Toward dynamic Monte Carlo source modeling: measurement variability and primary electron source-width sensitivity in very small photon fields

Background Very small photon fields are sensitive to both measurement conditions and radiation-source modelling, but the connection between effective primary-electron source width, downstream photon angular phase space, and field-size-dependent dosimetric response remains incompletely characterized. Methods The study was restricted to 6 MV. Empirical measurement variability was evaluated using five ionization-chamber datasets acquired on a clinical Elekta Versa HD for fields from 20×20 to 3 × 3 cm 2 . A generalized Versa HD/Agility EGSnrc/BEAMnrc model was evaluated with two prescribed stationary Gaussian source endpoints, M1 (FWHM x = FWHM y = 0.15 cm) and M2 (FWHM x = 0.35 cm, FWHM y = 0.60 cm). DOSXYZnrc dose responses and final pre-phantom photon phase-space data were analysed for 3×3, 2×2, 1×1, and 0.5×0.5 cm 2 . The 3×3 cm 2 field provided the experimental–Monte Carlo overlap; the smaller fields were controlled Monte Carlo endpoint-sensitivity analyses. M1 was additionally benchmarked against a 10×10 cm 2 CC13 commissioning PDD. Results Mean experimental coefficient of variation increased from 2.46% at 20×20 cm 2 to 15.45% at 3×3 cm 2 . The M2 central-axis reductions were 8.48%, 8.27%, 29.59%, and 56.34% for 3×3, 2×2, 1×1, and 0.5×0.5 cm 2 , respectively. The corresponding M2/M1 ratios of the 95% angular-containment solid angle were 1.275, 1.498, 2.415, and 3.835. The four-field comparison demonstrates clear field-size progression in angular containment, while showing that amplification is observable-dependent rather than universally monotonic. The M1 PDD benchmark supported reference depth-dose behaviour, whereas lateral-profile findings were restricted to controlled numerical sensitivity because absolute lateral agreement was not independently established in this study alone. Conclusion The demonstrated result is a stationary endpoint-sensitivity chain linking effective primary-electron source representation, linac-head transport, photon angular phase space, and small-field dose/profile response. A reduced covariance formulation is retained only as a prospective framework for testing whether a stationary effective source can represent unresolved accelerator-source states; no time-resolved source dynamics were measured or simulated.