Knot–Quiver Correspondence

• Knot–quiver correspondence is a framework that maps knots to symmetric quivers, where nodes represent HOMFLY-PT homology generators and arrows encode interaction data.
• The approach employs generating functions and motivic Donaldson–Thomas series to reproduce colored HOMFLY-PT polynomials and LMOV/BPS invariants via explicit variable changes.
• This framework integrates knot theory, quiver representations, and physical models, offering actionable insights for topological strings, M-theory, and 3d N=2 gauge theories.

The knot–quiver correspondence is the identification of a knot $K$ with a symmetric quiver $Q(K)$ such that the generating functions of colored HOMFLY-PT polynomials, colored superpolynomials, and associated LMOV/BPS invariants are reproduced by the motivic Donaldson–Thomas generating series of $Q(K)$ after an explicit specialization of quiver variables. In this formulation, nodes encode generators of uncolored HOMFLY-PT homology, arrows encode interaction data such as linking and self-linking, and motivic DT invariants of the quiver control the product decomposition of knot generating functions. The correspondence was formulated in 2017 and has since developed into a framework connecting knot homology, quiver representation theory, topological strings, M-theory, and $3\text{d }\mathcal N=2$ gauge theories (Kucharski et al., 2017, Sachdeva, 4 May 2025).

1. Formal statement and scope

In its original form, the correspondence asserts that for each knot $K$ there exists a finite symmetric quiver $Q$ whose motivic generating series reproduces the generating function of colored HOMFLY-PT polynomials in symmetric representations. This identifies LMOV invariants of $K$ with motivic DT invariants of $Q$, and therefore turns the integrality of LMOV numbers into a consequence of integrality theorems for symmetric quivers (Kucharski et al., 2017).

The knot side is organized by the Ooguri–Vafa generating function

$$P(x)=\sum_{r=0}^\infty \overline{P}_r(a,q)\,x^r =\exp\!\Big(\sum_{r,n\ge1}\frac{1}{n}\,f_r(a^n,q^n)\,x^{nr}\Big),$$

with

$$f_r(a,q)=\sum_{i,j}\frac{N_{r,i,j}\,a^i\,q^j}{q-q^{-1}}.$$

Equivalently,

$Q(K)$0

The quiver side is the motivic DT series of a symmetric quiver, written either as a sum over dimension vectors or as a quantum-dilogarithmic product with DT exponents. The central claim is that these two structures are the same after a monomial change of variables (Kucharski et al., 2017).

The scope of the correspondence in the original papers includes colored HOMFLY-PT polynomials, refined superpolynomials, classical and refined LMOV invariants, many explicit quiver assignments, and infinite knot families. For all knots for which the corresponding quiver is identified, LMOV integrality for all symmetric representations is automatic. Subsequent surveys emphasize that the same mechanism also organizes quadruply graded homology, knot complement invariants $Q(K)$1, and branch-dependent quiver descriptions for $Q(K)$2-polynomial sectors (Kucharski et al., 8 May 2025).

2. Generating series, plethystics, and the variable map

The quiver generating series for a symmetric quiver with adjacency matrix $Q(K)$3 and $Q(K)$4 nodes can be written as

$Q(K)$5

Its plethystic decomposition defines motivic DT invariants $Q(K)$6: $Q(K)$7 On the knot side, LMOV/BPS integers are extracted by the plethystic logarithm of $Q(K)$8. On the quiver side, DT invariants are extracted by the plethystic logarithm of $Q(K)$9. The correspondence works by identifying these two plethystic structures (Kucharski et al., 2017).

The explicit knot–quiver identification is

$Q(K)$0

with specialization

$Q(K)$1

Thus

$Q(K)$2

In the refined setting, the same quiver controls the quadruply graded Poincaré polynomials

$Q(K)$3

and the colored superpolynomial is recovered by

$Q(K)$4

This is the refined form in which a single symmetric quiver controls both unrefined and homological knot data (Kucharski et al., 2017).

The integrality mechanism is immediate. Since the exponents $Q(K)$5 are integers for symmetric quivers, the coefficients in the plethystic logarithm of $Q(K)$6 are integer linear combinations of quiver DT invariants. Surveys emphasize this as the conceptual explanation of LMOV integrality in the KQ framework (Sachdeva, 4 May 2025).

3. Quiver data from HOMFLY-PT homology

A basic structural statement of the correspondence is that nodes of $Q(K)$7 correspond to generators of the reduced uncolored HOMFLY-PT homology. If the homology generators have degrees $Q(K)$8, then the quiver data satisfy

$Q(K)$9

and $3\text{d }\mathcal N=2$0 is the $3\text{d }\mathcal N=2$1-degree of the corresponding generator. The sign $3\text{d }\mathcal N=2$2 appears precisely for generators of odd homological degree. The number of nodes $3\text{d }\mathcal N=2$3 equals the number of generators of the uncolored reduced HOMFLY-PT homology (Kucharski et al., 2017).

The original paper further proposes a block description of many knot quivers in terms of two primitive homological patterns. A knot has precisely one zig-zag block of odd length $3\text{d }\mathcal N=2$4, matching the torus knot $3\text{d }\mathcal N=2$5 block up to permutation and framing shifts, and each diamond contributes a $3\text{d }\mathcal N=2$6 block of prescribed form. Off-diagonal blocks encode interactions between the zig-zag and diamond pieces. This block language is particularly effective for torus knots, twist knots, and thick homology examples (Kucharski et al., 2017).

Framing acts uniformly on the quiver. A change of framing by $3\text{d }\mathcal N=2$7 shifts the whole matrix by the all-ones matrix,

$3\text{d }\mathcal N=2$8

which corresponds to adding $3\text{d }\mathcal N=2$9 loops at every vertex and $K$0 pairs of oppositely oriented arrows between every pair of vertices. This operation leaves the specialized knot generating series unchanged after the standard framing factor is taken into account (Sachdeva, 4 May 2025).

The practical reconstruction pipeline begins with the reduced uncolored HOMFLY-PT homology $K$1, assigns one quiver node to each monomial, sets $K$2 and $K$3, and then rewrites known colored formulas into the quiver form by repeated use of $K$4-series identities, in particular expansions of Pochhammer symbols and multi-binomial decompositions. The resulting quadratic exponent determines the full adjacency matrix $K$5, after which framing shifts and low-color consistency checks fix the remaining ambiguities (Kucharski et al., 2017).

4. Physical and geometric realization

The physical origin of the correspondence is the D-brane realization of knot invariants in topological string theory. In the A-model on $K$6, $K$7 branes wrap $K$8 and a Lagrangian brane intersects them along the knot $K$9. The Ooguri–Vafa operator

$Q$0

encodes open strings stretched between the two stacks. After the large-$Q$1 transition, the geometry becomes the resolved conifold with a Lagrangian brane on the knot conormal. In M-theory, LMOV invariants are interpreted as degeneracies of open M2-branes ending on M5-branes, and dimensional reduction produces an $Q$2 quiver quantum mechanics whose BPS counting matches the motivic DT invariants of a symmetric quiver (Kucharski et al., 2017).

The geometric dictionary developed in later work identifies quiver nodes with basic embedded holomorphic disks on the knot conormal $Q$3, and arrows with linking data of their boundaries. More precisely, the intersection of bounding chains and disk boundaries determines $Q$4, diagonal entries encode self-linking, and the $Q$5- and $Q$6-weights record the homology class and four-chain intersection of the corresponding disk. This gives a geometric explanation for why the quiver matrix simultaneously controls knot homology gradings and BPS interactions (Ekholm et al., 2018).

The same quiver also defines a $Q$7 theory $Q$8. Its gauge group is

$Q$9

with one chiral field per node, mixed Chern–Simons couplings $K$0, and FI terms determined by the quiver variables. In the semiclassical limit, the quiver partition function yields a twisted superpotential built from dilogarithms and logarithmic couplings, while its vortex partition functions and holomorphic blocks reproduce the quiver DT series and hence the knot generating functions after KQ specialization (Sachdeva, 4 May 2025).

This physical picture has been extended to knot complements. For fibered knots, the skein-valued partition function of the Lagrangian complement $K$1 localizes on holomorphic annuli associated to flow loops, and for $K$2-torus knots the resulting partition function admits a generalized quiver structure that includes annulus and anti-annulus nodes. In this setting the complement branch gives another coordinate chart of the quantized augmentation polynomial, compatible with the conormal branch after a simple $K$3 shift (Chauhan et al., 30 Jan 2026).

5. Explicit families and computations

The unknot is the basic test case. In reduced normalization, the generating function

$K$4

is the series of the one-vertex, $K$5-loop quiver. In unreduced normalization,

$K$6

which corresponds to a two-node quiver with one loop, and the LMOV invariants are concentrated in two quantum dilogarithms (Kucharski et al., 2017).

For the trefoil $K$7, the reduced uncolored superpolynomial is

$K$8

so there are three quiver nodes with $K$9-degrees $Q$0. The corresponding quiver matrix is

$Q$1

with $Q$2 and $Q$3. This is the standard explicit example showing that diagonal entries reproduce homological degrees and that the colored generating series takes the quiver form exactly (Kucharski et al., 2017).

For the figure-eight $Q$4, one explicit quiver is

$Q$5

which decomposes into a zig-zag plus one diamond block up to permutation and framing shift. The paper identifies quivers for all knots up to six crossings, including $Q$6, $Q$7, and $Q$8, and derives previously unknown closed forms for colored HOMFLY-PT polynomials of $Q$9 and $$P(x)=\sum_{r=0}^\infty \overline{P}_r(a,q)\,x^r =\exp\!\Big(\sum_{r,n\ge1}\frac{1}{n}\,f_r(a^n,q^n)\,x^{nr}\Big),$$0 from the quiver data (Kucharski et al., 2017).

Infinite families are central. Torus knots $$P(x)=\sum_{r=0}^\infty \overline{P}_r(a,q)\,x^r =\exp\!\Big(\sum_{r,n\ge1}\frac{1}{n}\,f_r(a^n,q^n)\,x^{nr}\Big),$$1 admit recursive and closed-form quiver matrices built from blocks $$P(x)=\sum_{r=0}^\infty \overline{P}_r(a,q)\,x^r =\exp\!\Big(\sum_{r,n\ge1}\frac{1}{n}\,f_r(a^n,q^n)\,x^{nr}\Big),$$2, $$P(x)=\sum_{r=0}^\infty \overline{P}_r(a,q)\,x^r =\exp\!\Big(\sum_{r,n\ge1}\frac{1}{n}\,f_r(a^n,q^n)\,x^{nr}\Big),$$3, and $$P(x)=\sum_{r=0}^\infty \overline{P}_r(a,q)\,x^r =\exp\!\Big(\sum_{r,n\ge1}\frac{1}{n}\,f_r(a^n,q^n)\,x^{nr}\Big),$$4. Twist knots of both signs admit systematic block constructions, and later work extends this to double twist knots $$P(x)=\sum_{r=0}^\infty \overline{P}_r(a,q)\,x^r =\exp\!\Big(\sum_{r,n\ge1}\frac{1}{n}\,f_r(a^n,q^n)\,x^{nr}\Big),$$5, for which the quiver has

$$P(x)=\sum_{r=0}^\infty \overline{P}_r(a,q)\,x^r =\exp\!\Big(\sum_{r,n\ge1}\frac{1}{n}\,f_r(a^n,q^n)\,x^{nr}\Big),$$6

nodes and a universal block form determined by $$P(x)=\sum_{r=0}^\infty \overline{P}_r(a,q)\,x^r =\exp\!\Big(\sum_{r,n\ge1}\frac{1}{n}\,f_r(a^n,q^n)\,x^{nr}\Big),$$7 seed blocks $$P(x)=\sum_{r=0}^\infty \overline{P}_r(a,q)\,x^r =\exp\!\Big(\sum_{r,n\ge1}\frac{1}{n}\,f_r(a^n,q^n)\,x^{nr}\Big),$$8 together with the recursion $$P(x)=\sum_{r=0}^\infty \overline{P}_r(a,q)\,x^r =\exp\!\Big(\sum_{r,n\ge1}\frac{1}{n}\,f_r(a^n,q^n)\,x^{nr}\Big),$$9 (Singh et al., 2023).

The correspondence also covers thick homology examples. The original paper gives bottom-row quivers for $$f_r(a,q)=\sum_{i,j}\frac{N_{r,i,j}\,a^i\,q^j}{q-q^{-1}}.$$0, $$f_r(a,q)=\sum_{i,j}\frac{N_{r,i,j}\,a^i\,q^j}{q-q^{-1}}.$$1, and a new $$f_r(a,q)=\sum_{i,j}\frac{N_{r,i,j}\,a^i\,q^j}{q-q^{-1}}.$$2, while later generalized formulations allow higher-level generators $$f_r(a,q)=\sum_{i,j}\frac{N_{r,i,j}\,a^i\,q^j}{q-q^{-1}}.$$3 and thereby accommodate thick knots with super-exponential growth, such as $$f_r(a,q)=\sum_{i,j}\frac{N_{r,i,j}\,a^i\,q^j}{q-q^{-1}}.$$4, $$f_r(a,q)=\sum_{i,j}\frac{N_{r,i,j}\,a^i\,q^j}{q-q^{-1}}.$$5, $$f_r(a,q)=\sum_{i,j}\frac{N_{r,i,j}\,a^i\,q^j}{q-q^{-1}}.$$6, and several 10-crossing examples (Stošić, 2024).

6. Equivalences, generalizations, and open directions

A recurring misconception is that the map $$f_r(a,q)=\sum_{i,j}\frac{N_{r,i,j}\,a^i\,q^j}{q-q^{-1}}.$$7 should be canonical. It is not. Equivalent quivers can encode the same knot data, and this non-uniqueness is systematic rather than exceptional. The set of equivalent quivers is organized by graphs made of permutohedra glued along common vertices, and the local moves relating them are specific transpositions satisfying explicit “center-of-mass” and arrow-sum constraints. These equivalences can also be interpreted as webs of dual $$f_r(a,q)=\sum_{i,j}\frac{N_{r,i,j}\,a^i\,q^j}{q-q^{-1}}.$$8 theories (Jankowski et al., 2021).

A second important development concerns quiver operations. Multi-cover skein relations imply linking and unlinking moves on symmetric quivers that preserve the generating series after introducing an auxiliary node and a prescribed specialization of its variable. These operations are not cluster mutations. They admit a quantum torus formulation parallel to Kontsevich–Soibelman wall-crossing identities, and they have now been categorified: the moduli spaces of a quiver and its linking or unlinking are related by homotopy equivalences, cohomological decompositions, and compatible CoHA-module structures (Ekholm et al., 2019, Garderen, 2024).

Generalized KQ extends the variable map to arbitrary powers of the knot variable,

$$f_r(a,q)=\sum_{i,j}\frac{N_{r,i,j}\,a^i\,q^j}{q-q^{-1}}.$$9

so that quiver nodes may have higher level $Q(K)$00. This extension captures knots lacking the exponential growth property, including $Q(K)$01, and leads to compact quiver-like formulas in which the structure of colored differentials becomes manifest. The combinatorics of these higher-level generators is related to lattice paths, Raney numbers, and explicit closed formulas for classical coefficients of quiver series (Stošić, 2024, Đorđević et al., 2024).

The framework has also expanded from knot conormals to knot complements and branch-dependent $Q(K)$02 invariants. Branches of the $Q(K)$03-polynomial admit quiver realizations, together with $Q(K)$04- and $Q(K)$05-deformed $Q(K)$06-ideals encoding recursion in the rank parameter. For closed 3-manifolds obtained by surgery, the same quiver formulation leads to the twisted superpotential of the $Q(K)$07 theory $Q(K)$08 and to data of the associated modular tensor category $Q(K)$09 (Ekholm et al., 2021).

Recent work relates stability of colored HOMFLY-PT polynomials under adding full twists to stability of augmented quivers under unlinking or linking of the extra node. The conjectural dictionary states that a full twist in a knot diagram corresponds to a prescribed sequence of unlinkings or linkings in the augmented quiver. This has been confirmed for all twist knots, $Q(K)$10 torus knots, and all pretzel knots up to 15 crossings with an odd number of twists in each twist region (Chauhan et al., 25 Aug 2025).

Several open problems remain standard. The correspondence beyond symmetric representations is not yet incorporated in full generality. Canonical quiver assignment, uniqueness up to permutations or mutations, full quadruply graded data for wider classes of thick knots, systematic quiver constructions for arbitrary links and satellites, and a complete refined correspondence for generalized and complement-type quivers all remain active directions (Sachdeva, 4 May 2025).