Kepler eclipsing binary stars – VI. Identification of eclipsing binaries in the K2 Campaign 0 data set

Abstract

The original Kepler mission observed and characterized over 2400 eclipsing binaries (EBs) in addition to its prolific exoplanet detections. Despite the mechanical malfunction and subsequent non-recovery of two reaction wheels used to stabilize the instrument, the Kepler satellite continues collecting data in its repurposed K2 mission surveying a series of fields along the ecliptic plane. Here, we present an analysis of the first full baseline K2 data release: the Campaign 0 data set. In the 7761 light curves we have identified a total of 207 EBs. Of these, 97 are new discoveries that were not previously identified. Our pixel-level analysis of these objects has also resulted in identification of several false positives (observed targets contaminated by neighbouring EBs), as well as the serendipitous discovery of two short-period exoplanet candidates. We provide catalogue cross-matched source identifications, orbital periods, morphologies and ephemerides for these eclipsing systems. We also describe the incorporation of the K2 sample into the Kepler Eclipsing Binary Catalog,^§ present spectroscopic follow-up observations for a limited selection of nine systems and discuss prospects for upcoming K2 campaigns.

INTRODUCTION

Between 2009 and 2013, the Kepler space telescope recorded high-precision photometry almost continuously for more than 150 000 stars near the constellation Cygnus and Lyra (Koch et al. 2010; Borucki et al. 2011a,b; Batalha et al. 2013; Burke et al. 2014). This unprecedented data set was used to detect more than 4000 exoplanet candidates (of which 1031 Kepler planets have been confirmed to-date) and revolutionized our understanding of the statistical occurrence of exoplanets and stellar astrophysics. The light curves of more than 2400 eclipsing binary (EB) stars were also measured (Prša et al. 2011; Slawson et al. 2011), resulting in the detection of over 70 triple systems (Rappaport et al. 2013; Conroy et al. 2014b; Borkovits et al. 2015) and new astrophysical phenomena such as tidally excited Heartbeat binaries (Thompson et al. 2012), self-lensing binaries (Kruse & Agol 2014) and Doppler boosting in ordinary stars (Kerkwijk et al. 2010).

Most of these discoveries were made by the Kepler science teams, using an automated pipeline (Jenkins, Caldwell & Borucki 2002; Jenkins et al. 2010) to analyse the photometric data. In addition, the Planet Hunters Citizen Science project¹ was established to engage the public in helping to analyse the wealth of light curves (Fischer et al. 2012). This turned out to be a complementary scientific asset: to-date, the visual inspection of light curves from Planet Hunters has resulted in the recovery and characterization of many of the transiting systems detected by the Kepler team and more than 50 unique exoplanet candidates. This includes dozens of giant planet candidates residing in the putative habitable zones of their host stars, and a circumbinary exoplanet orbiting one of a pair of binaries in a widely separated quadruple system (Schwamb et al. 2013; Wang et al. 2013; Schmitt et al. 2014).

Kepler was about to begin an extended mission when the loss of the second of four reaction wheels ended the primary mission. At this juncture, a new plan of operation was devised to point the spacecraft along the ecliptic plane. This orbit minimizes the perturbations from solar illumination and pointing is maintained with the remaining two reaction wheels (Howell et al. 2014). This repurposed mission, known as K2, will survey 10 new fields along the ecliptic plane over the next 2–3 yr. Each of these ecliptic pointings constitutes one campaign, and the duration of observations for each field is limited to about 75 d because of solar radiation pressure constraints. All data are immediately non-proprietary, and will be released from the Mikulski Archive for Space Telescope (MAST) archive about 3 months after observations to engage the community in rapid follow-up. The first K2 observations began in 2014 March to assess the quality of the photometric precision. This Campaign 0 (C0) data were uploaded to MAST in 2014 September.

The K2 C0 data set consists of observations from an ∼80 d long engineering run, designed to assess whether a long-term two-wheeled, ecliptic orientated mission was viable. Significant differences exist between the Kepler and the K2 data sets in both quantity and quality. While the Kepler targets were carefully vetted before the launch of the mission and nominally selected based on their magnitude and luminosity class, the K2 C0 targets were requested by the community and selected from 89 diverse Guest Observer (GO) proposals that aimed to carry out a broad range of science goals. This method of K2 target selection has introduced different selection biases than for the stars observed in the Kepler field. The K2 target information for each campaign is stored in the Ecliptic Plane Input Catalog² (EPIC).

Unlike the Kepler mission, the quality of the K2 C0 photometry is compromised by quasi-periodic, 6 h thruster firings, a more rapidly changing thermal environment and the emergent failure of detector Module 7. These effects have introduced significant and non-uniform systematics into the data. Targets that are farther from the bore sight sweep out larger arcs on the detectors than targets that are in the centre of the field. Because of complications extracting and cleaning light curves, the K2 mission is currently only providing MAST with a calibrated Target Pixel File (TPF) for each target.

Despite these new challenges, the precision photometry offered by the K2 mission will continue to make significant contributions to the field of EBs. This has already been evidenced by the ‘Two-Wheel Concept Engineering Test’ data set, a 9 d target limited observation performed in 2014 February that resulted in the detection of 31 EBs (Conroy et al. 2014a). Upon release of the C0 TPFs, we developed a custom light-curve extraction pipeline in order to reduce and review the data. We present results from our pilot study to identify light curves that exhibit periodic variations, in particular, EBs in the C0 data set. These results join EBs detected from Kepler and the K2: Two-Wheel Concept Engineering Test within the Kepler Eclipsing Binary Catalog.³

CAMPAIGN 0 LIGHT CURVES

We initially extracted 7761 light curves in an automated fashion and independently decorrelated each of them against the spacecraft's known pointing motions, using the same technique described in Vanderberg & Johnson (2014). In addition, the aperture extraction mask size was adjusted as a function of the EPIC catalogue listed source magnitude, with minimum mask dimensions of 3 × 3 pixels designated for the faintest observed C0 targets (K_p of 19 or less) ranging up to 15 × 15 for the brightest (K_p ∼ 9). Unfortunately, a systemic spacecraft pointing error resulted in degradation for the first 40 d of observations for all targets (about half of the full C0 observation time baseline). While this early data range was not entirely compromised, we only used the second half of C0 observations in our bulk analysis. However, in our final analysis we were able to recover some of this early data to aid in characterizing several long-period EBs.

Visual inspection of the extracted C0 light curves produced an initial list of 258 EPIC targets that were flagged to be investigated for transient or periodic variations suggestive of an EB. A fast Fourier transform routine subsequently recovered an additional 53 candidate signals of varying periods.

A 384 000 pixel ‘superstamp’ was used in C0 to capture flux from the open clusters NGC 2168 (M35) and NGC 2158. The superstamp is coverage oriented, and no explicit EPIC target identifications are associated with each of the 153 uniform dimension TPFs contained within the superstamp. We assessed each of the superstamp TPFs for crowding. In this selection, we excluded dim background sources comprised of fewer than 4 pixels and any brighter source truncated by the borders of the TPF mask. We tailored our photometric extraction routine to identify source centroids and process light curves, again adjusting aperture masks as a function of source magnitude. These superstamp extracted light curves were then visually reviewed and vetted at the pixel level in the same manner as the regular C0 target sample. Although there is increased incidence of source confusion for superstamp TPFs covering the nominal centre of each cluster, in many cases the exterior areas of the superstamp yielded high-quality light curves at an average of 23 extractions per TPF. Examination of the output from this superstamp pipeline resulted in 49 additional periodic variables which were added to our initial investigation pool, bringing the total number of flagged targets to 360.

Because of concern about the quality of the K2 C0 data, we employed open source GO software pyke (Still et al. 2012)⁴ to perform a second series of extractions for the flagged point sources, reducing the data into two new light curves using two different reduction techniques: self-field flattening (SFF; Vanderberg & Johnson 2014) and principal component analysis (PCA). These two reduction routines are included in the pyke software package as kepsff and keppca and have been updated for use with K2. The SFF method corrects for the K2 aperture photometry artefacts by isolating short-term variability and calculating a position–flux relation for each detrended iteration of the extracted light curve. PCA accomplishes a similar isolation of systematics by assessing the time series variability of each individual pixel in the chosen extraction mask and then segregating these trends into component groups that are correlated or anticorrelated to instrument movement (Harrison et al. 2012).

Aperture photometry pixel masks common to this second series of reductions and pixel-level analysis were manually defined on a case by case basis in order to minimize flux losses from the intended target and to exclude neighbouring sources wherever possible. We also extracted and inspected light curves from bright neighbouring sources that were potentially contaminating our target of interest.

The light curves extracted with keppca and kepssf were compared to assess the degree each routine had muted or obfuscated the intrinsic astrophysical signal. Testing this across a variety of point sources showed that neither reduction technique offers a clear advantage over the other in all situations. As such, we chose to employ light curves derived from both kepsff and keppca for each of the targets reviewed in the course of our dedicated pixel-level analysis.

CULLING

Due to Kepler's sensitivity, pixel size and observing fields, the likeliness for source confusion common to any given target is high. The most common forms of contamination occur as a result of Kepler's pixel response function (PRF), which is quantified by the spacecraft's pointing precision, pixel resolution and point-spread function (Bryson et al. 2010). When the PRF of two or more point sources directly overlap and blend, or when the PRF wings of sufficiently bright stars encroach into the photometric apertures of separated targets, the variable signal from one point source can be introduced into another point source as a diluted alias (Van Cleve & Caldwell 2009; Caldwell et al. 2010). For this reason, we take a conservative approach in the vetting of EBs for our final C0 sample. If the source of the eclipsing signal for a particular EB could not confidently be determined during a pixel-level review, it was excluded from our final sample. Of the 360 light curves we flagged for eclipsing or ellipsoidal activity, 153 were ultimately rejected as spurious detections or blended binaries that could not be confidently associated with a specific point source. The itemized breakdown of these rejected EB candidates is detailed as follows:

1. Fifteen flagged EPIC targets were rejected as suspected cases of direct PRF contamination. Examination of the TPF in conjunction with review of optical all-sky data bases (e.g. Two Micron All Sky Survey (2MASS), Skrutskie et al. 2006; DSS;⁵WISE, Wright et al. 2010) indicated that multiple point sources were blended or crowded within our initially employed photometric aperture. Subsequent efforts to re-extract and isolate the eclipse signal failed to tease apart the originating pixel location and as such these low signal-to-noise (SNR) candidates were culled from our sample.

2. Twenty-nine additional cases where a point source in the halo of an EPIC's TPF was found to contaminate the intended EPIC target's light curve. Aliases were identified where it was evident that the halo source displayed the same ephemerides but at a much greater amplitude. We have cross-matched the 2MASS identifications of these non-EPIC EBs and appropriately noted the victimized EPICs as false positives in the Kepler Eclipsing Binary Catalog. In eight cases where the suspected source of contamination was truncated by the edge of the TPF definition or itself appeared to be a blend, the target was removed from our final sample.

3. Nine targets were culled after an investigation for periods and epochs that matched or were closely similar to another EB candidate in our sample. Ephemerides matching has previously proven an effective tool to root out false positives Kepler data (e.g. Batalha et al. 2013; Coughlin et al. 2014). With the small size of our vetting sample, it was possible to perform this search manually with large allowances for period error. No maximum arcsecond search radius was imposed in order to safeguard against potential instrument related forms of contamination that allow variable signals to afflict victim sources positioned on entirely different area of the detector, such as CCD Crosstalk and Antipodal Reflection (Caldwell et al. 2010; Coughlin et al. 2014). Although no clear cases of such exotic contamination were clearly identified, we identified nine cases of EPICs that were contaminating other observed EPICs.

4. Fourteen pairs of EPICs were noted during ephemerides matching and other vetting processes as apparent duplicate observations: the TPFs for each EPIC centre on the same apparent point source, but with minutely differing target catalogue coordinates listed. After reviewing the data to ensure that a dim field companion star was not actually the source of either EPIC proposal, we retained the EPIC that appeared best centred on the intended target and removed the other seven EPICs from our sample.

5. Twenty-eight targets were rejected as cases of mistaken identity with variable stars displaying periodic or near-periodic pulsations or rotational modulation.

6. Sixty-five flagged EPICs were rejected without conclusive final dispositions and are suspected to be largely spurious detections from two general categories: faint background EBs unresolved in the TPF pixel data and available all-sky data or artefact escapements common to our reduction processes.

The 207 survivors of these culling reviews were selected for further characterization and calculation of ephemerides. A serendipitous by-product of these analyses was the identification of two exoplanet candidates.

EPHEMERIDES, PERIODS, PHASE CURVES, CROSS-MATCHED SOURCE IDENTIFICATIONS AND META DATA

The periodicity for each target displaying three or more eclipses was initially determined based on the phase dispersion minimization (PDM) method (Stellingwerf 1978). The PDM variant method described by Plavchan et al. (2008) was also used to find periods. The PDM methods essentially fold the light curve at different periods and assign a ‘dispersion score’ for each detected period. These scores are then sorted, the minima associated with the best scores are picked and iteratively phase folded at increasingly narrower period ranges to find the best local solution. Our routine based on the original Stellingwerf (1978) work proved effective at correctly identifying the estimated period with no constraints on initial period range. However, we also utilized the PDM variant by Plavchan et al. (2008) as it assigns scores differently: a boxcar average is subtracted from the flux for each test fold and the residuals are squared and summed in order to provide ranking.

We evaluated mid-eclipse times by employing the methods of ‘KWM’ (Kwee & van Woerden 1956), barycenter (e.g. Oshagh et al. 2012) and least squares (e.g. Smith et al. 2012). The output parameters were then vetted by eye to ensure that BJD0 coincides with the deeper eclipse in cases where secondary eclipses are also present in the light curve.

We use these ephemerides to plot the EB phase curves, presented in Figs 1–12. The phase of each system is arranged for illustrative purposes such that the secondary eclipse, if present, is visible. The phase-folded light curves were then used to determine a system morphology. Morphology values range between 0 (detached) and 1 (overcontact). Morphologies are determined by first fitting a chain of four quadratic functions to the shape of the phased light curve and then determining the morphology through application of a Local Linear Embedding (Matijevič et al. 2012) routine. With the exception of Heartbeat binaries and EBs lacking three observed eclipses, morphologies are computed for the entire C0 sample.

All targets observed eclipsing in the C0 data set were crossed matched against all-sky catalogue listings in The International Variable Star Index (VSX; Watson 2006) and the SIMBAD data base (Wenger et al. 2000) within a 20 arcsec search radius using RA and Dec. coordinates from MAST. Results for 2MASS source identifications were found to be in good agreement with the K2-TESS stellar properties catalogue (Stassun et al. 2014), which includes EPIC to 2MASS cross-matches accurate to within 1 arcsec. For M35 superstamp related targets, we also performed a cross-check against recently published ground survey data from the Asiago Pathfinder for HARPS-N programme, which has performed a long-term, high-precision (5 mmag) search of open clusters M35 and NGC 2158 for variable stars (Nardiello et al. 2014).

These cross-matches allowed us to determine known stellar companions to our targets as well as alternate source identifications. We also checked and reviewed our results against the target proposal lists of the C0 GO proposals in order to determine which targets were known binary or multiple star systems. Our search of the GO proposals revealed that 110 of the 207 eclipsing targets we identified and retained in our final sample had been requested for K2 C0 observation by several different GO programmes as known or suspected binaries. An additional balance of 158 EPICs (approximately 60 per cent) proposed by these binary star related GOs were not found to eclipse or possess ellipsoidal variations that could be attributed to a binary as per our culling criteria (Section 3). However, a considerable portion of this balance may represent non-eclipsing objects [e.g. spectroscopic binaries, Ellipsoidal Light Variation (ELV) binaries, cataclysmic variables, etc.] and thus the balance of non-detections in our K2 C0 photometry is not unexpected.

Our analysis resulted in 97 new EBs detected in the C0 field. The orbital periods, ephemerides, morphologies and cross-matched identifications for all EBs in our sample are organized and presented in five tables. The 110 previously identified or suspected EBs in C0 are catalogued in Table 1. The 97 new EB detections are presented in Tables 2–4 (EPIC targets), 3 (M35 superstamp targets) and 4 (non-EPIC targets). The non-EPIC associated EBs were detected via their contamination of GO proposed target masks, creating a false positive EB signal. Due to the large size of C0 TPF stamps, we were able to extract and analyse light curves from these listed contaminators (see Section 3). A separate collection of false positive cases caused by direct PRF contamination (Bryson et al. 2010) between pairs of observed EPIC targets is listed in Table 5.

The sky plotted distribution for all 207 EBs in our final sample is presented in Fig. 13.

SPECTROSCOPIC OBSERVATIONS

We observed nine selected objects identified as interesting in our analysis (for details refer to Sections 7 and 8) on the nights of 2014 December 15, 19 and 20 using the Observatoire Astronomique du Mont-Mégantic's (OMM) 1.6 m telescope and long-slit spectrometer. The spectrometer employed a 1200 line mm⁻¹ grating which yielded a dispersion of 0.89 Å pixel⁻¹ and a wavelength range of ∼1800 Å (3750–5600Å). Depending on the brightness of each object, up to three exposures were co-added to improve the signal to noise; the total integration time for each object ranged between 20 and 90 min. All of the spectra were background subtracted, corrected for the quantum efficiency of the CCD detector and the reflectance of the grating, and corrected for Rayleigh scattering in the atmosphere before being co-added.

We classified the spectrum for each source by comparison to the Gray Digital Spectral Classification Atlas⁶ (Gray & Corbally 2009). We adopt a conservative uncertainty in the spectral classifications of two sub-classes. After a match in spectral type and luminosity class was found, we used the reference table in Boyajian et al. (2013) to estimate the fundamental stellar properties for each classification. The spectra are presented in Figs 14 (EBs) and 15 (planet candidates). The spectrum for each target is discussed where appropriate in the sections to follow.

Nine stars for which we obtained spectroscopic T_eff estimate overlap in the K2-TESS portal. Eight of these nine stars have T_eff estimates in the K2-TESS data base (Stassun et al. 2014). In all cases, the photometric K2-TESS T_eff determination is significantly cooler than the spectroscopic based estimate of T_eff. Two of these stars are also present in the photometric catalogue of all Tycho stars (Ammons et al. 2006), also having similar T_eff estimates as the K2-TESS estimates, though the Ammons et al. (2006) errors are typically on the order of several thousands of degrees (e.g. EPIC 202073097 has a |$T_{\rm eff} = 7418^{+4695}_{-447}$| K). We suspect that the reason for this disagreement is that reddening is not accounted for in the K2-TESS Catalog. Since the C0 field points to the galactic anticentre (l ∼ 190°, b ∼ 5°) and thus much through the galactic plane, reddening makes considerable differences in the target colours which are used in the colour–temperature relations by the K2-TESS Catalog.⁷ As such, any of the temperatures listed in this paper (not estimated from our spectra) should be used with caution. It is also worth noting that neither method makes correction for binarity, however this task is not trivial with the small amount of data at hand.

In any case, for this set of nine of spectroscopically observed stars, it appears that the K2-TESS dwarf/giant estimator method is reliable. Two stars that were observed spectroscopically were classified as likely giants (luminosity class III), and both of these were identified as giants also in the K2-TESS data base on the basis of the reduced proper motion (see Stassun et al. 2014).

GROUND-BASED PHOTOMETRY

With a time baseline of about 80 d, the K2 light curves can be used to identify EBs with periods out to ∼30 d. While K2 has greater detection capabilities than ground-based photometry, due to greater photometric precision and duty cycle, there is value in combining these results with those from ground-based photometric surveys and transit searches. Projects including SuperWASP (Pollacco et al. 2006; Hellier et al. 2011), HATNet (Bakos et al. 2004, 2013), XO (McCullough et al. 2005) and KELT (Pepper, et al. 2007; Pepper et al. 2012) have accumulated observations over many years, and provide greater time baselines than K2. The addition of longer time baselines to the K2 light curves makes it possible to search for eclipse timing variations (ETVs) or amplitude variations of the EBs.

To explore the potential of that approach, we have cross-matched the K2 EBs in Tables 1 and 2 with data from the KELT survey. KELT is a wide-field, small-aperture photometric survey, with a 23°× 23°field of view, and a non-standard broad V + I filter. Two of the currently reduced fields from KELT overlap with K2 C0 north of Dec. = +19°. The KELT light curves were reduced and extracted according to the procedures described in Siverd et al. (2012), and include about 7700–8600 observations from 2006 October to 2013 March. Additional observations after that time have been gathered but not yet reduced. Typical photometric precisions are 0.5 % RMS around V = 9^th magnitude, and 2 % RMS around V = 12^th magnitude. North of +19°, KELT has light curves for all K2 C0 targets brighter than 12^th magnitude, and has light curves for at least 50 % of the K2 C0 targets in the range 12 < V < 13.

The long time baseline and good photometric precision of the KELT light curves allow us to measure the properties of many of the EBs found in K2. A full analysis is ongoing, but we show an example of combining the KELT and K2 data in Fig. 16. Table 6 displays the overlap of targets with K2 C0 and the currently available KELT data in hand. Potential applications include searching for eclipse depth, duration, and timing variations indicative of outer companions, or placing upper limits on such variations.

NOTEWORTHY EBs

Noteworthy EBs within our final sample include 10 detached EBs with periods longer than 20 d (EPICs: 202064253, 202071945, 202072282, 202084588, 202086225, 202090938, 202091197, 202091404, 202092842, 202071645) and six candidates possessing single eclipses of significant depth but for which no period is yet determined (EPICs: 202060921, 202071902, 202072917, 202085278, 202135247, 202137580). We have also discovered six ‘Heartbeat binaries’ (e.g. Welsh et al. 2011; EPICs: 202060503, 202064080, 202065802, 202065819, 202071828, 202072282); see Fig. 17. The most eccentric binaries, identified from the phase difference of the observed primary and secondary minima, are presented in Table 7.

The following is a brief discussion on 10 notable EBs in our sample. Their full light curves are presented in Figs 18 and 19. Fig. 14 shows the spectrum of the object, if available (Section 5).

• EPIC 202073097. A GO proposed EB with a period of 0.97 d and associated with a point source of Kepler magnitude K_p = 9.6. Alternate designations for EPIC 202073097 include HD 251042 and 2MASS 06042191+2032032. The C0 light curve shows a second set of eclipses of similar depth and duration with a period of 5.33 d. The 5 d period is not observed in extractions from the halo pixel sources within the TPF for EPIC 202073097. It is unclear if this anomaly represents two independent binaries or a gravitationally bound pair of binaries in a quadruple system. Spectral analysis indicates a late A1 or A0 component is present, possibly evolved as Fe ii and Si ii are observed. This is consistent with The Henry Draper Catalogue classification of A2 (Cannon & Pickering 1918).

• EPIC 202073145. A GO proposed EB with a period of 1.52 d, 30 per cent depth, and K_p = 11.3. The light curve also displays an additional anomalous signal with a depth of 15 per cent and a period of 4.44 d. Alternate designations for EPIC 202073145 include TYC 1323-169-1 and 2MASS 06185463+2036038. It is unclear solely from a pixel-level review if the longer period represents a close background contaminator or a second binary in a quadruple system. The K2-TESS Catalog lists T_eff = 6208 K for this source; our spectrum indicates a considerably hotter B3V type star, which is assumed to be un-evolved due to the absence of O ii.

• EPIC 202088178. A new EB from this paper with a period of 2.37 d, 15 per cent depth, and K_p = 13.3. EPIC 202088178 is also known as 2MASS 06230702+1828149. An additional series of eclipses are observed with a shorter period of 0.49 d, and a depth of only 0.5 per cent. A pixel-level review was inconclusive but cannot rule out with confidence that the light curve contains two unresolved binary systems. The K2-TESS Catalog lists T_eff = 5270 K for this source. Our OMM spectrum indicates the star has a true temperature much higher with a spectral type of F5 II.

• EPIC 202092613. A new EB with two periodic signals: a 3.18 d period with a primary eclipse depth of 5 per cent, and a 0.25 d periodic modulation with an amplitude of 0.5 per cent.. The additional short-period signal is not resonant with the 3.18 d period, and we suspect the cause of this 0.5 per cent amplitude modulation is attributed to either pulsations or star-spots common to one of the binary components. We classify the OMM spectrum of this object as an A0V based on Balmer line strengths, weak Mg ii ratio and lack of Si ii in the OMM spectrum. If the observed short-period variability is driven by star-spots, it could indicate a rapid rotation period of ∼6 h. However, given the stellar effective temperature and ratio of eclipse depths it is more likely that the short-period signal represents pulsations.

• EPIC 202088387. A new EB with a period of 3.55 d, depth of 15 per cent, and K_p = 12.9. Alternate designations for EPIC 202088387 include 2MASS 06064169+2234547 and SDSS J060641.69+223454.8. The light curve also exhibits asymmetrical brightening events of ∼2 d duration with an amplitude ∼5 per cent, which we assume to be associated with nearly corotating star-spots or pulsations in one of the stellar components. The K2-TESS Catalog T_eff = 5652 K; we do not have a spectrum of this object to confirm this temperature.

• EPIC 202135247. A new EB (K_p = 14.4; also known as 2MASS 06195142+1747474) with unknown period – a single, flat-bottomed, 9 per cent eclipse depth with a duration of 35 h is observed. The K2-TESS Catalog lists T_eff = 3919 K. Our OMM spectrum suggests an earlier-type primary component of G1.5. While the G-band strengths in our spectrum could be representative of either a luminosity class of V or III, the eclipse profile suggests an evolved star and annular eclipse of a smaller stellar companion.

• EPIC 202137580. A new EB (K_p = 13.1; also known as 2MASS 06031044+2330174) with unknown period. The light curve has a single 7 per cent flat-bottomed eclipse with a 37 h duration. While it has a similar eclipse profile to EPIC 202135247 (above) which may suggest an evolved component, the observed impact durations are considerably different as evidenced by shorter ingress and egress coupled to a longer disc passage of 28 h. With our OMM spectrum, we classify the primary component as G0, much hotter than T_eff = 3755 K from the K2-TESS Catalog. The evolutionary state of this target is unclear as the strength of the G-band in our spectrum rules out a supergiant, but does support either luminosity class V or III. An unexplained long-term trend in the light curve is visible and may be associated with orbital modulation and a period of under ∼100 d.

• EPIC 202062176. An eccentric (Table 7), new EB also known as 2MASS 06095262+2030273 having a period of 4.35 d, eclipse depth of ∼3.5 per cent, and K_p = 11. Independently proposed by two GO programmes for massive stars and a third for exoplanet detection, this EPIC may be a member of NGC 2174/2175.

• EPIC 202072430. Also known as TYC 1330-2152-1 and 2MASS 06402451+1508324, EPIC 202072430 is one of the 10 longest period EBs identified in our sample with P = 20.03 d. This eccentric (Table 7) EB has a K_p = 10.9 and 45 per cent primary eclipse depths. Two orbital cycles are captured in the C0 data; the light curve exhibits additional quasi-periodic variability in the out-of-eclipse regions on the order of ∼8 h.

• EPIC 202068807. EPIC 202068807 (K_p = 12.01), also known as TYC 1878-947-1 and 2MASS 06204185+2317264 is a new, eccentric (Table 7), EB with a 4.24 d period. The light curve has primary and secondary eclipse depths of ∼12 and 4 per cent, respectively. This eccentric binary displays out-of-eclipse asymmetrical ellipsoidal variation at the 0.5 per cent level.

EXOPLANET CANDIDATES

We report the detections of transiting planet candidates in the K2 C0 data set:

EPIC 202072704, and 2MASS 06101557+2436535 (no EPIC assignment as of this writing). The light curves for these objects were initially detected via the processing methods described in Section 2. After a detailed search of the TPF to rule out obvious contamination effects from other sources, the light curves were normalized and detrended specifically for analysis using the Transit Analysis Package (Gazak et al. 2012) in order to make a preliminary validation of their sub-stellar nature. HAT-P-54b, recently discovered by Bakos et al. (2015), was also identified in our analysis, but we do not elaborate upon it here.

In an effort to better characterize the candidate host stars in our orbital fits, we obtained a spectrum of each of the new planet candidate host stars (excluding HAT-P-54b) with the OMM spectrograph (see Section 5 for details). The spectra reveal that the hosts are early-type stars. Our best-fitting transit model retains a sub-stellar nature for each of the transiting companions, and we consider them new K2 objects of interest worthy of further follow up. The candidates common to these systems are of particular interest as there are currently few confirmed examples in the literature of exoplanets transiting early-type stars (e.g. WASP-33, Collier-Cameron et al. 2010; Kepler-13, Shporer et al. 2011, 2014).

The orbital solutions for each C0 candidate are shown in Table 8 and the phase-folded transit fits are shown in Fig. 15 with the spectrum of each point source.

• EPIC 202072704. This host star candidate has a K_p = 11.4, with alternate designations including HD 263309 and 2MASS 06455102+1712250. We identify 12 transit events of 3.16 h duration, depths of 6588 ppm, and a period of 2.65 d. There are no apparent contaminating sources within the TPF halo. The OMM spectrum suggests the host star is an A3V, consistent with the Henry Draper Catalogue classification of A5 (Cannon & Pickering 1918). Using the tables from Pecaut & Mamajek (2013), we assume the host has a nominal radius of 1.7 R_⊙. A best-fitting transit model based on these parameters indicates the transiting object has a radius of 1.26 R_Jup(14.1 R_⊕) and orbital radius of 0.04 AU.

• 2MASS 06101557+2436535. Located within the first listed TPF (200000811) for the M35 star cluster super stamp, a transit signal is recovered from a V = 12.6 point source associated with pixel coordinates corresponding to column number 52 and row number 36. This source has been cross-matched to 2MASS 06101557+2436535, as well as KIC 27058 from the original Kepler Input Catalog, although it lies out of the Kepler field of view (Brown et al. 2011). The light curve for this source displays five transits with a period of 7.5559 d, and depths of 6928 ppm. We are able to recover two additional transits of acceptable quality in the early coarse point data. Using the OMM spectrum, we classify this source as an A2IV/V star. Assuming a stellar radius of 1.8 R_⊙ (Pecaut & Mamajek 2013), our best-fitting transit model results in a planet radius of 1.37 R_Jup(14.50 R_⊕) object orbiting at a separation of 0.08 AU. Further stellar characterization is needed to determine the evolutionary state of the host star, as its radius is quite uncertain.

In addition to these planet candidates, EPIC 202066192 and EPIC 202090723 were initially noted as potential candidates. However, EPIC 202066192, was ultimately rejected as a false positive due to a V-shaped eclipse profile, high impact parameter, and early-type A0III-IV OMM spectral classification. EPIC 202090723's spectral type is determined from our OMM spectrum to be F5V – perfect agreement with the F5V classification from McCuskey (1967). Although the best-fitting transit model resulted in a planet radius of ∼0.5 R_Jup, a single high-resolution TRES spectrum showed signs of a secondary in the line profile (D. Latham, private communication). Note that our spectral classification infers a source about a thousand degrees hotter than the K2-TESS Catalog estimate of T_eff = 5442 K, similar to other discrepancies identified in this paper.

ECLIPSE TIMING VARIATIONS (ETVs)

ETVs can be induced by the gravitational presence of a third object in a system (Conroy et al. 2014b). Particularly so for short-period binaries, the hierarchical triple system occurrence rate is estimated to be 40 per cent or greater (Tokovinin et al. 2006). The Kepler Eclipsing Binary Catalog provides comprehensive eclipse timings from which over 100 confirmed and candidate triple systems have been identified (see also Rappaport et al. 2013; Borkovits et al. 2015).

Using the periods and ephemerides derived in Section 4, we analysed the C0 light curves for a deviation from a strictly linear ephemeris within the threshold of K2's 30 min observation cadence. This analysis resulted in several C0 EBs that display intriguing low amplitude trends and sinusoidal variations. However, these observed effects could be driven by interactions that do not require a third object (e.g. dynamical mass transfer, quadrupole coupling, apsidal motion; Conroy et al. 2014b). We are further burdened by the limited baseline for C0, in that most third body ETVs exhibit an O − C residual period greater than 45 d (Rappaport et al. 2013). This makes it unlikely for a full ETV cycle to have been serendipitously captured for a given target. We conclude that our analyses returned no high-confidence ETV periods that clearly suggest the presence of a third body companion.

SUMMARY AND DISCUSSION

Despite the new challenges inherent in working with currently available K2 data, the overall quality of these light curves, when properly decorrelated, is still well suited for the detection of EBs. A total of 207 EBs are recovered by visual inspection of the K2 C0 data set. For the 97 newly identified EBs, we provide periods, ephemerides, and cross-matched catalogue identifications. Due to the shorter baseline of C0 compared to the Kepler data set, the balance of detected eclipsing systems is unsurprisingly dominated by short-period binaries, with a steep rise in EBs with periods of less than 5 d. Of 207 systems identified in the K2 C0 data, 58 have periods in excess of 5 d. The majority of these (42 systems) are previously undetected with ground photometry (Table 2). The Kepler Eclipsing Binaries Catalog has been expanded to include all EPIC EBs⁸ in the C0 field with three or more observed eclipses. The EPICs contaminated by neighbouring EBs are catalogued as false positives in the Kepler Eclipsing Binary Catalog, having identifiers associated with the EPIC target that led to their detection.

The period distribution for the K2 C0 EBs with orbital periods less than 25 d is displayed in Fig. 20. Due to unknown selection biases it is not possible to evaluate incompleteness or draw firm conclusions regarding the fractional occurrence of EBs found in the C0 data set. Overall occurrence rate for all EBs in our K2 C0 sample is noted at 2.5 per cent, which is slightly higher than the 1.6 per cent determined for Kepler. Despite this, we detect an underrepresented range of EBs with periods of less than 1 d; the cause of this shortage is currently unknown but likely to be influenced by GO target selection bias and the self-imposed limitations of our own selection processes. We further recognize that the decorrelation methods utilized in our survey have played a major role in our recovery ratios and overall completeness. While kepsff and keppca provide an excellent foundation from which to reduce K2 data, intrinsic astrophysical signals can be obfuscated or removed by such solutions and care must be taken in their application. Alternative approaches to mitigating the effect of spacecraft pointing jitter are being explored, such as those described by Aigrain et al. (2015) and Foreman-Mackey et al. (2015).

There are compelling incentives that encourage the community to analyse each incoming K2 Campaign data set in a diligent and timely manner in order to characterize new EB and exoplanet candidates, particularly in cases where the host star is bright or close. Many such systems are within reach from the ecliptic and K2 stands uniquely poised to provide high-precision photometry for a limited selection of targets with overlapping coverage common to the upcoming TESS (Ricker et al. 2014) mission, which will begin an all-sky survey little more than 1 yr after the currently scheduled conclusion of K2 operations. For such systems meeting TESS selection criteria, the K2 campaign baseline of 80 d will provide expanded pre-coverage that should serve to make prospective detections of long-period objects (e.g. objects that transit only once or twice in a K2 baseline) whose signals might conceivably manifest themselves as transient dimming events during ecliptic-orientated TESS observations. While K2 observations alone cannot fully characterize such long-period events, their initial identification will allow potential spectroscopic or photometric follow-up from ground based or in-orbit facilities.

In addition, any M or K dwarfs local to Earth (e.g. within 20 pc) that can be shown to be binaries or blended binaries via K2 photometry (or subsequent ground-based follow-up) constitute a limited yet important early sample from a population of high priority targets for the upcoming JWST mission. Owing to their reduced stellar radii, transits and eclipses occurring in such systems can yield significantly higher final average SNR values and in some cases allow limited transmission spectroscopy. However, by consequence, the deeper transits noted for such stellar types must be carefully vetted against the host of false positive possibilities. The JWST Continuous Viewing Zone will have some degree of overlap with K2 ecliptic fields, but the instrument will also possess a finite supply of thruster fuel. Thus, it is critical that no exoplanet follow-up observations be squandered on a previously unidentified EB or BGEB.

Note added in manuscript. Since our paper was essentially completed, we have learned of a related study of the K2 Field 0 by (Armstrong et al. 2015). The K2 Variable Catalog II lists 2619 Field 0 variable objects including 137 EBs, but used a qualitatively different detrending algorithm than we have employed and does not include full ephemerides or cross-matched identifications.

Funding for the Kepler and K2 Discovery missions is provided by NASAs Science Mission Directorate. TSB, DAF, JW and JRS acknowledge support provided through NASA grant ADAP12-0172 and ADAP14-0245. LN thanks the Natural Sciences and Engineering Research Council (NSERC) of Canada for financial support, and the staff at the Observatoire Astronomique du Mont-Mégantic for their technical assistance. 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 research has made use of the NASA Exoplanet Archive, which is operated by the California Institute of Technology, under contract with the National Aeronautics and Space Administration under the Exoplanet Exploration Program. This research has made use of the VizieR catalogue access tool, CDS, Strasbourg, France. The original description of the VizieR service was published in Ochsenbein, Bauer & Marcout (2000). This research has made use of the SIMBAD data base, operated at CDS, Strasbourg, France (Wenger et al. 2000). This research has made use of the VSX data base, operated at AAVSO, Cambridge, Massachusetts, USA. 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. This publication makes use of data products from the Wide-field Infrared Survey Explorer, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, funded by the National Aeronautics and Space Administration. The Digitized Sky Surveys were produced at the Space Telescope Science Institute under U.S. Government grant NAG W-2166. The images of these surveys are based on photographic data obtained using the Oschin Schmidt Telescope on Palomar Mountain and the UK Schmidt Telescope. The plates were processed into the present compressed digital form with the permission of these institutions.

The authors gratefully acknowledge Ball Aerospace, the Kepler/K2 team and the Guest Observer community for making the K2 mission possible.

Facilities: Kepler/K2, Observatoire Astronomique du Mont-Mégantic (OMM)

REFERENCES