Yang-Baxter integrability of the 3D quantum crystal melting Hamiltonian: Kagome-lattice Boltzmann weights, a commuting transfer matrix, and the mass-gap conjecture

The quantum crystal melting Hamiltonian introduced by Dijkgraaf, Orlando and Reffert acts on states labelled by plane partitions and has the MacMahon-weighted crystal-melting state as its ground state; its 1D and 2D analogues are exactly solvable (the 2D problem is the XXZ spin chain in disguise) and share the same mass gap, which led to the conjecture that the genuine 3D problem is likewise integrable with the same gap. Araujo recast the 3D problem as an occupation model on the Kagome lattice and initiated a Yang-Baxter program: plane-partition states are encoded in 18 local hexagon configurations with undetermined Boltzmann weights; the associated Lax operators organize into two transfer-matrix families that commute within but not across families; and modified connectivity rules yield two integrable 6-vertex-type descendant subsystems with RTT (fundamental commutation relation) structure. The source leaves open the explicit weights reproducing the quantum spectrum, a global commuting transfer matrix for the full Kagome lattice, the reassembly of the descendant subsystems into the full plane-partition problem, a definitive proof or disproof of 3D integrability, and the mass-gap conjecture together with the first-excited-state coefficients. The problem is open.

Background

Plane partitions — arrays of non-negative integers weakly decreasing along rows and columns, equivalently finite stacks of unit boxes piled against the corner of a three-dimensional room (3D Young diagrams) — are the natural state labels of the crystal melting problem: the classical statistical model whose partition function, the MacMahon function $\prod_{k\ge1}(1-q^k)^{-k}$ generating plane partitions by box number, counts molten configurations of a three-dimensional crystal and arises in Donaldson-Thomas type counting on toric Calabi-Yau threefolds. Dijkgraaf, Orlando and Reffert (DOR) promoted this counting problem to a quantum dynamics by introducing a Hamiltonian whose kinetic terms create and annihilate boxes at admissible corners of the growing plane partition and whose diagonal terms are asymmetric diffusion weights controlled by real couplings $J$, $V$ and $q$ (the Hamiltonian is written explicitly in the problem statement). The ground state in the plane-partition basis is exactly the MacMahon-weighted melting state, so the model ties quantum many-body dynamics to the enumerative geometry of the classical problem.

DOR solved the lower-dimensional analogues: the 1D problem (growth confined to a half-line) and the 2D problem (growth in a quadrant, i.e. ordinary partitions) are exactly solvable — the 2D Hamiltonian maps, via a Jordan-Wigner transformation (a fermionization rewriting spins in terms of nonlocal fermionic variables), to the XXZ spin chain (the anisotropic spin-$1/2$ Heisenberg chain) with domain-wall (kink) boundary conditions, diagonalizable by Bethe ansatz. Strikingly, the 1D and 2D problems have the same mass gap, the ground-state-to-first-excited-state energy difference in the infinite-volume limit. These facts led DOR to conjecture that the genuine 3D problem — the same Hamiltonian acting on plane partitions — is likewise integrable and shares the same mass gap. Both conjectures have remained open since 2009.

The source attacks the 3D problem through a two-dimensional lattice reformulation. Araujo, Orlando and Reffert first established a fermion-boson duality for plane partitions under which the dimer configurations (matchings pairing nearest-neighbour sites) on a hexagonal lattice equivalent to plane partitions map to an occupation problem on the Kagome lattice (the trihexagonal lattice of corner-sharing triangles). The Hilbert space is then built from two families of coupled spin chains, called X-rows and Y-rows, constructed from two free fermions, and the plane-partition selection rules become local: exactly 18 local hexagon configurations are allowed. Assigning to each of them a Boltzmann weight $\mathcal{W}_i$ — at this stage an unknown function, conjecturally of a spectral parameter $u$ (the auxiliary complex variable on which vertex-model weights customarily depend) through $J=J(u)$, $V=V(u)$, $q=q(u)$, by analogy with the 1D/2D cases where the mass gap fixes $J$ in terms of $q$ — defines a classical two-dimensional statistical model of vertex type whose weights should encode the quantum spectrum: the original quantum problem is, in the source's words, completely established once the correct weights, appropriately combined, give the Hamiltonian eigenenergies.

Each hexagon defines a Lax operator — an S-matrix-like linear map acting on one two-dimensional auxiliary space and two two-dimensional quantum spaces, with matrix elements built from the local weights — and ordered products along rows define two row-to-row transfer matrices $\mathbf t(u)$ (even rows) and $\tilde{\mathbf t}(u)$ (odd rows). Integrability of the classical model would follow from the Yang-Baxter equation, the consistency condition guaranteeing commuting transfer matrices and hence an extensive family of conserved quantities. The source proves that the two families commute within themselves, $[\mathbf t(u),\mathbf t(v)]=[\tilde{\mathbf t}(u),\tilde{\mathbf t}(v)]=0$, but not across, $[\mathbf t(u),\tilde{\mathbf t}(v)]\neq0$, so the naive staggered monodromy fails and the Yang-Baxter constraints on the weights cannot even be written globally. Modifying the rules that connect hexagons (linking rows vertically rather than diagonally) decouples the lattice into two integrable descendant subsystems of 6-vertex-model type (the six-vertex model being the square-lattice ice-type model, the archetype of an exactly solved vertex model): their Lax operators $\mathcal{L}^{(1/2,1)}$ act on an auxiliary space in the spin-1/2 fundamental representation of $su(2)$ and on a three-dimensional quantum space (the adjoint of $sl(2,\mathbb C)$, i.e. spin 1), and satisfy the RTT relation (also called the fundamental commutation relation of the quantum inverse scattering method). These subsystems are integrable but live on a restricted physical space; how their solutions relate to the original plane-partition problem is itself one of the open questions. The known algebraic structure of the ground state — its relation to the affine Yangian of $gl(1)$ (the infinite-dimensional symmetry algebra whose representation on the plane-partition Fock space is the MacMahon representation) — signals an underlying integrable structure, but, as the companion paper stresses, a pragmatic relation between that algebra and the melting dynamics remains to be found.

Problem Statement

Settle the exact solvability of the 3D quantum crystal melting Hamiltonian

$$H\;=\;-J\sum\left(|\blacksquare\rangle\langle\square|+|\square\rangle\langle\blacksquare|\right)\;+\;V\sum\left(\sqrt q\,|\square\rangle\langle\square|+\tfrac{1}{\sqrt q}\,|\blacksquare\rangle\langle\blacksquare|\right),$$

acting on the Hilbert space spanned by plane partitions, where $|\square\rangle$ ranges over the local configurations in which a box can consistently be added to the growing crystal and $|\blacksquare\rangle$ over the boxes that can be removed, through the Kagome-lattice vertex-model correspondence, by determining:

1. the explicit local Boltzmann weights $\mathcal{W}_i(u)$ of the 18 allowed hexagon configurations of the Kagome reformulation, as functions of the spectral parameter $u$ (with $J=J(u)$, $V=V(u)$, $q=q(u)$), whose transfer-matrix combinations reproduce the eigenenergies of $H$;

2. a global commuting transfer-matrix (or alternative monodromy) construction for the full Kagome lattice that covers both parity sectors and overcomes the obstruction $[\mathbf t(u),\tilde{\mathbf t}(v)]\neq0$ of the staggered row construction — or a proof that no such construction exists within this framework;

3. whether the integrable even/odd descendant subsystems defined by the $\mathcal{L}^{(1/2,1)}$ RTT structures can be consistently assembled, respecting the original Kagome connectivity and the plane-partition selection rules, to reconstruct eigenstates and eigenvalues of the full plane-partition problem;

4. on this basis, or by any equivalent rigorous route, a definitive proof or disproof of the Yang-Baxter integrability of the full 3D quantum crystal melting Hamiltonian (the first Dijkgraaf-Orlando-Reffert conjecture); and

5. the spectral consequence explicitly targeted by the source: whether the 3D mass gap equals the common 1D/2D value (the second Dijkgraaf-Orlando-Reffert conjecture), for instance through the coefficients $f_\Lambda(q)$ of the first excited state $|\Psi_1\rangle=\sum_\Lambda f_\Lambda(q)|\Lambda\rangle$ in the plane-partition basis.

Items 1-3 are the proposed route to item 4, and items 4-5 together settle the two Dijkgraaf-Orlando-Reffert conjectures for the 3D problem; a resolution of items 4-5 by an alternative rigorous method supersedes items 1-3 but must address the full model with its plane-partition selection rules, not merely the known integrable descendant subsystems. The source's sixth direction — precise connections to affine Yangians, $\mathcal N=2$ supersymmetric theories, topological strings, Alday-Gaiotto-Tachikawa (AGT)-type correspondences and AdS$_3$/higher-spin duality — is announced research context rather than part of the completion boundary.

References

1. T. Araujo, 'Towards Yang-Baxter integrability of quantum crystal melting: From Kagome lattice to vertex models', Nucl. Phys. B 963 (2021) 115287. DOI: 10.1016/j.nuclphysb.2020.115287; arXiv:2009.06812. https://arxiv.org/abs/2009.06812
2. R. Dijkgraaf, D. Orlando, S. Reffert, 'Quantum Crystals and Spin Chains', Nucl. Phys. B 811 (2009) 463-490. DOI: 10.1016/j.nuclphysb.2008.11.027; arXiv:0803.1927. https://arxiv.org/abs/0803.1927
3. T. Araujo, D. Orlando, S. Reffert, 'Quantum crystals, Kagome lattice and plane partitions fermion-boson duality', Phys. Rev. D 103 (2021) 026020. DOI: 10.1103/PhysRevD.103.026020; arXiv:2005.09103. https://arxiv.org/abs/2005.09103
4. D. Orlando, S. Reffert, N. Reshetikhin, 'On domain wall boundary conditions for the XXZ spin Hamiltonian', arXiv:0912.0348 (2009). https://arxiv.org/abs/0912.0348
5. N. Wang, '3D fermions and affine Yangian', Nucl. Phys. B 969 (2021) 115461. DOI: 10.1016/j.nuclphysb.2021.115461. https://doi.org/10.1016/j.nuclphysb.2021.115461
6. D. Galakhov, W. Li, 'Charging solid partitions', JHEP 01 (2024) 043. DOI: 10.1007/JHEP01(2024)043; arXiv:2311.02751. https://arxiv.org/abs/2311.02751
7. F. Balducci, A. Gambassi, A. Lerose, A. Scardicchio, C. Vanoni, 'Localization and Melting of Interfaces in the Two-Dimensional Quantum Ising Model', Phys. Rev. Lett. 129 (2022) 120601. DOI: 10.1103/PhysRevLett.129.120601; arXiv:2203.09495. https://arxiv.org/abs/2203.09495
8. F. Balducci, A. Gambassi, A. Lerose, A. Scardicchio, C. Vanoni, 'Interface dynamics in the two-dimensional quantum Ising model', Phys. Rev. B 107 (2023) 024306. DOI: 10.1103/PhysRevB.107.024306; arXiv:2209.08992. https://arxiv.org/abs/2209.08992
9. A. Prakash, J. S. Rao, S. A. Parameswaran, A. Lerose, 'Interface phases and dynamics in two-dimensional quantum magnets: A "holographic" approach from universality to quantum simulation', arXiv:2608.12312 (2026). https://arxiv.org/abs/2608.12312
10. M. Ogura, Y. Imamura, N. Kameyama, K. Minami, M. Sato, 'Geometric Criterion for Solvability of Lattice Spin Systems', Phys. Rev. B 102 (2020) 245118. DOI: 10.1103/PhysRevB.102.245118; arXiv:2003.13264. https://arxiv.org/abs/2003.13264
11. T. Araujo, 'Q-boson model and relations with integrable hierarchies', Nucl. Phys. B 1006 (2024) 116640. DOI: 10.1016/j.nuclphysb.2024.116640; arXiv:2405.01213. https://arxiv.org/abs/2405.01213
12. T. Araujo, 'Comments on Slavnov products, Temperley-Lieb open spin chains, and KP tau functions', Nucl. Phys. B 972 (2021) 115566. DOI: 10.1016/j.nuclphysb.2021.115566; arXiv:2107.13060. https://arxiv.org/abs/2107.13060
13. C. Yang, 'A remark on certain restricted plane partitions and crystal melting model', Math. Res. Lett. 33 (2026) 115-160. DOI: 10.4310/mrl.260502144621; arXiv:2312.02749. https://arxiv.org/abs/2312.02749