Contemporary Physics Top 100 Dilemmas, Episode 99: many-body localization and the breakdown of thermalization. Picture a quantum city sealed inside a glass box. Its residents are electrons, spins, atoms, and local excitations. They are not frozen. They still hop, turn, shake, and affect one another. In the ordinary story, if the city is busy enough, information spreads like ink in water. A hot spot, bright spot, or special starting pattern should fade into a shared background. Later, one district should remember only coarse numbers: temperature, density, and average energy. That is the usual intuition of thermalization. Many-body localization is strange because some disordered, interacting, nearly closed quantum systems refuse that script. They keep moving, yet they do not become one well-stirred soup. Certain local observables can remember the initial state. Transport can be strongly suppressed. Entanglement can grow, but unusually slowly. It is like a city whose lights still flicker while main roads, elevators, and pipes have collapsed into short local segments. Every building is active. The city still refuses to become one common bath. Mainstream physics does not struggle because the phenomenon is invisible. The hard part is deciding what it truly is. In low-dimensional, short-range, well-isolated systems, the behavior can look persuasive. But in higher dimensions, with long-range interactions, or with weak environmental coupling, the question becomes slippery. Is this a strict long-time phase, or only a long prethermal plateau? Is it a real thermodynamic-limit phenomenon, or a finite-size illusion? Is it a genuine breakdown of thermalization, or ordinary disordered slow dynamics that would eventually thermalize if we waited longer? It also challenges the eigenstate thermalization hypothesis, or ETH: many complex isolated systems should make local regions look thermal through their own many-body eigenstates. Many-body localization replies: not always. So the checklist becomes severe: what is truly thermal, what is merely slow, when does ETH fail, and when does it return? EFT begins by taking apart the word thermalization. Thermalization is not magic heat dust sprinkled over a system, and it is not just the slogan that time makes everything forget. In EFT language, a quantum state is first a menu of available channels. Under the current object structure, sea condition, and boundary condition, which local channels can close? Which channels can connect far away? Which beats can cross-check? Thermalization happens when many local settlements repeat, information leaks into wider degrees of freedom, accessible channel volume keeps reshuffling, and the whole system is pulled toward a statistical attractor. In plain language, many small rooms keep opening their doors, inventories cross corridors, labels get erased, and the building finally carries one shared thermal account. In that frame, many-body localization is not the second law being overthrown, and it is not a quantum system receiving eternal memory as a supernatural gift. It is an extreme working state of channel fragmentation. Disorder scrambles corridor addresses, so some doors no longer line up. Interactions make the furniture in each room jam against other furniture: moving one chair may be possible, but rearranging the whole floor becomes expensive. Boundaries and weak environmental couplings act like gate controls, deciding which information can truly leak out and which information only circles locally. The channel network that might have grown into a system-wide thermal bath is chopped into many local pockets. Inside each pocket, real rearrangement still happens. But phase, memory, energy inventory, and structural labels cannot easily escape through long, coherent, low-cost shared corridors, so they do not merge into one global coarse-grained thermal ledger. This explains why many-body localization can look active and nonthermal at the same time. It is not motionless ice. It is motion that stays too local. It is not absence of information. It is information unable to find sufficiently open, coherent, low-cost routes for global diffusion. Imagine an apartment building where every room contains people talking, but thick walls, narrow doors, and broken hallways keep every voice local. Local memory survives because labels are not washed flat by the full network. Transport is suppressed because shared corridors do not connect. Entanglement grows slowly because pockets leak into one another bit by bit instead of opening into a whole-building communication grid. One guardrail matters. EFT does not declare that every MBL experiment is finally classified. It does not deny prethermalization, finite-size effects, rare thermal inclusions, avalanches, decoherence, or the possibility that higher dimensions and long-range interactions may reopen the network. EFT puts those warnings onto the same map: how long can fragmentation survive, do local pockets have slow leaks, can weak coupling reopen corridors, and can high-dimensional connectivity regrow a bath? If the big network eventually grows back, the system was not a strict long-time MBL phase. If fragmentation survives hard tests of isolation, scale, and time, then it earns the name. The deepest meaning of many-body localization is not that thermodynamics has failed. It is that thermalization needs roads. Without a connected network that erases microscopic labels and merges local ledgers into one coarse-grained basin, the system can stay busy in many small account books without one shared thermal statement. Open the playlist for more. Next episode: fusion plasma turbulence and anomalous transport. Follow and share, and let this series of new-physics explainers help you see the universe more clearly.
EFT
Language