When échelle observations are completed, the data are sent by FTP to a Sun workstation. There, the reduction of the data is done in a semi-automated manner by means of an IRAF1 script. Processing consists of bias subtraction, aperture extraction, flat fielding (in that order), and wavelength calibration by means of a Th-Ar lamp. In almost all cases, the calibration solution is linearly interpolated in time between the solutions for lamp exposures taken before and after the stellar spectrum. The data product is formatted in FITS. Header keywords for heliocentric correction are supplied, but the correction is not done.The reduction procedure is described in detail in a 1996 M. S. thesis by Christopher L. Mulliss.
In fall 2011, reprocessing of spectra taken between about mid-2002 and March 2007, when camera CCD1 was taken out of service, was begun by the author, Nancy Morrison. Initially, I used IRAF 2.15.1a in the 32-bit version (macosx) on a MacBook Pro. Since mid-2015, I have been using the 64-bit version, IRAF 2.16.1. Spectra in the public archive will be updated as the reprocessing is done.
The main goal is to incorporate the improved wavelength calibration developed for the larger-format camera, CCD3. The old calibration became suspect after 2002 March because a new Th-Ar lamp was installed, and its spectrum turned out to be so different from the old one that line identification became difficult. The new line identifications are summarized here. In that atlas are only 29 lines within the range of CCD1, not enough for proper constraint of a two-dimensional polynomial fit as performed by ecidentify. Therefore, a new Th-Ar line list has been constructed for CCD1 with additional lines from several sources: the NIST database of atomic spectra; the Las Campanas Observatory Th-Ar atlas for MIKE; and the on-line supplement to Murphy, M. T. et al. 2007, MNRAS, 378, 221. The final selection of 53 lines, with identifications, is given here. If no source is listed, the line comes from the original NOAO list.The new calibration has been tested for CCD1 on the radial-velocity standards beta Gem, alpha Cas, and epsilon Leo, with the most detailed test being done on beta Gem. Radial velocities were measured for a list of lines taken from the Arcturus atlas2 (Hinkle et al. 2000) with wavelengths extended to three decimal places by reference to the NIST database of atomic spectra.
|Order||Wavelengths (Å)||Mean RV||St. Dev.|
|42||5280 - 5340||3.56||0.49|
|41||5410 - 5470||3.66||0.57|
|40||5550 - 5610||3.33||0.35|
|39||5690 - 5750||3.46||0.53|
|38||5840 - 5900||3.55||0.36|
|37||6000 - 6060||3.47||0.46|
|36||6160 - 6230||3.59||0.37|
|35||6340 - 6410||3.45||0.58|
|34||6530 - 6600||3.76||0.70|
Additional improvements in the results of the reprocessing procedure:
Note added March 2, 2012: When the previous sentence was written, I was using the 'fit1d' option for determining the profile function in the process of extracting the stellar spectra. This procedure is apparently not robust against cosmic rays when the signal-to-noise ratio is low. I am now experimenting with the two-dimensional fitting option, 'fit2d', and it seems to be working better. If I continue to like it, I'll reprocess again.Note added July 24, 2012: The 'fit2d' option sometimes causes significant distortions in the continuum shape in response to a major cosmic ray hit. In such cases, I am providing both the 'fit1d' and the 'fit2d' product. Otherwise, I am using 'fit2d' only for weak exposures with exposure times longer than about 30 minutes.
A reduction pipeline for the large-format camera, CCD3, is still under development.
The Roper Scientific camera's images are saved in TIFF format and converted to FITS by means of the Mac OS X version of ImageMagick. Then, the data are processed with IRAF1. An average dark frame with the same integration time as the stellar spectrum, or interpolated to the same exposure time, is subtracted for first-order removal of stray light. Then, standard background subtraction is done. The stellar spectrum is divided by the background-subtracted flat lamp, which is set equal to 1 below a fairly high threshold, and then extracted with variance weighting. Because of strong vignetting and optical distortion at the ends of the spectrum, the signal-to-noise ratio is inferior there. Wavelength calibration utilizes a Ne lamp (in the red) or an Fe-Ar lamp (in the yellow) and is carried out by standard methods. The reduced data are saved in FITS format, and header keywords giving heliocentric velocity correction, heliocentric julian date, and airmass are provided.
Correction for telluric lines is the responsibility of the user of the data, but a small collection of artificial telluric spectra is provided in the on-line data archive for both spectrographs. The templates are created from spectra of rapidly rotating A- and B-type stars taken at high airmass and/or high humidity.
All the telluric lines in the spectrum of the standard star are fitted with Gaussian or Voigt functions with the IRAF task
splot. Then, the parameters of the fitted functions are read into the IRAF task
mk1dspec and superimposed on a flat, unit continuum. Since the templates are noise-free, noise is not introduced when the stellar spectrum is divided by them.
Telluric line removal can be done with the IRAF task
telluric or by any other suitable method. Because
telluric scales the template to secure a least-squares best fit to the data, it is not necessary for the standard star to have been observed on the same night or at the same airmass as the program star, as long as the telluric lines are not saturated. Saturation is very rare in the Hα region, although it occurs in the Na D line region.
The telluric standard spectra themselves are also in the archive, so that users can make their own templates if desired.
2Hinkle, K., Wallace, L., Valenti, J., and Harmer, D. 2000, Visible and Near Infrared Atlas of the Arcturus Spectrum, 3727--9300 Å (San Francisco: Astronomical Society of the Pacific)
3Eaton, J. A. and Williamson, M. H. 2007, PASP, 119, 886