Sparse-view 3D tomographic reconstruction of MagLIF stagnation plasmas with learned basis functions

Uncovering 3D structure of the stagnated fusion fuel and liner mix in Magnetized Liner Inertial Fusion (MagLIF) experiments is critical to understanding target performance. However, accurate diagnosis of 3D structure in MagLIF experiments on the Z Machine has been limited by extremely sparse data, often two or less 2D images of self-emission at stagnation. Here we present an approach to accurately reconstruct the morphology of stagnation columns with just two orthogonal views using a set of basis functions learned from training volumes with quasi-helical structures relevant to stagnation plasmas. Reconstructions of emission volumes generated from radiation-magnetohydrodynamic simulations show more accurate morphology compared to when using other bases and forms of regularization. In addition, the learned basis provides accurate estimates of fuel volume, which are necessary for inferences of stagnation pressures. We also present the first 3D reconstructions of experimental MagLIF stagnation plasmas. The approach is applicable to sparse-view 3D reconstruction of other Z-pinch and HED plasmas.