Spin Glass

A spin glass is a disordered magnetic system in which the interactions between spins are randomly distributed — some ferromagnetic (favoring alignment), some antiferromagnetic (favoring opposition) — producing a rugged energy landscape with exponentially many local minima. Spin glasses were introduced as models of magnetic alloys but have become canonical examples of complex systems with frustrated interactions, serving as the physical substrate for understanding critical phenomena, Hopfield networks, and combinatorial optimization.

The theoretical analysis of spin glasses, particularly the Sherrington-Kirkpatrick model, required the development of replica symmetry breaking — a mathematical technique that revealed the hierarchical organization of states in disordered systems. This structure, in which pure states are organized into clusters at multiple scales, has direct analogues in the energy landscapes of neural networks and the solution spaces of hard constraint-satisfaction problems.