Grain boundary diffusion in ceramics is modeled as movement along a fixed channel. The boundary has a structure; the dopant moves through it; the diffusion coefficient is a property of the boundary. This is straightforward transport theory. The channel determines the flow.
Researchers at the University of Tokyo found that Ti atoms diffusing along alumina grain boundaries transform the boundary structure itself. The boundary begins asymmetric. As Ti concentration increases at the diffusion front, the grain boundary transitions to a symmetric configuration. This structural transformation is not a side effect — it changes the diffusion coefficient by an order of magnitude.
The diffusion creates the conditions for more diffusion. The first atoms to arrive restructure the boundary, making it easier for subsequent atoms to follow. The channel widens as traffic passes through. This is a self-reinforcing loop: diffusion → structural transformation → enhanced diffusion → further transformation. The diffusion front isn't advancing through a static medium. It's advancing through a medium it has already modified.
The standard sintering models assume constant grain boundary properties — a diffusion coefficient measured at one concentration, applied across all concentrations. This paper shows that the coefficient depends on the accumulated history of diffusion at that point. The boundary remembers what has passed through it, and that memory changes what can pass through next. Two boundaries with identical initial structures but different diffusion histories will have different diffusion coefficients, because the history of traffic restructured the channel differently.
The engineering implication is direct: optimal sintering conditions cannot be calculated from initial grain boundary properties alone, because those properties change during sintering. The measurement that predicts the process destroys its own validity by initiating the process. The parameter is consumed by the phenomenon it describes.