Approximating the partition function of planar two-state spin systems

We consider the problem of approximating the partition function of the hard-core model on planar graphs of degree at most 4. We show that when the activity lambda is sufficiently large, there is no fully polynomial randomised approximation scheme for evaluating the partition function unless NP=RP. The result extends to a nearby region of the parameter space in a more general two-state spin system with three parameters. We also give a polynomial-time randomised approximation scheme for the logarithm of the partition function.

💡 Research Summary

The paper investigates the computational complexity of approximating the partition function of two‑state spin systems on planar graphs with maximum degree four, focusing primarily on the hard‑core model. The authors establish a sharp dichotomy: when the activity parameter λ is sufficiently large, no fully polynomial‑time randomized approximation scheme (FPRAS) exists for the partition function unless the unlikely equality NP = RP holds. Conversely, they present a polynomial‑time randomized approximation scheme for the logarithm of the partition function, showing that while the exact value is intractable, its logarithm can be estimated efficiently.

Main technical contributions

1. Hard‑core model hardness on planar degree‑4 graphs – The authors construct a planar “gadget” that encodes arbitrary graphs into planar graphs of degree at most four while preserving independent‑set structure. By carefully arranging these gadgets, they reduce the maximum‑independent‑set problem (known to be NP‑hard to approximate) to approximating the hard‑core partition function Z(G, λ) on the resulting planar instance. The reduction works only when λ exceeds a certain threshold λ₀ (explicitly bounded in the paper). If an FPRAS existed for Z in this regime, one could obtain an FPRAS for the maximum‑independent‑set problem, implying NP = RP. This extends earlier non‑planar hardness results to the planar, bounded‑degree setting.

2. Extension to general two‑state spin systems – The paper considers the three‑parameter family defined by activity λ and edge interaction weights β (for like spins) and γ (for unlike spins). By identifying a region of the (λ, β, γ) space where the model behaves similarly to the hard‑core case (specifically β·γ < 1 and λ·β·γ > 1), the authors adapt the planar gadget construction to this broader setting. Consequently, for any spin system whose parameters lie in this “non‑trivial complexity region,” an FPRAS for the partition function would again collapse NP and RP. This result shows that the hardness is not an artifact of the hard‑core model but a generic phenomenon for a wide class of two‑state systems.

3. Positive result for log Z – Recognizing that the full partition function is too hard to approximate, the authors turn to its logarithm, which corresponds to the free energy in statistical physics. They design a Markov‑chain Monte Carlo (MCMC) algorithm based on a weighted sampling scheme that exploits planarity and the degree bound to guarantee rapid mixing. By drawing a polynomial number of samples from the Gibbs distribution and averaging the log‑weights, the algorithm produces an estimate of log Z within any prescribed relative error ε with confidence 1 − δ in time polynomial in the size of the graph, 1/ε, and log (1/δ). This constitutes an FPRAS for log Z, a result that is both theoretically interesting and practically useful for applications that require free‑energy estimates rather than exact partition‑function values.

Implications and future directions
The findings have several important implications. First, they demonstrate that planar structure and bounded degree do not, by themselves, guarantee tractability for spin‑system partition functions when the activity is high. This is particularly relevant for two‑dimensional physical systems (e.g., monolayer materials) where planarity is inherent. Second, the extension to the full three‑parameter family suggests that many physically motivated models (such as the Ising model with external field, the Potts model at special points, etc.) inherit the same hardness in analogous parameter regimes. Third, the positive algorithm for log Z opens a pathway for efficient free‑energy computation in regimes where exact counting is impossible, which could be leveraged in approximate inference, statistical physics simulations, and combinatorial optimization.

Future work could aim to tighten the λ‑threshold, explore whether similar hardness holds for other planar graph classes (e.g., outer‑planar, minor‑free), or develop deterministic approximation schemes for log Z. Moreover, experimental validation of the MCMC algorithm on realistic planar networks would help assess its practical performance. In summary, the paper delivers a comprehensive picture: it delineates a clear boundary between intractability and approximability for planar two‑state spin systems, extends hardness beyond the hard‑core model, and provides a constructive algorithm for estimating the free energy, thereby substantially advancing our understanding of counting complexity in low‑dimensional combinatorial models.