Testing halo models for constraining astrophysical feedback with multi-probe modeling: I. 3D Power spectra and mass fractions

Abstract

Upcoming Stage-IV surveys will deliver measurements of distribution of matter with unprecedented precision, demanding highly accurate theoretical models for cosmological parameter inference. A major source of modeling uncertainty lies in astrophysical processes associated with galaxy formation and evolution, which remain poorly understood. Probes such as the thermal and kinematic Sunyaev-Zel'dovich effects, X-rays, and dispersion measure from fast radio bursts offer a promising avenue for mapping the distribution and thermal properties of cosmic baryons. A unified analytical framework capable of jointly modeling these observables is essential for fully harnessing the complementary information while mitigating probe-specific systematics. In this work, we present a detailed assessment of existing analytical models, which differ in their assumptions and prescriptions for simultaneously describing the distribution of matter and baryons in the universe. Using the Magneticum hydrodynamical simulation, we test these models by jointly analyzing the 3D auto- and cross-power spectra of the matter and baryonic fields that underpin the above probes. We find that all models can reproduce the power spectra at sub-percent to few-percent accuracy, depending on the tracer combination and number of free parameters. Their ability to recover underlying halo properties, such as the evolution of gas abundance and thermodynamic profiles with halo mass, varies considerably. Our results suggest that these models require further refinement and testing for reliable interpretation of multi-wavelength datasets.

Figures (12)

• Figure 1: Correlations between halo model parameters for combinations of different matter fields. Left to right: Parameter correlations for matter--matter, electron density--electron density, matter--pressure 3D power spectra and when analyzing all of them combined. Top to bottom: Parameter correlations for the Mead20+, Schneider19+, and Arico24+ models, respectively. We see that combining information from different probes helps in weakening parameter correlations, however, the degree of decorrelation varies across models. Note that the parameters $\{\alpha,T_\text{w} \}$ do not impact $P_{\rm mm}(k)$ and $P_{\rm ee}(k)$, hence they are not correlated with the other model parameter for the Mead20+ model. The same applies to $\{\alpha_\text{nt}, \gamma_\text{nt}\}$ in the case of Schneider19+ and $\{A_{\rm th},T_\text{w}\}$ for Arico24+.
• Figure 2: Projection of parameters across halo model implementations. Each panel illustrates how well the impact of a single parameter on the concatenated data vector ($P_{\rm mm}$, $P_{\rm ee}$, $P_{\rm mp}$) can be represented within other models. The dashed black line denotes when a parameter can be perfectly represented. Panels from left two right show projections of the Mead20+, Schneider19+, and Arico24+ models, respectively.
• Figure 3: Accuracy of predicted thermodynamic profiles from models calibrated on power spectra and mass fractions. Columns from left to right show results for the Mead20+, Schneider19+, and Arico24+ models, respectively. The horizontal axis shows the tracer combination on which the model is calibrated. Tracer combinations marked by $\dagger$ are fit using a six-parameter version of the models while the remaining combination are analyzed with the extended set (Table \ref{['tab:model_params']} ). The vertical axis shows the RMS error on the electron pressure ($\mathcal{P}_{\rm e}$), temperature ($T_{\rm e}$), and density ($n_{\rm e}$) profiles. We compute the RMS error between the best-fit prediction of the profile and simulation measurement, over the range $0.12<r/r_{\rm 200c}\leq1$ for cluster-sized halos (top row) and $0.25<r/r_{\rm 200c}\leq1$ for group-sized halos (bottom row).
• Figure 4: Results from joint fits to $R_{\rm mm}(k)$ and $R_{\rm mp}(k)$ from Magneticum (at $z=0.25$) using a six parameter model (parameters marked with $\dagger$ in Table. \ref{['tab:model_params']} ). Left: Best-fit predictions from the three models considered here and simulation measurements. The light shaded region represents the assumed standard deviation of 5%. Right: Panels (a) through (d) show the predicted bound gas fraction and electron pressure, temperature, and density profiles, respectively, and the colored shaded regions represent the $16^{\rm th}-84^{\rm th}$ percentile regions. While we show full the inferred profiles down to $x=(R/R_{\rm 200c})=0.1$, note that the power spectra responses up to $k<4\,{\rm Mpc}^{-1}h$ that we fit to are only sensitive to physical scales of $x\gtrsim0.25$ for group mass halos ($13\leq\log (M_{\rm200c}h/\mathrm{M}_\odot) <13.5$) and $x\gtrsim0.12$ for cluster mass halos ($14.5\leq\log (M_{\rm200c}h/\mathrm{M}_\odot)<15$). We use downward arrows to indicate where the model prediction differs from the simulation by $\gtrsim10\times$. See Sec. \ref{['sec:res_min']} for related discussion.
• Figure 5: Same as Fig. \ref{['fig:fit_min']} but fitting $R_{\rm mm}(k)$ and $R_{\rm mp}(k)$ from the Magneticum simulation with an extended parameter set (Table \ref{['tab:model_params']} ). See Sec. \ref{['sec:mm_mp_ext']} for related discussion.