Statistical mechanics of classical and quantum computational complexity

The quest for quantum computers is motivated by their potential for solving problems that defy existing, classical, computers. The theory of computational complexity, one of the crown jewels of computer science, provides a rigorous framework for classifying the hardness of problems according to the computational resources, most notably time, needed to solve them. Its extension to quantum computers allows the relative power of quantum computers to be analyzed. This framework identifies families of problems which are likely hard for classical computers (NP-complete'') and those which are likely hard for quantum computers (QMA-complete’’) by indirect methods. That is, they identify problems of comparable worst-case difficulty without directly determining the individual hardness of any given instance. Statistical mechanical methods can be used to complement this classification by directly extracting information about particular families of instances—typically those that involve optimization—by studying random ensembles of them. These pose unusual and interesting (quantum) statistical mechanical questions and the results shed light on the difficulty of problems for large classes of algorithms as well as providing a window on the contrast between typical and worst case complexity. In these lecture notes we present an introduction to this set of ideas with older work on classical satisfiability and recent work on quantum satisfiability as primary examples. We also touch on the connection of computational hardness with the physical notion of glassiness.

💡 Research Summary

The lecture notes present a unified perspective on computational complexity and statistical mechanics, focusing on both classical and quantum satisfiability problems as paradigmatic examples. The authors begin by reviewing the standard hierarchy of complexity classes. In the classical setting, problems are organized into P, NP, and the NP‑complete family, the latter representing the hardest problems in NP under polynomial‑time reductions. In the quantum realm, the analogous hierarchy consists of BQP (efficient quantum computation), QMA (quantum analogue of NP, where a quantum proof can be verified efficiently), and QMA‑complete problems, which are believed to be intractable even for quantum computers. These classifications are fundamentally worst‑case notions: they guarantee that some instances of a problem are hard, but they say nothing about the typical difficulty encountered by algorithms on randomly generated inputs.

To bridge this gap, the authors introduce statistical‑mechanical methods. By treating a random ensemble of problem instances as a disordered physical system, one can map logical constraints onto interaction terms in a Hamiltonian and study macroscopic observables such as free energy, entropy, and order parameters. For classical SAT, each Boolean variable becomes an Ising spin and each clause becomes a multi‑spin interaction; the resulting Hamiltonian counts unsatisfied clauses, and its ground‑state energy corresponds to the optimal number of satisfied clauses. Varying the clause‑to‑variable ratio (the “density” α) induces a phase transition from a SAT phase (zero ground‑state energy) to an UNSAT phase (positive ground‑state energy). This SAT‑UNSAT transition mirrors a thermodynamic phase transition and is accompanied by the emergence of a rugged energy landscape with an exponential number of metastable states—a hallmark of glassy behavior. The authors argue that such glassiness explains why local search algorithms become trapped and why typical instances near the critical α exhibit dramatically increased running times.

The quantum extension, quantum SAT (QSAT), replaces classical clauses with projectors acting on subsets of qubits. The Hamiltonian H = Σ_i Π_i encodes the constraints, and a satisfying assignment exists iff the ground‑state energy is zero. Random QSAT ensembles are constructed by selecting projectors uniformly at random at a given density α. Using replica‑symmetric mean‑field theory and numerical exact diagonalization, the authors locate a quantum SAT‑UNSAT threshold α_c^Q. Below α_c^Q, almost all instances are satisfiable and the ground state typically exhibits low entanglement; above the threshold, the ground‑state energy becomes extensive, and the system enters a highly entangled, frustrated phase. Notably, the entanglement entropy shows a sharp increase at the transition, indicating that quantum correlations play a central role in the hardness of QSAT, beyond the combinatorial structure captured by classical SAT.

Algorithmic implications are explored in depth. Classical algorithms such as DPLL, backtracking, and local search experience exponential slowdown near the classical critical point because the search space fragments into many isolated basins. Quantum algorithms—adiabatic quantum optimization, quantum annealing, and variational quantum eigensolvers—can in principle tunnel through energy barriers, offering a potential advantage in the glassy regime. However, the authors caution that for QMA‑complete problems, even quantum algorithms are unlikely to achieve polynomial‑time performance on worst‑case instances; the typical‑case analysis merely shows that some random instances may be easier, but the hardness gap remains.

The concluding section emphasizes the synergistic relationship between complexity theory and statistical mechanics. Complexity theory provides a rigorous classification of problem hardness, while statistical mechanics supplies tools to quantify the average‑case behavior of large ensembles. By combining these perspectives, one can predict regions of parameter space where algorithms are expected to succeed, identify glassy regimes that cause algorithmic failure, and guide the design of both quantum hardware (e.g., setting interaction strengths to avoid glassiness) and software (e.g., choosing problem encodings that stay below critical densities). The authors suggest that this interdisciplinary framework will be essential for assessing the practical power of forthcoming quantum computers and for developing robust algorithms that can cope with the intrinsic randomness of real‑world problem instances.