Title: Experimentally Validated Simulations of 50 μm X-ray PIV Tracer Particles

Abstract: We evaluate two simulation methods for prediction and comparison of X-ray Particle Image Velocimetry (XPIV) system performance: Beer-Lambert (BL) ray tracing and Monte Carlo N-Particle (MCNP) photon tracking. Particle image signal-to-noise ratio (SNR) is used as the metric of XPIV system performance. These simulation tools enable the design and methodical improvement of XPIV experiments and tracer particles. Simulated and experiment data of hollow, silver-coated, glass sphere tracer particles (AGSF-33) are compared. As predicted by the simulations, the AGSF-33 particles are visible with a SNR greater than unity in 100 ms exposure time images, demonstrating their potential as X-ray PIV or particle tracking velocimetry (XPTV) tracers. The BL ray-tracing method, as implemented, sacrifices accuracy for computational efficiency. Nevertheless, the BL approach predicts the system behavior, is computationally inexpensive, and enables the exploration of a vast parameter space for system design. MCNP simulations, on the other hand, are able to more accurately predict experiment images. For most practical XPIV system applications, however, the higher computational expense of MCNP may not be justified by the modestly improved performance compared to BL.