Bayesian soft X-ray tomography using non-stationary Gaussian Processes.

Area of Science:

• Plasma Physics
• Computational Physics

Background:

• Accurate inference of soft X-ray emissivity is crucial for understanding nuclear fusion plasma dynamics.
• Limited, noisy line integral measurements pose an ill-posed inversion problem for spatial profile reconstruction.

Purpose of the Study:

• Develop a Bayesian non-stationary Gaussian Process (GP) method for inferring soft X-ray emissivity.
• Quantify uncertainties associated with the inferred emissivity distribution.
• Improve the analysis of equilibrium and magnetohydrodynamic behaviors in fusion plasmas.

Main Methods:

• Utilized a Bayesian-based non-stationary Gaussian Process (GP) for modeling emission.
• Applied Bayesian probability theory to handle the ill-posed inversion problem.
• Optimized model complexity using Bayesian Occam's Razor formalism.

Main Results:

• The developed GP method provides convincing reconstructions of soft X-ray emissivity.
• The method effectively quantifies uncertainties in the inferred profiles.
• Results show good agreement with established methods like Maximum Entropy and Equilibrium-Based Iterative Tomography.

Conclusions:

• The Bayesian non-stationary GP method offers a robust approach for soft X-ray emissivity inference in fusion plasmas.
• Enhanced uncertainty analysis improves the reliability of scientific findings.
• This method facilitates better investigation of plasma equilibrium and dynamics.