How does Geant4's object-oriented design specifically enable swapping physics models without rewriting detector geometry code?

This is a great technical question, but the archive doesn't directly address Geant4's object-oriented design patterns for physics-model swapping—the provided papers discuss *applications* of Geant4 and alternatives to it, not its internal architecture.

That said, I can give you the conceptual answer from what's implied:

Geant4 is described as "an object-oriented toolkit for the simulation of the passage of particles through matter" [AG-2024.05-1380], and its widespread adoption across detector geometries, physics processes, and fields (space, nuclear, medical, archaeology) suggests a clean separation of concerns—a hallmark of good OO design.

The key principle is polymorphism: physics models (electromagnetic interactions, hadronic processes, etc.) and detector geometry are encapsulated as independent class hierarchies. A physicist writes geometry once as a detector description—solids, volumes, materials arranged in a hierarchy. Then, independently, they configure which physics *processes* act on particles in those volumes by plugging in different model implementations (Geant4 provides multiple options: QGSP, FTFP, etc.). The geometry code never needs to know *which* physics model is active; it just calls virtual methods on a generic "physics process" interface.

The DMG4 package illustrates this: they "integrate this simulation into the same program that performs the full simulation of particles in the experiment setup" by adding dark-matter physics processes to Geant4 without rewriting the detector geometry [AG-2024.01-1120]. The detector geometry is agnostic to what particles it's simulating.

To see the actual design (member classes, virtual function tables), you'd need Geant4's source code or design documentation, not these research papers.