Table 2 Simulation conditions used in this work, displayed as per the scheme proposed by the RECORDS report [12]

Code

PRIMO v. 0.3.1.1600, based on penEasy/penelope 2011, and DPM algorithm

Timing PSF

Simulation time: ≈10 d on an Intel Xeon E5-2670 v3, 24 cores @ 2.3 GHz, 64 GB RAM, Windows Server 2016

Timing 10×10 cm² field

Simulation time with DPM: 5.5 h on an Intel Xeon E5-2620. 2 CPU (×6 cores) @ 2.00 GHz, 32 GB RAM, Windows 7. Other fields: CPU time linearly proportional to the field area.

Source description Clinac 2100

PSF stage 1 simulated with PRIMO, 6 MV, initial energy: 5.4 MeV, energy FWHM: 0, focal spot FWHM: 0, beam divergence: 0, field size: 40 ×40 cm². Simulation engine: penEasy/penelope. Histories: 850 ×10⁶. PSF size: 150 GB

Source description TrueBeam

PSF stage 1 simulated with PRIMO, FakeBeam, 6 MV FFF, initial energy: 5.8 MeV, energy FWHM: 0.058 MeV, focal spot FWHM: 0.15 cm, beam divergence: 0, field size: 40 ×40 cm². Simulation engine: penEasy/ penelope. Histories: 850 ×10⁶. PSF size: 237 GB

Cross sections

penelope 2011 and DPM.

Transport parameters

PRIMO default transport parameters for 6 MV from Clinac 2100 and for 6 MV FFF from FakeBeam

Variance-reduction techniques

PSF simulations: splitting roulette. Movable–skins technique applied to the simulation of primary collimator, jaws and MLC. Water phantom simulations: particle splitting (×170, factor empirically determined)

Scored quantities

Absorbed dose to a voxelized water phantom of 30.2 ×30.2×30 cm³, voxel size 0.2 ×0.2×0.2 cm³. Simulation engine: DPM and penEasy/ penelope 2011

# histories/ statistical uncertainty

850 ×10⁶ histories. Statistical uncertainty of the calculated dosimetric parameters typically below 2% (k=2), estimated with the history–by–history method of penelope 2011

Post-processing

No smoothing or de–noising was applied to the simulation results.