Objective
Most interesting processes happen at more than one scale at a time. A battery electrode degrades atom by atom and fails millimetre by millimetre; a snowpack metamorphoses grain by grain and avalanches by the slope. Tomography usually forces a choice between the two: a wide field at coarse resolution, or a fine field that misses the context.
MUSITOX develops methods for imaging both scales simultaneously, in three dimensions and in real time. That means extending X-ray microtomography in three directions at once: towards time-resolved acquisition, towards phase contrast — so that weakly absorbing phases stop being invisible — and towards unconventional acquisition geometries that a classical reconstruction would not accept. Machine learning enters on the acquisition and reconstruction side, where it buys back the signal that shorter exposures give away.
Why it is cross-disciplinary
The programme is deliberately built across communities that do not usually share an instrument:
- materials — composites, metals, batteries;
- public health — bone architecture and its disorders, osteoarthritis in particular;
- environment and engineering — flow in porous media, suspension mechanics, snow.
Around fifty researchers from nine Grenoble laboratories take part, with the CMTC and GRICAD platforms, spanning the PEM, PAGE and CBS research clusters. For our group it is the project that keeps the methods honest: an algorithm tuned on soft tissue meets a metal foam and a wet granular medium, and its hidden assumptions surface quickly.