A proposal to construct the dark-matter-only counterpart of the observed Universe combining weak lensing and baryon censuses

Abstract

Baryonic effects such as AGN feedback can significantly impact the matter clustering, are harder to model from first principles, and emerge as a severe limiting factor in weak lensing cosmology. To tackle this issue, we propose a generic relation of mapping the observed matter clustering to its counterpart in a dark-matter-only universe. We verify this relation to be accurate at better than $1\%$ level at $k<1\,h/$Mpc and $z\in [0,3]$ in both TNG and Illustris simulations, demonstrating its model-independence to the underlying baryonic physics. Implementing this relation in observations will be made possible by the specifically designed cross-correlation statistics and baryon census (ionized diffuse gas through localized fast radio bursts, stellar mass through galaxy surveys, and neutral hydrogen through 21cm mapping). It is capable of correcting the baryonic effect not only in the matter power spectrum, but also at the field level, as demonstrated by tests on the scattering transform statistics. This approach paves the way for constructing the dark-matter-only counterpart of the observed Universe, establishing an ideal cosmological laboratory for probing the dark universe.

Figures (8)

• Figure 1: Validation of the assumptions ( I) and ( II) using TNG300-1 (solid lines) and Illustris-1 (dashed lines) simulations. The shaded regions mark the $\pm 1\%$ deviation. Top: Cross-correlation coefficients between total matter overdensity $\delta_m$ in hydrodynamical simulation and its DMO counterpart $\delta_{\rm DMO}$. Middle: Cross-correlation coefficients between cold dark matter overdensity $\delta_c$ and $\delta_{\rm DMO}$. Bottom: Ratio of the power spectrum $P_{cc}$ of $\delta_c$ to $P_{\rm DMO}$ of $\delta_{\rm DMO}$. Despite the different subgrid physics implemented in two simulations, both confirm assumption ( I) to be $<0.1\%$ accuracy and assumption ( II) to be $<1\%$ accuracy at scales $k<1{\, h/{\rm Mpc} }$.
• Figure 2: Validation of the key relation Eq. (\ref{['equ:T']}). We show the ratio of the estimated transfer function $\hat{{\mathcal{T}}}$ to the ground truth ${\mathcal{T}} = \sqrt{P_{\rm DMO} /P_{mm}}$, measured in TNG300-1 (top) and Illustris-1 (bottom) simulations. Luminous subhalos are identified as galaxies and cross-correlated with the ionized electrons ($\delta_e$), stars $+$ black holes ($\delta_*$), and neutral hydrogen ($\delta_ {\rm HI}$) to measure $\{ f_e b_e, f_* b_*, f_ {\rm HI} b_ {\rm HI} \}$, and hence $\hat{{\mathcal{T}}}$. These results indicate that $\hat{{\mathcal{T}}}$ agrees with the true ${\mathcal{T}}$ within $< 1\%$ accuracy on scales $k< 1{\, h/{\rm Mpc} }$ across redshift range $0<z<3$, regardless of the strength of baryonic effects.
• Figure 3: Validation of the field-level correction capability. We take the scattering transform as an example, which contains higher-order correlations and phase information. Left: Ratio of coefficients from $\delta_m$ in Illustris-1 simulation at $z=0.5$ to its DMO counterpart $\delta_{\rm DMO}$. Right: Ratio of coefficients from those with the ${\mathcal{T}}$ correction, $\hat{\delta}_{\rm DMO}({\bf x}) = \hat{{\mathcal{T}}}* \delta_m({\bf x})$, to $\delta_{\rm DMO}$. We measure the first two order coefficients, $S_{jl}$ and $S_{j_1l_1;j_2l_2}$, and set $j_1=j_2 \;\&\; l_1> l_2$ in $S_{j_1l_1;j_2l_2}$ to reduce data size cheng2020new. In each panel, different lines represent spatial index $j \in\{ 0, 1, 2, 3, 4, 5\}$, corresponding to the filter size $k_\sigma = 2\pi/\sigma_j$ in Fourier space. The horizontal ordinates represent angular indexes $\ell\in\{ 0, 1, 2, 3, 4 \}$, sorted as $\{ {\ell=0},\; {\ell=1},\; \cdots,\; (\ell_1,\ell_2)=(1, 0),\; (\ell_1,\ell_2)=(2, 0),\; \cdots \}$. The accurate removal of baryonic effects down to resolution $k_\sigma \lesssim 2{\, h/{\rm Mpc} }$ demonstrates the effectiveness of the $\hat{{\mathcal{T}}}$ correction at the field-level.
• Figure 4: Baryonic effects on the matter power spectrum modeled with the Zel'dovich approximation. We model the baryonic effect as an additional displacement ${\bf\Psi}_s({\bf k}) = B(k) {\bf\Psi}^Z({\bf k})$, where the window function is given by $B(k) = \alpha\, e^{- \left(k_l/k\right)^2 - \left(k/k_s\right)^2 }$, with $\alpha=-0.2$, $k_l = 0.5{\, h/{\rm Mpc} }$ and $k_s = 10{\, h/{\rm Mpc} }$. The small-scale power is further damped by a factor $e^{- \left(k/k_s\right)^2 }$ to localize the ${\bf\Psi}_s$ impacts in Fourier space. We perform calculations that evolve the DMO density field from $z=1$ to $z+\Delta z$ with/without the baryonic shift ${\bf\Psi}_s$, for two cases: $\Delta z=0.1$ (blue) and $\Delta z=0.5$ (red). Top panel: The adopted window function $B(k)$. Bottom panel: The resulted baryonic suppression of the matter power spectrum (dashed) and cross-correlation coefficients between $\delta_{\rm s}$ and $\delta_{\rm z}$ (solid). The results verify the conclusion that the matter clustering is sensitive to baryonic effects in amplitude than in phase under the Zel'dovich approximation.
• Figure 5: Baryonic suppression (left panel) and cross-correlation coefficient (right panel) measured in TNG300-1 (solid lines) and Illustris-1 (dashed lines) simulations. Different colors correspond to different redshifts, covering the range $0<z<3$.