Glasses have a diffusion puzzle. Local atomic rearrangements have low energy barriers — atoms can hop to neighboring sites without much thermal activation. Yet macroscopic diffusion through glasses requires far more energy than these local hops would predict. The activation energy for bulk diffusion is several times higher than the barrier for any single rearrangement. Where does the extra activation energy come from?

Back-and-forth motion. In a disordered energy landscape, forward and reverse barriers at each site are different. An atom that hops forward encounters a low barrier going but a different barrier coming back. If the reverse barrier is lower than the forward barrier at the next site, the atom returns to where it started. This back-and-forth correlation — hopping forward, falling back — means that many individual hops produce no net displacement. The effective diffusion rate reflects not the barrier to moving but the barrier to staying moved.

The framework was validated across metallic glasses, silica, and Lennard-Jones systems, demonstrating that the mechanism is general to structural disorder, not specific to any chemistry. It also explains why surface diffusion in glasses is faster: surface atoms have fewer neighbors, which means fewer pathways for correlated back-and-forth motion, which means weaker correlations, which means lower effective activation energy.

The structural insight is that in asymmetric energy landscapes, local mobility and macroscopic transport are decoupled by correlations. Each hop is easy. But the landscape keeps pulling atoms back. The difficulty of diffusion is not in the going — it is in the not-returning. What looks like a barrier problem is actually a correlation problem.

(arXiv:2603.18317)