Characterization of the visit-to-visit Stability of the GR700XD Wavelength Calibration for NIRISS/SOSS Observations

TL;DR

This study characterizes the stability of the wavelength calibration for NIRISS/SOSS observations on JWST, showing that calibration can vary predictably between visits and can be accurately modeled using polynomial regressions.

Contribution

The paper introduces a model-based approach to predict wavelength solution variations for NIRISS/SOSS, improving calibration accuracy across visits.

Findings

Wavelength solutions can fluctuate by a few pixels between visits.

Polynomial regression models predict wavelength solutions with RMS errors within a few tenths of a pixel.

The models are implemented in the PASTASOSS Python package for practical use.

Abstract

When utilizing the NIRISS/SOSS mode on JWST, the pupil wheel (tasked with orienting the GR700XD grism into the optical path) does not consistently settle into its commanded position resulting in a minor misalignment with deviations of a few fractions of a degree. These small offsets subsequently introduce noticeable changes in the trace positions of the NIRISS SOSS spectral orders between visits. This inconsistency, in turn, can lead to variations of the wavelength solution. In this report, we present the visit-to-visit characterization of the NIRISS GR700XD Wavelength Calibration for spectral orders 1 and 2. Employing data from Calibration Program 1512 (PI: Espinoza), which intentionally and randomly sampled assorted pupil wheel positions during observations of the A-star BD+60-1753, as well as data from preceding commissioning and calibration activities to model this effect, we demonstrate that the wavelength solution can fluctuate in a predictable fashion between visits by up to a few pixels. We show that via two independent polynomial regression models for spectral orders 1 and 2, respectively, using the measured x-pixel positions of known Hydrogen absorption features in the A-star spectra and pupil wheel positions as regressors, we can accurately predict the wavelength solution for a particular visit with an RMS error within a few tenths of a pixel. We incorporate these models in PASTASOSS, a Python package for predicting the GR700XD spectral traces, which now allows to accurately predict spectral trace positions and their associated wavelengths for any NIRISS/SOSS observation.