Many gauge theories in string theory are described by a quiver, a diagram that lists several gauge groups and the matter fields connecting them. When such a theory is placed in three spacetime dimensions, each gauge group acquires a "magnetic" symmetry that counts magnetic flux. Adding monopole operators (operators that insert a unit of magnetic flux) to the theory breaks this continuous symmetry down to a finite cyclic group. A 't Hooft anomaly is a conserved obstruction that forbids a trivial, featureless vacuum. For a single-node model, the $\mathbb{CP}^{N-1}$ sigma model, it is known exactly when this obstruction survives the breaking: when the monopole charge $n$ shares a factor with $N$, that is $\gcd(n,N)>1$ (Global anomaly matching in the higher-dimensional CP^{N-1} model). The corresponding rule for quivers with many gauge groups $\prod U(N_i)$ is unknown, and that paper names it as an open problem.
This proposal finds that rule for toric quivers. These are the quivers read off from D-branes probing toric Calabi-Yau singularities. They are drawn as bipartite graphs on a torus called dimer models, and they come in several equivalent "phases" related by Seiberg duality (a change of description that preserves the low-energy physics). Seiberg duality replaces a node's rank $N_k$ by $N_{f,k}-N_k$, where $N_{f,k}$ is the node's number of flavors. So any node-by-node gcd rule would give different answers in dual phases that describe the same physics. The core claim is that the correct multi-node rule must be a gcd of the monopole charge vector against an integer lattice built from the quiver's arrows and its anomaly-free rank vectors. A lattice invariant of that kind is automatically unchanged under the integer, invertible map that Seiberg duality induces on ranks. We will derive this criterion, prove its duality invariance, and test it across every known dual phase of the standard benchmark geometries.
Core Claim
Label the quiver's nodes by $i$. Let $a_{ij}$ be the number of arrows from node $i$ to node $j$, $N=(N_i)$ the rank vector, and $m=(m_i)$ the magnetic charges of the monopole added to the superpotential, the term in a supersymmetric theory's action that fixes its interactions. Two ingredients determine which discrete symmetries act faithfully and which anomalies they carry.
The first is gauge anomaly cancellation. In four dimensions this is the condition $\sum_j (a_{ij}-a_{ji})N_j=0$ at every node. It confines the allowed rank vectors to an integer sublattice $\Lambda$, spanned by the regular brane and the fractional branes. In three dimensions, the matching condition on Chern-Simons levels, the integer couplings of three-dimensional gauge fields, restricts $\Lambda$ further.
The second is the center of $\prod U(N_i)$, the elements that commute with everything in the group, together with the charges of the bifundamental fields under it. These fix the group that actually acts on the theory. That group is also where the one-form symmetry lives, a symmetry that acts on line operators rather than point operators.
We conjecture that the surviving anomaly is controlled by a single integer
$$ g(m;N) = \gcd\big(\text{elementary divisors of } [\,A^{\top}\,|\,N\,]\ \text{paired with } m\big), $$
Here $A$ is the antisymmetrized arrow matrix. The elementary divisors are the diagonal entries of the Smith normal form, the canonical diagonal form of an integer matrix under invertible integer row and column operations. "Paired with $m$" means the monopole charges are expressed in the basis in which that form is diagonal. The anomaly survives iff $\gcd(n,g)>1$. The exact form of $g$ is the output of the derivation, not an input. What the proposal commits to in advance are three properties.
First, $g$ reduces to $N$ for the single-node model with $N$ flavors, which recovers Global anomaly matching in the higher-dimensional CP^{N-1} model.
Second, $g$ is invariant under urban renewal, the local move on the dimer graph that implements Seiberg duality, once $m$ is transported by the duality map. The reason is that $N_k\to\sum_j a_{jk}N_j-N_k$ is an integer involution with determinant $\pm1$, and elementary divisors do not change under such maps.
Third, a naive node-local criterion $\min_i\gcd(n,N_i)$ does change under duality, so it must fail somewhere on the benchmarks.
Research Program
The work has three stages.
The first stage is analytic. We reduce four-dimensional toric quivers to three dimensions and study them with monopole superpotentials. We couple the quiver to background gauge fields for its flavor symmetries, the residual $\mathbb{Z}_n$ magnetic symmetries, and the one-form center symmetry. We then compute the obstruction to gauging them, as Global anomaly matching in the higher-dimensional CP^{N-1} model does, including the $\mathbb{RP}^2$ non-orientable background that detects the reflection-magnetic anomaly.
The method for reading global forms off monopole center charges comes from Generalized symmetries and anomalies of 3d N=4 SCFTs. That paper computes the global form of the symmetry group and its mixed anomalies from node balance, but only for eight-supercharge "good" quivers, and it does not deform by monopoles. We carry its center-charge bookkeeping over to four-supercharge chiral toric quivers.
The central technical question is how a monopole deformation at a dualized node maps across the duality. In three-dimensional Seiberg-type dualities, monopoles of the electric theory map to gauge-singlet fields of the magnetic theory. So a pure monopole deformation on one side can become an ordinary mass-like term on the other. Tracking this map is where the conjecture could break, and it is what the node-balance reading alone does not supply.
The second stage is computational. We build a pipeline that starts from each dimer phase. From the permutation-triple encoding used in Metaheuristic Generation of Brane Tilings, we extract $a_{ij}$, $\Lambda$ and the dualizable nodes. We then compute $g$ by Smith normal form for all monopole charge vectors up to a cutoff. Finally, we propagate $(N,m)$ through every urban renewal move to test invariance.
Anomaly cancellation enters here as a constraint rather than as the object of study. D-brane gauge theories from toric singularities and toric duality guarantees that rank vectors from partial resolution are anomaly-free, and Toric duality as Seiberg duality and brane diamonds shows that local anomaly cancellation is preserved across moves. We rely on both results and ask the harder, global question.
The third stage places the new invariant beside the known ones. INVARIANTS OF TORIC SEIBERG DUALITY shows two things stay fixed under duality: the Klein $j$-invariant of the dimer's isoradial embedding (a drawing of the graph fixed by the R-charges, the charges under the symmetry that rotates supersymmetry), and the number field generated by those R-charges. Both are computed at the four-dimensional conformal point. We ask whether $g$ is independent of these continuous-geometry invariants or correlated with them. One example is whether theories with equal $j(\tau_R)$ but different fractional-brane content separate by $g$. Unlike the $j$-invariant, $g$ needs no isoradial embedding. It should therefore remain computable in non-minimal phases such as those of $L^{2,2,2}$, where that paper reports the embedding degenerates.
Baselines
The naive reference point is to apply the single-node criterion of Global anomaly matching in the higher-dimensional CP^{N-1} model node by node. It declares the anomaly surviving if $\gcd(n,N_i)>1$ at any node carrying monopole charge. It is the natural first guess and costs nothing. We run it on exactly the same phases, rank vectors and monopole charges as the proposed criterion. The proposal beats it if the naive rule gives different verdicts in at least one pair of dual phases while $g$ gives the same verdict in both, and if $g$ agrees with the direct anomaly computation in every checked case.
The strongest existing method is the center-charge and node-balance analysis of Generalized symmetries and anomalies of 3d N=4 SCFTs, applied directly to each phase and read before any monopole deformation. It is the only method in the file that extracts global forms and mixed anomalies from quiver data. It is run on each phase separately, with the same backgrounds. Beating it means two things. First, predicting how its undeformed mixed anomalies descend to the $\mathbb{Z}_n$ subgroup, which it does not address. Second, showing that its phase-by-phase answers are tied together by one invariant, rather than needing a separate computation per phase. No reported result on Lacuna covers discrete anomaly survival across toric dual phases, so there is no prior number to beat. The benchmark is agreement with first-principles anomaly computations on the listed geometries.
The key ablation removes the anomaly-cancellation lattice. It computes $g$ from the bare rank vector without restricting to $\Lambda$ or imposing three-dimensional level matching. If this version is equally duality-invariant and equally accurate, then gauge anomaly cancellation plays no role, and the proposal shrinks to a statement about rank gcds.
Why not just…?
Why not just match continuous flavor and R-symmetry anomalies, as in orientifold QCD? Nonsupersymmetric Seiberg duality, orientifold QCD, and noncritical strings verifies dualities by matching $SU(N_f)^3$ and $U(1)_R$ anomalies. The monopole deformation explicitly breaks the magnetic $U(1)$, so the continuous anomalies it relies on no longer exist. The single-node analysis shows the discrete remnant can survive or vanish depending on arithmetic that continuous matching cannot see.
Why not just extend the node-balance reading and skip the deformation analysis? This is a real threat. If the center-charge method of Generalized symmetries and anomalies of 3d N=4 SCFTs carries over to four-supercharge quivers unchanged, much of the global-form computation is inherited. What it does not provide is the map of monopole deformations across dual phases, where monopoles turn into singlets. The strongest-baseline comparison measures exactly how much the deformation analysis adds. If it adds nothing, we will report that the invariant is simply the inherited global form.
Why not just Higgs or add mass terms to break the symmetry? Giving a field an expectation value is partial resolution: it changes the toric diagram and hence the geometry (D-brane gauge theories from toric singularities and toric duality). The question here is robustness at fixed geometry, which Higgsing cannot pose.
Why not just test by brute-force enumeration of all dual quivers? Enumeration is a check on the invariant, not a replacement for it, and it hits the factorial wall at $d=24$ fields reported in Metaheuristic Generation of Brane Tilings. An invariant computed from one phase replaces searching over all phases.
Decisive Checks
The benchmarks are those used across the file. They are the three phases of the Suspended Pinched Point; the cones over $dP_0$ through $dP_3$, including all four $dP_3$ phases shown in Symmetries of Toric Duality to come from one toric diagram; $F_0$; and a sample of the larger tilings reachable by the metaheuristic, up to the $d=26$ example.
The first check is the single-node limit. The criterion must reproduce $\gcd(n,N)>1$ exactly.
The second check covers every benchmark phase, every anomaly-free rank vector with ranks up to a cutoff, and every monopole charge vector with $|m_i|\le 6$. For each case we compare the verdict from $g$ with a direct anomaly-inflow computation in that phase. The conjecture fails outright if any dual pair disagrees in its direct computations under our monopole map. Such a disagreement would mean the map itself is wrong, which is a result worth publishing.
The third check is whether the naive node-local rule is falsified in at least one dual pair. We expect this first in $dP_3$, where duality produces large rank changes. If it is never falsified, the proposal's novelty over the naive rule collapses, and we will say so.
The fourth check is whether $g$ separates theories that the $j$-invariant and the R-charge number field identify. If it does, the discrete anomaly is independent information about the duality class.
References
- Global anomaly matching in the higher-dimensional <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mrow><mml:mi mathvariant="double-struck">C</mml:mi><mml:msup><mml:mi mathvariant="double-struck">P</mml:mi><mml:mrow><mml:mi>N</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msup></mml:mrow></mml:math> model
- Generalized symmetries and anomalies of 3d $\mathcal{N}=4$ SCFTs
- Metaheuristic generation of brane tilings
- D-brane gauge theories from toric singularities and toric duality
- Toric duality as Seiberg duality and brane diamonds
- INVARIANTS OF TORIC SEIBERG DUALITY
- Nonsupersymmetric Seiberg duality, orientifold QCD, and noncritical strings
- Symmetries of Toric Duality