Wavelength Accuracy of the Keck HIRES Spectrograph and Measuring Changes in the Fine Structure Constant

TL;DR

This study assesses the wavelength calibration accuracy of the Keck HIRES spectrograph and its implications for measuring potential variations in the fine structure constant, revealing significant systematic uncertainties.

Contribution

The paper introduces new calibration techniques using iodine cells and highlights the impact of calibration errors on constraining the fine structure constant's variability.

Findings

Wavelength offsets of 500-1000 m/s were found in Keck HIRES data.

Calibration drifts over nights can reach nearly 2000 m/s.

Systematic uncertainties may hinder precise measurements of the fine structure constant.

Abstract

We report on an attempt to accurately wavelength calibrate four nights of data taken with the Keck HIRES spectrograph on QSO PHL957, for the purpose of determining whether the fine structure constant was different in the past. Using new software and techniques, we measured the redshifts of various Ni II, Fe II, Si II, etc. lines in a damped Ly-alpha system at z=2.309. Roughly half the data was taken through the Keck iodine cell which contains thousands of well calibrated iodine lines. Using these iodine exposures to calibrate the normal Th-Ar Keck data pipeline output we found absolute wavelength offsets of 500 m/s to 1000 m/s with drifts of more than 500 m/s over a single night, and drifts of nearly 2000 m/s over several nights. These offsets correspond to an absolute redshift of uncertainty of about Delta z=10^{-5} (Delta lambda= 0.02 Ang), with daily drifts of around Delta z=5x10^{-6} (Delta lambda =0.01 Ang), and multiday drifts of nearly Delta z=2x10^{-5} (0.04 Ang). The causes of the wavelength offsets are not known, but since claimed shifts in the fine structure constant would result in velocity shifts of less than 100 m/s, this level of systematic uncertainty makes may make it difficult to use Keck HIRES data to constrain the change in the fine structure constant. Using our calibrated data, we applied both our own fitting software and standard fitting software to measure (Delta alpha)/alpha, but discovered that we could obtain results ranging from significant detection of either sign, to strong null limits, depending upon which sets of lines and which fitting method was used. We thus speculate that the discrepant results on (Delta alpha)/alpha reported in the literature may be due to random fluctuations coming from under-estimated systematic errors in wavelength calibration and fitting procedure.