Testing Matter Diffusion with Late-Time Cosmological Observations

Abstract

We investigate a class of late-time cosmological models derived from the phenomenological framework of variable matter diffusion. In these scenarios, energy-momentum conservation requires a continuous energy exchange between matter and an effective scalar-field dark-energy component, $ϕ$. We consider a baseline constant-diffusion scenario alongside four non-linear power-law parametrizations, in which the diffusion coefficient evolves as a function of the scale factor, matter density, scalar-field density, or Hubble expansion rate. To assess their cosmological viability, we implement these models within \texttt{SimpleMC} and constrain them using late-time observations, including cosmic chronometers, baryon acoustic oscillation measurements, Type Ia supernovae with and without the local SH0ES calibration. In the absence of the local calibration, the diffusion models yield only modest improvements in the fit ($Δχ^2 \approx -4$), performing comparably to the CPL parametrization while exhibiting negative Bayesian log-evidence differences relative to $Λ$CDM. In contrast, including SH0ES leads to a dramatic reduction in the minimum $χ^2$ ($Δχ^2 \approx -22$) and decisive Bayesian evidence in favor of the diffusion framework ($Δ\ln\mathcal{Z} \approx 9$). Remarkably, the single-parameter constant-diffusion model accounts for virtually all the statistical improvement, outperforming the CPL parametrization by more than 10 units in $χ^2$ and more than 8 units in $Δ\ln\mathcal{Z}$. The four non-linear extensions provide only negligible additional improvements ($Δχ^2 \approx -1$), indicating that current late-time background observations favor the presence of a non-vanishing matter-diffusion interaction over $Λ$CDM, while providing no statistically significant evidence for a specific time-dependent functional form of the diffusion coefficient.