HR8799: Imaging a System of Exoplanets
The HR 8799 planetary system is the most intriguing and spectacular system yet discovered by direct imaging. With four gas giant planets (5–7 MJup) orbiting at wide separations (15–70 AU) from an unusual, young A star, HR 8799 serves as a Rosetta Stone for atmospheric and planet formation physics. With direct access to the light from the planets themselves, extensive photometric and spectroscopic studies of the system have been undertaken, painting a fascinating picture of cloud formation, nonequilibrium chemistry, and potential elemental abundance measurements that can constrain formation pathways. The dynamical structure of the system is complex, with the four massive planets likely stabilized through participation in mean-motion resonances. Two belts of debris flank the planets, with a warm belt analogous to the solar system asteroid belt and a wide, cold ring comparable to the Kuiper belt. HR 8799 will continue to be studied extensively from the ground and space in the coming years, with the exciting possibility that additional planets remain to be discovered in the system.
With the advent of adaptive optics (AO) on large, ground-based telescopes, the possibility of imaging close, low-mass companions to young, nearby stars was immediately recognized. The assumed proximity of gas giant planets to their host stars requires high spatial resolution. The expected flux ratio between the star and the planet ( ∼ 104–108) necessitates careful control of the stellar point spread function (PSF). It also pushed for the development of precision coronagraphs to further suppress the light of the star (e.g., Roddier and Roddier 1997; Mawet et al. 2005; Soummer 2005).
Paving the way for discoveries of planetary mass companions via direct imaging were early discoveries of brown dwarf companions to stars. Becklin and Zuckerman (1988), using newly developed infrared arrays, imaged a low-mass companion to the white dwarf GD 165. The separation of the companion is 4.2″, and with GD 165A being a white dwarf, the magnitudes of the two objects in the near-infrared were quite comparable. GD 165B is now recognized as the first L spectral type object discovered, ushering in a new era of discovery and classification of substellar objects with more large-scale near-infrared surveys. Several years later, Nakajima et al. (1995) used primarily seeing limited observations coupled with an image-stabilizing coronagraph to image a brown dwarf companion 7.6″ away from the nearby M star GL 229. This discovery marked the first identification of an object of T spectral type and set the stage for direct imaging in the high-contrast regime. Thereafter, surveys using newly commissioned AO systems commenced with sensitivity to detect primarily brown dwarf companions but in the best cases planetary mass objects (e.g., Macintosh et al. 2003; McCarthy and Zuckerman 2004; Metchev and Hillenbrand 2004). A relative paucity of brown dwarf companions to Sun-like stars emerged from these surveys, representing a possible extension of the so-called brown dwarf desert found in radial velocity surveys (e.g., McCarthy and Zuckerman 2004).
An important aspect of these surveys is, and remains, proper target selection. The contrast between a star and a gas giant planet is generally improved when the planet is younger, as these planets cool as they age. For instance, in the near-infrared H band, where many surveys were conducted, the difference between a 50 and 500 Myr planet of 5 MJup is nearly 10 magnitudes (Baraffe et al. 2003). This necessitates identification of targets that are both youthful and relatively nearby to offer greater access to closer physical separations. In the late 1990s and early 2000s, pioneering work at identifying nearby, young stellar associations dramatically improved the number of stars accessible to direct imaging surveys (e.g., Kastner et al. 1997; Barrado y Navascués et al. 1999; Mamajek et al. 1999; Zuckerman and Song 2004; Torres et al. 2008). Thus, the stage was set for the first images of planets orbiting other stars.
Discovery of a Directly Imaged Planetary System
The star HR 8799 was originally identified as an unusual A-type star by a number of photometric and spectroscopic observations in the 1970s that indicated youth and an unusually low abundance of certain metals (e.g., Barry 1970; Oblak et al. 1976; Gehren 1977). It was later identified to be a “Vega-like” star by Sadakane and Nishida (1986), who noted an infrared excess as measured by the Infrared Astronomical Satellite (IRAS). Subsequent spectroscopic observations have further classified the source as both a γ Doradus-type variable and λ Bootis-type metal-depleted star (e.g., Kaye et al. 1999; Gray and Kaye 1999; Sadakane 2006). These unusual spectrophotometric properties, coupled with suggestive UVW space motions, led Zuckerman and Song (2004) to conclude that the star is youthful and suggest it as a target for direct imaging surveys. Current estimates suggest that the star is a member of the Columbia Association ( ∼ 30 Myr, Zuckerman et al. 2011) with a radius of 1. 44 ± 0. 06 R⊙, a mass of 1. 516−0. 024+0. 038 M⊙, and a temperature of 7193 ± 87 K (Baines et al. 2012). The distance to HR 8799 is 39. 4 ± 1. 1 pc (van Leeuwen 2007).
The first data taken on the source with the capability of seeing the planetary companions was gathered with HST NICMOS in coronagraphic mode in 1998. Lowrance et al. (2005) first reported on the data but noted only the presence of potential very wide companions with separations over 13″. Other ground-based AO datasets were taken with Subaru, Keck, and Gemini between 2002 and 2005. However, the planets were also not reported based on these datasets, or the early HST dataset, until many years later (Fukagawa et al. 2009; Metchev et al. 2009; Lafrenière et al. 2009; Soummer et al. 2011; Currie et al. 2012).
LOCI is a least-squares algorithm that takes advantage of data taken in ADI mode to perform a more optimized PSF subtraction. In simple ADI subtraction, all frames taken a long enough time apart to minimize planet self-subtraction are combined to create the reference PSF for subtraction. With LOCI, small regions are considered independently, and the best combination of frames is used to maximize the quality of the PSF subtraction across the whole field of view (Lafrenière et al. 2007b). Using this algorithm results in at least a factor of ∼ 3 improvement in signal-to-noise over ADI alone.
The first ADI datasets on HR 8799 were taken with both the Gemini NIRI and Keck NIRC2 between 2007 and 2008. As reported in the discovery paper by Marois et al. (2008), they first spotted the outer two planets, HR 8799b and HR 8799c, in the NIRI data, followed by the detection of HR 8799d with NIRC2. The Keck data taken in 2004 was then reanalyzed and combined with the other datasets in the announcement of the discovery of the planetary system (Marois et al. 2008). The planets were thus named in the order of their discovery.
Upon discovering the planet, Marois and collaborators continued to monitor the system, switching from the H and K near-infrared bands to the Lp band when it became apparent that the planets were quite red. Continued algorithmic improvements at tight separations coupled with deeper observations led to the identification of a fourth planet interior to the three initially published in Marois et al. (2008). This new planet, HR 8799e, was announced 2 years after the first discovery paper by Marois et al. (2010).
Figure 1 shows the image of all four known planets taken with Keck NIRC2 in the Lp band from Marois et al. (2010). The projected separations of the planets in their discovery epochs was 68, 38, 24, and 15 AU (b, c, d, and e, respectively). The estimated masses of the planets from evolutionary models combined with the estimated age range for the star are 5–13 MJup, though dynamical stability arguments imply they may be at the lower end of this range (between 5 and 7 MJup).
Atmospheric Properties of the HR 8799 Planets
The power of the direct imaging techniques lies in the ease of obtaining photometric and spectral information about the planets, which allows for detailed studies of their atmospheres. Since gas giant planets are generally thought to have atmospheric properties similar to brown dwarfs, much of the focus of initial direct imaging surveys was at the H band (e.g., Lafrenière et al. 2007a; Liu et al. 2010), where T dwarfs are bright due to methane absorption at other wavelengths. The expected temperatures of young, gas giant planets were predicted to be ∼ 500–1200 K, the range covered by T dwarfs in the field (Burgasser et al. 2002).
Initial assessment of the atmospheric properties of the HR 8799 planets came from photometry at the near-infrared wavebands J, H, Kp, and Lp (Marois et al. 2008). Narrowband photometry in these wavelength regimes was also obtained in order to provide initial hints at spectroscopic features in the planets. Though the earliest datasets were in the H band, it was immediately apparent that the planets were very red, behaving much more like the warmer L dwarfs than the T dwarfs. Figure 2 shows the location of the HR 8799 planets on a J − K vs. J color-magnitude diagram. At the time of their discovery, the planets were among the reddest substellar objects that had been seen, drawing comparisons to the extremely red planetary mass companion 2M1207b (Chauvin et al. 2004) and free-floating planetary mass objects in star-forming regions (e.g., Casewell et al. 2007).
Since the initial discovery, photometric observations have been obtained at a large number of wavelengths between ∼ 1 and 5 μm. With these data, the bolometric luminosity is estimated. A comparison of the age of the system and Lbol to the predictions of evolutionary models yields the primary mass estimates for the system. From there, spectral models are fit to the data in order to assess parameters such as temperature, surface gravity, and, in the case of substellar objects, cloud properties. Since the HR 8799 planets were obviously quite red, it was immediately apparent that models with clouds were more appropriate than models with a clear photosphere such as those typically used for T dwarfs (e.g., Chabrier et al. 2000). Marois et al. (2008) were the first to note that fitting with typical atmospheric models with thick clouds would yield preferred spectra with temperatures in the range of 1400–1700 K, which when combined with Lbol yielded unphysically small radii for the planets. As a possible solution to this problem, Marois et al. (2008) found that patchy clouds would allow model atmospheres to agree reasonably well with the observed photometry while simultaneously matching the radii and effective temperatures implied by the age, bolometric luminosities, and evolution models.
While photometric measurements at multiple wavelengths are powerful diagnostics for the atmospheres of the HR 8799 planets, spectroscopic observations offer a level of detail that are currently unique to the directly imaged systems. Spectroscopy reveals the fingerprints of molecular features that can be used to further quantify effects like nonequilibrium chemistry and to probe chemical abundances.
The first spectroscopic observations of the planets in this system were focused on HR 8799b, which as the most widely separated planet suffers less from the impact of speckle noise than the others. Bowler et al. (2010) obtained a narrowband spectrum spanning ∼ 0.1 μm of a K band methane feature for HR 8799b. They found that there was unlikely to be strong methane absorption in the planet. The following year, Barman et al. (2011) published broad H and K band spectra of the planet at low resolution (R ∼ 60). In a comprehensive modeling effort on both planets, Barman et al. (2011) found that the spectra also had only weak evidence for both methane and carbon monoxide. They also found a strong need for nonequilibrium chemistry and potentially nonsolar abundances to fit the spectra, in addition to a patchy cloud model as described in photometric fits. In both works, the spectra were compared to L- and T-type object spectra and found to match late L and early T dwarfs most closely. This is consistent with the picture of thinning clouds that characterizes the L/T transition (see discussion in Marley et al. 2010).
In order to fully constrain metallicity or elemental abundances, high-resolution spectra are required. For the HR 8799 system, this has been achieved for two planets – HR 8799b and HR 8799c. In Konopacky et al. (2013), K band spectra are presented for HR 8799c from OSIRIS at the full resolution offered by the instrument, R ∼ 4000. In these data, molecular lines from water and carbon monoxide are resolved and used to measure the ratio of carbon to oxygen in the planetary atmosphere via models. This ratio has been suggested as a possible diagnostic of the mode or location of formation within a protoplanetary disk (e.g., Öberg et al. 2011; Piso et al. 2015; Ali-Dib 2017). HR 8799c was found to have a slightly elevated C/O ratio from the assumed value for HR 8799A, which could be suggestive of a core accretion formation scenario (Konopacky et al. 2013). More recently, Barman et al. (2015) presented medium-resolution OSIRIS spectra for HR 8799b in both K and H bands. They found a similar C/O ratio for this planet of slightly above the expectation for the star and also the strongest evidence of methane absorption in its atmosphere to date. No methane is detectable in HR 8799c, suggesting the onset of strong methane absorption in the near-IR for the mass and age of these planets is near the effective temperature of HR 8799b ( ∼ 900 K).
The final verdict on the C/O ratio in these planets has not yet been reached, however. Recently, Lavie et al. (2016) performed a Bayesian spectral retrieval modeling analysis of the low-resolution spectra of all four HR 8799 planets and found supersolar C/O ratios for HR 8799b and c, but subsolar values for HR 8799d and e. However, the values they derived for HR 8799b and c are significantly higher than those found in Konopacky et al. (2013) and Barman et al. (2015). Further modeling of all available data, higher-resolution spectra of HR 8799d and e, and more precise measurements of the abundances in HR 8799A are necessary to more definitively determine whether the C/O ratios are different among the planets and different from their host star. The work done thus far highlights the power and potential of directly imaged planets to yield new insight into planetary atmospheres through photometry and spectroscopy. Figure 4 shows a compilation of all spectroscopic measurements of the four HR 8799 planets.
System Architecture and Dynamics in HR 8799
With multiple planets in the system, in addition to circumstellar material, HR 8799 is ripe for dynamical study and characterization. In order to fully assess the overall structure of the system, several measurements are required: the spin axis of the star, the orbital structure of the planets, and the orientation and extent of any circumstellar debris.
Stellar Spin Axis Orientation
For HR 8799A, the main parameter of interest for overall system architecture is the orientation of its spin axis. This can be compared to the orientations of the planetary orbits or debris disks to possibly shed light on the dynamical evolution of the system. The favored method for estimating this value comes from asteroseismology, which uses the known γ Dor-type variability of the star to constrain pulsation modes which in turn yield a spin orientation.
Moya et al. (2010) performed an asteroseismology analysis on HR 8799A using photometric data and the frequency ratio method. These analyses rely on knowledge of the rotational velocity of the star. The star has a measured vsini of 37.5 ± 2.0 km s−1 (Kaye and Strassmeier 1998), with inclination unknown. Moya et al. (2010) were interested in what these results implied about the age of the star. Depending on the inclination, ages as high as 1 Gyr were estimated, in contrast to the more commonly quoted 30 Myr age. In this case, the masses of the planets would be in the brown dwarf regime. This possibility is incompatible with dynamical stability simulations for the system (see next section). However, Moya et al. (2010) did explore various possibilities for inclination, finding required values of > 18∘ for the star to be λ Boo type and > 36∘ for the other portion of their analysis to be valid.
In a subsequent work, Wright et al. (2011) performed a spectroscopic asteroseismological analysis using high-resolution optical spectra from the SOPHIE spectrograph at Observatoire de Haute-Provence. They identified a pulsation mode in the star at frequency 1.98 day−1 and explore possible inclinations based on this mode. They perform fits for allowed inclinations and find that while their best fit is 65∘, inclinations of > 40∘ are roughly equivalently preferred by their data.
In a recent analysis using data from the MOST satellite, Sódor et al. (2014) identified a number of possible new pulsation modes. Though they did not extensively explore possible spin axis inclinations, they did note that a newly identified frequency peak at 1.10 day−1 could correspond to the stellar rotation, as it is the second strongest signal after the 1.98 day−1 frequency found in Wright et al. (2011). If this is the case, the stellar inclination could be between 23∘ and 37∘. These values conflict with the earlier estimate, and the inclination determined by Wright et al. (2011) remains the preferred value for the stellar axis.
Planet Orbits and Stability
The initial publication of the HR 8799 orbits demonstrated that based on their separations, their orbital periods were consistent with integer multiples of each other (Marois et al. 2008) and that the system could potentially be seen in a face-on geometry with an inclination near 0∘. The first dynamical study to explore this was Fabrycky and Murray-Clay (2010), who noted that given the nominal masses for the planets published in Marois et al. (2008), the projected separations as the semimajor axes, and assuming circular orbits, the system would become unstable in a timescale shorter than the system age. They found with the three planets initially known that system stability could be achieved via mean-motion resonances. In particular, the system could have either a 2:1 mean-motion resonance between just HR 8799c and d or a larger 4:2:1 resonance, including HR 8799b, and remain stable with planetary masses as high as 20 MJup. Similarly, Reidemeister et al. (2009) explored possible orbital configurations for the planets and looked in detail at possible masses and orbital plane configurations of the then three-planet system. They also concluded that resonances were necessary for system stability. A similar conclusion was reached by Goździewski and Migaszewski (2009) for a three-planet system. These authors further suggested that perhaps the system is destined for instability in the future, as many stable configurations they explored could last for ∼ 30 Myr but not ∼ 100 Myr if the planets were indeed as large as 10 MJup. A similar conclusion was reached by Moro-Martín et al. (2010), who explored stability for high masses as suggested by an older age for the system (Moya et al. 2010).
The addition of HR 8799e to the system led to an extra dimension of dynamical complexity. This planet was found to have roughly the same luminosity as HR 8799c and d, and, thus, three of the four planets likely have similar masses (Marois et al. 2010). Using the results of Fabrycky and Murray-Clay (2010) as a starting point for potentially stable orbits, Marois et al. (2010) performed dynamical simulations in order to determine whether stability of the system could be achieved with the original masses of the planets of 10, 10, 10, and 7 MJup based on an age of 60 Myr. They found that it was extremely challenging to make the planetary system stable with these masses, but using masses based on an age of 30 Myr (7, 7, 7, and 5 MJup), stability was much more easily achieved. In the latter case, the inner three planets were necessarily in a 4:2:1 mean-motion resonance, but HR 8799b could either participate or not in the resonance chain.
Since then, the exploration of the potential dynamical configuration and system architecture has taken two potential approaches: n-body integration of the planetary bodies over the lifetime of the system to hunt for regions of stability or fits to astrometric measurements of the system to determine potential orbital parameters for each planet. In some cases, both approaches are performed simultaneously, or based on the results from the opposite approach.
When fitting orbits to astrometry, the exercise is helped considerably by the continued monitoring of the system which results in an increased time baseline of measurements. The challenge for the system, and indeed for most directly imaged planetary systems, is that the fraction of orbital coverage for each planet is typically less than a few percent of the period. This means that orbital fits rely on the measurement of positions and velocities and typically acceleration upper limits only. As with most resolved multiple systems, relative astrometry of the star and companion is measured, and typical fit parameters include period, semimajor axis, eccentricity, inclination, longitude of the ascending node (Ω), time of periastron passage, and longitude of periastron passage (ω). The mass and distance of the star are estimated from other methods and held fixed in these analyses. Since the angular momentum vector of the orbit defines the orbital plane parameters, Ω, and inclination, these parameters can typically be constrained fairly well without the measurement of acceleration. For other parameters, generally priors for the distribution of each are required and often play a strong role in shaping the output distributions of allowed orbits. Thus, orbit fitting approaches tend to be some flavor of Monte Carlo, with Markov chain Monte Carlos (MCMCs) being a popular approach (e.g., Ford 2005). Output distributions are then compared between planets or to other features in the system.
Initial fits to the data were not fully unconstrained and often tested particular configurations of orbits. For instance, in Lafrenière et al. (2009), the recovery of HR 8799b in archival HST NICMOS data expanded the time baseline of observations by a factor of ∼ 2. In their orbit analysis, Lafrenière et al. (2009) restricted their parameter space to circular orbits with inclination of < 45∘ and found preferred inclinations of ∼ 13∘–23∘. In Bergfors et al. (2011), new VLT/NACO imaging data were presented providing astrometry for the outer three planets. With the expanded time baseline for HR 8799d, Bergfors et al. (2011) explored whether a circular, face-on orbit was consistent with all astrometry. They found that to be face-on, the eccentricity would need to be > 0.4 for the planet, inconsistent with stability requirements.
In a reanalysis of the 1998 HST dataset, Soummer et al. (2011) were able to detect HR 8799c and d in addition to HR 8799b. They combined these new measurements with available astrometry to explore consistency with various stable configurations, looking at coplanar orbits and fixing the eccentricity to circular for HR 8799c and b. They also determined that a face-on configuration for the system was unlikely and found a favored eccentricity of 28∘ in the presence of a 4:2:1 resonance between the outer three planets. They also found a modest eccentricity of 0.115 for HR 8799d was preferred from their MCMC analysis. Soummer et al. (2011) also noted that no solutions were found within their rejection criteria with i < 27. 3∘ or i > 33. 9∘ for HR 8799d with reasonable eccentricities.
Many orbital studies performed after Soummer et al. (2011) have relied on the HST dataset to extend their time baseline and have come to similar conclusions regarding the inclination of the system. For instance, Currie et al. (2012) used archival data from Keck NIRC2 taken in 2005 along with published astrometry to assess likely orbital configurations for the system, exploring allowed orbits in the vicinity of a 4:2:1 resonance. Since fitting highly undersampled orbits often leads to biased predictions for some orbital parameters such as eccentricity or time of periastron passage, Currie et al. (2012) developed a weighting scheme to try to account for these biases, finding that all planets preferred non-face-on orbits. They also found evidence for a possible mutual inclination between HR 8799d and the other planets.
In two papers utilizing the Large Binocular Telescope AO system, Esposito et al. (2013) and Maire et al. (2015) extended the time baseline to 2011 and provided strong new detections of HR 8799e that allowed for constraints of its orbital properties as well. In their orbital analysis, Esposito et al. (2013) considered noncoplanar solutions with eccentricity of zero. They found that an 8:4:2:1 orbital period ratios were allowed by the astrometry. However, these solutions required noncoplanar configurations, particularly for HR 8799d. This was not the case if they allowed HR 8799d to have a noncircular orbit. They found smaller deviations from coplanarity when looking at a 5:2 orbital period ratio between HR 8799d and e. Using data through 2013, Maire et al. (2015) used the same fitting procedure as Esposito et al. (2013) and considered the consistency of new astrometry with the orbits from that work. They found that the favored solution is one in which the planets have an 8:4:2:1 period ratio and noncoplanar orbits, though the other solution considered was not ruled out to any statistical significance. In a newer analysis that takes a similar approach, Zurlo et al. (2016) provided new astrometry for the planets in 2014 taken with SPHERE. They also used the orbit fitting method of Esposito et al. (2013) and found that the orbital period ratio of 5:2 between HR 8799d and e is ruled out by their data, preferring instead ratios of 2:1 or 3:1.
Taking a different approach, Pueyo et al. (2015) performed an MCMC analysis without forcing coplanarity or circularity. Adding astrometry from P1640 on Palomar to the set of available astrometry in the literature, Pueyo et al. (2015) also found that astrometry for HR 8799d seems to favor a noncoplanar orbit, with a misalignment of up to 20∘ from the other three planets. Konopacky et al. (2016) took a similar approach, considering all orbits consistent with the data, but instead used astrometry only from Keck NIRC2 to minimize possible systematic errors between different instruments. This included a reanalysis of the previously published astrometry from NIRC2 used in the majority of the works previously described. Konopacky et al. (2016) found a preferred inclination for the system of around ∼ 40∘ and that the distributions of inclinations were consistent with coplanarity of all four planets. They also found that the data was consistent with a 8:4:2:1 period ratio. Most recently, Wertz et al. (2017) added new precise astrometry from SPHERE through the end of 2014 and using an MCMC found that eccentric orbits were strongly preferred for HR 8799d, with a preferred inclination for all planets of ∼ 30∘, and some tentative evidence for noncoplanarity between HR 8799b and c. Continued astrometric monitoring of the system will be essential to determine whether these orbital trends hold or if they are the result of systematic biases in astrometry.
Concurrent with these studies, several authors pressed forward on large-scale dynamical simulations of the system including all four planets. Currie et al. (2011) considered the stability of the system when the masses were the same as those considered in Marois et al. (2010), expanding the analysis to include masses up to 13 MJup in the case where a 4:2:1 resonance is present between the inner three planets. They found that stability was never achieved for the lifetime of the system with masses as large as 13 MJup, favoring the lower masses preferred with a 30 Myr age. Sudol and Haghighipour (2012) carried out a very extensive simulation of the system, concluding that the system was almost never stable with masses between 7 and 10 MJup, and thus the masses were more likely lower than this. They also found evidence for a divergence between the inclinations of the outer two planets from HR 8799d or that some eccentricity is required. They also found that stability was easier to achieve if HR 8799e was closer to the star than observed.
In an extensive simulation that expanded upon their original work with a three-planet system, Goździewski and Migaszewski (2014) performed simulations that included the available astrometry of the system, migration of the planets, and long-term stability outlook. They determined that the 8:4:2:1 mean-motion resonance configuration was the most likely scenario, with an on-sky inclination of ∼ 25∘. They could accomplish stability with masses between 5 and 11 MJup for the planets for up to 160 Myr.
Taking a different approach from other works, Götberg et al. (2016) explored the possibility that a mean-motion resonance would not be required for system stability. They found this could be accomplished with masses of 7, 7, 7, and 5 MJup for a narrow range of possible semimajor axes. This is generally true if the separations of the planets are on the wider end of the allowed range from astrometry and the eccentricities are low. They postulate that it is potentially the youth of the HR 8799 system that allows for its currently observed state of apparent stability and that in the future, it may become unstable and evolve dramatically.
Debris Disk Properties and Imaging
HR 8799 has been known to have an infrared excess since the mid-1980s (Sadakane and Nishida 1986), when it was determined to host a Vega-like debris disk. However, it was not until observations with Spitzer and other higher-resolution facilities that the structure of the debris became elucidated. Chen et al. (2006) obtained Spitzer IRS spectra from 5.5 to 35 μm of a large sample of debris disk hosting stars, including HR 8799. In modeling these disks, they noted that HR 8799 was different from the other sources in that the data was better fit by a continuous disk model than a single temperature blackbody. However, they noted that the signal-to-noise of their detection was low, and thus they did not necessarily expect it to possess a continuous disk. They found a dust temperature of 150 ± 30 K, corresponding to a separation of 8 AU, and a fractional dust luminosity of 4. 9 × 10−5. This is slightly higher than a value derived around the same time by Moór et al. (2006), who analyzed SEDs based on IRAS, ISO, and Spitzer photometry to derive a luminosity in dust of 2. 3 × 10−4. Pushing to longer wavelengths, Williams and Andrews (2006) measured the flux of HR 8799 with JCMT at 850 μm. They found a dust temperature of 50 K, corresponding to a radius of 66 AU, and a luminosity of 2. 8 × 10−4. Interestingly, they also detect strong CO in their measurement of HR 8799 but attribute this to an unrelated high-latitude cloud behind the source (this was later confirmed to be the case by Booth et al. 2016).
After the discovery of the planetary companions in the system, several authors worked to understand the possible location of the dust in the context of the location of the planets. Reidemeister et al. (2009) reanalyzed the data presented in Chen et al. (2006) and combined it with other photometric data from IRAS and ISO plus submillimeter points from the literature (Sylvester et al. 1996; Williams and Andrews 2006). They modeled the disk as multiple rings either interior to the then innermost known planet HR 8799d and exterior to HR 8799b. They found a best fit with an outer disk between 75 and 125 AU and an inner disk between 2 and 10 AU.
Su et al. (2009) obtained a full complement of Spitzer observations of HR 8799, including a deeper spectrum with IRS. With greatly improved signal-to-noise, coupled with longer wavelength observations from previous works, Su et al. (2009) performed new modeling of the SED. They also find that there are at least two components to the disk, one at ∼ 150 K and another at ∼ 45 K, corresponding to blackbody temperatures at ∼ 9 and ∼ 90 AU, consistent with the results from Reidemeister et al. (2009). They find the warm component could have a full extent of ∼ 6–15 AU, unresolvable in their images. The outer disk in their model extends from ∼ 90 to 300 AU and is also unresolved. There is also a halo of grains extending out to ∼ 1000 AU inferred from the measurement of extended emission at long wavelengths. They find that the inclination of the disk must be < 25∘ based on their data. For the inner component, the fractional dust luminosity derived by Su et al. (2009) is 2. 2 × 10−5, derived from micron size grains with a total mass of 2. 3 × 10−7 M⊕. Conversely, they estimate the outer disk to have 0.12 M⊕ of material composed of larger grains that lead to a collision-dominated disk. The extent of both disk components, as well as their inner and outer radii, is consistent with sculpting from the planets. Su et al. (2009) also suggest that the presence of the halo of grains indicates strong dynamical stirring ongoing in the system.
Several subsequent observations at longer wavelengths aimed to resolve the outer disk component. Patience et al. (2011) obtained observations with the CSO at 330 μm and were able to resolve the outer edge of the disk. They found that the structure appeared somewhat clumpy, suggestive of possible resonant trapping of material in the disk. Hughes et al. (2011) observed the source with the Submillimeter Array and were able to resolve the inner edge of the outer disk for the first time, finding it to be at roughly ∼ 150 AU. Their results showed that the outer disk material was likely extended rather than concentrated in a narrow ring. Hughes et al. (2011) do not see evidence for asymmetry in the disk structure in their data. In Matthews et al. (2014), new Herschel images between 70 and 500 μm were obtained, with strong detections at each band. The disk was resolved at the shorter wavelengths, and modeling was performed to find the full extent and orientation of the disk. They found the disk extended from ∼ 100 to 310 AU, closer than seen by Hughes et al. (2011), with an outer halo out to ∼ 2000 AU. They measured a relatively precise inclination for the disk of 26 ± 3∘. They also found that the disk was populated by larger grains, whereas the halo may have small grains.
Most recently, Booth et al. (2016) used ALMA to fully resolve the outer disk in HR 8799 at 1.34 mm. With their new map, they are able to precisely fit the inner edge to be 145 ± 12 AU and the outer edge 429−32+37 AU. They also find an inclination of 40 ± 5∘. This is inconsistent with the values near ∼ 25∘ preferred by Su et al. (2009) and Matthews et al. (2014). Interestingly, the astrometric fits to the orbits described in the previous section are often similarly discrepant (e.g., Esposito et al. 2013; Konopacky et al. 2016). The Booth et al. (2016) fits are the first that show the disk may be consistent with the spin orbit axis orientation of HR 8799A from asteroseismology (Wright et al. 2011). As a possible explanation for the disk modeling differences, Booth et al. (2016) suggest that there may actually be a difference in the populations traced by the submillimeter observations versus the far infrared, analogous to the warps seen in other debris disks.
Based on the best estimates of the disk orientation from ALMA, Fig. 5 shows our current best understanding of the components of the HR 8799 planetary system. The outer cyan region represents the location of the outer disk. Red, orange, green, and blue lines show notional orbits that are consistent with the planet astrometry, the orientation of the outer disk, and a 8:4:2:1 period ratio. The selected orbits are coplanar and low eccentricity and are drawn from the allowed orbits derived in Konopacky et al. (2016). We also show in brown the location of the inner debris ring. The halo component is not shown.
Summary of System Architecture
In summary, the combination of empirical and dynamical constraints on HR 8799 indicate that we are viewing the system at an inclination roughly ∼ 30∘–40∘ from face-on. The evidence suggests that the planets and an outer debris disk orbit the star in roughly the same plane, and this plane is perpendicular to the stellar spin axis. Astrometric data obtained up-to-date samples only ∼ 5–15% of the full orbits of the planets but is currently consistent with dynamical predictions that the planets are in an 8:4:2:1 mean-motion resonance, although other resonant configurations are also possible. It is generally believed that some resonances must be in place in HR 8799 for the planets to remain stable.
Limits on Additional Planets in the HR 8799 System
With the discovery of three planets, followed by a fourth a few years later (Marois et al. 2008, 2010), there has been speculation on the existence of additional planets in the HR 8799 system. The qualitative similarity of the structure of HR 8799 to the solar system, with four gas giant planets flanked by two rings of debris, begs the question of potentially lower mass planets interior to HR 8799e. In their discussion of the inner debris disk, Su et al. (2009) suggest that the lack of debris between the star and this disk was indicative of planets in that region. This region ( < 6 AU) is inaccessible with current technology and thus can only be explored with dynamical models. Goździewski and Migaszewski (2014) explored test particles interior to HR 8799e with very small mass, finding that in this innermost region, small particles could be stable for the lifetime of the system. In further trying to elucidate the structure of the inner debris disk, Contro et al. (2016) simulated one million test particles in the inner region of the system. They find that the likely extent of the disk is 6–8 AU (Fig. 5) and that there are likely pileups of material at resonances with HR 8799e, similar to the structure of the asteroid belt. They found that the region interior to 6 AU should be relatively clear, including the notional “habitable zone” in this system between 1.97 and 3.41 AU. They suggest that small planets could potentially reside in this region, though likely subjected to a significant amount of bombardment.
Between HR 8799e and the outer edge of the debris belt, there is dynamical room for an additional larger planet. Goździewski and Migaszewski (2014) use stability arguments to determine where a fifth giant planet could reside and still be compatible with the structure of the inner debris ring. They find that a close in “HR 8799f” could survive in resonance with HR 8799e at either 2:1 ratio with a mass between 2 and 4 MJup or a 3:1 ratio with a mass between 2 and 8 MJup. Some of this parameter space can be probed with the current generation of high-contrast imaging systems. Hinkley et al. (2011) observed HR 8799 using nonredundant masking, an interferometric technique where the telescope pupil is divided into a series of pinholes via a mask in order to enhance sensitivity and contrast at very small angular separations (e.g., Tuthill et al. 2000). Hinkley et al. (2011) found that no object more massive than 11 MJup could reside between 3 and 10 AU in the system. Skemer et al. (2012) hunted for large planets up to the 2:1 resonance location and found no evidence for objects comparable to or larger than HR 8799e at > 9 AU. In observations with SPHERE, Zurlo et al. (2016) could specifically test the predictions of Goździewski and Migaszewski (2014). They were able to rule out a planet between 4.5 and 7 MJup at the 3:1 resonance position and between 3 and 5 MJup at the 2:1 resonance. This leaves only a limited parameter space primarily around ∼ 2 MJup yet to be explored for an inner gas giant.
A few studies have searched for giant planets beyond HR 8799b. This is particularly relevant in light of the ALMA observations from Booth et al. (2016) showing the inner edge of the wide debris disk. Through modeling, Booth et al. (2016) show that HR 8799b cannot be responsible for sculpting the inner edge of the this disk. They suggest that another planet at wider separations could be driving the disk morphology. This planet could be roughly ∼ 1.25 MJup at a distance of 110 AU ( ∼ 3″), or smaller than this with a more eccentric orbit. The location of this putative planet is denoted on Fig. 5. In terms of observational constraints on wider planetary companions, Close and Males (2010) performed a search using Gemini and NICMOS images of the system which explored the separation regime between 5 and 15″, or roughly ∼ 200 and 600 AU. They found two-point sources at wide separations that proved to be background stars in proximity to the system and placed limits of ∼ 3 MJup for any wide companion in this region. They also searched out to several arcminutes for common proper motion companions and found no evidence for any at very wide separations. Hinz et al. (2010) also searched for very wide companions within 10∘ of HR 8799 using the Tycho-2 and UCAC2 catalogs cross-referenced with X-ray sources, again finding no evidence of co-moving companions.
Though many high-contrast imaging systems have fields of view wide enough to hunt for the putative companion suggested by Booth et al. (2016), most works have not explored this region extensively, as typical depths make them only sensitive to companions larger than a few Jupiter masses, and it is clear there are no planets this large beyond 100 AU. In order to probe to ∼ 1 MJup or below, very deep images are required. Since contrast concerns are less severe at these wider separations, it is likely that new observations in the next few years will probe much of the parameter space allowed for a companion beyond HR 8799b.
The HR 8799 planetary system remains the only directly imaged multiplanet system at the time of writing. The four coeval gas giants amenable to direct spectroscopy provide an opportunity not possible for any other system. Outstanding questions remain about the atmospheric properties such as the clouds in these systems, which can potentially be probed by long-term monitoring for variability (e.g., Apai et al. 2016). Further spectroscopy at all wavelengths, including higher-resolution spectroscopy that can probe detailed chemical abundances, could hold the keys to the formation pathways for these planets.
The rarity of systems like HR 8799 based on surveys of hundreds of young stars begs the question of how unusual the system truly is. These surveys have found that the frequency of massive planets between ∼ 1 and 14 MJup at separations between 10 and a few 100 AU is at most a few percent (e.g., Nielsen et al. 2013; Biller et al. 2013; Brandt et al. 2014; Chauvin et al. 2015; Galicher et al. 2016). Given the complex dynamical structure of the system, including the difficulty of achieving stability, it is possible that HR 8799-like systems form often but then rapidly become unstable. This is an intriguing possibility in light of the number of free-floating planetary mass objects identified in microlensing surveys (e.g., Clanton and Gaudi 2017). As such, the potential for HR 8799 itself to become dynamically unstable in the future is certainly not negligible. Alternatively, it could be that HR 8799 represents a rare “massive” end of the planet formation process, with most disks tending to form smaller planets at closer separations. Further study of all aspects of the system, including the star, planets, and debris, with increasingly advanced instrumentation, will provide the tools needed to unlock the mysteries of this fascinating planetary system.
The authors wish to acknowledge Christian Marois, whose persistence made HR 8799 the system that keeps on giving.
- Barrado y Navascués D, Stauffer JR, Song I, Caillault J-P (1999) The Age of β Pictoris. ApJ 520:L123Google Scholar
- Esposito S, Mesa D, Skemer A et al (2013) LBT observations of the HR 8799 planetary system. First detection of HR 8799e in H band. A&A 549:A52Google Scholar
- Lavie B, Mendonça JM, Mordasini C et al (2016) HELIOS-Retrieval: An Open-source, Nested Sampling Atmospheric Retrieval Code, Application to the HR 8799 Exoplanets and Inferred Constraints for Planet Formation. arXiv:1610.03216Google Scholar
- Oppenheimer BR, Baranec C, Beichman C et al (2013) Reconnaissance of the HR 8799 Exosolar System. I. Near-infrared Spectroscopy. ApJ 768:24Google Scholar
- Pueyo L, Soummer R, Hoffmann J et al (2015) Reconnaissance of the HR 8799 Exosolar System. II. Astrometry and Orbital Motion. ApJ 803:31Google Scholar
- Torres CAO, Quast GR, Melo CHF, Sterzik MF (2008) Young Nearby Loose Associations. Handbook of star forming regions, vol II, 5, 757. ASP Monograph Publications, San FranciscoGoogle Scholar