[Paper Review] SDSS-III Baryon Oscillation Spectroscopic Survey: Analysis of Potential Systematics in Fitting of Baryon Acoustic Feature

M. Vargas-Magaña, Shirley Ho+13 more|arXiv (Cornell University)|Dec 17, 2013

Galaxies: Formation, Evolution, Phenomena6 references9 citations

TL;DR

This paper evaluates systematic errors in fitting the baryon acoustic oscillation (BAO) feature using anisotropic clustering in the SDSS-III BOSS survey. Using DR10 and DR11 data and mock galaxy catalogs, it demonstrates that the fiducial multipole fitting method is robust at the 0.1–0.2% level, with systematic uncertainties in distance measurements well below the 1% statistical error, ensuring reliability for precision cosmology.

ABSTRACT

Extraction of the Baryon Acoustic Oscillations (BAO) to percent level accuracy is challenging and demands an understanding of many potential systematic to an accuracy well below 1 per cent, in order ensure that they do not combine significantly when compared to statistical error of the BAO measurement. Sloan Digital Sky Survey (SDSS)-III Baryon Oscillation Spectroscopic Survey (BOSS) SDSS Data Release Eleven (DR11) reaches a distance measurement with $\sim 1\%$ statistical error and this prompts an extensive search for all possible sub-percent level systematic errors which could be safely ignored previously. In this paper, we analyze the potential systematics in BAO fitting methodology using mocks and data from BOSS DR10 and DR11. We demonstrate the robustness of the fiducial multipole fitting methodology to be at $0.1\%-0.2\%$ level with a wide range of tests in mock galaxy catalogs pre- and post-reconstruction. We also find the DR10 and DR11 data from BOSS to be robust against changes in methodology at similar level. This systematic error budget is incorporated into the the error budget of Baryon Oscillation Spectroscopic Survey (BOSS) DR10 and DR11 BAO measurements. Of the wide range of changes we have investigated, we find that when fitting pre-reconstructed data or mocks, the following changes have the largest effect on the best fit values of distance measurements both parallel and perpendicular to the line of sight: (a) Changes in non-linear correlation function template; (b) Changes in fitting range of the correlation function; (c) Changes to the non-linear damping model parameters. The priors applied do not matter in the estimates of the fitted errors as long as we restrict ourselves to physically meaningful fitting regions.[abridged]

Motivation & Objective

• To assess potential systematic errors in BAO fitting methodologies at sub-percent levels, critical for precision cosmology with BOSS DR10 and DR11.
• To test the robustness of the fiducial multipole fitting method for anisotropic clustering against variations in fitting parameters and model assumptions.
• To compare alternative methodologies—multipoles and clustering wedges—for consistency and bias in BAO distance scale estimation.
• To quantify systematic uncertainties in α∥ and α⊥ (parallel and perpendicular distance shifts) due to changes in correlation function templates, fitting ranges, and non-linear damping models.
• To incorporate the resulting systematic error budget into the final BAO measurements of BOSS DR10 and DR11, ensuring sub-1% total error control.

Proposed method

• Employs anisotropic clustering analysis using multipole decomposition of the galaxy correlation function, following the methodology of Xu et al. (2012), as the fiducial fitting approach.
• Applies both Markov Chain Monte Carlo (MCMC) and grid-based parameter estimation methods to assess numerical stability and convergence of the fitting process.
• Uses pre- and post-reconstruction mock galaxy catalogs from SDSS-III BOSS DR10 and DR11 to test sensitivity to reconstruction effects and model assumptions.
• Systematically varies key fitting components: non-linear correlation function templates, fitting ranges, and non-linear damping model parameters, while holding priors within physically meaningful bounds.
• Compares results from the multipole method with an alternative clustering wedges approach to evaluate consistency and potential bias across methodologies.
• Quantifies systematic errors in the best-fit α∥ and α⊥ parameters by computing median differences and dispersions across multiple test configurations.

Experimental results

Research questions

• RQ1 How robust is the fiducial multipole fitting method for BAO extraction in anisotropic clustering at the 0.1–0.2% level?
• RQ2 Which fitting parameters—non-linear template, fitting range, or damping model—have the largest impact on the measured α∥ and α⊥?
• RQ3 How do different numerical implementations (MCMC vs. grid) affect the recovered BAO distance scale and its uncertainty?
• RQ4 To what extent do the clustering wedges and multipole methods agree in their best-fit α values and errors?
• RQ5 What is the total systematic error budget for BAO distance measurements in BOSS DR10 and DR11, given variations in fitting methodology?

Key findings

• The fiducial multipole fitting method exhibits robustness at the 0.1–0.2% level in both α∥ and α⊥, with median differences between MCMC and grid methods below 0.3% for pre-reconstruction and below 0.1% for post-reconstruction.
• The fitted errors from MCMC and grid methods are well correlated, with dispersions of 0.2% in α∥ and 0.1% in α⊥ post-reconstruction, indicating high numerical stability.
• The clustering wedges and multipole methods show consistent results, with median differences of 0.7% in α∥ and 0.4% in α⊥ pre-reconstruction, reducing to 0.3% and 0.1% post-reconstruction.
• The differences in fitted errors between the two methodologies are small, with Δσα∥ = 0.8% and Δσα⊥ = 0.3% pre-reconstruction, decreasing to ≤0.5% post-reconstruction.
• The largest systematic effects on α come from changes in the non-linear correlation function template, fitting range, and non-linear damping model parameters, while priors do not affect results as long as they remain physically meaningful.
• The total systematic error budget is estimated at 0.3% (0.2% + 0.2% in quadrature), which is well within the 1% statistical error, confirming the method's reliability for current and near-future cosmological analyses.