DISCOVERY OF A NOVA-LIKE CATACLYSMIC VARIABLE IN THE KEPLER MISSION FIELD

Potential pulsating WDs were selected via their location in the reduced proper motion diagram based on photometric and astrometric data in the Kepler Input Catalog (Latham 2008). In particular, KIC J1924+4459 was flagged as a potential ZZ Ceti pulsator; its astrometric and photometric properties from the Kepler Input Catalog are given in Table 1; a finder chart is given in Figure 1. Initial time-series observations were obtained 2008 August 24 on the 3.6 m TNG telescope with the DOLORES imaging spectrograph. The imaging data were taken with the g-band filter with exposure times of 8 s; a short 0.98 hr run revealed flux variations of ≳10%. Photometric data were reduced using standard procedures in IRAF,⁹ including bias, flat-fielding, and sky subtraction. Aperture photometry was obtained for the target and several comparison stars. The flux ratios were obtained by dividing the counts of the target by the best combination (weighted mean) of three reference stars. These ratios were converted to fractional intensities by dividing the mean flux ratio. Barycentric corrections were also applied.

Zoom In Zoom Out Reset image size

Table 1. Astrometric and Photometric Properties of KIC J1924+4459

Parameter	Value	Date	Source
R.A. (J2000)	19h24m1082	...	KIC10a
decl. (J2000)	+44°59'349	...	KIC10a
μRA	24 ± 8 mas yr−1	1980.90	USNO-B1.0
μDec	34 ± 17 mas yr−1	1980.90	USNO-B1.0
Bp	15.22	1951.5195	USNO-B1.0
Rp	...	1951.5195	USNO-B1.0
Bp	15.59	1976.56	YB6b
Vp	15.79	1976.56	YB6b
Bp	15.30	1988.4695	USNO-B1.0
Rp	15.36	1991.5795	USNO-B1.0
Ip	15.72	1990.4463	USNO-B1.0
J	16.839 ± 0.137	1998.436	2MASS
H	⩾16.736	1998.436	2MASS
K	⩾16.536	1998.436	2MASS
g	16.182	2007.1c	KIC10a
g − r	−0.116	2007.1c	KIC10a
r − i	−0.091	2007.1c	KIC10a
r − z	−0.214	2007.1c	KIC10a

Notes. ^aKepler Input Catalog v. 10, Latham (2008). ^bUSNO Yellow–Blue Catalog v. 6, D. Monet (2009, private communications). ^cKIC10 photometry is a weighted average of data obtained 2006.281, 2007.412, and 2007.486.

Download table as:  ASCIITypeset image

Additional time-series data were obtained at the McDonald Observatory and the Bologna Astronomical Observatory. The McDonald observations were obtained with the Argos high-speed photometer on the 2.1 m Struve Telescope; Argos uses a prime-focus, back-illuminated frame-transfer CCD to obtain uninterrupted time-series photometry (Nather & Mukadam 2004). Data were obtained through a BG40 filter with individual exposure times of 10 s; runs of 2.57 hr and 2.06 hr were obtained on UT 2008 September 5 and 6, respectively. Data were reduced in the manner described by Mullally et al. (2005) and Mullally et al. (2008) using the IRAF package ccd_hsp (O'Donoghue et al. 2000).

The Loiano/Bologna observations were obtained on UT 2008 September 5 and 7 at 1.5 m Loiano telescope equipped with a BFOSC detector operated without a filter. Integration times were 40 s on September 5 and 45 s on September 7. A total of 332 and 269 frames were acquired during the 5.3 hr and 4.7 hr runs of the two nights, respectively. The data were reduced in the same manner as the TNG data.

The time-series data are shown in Figure 2.

Zoom In Zoom Out Reset image size

2.2.1. Short-term Variability

The light curves display large-amplitude variations of 0.12 mag that appear to be periodic. The light curve shape is not sinusoidal, but shows sharp rises and structured decays. We then performed a Scargle analysis of the Loiano data that revealed a strong peak at 94.4 μHz and a secondary peak at 188 μHz. The latter peak is close to twice the frequency of the strong peak. We therefore fitted a composite sinusoidal function with a fundamental period of 2.9422 ± 0.0016 hr (frequency ν = 8.1560 ± 0.0051 day⁻¹) and the first harmonic.

As the amplitudes of the variability were similar in both the unfiltered Loiano photometry and the BG40 McDonald photometry, and as the two data sets were interleaved, we combined the two data sets to more precisely constrain the periodicities in the data. Due to the large gap between the discovery photometry and our subsequent follow-up and the short duration of observation, we did not include the discovery TNG data in the combined analysis. A discrete Fourier transform (DFT) of the combined data (Figure 3) shows a strong peak very near the frequency of the Loiano data peak. Again, we simultaneously fit a fundamental frequency and its first harmonic

$$\begin{equation} f=A_0\sin \left(2\pi \frac{t-t_{0,\rm phot}}{P_{\rm phot}}\right) + A_1\sin \left(4\pi \frac{t-t_1}{P_{\rm phot}}\right)\,, \end{equation} \tag{ 1 }$$

where f is the fractional amplitude. For the fundamental frequency, we derive a photometric period P_phot = 2.9358 ± 0.0017 hr (ν = 8.1749 ± 0.0070 day⁻¹) with an amplitude A₀ = 0.0556 ± 0.0014 and t_0,phot(BJD) = 2454715.52547(74). For the first harmonic, A₁ = 0.0133 ± 0.0014 and t₁(BJD) = 2454715.57688(148). Figure 2 shows this fit overplotted on the photometric data.

Zoom In Zoom Out Reset image size

Although the formal errors indicate a significant difference between this period and that derived from the Loiano photometry alone, we believe the true error in the period is larger than the formal errors. In Figure 2, the model light curves for both periods are plotted over the observed photometry. Even when considering the TNG data, neither period is obviously preferred. In addition, light curves folded at both periods look qualitatively similar. We therefore consider a realistic estimate on the systematic error in the period to be 0.0064 hr, the difference between the two photometric periods. Additional observations will be needed to determine the photometric period more precisely. If the photometric period of 2.94 hr is the orbital period of the system, then KIC J1924+4459 lies within the 2–3 hr CV period gap.

No deep eclipses are observed. Small dips, which could be grazing eclipses, are visible just after minimum light in the Loiano data. These have been indicated in Figure 2 with vertical tick marks. The location of the tick mark at T = 24.175 hr after BJD 2454714.5 was set by visual inspection; all other tick marks are spaced by integer multiples of the photometric period of 2.9358 hr. These dips are consistent with the photometric period throughout both Loiano data sets, but are not seen in the McDonald data. A longer data set, such as that being obtained by the Kepler Mission, therefore will be required to address this issue.

After removing this long-period variability, the residual light curves from all four nights of data also show significant amplitude (∼5%) variability on time scales of ∼650 s and ∼1200 s. Significant power is visible at these frequencies in the DFTs of the combined data (Figure 3) and of each individual run. However, this variability is not fully coherent, even within a single observation, indicating that these are not non-radial pulsations on the surface of the WD primary.

2.2.2. Long-term Variability

Spectroscopic observations of KIC J1924+4459 were obtained with the Blue Channel spectrograph on MMT on UT 2008 September 24 and 25. On the first night, a single 300 s exposure revealed broad hydrogen absorption with narrow emission cores; this led us to an initial interpretation that this object could be a WD+M binary.

On the second night, significant additional spectroscopy was obtained when high winds precluded observations of the spectroscopic run's primary targets. Data were obtained with a 500 grooves per millimeter grating blazed at 5410 Å and a 1''-wide slit; no order blocking filter was used. Wavelength coverage extends from 3200 Å to 6300 Å. The resulting spectral resolution, measured from isolated night-sky emission lines, gives a Gaussian FWHM of 3.8 Å.

Nine 480 s exposures were obtained over 1.4 hr before winds forced dome closure, corresponding to nearly half of the suspected orbital period. In Figure 4, we show the average spectrum from these nine exposures. The spectrum is dominated by broad Balmer absorption with emission lines at the line core. He i λ4471 is also observed as a broad absorption line with an emission core. Numerous other weak He i lines are observed in emission only. Strong emission lines of He ii λ4686 and the C iii/N iii Bowen blend emission near 4648 Å are also present. Equivalent widths (EWs) of prominent absorption and emission lines are given in Table 2.

Zoom In Zoom Out Reset image size

Strong, narrow Ca ii K, and Na i D lines are observed in absorption. As the source is located near the galactic plane (b = +1346), these absorption features seem likely to be interstellar in origin. Using the dust maps of Schlegel et al. (1998), the estimated color excess due to dust extinction is E(B-V) = 0.144 mag, assuming all extinction is foreground to the system. Along the line of sight toward the CV, the total hydrogen column density is 1.21 × 10²¹ cm⁻² (Dickey & Lockman 1990), implying an upper limit E(B-V) ⩽ 0.18. The EW of the Na D lines measured from our spectroscopy is 0.57 ± 0.02 Å; based on the empirical relation of interstellar Na D strength to E(B-V) of Munari & Zwitter (1997), our measurement implies a significantly higher E(B-V) ≈ 0.45 ± 0.15. This suggests that KIC J1924+4459 is either reddened substantially by patchy interstellar extinction, or that significant Na D absorption is intrinsic to the system.

We searched for evidence of radial velocity variations by cross-correlating each individual spectrum with the first spectrum. Continuum, including the broad Balmer absorption, was fit and subtracted, so that the cross-correlation results in the emission-line velocity curve. These relative velocities are listed in Table 3, and the resulting radial velocity curve is shown in Figure 5.

Zoom In Zoom Out Reset image size

We detect small but significant radial velocity variations. Working under the assumption that these variations are intrinsic orbital motions, that the orbital period is equal to the primary photometric period of P_phot = 2.9358 hr, and that the orbit is circular, we use least-squares fitting to determine an orbital solution of the form:

$$\begin{equation} v_{r,{\rm rel}} = \gamma + K\cos \left(2\pi \frac{t-t_{0,\rm spec}}{P}\right)\,. \end{equation} \tag{ 2 }$$

The resulting fit (shown in Figure 5) gives K = 30.1 ± 11.8 km s⁻¹, t_0,spec(BJD) = 2454734.63588(383), and γ = −6.4 ± 9.2 km s⁻¹. The goodness of fit is very high, with χ²/ν = 0.4. We note that the photometric period may not be the orbital period, but the photometric modulation could be due to some other phenomenon such as a superhump. To confirm whether the photometric period traces the true orbital motion or a superhump period, longer spectroscopic coverage is needed.

The absolute systemic radial velocity γ_abs is obtained by cross-correlating the continuum-subtracted emission-line spectrum with an artificial zero-velocity Balmer emission spectrum; this velocity was corrected to the local standard of rest using the standard solar motion. This exercise gives γ_abs = −32.14 ± 9.2 ± 17.7 km s⁻¹, where the first error is the error arising from the orbital solution and the second error is the uncertainty in the absolute heliocentric velocity determination.

Defining zero spectroscopic phase as the red-to-blue velocity crossing, the time of zero phase T_0,spec(BJD) = 2454734.66646(383). Based on Equation (1) and including the accumulated systematic uncertainty in the photometric period determination, this corresponds to a photometric phase (t − t_0,phot)/P = 0.47 ± 0.30. The relatively large accumulated uncertainty in photometric phase therefore precludes precise comparison of these spectroscopic data with the photometry at this time.

Many CVs show different γ and K values for different emission lines due to their different origins within the system. Therefore, we also calculated the radial velocity curves subsets of the emission lines, including Hβ, the higher-order Balmer lines (Hδ–H8), He ii, and the Bowen blend line. The relative velocities for these subsets of lines are listed in Table 3 and plotted in Figure 5.

Radial velocity curves were determined by fitting these velocities with a sinusoid of a fixed period of 2.9358 hr. From Hβ velocities alone, the resultant K = 51.1 ± 4.8 km s⁻¹, significantly higher than the fit using all the lines, though the fit is significantly worse (χ²/ν = 4.4). Within the errors, γ is unchanged. The fit to the higher-order Balmer lines are, within the errors, consistent with the initial fit obtained using all emission lines.

The individual sinusoidal fits to He ii and the Bowen blend were significantly different than the initial and Balmer line fits. For He ii, K = 33.9 ± 13.8 km s⁻¹ and γ = -34.0 ± 11.8 km s⁻¹ with χ²/ν = 4.3. For the Bowen blend, K = 15.2 ± 9.1 km s⁻¹, γ = −24.4 ± 6.6 km s⁻¹, and χ²/ν = 9.0. Within the errors, these amplitudes are consistent with one another, yet of lower amplitude than the Hβ variations, as would be expected since higher-excitation lines generally originate much closer to the accreting WD. However, these fits are only slightly favored over the weighted means (constant velocity solutions) of −11.0 ± 3.6 km s⁻¹ (He ii, χ²/ν = 4.7) and −17.8 ± 2.7 km s⁻¹ (C iii/N iii, χ²/ν = 8.1). As the radial velocities of these high-excitation lines are observed to vary rapidly in some other CVs, particularly in some SW Sex systems, we again caution that additional observations are necessary and that these sinusoidal fits may not have any physical meaning.

The emission lines, including the Balmer series lines, are resolved in the individual spectra (Table 4). For comparison purposes, the measured FWHM of the unresolved [O i] λ5577 night skyline is included in the table. Mean velocity widths (FWHM) of the Balmer lines are 330 ± 20 km s⁻¹ after correcting for the instrumental resolution; He i lines are similarly narrow. The width of He ii λ4686 may be somewhat broader; ≈375 ± 15 km s⁻¹. In the summed spectrum, no evidence for double-peaked emission lines is seen.

The emission-line strengths of all emission lines, with the exception of the Bowen blend, are observed to vary significantly as a function of time (Table 2, Figure 6). In particular, the He i lines (λ4471 and λ5876) both vanish toward later times, and the He ii λ4686 line is significantly stronger in the first two exposures than in the later exposures; the Hβ emission line may show similar behavior. The Bowen blend emission has a mean EW of −1.29 ± 0.04 Å; all individual data points are within 2σ measurement errors of this mean value, so any intrinsic variability in the EW of this line must be ≲0.2 Å in amplitude.

Zoom In Zoom Out Reset image size

Figure 7 compares EWs for pairs of emission lines in four cases where a significant correlation is observed. Hβ and Hγ emission is correlated, as would be expected, though there is significant scatter at weaker Hβ strengths. This could be indicative of different emission mechanisms being observed at different phases, or of systematic problems in measuring the emission-line strengths in one or both lines. Hβ and He i λ4471 show a nearly linear correlation with EW(λ4471)/EW(Hβ) = 0.26. Likewise, the EWs of He i λ4471 and He i λ5876 are linearly correlated; EW(λ4471)/EW(λ5876) ≈ 0.36. Finally, the two spectra with strongest He ii emission also exhibit the strongest Hβ emission.

Zoom In Zoom Out Reset image size

Lastly, the strong He ii line might suggest that KIC J1924+4459 could be a source of X-ray emission. Searches of the available X-ray catalogs via the HEASARC archive fail to result in any coincident X-ray sources within 10' of the star.

To summarize our observations, we find that the star KIC J1924+4459 is a photometric variable with a period of ≈2.94 hr and shows spectral characteristics of CVs; hence it is a CV in the 2–3 hr period gap. Time-series photometry shows no deep eclipses. Quasi-periodic variations are also observed on time scales of ≈1200 s and ≈650 s. The spectrum exhibits Balmer absorption with resolved, single-peaked emission-line cores, as well as strong He ii and Bowen blend emission. Radial velocity variations are observed that are consistent with low-amplitude motion and have periods either close to the photometric period or close to half of the photometric period. However, the phase coverage is currently too small to study the spectral and radial velocity variability in detail.

KIC J1924+4459 is clearly an NL variable of the UX UMa class. The strong, persistent Balmer absorption features indicate that this is a system accreting at a high rate. The strength of UX Uma emission features varies from object to object (Dhillon 1996). Quasi-periodic short-term photometric variability is also a common feature observed in most NLs. KIC J1924+4459 is consistent with all of these descriptors. We roughly estimate the accretion rate via two methods. First, the small EW of Hβ emission implies a limit on the disk luminosity of M_V ⩽ 6, or a lower limit on the accretion rate of $\dot{M}\gtrsim 3\times 10^{-10}\,M_\odot \,{\rm yr}^{-1}$ (Patterson 1984). Based on magnetic braking calculation by McDermott & Taam (1989), the secular accretion rate for an orbital period of 2.94 hr is $\dot{M} = 1.1\times 10^{-9}\,M_\odot \,{\rm yr}^{-1}$. This should be considered a lower limit, as most CVs with periods ∼3–4 hr have higher secular accretion rates of $\dot{M}\gtrsim 2\times 10^{-9}\,M_\odot \,{\rm yr}^{-1}$ (e.g., Howell et al. 2001; Townsley & Bildsten 2003).

Archival photographic plate photometry suggests that this star may have been up to a magnitude brighter in the past than when imaged as part of the Kepler Input Catalog survey, hinting that this star may have been (or even still be) in a VY Scl-type low state. However, the data are sparse, and the historical magnitudes could be in error due to the close optical companion star.

We calculate a lower limit on the distance to KIC J1924+4459 using the archival 2MASS photometry, under the assumption that the J-band detection is due totally to the secondary (donor) star. Knigge (2006) calculates that the donor would be of spectral type M4.2 with M_J = 7.88 at the edge of the period gap. Accounting for the Galactic extinction from Schlegel et al. (1998), the intrinsic distance modulus is (m − M)₀ = 8.83 ± 0.14, or 583⁺³⁹₋₃₆ pc. If, instead, we adopt the reddening of E(B-V) = 0.45 ± 0.15 derived from the Na D EW measurements, the intrinsic distance modulus is 8.55 ± 0.19, or 513⁺⁴⁷₋₄₃ pc.

The photometric properties of KIC J1924+4459 are consistent with, though not exclusive to, NLs of the SW Sex class. The photometric period is within the CV period gap, and the light curve does not show deep eclipses. SW Sex systems account for 55% of CVs in the period gap (Rodríguez-Gil et al. 2007a), and 37% of SW Sex systems do not show eclipses (Rodríguez-Gil et al. 2007b). The time scale and amplitude of the quasi-coherent variability is similar to that observed in a number of NLs of the SW Sex class (Rodríguez-Gil et al. 2007a, 2007b).

One defining spectral characteristic of SW Sex stars are single-peaked Balmer and He i emission lines (Warner 1995), such as we observe in KIC J1924+4459. Additional defining spectral characteristics of the SW Sex stars include a higher-velocity emission S-wave and variable absorption components in the trailed spectrum. We are, unfortunately, unable to identify these features in our data due to the combination of the low velocity amplitude, the low resolution of these spectra, and the small phase coverage of these data. A longer set of time-series data at higher spectral resolution are needed to search for this diagnostic signature. Likewise, the phase delay in the red-to-blue crossing of the Balmer lines that is characteristic of many SW Sex systems cannot be tested with these data due to the inability to determine the absolute phase of the system in this short run of data.

In summary, we speculate that KIC J1924+4459 is a member of the rapidly growing SW Sex class of CV based on our data, which are consistent with common definitions of SW Sex stars. However, the data are not sufficient to make a definitive identification at this time.

The discovery of a new CV, and in particular a potential SW Sex-type system, in the 2–3 hr period gap is not only important to increasing the statistics of this emerging class of accreting CVs, but also to understanding the crucial role of angular momentum loss mechanisms that act at these periods (for details, see Townsley & Gänsicke 2009, and references therein). Systems in the period gap are believed to be transition systems. Angular momentum loss is believed to be driven by magnetic breaking for CVs above the gap and driven by gravitational radiation in and below the gap. The fact that many CVs in the gap, at least half of which are SW Sex objects, display features indicating high $\dot{M}$ when this is not expected by angular momentum loss theory is a problem that needs to be solved. Therefore, the study of these systems is crucial to a better understanding of CV evolution.

This system requires further careful observations in order to classify the CV securely. In particular, confirming whether KIC J1924+4459 is an SW Sex-class system will require time-resolved, high-resolution spectroscopy. (Spectro-)polarimetric and X-ray observations can provide evidence of the presence or absence of strong magnetic fields on the accreting WD. Long-term photometric monitoring and a more careful analysis of archival imaging may be able to determine if this system enters occasional VY Scl-like low states.

The identification of a CV in the field of view of the Kepler Mission provides the unprecedented opportunity to obtain long-term, uninterrupted photometric monitoring of such a system. First and foremost, coordinated observations combining Kepler data with ground-based spectroscopy and/or polarimetry could help in furthering our understanding of this system. Additionally, Kepler's superb photometric precision and long mission lifetime can provide a unique opportunity to search for very low-amplitude and low-frequency phenomena, and to open new avenues of data analysis for both this object and CV stars in general. Indeed, KIC J1924+4459 is targeted for short-cadence observations by the Kepler Asteroseismic Science Consortium during roll 3 of the Kepler Mission; these data will soon be available.