Experimental validation of the scintiPulses shot-noise model for coincidence resolving-time dependencies in liquid scintillation counting.

In Liquid Scintillation Counting (LSC), the accurate standardization of pure beta and low-energy electron emitters relies heavily on coincidence detection architectures, where the coincidence resolving time is a highly sensitive parameter. Signal quantization and delayed fluorescence cause a delay in the initial detected charge, leading to significant event loss if the coincidence window is not appropriately set. To elucidate the physical mechanisms driving this resolving-time dependence, this study employs a comprehensive Monte Carlo framework utilizing the open-source Python packages TDCRPy and scintiPulses to emulate the entire detection chain. Experimental measurements of 3H, 55Fe, and 14C utilizing a Triple-to-Double Coincidence Ratio (TDCR) system were compared against single- and double-exponential scintillation decay models. The results demonstrate that while a single-exponential model accurately captures prompt stochastic jittering for short resolving times (under 20 ns), it significantly underestimates detection efficiencies at larger windows. Introducing a delayed fluorescence component successfully corrects this drift, achieving agreement between the model and experimental data within ±1% for double coincidences across all tested resolving times. These findings validate the shot-noise model implemented in scintiPulses to emulate the scintillation signal and perfectly corrects that delayed fluorescence, driven by triplet-triplet annihilation, is the primary physical mechanism responsible for resolving-time-dependent efficiency variations in LSC systems.