The GALEX View of "Boyajian's Star" (KIC 8462852)

KIC 8462852, also known as "Boyajian's Star," is an unusual F3 dwarf in the Kepler field that has exhibited unexplained optical variability on a variety of timescales. The initial discovery was of several dramatic, short timescale (days) dimming events with amplitudes of up to 20% in the Kepler 30-minute cadence data (Boyajian et al. 2015). Though the Kepler mission (Borucki et al. 2010) obtained data at a 30-minute cadence for ∼4 years on this star, no definitive pattern or cycle was found, nor has any single explanation for this variability been accepted by the community (Wright & Sigurdsson 2016).

An analysis of archival optical photographic plates has found that KIC 8462852 may have additionally faded nearly 16% over the past century (Schaefer 2016). Such a precise measurement for a single star is difficult, and the result has been debated (Hippke et al. 2016). However, using the 53 "Full Frame Images" (FFIs) spread over the 4-year Kepler mission, Montet & Simon (2016) were able to trace the brightness of KIC 8462852 using an independent flux calibration. The resulting flux-calibrated FFI light curve showed definitively that KIC 8462852 faded by more than 3% over 4 years. A years-long timescale variability, with possible periodicity, has recently been confirmed with an analysis of archival ground-based optical photometry (Simon et al. 2017).

The short (days) and long (years) timescale variability discovered for KIC 8462852 has presented a unique set of observational constraints for any single model used to describe the system. For example, if variable dust extinction is responsible for both temporal features, then the dust must have a wildly variable density distribution on small spatial scales, and a small density gradient over large spatial scales. Searches for an infrared flux excess consistent with a foreground or circumstellar dust shell have found no strong detection (e.g., Marengo et al. 2015), further complicating attempts to attribute the variability to dust structures.

Since optical variability and infrared follow-up has not produced a robust explanation for KIC 8462852, further multi-wavelength studies are needed to constrain the nature of the long timescale fading and short timescale dimming. Multi-band photometric and spectroscopic campaigns are underway,¹³ which will provide an improved understanding of any future "dips." However, no multi-wavelength measurement of the mysterious variability for KIC 8462852 contemporaneous with the Kepler observations has been analyzed.

Archival photometry at ultraviolet wavelengths from the GALEX mission (Martin et al. 2005) spanning a range of timescales from seconds to more than a year is now available. This unique set of observations occurred during the Kepler mission, providing an independent constraint on the variability discovered in the Kepler photometry for KIC 8462852. The wide wavelength range probed by GALEX and Kepler also allows us to explore models of the variability based on dust extinction and thermal cooling.

The various GALEX data products used in our analysis are introduced in Section 2. In Section 3, we analyze the NUV data over 10–100 s timescales. In Section 4, we explore the long timescale evolution of KIC 8462852 between the 2011 and 2012 visits, and compare directly to the observed fading by Montet & Simon (2016). In Section 5, we discuss possible interpretations for the nature of KIC 8462852 that the combined Kepler and GALEX observations provide, including an estimate of the dust extinction properties necessary to reproduce the long timescale NUV observations. Finally, in Section 6, we summarize this work, and discuss the potential utility of GALEX in the study of other rare and unusual variable Kepler objects.

Time-tagged photon data has recently become available for GALEX (Million et al. 2016a), including a Python toolkit to search for and interact with this high-cadence data product called gPhoton (Million et al. 2016b). This allows us to resample the GALEX main survey data into any desired cadence. In the case of KIC 8462852, the primary GALEX survey obtained ∼1600 s of data during four separate visits spread across a ∼70-day baseline in 2011. These high-cadence data from 2011 are also coadded as part of the GALEX "GR6" data release (Bianchi et al. 2014). In this work, we analyze only the NUV data (${\lambda }_{\mathrm{eff}}=2315.7\,\mathring{\rm A}$), as KIC 8462852 is too faint in the GALEX FUV band. Note that the GALEX NUV band is similar in wavelength coverage to the Swift uvm2-band analyzed for KIC 8462852 by Meng et al. (2017).

As part of the GALEX Complete All-sky UV Survey Extension (CAUSE) program, 104 square degrees within the Kepler field were reobserved in the NUV, creating the GALEX-CAUSE Kepler survey (hereafter GCK). This survey occurred in 2012, and overlapped a portion of the Quarter 14 operations from the original Kepler mission. The GCK data was obtained using scan-mode observing that differed from the standard GALEX survey. A catalog of the integrated fluxes and uncertainties for 475, 164 Kepler targets observed in GCK, including for KIC 8462852, was made available by Olmedo et al. (2015). In the case of KIC 8462852, the GCK catalog utilizes 1413.8 s of integration in 2012. Unfortunately, since the observing mode differed from the standard GALEX survey, GCK data is not available for time-series analysis with gPhoton presently.

Within each of the four primary mission GALEX visits available for KIC 8462852, we searched for short timescale variability using gPhoton.¹⁴ While nanosecond optical variability has been investigated for this target (Abeysekara et al. 2016), few other studies have looked at variability on timescales shorter than the 30-minute cadence available with Kepler. The four GALEX visits in 2011 ranged from ∼70 to ∼1400 s in duration. Data for each visit was sampled at a 10 s cadence with gPhoton, as shown in Figure 1. Small amplitude variability is apparent in several of the visits, with coherent structure over durations of approximately 60–100 s. Computing a Lomb–Scargle periodogram using gatspy (VanderPlas & Ivezic 2015) on the entire gPhoton light curve, we find moderate power with a broad peak at around 80 s. This appears to be due to the ∼120 s observing cycle of the GALEX instrument in the standard "Petal Pattern" observing mode, and we believe it is not astrophysically significant.

Zoom In Zoom Out Reset image size

A periodic signal of 0.88 days was also found in the Kepler photometry, which was presumed by Boyajian et al. (2015) to be due to the rotation of starspots in- and out-of view on the surface of KIC 8462852. Each of the four GALEX visits shown in Figure 1 are too short to entirely capture this rotation signature. Our periodogram, computed using all four of the gPhoton light curves together, also does not show any signs of this 0.88-day period.

Since the standard GALEX data for this target was spread over four separate visits, we also examined the medium-timescale variability over ∼70 days. In Figure 2, we show the median flux from each of the four gPhoton-processed visits. The uncertainties shown are computed as the standard deviation in the 10 s sampled data within each visit, and are ∼10× larger than the statistical error on each visit's median flux. Though there is scatter between these four visits in Figure 2, no significant coherent variability is seen on this intermediate timescale with GALEX. Unfortunately, this 70-day time window also did not correspond to any of the previously identified dimming events from Boyajian et al. (2015).

Zoom In Zoom Out Reset image size

In Figure 3, we present the GALEX data for this target as observed in 2011 and 2012. The 2011 data represents the final GALEX GR6 catalog flux value for KIC 8462852 of 16.46 ± 0.01 mag from Bianchi et al. (2014), and is the integrated flux from all four visits described in Section 3. The 2012 data is from the GCK data from Olmedo et al. (2015), and measured an NUV brightness for KIC 8462852 of 16.499 ± 0.006 mag. Both apparent NUV magnitudes were converted to fluxes, and then normalized to the flux of the 2011 visit. This results in a measured NUV fading of 3.5% ± 1.0%.

Zoom In Zoom Out Reset image size

In Figure 3, we also show the slow fading discovered in the Kepler FFI's by Montet & Simon (2016). Note that the fact that the GALEX and Kepler FFI data are normalized to a relative flux of 1 around 2011 (MJD ∼ 55,700) is a coincidence. However, the observation that the GALEX flux decays coherently with the Kepler FFI flux over this time baseline is significant.

To determine what the typical variation in NUV flux is for F stars, we analyzed the variability for over 140,000 GCK stars in common with the Kepler Stellar Catalog (Mathur et al. 2017), observed, on average, 15 times over a time interval of 40 days in 2012. Note that the original GALEX data was not available over the entire Kepler footprint, and was not taken over a single 70-day observing window in 2011 for all targets as for KIC 8462852. A full analysis of this variability, while beyond the scope of our work here, is underway (D. Olmedo et al. 2017, in preparation). We found that stars with temperatures near KIC 8462852 (6750 K) have an average variation of 3.5%. GALEX was calibrated using the white dwarf LDS749b, which was repeatedly observed during normal operations. Figure 6 from Million et al. (2016a) finds visit-to-visit scatter of the photon-level data for LDS749b of 2%–3% using gPhoton. The GALEX calibration work done by Morrissey et al. (2007) finds the photometric repeatability for stars at the brightness of KIC 8462852 is ±1.5%. GALEX also provided an estimate for longer-exposure repeatability as a function of magnitude, which indicates an expected ∼3% uncertainty for our target.¹⁵ The NUV variation seen for KIC 8462852 is therefore not necessarily abnormal for stars of this spectral type.

As an aside, we also searched for long timescale variability for KIC 8462852 in the infrared from the WISE single-exposure source database using the W1-band (3.4 μm). This data set from the original WISE mission (Wright et al. 2010), and the NEOWISE extended mission (Mainzer et al. 2014) provides ∼2 day clusters of photometry spaced every 6 months due to the spacecraft roll pattern. Unfortunately, the GALEX observations for KIC 8462852 occurred during the observation gap between WISE and NEOWISE, and thus a direct comparison between the NUV and IR is not possible here. We found no clear long-term variability spanning 2009 through 2017 for KIC 8462852 in the W1-band. However, a more detailed comparison of this rich IR data set to the recently published work from Simon et al. (2017) and Meng et al. (2017) is warranted.

While many explanations for the nature of KIC 8462852 have been proposed, there is effectively no consensus on the nature of the years-long timescale fading (or variability) observed by Montet & Simon (2016) and confirmed here in the NUV. Critically, with only a single wavelength band available and no apparent characteristic timescale for this variation with the 4-year observing window, little can be constrained from the Kepler data alone. Metzger et al. (2017) have argued that the long timescale fading could be due to stellar atmosphere recovery after a planetary in-spiral, and possibly the short timescale dips are due to remaining debris. Montet & Simon (2016) note that the fading in the Kepler FFI's may be due to the transit of a dust cloud. However, none of these models definitively explain the long timescale variability observed in Montet & Simon (2016).

By combining the optical Kepler FFI light curve with the long timescale GALEX NUV data presented here, we can place the first multi-wavelength constraints on KIC 8462852. A natural model to compare the simultaneous variability in the NUV and optical is that of a dust cloud. Extinction by dust in the interstellar medium is well studied, and several models with varying dust compositions are available at these wavelengths. Regardless of where the dust originates (i.e., circumstellar versus interstellar), such extinction models are a useful path forward in exploring the fading of KIC 8462852.

To demonstrate the impact dust would have in these two bands, we computed the extinction in the GALEX NUV band that would be predicted given the fading observed by Montet & Simon (2016) within the 2011 and 2012 time windows observed by GALEX. We used a standard Cardelli et al. (1989) dust model with R_V = 3.1, computed using the Python code from Barbary (2016). The comparison of this R_V = 3.1 prediction with the flux decrease observed by GALEX is shown in Figure 4. The Cardelli et al. (1989) model overpredicts the fading found in the NUV, indicating the fading is more gray (less wavelength dependent) than a standard R_V = 3.1 dust model. The NUV decrease measurement is 1.7σ away from the R_V = 3.1 model, marginally inconsistent with "normal" interstellar dust as the culprit of the fading observed by Montet & Simon (2016), and supports a circumstellar origin. Given the lack of warm circumstellar dust detection by Thompson et al. (2016), this material must be very cool.

Zoom In Zoom Out Reset image size

However, the NUV response of dust models is highly dependent on grain composition. This can be explored in standard dust models by modifying the R_V parameter. We then tuned a dust model to match both the observed Kepler optical and GALEX NUV dimming by varying the R_V and specific extinction (A_V) parameters. To fit the fading in both wavelengths simultaneously requires a dust model with ${R}_{V}=5.0\pm 0.9$. While this is not typical for interstellar extinction material, such a high R_V has been reported, for example, around young protostars (e.g., Hecht et al. 1982). Competing dust models can produce significantly different NUV extinctions. For example, by modeling the Kepler and GALEX fading for KIC 8462852 shown in Figure 3 with a Fitzpatrick & Massa (2009) dust model, we find a best-fit parameter of ${R}_{V}=5.8\pm 1.6$. According to Simon et al. (2017), the dimming is weaker at redder optical wavelengths, broadly consistent with either dust extinction or temperature variations. A similarly large value for the reddening law (${R}_{V}\gt 5$) was recently reported for KIC 8462852 over a comparable timespan after the Kepler mission using follow-up near-ultraviolet, optical, and NIR monitoring by Meng et al. (2017). However, the long timescale variation in uvm2-band flux was comparable to the intrinsic light curve uncertainty in their comparison stars. A joint analysis of the GALEX data presented here and the Swift/UVOT data from Meng et al. (2017) may provide an improved understanding of the extinction properties of circumstellar dust around KIC 8462852.

Besides dust extinction, another simple model that can be invoked to fit the long timescale variations of KIC 8462852 is changes in the star's effective surface temperature. We carried out a toy model calculation of this cooling, assuming a quiescent blackbody temperature of 6750 K for the star, and flux variations in each wavelength band due to changes in blackbody temperatures. The Kepler FFI fading seen by Montet & Simon (2016) over the same time windows as our GALEX observations requires a temperature change of 41 ± 3 K. This temperature change, in turn, predicts a drop in the NUV flux of 5%, which is in weak tension with our observed flux change of 3.5% ± 1.0%.

We have undertaken the first exploration of the NUV variability for KIC 8462852, using GALEX data on a range of timescales. No significant variability is found on 10–100 s timescales using NUV light curves produced with gPhoton. Over four visits spanning 70 days in 2011, we also find no significant medium-term variability.

Comparing coadded data from 2011 with the follow-up GCK study of the Kepler field in 2012, we find that KIC 8462852 faded by 3.5% ± 1.0% in the NUV. This fading coincides with the slow variation reported by Montet & Simon (2016), and is the first verification that this star is variable in the NUV. A preliminary examination of the typical variance between the GALEX and GCK data shows an average NUV change of ∼3.5% for bright F-type stars. Thus we believe the NUV fading observed for KIC 8462852 is real, but not necessarily atypical at these wavelengths.

Though the long timescale NUV light curve is very sparsely sampled, the combination of NUV and optical wavelengths provides a powerful constraint on the nature of this slow dimming. We explored both dust extinction and thermal variations as possible causes for the long timescale fading. Our favored explanation from these NUV data is that KIC 8462852 may be occulted by a slowly changing column density of dust with R_V = 5.

Finally, GALEX provides us with a valuable new data set for use in the search for other objects of this class. We are able to expand the search criteria beyond dramatic short timescale events and slow dimming as observed with Kepler, to now include slow variability in the NUV. If other F-type stars are found with similar multi-wavelength variability over long timescales, it will shed light on the occurrence rate and the possible lifetime of "Boyajian's Star" type variables.

The authors wish to thank the anonymous referee, whose comments and suggestions improved the quality of this manuscript. We thank Joshua Simon for sharing a draft of their paper prior to publication.

J.R.A.D. is supported by an NSF Astronomy and Astrophysics Postdoctoral Fellowship under award AST-1501418.

Work by B.T.M. was performed under contract with the Jet Propulsion Laboratory (JPL) funded by NASA through the Sagan Fellowship Program executed by the NASA Exoplanet Science Institute.

D.J.W. has received funding from the European Research Council under the European Unions Seventh Framework Programme (FP/2007–2013)/ERC Grant Agreement No. 320964 (WDTracer).

Part of this research was carried out at the Jet Propulsion Laboratory, California Institute of Technology under a contract with the National Aeronautics and Space Administration.

Software: gPhoton (v1.28.2; Million et al. 2016b), gatspy (VanderPlas & Ivezic 2015), f3 (Montet et al. 2017).