Spectral Energy Distributions of Southern Binary X-Ray Sources

The data were taken with the Australia Telescope Compact Array in the 6C configuration, that has 15 East-West baselines with lengths from 150 m to 6 km, giving angular resolution of about 2″ at 4.8 GHz and 1.1″ at 8.64 GHz. These two frequencies were observed simultaneously with total bandwidth 128 MHz for each, of which the central 96 MHz was used to avoid filter edge-effects (correlator configuration full_128_2).

Flux densities were calibrated by daily observations of 1934-638, the absolute flux standard for southern hemisphere observations at cm wavelengths. The standard calibration procedure using the MIRIAD software package was followed (Sault et al., 2004).

As described in Dickey et al. (2021), we did not attempt to map every source. The standard approach of aperture synthesis imaging is not well suited to monitoring a rapidly varying source, as the aperture or u,v plane, which is the Fourier Transform of the image or sky plane, is not well sampled until Earth rotation has caused the baselines between telescopes to follow tracks that cover a large fraction of the aperture. This typically takes 12 hours. For this project we could spend only a few minutes of telescope time on each source each day, and the sources might change their flux densities from day to day, or even hour to hour. Analysis of short scans on each XRB taken on the first three days was done using the complex fringe visibility function as measured at each u,v point (baseline and time). These visibilities are vector-added by the MYRIAD task UVFLUX to determine the flux density of a point source at the phase center (the real part of the vector sum), and the rms error on the flux density. Sources that were not detected on any of the first three days of observations were not reobserved, upper limit flux densities are in most cases between 2 and 3 mJy at both frequencies. Starting on MJD 54284 (3 July 2007) we concentrated on the sources that showed detections or tentative detections, i.e. with flux density above about 1.5 mJy, For these detected sources, combining the data from all days gave sufficient u,v coverage for mapping and cleaning. We then fitted two dimensional Gaussians to each source. Table 2 gives the Gaussian parameters for each detected source at 8.64 and 4.80 GHz, and upper limits for the non-detections.

The source that shows by far the strongest variation from day to day during the observation is Sco X-1, with an increase from about 1 to 67 mJy between 30 June and 1 July (MJD 54281 - 54282), then back down to about 8 mJy on 4 July (MJD 54285) before increasing again to 19 mJy on 5 July (MJD 54286), as shown on figure 3.

The radio variability of Sco X-1 has been studied by Fomalont et al. (2001). Cir X-1 also shows significant variations at 8.64 and 4.8 GHz, as shown on table 2. In this case the source is much weaker, it increases from about 1 mJy to over 6 mJy at 8.64 GHz from 29 June to 1 July (MJD 54280 - 54282), then drops below the detection limit ($\sim$1 mJy) from 3 July on (MJD $>$54283).

Observations in the near infrared and optical bands were made with the SMARTS (Small and Moderate Aperture Research Telescope System) 1.3m telescope at CTIO and the ANDICAM-IR instrument with a NICMOS 4 detector. The near infra-red channel (NIR) has pixel scale 0.275″ and field of view (FoV) 2.4′, while the optical channel has pixel scale 0.372″and FoV 6′ (Muzerolle et al., 2019). The infrared observations were first flat-fielded using flat-field exposures taken one per night at J, H, and/or K band. These are used to flag bad pixels and to determine a scale factor for each good pixel that is applied to every exposure before aligning all exposures to a common center. All exposures for each night are then combined by taking the median for each pixel. A single frame for each night is saved, and an overall median image is computed for each source, an example is shown in figure 4. Coordinates and magnitudes are taken from comparison with star images in each band from the 2MASS survey (Skrutskie et al 2006). Simple geometric scale and rotation parameters were derived using the KOORDS utility in the KARMA package (Gooch, 1996). Comparison of the derived star centroid positions with those in the 2MASS catalog gives an estimate for the precision of the coordinates of better than 1″ in most cases. Magnitudes and fluxes for each source observed in the IR and optical are given on Table 3. In a few cases, source positions in the 2MASS catalog differ from those in the Gaia eDR3 (Gaia Collaboration et al., 2018) catalog (listed on Table 1 columns 2 and 3) by more than 3″. These are noted in the footnotes to Table 3.

Fig. Set4. SMARTS K,H,J band images

In each field, 50 comparison stars from the 2MASS point source catalog (Cutri et al., 2003)¹¹1 https://irsa.ipac.caltech.edu/cgi-bin/Gator/nph-dd are considered, and those that are unconfused are used to calibrate the photometry of the program sources. Two dimensional Gaussians are fitted to the star images, and their integrals are assumed to be proportional to the flux in the band. The catalog magnitudes are fitted by a linear function of the logarithm of the flux, and the result is used to convert the program source flux to magnitude. An example of the flux vs. magnitude calibration is shown in Fig. 5. The rms of the residuals after subtracting the best fit power law (linear fit on the log scale of Fig. 5) gives the the magnitude errors listed with the magnitudes on column 4, Table 3. Columns 5 and 6 give the corresponding flux density and flux error, using the zero point flux densities for the J, H, and K bands of 1594, 1024, and 667 Jy from Cohen et al. (2003). In the J and H bands both errors can be smaller than 0.1 magnitudes.

Although XRBs are variable stars, both due to rapid changes in the optical and infrared luminosities of the X-ray emission region, and to variability in the companion star, a comparison between the magnitudes measured in this project with the 2MASS catalog values shows that there is generally not much change between the two epochs. We compare the observed J band magnitudes with the values in the 2MASS point source catalog and the few available V band magnitudes from the APASS catalog for all sources on figure 6. There has been surprisingly little variation between the 2MASS epoch and the date of these observations, roughly ten years later, except for GX339-4, which is much brighter than in the catalog (J=13.75 vs. 15.91 in the point source catalog, and H=12.99 vs. 15.40). A similar result is found by Shidatsu et al. (2011) who measure J=13.66 with a slow brightening to J=12.9 over a month of observations in March, 2009.

The optical V band data were also taken with the SMARTS 1.3m at CTIO, with the ANDICAM-CCD based on the Fairchild 447 detector. The images were flat fielded and bad pixels flagged before aligning, and median filtered for each night. Coordinates were determined using five or six reference stars visible on the 2MASS J or H band images in the KARMA task KOORDS. Comparison stars near the X-ray source were used for photometric calibration. Magnitudes at V band for these stars were taken from the APASS catalog (Henden et al. , 2015). Stars were checked and rejected if there were nearby objects that could confuse the Gaussian fitting. Reference stars in the optical are illustrated on Fig. 7. The photometric calibration is done using these stars similarly to the 2MASS calibration for the IR observations described above. Fig. 8 illustrates the star flux vs. APASS catalog magnitudes for GX 339-4, similar to Fig. 5. The errors following the magnitudes in column 4 are the rms scatter of the reference stars about the best fit linear relation between their magnitudes and the integral of their fitted Gaussians on Fig. 8, or the formal error in the constant term of the fit to the reference star magnitudes, whichever is greater. Fluxes and flux errors are derived from the magnitudes and their errors assuming the Johnson V band zero magnitude flux density of 3780 Jy (Cox, 2000).

Fig. Set7. SMARTS V band images

Neil Gehrels Swift Observatory is a multi-wavelength observatory dedicated to the study of gamma-ray burst (GRB) science and transient sources. It has three instruments that observe simultaneously in the gamma-ray, X-ray, ultraviolet, and optical wavebands, which work in tandem to provide rapid identification and multi-wavelength follow-up of GRBs and transient sources. The three instruments are the (1) Burst Alert Telescope (BAT) observing at 15 - 150  keV, (2) X-ray Telescope (XRT) observing at 0.3 - 10  keV, and (3) UV/Optical Telescope (UVOT) observing at 170 - 600 nm. More details about the mission can be found in Gehrels et al. (2004).

We requested and acquired multiple pointed Swift observations for 15 sources during the week of our observing campaign. In addition, for all of the sources observed in our campaign, we made BAT lightcurves from the Swift/BAT hard X-ray transient monitor data.

The UltraViolet and Optical Telescope (UVOT) is a diffraction-limited 30 cm (12 inch aperture) Ritchey-Chrétien reflector, sensitive to magnitude 22.3 in a 17 minute exposure. The Ritchey-Chrétien telescope has a f/2.0 primary that is re-imaged to f/13 by the secondary. This results in pixels that are 0.502″ over its 17′ square FoV. The detector is an intensified CCD with seven broadband UVOT filters (white, U, V, B, UVW1, UVM2, and UVW2) covering the 170 - 600 nm wavelength range. For more details on UVOT see Roming et al. (2005).

For most of the Swift observations a UVOT observation in image mode was made in one or more of the filters. For these observations we analyzed the data using the standard UVOT analysis threads ²²2https://www.swift.ac.uk/analysis/uvot/ using the tools for Swift data analysis found in the Heasoft software package ³³3https://heasarc.gsfc.nasa.gov/docs/software/heasoft/. For most datasets we used the pipeline-reduced summed image in the data products. For a few observations for which pipeline-reduced products were not complete it was necessary to sum the images using the tool uvotimsum. For each source we created a circular 5″ radius extraction region centered on the source and a nearby background extraction region (with no obvious sources) with at least four times the area of the source. For each band observed we ran uvotsource using a background threshold of 3 sigma to obtain the fluxes and their errors. For the Swift/UVOT the PSF is 2.5 arcsec and the pixel scale is 0.502. This limits the resolution and sources that we include or exclude from the source detection region. Most of the Swift/UVOT observations are in one of the ultraviolet filters. In the ultraviolet most stars drop in flux and most of the XRBs have significant emission because of the accretion disk. This limits the number of sources that are detected. For the significant Swift/UVOT detections we find well defined sources at the positions of the XRBs. For fainter sources we exclude any confusing sources that overlap the source detection region. For detected sources we determine flux densities and errors given in table 4. For sources not detected we give upper limits. The conversion between UVOT band and frequency are given in table 5.