What physics processes does PYTHIA's event generation miss that Geant4 must then simulate in the detector?

Your question assumes a division of labor that doesn't quite match how these tools are actually used together.

PYTHIA and Geant4 operate at fundamentally different scales of the collision chain, so they don't really "divide up" the same physics problem [AG-2026.03-1244]. PYTHIA generates the *initial hard collision event*—the outcome when two particles collide head-on: which new particles are created, at what angles and energies [AG-2026.03-1244]. Geant4 then simulates what happens to those particles as they *travel through the detector material* [AG-2024.05-1380]. These are sequential, not overlapping, stages.

That said, there are some physics processes that PYTHIA handles incompletely and that Geant4 must then refine:

Intrinsic parton motion and soft radiation. PYTHIA models the "intrinsic transverse momentum" of quarks and gluons inside colliding hadrons, but this model has limitations at high collision energies [AG-2024.12-1320]. Geant4's detailed material interactions can capture some effects of these initial-state asymmetries through final-state particle distributions that deviate from PYTHIA's predictions.

Hadronic fluctuations in nuclear collisions. PYTHIA's Angantyr module for heavy-ion collisions includes only a "rudimentary model" of hadronic fluctuations [AG-2024.06-1222]. Geant4 can simulate nuclear interactions with finer granularity as particles traverse the detector.

The underlying event in certain beam configurations. While PYTHIA models multiparton interactions (the "background" activity in a collision), its tuning varies across different beam types (photon-hadron, photon-photon, hadron-hadron) [AG-2024.08-1270]. Geant4 provides independent constraints on particle multiplicities and spatial distributions through detector measurements.

However, the provided preprints don't systematically catalog what PYTHIA *misses* that Geant4 then *must* simulate—that inventory isn't their focus.