EVR-CB-001: An Evolving, Progenitor, White Dwarf Compact Binary Discovered with the Evryscope

We discovered photometric oscillations in EVR-CB-001 from analyzing 2.5 yr of data from the Evryscope, obtained from 2016 January to 2018 June. Data were taken through a Sloan g filter with 120 s integration times, providing a total of 53,698 measurements. The wide-seeing Evryscope is a gigapixel-scale, all-sky observing telescope that provides new opportunities for uncovering rare compact binaries through photometric variations. It is optimized for short-timescale observations with continuous all sky coverage and a multi-year period observation strategy. The Evryscope is a robotic camera array mounted into a 6 ft diameter hemisphere that tracks the sky (Law et al. 2015; Ratzloff et al. 2019b). The instrument is located at CTIO in Chile and observes continuously, covering 8150 sq. deg. in each 120 s exposure. Each camera features a 29MPix CCD providing a plate scale of 13''/pixel. The Evryscope monitors the entire accessible southern sky at 2 minute cadence, and the Evryscope database includes tens of thousands of epochs on 16 million sources.

Here we only briefly describe the calibration, reduction, and extraction of light curves from the Evryscope; for further details we point the reader to our Evryscope instrumentation paper (Ratzloff et al. 2019b). Raw images are filtered with a quality check, calibrated with master flats and master darks, and have large-scale backgrounds removed using the custom Evryscope pipeline. Forced photometry is performed using APASS-DR9 (Henden et al. 2015) as our master reference catalog. Aperture photometry is performed on all sources using multiple aperture sizes; the final aperture for each source is chosen to minimize light curve scatter. Systematics removal is performed with a custom implementation of the SysRem (Tamuz et al. 2005) algorithm.

We use a panel-detection plot that filters the light curves, identifies prominent systematics, searches a range of periods, and phase folds the best detections from several algorithms for visual inspection. It includes several matched filters to identify candidate hot subdwarfs for variability and is described in detail in Ratzloff et al. (2019a). EVR-CB-001 was discovered using Box Least Squares (BLS; Kovacs et al. 2002; Ofir 2014) with the same settings, pre-filtering, and daily-alias masking described in Ratzloff et al. (2019a). The discovery tools and settings were tested extensively to maximize recovery of the fast transits and eclipses characteristic of hot subdwarfs and white dwarfs. As part of our testing, we also recovered CD-30 (Vennes et al. 2012), the only known fast-period hot subdwarf + WD binary in our field of view and magnitude range. The BLS power spectrum revealed EVR-CB-001 to be a 2.34 hr binary exhibiting strong (12%) modulations due to the ellipsoidal deformation of the primary from the unseen, more massive companion. The detection power in terms of Signal Detection Efficiency (SDE; Kovacs et al. 2002) is 33.5, compared to an average SDE of 8 for targets in the hot subdwarf survey (Ratzloff et al. 2019, in preparation) that EVR-CB-001 was discovered in. Figure 1 presents both the BLS power spectrum and phase-folded light curve.

Zoom In Zoom Out Reset image size

Our detection tools also include Lomb–Scargle (LS; Lomb 1975; Scargle 1982) and interestingly, the LS detection of the short periods in both EVR-CB-001 and CD-30 are relatively weak and are overpowered by longer periods (the search range in our survey is 2–720 hr for LS in an effort to recover a wide range of variables). Narrowing the period search range and further filtering of low frequency signals recovers the same period from LS as the BLS discovery. The high amplitude photometric variability in EVR-CB-001 results in an asymmetric signal, with a difference in even versus odd phase that is significant enough to affect the LS (optimized for sinusoidal signals) recovery. In this work, we show the BLS signal as it is the algorithm that led to the discovery, and is confirmed to be the correct period in Section 3. In the Evryscope hot subdwarf survey (Ratzloff et al. 2019, in preparation), we compare the effectiveness of BLS and LS in detecting compact ellipsoidal systems with multiple light-curve features and discuss the modifications to our original search in an effort to maximize the recovery of these compact binary systems.

A subtle asymmetry in the light curve (a sub 1% difference in the height of alternating peaks) is observed, indicative of Doppler boosting with the higher peak corresponding to the orbital position where the pre-He WD is moving toward us most quickly. The difference in minima is due to gravitational darkening of the deformed primary, with the lower minimum corresponding to the orbital position where the pre-He WD is farthest from us.

In order to obtain a higher S/N light curve for modeling, we observed EVR-CB-001 on 2019 January 5 using the SOAR 4.1 m telescope at Cerro Pachon, Chile, with the Goodman spectrograph (Clemens et al. 2004) in imaging mode. We used the blue camera with Bessel-V blocking filter and took 409 images with 15 s integration times. The image Region of Interest (ROI) was reduced to 1700 × 1000 pixels with 1 × 1 binning, which resulted in a 60% duty cycle. The surrounding field is sparse, and so a larger-than-ideal ROI was needed to capture a sufficient number of comparison stars. For calibrations, we took 10 dome flats using 25% lamp power and 10 s integrations, 10 darks also with 10 s integrations, and 10 bias frames.

The SOAR frames were processed with a custom aperture photometry pipeline written in Python. The object images were bias-subtracted, dark-subtracted, and flat-field-corrected using master calibration frames. Five reference stars of similar magnitude were selected, and aperture photometry was performed on all frames using a centroid algorithm and range of aperture sizes. The reference stars were confirmed to be nonvariable. We also use the photometric aperture on dark areas of the image near the reference stars to capture background counts. The reference star counts are combined for the image and the background is subtracted (using the average per-pixel background times the pixels in the aperture). The background subtracted reference star counts are recorded for each image and normalized by the mean. The target star counts are background subtracted in the same way and recorded for each image. The background subtracted target star counts are divided by the normalized background subtracted reference star counts to remove sky variations. The result is normalized by the mean to produce the final light curve.

In order to choose the best aperture, we removed variability from each light curve and chose the aperture with the lowest residual rms values. At this juncture, we did not have an exact model of the astrophysical variability of the system, but needed a reliable estimate of the variability so that it could be removed to measure the residual rms and choose the best photometric aperture. We used a Savitzk–Golay filter from the scipy.signal module (Jones et al. 2001), holding the the filter settings constant for all apertures. We also explored different settings to confirm the filter was not biasing the results. The filter was only used in this step to determine the best aperture, and is in no way applied to the photometry. The solution converged nicely with the minimum rms corresponding to a photometric aperture of 36 pixels, as shown in Figure 11 in the Appendix. The resulting differential light curve from SOAR, which we use to model EVR-CB-001, is shown later in the manuscript, in Figure 7.

We observed EVR-CB-001 on 29 nights between 2018 December 19 and 2019 January 28 with the SMARTS 1.5 m telescope and CHIRON, a fiber-fed cross-dispersed echelle spectrometer (Tokovinin et al. 2013). Spectra were taken in image fiber mode (R ∼ 28,000) and covered the wavelength range of 4400–8800 Å. We used integration times of 600 s to obtain just enough S/N for radial velocity measurements; longer integrations would have resulted in too much phase smearing. Spectra were obtained every few days at specified epochs until full phase coverage was achieved. All raw spectra were reduced and wavelength-calibrated by the official CHIRON pipeline, housed at Georgia State University and managed by the SMARTS Consortium.⁸ In addition to Hα and Hβ, which span multiple orders, the spectra show four He i lines, including 6678, 5876, 5016, and 4922 Å. All of these lines are synced in phase, with no signs of absorption due to a companion, and we conclude they emanate from a single star.

The CHIRON spectra have too high a resolution to easily model atmospheric parameters using the H Balmer lines, which span multiple orders. As such, we also obtained low-resolution spectra on 2018 December 2 with the Goodman spectrograph using the 600 mm⁻¹ grating blue preset mode, 2 × 2 binning, and the 1'' slit. This configuration provided a wavelength coverage of 3500–6000 Å with spectral resolution of 4.3 Å (R ∼ 1150 at 5000 Å). We took four 360 s spectra of both the target and the spectrophotometric standard star BPM 16274. For calibrations, we obtained 3 × 60 s FeAr lamps, 10 internal quartz flats using 50% quartz power and 30 s integrations, and 10 bias frames.

We processed the spectra with a custom pipeline written in Python. The spectra were individually bias-subtracted and flat-corrected. A third-order polynomial was fitted to the brightest pixels in each row; the spectra are then extracted in a 10 pixel range and background subtracted. We identify 16 prominent lamp emission lines and compare with the known lines of the FeAr lamp using a Gaussian fit to each feature. We used a fourth-order polynomial to fit the wavelength solution and calibrate each spectrum. We used our observations of BPM 16274 to flux-calibrate the EVR-CB-001 spectra by removing prominent absorption features and fitting a seventh-order polynomial to the continuum. Each spectrum was then rest-wavelength calibrated using a Gaussian fit to the Hβ through H11 absorption features, as well as several prominent He absorption features. The resulting spectra were median-combined to form a final spectrum for atmospheric modeling. As shown in Figure 4, we detect strong H Balmer lines, from Hβ through H13, and one He i line at 4472 Å. As was the case for the CHIRON spectra, we find no evidence of absorption features due to the companion star; EVR-CB-001 appears to be a single-lined binary. Table 1 presents a brief overview of all of the photometric and spectroscopic data used in our analysis of EVR-CB-001.

6.1.1. Magnitude/Distance

We tested our interpretation of the primary as a pre-He WD by estimating the pre-He WD radius and mass independently from the light-curve modeling. Using the parallax from GAIA-DR2 (Gaia Collaboration et al. 2018) we determine the distance, and with the Johnson V-band magnitude from APASS (Henden et al. 2015) we use the distance modulus to determine the absolute magnitude (with the bolometric and extinction corrections described below). With the mass–radius relation (Equation (3)), we express the luminosity ($L=4\sigma \pi {R}^{2}{T}^{4}$ from the Stephan–Boltzman equation applied to a blackbody) as a function of mass and surface gravity instead of radius. Using the zero-point luminosity, we solve for the mass, combine constants, and simplify to the following formula:

$$\begin{eqnarray*}\begin{array}{rcl}{M}_{1}[{M}_{\mathrm{odot}}] & = & 4.06609\times {10}^{10}\times {10}^{\mathrm{log}g}\times {T}_{\mathrm{eff}}{\left[{\rm{K}}\right]}^{-4}\\ & & \times \,{10}^{-0.4\times ({\mathrm{BC}}_{{\rm{V}}}+{m}_{{\rm{V}}}-{A}_{{\rm{V}}}+5\times \mathrm{log}\varpi [\mathrm{arcsec}])}.\end{array}\end{eqnarray*}$$

In addition to the previously derived values for T_eff and $\mathrm{log}g$ from Section 3, and the Gaia parallax ϖ, we adopted the apparent magnitude in the Johnson V-band m_V = 12.619 ± 0.051 mag from the APASS catalog. To account for the significant variability of the star, we adopted a higher uncertainty of 0.12 mag. The bolometric correction BC_V = −1.76 ± 0.075 was interpolated from Vizier table J/A+A/333/231/table3 (Bessell et al. 1998) for the appropriate spectroscopic parameters. The extinction A_V = 0.3007 ± 0.027 mag was taken from the Stilism 3D maps of the local interstellar medium⁹ (Lallement et al. 2014) adopting the parallax distance from GAIA. To derive the mass uncertainty we used the Python Monte Carlo error propagation mcerp package assuming that all input parameters are normally distributed. From the resulting distribution we adopted the maximum value and the mass values at FWHM to derive the uncertainties.

In this way we derive ${M}_{1}={0.30}_{-0.10}^{+0.17}\,{M}_{\odot }$ consistent with the mass determination from the binary analysis and indicating a low-mass pre-He WD. Using the mass–radius relation the radius of the pre-He WD ${R}_{1}={0.30}_{-0.07}^{+0.09}$ is derived to be slightly larger than from the light-curve analysis, but still consistent within the uncertainties.

6.1.2. MESA Stellar Evolution Code

To understand the nature of the primary, we have constructed pre-helium WD models for different masses using the MESA stellar evolution code (Paxton et al. 2011, 2013, 2015, 2018), release version 10398. The models were constructed using an initially 1.0 M_⊙ star (thought to be the most likely progenitor when starting at the main-sequence stage) that ascends the RGB, building a helium core before it starts He-core burning. Once the helium core reaches a specified mass, all but 0.01 M_⊙ of the hydrogen envelope is stripped. Residual hydrogen shell burning then governs the timescale for evolution as the star contracts and evolves toward hotter T_eff as seen in the resulting tracks (solid lines) in Figure 8.

Zoom In Zoom Out Reset image size

Additionally, we also computed MESA models of 0.461 M_⊙ (understood to be the beginning of the helium burning stage post RGB) He-burning stars with two different hydrogen envelope masses: 1.0 and 3.0 × 10⁻³ M_⊙. Our models use the MESA "predictive mixing" scheme for core convection to allow for proper growth of the convective He core and yield the correct core burning lifetime and luminosity (Paxton et al. 2018). The tracks for these models extend from the beginning of core He burning through exhaustion of He in the core 150 Myr later. After this phase, the stars will begin He shell burning and evolve toward a hotter effective temperature. Our measured T_eff and log(g) intercepts tracks with masses in the range of 0.22–0.23 M_⊙, which is in agreement with the mass determined from our light-curve modeling (see Section 5) as well as the determination using Gaia DR2 (see the previous paragraph). Known hot subdwarfs from Kupfer et al. (2015) are shown for comparison, clearly hotter and with higher surface gravity than EVR-CB-001. Overall, the compact binary EVR-CB-001 appears to contain a pre-He WD that is ellipsoidally deformed due the gravitational presence of an unseen He WD companion.

There does not appear to be an exact known analog for EVR-CB-001. In the following discussion, we compare the prominent features and components to known systems.

The photometric variability of EVR-CB-001 most resembles one of the exceptional massive WD/hot subdwarf compact binaries such as KPD 1930+2752, KPD 0422+5421, or CD-30 11223 (Koen et al. 1998; Maxted et al. 2000; Vennes et al. 2012) but with a higher amplitude in light-curve variability. This is reasonable given the lower mass and surface gravity as well as the bloated nature of the pre-He WD of EVR-CB-001 compared to hot subdwarfs. CSS 41177 is a rare eclipsing WD/WD compact binary with deep eclipses and relatively low-mass WDs (Bours et al. 2015). However, both of the WDs are mature and there is no ellipsoidal deformation given the high surface gravity of each WD.

Short period WD/WD binaries with ELM secondaries have been recently discovered showing tidal distortions (Hermes et al. 2012b, 2014) and eclipses in the case of the exceptional system J0651+2844 (Hermes et al. 2012a). The ELM survey (Kilic et al. 2011) has discovered compact binaries and potential merger systems with ELM secondary components. The higher temperature, lower surface gravity, early evolutionary stage, and extreme light-curve variation of EVR-CB-001 are quite different compared to the ELM binaries, as is the mass ratio of the primary and secondary

The companion of HD 188112 (Heber et al. 2003) is perhaps most similar to the pre-He WD of EVR-CB-001; however, it has higher mass, surface gravity, and temperature. The system is quite different from EVR-CB-001 with a high mass WD primary, a longer period, and without photometric variation. WD 1242-105 (Debes et al. 2015) is an example double degenerate binary with similarly favorable conditions to EVR-CB-001 that will potentially merge into a single hot subdwarf B star. The higher total mass and similar primary and secondary WDs highlight some of the differences to the EVR-CB-001 system. OWJ074106.0-294811.0 (Kupfer et al. 2017a) is an ultra compact system with large photometric variability, but with quite different components (more massive, hotter, and higher surface gravity). EVR-CB-001 is best understood as combining interesting parts of each of these rare binaries to form a peculiar system.

EVR-CB-001 likely formed via two separate stages of mass transfer. The original binary consisted of two main-sequence stars. As the more massive of these evolved off the main sequence first and ascended the red giant branch, it filled its Roche lobe and started transferring mass onto its less massive companion. Whether this stage of mass transfer was dynamically stable (stable RLOF) or unstable (CE formation) depends on many unknown parameters, most notably the mass ratio at the time of transfer. Either way, enough mass was stripped from the red giant that its remnant was unable to fuse helium thereafter and formed a He WD.

The second phase of mass transfer, which commenced once the lower-mass main-sequence star reached the giant phase, was undoubtedly unstable and led to the formation of a CE. Its He WD companion was unable to accrete at a sufficiently high rate, and significant mass was ejected, further tightening the orbit. Once again, the stripped object was left with insufficient mass for fusing He, causing it also to bypass the horizontal branch and collapse onto the white dwarf cooling sequence as another He WD. We appear to have caught EVR-CB-001 fairly shortly after this second mass transfer stage: the object that was most recently stripped of its outer layers appears as a hot and bloated pre-He WD, on its way to becoming a fully degenerate WD. Assuming that the progenitor of the pre-He WD was a ∼1 M_⊙ star, we can calculate the orbital period of EVR-CB-001 at the moment when the progenitor filled its Roche lobe. Using the same MESA model as that used in Section 6.1.2 we find that a 1 M_⊙ progenitor has a radius of 4–9 R_⊙ when the helium core has built up a mass of 0.17–0.23 M_⊙. Assuming a 0.3 M_⊙ companion we find that the progenitor system consisting of a He-WD with a red giant had a period of ≈1–3 days when the red giant filled its Roche Lobe and started unstable mass transfer. This shows that the orbit must have shrunken substantially during the CE phase when the pre-He WD was formed.

EVR-CB-001 represents a viable candidate progenitor system for the He WD merger channel leading to single hot subdwarf B stars (e.g., Han et al. 2002; Schwab 2018). Eventually, the pre-He WD we now observe will evolve onto the white dwarf cooling sequence, and EVR-CB-001 will become a full-fledged double-degenerate system. At such a short orbital period, gravitational wave radiation will cause the system to shrink until the less massive He WD (currently a pre-He WD) fills its Roche lobe in about ≈1 Gyr, at an orbital period of a few minutes. If the initiated mass transfer is dynamically unstable, the less massive He WD will be dynamically disrupted and form an accretion disk around its companion (Schwab et al. 2012; Zhang & Jeffery 2012). Depending on the details of the evolution of the accretion disk and accretion rates, it is possible for the more massive He WD to increase its mass to the point where it ignites He shell burning and becomes a core He-burning hot subdwarf B star with ∼0.5 M_⊙. Unlike the other formation channels presented by Han et al. (2002, 2003), which all leave behind a binary hot subdwarf system, this He WD merger channel produces a single hot subdwarf B star.

Although EVR-CB-001 is a candidate to form a single hot subdwarf B star, the system has a mass ratio (.66 ± .07) that might prevent the merger and instead evolve into a stable accreting AM CVn type binary. We briefly discuss this possibility here. For double white dwarf systems, commonly in the literature a system with a mass ratio $q={M}_{2}/{M}_{1}\lt 2/3$, M₁ being the mass of the accretor, is assumed to prevent the merger. However, Marsh et al. (2004) and Gokhale et al. (2007) studied the effect of coupling of the accretor's and donor's spin to the orbit when the larger object starts to fill its Roche Lobe. They found that a strong coupling and therefore a strong feedback of angular momentum to the orbit can destabilize systems with mass ratios lower than q = M₂/M₁ < 2/3,M₁ being the mass of the pre-He WD. Most recently, Shen (2015) proposed that even accreting double WD binaries with extreme mass ratios will merge due to classical nova-like outbursts on the accretor. Dynamical friction within the expanding nova shell causes the binary separation to shrink and the donor to dramatically overfill its Roche lobe, resulting in highly super-Eddington mass transfer rates that lead to a merger. This result was supported by Brown et al. (2016b) who found that the merger rate of ELM white dwarfs exceeds the formation rate of AM CVn binaries by a factor of 40 concluding that most ELM white dwarf binaries merge. Thus, although we cannot definitively conclude either way, EVR-CB-001 is a viable candidate to merge and form a hot subdwarf B star.

The Evryscope is a new instrument, different from a conventional telescope, and potentially misunderstood. Comparison of the Evryscope EVR-CB-001 discovery light curve to the SOAR follow-up light curve gives a powerful example of the Evryscope potential. Figure 9 shows the binned-in-phase Evryscope light curve and SOAR light curve. The flux and residual scaling is the same in both plots. The astrophysical signal is fit with the best solution in Section 4 and removed from both curves leaving the residuals. The residual rms of the SOAR and Evryscope light curves is 0.00155 and 0.00354 respectively. Consider the following instrument comparisons: the Evryscope cameras are 6.1 cm diameter while the SOAR telescope is 4.1 m diameter. The Evryscope instrument cost ≈$300K, while the SOAR telescope cost ≈$28M. The competitive Evryscope light curve is made possible because the SOAR light curve took 2.5 hr of observing time, while the Evryscope light curve took 2.5 yr. SOAR observed 1 period, while the Evryscope observed over 1000. An individual Evryscope period observation has only a very modest precision (in this case ≈.05 rms), but with the proper photometric pipeline and systematics removal, the final combined and binned-in-phase light curve improves as $\approx \sqrt{\#\mathrm{periods}}$ (in this case $\approx \sqrt{1000}$).

Zoom In Zoom Out Reset image size

It is important to emphasize that SOAR (or any other large telescope) and Evryscope are very different instruments. SOAR has many capabilities that Evryscope does not—spectroscopy, radial velocity measurements, and multi-band photometry just to name a few. It offers rapid precision follow-up on high value targets that the Evryscope cannot match. However, the Evryscope has a 8150 sq. deg. field of view with continuous 2 minute cadence that provides light curves just like the one for EVR-CB-001, but for 9.3M targets brighter than m_v = 15. While some are better quality and some are worse depending on target brightness and location, EVR-CB-001 is a representative example. The Evryscope is a robotic system that requires minimal human intervention, with low construction and operating costs, and provides a data set that facilitates the discovery of rare, difficult to detect, fast event systems like EVR-CB-001. With the proper processing of the discovery light curve, very high levels of binned-in-phase precision can be reached.