Supersymmetric field theories have a special property. Certain quantities, such as the "twisted index" that counts protected ground states of a three-dimensional gauge theory on a circle times a surface, can be computed exactly by a finite contour integral instead of an infinite-dimensional path integral. This exact reduction is called supersymmetric localization. The index is not fully rigid, though. When a real parameter called the Fayet–Iliopoulos (FI) parameter $\tau$ crosses certain integer thresholds, the index jumps. This is called wall-crossing, and the contour-integral rule that produces the jumps becomes very hard to apply once the gauge group is non-Abelian, meaning a gauge group like $U(N)$ with $N>1$, which describes $N$ coincident branes. Topological charge pumping is the condensed-matter phenomenon in which slowly cycling a Hamiltonian's parameters around a point where the energy gap closes moves an exactly integer amount of charge per cycle. That integer is set by the Berry curvature of the ground states, a geometric quantity that can be computed numerically for any finite Hamiltonian.
The proposal treats each wall in $\tau$ as the gap-closing point of a pump. We conjecture, and then test numerically, that the jump of the flavor-graded twisted index across a wall equals the Chern number of the bundle of supersymmetric ground states over a small parameter loop that encircles the place where the wall's gap closes. If this holds, it gives a physical meaning to wall-crossing: it is charge pumped by an adiabatic cycle. It also gives a second way to compute non-Abelian wall-crossing jumps, through Berry curvature of a regulated quantum mechanics, in exactly the regime where the residue prescription is currently intractable.
Background and the gap
The 3d twisted index and wall-crossing computes the twisted index of 3D $\mathcal{N}=2$ theories on $S^1\times\Sigma$, where $\Sigma$ is a closed surface, as a function of the real FI parameter $\tau$. The index is a supersymmetric Witten index of an effective one-dimensional quantum mechanics. Because $\tau$ is not $Q$-exact, it deforms the integration contour and changes the Jeffrey–Kirwan (JK) residue rule at the contour's boundary. The JK rule is the prescription that selects which poles of the contour integral contribute. This change produces jumps at integer $\tau$, and the jumps match changes in stability conditions for moduli spaces of stable pairs. That paper derives a systematic formula only for Abelian groups, because non-Abelian poles collide. It treats $\tau$ as a static parameter.
Twisted indices of 3d N=4 gauge theories and enumerative geometry of quasi-maps supplies an independent check. It identifies the index with a virtual Euler characteristic of quasi-map moduli spaces. It is limited to theories with isolated vacua, but it gives an algebraic-geometry answer we can compare against.
On the pumping side, the role of each paper is as follows:
- Topological characterization of periodically driven quantum systems supplies the invariants we compute: Chern and winding numbers of states over a synthetic parameter torus.
- Quantized Pumping and Topology of the Phase Diagram for a System of Interacting Bosons is the closest conceptual precedent. It uses quantized pumping around gap-closing points to map the phase diagram of a strongly interacting system. We apply the same logic to the phase diagram of a supersymmetric quantum mechanics in $\tau$.
- Interplay between disorder and topology in Thouless pumping on a superconducting quantum processor and High-Order Topological Pumping on a Superconducting Quantum Processor constrain any hardware ambition. Coherence lasts roughly three cycles, and measured transport degrades sharply on complicated paths.
The research file finds no literature connecting pumping to localization. It also notes that nobody has studied $\tau(t)$ as a cyclic drive. That gap is real, but stated that way it is vague. What follows is a sharp version of it.
Core claim
Let $H(\tau,\theta)$ be the supersymmetric quantum mechanics obtained by reducing a 3D $\mathcal{N}=2$ gauge theory on $\Sigma$. Here $\tau$ is the real FI parameter and $\theta$ is the holonomy of the background field for the topological $U(1)$ symmetry around $S^1$. Together, $\tau+i\theta$ forms the complexified FI parameter.
According to the wall-crossing paper, deformations along the $\theta$ direction leave the index invariant wherever the system is gapped. The index can change only where the gap closes, on the walls $\tau=n$. Take any closed loop $\gamma$ in the $(\tau,\theta)$ plane that crosses the wall $\tau=n$ once in each direction. Such a loop winds once around $\theta$ on one side of the wall and once in reverse on the other side. Consider the bundle of supersymmetric ground states with fixed flavor charge $f$, which has rank $d_f$, and its Berry connection $\mathcal{A}$. We conjecture
$$C_f(\gamma)=\frac{1}{2\pi}\int_{D_\gamma}\mathrm{tr}\,\mathcal{F}=I_f(n^+)-I_f(n^-),$$
where $\mathcal{F}=d\mathcal{A}+\mathcal{A}\wedge\mathcal{A}$ is the Berry curvature, $D_\gamma$ is the region the loop encloses (including the singular gap-closing locus, suitably regulated), and $I_f(n^\pm)$ is the coefficient of flavor fugacity $y^f$ in the twisted index just above and just below the wall.
In plain terms, the integer number of ground states that the index gains or loses at a wall equals the integer charge pumped when the parameters are cycled once around the wall's gap closing. The left-hand side can be computed from a finite Hamiltonian. The right-hand side requires the JK residue prescription, and that prescription is unavailable in non-Abelian cases.
The physical intuition comes from what happens at a wall. In supersymmetric quantum mechanics, wall-crossing occurs when states delocalize into the non-compact Coulomb direction. That is exactly a gap closing into a continuum. A Thouless pump is likewise controlled by the topology of the ground-state bundle around a gap closing.
The claim is new in two respects:
- It reinterprets static index jumps as an adiabatic transport quantity, which the wall-crossing paper never considers.
- It turns the computation of non-Abelian jumps from combinatorics of colliding poles into numerical linear algebra on a regulated Hamiltonian.
Research program
Stage 1: Abelian theories with a known answer
We start with $U(1)$ theories with $N_f$ chirals on $\Sigma=S^2$ and $T^2$, where the Abelian wall-crossing formula is known. We construct $H(\tau,\theta)$ explicitly. Reducing on $\Sigma$ with gauge flux $m$ through it gives finitely many Landau-level-like zero modes per flux sector. The non-compact real scalar $\sigma$ is put on an interval $[-L,L]$ with supersymmetry-preserving boundary conditions.
This regulator is the main technical risk. Wall-crossing lives in the large-$\sigma$ asymptotics, so we must show that $C_f$ converges as $L\to\infty$, and in particular that the boundary at $\sigma=\pm L$ does not itself pump spurious charge. We will check convergence directly. We will also compare two different boundary conditions. Agreement between them is required before any stage-2 result is trusted.
We then compute $C_f(\gamma)$ by exact diagonalization on a discretized $(\tau,\theta)$ grid, using a gauge-invariant lattice evaluation of the Berry flux. We compare the result with the JK jumps.
Stage 2: non-Abelian theories and coincident branes
Next come $U(2)$ and $U(3)$ theories with $N_f$ fundamentals. In the brane language of the coincident-brane literature, $N$ is the number of coincident branes, and the vortex moduli spaces are their bound states. Here the JK prescription suffers from pole collisions, and the wall-crossing paper offers no formula. We compute the Chern numbers and cross-check them in two ways:
- Higgs-phase check. In any chamber with isolated vacua, we compare the indices on both sides of a wall against the quasi-map virtual Euler characteristics, using the method of the quasi-map paper for $\mathcal{N}=4$ points.
- Gluing check. We verify that summing $C_f$ over all walls between two chambers reproduces the total index difference between those chambers, wherever both chamber indices can be evaluated independently.
Stage 3: scaling with $N$
We tabulate how the jump sizes grow with the gauge rank $N$ and the flux degree. The aim is a conjectural closed-form non-Abelian wall-crossing formula, fit from exact integers rather than from approximate numbers.
We also compare our computational cost with the residue approach. Our cost grows polynomially in the truncated Hilbert-space dimension at fixed flux cutoff. The residue approach grows combinatorially in the number of colliding poles. Whether polynomial scaling actually wins at accessible sizes is an empirical question, and we will measure it rather than assume it.
The QFT-complexity direction enters only at this stage, as a method lens. Free-variable reformulations, as in Free variables and the two-matrix model, and master-field reformulations, as in Stochastic master fields, are candidates for computing the large-$N$ limit of the ground-state Berry curvature. We treat them as a speculative extension, not a deliverable.
Baselines
Do-nothing reference. We evaluate the index separately on each side of a wall, with the JK prescription applied as if no boundary contributions existed. Any jump this reference misses is something the method must detect.
Strongest existing method for Abelian theories. This is the modified JK residue prescription with boundary contributions from The 3d twisted index and wall-crossing. We implement it symbolically for the same theories, on the same $\Sigma$ and over the same flux range. Beating it on Abelian theories is not the goal. The goal is exact integer agreement on every wall and every flavor charge we can reach. For context on numerical performance, there is no reported result on Lacuna for computational cost or accuracy of JK wall-crossing evaluations; the paper is purely analytic. We will therefore report our own timing for both methods on identical hardware.
Strongest existing method for non-Abelian theories. This is the quasi-map enumerative computation of Twisted indices of 3d N=4 gauge theories and enumerative geometry of quasi-maps, applied in chambers with isolated vacua. The criterion for success here is producing jumps in cases where neither the residue formula nor the quasi-map construction gives them, while still matching both wherever either one applies.
Ablations isolating the key ingredient. These test that the integer really comes from the gap closing:
- A loop that does not enclose a wall must give $C_f=0$.
- A loop varying only Q-exact directions, such as $\theta$ at fixed $\tau$ inside one chamber, must give zero.
- Changing the $\sigma$-interval boundary condition must not change $C_f$ once $L$ is large.
If any of these ablations returns a nonzero integer, the identification fails.
Why not just…?
Why not just run the JK residue computation for non-Abelian groups by brute force? This is the most serious threat. For $U(2)$ with few flavors, careful symbolic residue evaluation may well succeed. The wall-crossing paper identifies pole collisions as the obstacle to a systematic formula, which does not mean small cases are impossible. We will run brute-force residues alongside our method for every $U(2)$ case and report where each one stops scaling. If brute force keeps pace through $U(3)$, the computational argument collapses, and the contribution reduces to the physical reinterpretation. We say so up front.
Why not just Abelianize, writing the non-Abelian index as a Weyl-symmetrized Abelian one? Abelianization handles the index inside a chamber. The difficulty, however, lies precisely in the boundary contributions at walls, where Abelian poles collide after symmetrization. Our gluing check is the direct test. If Abelianized boundary terms reproduce the jumps we compute, that is a valid and cheaper alternative, and the Chern-number method would then serve mainly as an independent confirmation.
Why not just treat wall-crossing as static and skip the pump language? For the index alone, the pump adds nothing new in Abelian cases. What it adds is a quantity that is local in parameter space and computable from a Hamiltonian, plus a transport observable. This mirrors how Quantized Pumping and Topology of the Phase Diagram for a System of Interacting Bosons uses pumping to diagnose phase boundaries that static invariants describe only indirectly. If the Chern number turns out to be merely a restatement of the index, stage 2 will show that it provides no cheaper access, and we will report that outcome.
Why not just simulate the pump on a quantum processor? The hardware evidence argues against making this central. Coherence lasts about three cycles in Interplay between disorder and topology in Thouless pumping on a superconducting quantum processor. In High-Order Topological Pumping on a Superconducting Quantum Processor, measured transport is $0.964$ against a simulated $0.985$ on a diagonal path, but falls to $0.555$ against $0.836$ on a non-diagonal path. Our loops are non-diagonal in exactly this sense. Classical exact diagonalization on the regulated quantum mechanics is therefore the primary tool. A one-chiral $U(1)$ demonstration with a few qubits is a stretch goal, and success would mean pumped charge within the $0.964$ diagonal-pump fidelity reported there.
Decisive checks
The idea survives or dies on three outcomes:
- Abelian agreement. For $U(1)$ with $N_f\le 4$ on $S^2$ and $T^2$, every wall's Chern number must agree exactly with the JK jump for every flavor charge, and must be stable as $L\to\infty$ and under a change of boundary condition. A single robust mismatch refutes the conjecture as stated.
- Non-Abelian consistency. For $U(2)$, the Chern numbers summed across walls must reproduce chamber-to-chamber differences that the quasi-map computation fixes independently.
- Cost scaling. Wall-clock comparisons with brute-force residues at $U(2)$ and $U(3)$ determine whether the method has computational value or only conceptual value.
A positive outcome on all three would yield the first conjectured non-Abelian wall-crossing data for the 3D twisted index. It would also establish a dictionary in which index walls are pump singularities.
References
- The 3d twisted index and wall-crossing
- Twisted indices of 3d $$ \mathcal{N} $$ = 4 gauge theories and enumerative geometry of quasi-maps
- Topological characterization of periodically driven quantum systems
- Quantized Pumping and Topology of the Phase Diagram for a System of Interacting Bosons
- Interplay between disorder and topology in Thouless pumping on a superconducting quantum processor
- High-Order Topological Pumping on a Superconducting Quantum Processor
- Free variables and the two-matrix model
- Stochastic master fields