Abstract
We consider a 2D quantum spin model with ring-exchange interaction that has subsystem symmetries associated to conserved magnetization along rows and columns of a square lattice, which implies the conservation of the global dipole moment. The model is not integrable, but violates the eigenstate thermalization hypothesis through an extensive Hilbert space fragmentation, including an exponential number of inert subsectors with trivial dynamics, arising from kinetic constraints. While subsystem symmetries are quite restrictive for the dynamics, we show that they alone cannot account for such a number of inert states, even with infinite-range interactions. We present a procedure for constructing shielding structures that can separate and disentangle dynamically active regions from each other. Notably, subsystem symmetries allow the thickness of the shields to be dependent only on the interaction range rather than on the size of the active regions, unlike in the case of generic dipole-conserving systems.
Introduction
In quantum statistical mechanics, isolated systems are generally expected to thermalize, reaching an equilibrium state described by the Eigenstate Thermalization Hypothesis (ETH). However, certain systems exhibit "stubborn" non-thermalizing behavior. Beyond traditional integrability and Many-Body Localization (MBL), a phenomenon known as Hilbert space fragmentation has emerged, where the Hilbert space "shatters" into many disconnected sectors. While previously studied primarily in 1D, this report examines the XY-plaquette model in 2D to understand how subsystem symmetries drive this fragmentation.
Subsystem Symmetries and Sector Counting
The model is defined on a square lattice with spin-1/2 particles interacting via "ring-exchange" on elementary plaquettes. A defining feature is the presence of $U(1)$ subsystem symmetries: the total magnetization along every individual row and column is conserved. This leads to a massive number of symmetry sectors.
As shown in Figure 2, the number of symmetry sectors grows exponentially with the system size. This conservation of row and column magnetization also implies the conservation of the global dipole moment, a key constraint that restricts the movement of excitations.
Fragmentation and the Scaling of Inert States
The most striking consequence of these constraints is the "shattering" of the Hilbert space. The system does not just split into symmetry sectors; it further fragments into an overwhelming number of tiny, isolated subsectors. Among these are "inert states"—configurations that are completely frozen under the Hamiltonian's dynamics.
Numerical analysis reveals that the number of inert states grows exponentially with the number of sites, scaling as approximately $1.80^{N^2}$. This growth is significantly faster than what would be predicted by subsystem symmetries alone, indicating that the locality of the interaction (the ring-exchange) plays a crucial role in trapping the system.
Key Insights: Shielding and Locality
A major theoretical breakthrough presented is the construction of "shielding structures." These are patterns of inert spins that act as perfect insulators, isolating active quantum regions. In this 2D model, subsystem symmetries allow these shields to have a fixed thickness—determined only by the interaction range—regardless of the size of the region they protect. This contrasts with generic dipole-conserving systems where shields typically scale with the size of the protected area.
The study also demonstrates that if interactions were infinite-range, the fragmentation effect would weaken. Thus, the combination of subsystem symmetry and local interaction is the essential "recipe" for this robust shattering.
Conclusion
The 2D XY-plaquette model provides a rich playground for studying non-thermalizing quantum dynamics. By leveraging subsystem symmetries, the authors demonstrate a form of Hilbert space fragmentation that is both extensive and robust. The ability to construct thin, effective shields suggests potential applications in quantum information, where such structures could protect qubits from decoherence. Future work may explore the transition between chaotic and integrable-like statistics under varying levels of disorder and the potential for experimental realization in platforms like Rydberg atom arrays.
Summary updated