The Lattice That Remembers: A New Way to Simulate Quantum Spins
26 Apr 2026, Yanjiang
Naively, one might think that simulating a quantum system is simply a matter of letting a computer do the math — plug in the interactions, run the algorithm, and read off the answer. But for the most interesting quantum materials, those where spins arrange themselves on frustrated lattices like the kagome pattern of corner-sharing triangles, this naive picture collapses. The computational cost grows exponentially with system size, like trying to predict the weather by tracking every air molecule in the atmosphere. You don’t run out of patience; you run out of universe.
A preprint (arXiv:1307.5696) from researchers at the Institute of Physics, Chinese Academy of Sciences proposes a new way to cut through this complexity. Led by T. Xiang, the team — including Z. Y. Xie, J. Chen, J. F. Yu, X. Kong, and B. Normand — has developed a class of tensor-network states called projected entangled simplex states (PESS) that capture the quantum correlations of lattice models with unprecedented efficiency. Their method, applied to the spin-1/2 antiferromagnetic Heisenberg model on the kagome lattice, produces ground-state energies that approach the lowest upper bounds yet estimated for this notoriously difficult system.
But to understand what they’ve built, we first need to understand the problem they’re solving.
The Curse of Entanglement
Think of a quantum lattice as a crowded dance floor. Each spin — a tiny magnetic moment — is a dancer, and the rules of quantum mechanics say that every dancer is simultaneously executing every possible move. The challenge of simulation is not tracking one dancer, but tracking how each dancer’s motion is entangled with every other dancer’s. This entanglement grows with system size, and for a generic quantum state, the amount of information required to describe it scales exponentially.
In particular, for two-dimensional lattices like the kagome — a pattern of hexagons and triangles that looks like a Japanese basket weave — the entanglement structure is rich enough to defeat most computational approaches. The ground state of such a frustrated magnet is not a simple ordered pattern but a “quantum spin liquid,” where spins remain fluid and entangled down to absolute zero. Simulating this state has been a holy grail of condensed matter physics for decades.
The standard approach, projected entangled pair states (PEPS), works by representing the quantum state as a network of tensors — mathematical objects that connect each lattice site to its neighbors through a web of entangled “bonds.” Imagine a fishing net where each knot is a tensor, and each strand is an entanglement bond. PEPS captures the two-dimensional entanglement structure naturally, but it has a limitation: each bond connects only pairs of sites. For systems where entanglement spreads across larger clusters — triangles, hexagons, entire plaquettes — PEPS struggles.
This is where PESS enters.
From Pairs to Simplexes
The key insight of Xie and colleagues is elegantly simple: if the physics of your system involves clusters of three or more sites, why not build your tensor network around those clusters directly? Instead of connecting sites pairwise, PESS uses simplexes — the generalization of a triangle to higher dimensions — as the fundamental building block.
Picture a dinner party where guests interact in pairs. PEPS is like assigning each pair a private conversation. But on the kagome lattice, spins on the same triangle interact collectively — it’s a three-way conversation, not three separate pair conversations. PESS captures this by placing a “simplex tensor” at the center of each triangle, which encodes the entanglement among all three spins simultaneously. The individual site tensors then connect these simplexes together, forming a hierarchical network that respects the lattice geometry.
Unlike dinner guests, quantum spins can occupy all choices simultaneously — a simplex tensor doesn’t pick one configuration but superposes all of them, weighted by the entanglement structure of the physical system. This is not a metaphor for social dynamics but a precise mathematical representation: the PESS wave function is an exact representation of what the authors call “simplex solid states,” and it satisfies the area law of entanglement entropy — the principle that entanglement scales with the boundary of a region, not its volume, which is the hallmark of gapped quantum phases.
The Update Method
Of course, proposing a new tensor-network state is only half the battle. You also need a way to find the right tensor entries — the numerical values that minimize the ground-state energy. The team introduces a “simple update” method based on imaginary-time evolution, a technique that simulates cooling a quantum system by evolving it in imaginary time until it settles into its ground state.
Here’s where the computational craftsmanship shines. The update method uses higher-order singular-value decomposition (HOSVD) — a generalization of the familiar matrix decomposition to higher-dimensional tensors. If ordinary singular-value decomposition is like cutting a photograph into strips to find the most important features, HOSVD is like slicing a cube into multiple orientations simultaneously, extracting the dominant correlations in every direction at once.
The team applied this machinery to the spin-1/2 antiferromagnetic Heisenberg model on the kagome lattice — a system that has resisted precise numerical treatment for decades. Their results for the ground-state energy approach the lowest upper bounds yet estimated, and the systematic convergence of their method suggests that PESS can achieve accuracy comparable to, or exceeding, the best available techniques.
What This Means
One might ask: why does a better ground-state energy for a model magnet matter? The answer is that the kagome Heisenberg model is a paradigmatic example of a quantum spin liquid — a state of matter where magnetic moments remain disordered and entangled even at absolute zero, with excitations that behave like fractional particles. Understanding this state is not merely an academic exercise; quantum spin liquids are candidates for topological quantum computation, where information is stored in the braiding of these fractional excitations, protected from decoherence by the topology of the system itself.
PESS provides a new tool for exploring this landscape. By building tensor networks that respect the cluster structure of frustrated lattices, the method opens a path to studying larger systems with higher accuracy. The same approach could be extended to other frustrated geometries — the triangular lattice, the honeycomb lattice, the pyrochlore — and to models with longer-range interactions or different spin symmetries.
The Architecture of Understanding
What makes this work compelling is not just the numerical results but the conceptual architecture. The team has built a bridge between two ways of thinking about quantum matter: the tensor-network approach, which encodes entanglement through geometry, and the cluster-expansion approach, which captures correlations through local building blocks. PESS is a synthesis that respects the physics of the system — it doesn’t force the lattice into a pair-bonded straitjacket but lets the entanglement structure of the simplex emerge naturally.
For a field that has long treated the kagome Heisenberg model as a benchmark for numerical methods, this work provides a new reference point. The ground-state energies reported here are not just numbers; they are constraints on what any theory of quantum spin liquids must explain. And the method itself — the combination of simplex tensors, imaginary-time evolution, and higher-order decomposition — is a template that can be adapted to other systems.
Perhaps, in the coming years, when physicists finally map the complete phase diagram of the kagome Heisenberg model — identifying the precise nature of its spin-liquid ground state, the spectrum of its fractional excitations, the response to external fields — they will look back at this work as one of the steps that made it possible. Not a revolution, but a refinement: a better tool for a hard problem, built on the insight that the right building blocks matter.
Yanjiang is an online editor of Loom Science
References
- Z. Y. Xie et al., Tensor renormalization of quantum many-body systems using projected entangled simplex states, arXiv:1307.5696