--- title: 'Quiver Subtraction: Operations in 3d N=4 Gauge Theories' url: https://www.emergentmind.com/topics/quiver-subtraction type: topic --- # Quiver Subtraction: Operations in 3d N=4 Gauge Theories Quiver subtraction denotes a family of diagrammatic operations on quiver gauge theories, developed primarily for \(3d\;\mathcal N=4\) moduli spaces. In the foundational construction, one starts from nested unitary quivers \(Q'\) and \(Q\) with \(\mathcal C_{Q'}\subseteq \mathcal C_Q\), defines a subtraction quiver \(D=Q-Q'\) by nodewise rank subtraction, and interprets \(\mathcal C_D\) as the transverse slice to \(\mathcal C_{Q'}\) inside \(\mathcal C_Q\). Physically, this describes singular loci where extra massless states appear and identifies the Higgs factor opening there [1803.11205]. Subsequent work generalized the idea in several directions: quotient quiver subtraction for Coulomb-branch hyper-Kähler quotients, orthosymplectic and classical-group variants, and a distinct Higgs-branch subtraction algorithm for simply-laced unitary quivers with loops [2308.05853], [2409.15419], [2503.19954], [2603.08774], [2409.16356]. ## 1. Formal definition and local geometric meaning The original definition applies to two **unitary** quivers \(Q\) and \(Q'\) with the same number of gauge nodes, the same connectivity, and an inclusion \[ \mathcal C_{Q'} \subseteq \mathcal C_Q. \] The defining conditions are as follows [1803.11205]. - **Gauge ranks**: the gauge-node ranks of \(Q'\) equal those of \(Q\), except on a connected subset where the ranks are strictly smaller. - **Flavor data**: the flavor nodes of \(Q'\) agree with those of \(Q\), except that nodes adjacent to the lowered ranks acquire extra flavor so that the quiver remains admissible. - **Nontrivial support**: at least one flavor is attached to a node where \(Q\) has larger rank than \(Q'\). The subtraction quiver \(D=Q-Q'\) is then defined on the same underlying graph, with gauge ranks \[ \mathrm{rank}(D_i)=\mathrm{rank}(Q_i)-\mathrm{rank}(Q_i'), \] and with flavor nodes inherited from \(Q\). The paper’s explicit example is \[ Q:\quad \bullet_1-\bullet_2^{\,f=1}-\bullet_2^{\,f=1}-\bullet_1,\qquad Q':\quad \bullet_1^{\,f=1}-\bullet_1-\bullet_1-\bullet_1^{\,f=1}, \] so that \[ D=Q-Q'=\bullet_1^{\,f=1}-\bullet_1^{\,f=1}. \] Its Coulomb branch is \[ \mathcal C_D=a_2, \] the closure of the minimal nilpotent orbit of \(\mathfrak{sl}_3\), equivalently the \(A_2\) Kleinian singularity. The geometric interpretation is local. If \(\mathcal C_{Q'}\) is a singular subvariety of \(\mathcal C_Q\), the conjectural statement is that \[ \mathcal C_{Q-Q'} \cong \text{slice transverse to } \mathcal C_{Q'} \text{ inside } \mathcal C_Q. \] Near a point \(x\in \mathcal C_{Q'}^\circ\), one has smooth-equivalence of singularity structure in the form \[ \mathcal C_Q \sim \mathcal C_{Q'} \times \mathcal C_D. \] The corresponding mixed branch is written as \[ \mathcal B(Q)=\mathcal C_{Q'}\times \mathcal C_D. \] In this sense, quiver subtraction turns a local singularity problem on \(\mathcal C_Q\) into a new quiver whose Coulomb branch records the transverse Higgs physics. ## 2. Kraft–Procesi transitions, nilpotent orbits, and infinite-coupling physics A principal motivation for quiver subtraction is the analysis of theories with eight supercharges at infinite coupling. In \(5d\), \(1/g^2\) has dimensions of mass, and taking \(1/g^2\to 0\) can produce new massless states, enlarge the Higgs branch, and enhance global symmetry. In \(6d\), \(1/g^2\) has the scale of a string tension, and infinite coupling can produce tensionless strings; in the small \(E_8\) instanton transition, the Higgs branch dimension can jump by \(29\) [1803.11205]. The central premise is that the Higgs branch at infinite coupling is often described by the Coulomb branch of a \(3d\;\mathcal N=4\) quiver, so subtraction becomes a tool for identifying the new local sector. The construction reformulates **Kraft–Procesi transitions** as quiver operations. For \(\mathfrak{sl}_5\), the paper considers \[ \mathcal C_Q = \overline{\mathcal O}_{(2^2,1)}, \qquad \mathcal C_{Q'} = \overline{\mathcal O}_{(2,1^3)}, \] and finds that \(D=Q-Q'\) has \[ \mathcal C_D=a_2, \] reproducing the transverse slice from \(\overline{\mathcal O}_{(2^2,1)}\) to \(\overline{\mathcal O}_{(2,1^3)}\). The same logic extends from the classical Lie algebras to exceptional cases. In \(E_6\), if \(Q\) has \(\mathcal C_Q=\overline{\mathrm{n.min}_{E_6}}\) and \(Q'\) has \(\mathcal C_{Q'}=\overline{\mathrm{min}_{E_6}}\), then \(D=Q-Q'\) satisfies \[ \mathcal C_D=a_5=\overline{\mathrm{min}_{A_5}}. \] The paper also emphasizes that the formalism reaches **non-special** nilpotent orbits, at least for the height-\(\le 2\) cases studied. A separate application concerns integrating out a massive quark in \(5d\;\mathcal N=1\) SQCD while keeping the gauge coupling infinite. For \[ H_9=SU(3)_{k=1/2}\text{ with }N_f=9,\qquad H_8=SU(3)_{k=1}\text{ with }N_f=8, \] the subtraction \(D_9=Q_9-Q_8\) is defined after padding with zero-rank nodes where needed and decoupling a \(U(1)\) if necessary. The resulting Coulomb branch is \[ {}^{3d}(D_9)=\mathbb C^{32}. \] Because the slice is smooth rather than singular, no extra massless hypermultiplets appear and no new Higgs branch opens at that step. In the paper’s formulation, quiver subtraction therefore distinguishes singular transitions, which produce emergent Higgs factors, from smooth deformations, which do not. ## 3. Quotient quiver subtraction and Coulomb-branch hyper-Kähler quotients A later development replaced the transverse-slice viewpoint by a direct diagrammatic realization of Coulomb-branch gauging. In **quotient quiver subtraction**, one computes an \(SU(n)\) hyper-Kähler quotient of a Coulomb branch by subtracting a special auxiliary quiver from a target unitary magnetic quiver [2308.05853]. The field-theoretic benchmark is the Weyl-integration formula \[ HS_{\mathcal C///SU(n)}(x_1,\dots,x_r;t) = \int_{SU(n)} d\mu_{SU(n)} \frac{HS_{\mathcal C}(x_1,\dots,x_r;y_1,\dots,y_{n-1};t)} {PE[\chi([1,0,\dots,0,1]_{SU(n)})t^2]}. \] When gauging is complete, \[ | \mathcal C///SU(n) | = |\mathcal C| - (n^2-1), \] where \(|\cdot|\) denotes quaternionic dimension. The subtraction object is the **quotient quiver** \[ (1)-(2)-\cdots-(n)-\cdots-(2)-(1), \] described in the paper as the \(U(n)\) quotient quiver. It is **bad** in the Gaiotto–Witten sense because its central node has balance \(-2\), so the monopole formula is not well-defined for its Coulomb branch as a standalone theory. This is a formal subtraction object rather than a physical target theory. The target quiver, by contrast, must be **good** or **ugly**, and the subtraction is constrained by explicit selection rules: long framing, an external leg \((1)-(2)-\cdots-(n)\), subtraction only along single-laced edges, a junction rule requiring alignment with a rank-\(2\) node if the quotient passes a junction, survival of nodes carrying adjoint hypers, and nonnegative resulting ranks after rebalancing. The conceptual workflow is: unframe the target quiver, identify an external leg matching \((1)-(2)-\cdots-(n)\), overlay the quotient quiver, subtract node by node, rebalance surviving nodes by attaching appropriate flavors, and interpret disconnected outputs as Cartesian products. A notable feature is that the answer need not be a single quiver. If several valid alignments exist, the hyper-Kähler quotient is the **union** of the corresponding Coulomb branches, \[ \mathcal C(Q_1)\cup \mathcal C(Q_2)\cup \cdots, \] with Hilbert series computed by inclusion–exclusion. The paper states that quotient subtraction and Kraft–Procesi subtraction do **not generally commute**, while quotient subtraction often commutes with folding and often commutes with discrete gauging in the examples considered. The method was tested on free-field moduli spaces, nilpotent orbit closures of types \(A,B,D\) and exceptional type, Slodowy slices, intersections, and affine Grassmannian slices. For example, the paper conjectures \[ \mathbb H^{2k}///SU(2)=\overline{\min D_k},\qquad k\ge 2, \] checked for \(k=2,\dots,6\), and \[ \mathbb H^{2k+1}///SU(2)=\overline{\min B_k},\qquad k\ge 2, \] checked for \(k=1,\dots,5\). These examples established quotient quiver subtraction as a distinct operation from Kraft–Procesi subtraction: it is designed to implement gauging rather than to describe adjacent symplectic-leaf transitions. ## 4. Orthosymplectic and classical-group generalizations The unitary \(SU(n)\) construction was extended in several non-equivalent directions. For **unframed orthosymplectic quivers**, orthosymplectic quotient quiver subtraction gauges a subgroup of the IR Coulomb-branch global symmetry by subtracting a specially constructed orthosymplectic quotient quiver [2409.15419]. The quotient quivers found in that work correspond to \[ SU(2),\qquad SU(3),\qquad G_2,\qquad SO(7). \] The algorithm requires alignment against a maximal leg, with \(D\)-type nodes aligned with \(D\)-type nodes and \(C\)-type nodes aligned with \(C\)-type nodes, followed by nodewise subtraction, positivity of the resulting ranks, nonnegative imbalance, and rebalancing **only the nodes not participating in the subtraction**, always by adding a \(C_1\) node. If the quotient quiver extends one node past a junction, the result is the **union** of all possible alignments. For **framed orthosymplectic quivers**, a different procedure was introduced under the name **orthosymplectic quotient quiver subtraction** [2503.19954]. Here the quotient quivers are identified with magnetic quivers for class \(\mathcal S\) theories on cylinders with maximal punctures: \[ G=\mathrm{SO}(2n),\quad \mathrm{SO}(2n+1),\quad \mathrm{Sp}'(n). \] The subtraction algorithm aligns the quotient quiver with a long leg beginning with a balanced maximal chain, subtracts ranks node by node, reevaluates the gauge-node types after subtraction, rebalances using **flavor nodes**, and, if the quotient extends one node past a junction, takes the **union** of the resulting cones. A distinctive feature is Lie-type conversion under subtraction: \[ D_n-D_m \Rightarrow B_{n-m},\qquad B_n-B_m \Rightarrow D_{n-m},\qquad C_n-C_m \Rightarrow C_{n-m}. \] The paper explicitly states that framed and unframed orthosymplectic subtraction are different procedures: framed orthosymplectic quivers do not admit the unitary-style overall \(\mathrm U(1)\) shift, the framed quotient quivers have rank exactly \(\dim(G)\), and rebalancing is done with flavors rather than a \(C_1\) gauge node. A further extension addresses **classical groups** acting on unitary magnetic quivers through Type IIB constructions with \(\mathrm{O5}\) planes [2603.08774]. In this setting, quotient quiver subtraction is no longer solely subtraction; one must also change the graph type. | Gauged subgroup | Quotient quiver tail | Additional post-subtraction step | |---|---|---| | \(Sp(n)\) | \(1-2-\cdots-(2n-1)-2n\) | split \(U(j)\) into \(U(\lceil j/2\rceil)\) and \(U(\lfloor j/2\rfloor)\) | | \(SO(2n)\) | \(1-2-\cdots-(2n-2)-(2n-1)\) | all edges attached to \(U(2n)\) become doubly laced | | \(SO(2n+1)\) | \(1-2-\cdots-(2n-1)-2n\) | all edges attached to \(U(2n+1)\) become doubly laced | | \(Sp(n)+\frac12\mathsf F\) | \(1-2-\cdots-(2n-2)-(2n-1)\) | surviving \(U(2n)\) is halved to \(U(n)\), with doubly laced attached edges | For \(Sp(n)\), the total rank of the quotient quiver is \[ n(2n+1)=\dim Sp(n), \] and no extra rebalancing node is needed. For \(SO(2n)\), the total rank matches \[ \dim SO(2n)=n(2n-1), \] while for \(SO(2n+1)\) it matches \[ n(2n+1)=\dim SO(2n+1). \] For \(Sp(n)+\frac12\mathsf F\), the procedure combines subtraction with rank halving and lacing change. The physical backbone of these rules is the Type IIB brane system with orientifold 5-planes: \(\mathrm{O5}^-\) yields \(Sp(n)\), \(\mathrm{O5}^+\) yields \(SO(2n)\), \(\widetilde{\mathrm O5^+}\) yields \(SO(2n+1)\), and \(\widetilde{\mathrm O5^-}\) yields \(Sp(n)\) with a half-hyper. ## 5. Higgs-branch subtraction, minimal transitions, and global data A distinct development introduced **quiver subtraction on the Higgs branch** for simply-laced unitary \(3d\;\mathcal N=4\) quivers with loops [2409.16356]. The target theories have gauge group \[ G=\prod_i U(n_i)/U(1), \] with bifundamental edges and adjoint loops only. The purpose is to reconstruct the **Hasse diagram of Higgs-branch strata** and the associated minimal transverse slices directly from the quiver. In this setting, a Higgs-branch leaf corresponds to a partially Higgsed residual theory, and the slice between two leaves is the Higgs branch of a transverse quiver. The algorithm classifies all minimal Higgsing transitions by three local rules. The first rule is **adjoint Higgsing on a single node**: \[ U(n)\to U(n-m)\times U(m),\qquad 1\le m\le \frac n2. \] If \(m=n-m\), the reduced normalizer is \(S_2\subset S_n\), and the slice is \[ c_g \cong \mathbb C^{2g}/\mathbb Z_2. \] If \(m\neq n-m\), the slice is the non-normal variety \(m_g\), whose normalization is \(\mathbb C^{2g}\). The second rule is **bifundamental Higgsing between two nodes**: \[ U(n_1)\times U(n_2)\to U(n_1-m)\times U(n_2-m)\times U(m), \] with \(1\le m\le \min(n_1,n_2)\). The transverse slice is the Kleinian singularity \[ a_g. \] The paper stresses that loops constrain minimality: if the higher-rank node has loops, Rule 2 is not minimal and Rule 1 must be applied first; if the lower-rank node has loops, Rule 2 is minimal only for the maximal subtraction \(m=\min(n_1,n_2)\). The third rule is **affine \(ADE\) Higgsing**. If the quiver contains \(n\) copies of a minimal balanced affine \(ADE\) quiver \(\mathsf Q_{ADE}\), one may subtract \(m\) copies, \[ U(n_0)\times\cdots\times U(n_k) \to U(n_0-mh_0^\vee)\times\cdots\times U(n_k-mh_k^\vee)\times U(m), \] where \[ \sum_{j=0}^k c_{ij}h_j^\vee=0,\qquad h_0^\vee=1. \] The transverse slice is the corresponding affine \(ADE\) quiver, hence the \(ADE\) Kleinian singularity. Again, loops can obstruct minimality: if a node with dual Coxeter label \(>1\) carries loops, Rule 3 is not minimal and Rule 1 must precede it. A major feature of this Higgs-branch algorithm is sensitivity to **global data**. The local rules alone do not determine the effective slice in general, because monodromy around a leaf may act nontrivially on the slice. This is encoded by **decorations** on quivers. A decorated subquiver records a nontrivial global identification; for example, the paper states that in an \(A_3\to C_2\) example the local slice is \(A_3\), but the global decoration induces an \(S_2\) identification, changing the effective slice to \(C_2\). The same decorated data determine the **Namikawa–Weyl group** \[ \mathcal W^X=\prod_i \mathcal W_i, \] where the product runs over all quaternionic \(1\)-dimensional top slices in the Hasse diagram. One of the paper’s principal conclusions is that for simply-laced unitary quiver gauge theories, the Coulomb-branch global symmetry can only be of type \[ U(1),\ SU(n),\ SO(n),\ G_2,\ E_6,\ E_7,\ E_8, \] and **not** \(Sp(n)\) or \(F_4\). The stated reason is that the folded quivers that would produce \(C_n\) or \(F_4\) require decorations on multiple nodes or legs in a way that cannot arise from the single-node decoration mechanism available in the subtraction procedure. ## 6. Related notions: full subquivers, deletion, and mutation-theoretic reduction Outside the \(3d\;\mathcal N=4\) gauge-theory literature, subtraction-like language usually refers not to hyper-Kähler quotients or transverse slices, but to **full subquivers** and **vertex deletion**. These operations are related in spirit, because they isolate a controlled difference between quivers, but they are not the same formal construction. In cluster-theoretic mutation theory, a relevant theorem states that **every finite acyclic quiver is a full subquiver of a quiver mutation equivalent to a bipartite quiver** [1311.0711]. A quiver is written as \[ Q=(Q_0,Q_1,t,h), \] and a full subquiver on \(S\subseteq Q_0\) retains all arrows \(a\in Q_1\) with \(t(a),h(a)\in S\). The proof starts from a finite acyclic quiver \(Q\), repeatedly adds a vertex \(v_a\) on an arrow \(a\) not lying on a maximal path, inserts arrows \[ h(a)\to v_a,\qquad v_a\to t(a), \] mutates at \(v_a\), and then mutates repeatedly at all sources until every maximal oriented path has length \(1\), so the resulting quiver is bipartite. The original quiver persists as a full subquiver throughout. This gives a deletion-based realization of an arbitrary finite acyclic quiver inside a more structured mutation class. A second deletion-based framework arises in the study of **minimal mutation-infinite quivers** [1505.01735]. There the basic relation is \[ P