Mean-flux Regulated PCA Continuum Fitting of SDSS Lyman-alpha Forest Spectra

TL;DR

This paper introduces a new PCA-based continuum fitting method, MF-PCA, that significantly improves accuracy in noisy SDSS Lyman-alpha forest spectra, enabling better large-scale flux power spectrum measurements.

Contribution

The paper presents MF-PCA, a novel continuum fitting technique that combines PCA with mean-flux regulation, reducing errors and extending measurement scales in low S/N spectra.

Findings

Reduces continuum fitting errors to below 8% RMS in S/N ~ 2 spectra.

Decreases residual Fourier power, allowing larger scale flux power spectrum analysis.

Provides continuum fits for over 12,000 SDSS spectra, aiding community research.

Abstract

Continuum fitting is an important aspect of Lyman-alpha forest science, since errors in the estimated optical depths scale with the fractional continuum error. However, traditional methods of estimating continua in noisy and moderate-resolution spectra (S/N < 10 pixel^-1 and R ~ 2000, respectively, such as SDSS) using power-law extrapolation or the mean spectrum, achieve no better than ~ 10-15% RMS accuracy. To improve on this, we introduce mean-flux regulated/principal component analysis (MF-PCA) continuum fitting. In this technique, PCA fitting is carried out redwards of the quasar Lyman-alpha line in order to provide a prediction for the shape of the Lyman-alpha forest continuum. The slope and amplitude of this continuum prediction is then corrected using external constraints for the Lyman-alpha forest mean-flux. From tests on mock spectra, we find that MF-PCA reduces the errors to 8% RMS in S/N ~ 2 spectra, and < 5% RMS in spectra with S/N > 5. The residual Fourier power in the continuum is decreased by a factor of a few in comparison with dividing by the mean continuum, enabling Lyman-alpha flux power spectrum measurements to be extended to ~2x larger scales. Using this new technique, we make available continuum fits for 12,069 z>2.3 Lyman-alpha forest spectra from SDSS DR7 for use by the community. This technique is also applicable to future releases of the ongoing BOSS survey, which is obtaining spectra for ~ 150,000 Lyman-alpha forest spectra at low signal-to-noise (S/N ~ 2).