An Improved Fit for Linear Halo Bias at High Redshift

1. Problem Statement

• Context: Understanding the clustering of dark matter halos is fundamental to interpreting galaxy formation and cosmological models. The "halo bias" ($b$) quantifies how halos trace the underlying dark matter distribution.
• The Gap: While halo bias is well-calibrated at low redshifts ($z \lesssim 4$) using extensive N-body simulations and analytic fits (e.g., Tinker et al. 2010, or "T10"), it remains poorly constrained at high redshifts ($z \gtrsim 4$).
• The Issue: Current and upcoming surveys (JWST, Roman, HERA, SKA) are reaching percent-level precision in measuring galaxy clustering at $z = 6 - 19$. However, existing theoretical models are often extrapolated from low-$z$ calibrations without direct validation. High-$z$ halos form in a less evolved, more linear density field, potentially causing low-$z$ fits to fail.
• Specific Discrepancy: Preliminary studies have yielded diverse and conflicting results for high-$z$ bias, leaving a critical systematic uncertainty in the interpretation of early-Universe data.

2. Methodology

The authors utilized a high-fidelity simulation suite and rigorous statistical techniques to measure bias directly at high redshifts.

• Simulation Suite (GUREFT):
  • Used the GUREFT suite, consisting of four periodic boxes with varying resolutions (side lengths: 5, 15, 35, and 90 $h^{-1}$ Mpc) and $1024^3$ particles each.
  • Dynamic Range: The smallest box resolves halos down to $\sim 10^6 M_\odot$, crucial for high-$z$ studies where typical galaxy-hosting halos are less massive.
  • Cosmology: Flat $\Lambda$CDM with $\Omega_m = 0.307$, $\sigma_8 = 0.829$, $H_0 = 67.8$ km s$^{-1}$ Mpc$^{-1}$.
• Halo Identification:
  • Employed ROCKSTAR, a halo finder using adaptive hierarchical refinement in 6D phase space (position + velocity). This is critical at high $z$ where halos are less virialized and mergers are frequent, ensuring robust separation of bound structures.
  • Analyzed only host halos (excluding subhalos) to avoid contamination from backsplash effects.
• Bias Measurement:
  • Calculated the linear halo bias ($b_{lin}$) using two-point correlation functions: $\xi_{hh}$ (halo-halo), $\xi_{hm}$ (halo-matter), and $\xi_{mm}$ (matter-matter).
  • Used the Landy-Szalay estimator and the pycorr package.
  • Defined bias as $b_h = \xi_{hm} / \xi_{mm}$ (cross-correlation preferred over auto-correlation to minimize shot noise).
  • Scale-Dependence Mitigation: Measured the scale-dependent bias $b_h(r)$ and fitted it to the form $b_h(r) = b_{lin} + c_1/r$ to extract the scale-independent linear bias $b_{lin}$. Fitting was restricted to scales between $\sim 30 R_{vir}$ and $1/6$ of the box size to avoid non-linear effects and finite-volume suppression.
• Peak Height ($\nu$):
  • Bias was parameterized as a function of effective peak height $\nu_{eff} = \delta_c / \sigma(M, z)$.
  • Applied a correction to the power spectrum (truncating at $k_{box}$) to account for finite simulation volume effects, ensuring consistency across different box sizes.

3. Key Contributions & Results

• Measurement of High-$z$ Bias:
  • Measured halo bias for $z = 6 - 19$ across a peak height range of $1.31 \leq \nu \leq 3.88$ ($0.12 \leq \log_{10}\nu \leq 0.59$).
  • Found that high-$z$ halos are $\sim 3 - 4\%$ more biased than predicted by the canonical T10 fit (calibrated at $z \sim 0-2.5$).
  • The discrepancy is most pronounced at intermediate peak heights ($\log_{10}\nu \approx 0.2 - 0.5$).
• Updated Fitting Function:
  • The authors proposed an updated fit within the Tinker et al. (2010) functional framework.
  • By allowing parameters $B, C, a, b, c$ to vary (while keeping the constraint that the bias-weighted mass function integrates to unity), they derived a new formula (Equation 8):
    $$b_{lin} = 1 - \frac{\nu^{0.08059}}{\nu^{0.08059} + \delta_c^{0.08059}} + 0.4499\nu^{2.138}$$
  • Performance: This new fit reduces the mean systematic offset from the simulation measurements to $< 1\%$ (specifically $\sim -0.7\%$) and lowers the $\chi^2$ value from 81.5 (T10) to 26.9.
• Validation:
  • The results show no strong explicit redshift dependence in the $b_{lin}$-$\nu$ relation beyond the shift in $\nu$ values caused by cosmic evolution.
  • The updated fit is valid for $6 \leq z \leq 19$ and $0.12 \leq \log_{10}\nu \leq 0.59$.

4. Significance

• Enabling Precision Cosmology: As surveys like JWST and Roman achieve sub-percent precision in measuring galaxy clustering at high redshifts, theoretical uncertainties in halo bias must be equally small. The previous $\sim 4\%$ offset in standard models would dominate the error budget.
• Galaxy Formation Models: Accurate bias measurements allow for tighter constraints on the relationship between galaxies and their host halos (e.g., Halo Occupation Distribution models) during the "Cosmic Dawn."
• Methodological Rigor: The paper highlights the importance of using phase-space halo finders (ROCKSTAR) and correcting for finite-volume effects in small boxes, which are often sources of systematic error in previous high-$z$ studies.
• Future Outlook: The authors note that while this fit improves the situation, future work is needed to address assembly bias (dependence on formation history/environment) and to extend these measurements to even higher masses and different cosmologies.

In summary, this work provides the first robust, simulation-calibrated linear bias fit for the epoch of reionization and cosmic dawn, reducing theoretical systematic errors to a level compatible with next-generation observational capabilities.