Boltzmann Entropy and the Arrow of Time in Cosmological Models

The thermodynamic arrow of time is conventionally traced to the growth of Boltzmann entropy from an extraordinarily low-entropy initial state, but computing that entropy for a self-gravitating, expanding universe remains technically difficult because gravitational clustering, unlike ordinary gas mixing, drives systems away from spatial uniformity rather than toward it. We construct a coarse-grained Boltzmann entropy estimator for a simplified homogeneous-background N-body cosmological model, tracking both matter (clustering) entropy and radiation entropy across 30 simulated expansion histories that vary initial density-perturbation amplitude and background expansion rate, from an initial scale factor a = 0.01 to a = 1.0 (present day), each resolved at 400 time steps and 50,000 particles. Total coarse-grained entropy increased monotonically in all 30 simulated histories, rising by a mean factor of 8.4 x 10^3 (SD 1.1 x 10^3) from initial to final scale factor, with matter clustering entropy contributing 78% of the total entropy increase and radiation dilution entropy contributing the remaining 22%, a split that was stable to within 4 percentage points across all tested initial perturbation amplitudes. Matter entropy growth was strongly nonlinear in scale factor, remaining below 15% of its final value until a = 0.3 before rising steeply as nonlinear gravitational clustering set in, whereas radiation entropy grew smoothly and nearly linearly with the logarithm of the scale factor throughout. Varying the initial density-perturbation amplitude by a factor of 4 shifted the scale factor at which matter entropy growth became dominant by only Deltaa = 0.06, indicating a comparatively weak dependence of the onset of entropy-dominated clustering on initial condition amplitude within the range tested, while varying the background expansion rate by ±20% shifted total final entropy by 11%. Comparing our coarse-grained matter entropy estimator against the commonly cited order-of-magnitude estimate based on supermassive black hole entropy, our simulated late-time matter entropy reached 3.1% of the black-hole-dominated estimate, consistent with the expectation that black hole entropy, not included in our N-body model, dominates the entropy budget of the present-day universe by orders of magnitude. These results provide a reproducible, quantitative coarse-graining framework for tracking the growth of gravitational and radiative entropy separately across cosmic history and clarify how the relative timing of matter versus radiation entropy production shapes the thermodynamic arrow of time in an expanding, self-gravitating universe.