Wavelength Solution

During the day, we take a series of ThAr arc lamp exposures through both sets of fibers to use for wavelength calibration. These are stored in the archive under OBSTYPE=DOUBLE.

BANZAI-NRES stacks these frames and calibrates the wavelength solution using xwavecal as described in Brandt et al. 2020 DOI: 10.3847/1538-3881/ab929c.

Line List

NOTE: NRES wavelength calibrations are in vacuum wavelengths.

Our line list is the ThAr atlas from ESO which was fetched from http://www.eso.org/sci/facilities/paranal/instruments/uves/tools/tharatlas.html on 27 August 2020. We converted the air wavelengths to vacuum wavelengths using the original 1966 Edlen Equation (Murphy et al. 2007, DOI 10.1111/j.1365-2966.2007.11768.x).

Wavelength calibration metrics

BANZAI-NRES calculates metrics characterizing the quality of the wavelength solution. The metrics are only calculated on features that are matched to lines in the catalog. Features are matched to the closest feature in the line list and then outliers/non-matches are rejected using sigma clipping. The header keywords that contain the metrics and rule-of-thumb values for good wavelength calibrations are included in the list below. Note that these header keywords are only included in the stacked ThAr super-calibration frames.

  • ‘SIGLAM’: Standard deviation of measured wavelengths - line list wavelengths in Angstroms. Typically this should be between 0.003 and 0.01 Angstroms for NRES.

  • ‘RVPRECSN’: Standard deviation of measured wavelengths - line list wavelengths converted to velocity units via \(\frac{\Delta \lambda}{\lambda} = \frac{\Delta v}{c}\). This metric has the advantage that it is normalized accross the full wavelength range and is in the same units as the radial velocity. For NRES, this is typically between 4 and 10 m/s.

  • ‘WAVRCHI2’: Reduced \(\chi^2\) of the line list wavelengths - wavelength model of the detected features. This should be close to one. Much larger than one would suggest the wavelength solution is not describing the true dispersion well. Much less than one would suggest that error bars on the centroids of the detected features are over-estimated.

  • ‘NLINEDET’: Number of features detected on the frame. For NRES, this is typically ~4000.

  • ‘NLINES’: Number of matched lines to the line list. This should be ~75% of the features detected on the frame. If this is less than 1000, blind solves of the wavelength solution will likely fail.

To check for a failed wavelength solution, a good test is to compare adjacent orders and ensure the features line up in regions that the wavelengths overlap.