Fiber scrambling for high-resolution spectrographs. II. A double fiber scrambler for Keck Observatory

TL;DR

This paper presents a double fiber scrambler prototype for Keck HIRES that significantly stabilizes spectral line spread functions, reducing radial velocity scatter and improving measurement precision for high-resolution spectrographs.

Contribution

The study introduces a novel double fiber scrambler with pupil slicing that enhances spectral stability and radial velocity precision in high-resolution spectrographs.

Findings

Spectral line spread function is 18 times more stable with the double scrambler.

Radial velocity scatter reduced from 2.1 m/s to 1.5 m/s using the stabilized SLSF.

Inaccurate SLSF modeling limits measurement precision more than stellar jitter.

Abstract

We have designed a fiber scrambler as a prototype for the Keck HIRES spectrograph, using double scrambling to stabilize illumination of the spectrometer and a pupil slicer to increase spectral resolution to R = 70,000 with minimal slit losses. We find that the spectral line spread function (SLSF) for the double scrambler observations is 18 times more stable than the SLSF for comparable slit observations and 9 times more stable than the SLSF for a single fiber scrambler that we tested in 2010. For the double scrambler test data, we further reduced the radial velocity scatter from an average of 2.1 m/s to 1.5 m/s after adopting a median description of the stabilized SLSF in our Doppler model. This demonstrates that inaccuracies in modeling the SLSF contribute to the velocity RMS. Imperfect knowledge of the SLSF, rather than stellar jitter, sets the precision floor for chromospherically quiet stars analyzed with the iodine technique using Keck HIRES and other slit-fed spectrometers. It is increasingly common practice for astronomers to scale stellar noise in quadrature with formal errors such that their Keplerian model yields a chi-squared fit of 1.0. When this is done, errors from inaccurate modeling of the SLSF (and perhaps from other sources) are attributed to the star and the floor of the stellar noise is overestimated.