Dichroic X-ray tomography resolves a sample’s 3D crystal orientation (linear dichroism) or magnetization (circular dichroism) from projections taken at varying beam directions and polarizations. Like most tomographic methods, its reconstruction assumes that absorption along a ray is a simple line integral of a local, voxel-independent absorption coefficient. This work showed that for linear dichroism that assumption can fail: as the beam crosses an anisotropic material whose optic axis is misaligned with the polarization, the polarization itself rotates dynamically. Using a finite-element Maxwell solver, full-field tomography of a 2000 Å aragonite Voronoi polycrystal was simulated at the O K-edge. Reconstructions indicated that polarization rotation through propagation (PRP) introduces real artifacts, including spurious grain boundaries, false orientation gradients, and the apparent splitting of single grains. The authors found that the reconstruction error increased by three orders of magnitude as the dichroism was tuned from weak (0.1) to full strength (1.0). Qualitatively, the observed reconstruction artifacts mimic genuine microstructure, demonstrating a potential for false conclusions to be drawn about a material’s orientational statistics or grain-boundary structure.
These results identify a previously underappreciated, physics-based limitation of linear dichroic orientation tomography, a technique of growing importance for imaging anisotropic functional materials. This work also supplies practical diagnostics, including degraded sinogram fit residuals and a rotationally invariant biaxiality measure that flags voxels where the uniaxial reconstruction is likely to break down. The authors chart concrete mitigation paths: acquiring redundant data across additional tilt and polarization conditions so the reconstruction can down-weight the worst-affected angles, operating at energies where dichroism is weak, or folding a beam-propagation model directly into the reconstruction. Additionally, they outline a fixed-point iterative scheme and a more robust model-based iterative reconstruction (MBIR) framework for performing propagation-aware reconstructions that keep dichroic tomography quantitatively reliable.