TWO MORE CANDIDATE AM CANUM VENATICORUM (AM CVn) BINARIES FROM THE SLOAN DIGITAL SKY SURVEY^*

AM Canum Venaticorum (AM CVn) systems are a select group of ultracompact binary systems, with orbital periods extending down to tens of minutes (e.g., see reviews by Warner 1995; Cropper et al. 2004; Nelemans 2005; Ramsay et al. 2007). The ultrashort orbital periods (the shortest of any known binary class) suggest that both binary components are at least partially degenerate, with a popular model invoking accretion driven by gravitational radiation from a fully- or semi-degenerate helium-rich donor (perhaps of very low mass) onto a more typical white dwarf primary. A hallmark spectroscopic characteristic of an AM CVn is a spectrum dominated by helium lines, and essentially devoid of hydrogen.

The unusual character of the prototype, AM CVn, was discovered 40 years ago (Smak 1967; Paczynski 1967), but this remarkable class has grown only slowly in membership over the intervening decades, e.g., with about a dozen members discussed in the review by Nelemans (2005). New cases of AM CVn systems are eagerly sought, as these systems may provide insights into a prior common-envelope phase, are discussed as possible SN Ia progenitors, and are plausible sources of gravitational waves (e.g., see Liebert et al. 1997; Livio & Riess 2003; Nelemans et al. 2004).

In the last few years the Sloan Digital Sky Survey (SDSS) has discovered many new cataclysmic variables (Szkody et al. 2002, 2007), and even five new AM CVn candidates. The first such SDSS AM CVn system, SDSS J1240-0159, was discovered in an early SDSS data release by Roelofs et al. (2005). In Anderson et al. (2005), hereafter Paper I, we reported the discovery of four additional AM CVn candidates (including the first eclipsing AM CVn) from the then-available SDSS spectroscopic database. All five of those earlier SDSS AM CVn candidates showed optical spectra dominated by strong helium emission, and this characteristic was essential in their recognition in the SDSS spectroscopic database.

In this paper, we report on another similar case, SDSS J120841.96+355025.2 (hereafter, SDSS J1208+3550), also recognized from SDSS spectra because of its strong helium emission. We also provide an updated surface density estimate for such emission-line AM CVns based on our expanded DR6 spectroscopic search area. Additionally, we report a second new AM CVn candidate, SDSS J204739.40+000840.3 (hereafter, SDSS J2047+0008), that was discovered via time-domain imaging data as part of the SDSS-II supernova survey (Frieman et al. 2007). As initially announced in a telegram by Prieto et al. (2006), SDSS J2047+0008 shows novae-like outbursts in multi-epoch SDSS imaging. Moreover, the optical spectrum of SDSS J2047+0008 at outburst shows helium absorption lines, in contrast to the spectra of the other SDSS discoveries, and more similar to some other well-known AM CVn systems such as KL Dra and 2003aw (Jha et al. 1998; Filippenko & Chornock 2003). Including the two additions discussed herein, the total SDSS yield thus far is seven new AM CVn candidates, a substantial contribution to expanding the membership of this rare subclass, compared to the approximately dozen cases known prior to SDSS.

The SDSS is a multi-institutional and international project creating an optical digital imaging and spectroscopic data bank of a region approaching ∼10⁴ deg² of sky, primarily centered on the north Galactic polar cap. Imaging and spectroscopic data are obtained by a special purpose 2.5 m telescope, at Apache Point Observatory, New Mexico, equipped with a large-format mosaic camera that can image ∼10² deg² per night in five filters (u,  g,  r,  i,  z), along with a multifiber spectrograph that obtains 640 spectra within a 7 deg² field. The imaging database is used to select objects for the SDSS spectroscopic survey, which includes (λ/Δλ ∼ 1800) spectrophotometry covering 3800–9200 Å for 10⁶ galaxies, 10⁵ quasars, and 10⁵ stars. Technical details on SDSS hardware, software, and astrometric, photometric, and spectral data may be found in a variety of papers, e.g., Fukugita et al. (1996), Gunn et al. (1998), Lupton et al. (1999), York et al. (2000), Hogg et al. (2001), Stoughton et al. (2002), Smith et al. (2002), Pier et al. (2003), Ivezić et al. (2004), and Gunn et al. (2006). A description of the most recent SDSS Public data release (Data Release 6; hereafter, DR6) is given by Adelman-McCarthy et al. (2008).

As discussed in Paper I, the SDSS collaboration includes a "Serendipity Working Group" which is engaged in selection and examination of a large number of atypical SDSS spectra; such spectroscopic selection and visual examinations have led to the initial discovery of the bulk of the previous SDSS AM CVns. Their distinctive spectra, similar to cataclysmic variables but with helium rather than hydrogen emission lines, drew our special attention to these cases.

Similarly, SDSS J1208+3550, the first new AM CVn candidate discussed in this paper, was initially found as a serendipitous discovery during a search of the DR5 SDSS spectroscopic database for DQ white dwarfs (Halford et al. 2005).¹⁶ The discovery spectrum of SDSS J1208+3550 is displayed in Figure 1, and strongly resembles the previous SDSS AM CVn finds, with prominent broad emission at He i 3888, 4026, 4471, 4921, 5015, 5875, 6678, 7065, and 7281 Å; for example, the equivalent width and full width at half maximum of He i λ5875 are, respectively, 28 Å and 1600 km s⁻¹. There is also weaker He ii λ4686 emission. Though the signal to noise is modest, the Figure 1 spectrum again suggests multi-peaked helium emission, presumably from an accretion disk. As with most previous SDSS AM CVn discoveries, SDSS J1208+3550 was targeted for a spectrum because of its unusually blue color in SDSS imaging data, having been selected by several independent (but overlapping) target selection methods including "hot-standard star," "quasar," "serendipity," and "white-dwarf" algorithms. Basic astrometric and photometric data from SDSS are provided in Table 1, though, of course, AM CVns may be variable, and so the tabulated SDSS J1208+3550 magnitudes are only indicative of its character at the epoch of the SDSS observations.

Zoom In Zoom Out Reset image size

Because visual and other serendipitous searches of spectra, such as those which initially revealed SDSS J1208+3550, could be biased and incomplete, we have also algorithmically sifted through the SDSS DR6 spectroscopic database for any further new SDSS AM CVn candidates showing strong helium emission. This algorithmic search is directly analogous to that discussed in detail in Paper I (which was based largely on the smaller DR4 spectroscopic area). In the current paper, this algorithmic search was applied to DR6 plates encompassing more than 7000 deg² and a million spectra.

In particular, we queried the DR6 spectroscopic database to return a list of objects with spectra for which the SDSS pipeline data reduction algorithms (Stoughton et al. 2002) found any emission line with equivalent width >3 Å, with wavelength centered within 20 Å windows around either He ii λ4686 or He i λ5875. These are typically the highest equivalent-width lines among the Paper I AM CVn candidates (initially found in the visual search of earlier SDSS spectroscopic plates). The choice of minimum equivalent width in this algorithmic search, of course, limits such discoveries to AM CVns with helium emission lines, but does encompass all previous SDSS AM CVn candidates: all earlier SDSS cases found in visual searches of spectra have EW>13 Å for one of the above two helium emission lines. The algorithmic SDSS spectral database queries initially returned a list of 23,100 spectra potentially having such emission at either 5875 Å or 4686 Å. Each algorithmically-chosen spectrum was then reviewed by eye to assess whether it might be an AM CVn candidate. This algorithmic search process recovered SDSS J1208+3550 (originally in DR5) and all five previously published SDSS AM CVn candidates (Roelofs et al. 2005; Anderson et al. 2005); however, no additional convincing AM CVn candidates were identified in DR6 based on this algorithmic emission-line search.

That the known AM CVns occupy small and relatively sparsely-populated regions of SDSS color–color diagrams (Figure 2) also permits us a secondary search technique for AM CVn candidates that might have slipped through the pipeline emission-line algorithmic search discussed above. A total of 13 AM CVns fall within the SDSS DR6 imaging survey region, among which are ten whose magnitudes lie in the range 15 < g < 20.5; this is an approximate range in which SDSS provides high quality imaging photometry and spectroscopy. We construct a three-dimensional (hereafter, 3D) box in u − g, g − r, and r − i multicolor space that encompasses those ten AM CVns; we enlarge the boundaries of this box a little in each dimension, as shown in Figure 2, to encompass not just the SDSS colors of these ten AM CVns, but their colors plus 3σ color errors as well. (In constructing this box, we do not consider three additional AM CVns in the SDSS area that were either too bright or too faint for reliable photometry. Nor do we consider the i − z color distribution of the AM CVns, as the z-band photometric errors are large for most of these blue AM CVns imaged in SDSS.) We then visually examined the 8300 objects with spectra in SDSS DR6, whose colors in SDSS imaging fall within this 3D multicolor box. This additional perusal of SDSS DR6 color-selected spectra again recovered all five previously published SDSS AM CVns, but revealed no additional strong emission-line AM CVn candidates aside from SDSS J1208+3550. Thus, SDSS J1208+3550 (originally from DR5) is the solitary new emission-line candidate AM CVn discussed herein.

Zoom In Zoom Out Reset image size

Following our initial 2005 discovery of SDSS J1208+3550, in 2006-2007 we then obtained follow-up exploratory optical lightcurve and X-ray flux observations of this new AM CVn system. We used the SPIcam CCD imager with an SDSS g-band filter on the ARC 3.5 m to obtain optical lightcurve information from brief time-series sequences (spanning about 1–2 h each) on four distinct nights: 2006 February 1, 2006 March 3, 2006 April 1, and 2007 January 12 (all UT). Several of these nights were not photometric, but differential photometry was obtained relative to (the same) three brighter comparison stars in each image. The time resolution between individual exposures was in the range 1.4–2.4 min for all nights. Unfortunately, these data reveal no convincing periodic optical modulation for periods in the range of about 4–40 min, and with an amplitude limit of a few hundredths of a magnitude. Longer-timespan observations would be especially useful to obtain similar sensitivity to photometric modulations at (plausibly) longer orbital periods.

We also obtained a ∼3000 s X-ray exposure of SDSS J1208+3550 from the Chandra X-ray Observatory on 2007 March 23 (UT). These data were collected in VFAINT mode allowing us to utilize the full 5 × 5 pixel event island and thus reduce ACIS particle background. We reprocessed the level 1 events file in order to leverage this additional information and thereby filter for hot pixels, cosmic rays, background events, bad grades, and good time intervals. SDSS J1208+3550 is strongly detected in our Chandra data, with 45 counts in a medium energy (0.5–4.5 keV) X-ray band; background contributions are evaluated with a monoenergetic (0.5 keV) exposure map. For an assumed blackbody X-ray spectrum with kT = 0.7 keV, e.g. similar to that inferred for ES Cet (Strohmayer 2004), these Chandra medium band data imply an unabsorbed 0.2–10 keV flux of about 1.0 × 10⁻¹³ erg s⁻¹ cm⁻². This measured X-ray flux for SDSS J1208+3550 is reassuringly similar (within about a factor of 2) to that we predicted based on the typical X-ray to optical flux ratios of other AM CVns.

Since the completion of Paper I, an entirely new opportunity for AM CVn searches has emerged from SDSS, and here we also report a new AM CVn, SDSS J2047+0008, discovered in that very different manner. SDSS J2047+0008 (originally known internally as "candidate 15204") was instead discovered in the course of the SDSS-II supernova search (Frieman et al. 2007), which surveys the 300 deg² SDSS "Stripe 82" region via multi-epoch imaging. Basic data on SDSS J2047+0008 are displayed in Table 1, which reflect the object's state (in outburst) on UT 2006 October 12 (MJD 54020.125); its color on this epoch also falls nicely within the 3D multicolor box found for other AM CVns shown in Figure 2.

This new object is markedly variable. As may be discerned from the Figure 3 long-term lightcurve, SDSS J2047+0008 has been imaged with SDSS-I and SDSS-II Stripe 82 observations at 60 epochs over an eight-year timespan (Ivezić et al. 2007; Holtzman et al. 2008), but was unnoticed until its outburst in the fall of 2006, when caught as part of the SDSS-II supernova discovery program. The large amplitude variations strongly suggests nova-like outbursts in this system.

Zoom In Zoom Out Reset image size

The SDSS-II lightcurve of SDSS J2047+0008 shown in Figure 4 has been expanded to highlight the fall 2006 discovery and outburst observations. SDSS J2047+0008 (not too surprisingly) appears to have a stronger ultraviolet excess in its brighter state. A potentially interesting aspect of the lightcurve is that this object may show some cycling between high and intermediate states while fading; such behavior, for example, has been reported for the (outbursting) AM CVn system V803 Cen (Patterson et al. 2000; Kato et al. 2004).

Zoom In Zoom Out Reset image size

Prompted by its dramatic variability in SDSS-II imaging, an optical spectrum of SDSS J2047+0008 was taken on UT 2006 October 19 (MJD 54027) with the MDM 2.4 m Hiltner telescope (plus OSU Boller and Chivens Spectrograph), and is shown in Figure 5. There is clearly a blue continuum with shallow helium absorption lines of He i 3820, 3867, 3927, 4009, 4026, 4144, 4388, 4471, 4921, and perhaps 5875 Å. This spectrum is unlike traditional novae, but similar to the other objects such as KL Dra (also known as 1998di; Jha et al. 1998) and 2003aw (Filippenko & Chornock 2003), which were also noticed as hydrogen deficient novae, and subsequently classified as likely AM CVns.

Zoom In Zoom Out Reset image size

The absorption-line spectral character of this new variability-selected AM CVn system, SDSS J2047+0008, is in strong contrast to the other SDSS discovered AM CVns, which were all recognized in the SDSS database by virtue of their helium-dominated emission line spectra. The SDSS databases (including the SDSS-II Stripe 82 data) may include interesting additional AM CVns similar to SDSS J2047+0008, still awaiting recognition.

A follow-up lightcurve of SDSS J2047+0008 in outburst, and with time resolution potentially sensitive to a strong orbital modulation, was obtained with the WIYN telescope and MiniMosaic CCD array on 2006 November 1 (UT). Exposures were 300 s long through a Harris R-band filter, and each exposure had a 2 min readout time, yielding 19 images over a 2.5 h timespan. Differential photometry was performed by comparison to nearby stars with known SDSS calibrations. A star 15'' to the southeast of SDSS J2047+0008 is estimated to have a R-band magnitude of 16.46 ± 0.02 mag by converting from SDSS magnitudes. From this approximate calibration, the average magnitude of SDSS J2047+0008 over the WIYN observing run was found to be R = 20.58, but with significant variations of 0.2 mag over the 2.5 h span. Additional lightcurve data taken at good time-resolution may be useful in determining an orbital period; however, a quality lightcurve outside of outburst may prove difficult, given the faintness of the object in quiescence (Figure 3).

Our spectroscopic search for emission-line AM CVn candidates from SDSS now includes spectroscopic plates extending through DR6, and provides some updated, though approximate, constraints on the surface density of AM CVns. These constraints remain uncertain as SDSS spectroscopy (purposefully emphasizing the main galaxy and quasar surveys) is highly incomplete and non-uniform across the sky even for emission-line AM CVn candidates (see details in Paper I).

The main SDSS spectroscopic survey, along with the SDSS-II SEGUE survey, include spectroscopic plates which encompass 7400 deg² of sky and 1.3 million spectra. SDSS has discovered six emission-line AM CVn candidates in this region, including SDSS J1208+3550 presented herein; all six were initially selected as interesting for spectroscopic follow-up based on their blue colors in SDSS imaging, with subsequent recognition as AM CVns in the SDSS spectroscopic database because of their helium-dominated emission line spectra (Roelofs et al. 2005; Paper I; this paper). All six of these had 15 < g < 20.5 at the epochs of both SDSS imaging and SDSS spectrophotometry, and this is also a plausible range for both high-quality SDSS imaging and spectra. (For example, SDSS spectroscopy is restricted to m > 15 to avoid fiber cross-talk; and for blue stellar objects, m < 20.5 is the typical faintest limit for routine spectroscopic target selection.) Thus, an approximate, conservative lower limit on the surface density of AM CVns is >1/1200 deg⁻² for 15 < g < 20.5.

To proceed further, albeit less securely, requires an estimate of the incompleteness of SDSS spectroscopy for similar emission-line AM CVn candidates. The incompleteness even for very similar AM CVn candidates arises from several complicating factors, including: the emission-line AM CVn candidates found in SDSS were selected for spectroscopy by several distinct target selection algorithms, each differing in their specific color-selection criteria and limiting magnitudes; and some of these target selection algorithms actually receive spare spectroscopic fibers only rarely and non-uniformly over the sky, when the main galaxy and quasar surveys do not consume all available fibers for a given spectroscopic plate.

But, as a rough estimate of such spectroscopic incompleteness for similar AM CVn candidates, we consider again the same 3D box in u − g, g − r, r − i multi-color imaging space described in Section 2 (also see Figure 2). In the DR6 imaging area there are about 2.97 deg⁻² stellar objects with such colors and 15 < g < 20.5; in the DR6 spectroscopy area, the surface density of such objects actually having SDSS spectra taken is about 1.14 deg⁻². Thus, SDSS spectroscopy would be of order 38% complete for such objects, naively assuming uniform spectral coverage throughout the 3D box.

Alternatively, one might consider a somewhat smaller box in multicolor space that extends only to objects as red as about u −  g < 0.11; as may be discerned from Figure 2, this smaller 3D box excludes only one of the (ten) known AM CVns with reliable photometry in DR6. In the DR6 imaging area, such a smaller 3D multicolor box includes a typical surface density of about 1.04 deg⁻² of such objects. The surface density of such SDSS objects also having spectroscopy is about 0.40 deg⁻², suggesting (naively) a similar SDSS spectroscopic overall completeness of order 39%. In fact, the situation is more complicated, as such spectroscopic completeness depends on various magnitude limits and may vary across even this smaller multicolor box, due to the differing target selection algorithms used. More elegant estimates may be tied to AM CVn binary period and temperature/color relations; see Roelofs et al. (2007) for an excellent example of this approach.

As an intermediate improvement on our simple estimate above, we have considered the SDSS spectroscopic completeness (accounting for differences in DR6 imaging and spectroscopic survey areas) in these same 3D multicolor boxes for each 0.5 magnitude bin in the range 15.0 < g < 20.5. Our spectroscopic completeness estimates are very similar when considering either the larger or smaller 3D multicolor boxes, so the precise box color boundaries may not be critical; completeness (for morphologically stellar objects) ranges from a low of 10–20% at the very brightest and faintest ends, maximizing at just under 70% spectral completeness at g = 18 − 19. Within the faintest four bins spanning g = 18.5 −  20.5, where the great bulk of SDSS 3D color-selected candidates would in any case fall, and which also encompasses all the emission-line selected AM CVns from SDSS, the lowest completeness estimate is 24%. We thus adopt the latter value as a representative lower limit to the SDSS spectroscopic completeness within the 3D boxes shown in Figure 2. This updated completeness estimate is also quite similar to the ∼20% we suggested in Paper I, and even to the more sophisticated completeness estimates of Roelofs et al. (2007) in a similar magnitude regime (and for a wide range in AM CVn binary periods in their model).

Broadly then, the surface density of AM CVns with 15 ≲ g ≲ 20.5, similar to those with strong helium emission as found in the SDSS spectroscopic database, might be expected to be approximately in the range from 10^−3.1 deg⁻² (with no spectroscopic completeness correction) to about 10^−2.5 deg⁻² (adopting representative spectroscopic completeness of >24%). Of course, there are additional AM CVns aside from those typically showing strong emission lines; indeed, SDSS J2047+0008 discussed in Section 3 is the most recent example.

In one commonly discussed scenario (e.g., Nelemans 2005), the youngest members of the AM CVn class are thought to be those with the highest mass-transfer rates and shortest periods (<20 min), and these "high-state" systems show predominantly helium in absorption in their optical spectra. In that scenario, they then evolve into somewhat longer periods as the binary widens and the mass transfer rate continues to decrease; in this intermediate stage, perhaps with 20–40 min periods typical, the objects may show nova-like outbursts, and display both high-state (absorption) and low-state (emission) spectra. Then finally there is a late stage (perhaps with typical 40–60 min binary periods) in which AM CVns are primarily low-state systems with predominantly emission-line spectra. Other scenarios similarly place an approximate dividing line between mainly emission-line and mainly absorption-line systems at binary periods of about 30 min (Roelofs et al. 2007). The relative populations in these various stages are not yet well constrained empirically, but at least some models (Roelofs et al. 2007) predict significantly fewer of the shorter-period systems, and so the surface density of all AM CVns might be comparable to the estimate above based on emission-line AM CVns only.

At first glance these fiducial surface density limits are also compatible with our discovery of the one variability-selected case, SDSS J2047+0008, from the SDSS Stripe 82 survey region, given: (i) the uncertainty in the above surface density estimates; (ii) the 300 deg² areal coverage of time-domain imaging of the Stripe 82 supernova survey region; (iii) the faintness in quiescence (often g > 22) of SDSS J2047+0008, a regime for which presumably the surface densities are even higher; and (iv) the large uncertainty in the relative populations among high, intermediate, and low-state AM CVns.

Despite recent down-sizing in population model estimates for AM CVns based on essentially this same set of SDSS objects (Roelofs et al. 2007), our current surface density estimates suggest that upcoming large-area multicolor plus time-domain surveys such as the Large Synoptic Survey Telescope (LSST)—with their excellent time-sampling and color range, and extending to fainter magnitudes than the SDSS—may still provide an excellent opportunity to discover dozens or even hundreds of additional AM CVns. Such a large area, multicolor plus time-domain sample might especially extend uniform surveys to the intermediate- and high-state AM CVn populations that (often) lack emission lines, and which may be much more difficult to efficiently identify in multicolor-only surveys.

In any case, along with SDSS J1208+3550 and J2047+0008 presented herein, SDSS-I and SDSS-II have thus far yielded a total of seven new AM CVn candidates (Roelofs et al. 2005; Anderson et al. 2005; this paper). The SDSS is thus providing a substantial expansion to this rare object class, compared to the approximately dozen AM CVns previously known.

Funding for the SDSS and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the U.S. Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, the Max Planck Society, and the Higher Education Funding Council for England. The SDSS Web Site is http://www.sdss.org/.

The SDSS is managed by the Astrophysical Research Consortium for the Participating Institutions. The Participating Institutions are the American Museum of Natural History, Astrophysical Institute Potsdam, University of Basel, University of Cambridge, Case Western Reserve University, University of Chicago, Drexel University, Fermilab, the Institute for Advanced Study, the Japan Participation Group, Johns Hopkins University, the Joint Institute for Nuclear Astrophysics, the Kavli Institute for Particle Astrophysics and Cosmology, the Korean Scientist Group, the Chinese Academy of Sciences (LAMOST), Los Alamos National Laboratory, the Max-Planck-Institute for Astronomy (MPIA), the Max-Planck-Institute for Astrophysics (MPA), New Mexico State University, Ohio State University, University of Pittsburgh, University of Portsmouth, Princeton University, the United States Naval Observatory, and the University of Washington.

D.H. acknowledges support from the NASA Harriett G. Jenkins Predoctoral Fellowship. Support for this work was provided by the National Aeronautics and Space Administration through Chandra Award Number GO7-8028X issued by the Chandra X-ray Observatory Center, which is operated by the Smithsonian Astrophysical Observatory for and on behalf of the National Aeronautics Space Administration under contract NAS8-03060.