MOTIVATION
Geant4 is the simulation backbone behind ATLAS, CMS, LHCb, ALICE, and hospital-based
proton/carbon-ion therapy centres — any error in its physics core propagates into real
detector design and real patient dose plans.
The toolkit offers 10 EM physics-list constructors, each tuned to a different accuracy-vs-
speed trade-off, yet no published reference tests them across six particle species (e⁻, e⁺,
γ, p, π±, K±) under one fixed geometry.
A wrong physics-list choice is not a minor rounding issue — it produces 5–15%
systematic error, large enough to be clinically significant in dosimetry and
experimentally significant in calorimeter design.
This poster is built to let a fellow HEP researcher — detector physicist, medical physicist,
or Geant4 developer — see in seconds which physics list fits their own energy range and
material, without reading the full thesis.
RESEARCH OBJECTIVES
Characterise the Geant4 toolkit architecture and simulation workflow well enough that
any reader can reproduce every result independently.
Document the physical basis of every EM process Geant4 models — ionisation,
bremsstrahlung, multiple Coulomb scattering, photoelectric effect, Compton scattering,
pair production — with the cross-section formula behind each.
Compare all major EM constructors (Opt0, Opt3, Opt4, Livermore, Penelope) by energy
range, accuracy, and computational cost, using identical detector geometry and scoring
code for every particle — this control is what makes the comparison valid rather than
anecdotal.
Investigating this matters because Geant4 is used across HEP, medical physics, and space
science by groups who rarely benchmark their own physics-list choice — this thesis gives
them a quantitative basis instead of a default assumption.
Extend validated results to charged pions and kaons, a comparison largely absent from
existing Geant4 EM literature, to test whether accuracy trends seen in electrons and
protons hold for heavier, less-studied particles.
CONCLUSIONS
The central question this thesis answers: does EM physics-list choice produce
measurable, application-relevant differences across particle species — and can that be
turned into a quantitative rule? The answer is yes, on both counts.
Opt4 held <1% agreement with NIST ESTAR for electron stopping power across nine
orders of magnitude in energy (10 keV–10 GeV); Opt0 drifted 5–8% low-energy and
missed the lead photoelectric K-edge by 14% — Livermore corrected that to <0.3%.
The Bragg-peak height computed with Opt0 differs from Opt4 by 5–7%, exceeding the 3–
5% margin hadron-therapy dose planning allows — turning a physics benchmark directly
into a clinical safety finding.
Extending to kaons produced a result unavailable in the existing literature: the kaon/pion
model crossover-energy ratio equals 3.53, matching the K/π mass ratio — direct
confirmation that Geant4's Bethe-Bloch treatment scales correctly by particle mass, not
only for light leptons.
Wider implication: any Geant4 user — HEP detector group, hadron-therapy centre, or
shielding/spectroscopy team — now has a quantitative, cross-particle reference for
physics-list selection, rather than relying on the toolkit's default configuration.
FUTURE WORK
Convert the validated benchmark curves into training data for a lightweight ML
surrogate that predicts stopping power and Bragg-peak position without a full Monte
Carlo run.
Submit the curves to geant-val, the Geant4 Collaboration's own validation platform,
where hadron-therapy benchmarking is an active 2026 development target.
Re-run the kaon/pion comparison once planned higher-order QED bremsstrahlung
corrections are merged into the EM physics lists.
Test whether the same cross-particle accuracy holds once Geant4's EM transport is
ported to GPU hardware via AdePT and Celeritas.