AN ATLAS OF BRIGHT STAR SPECTRA IN THE NEAR-INFRARED FROM CASSINI-VIMS

The recovery of high-quality stellar spectra underpins an integral part of our understanding of stellar composition, evolution, and temporal behavior. Unfortunately, important molecular spectral features expected to be present are masked by the existence of the same molecular species in our own atmosphere. This is especially true for cooler evolved stars, where substantial quantities of H₂O and CO₂ are known to form. This results in large parts of the infrared region being effectively opaque to ground-based observations. Piercing this veil is only possible by observing beyond the Earth's atmosphere from space-based platforms.

There has yet to be a space-based spectrometer dedicated to surveying bright stars over a broad range of infrared wavelengths, despite the potential scientific merit and the severe obstacles to observing from the ground. However, the Visual and Infrared Mapping Spectrometer (VIMS) instrument on the Cassini spacecraft has measured the spectra of many bright stars in the wavelength range of 0.8–5.1 μm. This broad wavelength range is distinct from any other existing space-based observatory platform. The International Ultraviolet Explorer (IUE) provides a survey of spectra shorter than this range, spanning 0.11–0.32 μm (Boggess et al. 1978). The longer wavelengths overlap with surveys by the Short Wave Spectrometer (SWS) on ESA's Infrared Space Observatory (ISO) covering 2.38–45.2 μm (de Graauw et al. 1996). The Space Telescope Imaging Spectrograph (STIS) on the Hubble Space Telescope (HST) can observe spectra between 0.11 and 1.0 μm (Woodgate et al. 1998), while the Near Infrared Camera and Multi-Object Spectrometer (NICMOS) on HST was capable of observing the remaining 1.0–2.5 μm range at the shorter end of the infrared  (Thompson et al. 1998). However, these HST instruments have not been used for survey purposes and therefore only specifically targeted objects have been observed. Photometry in the H, K, L, and M bands was acquired for 92 of the brightest stars by the Diffuse Infrared Background Experiment (DIRBE) on COBE (Arendt et al. 1998). A much more extensive catalog of ground based photometric observations was produced from 2MASS observations in the J, H, and Ks bands (Skrutskie et al. 2006). Observations at longer wavelengths have been performed by SPITZER (>5.3 μm), MSX (>8.3 μm), IRAS (>12 μm), AKARI (>50 μm), and Herschel (>55 μm). None of these instruments span the entire range accessible to VIMS and many of them have been unable to observe the brightest stars as they are beyond the instrumental saturation limit. VIMS has previously been used for stellar studies using stellar occultations by Saturn's rings (Stewart et al. 2013, 2014, 2015a, 2015b, 2015c), but until now has not been used for studies of stellar spectra.

The atlas provides unique access to broadband, space-based spectra of many of the brightest stars in the sky, albeit at a relatively low spectral resolution. This paper starts with a description of the experiment and instrument in Section 2, followed by details on the data reduction processes in Section 3. Sample spectra are presented in Section 4, and compared with those obtained by ISO and IRTF, and photometry from DIRBE and 2MASS. The full list of stars in the atlas is provided in appendix and reduced spectra are available online at http://www.physics.usyd.edu.au/sifa/caoss/ and Vizier. By making these spectra available, we anticipate scientific benefits beyond the goals of the Cassini mission, and hope to enable new sections of the astronomical community to gain directly from a planetary exploration mission.

The observations were originally acquired for the purposes of monitoring for any temporal variations within the VIMS instrument and to provide a baseline for the interpretation of stellar occultation results. Their acquisition started in the early stages of the mission, while the spacecraft was in "cruise" and on its way to Saturn, with the program expected to continue through to the end of the mission. The observations from this monitoring program were not intended to be used for stellar science. However, they fill a void in spectral coverage for a sufficient number of bright stars so as to warrant dissemination to the wider astronomical community.

2.1. The VIMS Instrument

The observations were obtained from the data archive at the Planetary Rings Node of NASA's Planetary Data System (PDS; NASA PDS 2015a). The raw format stores the intensity across the spectrum in background-subtracted detector counts together with the background itself. The parameters of the observation and instrumental operation are recorded in the data file headers.

These observations were performed in the raster-scanning mode described above and stored as three-dimensional (3D) image "cubes." The full 0.8–5.1 μm spectrum was obtained simultaneously for each signal pixel individually, with typical integration times of 320 or 640 ms per pixel. The images recorded by VIMS usually cover a 8 × 8 or 12 × 12 pixel field in all 352 spectral channels using measurements from both VIMS-VIS and VIMS-IR, but only the 256 channels recorded by the infrared instrument are used here. From the recorded data (Figure 1(a)), VIMS subtracts the measured background (Figure 1(b)) before transmitting them both back to Earth. The first step in the extraction of the stellar spectra from these cubes is to sum over a 3 × 3 pixel square, centered on the brightest pixel, producing a base spectral measurement as shown in Figure 1(c). In order to determine if any spectral channels saturated during the exposure, the measured and separately recorded background spectrum (Figure 1(b)) is added to the brightest pixel's counts (Figure 1(a)). Any spectral channels which now show 4095 counts are flagged as saturated and are omitted. These occur frequently at the hot pixels shown as spikes in Figure 1(b), such as the saturated pixel at 1.25 μm in Figure 1(a). Similarly, pixels affected by cosmic rays are identified and flagged for omission. These base spectra are divided by the geometric area of the telescope's aperture and the exposure time of the observation (extracted from data headers) to give the measured spectra in counts m⁻² s⁻¹. The instrumental sensitivity function (Figure 1(d)), combining both the electronic gain and the detector sensitivity as a function of wavelength, is used to convert this into photons m⁻² s⁻¹ (Figure 1(e)). By dividing by the instrumental bandpass shown in Figure 1(f) (Δλ), we arrive at a spectrum in photons m⁻² s⁻¹ μm⁻¹ (Figure 1(g)) which is multiplied by the photon energy (Figure 1(h)) to give J m⁻² s⁻¹ μm⁻¹. Finally this is converted into Janskys as shown in Figure 1(i) ($={10}^{-26}\;{\mathrm{Wm}}^{-2}\;{\mathrm{Hz}}^{-1}$) in order to compare with previously published spectra from the literature.

Zoom In Zoom Out Reset image size

The sensitivity and bandpass functions were derived as part of the VIMS-IR calibration for extended sources. Clark et al. (2012) detail the development of these functions through to the latest version (RC17) which is used in this work. For stellar sources which do not fill the spectrometer's pixel, the sensitivity function is expected to differ somewhat from these standard values. A wavelength-dependant systematic deviation between the recovered VIMS spectra and those previously published was noted in this work, and was found to be consistent across a range of stars of various types and spanning multiple epochs. A correction ratio was constructed based on a comparison of VIMS spectra for standard stars to corresponding literature spectra from Engelke et al. (2006) and Rayner et al. (2009). The ratios of sections of the literature spectra to VIMS observations were averaged to produce a mean brightness ratio by which VIMS observations were multiplied. Variable stars known to exhibit strong wavelength-dependent temporal variations which have the potential to introduce spectral artefacts, and those with an amplitude of greater than one magnitude in any part of the near-infrared have been omitted. These relevant reference spectra are summarized in Table 1. The final correction function is available online with the atlas, and is provided to enable the application of alternative calibrations.

Figure 2 demonstrates the effectiveness of this correction as applied to α Boo from sequence S16. The resultant corrected spectrum provides a good match to the general shape of the literature data, as well as some of the finer details, especially through 2.2–2.6 μm and the narrow absorption feature near 1.1 μm. For comparison purposes, the literature spectra have been downsampled to match the spectral resolution of VIMS. In order to to ensure the robustness of this comparison we also produced a calibration curve omitting the "test" epoch and found the resultant curve to be indistinguishable from that produced using all observations which have literature spectra. Application of this correction function is only necessary when observing point-sources, and calibrating filled-pixel spectra does not require this step.

Zoom In Zoom Out Reset image size

The spectra are presented without any correction for interstellar reddening for the following reasons. The majority of sources in the atlas are close to the Sun and thus have ${A}_{{\rm{V}}}$ values of approximately zero, with half the sample within 110 pc and only 10 stars further than 500 pc. Extinction corrections are much smaller in the infrared than in the optical, reducing the impact of interstellar reddening on stars in the atlas. It is not envisioned that the calibration pipeline presented here will change, whereas extinction corrections, dependent on poorly known distances and estimated extinctions to that distance along a given line of sight, will almost certainly change in the future. This is especially likely as 3D extinction models of the Galaxy improve over the coming decades.

3.1. Data Products

The spectra produced in this work include stars from many spectral classes including A and G main-sequence stars and K giants but, due to engineering requirements for bright targets, the atlas is dominated by evolved M giants. Many of these targets have complex molecular atmospheres and exhibit variability. This section shows some of the recovered spectra and compares them, where possible, to existing literature measurements. These literature values come predominately from space-based ISO SWS spectra (Engelke et al. 2006) and ground-based IRTF spectra (Rayner et al. 2009), with space-based photometry from DIRBE (Arendt et al. 1998). WISE photometry was found to be unreliable for many of these targets due to unflagged saturation. For accurate comparison, both the ISO and IRTF spectra have been downsampled to match the spectral resolution of VIMS.

To give a sample of the CAOSS data, Figure 3 shows reduced spectra for five different stellar targets. Each panel labelled (a) to (e), contains an observation at a single epoch for a single star. Panel (a) shows the A1 star Sirius (α Canis Majoris) with a spectrum comprising predominately the Rayleigh–Jeans side of a hot Planck curve (Sirius has an effective surface temperature of approximately 10,000 K with peak intensity in the ultraviolet). This shows very good agreement with the Engelke et al. (2006) spectrum, as well as both the DIRBE and 2MASS photometry. Panel (b) contains the blended spectrum of the binary system, α Centauri A and B and also shows good agreement to the 2MASS photometry. Panels (c) and (d) contain the early-mid M giants ρ Persei and γ Crucis. These spectra have broadly similar shapes, yet very different magnitudes due to the latter's relative proximity. They agree well with existing literature spectra and photometry from all four sources. The final panel (e), shows R Cnc, a Mira variable-type star with a spectral class of M6.5–9. Such stars are known to exhibit huge variations in magnitude over a period on the order of a year. The spectrum presented only partly agree with the published Rayner et al. (2009) IRTF spectra and 2MASS photometry, which in keeping with the known spectral variability of such objects.

Zoom In Zoom Out Reset image size

Figure 4 shows two more Mira variable stars for which there are multiple epochs presented in this atlas. Large changes in overall flux, accompanied by more minor shifts in spectral shape are noted between epochs. The variations in flux for these particular stars, as with all variable stars in the atlas, have been checked against the AAVSO lightcurves and found to be broadly consistent, but no detailed analysis of stellar variability has been performed in this work.

Zoom In Zoom Out Reset image size

We have produced a database of space-based near-infrared spectra for 73 stellar targets at 109 epochs. These spectra have been made publicly available online in flexible data formats. Where comparison was possible they have been shown to be consistent with existing spectra and photometry. In spite of the relatively low spectral resolution, we believe that these spectra are of sufficient quality to continuously bridge a vital yet unsurveyed part of the spectrum and to become a valuable resource for the stellar astrophysical community. As spectra from this data set span the telluric water bands they particularly enable the study of evolved stars, known to exhibit strong spectral structure from their own atmospheric water. Spectra presented in this atlas have already been used in a study of the atmosphere of Mira, including an assessment of models of its atmospheric behavior (Stewart et al. 2015c).

This research made use of the SIMBAD database, operated at CDS, Strasbourg, France.

Some of the data presented in this paper were obtained from the Mikulski Archive for Space Telescopes (MAST). STScI is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555. Support for MAST for non-HST data is provided by the NASA Office of Space Science via grant NNX13AC07G and by other grants and contracts.

This publication makes use of data products from the Two Micron All Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation.

The details of all stars contained in CAOSS are listed in Table 3, including coordinates, spectral type and variability type where relevant. A full list of flux density calibrated spectra to date is contained in Table 4. Observations processed identically to those in Table 4 yet containing significant deviations from the literature values are presented online with the atlas, but not listed here. These spectra are still expected to be of value in situations where an absolute flux density calibration is not required.