Characterization and Mitigation of Boundary Sampling Artifacts in Electron Phase Space Generation for Clinical FLASH-RT Deployment.

Background: Clinical implementation of Monte Carlo (MC)-based treatment planning for electron FLASH radiotherapy requires computationally efficient phase space (PHSP) generation strategies to meet the planning timelines of ultra-high-dose-rate (UHDR) delivery. Applicator-specific PHSP files, recorded at the aperture exit, reduce dose calculation time by 30-50% relative to traditional placement at the linac-exit window, enabling clinically practical planning cycles for FLASH treatment planning system (TPS) deployment. However, positioning PHSP scoring planes coincident with the applicator-air interface introduces boundary sampling artifacts that have not been previously characterized in the electron PHSP literature. Purpose: To identify and characterize a boundary sampling artifact arising in Geant4-based Monte Carlo simulations when PHSP scoring planes are positioned coincident with material-air interfaces, and to demonstrate its mitigation through systematic scoring plane offset, enabling clinical deployment of applicator-specific electron beam PHSP libraries for FLASH-RT treatment planning. Methods: Applicator-specific PHSP files were generated using GAMOS 6.2.0 for a commissioned 9 MeV Mobetron UHDR Monte Carlo model across twelve clinical aperture configurations (2.5-10 cm diameter) intended for deployment in the GMV Radiance treatment planning system. The scoring plane sensitivity analysis was performed on the A6 short cone (aperture exit at 397 mm) across three positions: coincident with the material-air interface (397 mm), 0.1 mm downstream (397.1 mm), and 1 mm downstream (398 mm). The corresponding 1 mm offset was subsequently applied to the A10 long cone (aperture exit at 406 mm, scoring plane at 407 mm) for global validation. Dose calculations were benchmarked against linac-exit PHSP references and experimental water phantom measurements via distance-to-agreement (DTA) analysis against the 3 mm clinical acceptance criterion established by AAPM TG-105. Results: Scoring at the exact interface produced systematic proximal R50 shifts of up to 2.2 mm and DTA values of 4-6 mm for small-aperture configurations, exceeding the clinical acceptance criterion and rendering the PHSP library unsuitable for clinical deployment. A 0.1 mm offset partially corrected the angular distribution at the boundary but failed to resolve the energy-loss artifact affecting the distal falloff region. A 1 mm offset fully resolved all artifacts, achieving mean DTA values of 1.45 ± 0.63 mm (A6 series PDDs) and 2.00 ± 0.61 mm (A10 series PDDs), equivalent to or better than the linac-exit reference across all twelve configurations. Conclusions: Boundary sampling artifacts in Geant4-based PHSP generation arise from the degenerate behavior of the fUseSafety step-limitation algorithm when safety = 0 at exact material boundaries, producing incomplete secondary electron equilibration and a systematic energy-loss artifact in the scored particle population that manifests as a proximal shift of the distal dose falloff. A 1 mm downstream offset fully mitigates these artifacts while introducing negligible perturbation to primary beam characteristics, enabling clinical deployment of the resulting applicator-specific PHSP library in the GMV Radiance TPS for FLASH-RT treatment planning. This offset requirement warrants investigation in other Geant4-based frameworks (including TOPAS and GATE) scoring PHSP files at applicator or collimator exit surfaces.