Kontsevich–Soibelman Wall-Crossing Invariant

• The Kontsevich–Soibelman invariant is an ordered product of BPS wall-crossing factors that remains constant even as individual indices jump at walls of marginal stability.
• It is derived using twistorial constructions, Riemann–Hilbert problems, and spectral network techniques to ensure continuity of hyperkähler metrics in Seiberg–Witten theory.
• The invariant underpins modern approaches in quantum torus algebras, motivic DT theory, and categorification, bridging discontinuous BPS spectra with smooth physical observables.

The Kontsevich–Soibelman wall-crossing invariant is the ordered product of wall-crossing factors attached to BPS charges in a chosen angular sector of the central-charge plane. Its defining feature is that, although the protected BPS indices $\Omega(\gamma;u)$ and the ordering of contributing charges jump at walls of marginal stability, the total product remains unchanged. In four-dimensional $\mathcal N=2$ Seiberg–Witten theory, Gaiotto, Moore, and Neitzke identified this invariant with the “spectrum generator” and showed that its constancy is equivalent to continuity of the exact hyperkähler metric on the moduli space of the theory on $\mathbb R^3\times S^1$ (0807.4723). In the abstract Kontsevich–Soibelman framework, the same object appears as a sector element in a pronilpotent group or motivic quantum torus, factorized over rays of the central charge (Kerr et al., 2017).

1. Seiberg–Witten data and the wall-crossing problem

The basic input is a local system of electromagnetic charges $\Gamma \cong \mathbb Z^{2r}$ over the Coulomb branch $B$, equipped with the integral antisymmetric Dirac pairing $\langle \gamma,\gamma' \rangle$. In a local electric–magnetic duality frame one may write $\Gamma=\Gamma_e\oplus\Gamma_m$ with bases $\{\alpha_I\}$ and $\{\beta^J\}$ satisfying $\langle \alpha_I,\beta^J\rangle=\delta_I^J$ and vanishing pairings among electric–electric and magnetic–magnetic basis elements. The central charge is the Seiberg–Witten period

$\mathcal N=2$0

with $\mathcal N=2$1 and $\mathcal N=2$2 in a duality frame. A one-particle state of charge $\mathcal N=2$3 satisfies $\mathcal N=2$4, with equality for BPS states. The protected index is the second helicity supertrace

$\mathcal N=2$5

and is locally constant away from walls where phases align, $\mathcal N=2$6, allowing decays $\mathcal N=2$7. The Support Property,

$\mathcal N=2$8

for all charges with $\mathcal N=2$9, ensures that only finitely many active charges occur in any bounded phase sector (0807.4723).

This data already contains the wall-crossing problem. Individual BPS indices jump on codimension-one walls of marginal stability, yet the low-energy theory is expected to vary smoothly except at loci where states become massless. The KS invariant is the mechanism by which those two facts are reconciled.

2. Twistorial construction and continuity of the hyperkähler metric

After compactification on a circle of radius $\mathbb R^3\times S^1$0, the low-energy theory on $\mathbb R^3\times S^1$1 becomes a sigma model whose target $\mathbb R^3\times S^1$2 is a hyperkähler $\mathbb R^3\times S^1$3-torus fibration over $\mathbb R^3\times S^1$4. The fiber coordinates are the electric Wilson lines and dual photons, combined into $\mathbb R^3\times S^1$5. The semi-flat Darboux coordinates are

$\mathbb R^3\times S^1$6

For each charge with $\mathbb R^3\times S^1$7, one defines the BPS ray

$\mathbb R^3\times S^1$8

Instanton corrections from 4d BPS particles wrapping $\mathbb R^3\times S^1$9 deform the semi-flat coordinates to corrected coordinates $\Gamma \cong \mathbb Z^{2r}$0, characterized as the solution of a Riemann–Hilbert problem: they are piecewise holomorphic in $\Gamma \cong \mathbb Z^{2r}$1, have prescribed jumps across the rays $\Gamma \cong \mathbb Z^{2r}$2, and match $\Gamma \cong \mathbb Z^{2r}$3 as $\Gamma \cong \mathbb Z^{2r}$4. Equivalently, they satisfy the TBA-like integral equations

$\Gamma \cong \mathbb Z^{2r}$5

The crucial physical statement is that the exact hyperkähler metric $\Gamma \cong \mathbb Z^{2r}$6 reconstructed from these Darboux coordinates must be smooth in $\Gamma \cong \mathbb Z^{2r}$7, except where states become massless. Single-instanton terms, however, depend explicitly on $\Gamma \cong \mathbb Z^{2r}$8 and therefore jump. The resolution is that higher multi-instanton terms jump in a compensating way, so that the full $\Gamma \cong \mathbb Z^{2r}$9, the holomorphic symplectic form $B$0, and the metric $B$1 remain continuous. In this formulation, the wall-crossing formula is precisely the condition that the total jump data of the Riemann–Hilbert problem does not change when BPS rays coalesce across a wall (0807.4723).

GMN also express the construction through “4d $B$2” equations: $B$3 obeys a flat meromorphic $B$4-connection with rank-1 irregular singularities at $B$5, and the Stokes factors of this connection are the KS symplectomorphisms. Isomonodromic deformation of that connection is the formal origin of the invariant product.

3. Definition of the invariant

In the classical algebraic-torus formulation, one introduces multiplicative coordinates $B$6 with Poisson structure determined by $B$7. The KS symplectomorphism associated to a charge $B$8 acts by

$B$9

Globally one may need a quadratic refinement $\langle \gamma,\gamma' \rangle$0, in which case $\langle \gamma,\gamma' \rangle$1 is replaced by $\langle \gamma,\gamma' \rangle$2. For a half-plane of phases $\langle \gamma,\gamma' \rangle$3, the 4d spectrum generator is

$\langle \gamma,\gamma' \rangle$4

Its invariance,

$\langle \gamma,\gamma' \rangle$5

is the KS wall-crossing formula in the GMN presentation (0807.4723).

An equivalent abstract form uses stability data $\langle \gamma,\gamma' \rangle$6 on a $\langle \gamma,\gamma' \rangle$7-graded Lie algebra $\langle \gamma,\gamma' \rangle$8. For a strict sector $\langle \gamma,\gamma' \rangle$9, one defines

$\Gamma=\Gamma_e\oplus\Gamma_m$0

and factorizes it clockwise over rays $\Gamma=\Gamma_e\oplus\Gamma_m$1,

$\Gamma=\Gamma_e\oplus\Gamma_m$2

The sector element $\Gamma=\Gamma_e\oplus\Gamma_m$3 is the KS invariant of $\Gamma=\Gamma_e\oplus\Gamma_m$4: under deformations of the stability condition that do not move BPS rays across the boundary of $\Gamma=\Gamma_e\oplus\Gamma_m$5, the ordered product remains unchanged (Kerr et al., 2017). In the numerical unrefined torus language of Seiberg–Witten integrable systems, the same object is written

$\Gamma=\Gamma_e\oplus\Gamma_m$6

with clockwise ordering in an admissible sector $\Gamma=\Gamma_e\oplus\Gamma_m$7 (Wang, 2019).

A recurrent misconception is that wall-crossing invariance means the indices $\Gamma=\Gamma_e\oplus\Gamma_m$8 themselves remain constant. The invariant is not the collection of indices but the ordered product of ray factors. Both the contributing charges and their multiplicities may change, and the order of factors may reverse, while the total product is fixed.

In refined or motivic form, the classical torus is replaced by a quantum torus,

$\Gamma=\Gamma_e\oplus\Gamma_m$9

and the KS factor is implemented by quantum dilogarithms rather than classical binomials. The invariant statement remains the same: a phase-ordered product over a sector is constant, but now in the noncommutative quantum torus algebra (Sen, 2011).

4. Fundamental identities and chamber examples

The elementary rank-two identity is the pentagon relation. If $\{\alpha_I\}$0, then

$\{\alpha_I\}$1

Physically, this expresses the appearance or disappearance of a single hypermultiplet of charge $\{\alpha_I\}$2 when the phases of $\{\alpha_I\}$3 and $\{\alpha_I\}$4 cross. In the simplest class-$\{\alpha_I\}$5 and Argyres–Douglas examples, it is the local model of wall crossing (0807.4723).

For pure $\{\alpha_I\}$6 Seiberg–Witten theory, the strong-coupling chamber contains a monopole and a dyon, while the weak-coupling chamber contains an infinite dyon tower together with the $\{\alpha_I\}$7-boson vectormultiplet. GMN encode this by the infinite product identity

$\{\alpha_I\}$8

and for $\{\alpha_I\}$9 with $\{\beta^J\}$0 massless flavors by

$\{\beta^J\}$1

making explicit that weak- and strong-coupling spectra differ while the total ordered product does not (0807.4723).

For pure $\{\beta^J\}$2, the KS invariant becomes a computational device. Using the wall-crossing structure formalism, split attractor flows, and ordered products over sectors, one derives the standard $\{\beta^J\}$3-type towers associated with the three simple-root embeddings and also an additional weak-coupling wall $\{\beta^J\}$4. The corresponding composite KS identity changes the degeneracy of the family

$\{\beta^J\}$5

from $\{\beta^J\}$6 to $\{\beta^J\}$7, yielding a new family of BPS states while keeping the sector product $\{\beta^J\}$8 fixed (Wang, 2019).

These examples show the general pattern. Wall crossing is not a failure of consistency; it is the refactorization of a fixed element. Finite products, infinite products, and mixed hypermultiplet–vectormultiplet identities are different manifestations of the same ordered-product invariance.

5. Spectral networks, WKB triangulations, and exact WKB

In class-$\{\beta^J\}$9 theories of type $\langle \alpha_I,\beta^J\rangle=\delta_I^J$0, the invariant acquires a concrete geometric realization through WKB foliations on the ultraviolet curve $\langle \alpha_I,\beta^J\rangle=\delta_I^J$1. For a fixed phase $\langle \alpha_I,\beta^J\rangle=\delta_I^J$2, the WKB condition

$\langle \alpha_I,\beta^J\rangle=\delta_I^J$3

defines a foliation whose generic trajectories determine a decorated ideal triangulation $\langle \alpha_I,\beta^J\rangle=\delta_I^J$4. Fock–Goncharov coordinates $\langle \alpha_I,\beta^J\rangle=\delta_I^J$5 attached to the edges satisfy log-canonical brackets

$\langle \alpha_I,\beta^J\rangle=\delta_I^J$6

When $\langle \alpha_I,\beta^J\rangle=\delta_I^J$7 crosses a critical phase, the triangulation changes by a flip or a juggle. A finite WKB arc yields the hypermultiplet transformation

$\langle \alpha_I,\beta^J\rangle=\delta_I^J$8

while a one-parameter family of closed WKB curves yields the vectormultiplet transformation

$\langle \alpha_I,\beta^J\rangle=\delta_I^J$9

These are precisely KS factors, with the sign determined by the quadratic refinement. The corresponding spectrum generator

$\mathcal N=2$00

is invariant under deformations that do not move BPS rays across the endpoints of the half-plane (0907.3987).

A different spectral-network construction produces the BPS monodromy directly from a finite graph $\mathcal N=2$01 at a maximally degenerate locus where all vanilla 4d central charges align. In that setting the monodromy operator is determined by equations

$\mathcal N=2$02

for the edges $\mathcal N=2$03 of the critical graph. The construction is manifestly KS-invariant in the sense that it uses no prior knowledge of the 4d BPS spectrum; the invariant monodromy is extracted from topological data of the graph and framed 2d–4d soliton transport (Longhi, 2016).

For meromorphic quadratic differentials, an analytic version of the same story identifies the sector BPS automorphism $\mathcal N=2$04 with the birational map relating Fock–Goncharov coordinate charts of the boundary WKB triangulations: $\mathcal N=2$05 Under the canonical quadratic refinement, the KS factors become cluster $\mathcal N=2$06-mutations with $\mathcal N=2$07, and the same automorphisms describe Stokes jumps of Voros symbols in exact WKB analysis (Allegretti, 2020).

6. Motivic, categorical, quiver-theoretic, and geometric extensions

The invariant admits several extensions that preserve its ordered-product character while changing the ambient algebra. In the motivic $\mathcal N=2$08–$\mathcal N=2$09 theory, the single charge lattice is replaced by a vacuum groupoid $\mathcal N=2$10, diagonal entries record 4d charges, off-diagonal entries record 2d soliton charges, and the wall-crossing element becomes a matrix $\mathcal N=2$11 built from motivic Hall data. Its defining factorization is

$\mathcal N=2$12

which simultaneously categorifies the Cecotti–Vafa and Kontsevich–Soibelman formulas (Kerr et al., 2017).

A different line of work clarifies that the KS product, the Joyce–Song formula, and the Manschot–Pioline–Sen formula are not competing prescriptions but equivalent expansions of the same invariant. Stoppa showed that the GMN integral equations yield asymptotic rooted-tree expansions whose discontinuities reproduce Joyce–Song wall crossing (Stoppa, 2011), while Sen proved the equivalence of KS, MPS, and Joyce–Song formulas in the refined rank-two setting (Sen, 2011). In the Hall-algebra approach to framed objects, the same invariant appears as a product of motivic dilogarithms $\mathcal N=2$13, and framed wall crossing becomes a left-right action of those KS factors on generating series of framed objects (Mozgovoy, 2011). Mozgovoy and Reineke further identified the relevant “halo” coefficients with abelian quiver invariants, proved positivity and geometricity properties, and showed that a motivic version of the MPS formula is equivalent to the KS identity in the quantum torus (Mozgovoy et al., 2012). At the foundational level, the motivic integration map from Hall algebras to the motivic quantum torus relies on the Kontsevich–Soibelman conjectural identity for motivic Milnor fibers, proved in several nontrivial cases by Le Quy Thuong (Thuong, 2010).

Open and relative enumerative theories furnish geometric realizations of the same invariant. For complex Lagrangians in hyperkähler four-manifolds, multi-disk homology produces rational numbers $\mathcal N=2$14 and a ray-ordered product

$\mathcal N=2$15

which is invariant across walls of the first type (Iacovino, 2017). On elliptic K3 surfaces, open Gromov–Witten invariants $\mathcal N=2$16 define automorphisms

$\mathcal N=2$17

and their primitive jumps realize the pentagon identity and the KS scattering-diagram consistency condition (Lin, 2014).

More recent reformulations reinterpret the invariant in other languages without changing its formal role. Morozov, Morozov, and Morozov relate the ordered product of Stokes/flip automorphisms to Reshetikhin–Turaev–Witten knot invariants, recovering Jones polynomials in the simplest degenerate conformal-block case and representing quantum flips by Faddeev’s quantum dilogarithm (Galakhov et al., 2014). For $\mathcal N=2$18-Kronecker wall crossing, the refined KS product has been rewritten as an identity between generating series of symmetric quivers,

$\mathcal N=2$19

with one side finite and the other infinite; diagonalization and “trees of unlinkings” then express closed 4d DT invariants in terms of open 3d DT invariants and loop-quiver data (Bryan et al., 11 Jun 2025). In 5d BPS-quiver settings, quiver symmetries and affine root lattices provide functional equations for the KS invariant in collimation chambers, yielding exact conjectural spectrum generators for local $\mathcal N=2$20, $\mathcal N=2$21, and $\mathcal N=2$22 geometries (Monte et al., 2021). At the categorical level, the bosonic pentagon identity has been promoted to an equivalence of chain complexes and differential graded algebras, quadratic-dual to the PBW presentations of the corresponding Cohomological Hall algebra; this gives a categorification of the bosonic wall-crossing formula via Koszul duality (Gaiotto et al., 2023).

Across all these frameworks, the same structural principle persists. The wall-crossing invariant is a sector-ordered product over rays, not a chamberwise list of BPS numbers. Its algebraic realization may be a classical Poisson torus, a motivic quantum torus, a Hall algebra, a cluster variety, a spectral-network monodromy, or a multi-disk envelope algebra; its physical realization may be instanton corrections to a hyperkähler metric, framed halos around a heavy core, WKB flips on a UV curve, or open enumerative counts. What remains fixed is the ordered product itself.