Draft version September 1, 2021 Typeset using LATEX twocolumn style in AASTeX631

Obliquity Constraints on the Planetary- Companion HD 106906 b

Marta L. Bryan,1 Eugene Chiang,1 Caroline V. Morley,2 Gregory N. Mace,2 and Brendan P. Bowler2

1Department of Astronomy 501 Campbell Hall University of California Berkeley Berkeley, CA 94720-3411, USA 2Department of Astronomy The University of Texas at Austin Austin, TX 78712, USA

ABSTRACT We constrain the angular momentum architecture of HD 106906, a 13 ± 2 Myr old system in the ScoCen complex composed of a compact central binary, a widely separated planetary-mass tertiary HD 106906 b, and a nested between the binary and tertiary orbital planes. We measure the orientations of three vectors: the companion spin axis, companion orbit normal, and disk normal. Using near-IR high-resolution spectra from Gemini/IGRINS, we obtain a projected rotational velocity of v sin ip = 9.5 ± 0.2 km/s for HD 106906 b. This measurement together with a published photometric rotation period implies the companion is viewed nearly pole-on, with a line-of-sight spin axis inclination ◦ ◦ of ip = 14 ± 4 or 166 ± 4 . By contrast, the debris disk is known to be viewed nearly edge-on. The likely misalignment of all three vectors suggests HD 106906 b formed by gravitational instability in a turbulent environment, either in a disk or cloud setting.

Keywords: systems – Exoplanet formation – Exoplanet evolution – High resolution spec- troscopy – Astrostatistics

1. INTRODUCTION These processes and others can apply to . Planetary obliquity measurements inform our under- Theoretical work shows that obliquities can be excited standing of how planets form and evolve. A planetary through secular spin-orbit resonances created by planet- obliquity is the mutual inclination between the planet’s planet or planet-disk interactions (e.g. Millholland & spin axis and its orbit normal. Up until last , only Laughlin 2019; Millholland & Batygin 2019; Li 2021). solar system planets had measured obliquities. In this Kozai-Lidov oscillations from an external perturber can single system we find a wide range of orientations — be expected to produce significant planetary and stellar Uranus is on its side, Venus is upside-down, and Earth obliquities (e.g. Storch et al. 2014; Martin et al. 2014). is tilted by 23◦ which gives us our seasons. From these Spin axes may also be tilted at the time of formation: spin rates and directions we can infer planet formation turbulence in self-gravitating disks is expected to pro- histories. The terrestrial and ice giant planets likely ex- duce a dispersion of spin axis directions for fragmenting clumps (Bryan et al. 2020b; Jennings & Chiang 2021).

arXiv:2108.13437v1 [astro-ph.EP] 30 Aug 2021 perienced giant impacts, tidal friction, and gravitational forcing (e.g. Dobrovolskis 1980; Lissauer & Kary 1991; A planet’s obliquity can be constrained from three Laskar & Robutel 1993; Touma & Wisdom 1993; Correia observables: the projected rotation rate v sin i of the 2006; Schlichting & Sari 2007; Reinhardt et al. 2019). planet, its photometric rotation period Prot, and a 3D While the spin axes of Jupiter and Saturn may have ini- orbit. Combining v sin i, Prot, and a radius estimate tially both been aligned with the angular momentum of yields the line-of-sight spin axis inclination of the planet, the broader circumstellar disk, secular spin-orbit reso- and the orbit plane gives the . At nances driven by orbital migration have been invoked to present, the only planets amenable to these measure- explain Saturn’s 27◦ obliquity (e.g. Ward & Hamilton ments are ∼25 young super-Jupiters discovered by direct 2004; Nesvorn´y 2018). imaging campaigns (e.g. Bowler 2016). Because these objects are young (. 100 Myr old) and massive (∼10 – 2 Bryan et al.

20 MJup), they are relatively bright, and their large sep- 2016). We seek to constrain three angular momentum arations from their host (& 50 AU, & 1 arcsec) help vectors: the planetary spin axis, planetary orbit nor- ensure that their fluxes are not buried beneath the glare mal, and debris disk normal. We do not include the of their host stars. It is thus feasible to extract spec- binary system in our analysis as the angular mo- tra and light curves for the planets themselves, thereby mentum vectors for the binary orbit and stellar spins measuring v sin i and Prot. are unknown. While recent work has shown that close, However, it is exceptionally rare to obtain all three circular binaries have orbital planes that are more likely of these observables for a single object. To date to be aligned with the planes of their circumbinary de- ∼15 planetary-mass companions have measured rotation bris disks (Czekala et al. 2019), the central binary in rates (see Table 3 in Bryan et al.(2020a)). Only com- HD 106906 has an that is too long and an panions 2M0122 b and VHS 1256-1257 b have both a eccentricity that is too high to safely make the assump- measured v sin i and Prot. Some objects with measured tion that the binary plane and the debris disk plane are Prot are too faint to extract a spectrum and measure coplanar. v sin i. Others that have measured v sin i’s do not have This paper is organized as follows. In Section 2 we detectable rotational modulations in their light curves, describe our high-resolution spectroscopic observations precluding a Prot constraint. Some of these companions, with IGRINS/Gemini. Section 3 lays out measurements such as VHS 1256-1257 b, are so far from their host stars of the line-of-sight inclinations of the planetary spin, or- that constraining the 3D orbit is not feasible. Prior to bital, and disk angular momentum vectors, and gives this work there was only one system with all three pieces constraints on the true 3D angles between each pair. We in hand. discuss what physical scenarios could account for these Bryan et al.(2020b) placed the first constraint on the constraints in Section 4, and present our conclusions in obliquity of a planetary-mass object outside the solar Section 5. system. This study focused on the 120 Myr old sys- tem 2MASS J01225093–2439505, which comprises a 0.4 2. OBSERVATIONS M host star with a 12–27 MJup companion (hereafter 2.1. IGRINS/Gemini High-Resolution Spectroscopy 2M0122 b) orbiting at 52 AU (Bowler et al. 2013). Line- Observations of HD 106906 b with the Immersion of-sight inclinations for the planetary spin, stellar spin, Grating Infrared Spectrometer (IGRINS: Yuk et al. and orbital angular momentum vectors were measured 2010; Park et al. 2014) on the Gemini South telescope using projected rotational velocities v sin i’s for the star (Mace et al. 2018) were completed 2020 February 04, and companion, rotation periods Prot’s for the star and 07, 08, 09 UT as part of program GS-2020A-Q-135 (PI: companion, and an astrometric orbit for the compan- Bryan). These observations simultaneously covered H ion. There is evidence that the true stellar obliquity and K bands from ∼1.45 – 2.52 µm. We observed the K is small and the true companion obliquity is large, al- = 15.5 mag companion with individual image exposure though there are large uncertainties because of the un- times of 1528 seconds. All observations were taken with known orientation of the spin axes in the sky plane. a slit orientation perpendicular to the 307.3◦ position A promising scenario that could account for these mu- angle between the host star and the companion (angu- tual inclinations is formation via instability in a gravito- lar separation 7.1”) in order to prevent a flux gradient turbulent disk, wherein turbulent eddies of gas spinning across the slit. On 2020 February 04 UT we acquired in a variety of directions collapse under their own self- three pairs of AB nodded exposures, amounting to three gravity, yielding a range of obliquities for the resulting epochs of observation in ∼2.5 hours of on-source integra- objects (Bryan et al. 2020b; Jennings & Chiang 2021). tion time. On the nights of 2020 February 07, 08, and 09 Here we present constraints on a second extraso- UT ABBA-nodded exposures were combined, producing lar planetary-mass companion obliquity. We study three additional epochs. In total, there were six epochs HD 106906, a 13 ± 2 Myr old system with a central of observation from four nights in early February 2020. close binary ( 1.37 and 1.34 M , orbital period 49.233±0.001 days, and eccentricity 0.669±0.002), or- 3. ANALYSIS +1.7 bited by an 11.9−0.8 MJup companion at a projected 3.1. Measuring v sin i for HD 106906 b separation of ∼737 AU (Bailey et al. 2014; Nguyen et al. We reduce all data with the IGRINS Pipeline Pack- 2021). This system also hosts an asymmetric debris disk age (PLP; Lee & Gullikson 2016). The package uses AB with a vertically thin eastern side that extends to over pairs of slit-nodded spectra to sky subtract and then op- 550 AU, and a vertically thick western side that reaches timally extract the target flux. Wavelength calibration a radius of ∼ 370 AU (Kalas et al. 2015; Lagrange et al. is carried out using both OH sky emission and telluric HD106906b Obliquity 3 absorption from the A0V star observed immediately be- ley et al. 2021, Morley et al. in prep.). These models are fore or after the target. The A0V star also serves as a tel- calculated assuming that the atmosphere is in radiative– luric standard, and when the target spectrum is divided convective and chemical equilibrium, following the ap- by the A0V spectrum the target is corrected for both proach of Marley et al.(1999); Saumon & Marley(2008); telluric absorption and the instrument profile. The final Morley et al.(2012), with updated chemistry and opaci- product of the PLP is a wavelength-calibrated spectrum ties as described in (Marley et al. 2021). We assume Teff of the target star, with flux in counts, and corresponding = 1820 K and log(g) = 4.0 for HD 106906 b, following signal-to-noise spectrum. While reduced spectra were measurements of Teff = 1820 ± 240 K and log(Lbol/L ) produced across the wavelength range ∼1.45–2.52 µm, = -3.65 ± 0.08 using medium resolution spectra from we only consider the K -band spectra in subsequent anal- VLT/SINFONI (Daemgen et al. 2017) and converting yses given the low signal-to-noise ratio (SNR) of the H - log(Lbol/L ) and system age to log(g) using hot-start band spectrum. This companion is brightest in K -band evolutionary models (Burrows et al. 1997). The Sonora (1.85 – 2.52 µm), although we find that some K -band models generated for HD 106906 b have solar orders are also unusable due to low SNR. and solar carbon-to-oxygen ratio (C/O), and include sil- Instrumental resolution and rotation both produce icate, iron, and corundum clouds with a sedimentation line broadening, and these two sources of broadening are efficiency fsed = 2 as described in Ackerman & Marley degenerate. It is thus important to accurately measure (2001). the resolution in order to accurately measure v sin i. Us- We compare this “data” CCF to “model” CCFs, ing observed standard star spectra, we first select four where each model is calculated by cross-correlating a IGRINS orders spanning wavelengths 2.293–2.325 µm, model atmosphere broadened by the instrumental line 2.236–2.267 µm, 2.182–2.212 µm, and 2.105–2.135 µm. profile with that same model additionally broadened by For each of the six epochs of data, we use the molecfit some rotation rate and offset by a (RV). routine, which simultaneously fits a telluric model and We perform this comparison in a Bayesian framework an instrumental profile defined by a single Gaussian ker- using MCMC to fit for three free parameters: v sin i, nel, to the spectrum (Smette et al. 2015; Kausch et al. RV, and instrumental resolution. We use uniform priors 2015). We take the median of these 24 instrumen- on v sin i and RV. For the instrumental resolution, we tal resolution measurements as the instrumental reso- use a Gaussian prior with a mean of 48599 and stan- lution to use in our analysis, and the standard deviation dard deviation of 1514 to account for uncertainties on to be the uncertainty. From this analysis we find R the measured resolution. = 48599±1514. We also check whether the resolution The log-likelihood function used in our MCMC frame- changes significantly within an order. We select one of work is given by the epochs taken on UT February 04 2020, and divide n  2 X mi − di each of the four orders into five parts (each ∼ 400 pixels log L = −0.5 , (1) σ across). We find that the resulting instrumental resolu- i=1 i tion values within each order are consistent with each other and with the global resolution measurement. where d is the “data” CCF, m is the “model” CCF, and In the wavelength-calibrated and telluric-corrected σi is the CCF error at position i. We calculate uncer- output spectra from the IGRINS reduction routine, we tainties on the “data” CCF using the jackknife resam- remove artifacts from strong sky lines that manifested pling technique. In this case, uncertainties are given by as spikes in the data. In addition, we find that the short n (n − 1) X 2 wavelength end of each spectrum contains less flux as the σ2 = (x − x) , (2) jackknife n j instrument blaze falls off. We cut the leftmost 20 – 70 j=1 pixels off of each reduced spectral order, where the cut- where n is the total number of samples. We define a sam- off value grew with increasing order number (decreasing ple as one of data – there are six for HD 106906 wavelength). b. x is the “data” CCF calculated using all epochs of With these spectra of HD 106906 b, we want to mea- j data except the jth epoch, and x is the “data” CCF sure the amount of line broadening due to the rota- calculated using all epochs of data. tional velocity. To do so, we calculate the “data” cross- Before undertaking a joint fit of multiple orders to de- correlation function (CCF) between the observed spec- termine v sin i, we first consider each order in K -band trum and a model atmosphere, where the model has individually. We compute “data” CCFs for 22 orders been broadened to the instrumental resolution. We use spanning wavelengths 1.85 – 2.42 µm (the first three an atmospheric model from the Sonora model grid (Mar- spectral orders extended to longer wavelengths than 4 Bryan et al. those covered by our models), and determine the signif- icance of the peak (if present) by calculating the ratio of the peak height to the standard deviation of the CCF outside the central peak. We exclude orders with peaks with less than 5σ significance. This cut leaves us with 15 orders running from 1.99 – 2.42 µm (excluded orders have significant telluric features and lower SNR spec- tra). We fit each order individually to get independent estimates for v sin i, and find that they are consistent within their uncertainties. We then perform a joint fit, calculating a “data” CCF using all 15 orders and fitting models across that entire 1.99 – 2.42 µm wavelength range. The measured projected rotation rate for HD 106906 b is v sin i = 9.5 ± 0.2 km/s (see Figures1-3 for reference). Figure 2. Cross correlation function between orders 4–19 (1.99 – 2.42 µm) of the observed spectrum with a model at- mosphere broadened to the instrumental resolution (black points), shown with 1σ uncertainties shaded in gray cal- culated using jackknife resampling technique. The cross- correlation function between a model atmosphere broadened to the instrumental resolution, and that same model addi- tionally broadened by the best fit rotation rate and shifted by the best fit velocity offset is shown in teal.

Figure 3. Cross correlation function between orders 4–19 (1.99 – 2.42 µm) of the observed spectrum with a model at- mosphere broadened to the instrumental resolution (black points), shown with 1σ uncertainties shaded in gray cal- culated using jackknife resampling technique. The cross- correlation functions between a model atmosphere broadened to the instrumental resolution, and that same model addi- tionally broadened by a series of rotation rates (0, 5, 10, 15, Figure 1. Orders 4 (top) and 7 (bottom) spectra for HD 20, 25, 30 km/s) are shown in color. 106906 b (red), overplotted with a model atmosphere broad- ened to the best-fit rotation rate. = 1820±240 K, and we converted log(Lbol/L ) to log(g) = 4.0±0.5 using hot start evolutionary models (Bur- We now consider how our modeling assumptions could rows et al. 1997). We take the 1σ errors on these values impact the measured v sin i. First we test our choice of and generate four models: (1580K, 3.5 dex), (1580K, Teff and log(g). From Daemgen et al.(2017) we have Teff 4.5 dex), (2060K, 3.5 dex), and (2060K, 4.5 dex). We HD106906b Obliquity 5 calculate new v sin i values with each of these models to Table 1. Model Tests and Resulting test the possible impact of the measured uncertainties v sin i’s on our adopted Teff and log(g), and found that all v sin i values were consistent with our original measurement at Model v sin i the ≤ 2σ level (see Table1). Original 9.53±0.24 km/s Another assumption we make when generating atmo- 1580 K, 3.5 dex 9.88 (+0.25 -0.24) spheric models is a C/O ratio. In our original model we 1580 K, 4.5 dex 8.85 ± 0.22 assume a solar (0.54) C/O value, and here we test three 2060 K, 3.5 dex 9.89 (+0.23 -0.26) 2060 K, 4.5 dex 9.20 (+0.22 -0.21) additional ones: 0.25×solar, 0.5×solar, and 1.5×solar. 0.25× solar C/O 9.63 (+0.24 -0.23) When we implement these models in our MCMC frame- 0.5× solar C/O 9.69 (+0.28 -0.24) work, we find that resulting v sin i values are consistent 1.5× solar C/O 8.73 (+0.25 -0.20) with the original value to ≤ 0.5σ for the sub-solar C/O 0.1×P 9.87 (+0.20 -0.22) ratios, and differ by 2.3σ for the 1.5×solar model (see 10×P 8.84 (+0.33 -0.28) Table1). While not significant, this tentative offset in v sin i due to higher C/O suggests that abundance as- sumptions can become important for v sin i’s given our measurement precision of 0.2 km/s. objects (i.e. closer to pole-on) have more attenuated Finally, we explore the impact that uncertainties in brightness changes (Vos et al. 2017). pressure broadening can have on the measured rota- The photometric rotation period for HD 106906 b was tional velocity. Pressure broadening is a degenerate ef- published by Zhou et al.(2020). The authors used the fect along with instrumental broadening and rotational Hubble Space Telescope Wide Field Camera 3 (WFC3) line broadening – higher pressure broadening with the near-IR channel in time-resolved direct imaging mode to same instrumental resolution leads to less rotational line observe HD 106906 b in three bands: F127M, F139M, broadening and a correspondingly smaller v sin i. To test and F153M. Using techniques such as two-roll differen- our pressure broadening assumptions, we run two mod- tial imaging and hybrid point-spread function model- els with modified molecular opacities, where molecular ing yielded ∼1% precision in the light curves across all cross sections were 10× and 0.1× the actual pressure three bands. Fitting the light curve with a sinusoid re- for the whole profile. This simulates a scenario where sults in a period of 4.1±0.3 hours and an amplitude of pressure broadening parameters that are used to cre- 0.49±0.12%. ate the molecular cross sections are off by an order of Zhou et al.(2020) present several caveats to the rota- magnitude. Collision-induced opacity of hydrogen and tion period measurement. First, they find only marginal helium is treated separately for all models, using the evidence of variability, with a significance of 2.7σ. Be- standard pressure for each layer. When we recalculate cause of this tentative detection, when fitting the light v sin i values using these new models, we find that for curves the authors applied a strict sinusoidal shape to the 0.1× model (which under predicts the amount of the photometric modulations, and could not investigate pressure broadening) the resulting rotation rate is only whether the light curve could have multiple peaks. Pre- 1.0σ away from the original value, and the 10× model viously, Apai et al.(2017) found that for 3 L/T tran- (which over predicts pressure broadening) produces a sition brown dwarfs with high signal-to-noise data and v sin i that is 1.7σ lower (see Table1). Both values are extremely long baselines (> 1 year), the power spectra consistent with the original rotation rate measurement. of their light curves produced peaks at both the full ro- tation period of the object as well as half the rotation period. A more recent analysis of long baseline high SNR photometry of Luhman 16 A and B found a simi- 3.2. Measuring Prot,p for HD 106906 b lar result, with peaks at both the full and half rotation Periodic features in substellar light curves can be period (Apai et al. 2021). While this raises the question produced by cloud patchiness or planetary-scale waves, of whether the detected 4 hour rotation period of HD which manifest as longitudinal bands produced by zonal 106906 b is the full or half rotation period, both higher circulation (Apai et al. 2017, 2021). Lower gravity ob- quality and theoretical light curves of brown dwarfs show jects typically have higher variability amplitudes and that the full rotation period dominates the signal in the higher intrinsic variability rates (Metchev et al. 2015; power spectrum for a given light curve (Zhang & Show- Vos et al. 2019). This observed variability appears to be man 2014; Apai et al. 2017, 2021). In addition, periodic- impacted by viewing geometry – more highly inclined ity in the light curves of Jupiter and Neptune correspond 6 Bryan et al. to their full rotation periods (Karalidi et al. 2015; Simon code (Augereau et al. 1999) as an optically thin, inclined et al. 2016; Ge et al. 2019). We thus assume that the ring centered on the host binary. The authors assumed a full rotation period of HD 106906 b is 4.1±0.3 hours. dust density distribution that peaks at radius r0 and has Another caveat to this rotation period measurement is a power law slope of αin inside of r0 and αout outside of that photometric modulations for HD 106906 b are only r0. In addition to r0 and αout, fitted model parameters detected in the bluest band (F127M), and not in the include the inclination of the disk id, the position angle other two bands. However, for the majority of substel- (PA), a scaling factor to match the total flux of the disk, lar objects, rotational modulations are wavelength de- and the Henyey-Greenstein coefficient g, which quanti- pendent and have higher amplitudes at shorter (bluer) fies how anisotropic the scattering is. With this model- ◦ wavelengths (e.g. Apai et al. 2013; Yang et al. 2015; Zhou ing, the authors find a disk inclination id = 85.3±0.1 . et al. 2016, 2019). Assuming a similar wavelength de- In Kalas et al.(2015), the authors estimate a disk in- ◦ pendence for the photometric modulations in HD 106906 clination of id ∼ 85 by assuming the disk is circular b as measured for 2M1207 b (Zhou et al. 2016), the clos- and translating the disk aspect ratio from their fitted est spectral type young companion analog to HD 106906 semi-major and minor axes to line-of-sight inclination. b with detected modulation, the authors predict that Because it is unclear whether the disk is rotating in a the modulation amplitude for the redder bands would prograde or retrograde fashion, there is a degeneracy in have been too small for them to observe. The detection id and Ωd pairs, where Ωd is the position angle (PA) of ◦ ◦ of modulation in only the bluest band is therefore con- the ascending node. Thus angles (id = 85 ,Ωd = 104 ) ◦ ◦ sistent with low overall amplitude variability and wave- and (id = 95 ,Ωd = 284 ) are equally likely. In this pa- length dependent modulations. per, we assume a bimodal distribution for id, with id = 85.3±0.1◦ and i = 94.7±0.1◦ defining the two Gaussian 3.3. Measuring i d o distributions. Given the wide projected separation of HD 106906 b (737 AU), detecting orbital motion and placing con- 3.5. Measuring ip straints on orbital parameters requires long baseline pre- cision (e.g. Bowler et al. 2020). Recently, We combine Prot and our measurement of v sin i to Nguyen et al.(2021) detected orbital motion of this com- determine the line-of-sight spin axis inclination of the panion using 14 of astrometry measurements be- companion ip. However, simply computing the inclina- tween 2004 and 2017. All data were taken with HST tion as: using a combination of the Advanced Camera for Sur-  v sin i  veys (ACS), the Space Telescope Imaging Spectrograph sin i = (3) (STIS), and the Wide Field Camera 3 (WFC3). By 2πR/Prot cross-registering background star locations in the HST does not account for correlations between relevant pa- images with the Gaia astrometric catalog, Nguyen et al. rameters (Masuda & Winn 2020). For example, v and (2021) calculated astrometry to high precision (at the v sin i are not statistically independent given that v sin i sub-pixel level). To obtain constraints on the orbital is always less than v. We therefore follow the method parameters of HD 106906 b, the authors used the open- described in detail in Masuda & Winn(2020) and sum- source Python package orbitize!1 (Blunt et al. 2019) marized for the application to HD 106906 b below. to perform an orbit fit to the assembled astrometric mea- Given v as the equatorial rotational velocity and u = surements. The line-of-sight orbital inclination i from o v sin i as the projected rotation rate, we have the follow- these fits is 56+12◦. In this paper we use the posterior −21 ing two likelihood functions: distribution shown in Fig. 10 of Nguyen et al.(2021) when incorporating io in our analyses.

3.4. Measuring id Lv(v) = p(dv|v) (4) Scattered light imaging of the debris disk in the HD Lu(u) = p(du|u) (5) 106906 system with both GPI and SPHERE constrained the morphology of the disk (Kalas et al. 2015; Lagrange where dv and du are the datasets from which these like- et al. 2016). In Lagrange et al.(2016), the disk de- lihood functions are calculated. In our case, Lu(u) is tection with SPHERE was modeled using the GRATER the probability distribution for v sin i that we deter- mined from our high-resolution spectra, a Gaussian with peak location 9.5 km/s and standard deviation 0.2 km/s. 1 https://github.com/sblunt/orbitize Lv(v) is the probability distribution for v, which we cal- HD106906b Obliquity 7 culate using v = 2πR/Prot. For the radius R we calcu- late the effective blackbody radius: s L R = 4 , (6) 4πσbTeff where L is the bolometric log(Lbol/L ) = - 3.65±0.08, σb is the Stefan-Boltzmann constant, and Teff is the effective temperature Teff = 1820±240 K (Daem- +0.37 gen et al. 2017). This yields a radius of 1.49−0.45 RJup. We produce a probability distribution for v by incorpo- rating uncertainties on Prot, log(Lbol/L ), and Teff in a Monte Carlo fashion. With Lu(u) and Lv(v) in hand, Masuda & Winn (2020) specify two key assumptions: Figure 4. Normalized posterior distribution of the line-of- 1. dv and du are independent, so the likelihood func- sight companion spin axis inclination ip (blue). This distri- tion for D = {dv, du} is separable: bution is bimodal and symmetric around 90◦ because we do not know whether this spin axis vector (which has direction- ality) is pointed towards us or away from us. Setting aside Lvu(v, u) = p(D|v, u) = p(dv|v, u) = the symmetric distribution above 90◦, just considering values (7) ◦ p(dv|v)p(du|u) = Lv(v)Lu(u) of ip < 90 we find that the mode and 68% highest proba- ◦ bility density interval of ip are 14 ± 4 . This distribution is 2. v and i are a priori independent, which means that compared to a random inclination distribution (black) whose values are drawn from a uniform distribution in cos i. the prior Pvi(v, i) is separable:

the companion spin axis, the companion orbit normal, Pvi(v, i) = Pv(v)Pi(i) (8) and the disk normal. These angles are given by: and

−1 p(v|i) = Pv(v); p(i|v) = Pi(i) (9) Ψop = cos (cos ip cos io + sin ip sin io cos(Ωo − Ωp)) (11) Given these assumptions, the posterior PDF for cos i can be written as: −1 Ψdp = cos (cos ip cos id + sin ip sin id cos(Ωd − Ωp)) Z (12) p 2 p(cos i|D) ∝ Pcosi(cos i) Lv(v)Lu(v 1 − cos i)Pv(v)dv (10) −1 where Pcosi(cos i) is uniform between 0 and 1 and the Ψod = cos (cos id cos io + sin id sin io cos(Ωo − Ωd)) prior on the rotation rate Pv(v) is uniform from 0 to (13) break-up speed. where Ψop is the true companion obliquity, Ψdp is the Converting this PDF in cos i to a PDF in i yields the true spin-disk mutual inclination, and Ψod is the true distribution shown in Figure4. We note that the poste- orbit-disk mutual inclination. The position angles Ωo, rior distribution for ip is bimodal and symmetric around Ωd, and Ωp measure how the orbit, disk, and companion 90◦ simply because we do not know whether this spin spin axis, respectively, are oriented on the sky plane. axis vector (which has directionality) is pointed towards The nodal angle Ωp is unknown. us or away from us. Therefore the mode and 68% high- The difference between the line-of-sight inclination of ◦ est probability density interval (HPDI) of ip is 14 ± 4 the companion spin axis ip and that of the orbit normal ◦ ◦ ◦ for ip < 90 , and 166 ± 4 for ip > 90 . io yields a lower limit on the true de-projected obliquity Ψop (Bowler et al. 2017): 3.6. Measuring the 3D Spin-Orbit Architecture of the HD 106906 b System Ψop > |ip − io| . (14) Our goal is to measure the true 3D angles between the three angular momentum vectors in this system – Similarly: 8 Bryan et al.

Ψdp > |ip − id| (15)

Ψod > |io − id| . (16)

Figure5 shows the posteriors for |ip − io|, |ip − id|, and |io − id|. We also plot a random distribution in black for comparison, where ip, io, and id are all drawn from uniform distributions in cos i. We find that the 68% highest probability density interval (HPDI) for |ip − io| lies between [32, 119] degrees. For |io − id| the 68% HPDI is [16, 48] degrees, and for |ip − id| we have the tightest 68% HPDI of [69, 83] degrees. Since these projected angles are all lower limits on the true 3D angles, we can say that our results tend to favor more “misaligned” orientations (defined here as mutual inclinations of 20–180◦). A schematic illustration of the line-of-sight architecture of the system is shown in Fig- ure6.

We now calculate full probability distributions for all Ψ’s using equations (11)-(13). For equations (11) and (12) we assume Ωp is randomly drawn from a uni- form distribution between 0 and 2π. For equation (13), +26◦ both Ωo and Ωd have been measured: Ωo = 99−28 or +25◦ ◦ ◦ 279−29 , and Ωd = 104.4 ± 0.3 or 284 ± 0.3 (Kalas et al. 2015; Lagrange et al. 2016; Nguyen et al. 2021). Figure7 shows the resulting probability distributions for Ψop,Ψdp, and Ψod, along with a random mutual incli- nation distribution Ψrandom in black for reference. Not knowing Ωp along with poor constraints on HD 106906 b’s orbital elements (a consequence of the com- panion’s wide orbital separation) leads to broad poste- rior distributions for the true de-projected angles Ψop and Ψod. By comparison, Ψpd is remarkably well- constrained. All of these posteriors are bimodal, re- flecting symmetries across 90◦. For each Ψ posterior we calculate the mode and 68% HPDI for each half of the distribution below and above 90◦. We find that the Figure 5. Top panel: Posterior distribution for the line- +22◦ +16◦ of-sight projected companion obliquity (blue), which has a true companion obliquity Ψop is 55−16 or 125−22 , the +6◦ +8◦ 68% highest probability density interval (HPDI) of [32, 119] true spin-disk angle Ψdp is 84 or 96 , and the true −8 −6 degrees. Middle panel: Posterior distribution for the line- mutual inclination between the orbit and disk normals +20◦ +15◦ of-sight projected spin-disk angle (purple), which has a 68% Ψod is 39−15 or 141−20 (Table2). Each of these de- HPDI of [69, 83] degrees. Bottom panel: Posterior distribu- projected mutual inclinations deviates distinctly from tion for the line-of-sight projected orbit-disk angle, which has the geometric prior – both Ψop and Ψod favor angles a 68% HPDI of [16, 48] degrees. These line-of-sight projec- ◦ away from 90 which is where the geometric prior peaks, tions are lower limits on the true 3D mutual inclinations Ψop, ◦ and while Ψdp peaks around 90 it is a much tighter con- Ψdp, and Ψod. The posterior distributions are compared to a straint than a random distribution. random projected inclination distribution (black), where io, We quantify the probability that each Ψ distribution id, and ip have all been drawn from uniform distributions in cos i. yields an “aligned” state, which we define as Ψ ∈ (0, 20) degrees, or a “misaligned” state where Ψ ∈ (20, 180) where p(m|D) is the probability of misaligned state m degrees. The Bayesian odds ratio is p(m|D)/p(a|D), HD106906b Obliquity 9

Table 2. Measured Parameters

Parameter Measured Value Ref

vp sin ip 9.5±0.2 km/s This work Prot,p 4.1±0.3 hrs Zhou et al.(2020) ip 14±4 or 166±4 deg This work +12 io 56−21 deg Nguyen et al.(2021) id 85.3±0.1 or 94.7±0.1 deg Kalas et al.(2015); Lagrange et al.(2016); Nguyen et al.(2021) |io − ip| [32, 119] deg This work |ip − id| [69, 83] deg This work |id − io| [16, 48] deg This work +22 +16 Ψop 55−16 or 125−22 deg This work +6 +8 Ψdp 84−8 or 96−6 deg This work +20 +15 Ψod 39−15 or 141−20 deg This work

Note—The three i inclinations presented here are all along our line-of-sight. The angle ip is symmetric about 90◦ due to the fact that we do not know whether this spin angular momentum vector is pointing towards us or away from us. The angle id has two solutions because it is unclear whether the disk is rotating prograde or retrograde, so there are two combinations of id and Ωd, where Ωd is the position angle of the ascending node, that are equally likely. The line-of-sight mutual inclinations |io − ip|, |ip − id|, and |id − io| are lower limits on the true de-projected angles Ψop,Ψdp, and Ψod. Here we quote the mode and 68% highest probability density intervals for these angles.

in Fig.7) from 20 - 180 ◦, while the latter probability is �# line of sight �# obtained by integrating the posterior distribution from 0 - 20◦. We compute these integrals over the posterior dis- � " �! tributions for all three mutual inclinations Ψop,Ψdp, and Ψod, as well as for the geometric prior distribution �" �! which is uniform in cos Ψrandom (Fig.7, black curve). In the absence of any data, the geometric prior yields an odds ratio favoring misalignment to alignment at 32:1 (2.2σ). The odds ratios for the true companion obliquity Ψop and the orbit-disk mutual inclination Ψod are only slightly larger, 39:1 (2.3σ) and 40:1 (2.3σ), respectively. Nevertheless, the fact that the shapes of both posteriors differ significantly from that of the prior indicates that the odds favoring misalignment for both Ψop and Ψod Figure 6. 3D architecture of the HD 106906 system showing are not entirely driven by the prior. We can say there is ~ how each of the three angular momentum vectors, namely Lp tentative evidence that both of these mutual inclinations ~ ~ for the companion spin, Lo for the orbit, and Ld for the debris are large. By comparison, the true spin-disk angle Ψ disk, are oriented along our line-of-sight, with corresponding dp has an odds ratio that favors misalignment much more angles ip, io, and id. Both ip and id are symmetric around ◦ strongly than does our random prior – 97081:1 (4.4σ). 90 (L~ p and L~ d could be pointing towards or away from us). Mutual inclinations between each set of vectors favor We can say with confidence that this system contains an misalignment. Note the purpose of this figure is to illustrate edge-on disk and a planet spinning nearly pole-on; even the orientation of the three angular momentum vectors – it allowing for sky-plane degeneracy, our measurements in- does not accurately depict other separate system properties dicate that the companion spin axis and disk normal are (i.e. disk asymmetry, disk inner and outer radius – see Fig. strongly misaligned. 9 in Nguyen et al.(2021) for a scattered light image of the system showing some of these properties). 4. DISCUSSION: POSSIBLE FORMATION HISTORIES given data D, and p(a|D) is the probability of an aligned For the HD 106906 system, we have strong evidence state a. The former probability is the integral of the that the spin axis of the planetary-mass companion posterior distribution p(Ψ|D) (the colored histograms (PMC) and the orbit normal of the debris disk are mis- 10 Bryan et al.

the PMC orbit and debris disk, are large. We proceed on the assumption that all three vectors are mutually misaligned and consider possible origin scenarios. As in the case of 2M0122 b, gravitational instabil- ity in a turbulent environment is a promising way to create tilted architectures. The setting can either be a gravito-turbulent disk around a protostar (Bryan et al. 2020b; Jennings & Chiang 2021), or a portion of a self- gravitating turbulent cloud that fragments into a binary (e.g. Bate et al. 2002; Bate 2009, 2018). Turbulent ed- dies spin in a variety of directions, and when those eddies gravitationally collapse, the objects they form should have a correspondingly wide range of spin directions. In situ formation of HD 106906 b in a disk would require a as large as 1000 au in radius; Class I disks range up to this size (Maury et al. 2019, see their Figure 9). Numerical simulations of fragmenting gravito-turbulent disks by Jennings & Chiang(2021) find that obliquities of nascent fragments can be as high as 45◦, and that sub- sequent collisions between fragments can raise or lower obliquities. Turbulence also imparts vertical velocities to disk fragments, driving them out of the midplane; or- bital inclinations up to 20◦ are suggested by these sim- ulations. In population synthesis calculations by Bate (2018) of stellar binaries fragmenting from a turbulent cloud, roughly 2/3 of circumstellar disks are inclined by more than 30◦ relative to the binary orbit, for bina- ries with semimajor axes between 100 and 1000 au (see their Figure 20). This result of frequent disk-orbit mis- alignments, in combination with their finding that most (80%) of stellar spin axes are within ∼45◦ of circumstel- lar disk normals (their Figure 23), would seem to imply that spin-orbit angles are typically large (obliquities are not explicitly calculated in Bate 2018). In sum, all the misalignments indicated by our observations seem possi- ble to account for by turbulent gravitational instability. Another scenario is that the PMC formed as an iso- lated object via molecular cloud fragmentation and was subsequently captured by a star. Binaries can form in fly-by events if sufficient energy is dissipated during the encounter; circumstellar disks can provide this energy sink (e.g. Clarke & Pringle 1991a,b; Moeckel & Bally 2006; Offner et al. 2016; Bate 2018). Assuming the disk Figure 7. Top panel: Normalized posterior distribution surrounding the stellar host is the dominant sink (as for the true companion obliquity Ψop (blue). Middle panel: opposed to any smaller and less massive disk orbiting Normalized posterior distribution for the true spin-disk angle the PMC), we have no reason to expect the planet’s Ψdp (purple). Bottom panel: Normalized posterior distribu- spin axis to be aligned with its orbit normal or the disk tion for the true orbit-disk angle Ψod (green). These pos- terior distributions are compared to geometrically random normal. The captured PMC’s orbital plane may also mutual inclination distributions (black). be randomly oriented with respect to the circumstel- lar disk plane, at least initially; this is borne out in aligned. In addition, we find tentative evidence that the simulations by Bate(2018, see their Figure 22). How- 5 true PMC obliquity, and the mutual inclination between ever, these simulations lasted only up to ∼10 yr; on HD106906b Obliquity 11 longer timescales the PMC orbit could be gravitation- tations (defined here as mutual inclinations of 20–180 ally torqued into alignment with the circumstellar disk. degrees). Thus two if not three misalignments can be accommo- We further constrained the true 3D mutual inclina- dated in this encounter scenario. tions between the companion spin, companion orbit, and A handful of other mechanisms can produce only one disk angular momentum vectors: Ψop (true companion or two of the three possibly large mutual inclinations obliquity), Ψdp (true spin-disk mutual inclination), and observed in this system. A stellar fly-by can tilt the Ψod (true orbit-disk angle). Since we have no constraints orbit of the PMC, generating a large obliquity as well as on the companion spin axis orientation in the sky plane, a mutual inclination between the orbit and disk planes to compute distributions for Ψop and Ψdp we assumed (e.g. Laughlin & Adams 1998; Kenyon & Bromley 2004; that nodal angle Ωp was randomly and uniformly dis- Zakamska & Tremaine 2004; Parker & Quanz 2012; De tributed between 0 and 2π (see equations 11 and 12). Rosa & Kalas 2019). However, the PMC spin axis is To calculate a distribution for Ψod we used measured not expected to be materially altered by the fly-by, and values of Ωd and Ωo for the disk and orbit, respectively thus the observed misalignment between the spin axis (see equation 13). and the disk normal in HD 106906 is unexplained in The lack of a constraint on how the companion spin this scenario. Another way to generate obliquity is by axis is oriented in the sky plane combined with poor con- using the stellar tidal potential to tilt a circumplanetary straints on HD 106906 b’s orbit (given its wide projected disk and by extension its host planet (Lubow & Ogilvie separation) leads to broad posterior distributions for 2000; Martin & Armitage 2021). This scenario does not Ψop and Ψod, whereas Ψdp is more tightly constrained address the misalignment between the orbit and disk (Figure7). Since these posteriors are bimodal (given normals. symmetries about 90◦), we calculated the mode and 68% HPDI for each Ψ both above and below 90◦. We found +22◦ +16◦ +6◦ +8◦ 5. CONCLUSIONS that Ψop is 55−16 or 125−22 ,Ψdp is 84−8 or 96−6 , and Ψ is 39+20◦ or 141+15◦. All three angles exhibit In this study we constrained the orientation of the od −15 −20 marked differences from the geometric prior, with Ψ planetary spin, orbital, and disk angular momentum op and Ψ peaking at values away from the 90 degree max- vectors for the HD 106906 system. HD 106906 is a 13 od imum of the geometric prior, and Ψ showing a much ± 2 Myr system in ScoCen composed of a close binary dp tighter distribution around 90 degrees than the prior. of two F stars, a widely separated planetary-mass ter- We assessed how likely the three angular momentum tiary 106906 b, and an asymmetric debris disk interior vectors were to be “misaligned” or “aligned” (defined as to companion b’s orbit. Line-of-sight inclinations for mutual inclinations between 20–180◦ and 0–20◦ respec- the companion orbit i and debris disk i were previ- o d tively) by calculating an odds ratio for each angle. Both ously published (Kalas et al. 2015; Lagrange et al. 2016; the true obliquity Ψ and mutual inclination between Nguyen et al. 2021). The debris disk is known to be op the orbit and the disk Ψ tentatively favor misalign- viewed edge-on. Here we measured the line-of-sight in- od ment with ratios of 39:1 (2.3σ) and 40:1 (2.3σ), respec- clination of the companion spin axis i . We used near- p tively. The orientation between the disk normal and the IR high-resolution spectra from IGRINS/Gemini South companion spin axis Ψ strongly favors misalignment to measure rotational line broadening for HD 106906 b, dp at 97081:1 odds (4.4σ). We are seeing an edge-on disk and found a rotation rate of 9.5 ± 0.2 km/s. Combining orbited by a planet spinning nearly pole-on. this measurement with the photometric rotation period Given strong evidence that the planetary-mass com- yielded a companion spin axis inclination of 14 ± 4◦ panion (PMC) spin axis and disk normal are misaligned, (when considering inclinations < 90◦). We are seeing and tentative evidence that the true PMC obliquity and this companion nearly pole-on (Fig.6). the mutual inclination between the PMC orbit and de- Differences between line-of-sight inclinations yield bris disk are large, we considered various origin scenar- lower limits on the true 3D mutual inclinations. We ios. As in the case of 2M0122 b (Bryan et al. 2020b), computed the projected inclinations |i − i |, |i − i |, o p p d gravitational instability in a turbulent medium, either in and |i −i |. The projected companion obliquity |i −i | d o o p a circumstellar disk or cloud setting, is viable. Gravita- has a 68% highest density probability interval (HDPI) of tional collapse of turbulent eddies naturally yields large [32, 119] degrees. The projected orbit-disk angle |i −i | d o obliquities, and subsequent fragment interactions (col- has an HDPI of [16, 48] degrees. The projected spin-disk lisions and mergers) can further increase obliquity dis- inclination |i −i | is the most strongly constrained, with p d persions (Jennings & Chiang 2021). Disk turbulence an HDPI of [69, 83] degrees. These lower limits on the also excites non-zero orbital inclinations. In simulations 3D mutual inclinations favor more “misaligned” orien- 12 Bryan et al. of fragmentation in a turbulent molecular We thank Ian Czekala, Gaspard Duchene and Yifan cloud, disks and binary orbits are frequently misaligned, Zhou for helpful conversations. M.L.B. is supported especially for wide binaries (Bate 2018). Thus all ob- by the Heising-Simons Foundation 51 Pegasi b Fellow- served misalignments in HD 106906 are potentially ac- ship. C.V.M. acknowledges the support of the National counted for in a turbulent gravitational instability sce- Science Foundation grant number 1910969. B.P.B. ac- nario. Another possibility is that the PMC formed via knowledges support from the National Science Founda- molecular cloud fragmentation initially isolated and un- tion grant AST-1909209 and NASA Exoplanet Research bound from the star, and was subsequently captured by Program grant 20-XRP20 2-0119. the star in a dissipative fly-by event. In this encounter Based on observations obtained at the international scenario there is no reason to expect the PMC’s spin Gemini Observatory, a program of NSF’s NOIRLab, axis would be aligned with its orbit normal or the disk which is managed by the Association of Universities for normal, and the orbital plane could also be randomly Research in Astronomy (AURA) under a cooperative oriented with respect to the disk plane. In all of the agreement with the National Science Foundation. on above scenarios, HD 106906 b forms top-down by grav- behalf of the Gemini Observatory partnership: the Na- itational instability, and thus shares kinship with stars. tional Science Foundation (United States), National Re- This is only the second obliquity constraint for a search Council (Canada), Agencia Nacional de Investi- planetary-mass companion outside the Solar System. gaci´ony Desarrollo (Chile), Ministerio de Ciencia, Tec- The work of measuring a planet’s rotation speed, spin nolog´ıae Innovaci´on(Argentina), Minist´erioda Ciˆencia, period, and 3D orbit remains challenging. But the in- Tecnologia, Inova¸c˜oese Comunica¸c˜oes(Brazil), and Ko- sights into planet formation enabled by obliquity are rea Astronomy and Space Science Institute (Republic new, powerful, and unique. The bane of sky-plane de- of Korea). This work used the Immersion Grating In- generacies may be banished by a larger sample of sys- frared Spectrometer (IGRINS) that was developed un- tems like 2M0122 b and HD 106906 b which will enable der a collaboration between the University of Texas at statistical constraints. Austin and the Korea Astronomy and Space Science In- stitute (KASI) with the financial support of the Mt. Cuba Astronomical Foundation, of the US National Sci- ence Foundation under grants AST-1229522 and AST- 1702267, of the McDonald Observatory of the University of Texas at Austin, of the Korean GMT Project of KASI, and Gemini Observatory.

REFERENCES

Ackerman, A. S., & Marley, M. S. 2001, ApJ, 556, 872 Blunt, S., Wang, J., Angelo, I., et al. 2019, arXiv e-prints, Apai, D., Nardiello, D., & Bedin, L. R. 2021, ApJ, 906, 64, arXiv:1910.01756. https://arxiv.org/abs/1910.01756 doi: 10.3847/1538-4357/abcb97 Bowler, B. P. 2016, PASP, 128, 102001, Apai, D., Radigan, J., Buenzli, E., et al. 2013, ApJ, 768, doi: 10.1088/1538-3873/128/968/102001 121, doi: 10.1088/0004-637X/768/2/121 Bowler, B. P., Blunt, S. C., & Nielsen, E. L. 2020, AJ, 159, 63, doi: 10.3847/1538-3881/ab5b11 Apai, D., Karalidi, T., Marley, M. S., et al. 2017, Science, Bowler, B. P., Liu, M. C., Shkolnik, E. L., & Dupuy, T. J. 357, 683, doi: 10.1126/science.aam9848 2013, ApJ, 774, 55, doi: 10.1088/0004-637X/774/1/55 Augereau, J. C., Lagrange, A. M., Mouillet, D., Papaloizou, Bowler, B. P., Kraus, A. L., Bryan, M. L., et al. 2017, AJ, J. C. B., & Grorod, P. A. 1999, A&A, 348, 557. 154, 165, doi: 10.3847/1538-3881/aa88bd https://arxiv.org/abs/astro-ph/9906429 Bryan, M. L., Ginzburg, S., Chiang, E., et al. 2020a, ApJ, Bailey, V., Meshkat, T., Reiter, M., et al. 2014, ApJL, 780, 905, 37, doi: 10.3847/1538-4357/abc0ef L4, doi: 10.1088/2041-8205/780/1/L4 Bryan, M. L., Chiang, E., Bowler, B. P., et al. 2020b, AJ, Bate, M. R. 2009, MNRAS, 392, 590, 159, 181, doi: 10.3847/1538-3881/ab76c6 doi: 10.1111/j.1365-2966.2008.14106.x Burrows, A., Marley, M., Hubbard, W. B., et al. 1997, ApJ, —. 2018, MNRAS, 475, 5618, doi: 10.1093/mnras/sty169 491, 856, doi: 10.1086/305002 Bate, M. R., Bonnell, I. A., & Bromm, V. 2002, MNRAS, Clarke, C. J., & Pringle, J. E. 1991a, MNRAS, 249, 588, 332, L65, doi: 10.1046/j.1365-8711.2002.05539.x doi: 10.1093/mnras/249.4.588 HD106906b Obliquity 13

—. 1991b, MNRAS, 249, 584, doi: 10.1093/mnras/249.4.584 Masuda, K., & Winn, J. N. 2020, AJ, 159, 81, Correia, A. C. M. 2006, Earth and Planetary Science doi: 10.3847/1538-3881/ab65be Letters, 252, 398, doi: 10.1016/j.epsl.2006.10.007 Maury, A. J., Andr´e,P., Testi, L., et al. 2019, A&A, 621, Czekala, I., Chiang, E., Andrews, S. M., et al. 2019, ApJ, A76, doi: 10.1051/0004-6361/201833537 883, 22, doi: 10.3847/1538-4357/ab287b Metchev, S. A., Heinze, A., Apai, D., et al. 2015, ApJ, 799, Daemgen, S., Todorov, K., Quanz, S. P., et al. 2017, A&A, 154, doi: 10.1088/0004-637X/799/2/154 608, A71, doi: 10.1051/0004-6361/201731527 Millholland, S., & Batygin, K. 2019, ApJ, 876, 119, De Rosa, R. J., & Kalas, P. 2019, AJ, 157, 125, doi: 10.3847/1538-4357/ab19be doi: 10.3847/1538-3881/ab0109 Millholland, S., & Laughlin, G. 2019, Nature Astronomy, 3, Dobrovolskis, A. R. 1980, Icarus, 41, 18, 424, doi: 10.1038/s41550-019-0701-7 doi: 10.1016/0019-1035(80)90157-8 Moeckel, N., & Bally, J. 2006, ApJ, 653, 437, Ge, H., Zhang, X., Fletcher, L. N., et al. 2019, AJ, 157, 89, doi: 10.1086/508145 doi: 10.3847/1538-3881/aafba7 Morley, C. V., Fortney, J. J., Marley, M. S., et al. 2012, Jennings, M., & Chiang, E. 2021, ApJ ApJ, 756, 172, doi: 10.1088/0004-637X/756/2/172 Kalas, P. G., Rajan, A., Wang, J. J., et al. 2015, ApJ, 814, Nesvorn´y,D. 2018, ARA&A, 56, 137, 32, doi: 10.1088/0004-637X/814/1/32 doi: 10.1146/annurev-astro-081817-052028 Karalidi, T., Apai, D., Schneider, G., Hanson, J. R., & Nguyen, M. M., De Rosa, R. J., & Kalas, P. 2021, AJ, 161, Pasachoff, J. M. 2015, ApJ, 814, 65, 22, doi: 10.3847/1538-3881/abc012 doi: 10.1088/0004-637X/814/1/65 Offner, S. S. R., Dunham, M. M., Lee, K. I., Arce, H. G., & Kausch, W., Noll, S., Smette, A., et al. 2015, A&A, 576, Fielding, D. B. 2016, ApJL, 827, L11, A78, doi: 10.1051/0004-6361/201423909 doi: 10.3847/2041-8205/827/1/L11 Kenyon, S. J., & Bromley, B. C. 2004, Nature, 432, 598, Park, C., Jaffe, D. T., Yuk, I.-S., et al. 2014, in Proc. SPIE, doi: 10.1038/nature03136 Vol. 9147, Ground-based and Airborne Instrumentation Lagrange, A. M., Langlois, M., Gratton, R., et al. 2016, for Astronomy V, 91471D, doi: 10.1117/12.2056431 A&A, 586, L8, doi: 10.1051/0004-6361/201527264 Parker, R. J., & Quanz, S. P. 2012, MNRAS, 419, 2448, Laskar, J., & Robutel, P. 1993, Nature, 361, 608, doi: 10.1111/j.1365-2966.2011.19911.x doi: 10.1038/361608a0 Reinhardt, C., Chau, A., Stadel, J., & Helled, R. 2019, Laughlin, G., & Adams, F. C. 1998, ApJL, 508, L171, arXiv e-prints, arXiv:1907.09809. doi: 10.1086/311736 https://arxiv.org/abs/1907.09809 Lee, J.-J., & Gullikson, K. 2016, plp: v2.1 alpha 3. Saumon, D., & Marley, M. S. 2008, ApJ, 689, 1327, https://doi.org/10.5281/zenodo.56067 doi: 10.1086/592734 Li, G. 2021, arXiv e-prints, arXiv:2103.15843. Schlichting, H. E., & Sari, R. 2007, ApJ, 658, 593, https://arxiv.org/abs/2103.15843 doi: 10.1086/511129 Lissauer, J. J., & Kary, D. M. 1991, Icarus, 94, 126, doi: 10.1016/0019-1035(91)90145-J Simon, A. A., Rowe, J. F., Gaulme, P., et al. 2016, ApJ, Lubow, S. H., & Ogilvie, G. I. 2000, ApJ, 538, 326, 817, 162, doi: 10.3847/0004-637X/817/2/162 doi: 10.1086/309101 Smette, A., Sana, H., Noll, S., et al. 2015, A&A, 576, A77, Mace, G., Sokal, K., Lee, J.-J., et al. 2018, in Proc. SPIE, doi: 10.1051/0004-6361/201423932 Vol. 10702, Ground-based and Airborne Instrumentation Storch, N. I., Anderson, K. R., & Lai, D. 2014, Science, for Astronomy, 10702 – 10702 – 18, 345, 1317, doi: 10.1126/science.1254358 doi: 10.1117/12.2312345 Touma, J., & Wisdom, J. 1993, Science, 259, 1294, Marley, M. S., Gelino, C., Stephens, D., Lunine, J. I., & doi: 10.1126/science.259.5099.1294 Freedman, R. 1999, ApJ, 513, 879 Vos, J. M., Allers, K. N., & Biller, B. A. 2017, ApJ, 842, Marley, M. S., Saumon, D., Visscher, C., et al. 2021, arXiv 78, doi: 10.3847/1538-4357/aa73cf e-prints, arXiv:2107.07434. Vos, J. M., Biller, B. A., Bonavita, M., et al. 2019, https://arxiv.org/abs/2107.07434 MNRAS, 483, 480, doi: 10.1093/mnras/sty3123 Martin, R. G., & Armitage, P. J. 2021, arXiv e-prints, Ward, W. R., & Hamilton, D. P. 2004, AJ, 128, 2501, arXiv:2104.06479. https://arxiv.org/abs/2104.06479 doi: 10.1086/424533 Martin, R. G., Nixon, C., Lubow, S. H., et al. 2014, ApJL, Yang, H., Apai, D., Marley, M. S., et al. 2015, ApJL, 798, 792, L33, doi: 10.1088/2041-8205/792/2/L33 L13, doi: 10.1088/2041-8205/798/1/L13 14 Bryan et al.

Yuk, I.-S., Jaffe, D. T., Barnes, S., et al. 2010, in Zhou, Y., Apai, D., Schneider, G. H., Marley, M. S., & Proc. SPIE, Vol. 7735, Ground-based and Airborne Showman, A. P. 2016, ApJ, 818, 176, Instrumentation for Astronomy III, 77351M, doi: 10.3847/0004-637X/818/2/176 doi: 10.1117/12.856864 Zhou, Y., Apai, D., Lew, B. W. P., et al. 2019, AJ, 157, Zakamska, N. L., & Tremaine, S. 2004, AJ, 128, 869, 128, doi: 10.3847/1538-3881/ab037f doi: 10.1086/422023 Zhou, Y., Apai, D., Bedin, L. R., et al. 2020, AJ, 159, 140, Zhang, X., & Showman, A. P. 2014, ApJL, 788, L6, doi: 10.3847/1538-3881/ab6f65 doi: 10.1088/2041-8205/788/1/L6