Objective
The chair takes on the two X-ray contrasts that carry the most information and resist imaging the hardest: fluorescence, which identifies elements at trace concentration but arrives at a detector with no positional information, and small-angle scattering, which describes structure far below the pixel but requires point-by-point acquisition to measure properly.
Both are usually addressed by scanning — a focused beam, one position at a time, days of beamtime for a single tomogram. FANFARES replaces the scan with structured illumination and reconstruction: pattern the beam, record a sequence of total intensities, and recover the map by solving an inverse problem. Done well, the number of measurements needed is far smaller than the number of pixels recovered.
That combination is why the chair sits where it does. Nicolas Ducros’ group at CREATIS built much of the single-pixel imaging methodology and its learned reconstruction; Nicola Viganò at CEA IRIG works on tensor tomography and the inversion of scattering data; this group brings the X-ray wave-optics and the experimental platforms. The method work is described on the single-pixel sensing page.
Doctoral contracts
Two three-year contracts run from October 2026 to September 2029:
- SAXS tensor tomography — reconstructing the three-dimensional orientation of sub-pixel structure from projections that mix scattering directions. Based at IAB in Grenoble.
- XRF ghost tomography — element-specific three-dimensional maps from a compressive acquisition and a single-element fluorescence detector. Hosted at CREATIS in Lyon.
Why it matters beyond the chair
MIAI is Grenoble’s artificial-intelligence institute, and the chair is explicit that the learning belongs in the reconstruction rather than in place of the physics. A learned prior that fills in what a compressive acquisition could not measure is doing legitimate work; a network that replaces the forward model is not. Keeping that line clear is part of the scientific programme, not a caveat to it.