MOVES – I. The evolving magnetic field of the planet-hosting star HD189733 Free

Abstract

HD189733 is an active K dwarf that is, with its transiting hot Jupiter, among the most studied exoplanetary systems. In this first paper of the Multiwavelength Observations of an eVaporating Exoplanet and its Star (MOVES) programme, we present a 2-yr monitoring of the large-scale magnetic field of HD189733. The magnetic maps are reconstructed for five epochs of observations, namely 2013 June–July, 2013 August, 2013 September, 2014 September and 2015 July, using Zeeman–Doppler imaging. We show that the field evolves along the five epochs, with mean values of the total magnetic field of 36, 41, 42, 32 and 37 G, respectively. All epochs show a toroidally dominated field. Using previously published data of Moutou et al. and Fares et al., we are able to study the evolution of the magnetic field over 9 yr, one of the longest monitoring campaigns for a given star. While the field evolved during the observed epochs, no polarity switch of the poles was observed. We calculate the stellar magnetic field value at the position of the planet using the potential field source surface extrapolation technique. We show that the planetary magnetic environment is not homogeneous over the orbit, and that it varies between observing epochs, due to the evolution of the stellar magnetic field. This result underlines the importance of contemporaneous multiwavelength observations to characterize exoplanetary systems. Our reconstructed maps are a crucial input for the interpretation and modelling of our MOVES multiwavelength observations.

1 INTRODUCTION

Hot-Jupiters (HJs), i.e. gas giant exoplanets orbiting close (≲0.1 au) to their host stars, are useful laboratories to study the complex interactions (e.g. magnetospheric, tidal, ionization) between exoplanets and their host-stars. Because of the short orbital distance of HJs, interactions are expected to be strong and could potentially be detected with current instrumentation. The high stellar XUV fluxes that HJs are subjected to can lead to enhanced heating and atmospheric escape (e.g. Lecavelier Des Etangs 2007; Davis & Wheatley 2009; Murray-Clay, Chiang & Murray 2009; Lammer et al. 2011; Jackson, Davis & Wheatley 2012; Bourrier & Lecavelier des Etangs 2013b; Koskinen et al. 2013), which can be probed through transmission spectroscopy technique (Vidal-Madjar et al. 2003; Fossati et al. 2010; Lecavelier Des Etangs et al. 2010; Ehrenreich et al. 2012; Lecavelier des Etangs et al. 2012; Bourrier et al. 2013a). In addition to the extreme radiation environment, HJs are also immersed in stellar winds, whose physical characteristics are unparalleled to those felt by the planets in our own Solar system (Preusse et al. 2005; Grießmeier et al. 2007a; Vidotto et al. 2009, 2015). The extreme particle and magnetic environments of the stellar wind could lead to powerful reconnection events between the stellar and planetary magnetospheres (e.g. Ip, Kopp & Hu 2004), which may have the potential to enhance the stellar activity (Cuntz, Saar & Musielak 2000; Shkolnik, Walker & Bohlender 2003; Shkolnik et al. 2008; Scandariato et al. 2013). In addition, as a result of the interaction between the planet's magnetosphere and the coronal material of the star, bow shocks might also form around HJs (e.g. Vidotto, Jardine & Helling 2010; Cohen et al. 2011; Llama et al. 2013; Matsakos, Uribe & Königl 2015), explaining asymmetric UV transit features (Fossati et al. 2010; Vidotto et al. 2010). Radio emission resultant from the interaction between the stellar magnetized wind and a magnetized exoplanet is expected to be several orders of magnitude larger than that of the largest emitter in our own Solar system, Jupiter (Grießmeier, Zarka & Spreeuw 2007b; Zarka 2007; Jardine & Cameron 2008; Fares et al. 2010; Vidotto et al. 2012, 2015; See et al. 2015). A varying stellar magnetic field has also implication for the atmosphere of the orbiting planet leading to variability of the cosmic ray flux reaching the planetary atmosphere (Helling et al. 2013; Rimmer & Helling 2013). Cosmic rays have been studied as one example for external ionization sources that also affect the chemical compositions by possibly opening reaction paths to carbohydrate molecules (Rimmer, Helling & Bilger 2014a; Rimmer, Stark & Helling 2014b).

HD189733 is an ideal system to study these different types of interactions. At a distance of 19.3 pc, the bright (V=7.7) and active K2V star hosts a transiting HJ orbiting at 0.031 ± 0.001 au, i.e. at an orbital distance of 8.84 ± 0.27R_⋆ (see Table 1). The interactions within this system have been the subject of many studies in the literature (e.g. Smith et al. 2009; Pillitteri et al. 2010; Berdyugina et al. 2011; Lecavelier Des Etangs et al. 2011; Ben-Jaffel & Ballester 2013; Bourrier et al. 2013a; Bourrier & Lecavelier des Etangs 2013b; Poppenhaeger, Schmitt & Wolk 2013; Louden & Wheatley 2015; Barnes et al. 2016; Bott et al. 2016; Brogi et al. 2016). Because of its active host, HD189733b is in fast-changing radiation, particle and magnetic environments. Recent UV observations of the planetary transit showed that the properties of the hydrogen exosphere surrounding the planet are varying over time (Lecavelier des Etangs et al. 2012). An X-ray flare was detected 8 h prior to the UV detection of atmospheric escape, suggesting that variations in the XUV emission of the host-star and/or its magnetized outflowing plasma might have been the cause of the observed variability.

Variations in the quiescent stellar wind are also expected for this system. Through spectropolarimetric observations performed by Moutou et al. (2007) and Fares et al. (2010), it was shown that the large-scale magnetic field of HD189733 varied for the observed epochs. Recently, using the large-scale magnetic maps reconstructed by Fares et al. (2010), Llama et al. (2013) showed that the stellar wind of HD189733 presents inhomogeneities at the position of the planet, which could cause short time-scale variations in the UV transit light curve (on the order of an orbital period), as well as longer time-scale variations (on the order of a year, due to the change in the stellar magnetic field). These authors concluded that multiwavelength data acquired simultaneously would provide the best tool for a comprehensive characterization of the system.

In this context, we started a multiwavelength observational campaign of this system, in the frame of the MOVES collaboration (Multiwavelength Observations of an eVaporating Exoplanet and its Star, PI V. Bourrier). Observations of the star and the planet were obtained at similar epochs with ground-based and space-borne instruments in X-rays with Swift and XMM–Newton, UV with HST and XMM–Newton, optical spectropolarimetry with NARVAL and ESPaDOnS, and radio with LOFAR. In this (first) paper of this collaboration, we use optical spectropolarimetric observations to reconstruct the surface magnetic field of HD189733 at five different epochs, contemporaneous to other sets of observations. These magnetic field maps will provide a crucial input to the analysis, modelling and interpretation of the multiwavelength data sets that will follow in forthcoming papers.

This paper is organized as follows. Section 2 presents our observations. In Section 3, we describe the magnetic imaging method we use. The results are shown in Section 4, where we present the reconstruction of the magnetic field of HD189733 at five different epochs. We discuss the magnetic field evolution of HD189733 in Section 5, based on this paper's results and results of Moutou et al. (2007) and Fares et al. (2010). Section 6 presents the summary and conclusions of this work.

2 OBSERVATIONS

Our spectropolarimetric data were obtained using both NARVAL spectropolarimeter at the TBL (2 m) and ESPaDOnS at CFHT (3.6 m). A spectropolarimeter is a spectrograph with a polarization section, allowing measurement of the polarization of the spectral lines. A circular polarization spectrum is extracted from four subexposures taken each at a different angle of the polarization waveplates. ESPaDOnS and NARVAL are twin instruments, both operating in the optical domain (370–1000 nm) at a resolution of 65 000 in the polarization mode. Data reduction is done automatically using Libre ESpRIT, a fully automated reduction tool installed at TBL and CFHT (Donati et al. 1997). The spectra are normalized to a unit continuum, their wavelengths referring to the heliocentric rest frame. Telluric lines are used to correct from spectral shifts due to instrumental effects. This correction secures a radial velocity (RV) precision of about 15 m s⁻¹ (Moutou et al. 2007).

Observations with both NARVAL and ESPaDOnS were obtained in service mode. Our NARVAL 2013 data (PI Bourrier) were collected as follows: 5 spectra in May (11–13), 2 spectra in June (12–15), 5 spectra in July (01–04–05–08–14), 14 spectra in August (02–23), 13 spectra in September (02–24) and finally 5 spectra in October (08–09–12–13–18). The ESPaDOnS data were obtained via a filling program targeting planet-hosting stars (PI Moutou), seven spectra were collected in 2013 September. Table 2 presents the log of observations of all our data in 2013. In 2014, 15 spectra were collected between September 01 and October 20. Another 18 spectra were collected between 2015 May 20 and 2015 July 20. The log of the 2014 and the 2015 (PI Bourrier) observations are presented in Table 3.

For cool stars, the polarization signature in single lines is typically within the noise level. Using the polarization information of many spectral lines simultaneously, we can extract the polarization signal of the spectrum, this is known as multiline technique least-square deconvolution (LSD; Donati et al. 1997). LSD assumes that all lines have the same polarization information, and extracts the polarization signature by deconvolving the observed spectrum with a line mask. LSD calculates intensity (Stokes I) profiles, circular polarization (Stokes V) profiles, as well as a null polarization profiles (labelled N). These profiles are calculated using a combination of the sub-exposures taken at different angles of the waveplates. The null profile is a check profile, it is calculated so as to cancel out the stellar polarization signature, and thus should contain no polarization. It helps check for spurious or instrumental signatures. For more details on how these profiles are calculated, see Donati et al. (1997) and Mengel et al. (2017).

We compute a line mask for HD189733 using Kurucz's lists of atomic line parameters (Kurucz CD-Rom 18) and a Kurucz model atmosphere with solar abundances, temperature set to 5000 K and logarithmic gravity (in cm s⁻²) set to 4.0. Only moderate-to-strong lines, featuring central depths larger than 40 per cent of the local continuum, are included in the mask (before any macroturbulent or rotational broadening); the strongest and broadest features (such as Balmer lines) are excluded. In the optical domain, this mask contains about 4000 lines. The LSD profiles calculated by deconvolving the stellar spectra with the mask have a S/N ∼ 30 times higher than the S/N in single lines with average magnetic sensitivity (see Tables 2 and 3), allowing for the detection of the polarization signature.

We calculated the RV of each Stokes I profile by fitting a Gaussian profile to it. Since the star hosts a HJ, the RV signatures vary over the planetary orbit. Our RV data agree (within the error bars) with the orbital solution of Boisse et al. (2009). There is an offset between our values and theirs, +0.06 km s⁻¹ for 2013 June–July, +0.13 km s⁻¹ for 2013 August, +0.10 km s⁻¹ for 2013 September, +0.13 km s⁻¹ for 2014 September and +0.11 km s⁻¹ for 2015 June. Such offsets are due to the use of different reduction pipelines (ESPaDOnS and Narval versus Sophie). For the tomographic imaging, we shift each profile by its RV to centre all profiles around 0 km s⁻¹.

3 MODEL DESCRIPTION AND IMAGING METHOD

Magnetic fields, when present in the photosphere, cause splitting of the magnetically sensitive spectral lines via the Zeeman effect. The lines that form in such environments are polarized. Measuring the wavelength shift between spectral components of the line (when possible, which depends, among others, on the amplitude of this shift, the rotational broadening and the spectrograph resolution) allows us to calculate the longitudinal magnetic field of the star (see e.g. Shulyak et al. 2014 for M dwarfs). The field topology (i.e. distribution, polarity, configuration), on the other hand, cannot be determined by the wavelength shift alone, but is determined using a tomographic imaging technique, Zeeman–Doppler imaging (ZDI; Donati et al. 1997). Polarization of the spectral lines depends on the orientation of the magnetic field relative to the line of sight (see fig. 2 of Reiners 2012). To map the field, spectra are collected during one or more stellar rotations. ZDI inverts these spectra/profiles into a magnetic topology that can produce the observed polarization signatures. Because this problem is ill-posed, ZDI uses maximum-entropy regularization to get the simplest magnetic map compatible with the data.

In the newest version of ZDI (Donati et al. 2006), the field is described by its poloidal and toroidal components (Chandrasekhar 1961), with all the components expressed in terms of spherical harmonics expansion. The highest degree of spherical harmonics expansion used to map the field represents the ZDI map resolution around the equator. For slow rotators, such as HD189733, we limit the spherical harmonics to the lowest degrees (l ≤ 5, see Fares et al. 2010 for more details). ZDI follows an iterative procedure, it compares synthetic Stokes V profiles to the observed profiles collected during the stellar rotational cycles. In practice, the stellar surface is divided into 9000 grid cells of similar area. The contribution of each grid cell is then calculated in the weak field regime, and a synthetic Stokes profile for each observed rotation phase is produced.

In addition to the field intensity and distribution, ZDI also gives an indication on the stellar inclination (up to ∼10° accuracy), on the stellar rotation period (see e.g. Alvarado-Gómez et al. 2015) and on the differential rotation (DR). We describe DR using a solar-like DR law, where the rotation rate at a latitude θ is defined by Ω(θ) = Ω_eq − dΩsin ²(θ), where Ω_eq and dΩ are, respectively, the rotation rate at the equator, and the difference in rotation rate between the pole and the equator. In practice, to measure DR, we reconstruct the magnetic image at a given magnetic energy for a pair of fixed (Ω_eq, dΩ), and obtain the reduced-chi squared (⁠|$\chi ^2_{\rm r}$|⁠) of the fit. We investigate the |$\chi ^2_{\rm r}$| values of the 2D parameter space of (Ω_eq, dΩ). The optimum DR parameters are the ones minimizing |$\chi ^2_{\rm r}$|⁠. They are obtained by fitting the surface of the |$\chi ^2_{\rm r}$| map with a paraboloid around the minimum value of |$\chi ^2_{\rm r}$|⁠.

The null profiles, for each epoch, are used as a test to check for spurious polarization. In practice, we fit these profiles with a zero magnetic field configuration. Since the null profiles should contain no polarization, the |$\chi ^2_{\rm r}$| of the fit should be one or less. A |$\chi ^2_{\rm r}$| greater than one indicates either a spurious polarization signature, or an underestimation of the error bars. If a systematic signature is found in the profiles, it indicates a spurious origin of the signature. In this case, we calculate a mean signature of the null profiles, and subtract it from the Stokes V profiles to eliminate this spurious feature.

4 RESULTS

ZDI requires the use of data covering one or more stellar rotations to derive the photospheric magnetic map. In some field configurations (see, e.g. Morin et al. 2008), the large-scale magnetic field of the star is stable for many stellar rotations and also over many years. In these cases, one can combine data collected over many rotations together as one data set. To estimate whether the field is stable over many stellar rotations, we compare the quality of the fit and the shape of the polarization profiles at the same rotational phases.

Our observations cover many stellar rotations at each observing epoch, from 2013 June to 2015 July (see Tables 2 and 3).

We performed a series of tests on the data, reconstructing the maps using a combination of data sets for each epoch. We find that using Stokes V profiles spread over more than two consecutive stellar rotations to reconstruct the map worsens the quality of the fit. We therefore adopted data sets of up to two stellar rotations for each reconstructed map. Some spectra were not used in the reconstruction because they were collected with a rotational phase gap of more than a stellar rotation in respect to the main data set (see Tables 2 and 3).

4.1 Differential rotation

DR distorts the magnetic regions on the stellar surface. Spectropolarimetric data can therefore be used to estimate the level of DR.

We applied the same technique as in Fares et al. (2010) to search for DR in our data (explained in Section 3). In this study, we were able to detect DR for 2013 August data set. The |$\chi ^2_{\rm r}$| map in the (Ω_eq, dΩ) space is shown in Fig. 1. A well-defined |$\chi ^2_{\rm r}$| minimum is found for dΩ = 0.11 ± 0.05 rad d⁻¹ and Ω_eq of 0.535 ± 0.004 rad d⁻¹, corresponding to a rotation periodof 11.7 ± 0.1d at the equator. The shear value dΩ is consistent with the one measured by Fares et al. (2010) at dΩ = 0.146 ± 0.049 rad d⁻¹. These measurements, within their uncertainties, do not reveal a variation of DR.

HD189733 is a slow rotator for which measuring DR using the Fourier transform of the intensity profile technique presented by Reiners & Schmitt (2002) cannot be applied. There are no measurements of the DR of this star in the literature, apart from the recent work of Cegla et al. (2016). Cegla et al. (2016) modelled the Rossiter–McLaughlin effect to probe planetary parameters and stellar DR, and found a dΩ > 0.12 rad d⁻¹ for HD189733, in agreement with our findings.

Stellar DR is not constant across spectral types and rotation rates. Many observational and theoretical studies have addressed the question of its variation (see e.g. Barnes et al. 2005; Reiners 2006; Collier Cameron 2007; Augustson et al. 2012; Reinhold, Reiners & Basri 2013; Balona & Abedigamba 2016; Distefano et al. 2016). Using Kepler photometric data, Reinhold et al. (2013) and Balona & Abedigamba (2016) studied the DR for a large sample of stars. HD189733 DR is within the range of the shear values of Kepler stars both as a function of temperature and as a function of rotation. Balona & Abedigamba (2016) propose an empirical relation between the shear value, the effective temperature of the star and its rotation. For HD189733, the predicted value of the dΩ is |$0.076^{+0.004}_{-0.001}$|⁠. Our value is slightly higher, their value is, however, within the error bars of our measurement.

4.2 Magnetic maps

We reconstruct the magnetic maps for five observing epochs. The observed Stokes V (in red) and a 1σ error bar are shown in Fig. 2, along with the ZDI reconstructed circular polarization (in black). The observed profiles show a detection of the polarization signature (false-alarm probability less than 10⁻⁵, see Donati et al. (1997) for the classification of the detections). The reconstructed maps are shown in Fig. 3. As mentioned previously, each data set used to reconstruct a magnetic map consisted of no more than two stellar rotations.

• 2013 June/July

  We reconstructed the map of 2013 June/July, although our phase coverage is not optimal. The |$\chi ^2_{\rm r}$| of the reconstruction is 1.15. The mean magnetic field is of 36 G. 61 per cent of the magnetic energy is in the toroidal component. For the poloidal component, we define axisymmetric modes as having m = 0 and m < l/2 (m and l being the order and the degree of the associated Legendre polynomial, used to describe the field as spherical harmonics expansion). About 38 per cent of the poloidal field is axisymmetric.

• 2013 August

  In 2013 August, the data samples the stellar rotation well. However, the null profile shows a systematic signature in the core of the profile. Fitting these null profiles with no magnetic field configuration leads to a |$\chi ^2_{\rm r}$| of 1.3. We calculated a mean profile for all the null profiles. Subtracting the mean null profile from the 2013 August null spectra leads to a |$\chi ^2_{\rm r}$| of 0.7 when fitting the corrected null profiles with no magnetic configuration. Given that the signatures in the null profiles show a systematic trend, it is more likely to be a spurious signature rather than an underestimation of the error bars. We subtract the mean null profile signature of the Stokes V profiles. The map is reconstructed with a |$\chi ^2_{\rm r}$| of 1.85. The average surface magnetic field is 41 G. 50 per cent of the total energy is in the toroidal component of the field. 2 per cent of the poloidal energy is in the axisymmetric modes and mainly in modes with m = 0. The poloidal component is thus mainly non-axisymmetric. Spherical harmonics modes with l > 3 (i.e. modes higher than the dipole, quadrupole and octupole) contribute to ∼38 per cent of the poloidal energy.

• 2013 September

  In 2013 September, we have data from ESPaDOnS and NARVAL, covering two stellar rotations. We reconstructed a map using those data with a |$\chi ^2_{\rm r}$| of 2.15. We notice a systematic difference in the intensity profile depth between the Narval spectra and the ESPaDOnS spectra. The difference in depth is not significant enough to require different modelling of the intensity profiles between instruments. This difference is due to the normalization of the spectra at the telescope.

  The average surface magnetic field is 42 G. 59 per cent of the energy in the toroidal component. The poloidal field is still mainly non-axisymmetric (88 per cent, see Table 4). About 50 per cent of the poloidal component is in the octupolar mode, higher orders contribute to ∼45 per cent.

• 2014 September

  In 2014 September, the average magnetic field drops to 32 G (see Fig. 3). The |$\chi ^2_{\rm r}$| of the fit is 1.25. The field remains mainly toroidal, with 78 per cent of the total energy stored in the toroidal components. The poloidal field is strongly non-axisymmetric, and only 10 per cent of the energy lies within axisymmetric modes.

• 2015 July

  In 2015 July, the magnetic field, and in particular the radial component, has changed significantly relative to 2014 September. The radial component changes polarity around the equator between 2014 September and 2015 July. The poloidal component contributes to 15 per cent of the total energy. The average magnetic field is of 37 G.

The characteristics of each reconstructed map are listed in Table 4. ZDI does not provide error bars for the reconstructed map. However, statistical errors can be calculated by varying the input parameters (e.g. v sin i, stellar inclination, Ω_eq, dΩ) within their error bars. Error bars on field characteristics are calculated by comparing the characteristics of the reconstructed field for the best set of input parameters to the characteristics of maps reconstructed by varying those best input parameters within their error bars. We follow Mengel et al. (2016) for error bar calculations: we fix (Ω_eq, dΩ) and vary v sin i within its error bars (2.75 km s⁻¹ to 3.2 km s⁻¹, Table 1). We then do a new set of maps fixing v sin i and varying (Ω_eq, dΩ) within their error bars. Error bars in Table 4 represent the highest values of our procedure.

4.3 Extrapolation of the magnetic field to investigate the corona and the planetary orbit

The interactions between the stellar wind and the planetary magnetosphere might trigger planetary emission, such as radio emission or bow shock formation (see e.g. Fares et al. 2010; Llama et al. 2013), and it might influence the ionization state of the planetary atmosphere (Rimmer & Helling 2013; Rimmer et al. 2014a). In this section, we examine the stellar magnetic field in the corona up to the planetary orbit using a potential field extrapolation of the surface magnetic field. For this purpose, we use the potential field source surface (PFSS) code of Jardine, Collier Cameron & Donati 2002, originally developed for the Sun (van Ballegooijen et al. 1998) based on Altschuler & Newkirk (1969). The potential field extrapolation assumes that there is no electric current in the corona. The components of the magnetic field in the corona are described using a spherical harmonic decomposition. This technique delivers satisfactory results when compared to wind modelling of the solar corona (Riley et al. 2006).

The PFSS model extrapolates the magnetic field considering two boundary conditions: the first one being that the field is purely radial at a surface called the source surface, the second boundary condition is the observed field geometry at the surface of the star. We assume that the source surface is at 3.4 R_⋆. Our radial magnetic maps are used as a boundary condition at the surface of the star, thus giving a realistic model of the radial field.

The extrapolated field in the stellar corona for the five maps presented in this paper are shown in Fig. 4. This figure shows how different surface field configurations produce different field line configurations in the corona. Since the magnetic field lines in the corona are not those of a very simple configuration, and since the planet and the star are not synchronized, the planet crosses different field configurations on its orbit, as well as from one orbit to the other (the planet crosses in front of the same stellar field configuration after one beat period (of rotation and orbital periods), rather than after one orbital period). We calculate, for each observing epoch, the footprints of field lines connecting the stellar surface to the position of the planet on its orbit. They are shown in Fig. 5.

The PFSS model allows the calculation of the field value and orientation at each position in the corona, out to the source surface. Since the planet is outside the source surface, we proceed as follows to calculate the field at its position. First, we calculate the position of the sub-planetary point on the source surface, and calculate the energy budget at this point. We remind the reader here that at the source surface, the stellar field is purely radial, the meridional and azimuthal components are negligible. Assuming that the magnetic flux is conserved over spherical shells from the source surface to the planetary orbit, we calculate the decay of the magnetic field between the source surface and the orbit. In Fig. 6, we plot the field value at the planetary orbit for each epoch of observations (including 2007 June and 2008 July). The maximum value the field reaches at the planetary orbit can change by 100 per cent between different epochs, which supports the importance of simultaneous observations when studying star–planet interactions. We investigate the effect the error bars on the maps could have on the calculated magnetic field at the planetary orbit. To do so, for each epoch, we use as boundary condition each of the maps calculated for a range of v sin i, Ω_eq, and dΩ (see Section 4.2), extrapolate the magnetic field for each map, and calculate the magnetic field at the position of the planet. We find that the mean difference between the field values presented in this paper and those calculated for the maps reconstructed for sets of v sin i, Ω_eq and dΩ is of the order of 3mG. The difference can reach up to 70 per cent of the mean value over a small fraction of the orbit (for one epoch, all other epochs showed a smaller maximum difference value). This shows the robustness of our results: at the position of the planet, the variation of the stellar magnetic field is real and not affected by the error bars on the maps.

5 DISCUSSION: MAGNETIC FIELD EVOLUTION

In this section, we take advantage of previous spectropolarimetric observations of HD189733 to investigate the evolution of its magnetic field (intensity and topology) over a 9 yr time span. The first reconstructed magnetic maps of HD189733, based on ESPaDOnS/CFHT data of 2006 June and August, were presented by Moutou et al. (2007). Later on, and with the aim of detecting star-planet interaction signatures, Fares et al. (2010) presented two additional reconstructed magnetic maps of HD189733, based on spectropolarimetric observations of 2007 June and 2008 July, as well as an update of the reconstructed map of 2006, merging June and August data as one data set.

The data presented in this paper (five observing epochs spanning 2 yr) shows that the field of HD189733 can evolve over a few stellar rotations. Combining data sets spanning more than two stellar rotations systematically reduces the quality of the fit. Fares et al. (2010) have merged data obtained in 2006, spread over five stellar rotations. We revisited the summer 2006 data and found that the map was overfitted. We adopt the results of Moutou et al. (2007) in this paper.

Due to the 5-yr gap in the observations, we cannot investigate the presence of cyclic variations in the stellar magnetic field, in a similar way as those reported for the planet-hosting star τ Boo (Donati et al. 2008; Fares et al. 2009, 2013). Nevertheless, an evolution in the stellar magnetic field intensity and topology can be seen. Variations in both the axisymmetric contribution to the poloidal field and the toroidal contribution to the total field are observed during this time span (see Fig. 7).

For all epochs (apart from 2006 June), the toroidal component dominates over the poloidal one. Petit et al. (2008) suggest, studying a sample of solar-like stars, that the toroidal energy dominates over the poloidal one for stars with rotation periods less than ∼12 d. HD189733, having an equatorial rotation period of ∼12 d, does not contradict their findings. On the other hand, Donati & Landstreet (2009) suggest that stars with Rossby number (Ro) <1 develop toroidal fields. HD189733 has a Ro = 0.403 (Vidotto et al. 2014). HD 189733 field geometry is therefore compatible with the geometries observed for stars with similar masses and Ro numbers. The fraction of axisymmetric field is almost always less than 50 per cent.

In addition, it is also interesting to compare the evolution of the stellar magnetic field within the 2013 data sets, that are separated by just nine rotation periods. We note that little variation was seen for the toroidal component. The percentage of the contributors (dipole, quadrupole and octupole) to the poloidal field, on the other hand, has changed. The main contributor to the field (i.e. azimuthal component) does not change polarity. The radial component evolved with negative and positive magnetic features appearing at the surface.

6 CONCLUSIONS AND FUTURE PLANS

This paper is part of the MOVES collaboration, which aims to characterize comprehensively the complex environment of the exoplanet HD189733b. Orbiting a bright and active K dwarf at a short distance, this transiting HJ has been subjected to many stellar and planetary atmosphere studies. The main objectives of MOVES are to probe the different regions of the extended planetary atmosphere, its interactions with the host-star and their temporal variability. The wider set of multiwavelength observations (X-ray with Swift and XMM–Newton, UV spectroscopy with HST and XMM–Newton and radio observations with LOFAR) were taken contemporaneously with the magnetic field mapping presented here.

In this first paper, we presented a detailed spectropolarimetric study of HD189733 and studied the evolution of its magnetism. Stellar magnetism is an important ingredient in stellar evolution, and also has important effects on planets surrounding these stars. The star was observed at five epochs (2013 Jul, 2013 Aug, 2013 Sept, 2014 Sept and 2015 Jun), during which we also collected X-ray and UV observations (Wheatley et al., in preparation). Using ZDI, we reconstructed the magnetic maps of the star. With a strength up to 42 G, the magnetic field is dominated by the toroidal component at the five epochs. The toroidal component is mainly axisymmetric during all observing epochs. In contrast, the poloidal component is mainly non-axisymmetric. We will continue monitoring this system to study the magnetic evolution on time-scales longer than 2 yr and look for a potential magnetic cycle. These reconstructed magnetic maps are crucial for analysing multiwavelength observations. They allow us, modelling the stellar wind in the corona and at the planetary orbit, to reconstruct the X-ray emission and irradiation of the planet, as well as the spatial distribution of X-ray that will be absorbed by the extended atmosphere of the exoplanet.

Acknowledgements

We thank an anonymous referee for their useful comments. This work is based on observations obtained with ESPaDOnS at the Canada–France–Hawaii Telescope (CFHT) and with NARVAL at the Télescope Bernard Lyot (TBL). CFHT/ESPaDOnS are operated by the National Research Council of Canada, the Institut National des Sciences de l'Univers of the Centre National de la Recherche Scientifique (INSU/CNRS) of France and the University of Hawaii, while TBL/NARVAL are operated by INSU/CNRS. We thank the CFHT and TBL staff for their help during the observations, and in particular R. Cabanac and P. Petit. We also thank J.-F. Donati and E. Hébrard for useful comments on the data analysis. RF acknowledges financial support by WOW from INAF through the Progetti Premiali funding scheme of the Italian Ministry of Education, University, and Research. VB and AL acknowledge the support of the French Agence Nationale de la Recherche (ANR), under programme ANR-12-BS05-0012 ‘Exo-Atmos’. Part of VB work has been carried out in the frame of the National Centre for Competence in Research ‘PlanetS’ supported by the Swiss National Science Foundation (SNSF). VB also acknowledges the financial support of the SNSF. AAV acknowledges partial support from an Ambizione Fellowship of the Swiss National Science Foundation. ChH highlights financial support of the European Community under the FP7 by an ERC starting grant number 257431. PW is supported by an STFC consolidated grant (ST/L000733/7).

REFERENCES