic sum formula (Tehrani et al., 2014). These refinements control vanishing cycles and enable the identification of term-by-term invariants for sums along divisors, which in the unrefined theory can only be obtained as sums over vanishing cycles. Applications include vanishing and rigidity phenomena: for example, all GW invariants of $K3$ vanish outside of fiber classes (Tehrani et al., 2014).

Correlated GW invariants in $\mathbb{P}^1$ -bundle settings introduce additional correlation data tracing values in torsors of roots of line bundles in Alb $(X)$ , leading to decomposition of virtual classes and refinement of degeneration formulas (Blomme et al., 2024). These constructions are expected to play critical roles in the study of reduced GW invariants for Abelian and K3 surfaces and connect to torsion phenomena in the Picard group.

7. Combinatorial Models and Mirror Symmetry Correspondences

Refined floor diagrams and rooted trees supply the combinatorial backbone of many refined GW enumeration schemes, especially for toric and Hirzebruch surfaces (Bousseau, 2019). After changing variables ( $q = e^{iu}$ ), these diagrammatic counts compute generating series for higher genus relative invariants with $\lambda$ -class insertions. Refined tropical correspondence matches these counts with log GW-invariants of toric surfaces; the resulting combinatorial identities extend to binomial transformations between Block–Göttsche invariants for distinct rational surfaces.

Mirror symmetry enters via correspondences between refined GW invariants and refined Donaldson–Thomas invariants of moduli spaces of sheaves (e.g., on $\mathbb{P}^2$ ), establishing generating function equalities after non-trivial change of variables ( $y = e^{i\hbar}$ ) linking cohomological grading and genus expansion (Bousseau, 2019). The underpinning wall-crossing and scattering diagram formalism achieves a refined sheaves/GW correspondence, contextualized by the interplay of tropical geometry, stability conditions, and quantum parameters.

Refined Gromov-Witten invariants thus integrate combinatorial, geometric, and representation-theoretic data into curve counting, enriching the formalism and extracting subtle arithmetic and modular behavior. Their construction is intimately tied to degeneration, tropical methods, integrable systems, and mirror symmetry, and they serve as the foundational enumerative framework in contexts ranging from orbifold geometry to higher-genus crepant resolutions, floor diagrams, and symplectic sum theory.