• NASA ADS Record
• Simbad Objects
• About Related Links

The detection of X-ray flares from HD 189733A either that are generated through magnetic reconnection events intrinsic to the primary star or that may be induced by a star–planet interaction leads to the possibility that radio emission associated with these events may be detected. Studies of solar and stellar magnetic reconnection provide the model for this supposition. The twisting of magnetic field lines causes reconnection events, which accelerate electrons both toward and away from the stellar surface. These coronal electrons spiral along magnetic field lines and emit bremsstrahlung-dominated hard X-rays and radio waves via the mildly relativistic gyrosynchrotron, plasma, and electron cyclotron maser emission (ECM) mechanisms. The electrons that impact the chromosphere lower in the atmosphere deposit their kinetic energy into the denser plasma, thereby producing soft X-rays and enhancing chromospheric emission (Benz & Güdel2010; White et al.2011).

Radio observations provide valuable information that complements observations of flares and additional types of stellar activity at other wavelengths. Radio emissions come from the rarefied layers of the stellar atmosphere and therefore probe flare conditions in the coronae of stars that would otherwise be inaccessible to remote observation. These observations allow us to directly examine the physics of nonthermal electron acceleration regions (Benz2017). Measurement of the flux density and circular polarization fraction at radio wavelengths can be used to determine the emission mechanism(s) involved, determine the wave propagation conditions, and reveal the magnetic field strength of the emitting region (e.g., Pick & Vilmer2008). The sign of the StokesV signature determines the direction of the magnetic field, which can be used to trace large-scale magnetic topology (e.g., Benz & Güdel2010) or, in spatially resolved images, reveal changes in magnetic field direction within active regions (e.g., Bogod & Yasnov2009). Radio emission mechanisms constrain the electron density in the emitting region (e.g., Route & Wolszczan2012). Finally, the drifting of radio emissions in dynamic spectra can reveal plasma motions within a magnetic field (e.g., Osten & Bastian2008).

Previous radio observations of the HD 189733 system sought low-frequency, ECM-induced radio emission from the auroral regions of the exoplanet, similar to that found at Jupiter (Treumann2006; Lecavelier des Etangs et al.2009,2011; Smith et al.2009). Instead, I searched for radio emission that may be associated with stellar flares or active regions and may be induced by star–planet interactions, as part of the ROME (Radio Observations of Magnetized Exoplanets) program. The system was observed at higher frequencies than previously used because ∼gigahertz frequencies are typical for the detection of radio bursts emitted by gyrosynchrotron and, potentially, ECM mechanisms associated with stellar flares and CMEs in the lower corona (Treumann2006).

In addition to the search for flaring radio emission, I also analyzed published photometric time series to examine how active regions on HD 189733A may be correlated with other signatures of potential star–planet interactions. Active regions are small areas of the stellar surface with strong, concentrated magnetic fields that host most of the stellar activity. At visible wavelengths, they are manifest as starspot groups, which darken the stellar photosphere due to their cooler temperatures. Typically, regions consist of two starspot groups: in the leading group starspots are geographically concentrated and of a single magnetic polarity, while the trailing group is more spread out and of the opposite polarity. Thus, the magnetic fields within an active region are generally closed. The emergence of magnetic flux in these regions is subject to magnetic reconnection that result in plasma heating, flares, and CMEs. Plasma heating in the chromosphere results in upflows of plasma in magnetically confined coronal loops that produce X-ray, EUV, soft X-ray, and radio emission (Aschwanden2005).

Since Pillitteri et al. (2014,2015) asserted that an active region migrates across the stellar surface in phase with the exoplanet orbit, we can test this hypothesis by searching for persistent starspot-hosting active regions that reappear at particular orbital phases. The starspots in these active regions would appear as local minima in the photometric light curve. Similarly, since active regions are sites of enhanced chromospheric activity, including flaring, the location of photometric minima will also allow us to test the hypothesis, suggested by Shkolnik et al. (2008), of increased chromospheric flaring induced by star–planet interactions at a preferred orbital phase. This investigation will, to my knowledge, constitute the first photometric search for signatures associated with star–planet interactions.

I observed the HD 189733 system using the 305 m William E. Gordon radio telescope at Arecibo Observatory on 2011 September 7, for ∼7.2 ks, the time it takes for the target to transit the fixed dish. I monitored the system with theC-band receiver operating at 4.25–5.25 GHz with the same instrumental setup that has been used in my surveys of magnetized brown dwarf radio emission (Route2013; Route & Wolszczan2013,2016). The antenna gain and system temperatures were approximately 8 K Jy⁻¹ and 30 K, respectively, at low zenith angles.⁴ TheC-band receiver has a half-power beamwidth of ∼1′ in both azimuth and zenith angles. The receiver collects dual-linearly polarized signals that the Mock spectrometer⁵ processes into full Stokes parameters. The Mock spectrometer consists of seven individually tunable field-programmable gate array (FPGA) equipped fast Fourier transform (FFT) boxes, each with a 172 MHz bandpass divided into 8192 channels that yield a frequency resolution of ∼20 kHz. The observing run consisted of 12 science scans collected at 0.1 s temporal resolution. Each 10-minute science scan is bracketed by 20 s of calibration-on/calibration-off scans that leverage a local oscillator.

A natively developed IDL software pipeline, as described in Route (2013), processes the signals from the Mock spectrometer. These routines compute theIQUV Stokes parameters, perform flux calibration and bandpass corrections, iteratively reduce radio frequency interference (RFI), and resample the data to ∼80 kHz spectral and 0.9 s temporal resolutions. Data analysis consists of examining StokesV dynamic spectra (time-frequency spectrograms) for bursts of left- or right-circularly polarized emission that do not correspond to previously identified RFI patterns. Examples of these spectra can be found in Route & Wolszczan (2013,2016). Due to the broadband nature of gyrosynchotron and ECM emission from stars and brown dwarfs, bursts of radio emission from HD 189733A should appear in more than one box, which aids in the dismissal of false positives.

I transformed the Modified Julian Dates (MJDs) of my observations into Heliocentric Julian Dates (HJDs) and then into host star rotational (E_Rot) and exoplanet orbital (E_Orb) cycles using the ephemerides provided in Fares et al. (2017):

The observations spanned HJD 2,455,811.508894 to 2,455,811.597663 and probedϕ_orbit = 0.568–0.608 (ϕ_orbit =0.0 during exoplanet transit), which includes a fraction of the orbital phase range for which Pillitteri et al. (2010,2011,2014,2015) potentially observed enhanced X-ray flaring. The experimental setup is only sensitive to rapidly varying StokesV circular polarization fractions of ≳10% owing to confusion limitations and calibration uncertainty in StokesI total intensity. However, this circular polarization sensitivity should be adequate to detect emission from gyrosynchrotron, ECM, and plasma emission mechanisms (e.g., Stepanov et al.2001; Topchilo et al.2010).

I analyzed the time series from three photometric data sets, including 21 days ofMicrovariability and Oscillations of Stars (MOST) satellite photometry collected in 2006 August (Miller-Ricci et al.2008), ∼31 days ofMOST data from 2007 July to 2007 August (Boisse et al.2009), and Automated Photoelectric Telescope (APT) and Wise Observatory photometric data from 2009 October to 2009 December and 2010 May to 2010 June (Sing et al.2011).MOST photometry was obtained in Direct Imaging mode through a single broadband filter (3500–7000 Å) that was binned in intervals of 21 s (Miller-Ricci et al.2008) or 101.43 minutes (Boisse et al.2009). APT, located at Fairborn Observatory, measured Strömgrenb andy photometry on approximately nightly timescales.R filter photometry acquired via the Wise Observatory was measured approximately every second night. From these three data sets, I measured the onset and termination of the local minima in the light curves, which were averaged to produce the time at which the active region was in the center of the disk. I then used Equations (1) and (2) to compute the rotational and orbital cycles associated with these active regions, as listed in Table1. The inherent precision of the data sets makes it unnecessary to convert the temporal measurements to HJD for comparison with other data sets.

No radio bursts were detected during my observations, which have a 3σ detection sensitivity of 1.158 mJy in StokesV. Although the theoretical 1σ sensitivity is ∼0.15 mJy over a ∼1 GHz bandpass, the erratic presence of RFI across individual boxes reduces this to a 1σ sensitivity of 0.386 mJy in the cleanest box. This limit on sensitivity corresponds to the ability to observe flares with energiesνL_ν ≥ 2.183 × 10²⁴ erg s⁻¹ for isotropic (nonbeamed) emission at 4.25 GHz. These results are 5–50× more sensitive than previous searches for flaring radio emission, with order-of-magnitude smaller integration times.

4.1.1. Radio Observations throughout the Main Sequence and Beyond

4.1.2. Incoherent Emission Processes: The Güdel–Benz Relationship

However, the radiation mechanisms potentially involved in any nonthermal emission processes provide some guidance in the interpretation of these results. Nonthermal emission, such as hard X-rays and radio, can be caused by either coherent or incoherent processes. We will start with an estimate of the anticipated radio luminosity from incoherent processes. The Güdel–Benz relationship describes the correlation between nonthermal, incoherent, gyrosynchrotron radio emission and the thermalized X-ray emission in flares witnessed on F–M stars, including the Sun. The Güdel–Benz relationship is given by

whereL_X is the X-ray luminosity,L_R is the radio luminosity, andκ varies by type of star but, in this case, is of order unity (Güdel & Benz1993; Benz & Güdel1994). Using the peak X-ray flare luminosity Pillitteri et al. (2014) reported (L_X =1.950 × 10²⁸ erg s⁻¹), the Güdel–Benz relationship yields a maximum peak radio flare luminosity at 4.25 GHz of 8.3 × 10²² erg s⁻¹. However, it is important to note that the Güdel–Benz relationship expectedL_X to measure soft X-rays down to 0.1 keV, whereas theXMM-Newton bandpass that Pillitteri et al. (2014) used spanned only 0.3–8 keV. This difference in expected energy ranges could result in an underestimate ofL_X for HD 1897333A by nearly two orders of magnitude (Benz & Güdel1994).

A reference star more similar to HD 189733A may better estimate the X-ray and radio luminosities involved in incoherent processes. Eri is another active K2V star with similar mass (0.86M_☉), radius (0.74R_☉), and rotation period (11.68 days) to HD 189733A, if somewhat smaller mean magnetic field strength (10–20 G), and should therefore have similar levels of activity (Jeffers et al.2014). Ness et al. (2002) give the maximum X-ray luminosity for Eri asL_X =2.09 × 10²⁹ erg s⁻¹ as measured by theChandra Low Energy Transmission Grating Spectrometer (LETGS) over a 0.07–2.5 keV bandpass. Inserting this luminosity into the Güdel–Benz relationship, we would anticipate a maximum flare luminosityνL_ν = 8.9 × 10²³ erg s⁻¹ for radio emission at 4.25 GHz.

Another means to estimateL_R is by leveraging the X-ray luminosity range determined for other BY Dra class stars. Figure 11 of Benz & Güdel (2010) gives their peak X-ray luminosity range asL_X ∼ 10^27.5–10³⁰ erg s⁻¹, which corresponds to 4.25 GHz radio luminosities ofνL_ν ∼ 4 × 10²¹–4 ×10²⁴ erg s⁻¹. Thus, these considerations indicate that for flares generated by incoherent gyrosynchrotron emission, as is typical in many stellar flares, my observations exclude only the most powerful flares.

4.1.3. Coherent Emission Processes: The Electron Cyclotron Maser

Alternatively, coherent emission processes, such as plasma and ECM emission, may preferentially beam radio emission toward Earth as their active regions rotate into view. Indeed, solar flares larger than C5 class always radiate some coherent radio emission (Benz2017 and references therein). The observational signatures of ECM emission include high circular polarization fractions, high brightness temperatures, and fine structure in the radio dynamic spectrum. For example, during Effelsberg observations of the dM3.5e flare star AD Leonis, Stepanov et al. (2001) detected an entirely right-circularly polarized, irregular sequence of pulses with peak flux densityF_ν ∼ 100 mJy. Assuming that the emitting region was the size of the stellar disk, they computed a brightness temperatureT_B ∼ 5 × 10¹⁰ K, while for reasonable flare loop sizesT_B ≥ 3 × 10¹³ K. Using this peak flux density and the distance to the star,d = 4.85 pc, the ECM mechanism would produce a flare with anisotropic (nonbeamed) radio luminosityνL_ν = 1.365 × 10²⁵ erg s⁻¹. Obviously, once the effects of beaming are taken into account, the real flare luminosity is much less. If such flares occurred on HD 189733A, their radio flux densities would be readily detectable at Arecibo Observatory. Thus, these observations exclude the presence of extremely large gyrosynchrotron flares and modest-sized coherent ECM flares in the lower corona of HD 189733A atϕ_orbit = 0.568–0.608 (Figure1).

Zoom InZoom OutReset image size

4.1.4. Detection Probability of Flaring

It is clear from my temporal analysis of the photometric data (Table1) that starspot features persist over several rotation periods and are better correlated with rotation phase than the orbital phase of the exoplanet. If the star–planet interaction from the exoplanet periodically, or quasi-periodically, enhances stellar activity, it may be supposed that this enhancement would result in an increase in starspots at certain orbital phases that would be manifest as photometric minima synchronized to a particular orbital phase. However, this effect is not observed, as can readily be discerned in Tables1 and3. We will analyze the statistics of these results, both separately and as part of the ensemble of multiwavelength observations of HD 189733A activity, in Sections6.1 and6.3, respectively.

Shkolnik et al. (2003) initially investigated the star–planet interaction by observing the chromospheric Caii H and K emission lines (3933.7, 3968.5 Å) of HD 179949. They found enhanced emission that appeared to coincide with the location of the subplanetary point (ϕ_orbit ∼ 0). The team then investigated the star–planet interaction in several additional exoplanet systems, including HD 189733. Keen interest in the magnetic properties of this system stems from theabstract of Shkolnik et al. (2008), which states that “variability in the transiting system HD 189733 is likely associated with an active region rotating with the star; however, the flaring in excess of the rotational modulation may be associated with its hot Jupiter.” The team collected four nights of spectra from the Echelle Spectropolarimetric Device for the Observation of Stars (ESPaDOnS) on the Canada–France–Hawaii Telescope (CFHT). They computed the mean absolute deviation ( forN spectra) for residuals of the normalized spectra containing the Caii K line. From this analysis, they found that “no correlation is seen between the planet’s orbit and the residuals to the rotational modulation,” as depicted in their Figure 11. However, they mention that the Caii K MAD shows a “clear increase in very short (≤30 minutes) activity atϕ_orb ∼ 0.8,” which is manifested in their Figure 12 as a maximum among their four nightly data sets atϕ_orbit ∼ 0.8.

Comparing Figures 11 and 12 reveals the perplexing behavior of this increase in the Caii K MAD, which measures chromospheric variability and flaring. If the maximum in MAD atϕ_orbit ∼ 0.8 represents an increase in chromospheric activity, there should be a rapid increase in residual Caii K emission above that indicated by the rotationally fitted model. However, close inspection of their Figure 11 indicates that the night with the most variation (and hence largest MAD) is the second set from the left, showing a large ∼6%decline in Caii K emission below the best-fit model to the rotationally modulated chromospheric emission. Such a precipitous decrease in chromospheric emission is difficult to explain; for comparison, it is even greater in magnitude than the ∼5% decline in Hi Lyα (1215.6 Å) flux absorbed by the evaporating exosphere of the planet as it transited the stellar disk (Lecavelier des Etangs et al.2010). Stellar flares generally exhibit the opposite behavior and result in a large increase in Caii K emission, as demonstrated by simultaneousKepler and Apache Point Observatory spectroscopic observations of flares on the dM4e star GJ 1243 (Silverberg et al.2016). On the Sun, a weakening in chromospheric Caii H and K emission is found at very quiet regions, while active regions exhibit increasing emission (Linsky2017). The peak-to-peak scatter in integrated residual Caii K flux, barring this single low measurement, is ∼0.03, the same as for the other nights. Shkolnik et al. (2008) analyzed several other spectral lines that are typically indicative of chromospheric activity, including Caii H, Ca IRT (8662 Å), Hα (6562.8 Å), and Hei D3 (5875.6 Å). Among these, only the other Caii lines were correlated with Caii K activity; the Hα and Hei D3 lines either were poorly correlated or were difficult to measure, respectively. Therefore, of the five spectroscopic activity indicators observed, only the three Ca measurements supported the assertion of enhanced chromospheric activity. Rather than the spectroscopic data demonstrating enhanced stellar chromospheric flaring phased with the orbital motion of HD 189733b, the low Caii emission atϕ_orbit ∼ 0.8 is instead consistent with an absence of flaring and the rotation of aquiet region into view (or an active region out of view), in contradiction to the hypothesis presented.

Subsequently, an arsenal of spacecraft and instruments conducted detailed, multiwavelength analyses of the HD 189733 system. Moutou et al. (2007) used ESPaDOnS to collect 10 nights of spectra over two observing epochs, in order to leverage Zeeman signatures to infer the star’s large-scale magnetic topology. Noting that previous work found rotation periods for HD 189733 ranging from 11.7 to 13.4 days, they suggested that differentially rotating bright and dark active regions at various latitudes observed over multiple epochs could explain this behavior (e.g., Winn et al.2007). They determined that Caii H and K core emission flux can vary by ±10% on timescales of ∼1 hr and that “longer-term variations in core emission are apparently not related to the orbital phase,” but were correlated to stellar rotation phase. Although their results do not support the presence of a star–planet interaction, they cautioned that denser temporal sampling of the magnetic activity as a function of both orbital and rotational phases would be required to conclude that the activity is entirely independent of HD 189733b’s orbital phase.

Similarly, Boisse et al. (2009) obtained 55 high-resolution spectra of HD 189733A with the SOPHIE spectrograph at the Observatoire de Haute-Provence from 2007 July to 2007 August and analyzed activity in Caii H and K, Hα, and Hei lines. They computed a battery of diagnostics, including the cross-correlation function (CCF) for all spectral lines, CCF bisector velocity spans (V_span) that measure line asymmetry, chromospheric line spectral indices, “observed-minus-calculated” (O–C) radial velocity residuals, and the measurement ofMOST photometry. Spectral indices for the Caii, Hα, and Hei lines were computed using Index = F_SL/(F_L +F_R), whereF_L (F_R) denotes the continuum flux in narrow windows on the left-hand (right-hand) sides of the spectral line for which a flux is measured (F_SL). The wavelength windows for theF_SL,F_L, andF_R components all varied based on the spectral feature being probed. The noisier Caii and Hei indices were much less useful for analysis than the Hα index, which was correlated with the stellar rotation but uncorrelated with exoplanet orbital motion. The O–C residuals,V_span,MOST photometry, and the location of two minima in the Hei index light curve indicated, as a first approximation, the presence of a single active region on the stellar surface that was phased with the stellar rotation period alone. Thus, the detailed, multifaceted analysis by Boisse et al. (2009) of HD 189733's chromospheric activity indicators also does not support the star–planet interaction hypothesis.

Fares et al. (2010) collected 44 spectra over 27 nights from 2007 June to 2008 June using the ESPaDOnS and NARVAL spectropolarimeters, the latter of which is located at the Telescope Bernard Lyot. They found that the Hα and Caii H and K indicators were modulated on 11.6–14.4 day and 11.9–12.2 day timescales, respectively. Differential rotation rates ranging from 11.94 ± 0.16 days at the equator to 16.53 ± 2.43 days at the poles explained both data sets well. To search for lower-amplitude signals from a potential star–planet interaction, they subtracted the best-fitting rotation solutions to these activity indicators and independently searched the Hα and Caii H and K residuals for periodicities. They hypothesized that any effect of the star–planet interaction would be observed as a periodicity of 2.5–2.7 days, corresponding to the beat period between the stellar rotation and exoplanet orbital periods. Only a single Hα observing epoch was within 1σ of the beat period, while none of the Caii H and K emission residuals conformed to this periodicity. The residual emission varied considerably at every orbital phase, indicating that intrinsic variability, not a star–planet magnetic interaction, was the cause.

Czesla et al. (2015) witnessed a flare on 2012 July 1 with the Very Large Telescope (VLT-UT2) Ultraviolet and Visual Echelle Spectrograph (UVES) during their study of center-to-limb variations of the HD 189733A disk. The onset of the flare coincided with a steep rise in Caii H and K flux at midtransit (ϕ_orbit = 0.005), followed by a decay phase that lasted until the end of the observation. This flare helps us to measure the intrinsic variability of HD 189733A and to evaluate whether its activity is enhanced near the subplanetary point (ϕ_orbit = 0.0).

During their Multiwavelength Observations of an eVaporating Exoplanet and its Star (MOVES) program (2013 June to 2015 July), Fares et al. (2017) used Zeeman Doppler Imaging (ZDI) of HD 189733A from the NARVAL and ESPaDOnS spectropolarimeters to measure a mean magnetic field strength that varied from 18 G in 2006 August to 36–42 G in 2015 July. Most of the energy was allocated to the toroidal field, while the weaker, radial field was found to be largely nonaxisymmetric. Although they did not witness a magnetic field polarity reversal associated with a magnetic activity cycle, they noted that the field evolved significantly on a monthly timescale. The nonuniform magnetic topology and temporal evolution of the stellar magnetic field caused order-of-magnitude variations in the magnitude of the field at the distance of the exoplanet, but the field never exceededB ∼ 0.1 G (e.g., their Figure 6). By comparison, the magnetic field strength at the top of solar flare loop arcades, derived from radio brightness temperatures and hard X-ray spectral electron energy spectral indices, isB ∼ 200–500 G (White et al.2011). Thus, ZDI results indicated that HD 189733A has strong intrinsic variability and provided an estimate of the magnetic field strengths that would be involved in any star–planet interaction. However, any exoplanet-induced magnetic perturbations are quite tiny compared to the magnetic field strengths required to initiate stellar flares.

In summary, despite five separate observing campaigns that searched for enhanced Caii H and K chromospheric emission that would be indicative of enhanced flaring or an exoplanet-synchronized active region, only a single campaign found evidence to support the hypothesis of a star–planet interaction at a preferred orbital phase. Three additional ZDI campaigns (Moutou et al.2007; Fares et al.2010,2017) determined that the magnetic field of HD 189733A evolves significantly on short timescales, but they found no evidence for a star–planet interaction.

Pillitteri et al. (2010) analyzed two epochs ofXMM-Newton observations of HD 189733 conducted with the European Photon Imaging Cameras (EPIC) pn at 0.3–8.0 keV. The ∼55 ks 2007 April 17 observation was recorded during primary eclipse (transit), while the ∼35 ks 2009 May 18–19 observation occurred during secondary eclipse. Although the authors noted a modest flare accompanied by spectral hardening at ∼30 ks (ϕ ∼ 0), alterations in the X-ray flux are unremarkable compared to variability throughout the time series. Pillitteri et al. (2010) witnessed anF_X = 1.3 × 10⁻¹³ erg s⁻¹ cm⁻² X-ray flare atϕ_orbit = 0.54 during the 2009 observation, which they speculated may “arise from active regions for which the activity is triggered by the complex magnetic interaction that should be present between the star and planet.” Pillitteri et al. (2010) linked their result to the potential Caii H and K detection of a star–planet interaction, noting that “Shkolnik et al. (2008) has also suggested some excess of flaring activity on HD 189733 in phase with the planet rotation superimposed to the main long lasting active regions on the stellar surface.” Follow-up ∼39 ksXMM-Newton EPIC observations on 2011 April 30 recorded another X-ray flare beginning atϕ_orbit = 0.52 with similar energy properties to the 2009 flare, but with longer duration (Pillitteri et al.2011). They concluded that “the recurrence of such flares is explained by the following scenario: an active region is present at the same location on the stellar surface of both observations. The magnetic interaction with the planet is inducing a flaring activity in this region.”

Swift X-ray Telescope observations simultaneous with UV exosphere observations (Section5.3) detected an X-ray flare in the 0.3–3 keV energy band ∼8.5 hr prior to transit, which tells us about the intrinsic variability of the source (Lecavelier des Etangs et al.2012). Poppenhaegar et al. (2013) observed HD 189733A during six transits from 2011 July 5 to 18 withChandra/ACIS and noted little X-ray variability for an active star and no large flares. Although they measured transit depths varying from 2.3% to 9.4%, in excess of the broadband transit depth of 2.4%, they argued that this effect was due to the existence of an extended exoplanetary atmosphere, not the transit of a permanent, large starspot or associated coronal hole. Thus, the observations by Poppenhaegar et al. (2013) do not support the existence of enhanced activity nearϕ_orbit ∼ 0, as might be expected in both tidal and magnetic models of the star–planet interaction.

AdditionalXMM-Newton observations on 2012 May 7 detected a small X-ray flare atϕ_orbit = 0.57 and a larger X-ray flare that resembled those spotted in 2009 and 2011, atϕ_orbit = 0.64 during a ∼62 ks exposure (Pillitteri et al.2014). Pillitteri et al. (2014) analyzed the temporal behavior of all X-ray flares reported in the literature, which represented ∼370 ks of observations fromChandra,XMM-Newton, andSwift. They reported “both an excess of flares at 0.45–0.65 phase interval and a deficit in the 0.9–0.1 range with significance >1σ, while theSwift flare is within 1σ of the expected number of flares.” I note that theSwift flare orbital phase is misreported in their Figure 9 and Table3 asϕ_orbit = 0.2, whereas it should beϕ_orbit = 0.84 since it occurs before transit (Lecavelier des Etangs et al.2012). This led Pillitteri et al. (2014) to hypothesize that “when the planet is at phase 0.55, a region on the surface leading about 70–75 deg from the subplanetary point emerges from the limb. This region could be the site of the X-ray flares and responsible for the enhanced chromospheric activity observed in Caii lines. Given the motion of the planet, it is expected that the active region should travel at the same rate of the orbital velocity of the planet on the stellar surface and produce strong flares while on the visible face of the star, or about when the planet is in the rangeϕ = 0.55–0.15.” Thus, Pillitteri et al. (2014) theorize that an active region synchronized to the exoplanet’s orbital motion moves across the stellar surface. However, it is noteworthy that the overabundance in flaring that they found at orbital phasesϕ_orbit = 0.52–0.65 does not correspond well to the putative chromospheric activity enhancement Shkolnik et al. (2008) found atϕ_orbit ∼ 0.8. Furthermore, the alleged flaring excess that Pillitteri et al. (2014) described is not statistically significant, nor does it fit logically into the scenario that they proposed, since an active region can hardly be responsible for adecline in flaring. Moreover, the case for an overabundance of magnetic activity at certain orbital phases is not compelling in light of the accumulated multiwavelength observations reported in the literature (Table2, Figure2).

Zoom InZoom OutReset image size

Following the Lyα detection of atomic hydrogen escaping from the exoplanet HD 209458b (Vidal-Madjar et al.2003), Lecavelier des Etangs et al. (2010) searched for similar signatures of an exosphere during the transit of HD 189733b. Using theHST Advanced Camera for Surveys (ACS) Solar Blind Camera, they observed three transits from 2007 June 10 to 2008 April 24. They measured a 5.05% ± 0.75% Lyα transit depth and detected a ∼30-minute flare at midtransit of the third epoch in Cii (1335.1 Å). This flare helps us to quantify the intrinsic variability of HD 189733A.

Pillitteri et al. (2015) observed HD 189733 with theHST Cosmic Origins Spectrograph (COS) on 2013 September 12 (ϕ_orbit = 0.500–0.626) in search of FUV signatures of the star–planet interaction that corresponded to their X-ray flares. They noted two distinct flaring episodes in the lines of Cii (1334.5–1335.7 Å), Nv (1238.8–1242.8 Å), Siii (1264.7 Å), Siiii (1201.0–1206.6 Å), and Siiv (1393.8,1402.8 Å) atϕ_orbit = 0.525 andϕ_orbit = 0.588 with significance of 8σ–10σ. They attributed these features to “a stream of gas evaporating from the planet is actively and almost steadily accreting onto the stellar surface, impacting at 70°–90° ahead of the subplanetary point.” This assertion of a steady accretion stream further elaborates on the persistent active region scenario proposed in Pillitteri et al. (2014). Although the X-ray observations of Pillitteri et al. (2010,2011,2014) are generally consistent with these scenarios, they are excluded by observations of the same orbital phase range at other wavelengths. A more thorough evaluation of this accretion scenario will be forthcoming in a later publication.

Previous searches for radio emission from the HD 189733 system sought beamed, ECM auroral emission from its orbiting exoplanet. The system was thought to be particularly promising since its “hot Jupiter” companion would encounter a dense stellar wind that would result in more powerful auroral emission at low frequencies. It was also thought that observing a decline in radio emission from the system during secondary eclipse would yield strong proof that exoplanet aurorae contributed to any radio emission. Smith et al. (2009) first observed the HD 189733 system for ∼5.7 hr on 2007 April 21 at 327 MHz in dual polarization mode using the Robert C. Byrd Green Bank Telescope. The 3σ sensitivity for the entire observation was 81 mJy. Although they detected 1.5σ and 2.0σ brightenings in 330–333 MHz and 335–339 MHz subbands, respectively, they noted that these were statistically insignificant.

Lecavelier des Etangs et al. (2009) conducted a ∼7.7 hr observation of the system on 2008 August 14 at 244 MHz (LL polarization only) and 614 MHz (RR polarization only) using the Giant Metrewave Radio Telescope. They reported a 3σ sensitivity threshold of 2 mJy at 244 MHz and 0.16 mJy at 614 MHz for the entire observation. This sensitivity decreases to ∼50 and ∼5 mJy, respectively, for a default integration time of 16.78 s, which is more useful for detecting flaring behavior. Lecavelier des Etangs et al. (2011) followed up on this effort with a ∼9 hr observation of HD 189733 on 2009 August 15 at a central frequency of 148 MHz. They collected RR/LL polarimetry that had a 3σ sensitivity of ∼25 mJy over an integration time of 339 s. A search for flaring activity on timescales of 1–30 minutes revealed no detections during either campaign.

These three observing campaigns were all centered onϕ_orbit ∼ 0.5 and would have been capable of detecting flares withνL_ν = 1.4 × 10²⁴ to 1.2 × 10²⁵ erg s⁻¹. Although the frequencies probed were aimed at determining the cyclotron frequency of auroral emission of the exoplanet, it is interesting to note that they are each useful in the context of solar radio emissions. Type III bursts from the Sun typically occur at frequenciesν ∼ 300 MHz, while type IV bursts have been observed atν ∼ 170 MHz andν ∼ 600 MHz (Pick & Vilmer2008). However, since every search for flaring radio emission atϕ_orbit = 0.41–0.61 ended in failure, these observations, together with my own, do not lend support to the notion of enhanced stellar activity at the orbital phase range Pillitteri et al. (2010,2011,2014,2015) indicated.

In a search for starspot-induced, quasi-periodic flux variations that would enable the measurement of the host star’s rotation period, Winn et al. (2007) measured Sloan Digital Sky Surveyz-band, Strömgrenb andy, orV- andI-band photometry that spanned eight transits and 2 yr of out-of-transit observations, as part of the Transit Light Curve project (2005–2006). The 1.3% peak-to-peak variation in their photometric light curves indicated a starspot covering fraction of ∼1% that changed substantially over each 13.4 ± 0.4 day rotation and thus lacked permanent features.

Pont et al. (2007) monitored HD 189733A withHST/ACS for three epochs spanning 2006 May 22 to 2006 July 14. They detected and characterized two photometric flux increases of ∼0.1% and ∼0.04%, respectively, that occurred during transit atϕ_orbit = 0.002 during epoch 1 andϕ_orbit = 0.007 during epoch 2. The first increase indicated that the planet occulted a large, cool starspot or starspot group that spanned ∼80,000 km in longitude and ∼12,000–165,000 km in latitude, depending on the temperature contrast between the spots and the surrounding photosphere. The second event consisted of a transit of a smaller active region. Photometry accumulated during the third epoch potentially indicated the transiting of a small starspot group during egress. These observations, therefore, rule out a persistent active region at the subplanetary point that may be induced by a magnetic or tidal star–planet interaction. Instead, they demonstrate that the starspots evolve with time and may have a decay timescale similar to the Gnevyshev–Waldmeier rule for sunspots (Solanki2003).

Miller-Ricci et al. (2008) analyzed six transit light curves collected during a 21-dayMOST broadband optical photometric monitoring campaign of HD 189733A. They found no indications that the exoplanet transited starspots >0.2R_J. They also compared their results to a “Croll et al. (2007) model,” which allegedly fit the HD 189733A light curve with two large permanent spots rotating at a 11.73-day period. Such a model was attractive since it would indicate the presence of star–planet-interaction-induced permanent magnetic structures on the stellar surface. Miller-Ricci et al. (2008) found that the transit events they observed would either occult the northern permanent spot in the missing Croll et al. (2007) model in four of six transits, which was not observed, or transit the large starspot south of the equator, which could not be discerned owing to uncertainties in the observations and unaccounted-for model. Although Moutou et al. (2007), Pont et al. (2007), Miller-Ricci et al. (2008), Shkolnik et al. (2008), and Sing et al. (2011) reference a two-starspot model for HD 189733A in Croll et al. (2007), these citations appear to be in error, as the paper neither mentions the detection of starspots nor presents a starspot model. Furthermore, Miller-Ricci et al. (2008) state that this model assumed a stellar rotational inclination anglei = 59°, whereas measurement of the Rossiter–McLaughlin effect yieldedi = 855 (Triaud et al.2009). Thus, mistaken citations, together with the measurement of the stellar rotational inclination angle, rule out the presence of the large, permanent starspots Miller-Ricci et al. (2008) depicted that may have been related to a star–planet interaction.

Sing et al. (2011) leveragedHST/STIS near-UV–optical spectra to characterize starspots near transits (ϕ_orbit = 0.995) that occurred on 2009 November 20 and 2010 May 18. The spectra indicated spot temperatures of ∼4250 K, which would constrain the size of the large active region in Pont et al. (2007) to be ∼12,000 km in latitude. These spot characteristics enabled them to reproduce the photometric light curve with 6–96 spots that cumulatively decrease the stellar flux by 1%–2.8%. Higher spot numbers may be supported by the exceeding rarity of spot-freeHST observations. These observations demonstrated that starspots observed at the subplanetary point are of normal size and evolve with time. Furthermore, they suggest that HD 189733A likely hosts many starspots and is intrinsically active.

Pont et al. (2013), through a comprehensive analysis of the HD 189733b transmission spectra stretching from UV to infrared wavelengths, confirmed a 1%–2% starspot fraction limit based on the failure of starspots to leave spectral signatures on stellar Na and Mg H lines. Their analysis also constrained any large, permanent starpots to be located at the poles and to have anomalous spectra that differ from a cooler stellar photosphere. The temporal properties of the minima in photometric light curves (Section4.2), which correspond to dark starspots, and the analysis of Pont et al. (2013) do not support the hypothesis that a persistent active region of high magnetic field strength exists nearϕ_orbit = 0.52–0.65 orϕ_orbit ∼ 0.8 (Shkolnik et al.2008; Pillitteri et al.2014,2015), nor do they support the permanent, large two-spot model of Miller-Ricci et al. (2008).

In addition to the Hα observations of the HD 189733 system by Shkolnik et al. (2003), Boisse et al. (2009), and Fares et al. (2010) described above, Cauley et al. (2017a) conducted a series of short-cadence, high-resolution, out-of-transit Hα observations from 2013 July 4 to 2016 September 19 to examine stellar variability on short timescales. These observations coveredϕ_orbit ∼ 0.0–0.25, 0.45–0.60, and 0.90–1.05. They found that relative changes in the Hα core flux varied most from median values before and after exoplanet transit. They hypothesized that this effect could be due to star–planet interactions near the subplanetary point, or could indicate the presence of absorbing circumplanetary material near the exoplanet, in agreement with the evaporating exosphere model of Lecavelier des Etangs et al. (2010).

Cauley et al. (2017b) examined seven Hα transits collected by the HARPS spectrograph at La Silla and the HIRES spectrograph at Keck Observatory from 2006 August 21 to 2015 August 4. Only two data sets showed significant correlation between Hα and Caii H and K emission, as occurs for solar activity (Livingston et al.2007). Thus, absorption by a planetary exosphere is more likely to contribute to the observed variability than stellar activity. However, a stellar activity model that includes spots, filaments, and faculae concentrated at latitudes ∼±40° that provide a covering fraction of ∼5%–10% could also explain the observed Hα variations. Such a model resembled the chromospheric activity of the active Sun and would not suggest any enhancement from star–planet interactions. Thus, although these Hα observations do not permit us to evaluate the hypothesis of enhanced chromospheric activity atϕ_orbit ∼ 0.8 (Shkolnik et al.2008), they do exclude persistent faculae atϕ_orbit ∼ 0 (Cuntz et al.2000) andϕ_orbit ∼ 0.52–0.65 (Cuntz et al.2000; Pillitteri et al.2011,2014,2015).

Independently, observations of stellar activity on HD 189733A do little to constrain the various claims of enhanced photospheric, chromospheric, transition region, and coronal activity. However, a holistic, multiwavelength approach to star–planet interactions strongly constrains the issue.

• 1.  
  Although Cuntz et al. (2000) hypothesized that “hot Jupiters” as a class may induce enhanced magnetic activity near the subplanetary point (ϕ_orbit ∼ 0), the X-ray and photometric results rule out a persistent active region near this orbital phase. Hα emission may indicate the presence of stellar activity near the subplanetary point, but this activity is similar to that found on the active Sun and would not be anomalous for an even more active star such as HD 189733A. Furthermore, a more likely explanation for Hα variability nearϕ_orbit ∼ 0 is the presence of the evaporating exosphere of the nearby exoplanet.
• 2.  
  Although Shkolnik et al. (2008) suggested that HD 189733A displays enhanced Caii H and K emission atϕ_orbit ∼ 0.8, the only support for this claim is to be found in theabstract of that paper and not within the data itself. Furthermore, five independent Caii H and K observing campaigns, three ZDI campaigns, and photometric monitoring campaigns failed to find any anomalous flaring or spotting at this orbital phase.
• 3.  
  Although Pillitteri et al. (2010,2011,2014,2015) reported that HD 189733A exhibits enhanced X-ray and FUV activity atϕ_orbit ∼ 0.52–0.65, Hα observations do not indicate any abnormal chromospheric activity, radio observations did not detect any significant coronal flaring, and photometric monitoring failed to find persistent or recurring active regions at these preferred orbital phases.
• 4.  
  Although the results of Miller-Ricci et al. (2008) may be suggestive of persistent spots associated with a star–planet interaction, other photometric results exclude their model. Furthermore, the inclination angle,i, that the model relies on has been falsified by measurements of the Rossiter–McLaughlin effect.

Moreover, the various claims of enhanced stellar activity in individual wavelength regimes at preferred orbital phases are excluded by multiwavelength observations that fail to observe related signatures of enhanced activity (i.e., flaring, starspots, faculae). Furthermore, the competing claims are not consistent with each other.

Using photometric data fromMOST, APT, and Wise, I found the timings, rotational cycles, and orbital cycles of 17 photometric minima that correspond to pronounced active regions in the photosphere. Leveraging the method described above on the data in Table1 results in the statistical analysis listed in Table3. Significantly, this analysis shows that the measured number of active regions does not differ from active regions randomly distributed in orbital phase by more than 0.87σ. When the data set is reduced so that photometric minima are located at least one rotational period apart, there is still no statistically significant preferred orbital phase range. Thus, the star–planet interaction does not enhance starspot formation in the photosphere of HD 189733A.

However, the interpretation of photometric data is not without its difficulties. Decreases in photospheric flux represent a higher-than-average covering fraction of cool, dark starspots and active regions. On the other hand, gradual increases in the photometrically measured photospheric flux denote regions with relatively few starspots and active regions. Sudden sharp, ∼tens-of-minutes-long increases in photometric flux denote white-light stellar flares (e.g., Doyle et al.2018). However, a stellar photosphere entirely devoid of starspots and active regions is indistinguishable from an axisymmetric distribution of these features from photometry alone. A stellar photosphere may be dotted with a multitude of small starspots and active regions, but because of their spatial distribution, it will be unclear to photometric instruments whether flares occurred near starspots. Yet a darker hemisphere should, overall, have more spots and active regions, thus increasing the likelihood of flaring activity. Thus, HD 189733A may well have flaring active regions located at preferred orbital phases, but sometimes particularly dense starspot distributions confound our observations.

Another consideration is that starspots evolve over both short and long timescales. Large sunspots evolve in accordance with the Gnevyshev–Walkmeier rule and endure for up to several months; starspots on HD 189733A may evolve similarly and should therefore exist for several rotation periods (e.g., Solanki2003). New active regions are also likely created during this time interval, while others disappear. Differential rotation will alter the photometry of a particular rotational phase as starspot groups are carried to different longitudes at differing rates as a function of their latitudes (e.g., Hall1976). Finally, long-term (≳year), phased photometric measurements of active regions are likely not comparable owing to the existence of stellar magnetic activity cycles that extend from F stars with convective outer envelopes potentially to fully convective stars (Baliunas et al.1995; Route2016). These alter the degree of spottedness and the latitudes at which they emerge. All of these considerations serve to complicate the analysis presented here. However, if either temporal evolution of the starspots or their spatial distribution can effectively confound this analysis, it implies that the star–planet interaction is exceedingly weak.

The association of stellar flares with starspots and active regions is the subject of debate. On the Sun, the best-studied laboratory for the observation of stellar activity, Lee et al. (2012) noted that the probability of observing flares markedly increased with sunspot area, which would imply that the darkest hemisphere of a star would likely emit the most flares. On the other hand, studies of correlations between starspots and flares on other stars have not produced straightforward results. Walkowicz et al. (2011), in their analysis of flares from 373 K and M dwarfs in theKepler Quarter 1 data, found that larger photometric variability is correlated with greater energy release from flares. Silverberg et al. (2016) examined flare occurrence rates and energies as a function of rotation phase inKepler photometry of the dM4e star GJ 1243 and found no clear correlation. However, they suggested that the star may host a long-lived polar starspot that, given the low inclination anglei ∼ 32°, would result in little change of the spot distribution with rotational phase. Doyle et al. (2018) analyzedKepler K2 photometry of 34 M0–L1 dwarfs and found that both low- and high-energy flares were uncorrelated with photometric minima and maxima. It is important to note that in each of these studiesKepler photometry is only sensitive to high-energy flares that would correspond to the most powerful observed on the Sun. Nevertheless, despite these concerns about relations between starspots and flares, Pillitteri et al. (2014,2015) explicitly posited an active region synchronized to a particular orbital phase, and the chromospheric variability of Shkolnik et al. (2008) may be related to active regions. Yet this statistical analysis demonstrates that the photometric minima do not correspond to either of these hypothesized orbital phase ranges.

The orbital phases of the X-ray, UV, Caii H and K, Hα, photometric, and radio activity listed in Table2 can likewise be analyzed for the signature of a star–planet interaction. For the purposes of this analysis, the claimed activity of Shkolnik et al. (2008) and Pillitteri et al. (2010) is included (see Table2 notes), but the Hα variability detected by Cauley et al. (2017a) is assumed to be from the evaporating exosphere and is therefore excluded, as are my radio and photometric observations. Our statistical analysis of these 18 flares and starspots represents a fourfold increase in observations over the analysis presented in Pillitteri et al. (2014) and is presented in Table4. However, despite the increase in observed activity, there is little evidence for a star–planet interaction, as the measured activity does not deviate from random activity by more than 1.408σ.

Observations that investigate the nature of the exoplanet HD 189733b and its evaporating exosphere have led to an abundance of observations near transit, atϕ_orbit ∼ 0.9–1.1. Similarly, the suggestion that enhanced activity may be induced by tidal forces motivated searches atϕ_orbit ∼ 0.5 and the subsequent detection of some X-ray flares near this orbital phase (e.g., Pillitteri et al.2010) spurred additional research near these orbital phases. However, it is clear that observations recorded in the literature have poorly sampled large portions of orbital phase, particularlyϕ_orbit ∼ 0.2–0.4. Thus, it becomes readily apparent that the more certain orbital phases are observed, the more stellar activity is witnessed. However, the amount of activity observed near certain orbital phases no longer seems outstanding, and the literature provides little statistical evidence for a star–planet interaction.

Finally, the entire ensemble of existing multiwavelength observations, spanning X-ray to radio wavelengths (Tables1 and2), is compiled and depicted in Figure2. Perhaps one of the more important contributions to the present study of stellar activity on HD 189733A is that my measurement of the local minima inMOST, APT, and Wise time series samples all orbital phases uniformly. Thus, these measurements of the location of concentrations of active regions provide much-needed activity coverage atϕ_orbit ∼ 0.2–0.4 and considerably improve the statistics of stellar activity in orbital phase regions that ought to beunassociated with star–planet interactions.

The measured activity in each bin does not deviate from the anticipated activity by more than 0.589σ, thereby demonstrating that the observed flaring and spotting are not statistically different from an average rate across orbital phase (Table5). These results improve on the statistical power of the findings of Pillitteri et al. (2014) in their Table 3 by a factor of 8.75. However, these results cast doubt on the accompanying textual description presented in that paper and reproduced in Section5.2. Thus, this statistical analysis does not support an overdensity of stellar magnetic activity that results from either magnetic or tidal effects from the orbiting planet atϕ_orbit ∼ 0 (Cuntz et al.2000),ϕ_orbit ∼ 0.52–0.65 (Pillitteri et al.2010,2011,2014,2015),ϕ_orbit ∼ 0.8 (Shkolnik et al.2008), or as described by the spot model of Miller-Ricci et al. (2008). Statistical arguments alone cast significant doubt on the existence of a star–planet interaction in the HD 189733 system.