Keck constraints on a varying fine-structure constant: wavelength calibration errors

TL;DR

This study assesses how wavelength calibration errors in the Keck/HIRES spectrograph could influence measurements of the fine-structure constant's variability, finding the effect to be small but emphasizing the need for careful calibration.

Contribution

It investigates the impact of intra-order wavelength distortions on previous measurements of alpha variation, providing a correction estimate and highlighting calibration challenges.

Findings

Calibration distortions cause a small (~0.4x10^{-5}) random effect per absorber.

Applying corrections changes the mean alpha variation by less than 0.05x10^{-5}.

High-precision measurements require detailed understanding of instrumental distortions.

Abstract

The Keck telescope's HIRES spectrograph has previously provided evidence for a smaller fine-structure constant, alpha, compared to the current laboratory value, in a sample of 143 quasar absorption systems: da/a=(-0.57+/-0.11)x10^{-5}. This was based on a variety of metal-ion transitions which, if alpha varies, experience different relative velocity shifts. This result is yet to be robustly contradicted, or confirmed, by measurements on other telescopes and spectrographs; it remains crucial to do so. It is also important to consider new possible instrumental systematic effects which may explain the Keck/HIRES results. Griest et al. (2009, arXiv:0904.4725v1) recently identified distortions in the echelle order wavelength scales of HIRES with typical amplitudes +/-250m/s. Here we investigate the effect such distortions may have had on the Keck/HIRES varying alpha results. We demonstrate that they cause a random effect on da/a from absorber to absorber because the systems are at different redshifts, placing the relevant absorption lines at different positions in different echelle orders. The typical magnitude of the effect on da/a is ~0.4x10^{-5} per absorber which, compared to the median error on da/a in the sample, ~1.9x10^{-5}, is relatively small. Consequently, the weighted mean value changes by less than 0.05x10^{-5} if the corrections we calculate are applied. Nevertheless, we urge caution, particularly for analyses aiming to achieve high precision da/a measurements on individual systems or small samples, that a much more detailed understanding of such intra-order distortions and their dependence on observational parameters is important if they are to be avoided or modelled reliably. [Abridged]