Wavelength Calibration of the VLT-UVES Spectrograph

TL;DR

This study investigates wavelength calibration errors in the VLT-UVES spectrograph using iodine cell spectra, identifying systematic shifts that could impact measurements of the fine-structure constant's variation over cosmic time.

Contribution

It provides a detailed analysis of calibration shifts in UVES, quantifies their potential impact on alpha variation measurements, and compares these errors to those found in other spectrographs.

Findings

Average velocity shifts of 100-500 m/s per exposure.

Intra-order distortions of 100-200 m/s within exposures.

Calibration errors can cause systematic uncertainties in alpha variation measurements.

Abstract

We attempt to measure possible miscalibration of the wavelength scale of the VLT-UVES spectrograph. We take spectra of QSO HE0515-4414 through the UVES iodine cell which contains thousands of well-calibrated iodine lines and compare these lines to the wavelength scale from the standard thorium-argon pipeline calibration. Analyzing three exposures of this z = 1.71 QSO, we find two distinct types of calibration shifts needed to correct the Th/Ar wavelength scale. First, there is an overall average velocity shift of between 100 m/s and 500 m/s depending upon the exposure. Second, within a given exposure, we find intra-order velocity distortions of 100 m/s up to more than 200 m/s. These calibration errors are similar to, but smaller than, those found earlier in the Keck HIRES spectrometer. We discuss the possible origins of these two types of miscalibration. We also explore the implications of these calibration errors on the systematic error in measurements of the relative change in alpha (current value - past value) / current value, the change in the fine-structure constant derived from accurate measurement of the relative redshifts of absorption lines in QSO absorption systems. The overall average, exposure-dependent shifts should be less relevant for fine-structure work, but the intra-order shifts have the potential to affect these results. Using either our measured calibration offsets or a Gaussian model with sigma of around 90 m/s, Monte Carlo mock experiments find errors in the relative change in alpha of between 1e-6 Nsys^(-1/2) and 3e-6 Nsys^(-1/2), where Nsys is the number of systems used and the range is due to dependence on how many metallic absorption lines in each system are compared.