The paper tackles a long-standing problem in statistical physics: proving rigorously that a spin glass can actually exist as an ordered phase at positive (nonzero) temperature in a three-dimensional setting. A spin glass is a disordered magnetic material where the interactions between neighboring atoms are random, some favoring alignment and others opposing it. This frustration leads to a frozen, disordered kind of order at low temperatures that is distinct from ordinary magnetism. Proving this order exists mathematically has been notoriously difficult, and before this work the best rigorous result required an absurdly large number of neighbors (more than a billion billion), making it physically meaningless.
The authors work on a class of hierarchical lattices, which are recursively built networks that mimic the geometry of real three-dimensional space well enough to share its critical behavior. They prove that on the physically relevant three-dimensional version of this lattice, a genuine spin-glass phase exists at sufficiently low temperatures. The core mathematical tool is tracking how an effective coupling strength grows under repeated renormalization, meaning as you zoom out and look at the system on larger and larger scales, the frozen glassy order becomes more and more pronounced rather than washing out. For standard random couplings they also use computer-assisted interval arithmetic to push the proof up to temperatures close to the numerically estimated freezing point.
A secondary result concerns how the spin-glass phase reveals itself in a percolation picture, a way of visualizing the phase transition through connected clusters. The authors prove that a single dominant cluster spans the entire system while all other clusters remain negligible in size, an imbalance that serves as a geometric fingerprint of spin-glass order. Strikingly, this is the first time such a cluster imbalance has been rigorously proven for any spin glass other than the fully connected mean-field model, making it a meaningful step toward understanding glassy phases in realistic spatial settings.