GW Orionis: A pre-main-sequence triple with a warped disk and a torn-apart ring as benchmark for disk hydrodynamics
aa r X i v : . [ a s t r o - ph . S R ] O c t GW Orionis:A pre-main-sequence triple with awarped disk and a torn-apart ring asbenchmark for disk hydrodynamics
Stefan Kraus, University of Exeter(email: [email protected])
Pre-main-sequence multiples as benchmarkfor disk hydrodynamics
Understanding how bodies interact with each other andwith disk material holds the key to understanding thearchitecture of stellar systems and of planetary systems.While the interactions between point sources can be de-scribed by simple gravity, interactions with disk materialrequire further knowledge about viscosity and gas+dustmicrophysics that need to be included when simulatingdisk-body interactions. As a result of our limited knowl-edge in these areas it is, for instance, still difficult to esti-mate the mass of a gap-opening planet from the morphol-ogy of a gap observed in a protoplanetary disk, or, to de-rive with certainty whether a gap is opened by a planet in-stead of by other processes. Furthermore, numerical sim-ulations continue to unveil new dynamical processes thatmight shape protoplanetary disk structures and affect theplanet populations forming from these disks. One exampleis disk tearing that might occur in the disks around mul-tiple stars whose orbital angular momentum vectors aremisaligned with respect to the rotation axis of the disk.Based on computer simulations it has been proposed in2012 that the resulting gravitational torques could tearthe disk apart and cause rings to separate from the diskand to precess independently around the central objects(Nixon et al. 2012, 2013, Facchini et al. 2013, 2018). Inorder to test such theories and to calibrate the fundamen-tal parameters involved in hydrodynamic simulations, pre- main-sequence (PMS) multiple systems provide us with aunique laboratory (for general reviews on PMS binariesand multiples see for instance Dˆuchene & Kraus 2013 andReipurth et al. 2014). For these systems we are able todirectly measure the 3-dimensional orbits and dynamicalmasses of the perturbing bodies and can image how thedisk responds to the perturbation.One system that has the potential to serve as such a“rosetta stone” for hydrodynamic studies, is the PMS tripleGW Orionis. This system is one of the brightest and best-studied T Tauri multiple systems. It is located in the λ Orionis star-forming region (388 pc; Kounkel et al. 2017)and has an age of ∼ ∼ The triple star system
GW Ori is long known to be a single-lined spectroscopicbinary with a 242 day period (Mathieu et al. 1991). Pratoet al. (2018) reported the detection of lines associated withthe secondary. Observations with the IOTA infrared inter-ferometer resolved the inner binary and discovered a thirdcomponent (Berger et al. 2011). Building on the workby Mathieu and coworkers, Czekala et al. presented in2017 an impressive set of spectra that were obtained over35 years with the Fred L. Whipple Observatory and OakRidge Observatory and that provide radial velocities forthe primary and secondary. Using these radial velocitiesand the IOTA astrometry, Czekala and colleagues derivedfirst orbit solutions for all 3 stars. These orbit solutions in-dicated a 11.5 year orbit period for the tertiary and hintedat a signficant misalignment between the stellar orbits andthe disk, although the small orbital arc covered by theIOTA astrometry resulted in degeneracies in the orbit fits.Between 2008 and 2019, the VLT Interferometer and theCHARA array were used to monitor the astrometric orbitof the inner binary and the tertiary (Kraus et al. 2020a).The resulting orbits are shown in Figure 1 and correspondto a near-circular ( e = 0 . ± . . ± .
05 day or-bit for the inner binary and a 4216 . ± . e = 0 . ± . . ± . ◦ .The precise masses of the components in the GW Ori sys-tem has long been a matter of debate. Mathieu et al.(1995) estimated the mass using evolutionary tracks, yield-ing 2.5 M ⊙ and 0.5 M ⊙ for the primary and secondary,respectively. Other workers estimated the mass from theH-band flux ratio and derived more equal mass ratios (e.g.1 A d D E C [ m a s ] dRA [mas]GW Ori B A B2015.0 -30-20-100102030 B -20-1001020 C R V A , A - B [ k m / s ] -20-1001020 D R V A , AB - C [ k m / s ] -30-20-100102030 0 0.2 0.4 0.6 0.8 1 1.2 1.4 E R V B , A - B [ k m / s ] PhaseA-B 0 0.2 0.4 0.6 0.8 1 1.2 1.4 -30-20-100102030 F R V B , AB - C [ k m / s ] PhaseAB-C
Figure 1:
Top:
GW Ori orbit fit to the astrometry of theinner binary (A-B; panel A) and the tertiary (AB-C; panelB), where the primary is in the origin of the coordinate sys-tem (Kraus et al. 2020a). Panels C-F show the fit to theradial velocity data presented by Czekala et al. 2017.
Left:
Aperture-synthesis image of GW Ori, obtained withthe MIRC-X instrument and CHARA on 2019 August 27.The white dot marks the location of the center-of-mass ofthe system, while the blue/brown/white curves give thebest-fit orbits derived for the component A/B/C. FromKraus et al. (2020).3.2 M ⊙ and 2.7 M ⊙ ; Prato et al. 2018). These discrepan-cies can likely be explained by the non-negligle & time-variable contributions from circumbinary/circumtertiarydust emission biasing the near-infrared flux ratio (Krauset al. 2020a). The dynamical masses resulting from theorbit solution are M A = 2 . ± .
33, M B = 1 . ± . C = 1 . ± . M ⊙ (Kraus et al. 2020a). A highly dynamical environment in the in-ner few au
The spectral energy distribution (SED) of GW Ori fea-tures strong excess emission from mid-infrared to millime-ter wavelengths, indicating the presence of circumstellardust. SED modeling suggested the presence of a circumbi-nary disk extending from around 1 . ±
2% of the H-band flux with circumbinarymaterial located 2 +2 . − . au from the inner binary (Kraus etal. 2020a). To fit the mid-infrared SED, Fang et al. (2014, 2017) de-rived a dust-depleted gap at ∼
45 au. Further evidencefor a truncated or gapped disk structure comes from theline profile of the CO fundamental lines, where Najitaet al. (2003) noted that the line profile exhibits a nar-row+broad emission component and that the line widthincreases towards the more energetic transitions. The sys-tem also shows signposts of active accretion, in particularBr γ -emission (Folha & Emerson 2001) and strong UV veil-ing ( ˙ M acc = 3 × − M ⊙ yr − , Calvet et al. 2004).There has been a long debate on the viewing geometryof the system: Based on measurements of the stellar rota-tion period ( P = 3 .
25 days), the rotation velocity ( v sin i =43 . ± . − ) and estimates of the stellar radius, Bou-vier & Bertout 1989 estimated the inclination of the sys-tem to ∼ ◦ , i.e. close to face-on. Using a similar method,but other observational data, Mathieu et al. 1995 obtainedan inclination of 30 ◦ . However, photometric observationsalso reported Algol-like eclipses (e.g. Shevchenko et al.1992, Lamzin et al. 1998, Czekala et al. 2017) that havebeen interpreted as evidence for a nearly edge-on disk ori-entation.Variability on longer time scales has been reported as well,2 N Thermal!dust!emission!1.3mm Thermal!dust!emission!1.3mm
R2 R3 R3 asym D AB D C R1 A B
C2 C1 Bi ! et ! al. ! ! mom. ! C D CO ! moment ! ! + ! continuum ! (contours) CO ! moment ! ! + ! continuum ! (contours) Figure 2: ALMA observations of GW Ori at 0.12” (right, Bi et al. 2020a) and 0.02” resolution (left, panels A-D,Kraus et al. 2020a). The top row shows continuum images, while the bottom row shows CO moment maps.including dramatic changes in the near-infrared SED ontimescales of ∼
20 yrs (Fang et al. 2014). This variabil-ity might be linked with a 0.2 mag-amplitude sinusoidalvariation in the V-band light curve (Czekala et al. 2017)that is phased with the orbital period of the tertiary. Theorigin of the long-term variability has not been answeredconclusively yet, but might be due to changes in the view-ing geometry or accretion rate on the circumtertiary disk.The circumtertiary disk has been detected as submillime-ter emission in the ALMA 0.02” images (Fig. 2B) and asnear-infrared excess emission near the location of the ter-tiary in infrared interferometry data (Kraus et al. 2020a).
Misaligned rings
The intermediate/outer dust disk has been probed withJCMT (Mathieu et al. 1995) and SMA millimeter interfer-ometry, where Fang et al. 2017 highlighted the large spa-tial extent ( ∼
400 au) and high mass (0.12 M ⊙ ) of the disk.Bi et al. 2020 and Kraus et al. 2020a acquired ALMA datawith different baseline configurations and detected threedust rings. The two outer rings, R1 and R2, have radiiof about 350 and 180 au and are oriented North-South and seen under intermediate inclination 37 ± ◦ (Fig. 2A),representing the angular moment vector of the cloud thatfeeds the disk. The Eastern side of the disk is tilted to-wards us, as indicated by the strong forward-scatteredlight from that side of the disk (Fig. 3).The inner submillimeter ring, R3, appears much more cir-cular in the images – which could be interpreted as a moreface-on viewing geometry. However, from extrapolatingthe center of the outer rings (Bi et al. 2020) and from directimaging (Kraus et al. 2020a) it is evident that the innerring is not centered on the position of the stars (Fig. 2B).This can be explained best if the ring is intrinsically ec-centric but seen under significant inclination and appearsnear-circular in projection. The following additional in-formation can be used to derive the 3-dimensional shapeand orientation of the ring:(a) Gas kinematics : CO moment 1 maps show thatthe rotation axis of the outer disk is oriented roughlyin East-West direction (position angle 90 ◦ ; e.g. Fanget al. 2017, Czekala et al. 2017) with a ’twist’ in thevelocity field in the inner 0.2” (Bi et al. 2020; Fig. 2,right). The twist might follow a spiral-arm pattern,with the position angle of the rotation axis changing3igure 3: Top-left: Overlay of the ALMA continuum image (blue) and the SPHERE scattered light image (red; credit:ESO/Exeter/Kraus et al.). Top-right: SPHERE H-band polarimetric image and model image. Bottom: Sketch of the3-dimensional disk geometry of GW Ori. From Kraus et al. 2020a.to 180 ◦ at 100 au and ∼ ◦ at ∼
30 au (Kraus etal. 2020a, Fig. 2D).(b)
Warm gas at the inner surface of the ring :The CO moment 0 map shows that the CO sur-face brightness is low at the location of the ring R3,which can be explained with the high optical depthand low gas temperature within the ring. However,there is strong CO emission near the inner edge ofring R3 on the Eastern side (labeled C1 in Fig. 2C),indicating that the Eastern side of the ring is fartheraway from the observer and that we see warm gasat the illuminated inner surface of the ring (Fig. 3,bottom-right; Kraus et al. 2020a).(c)
Shadows cast by the ring:
SPHERE scatteredlight imagery (Figs. 3 and 4, top) exhibits severalshadows, including narrow shadows in south-east andnorth-west direction (S1 and S2; Fig. 3, top) andbroader shadows extending in north and south-westdirection (S3 and S4). Remarkably, shadow S1 does not follow a straight line but changes direction at ∼
100 au separation (Fig. 3, top-right), indicatingthat the shadow falls onto a curved surface in theinner 100 au. Simultaneous modeling of the shadowmorphology and of the ALMA continuum geometryyields an eccentricity e = 0 . ◦ (Fig. 3; Kraus et al. 2020a). Broken & warped disk geometry
Shadows have been observed in several protoplanetary disks,but GW Ori is (to my knowledge) the first case where thering casting the shadow has been spatially resolved. Thisenables tight constraints on the shape and 3-dimensionalorientation of the misaligned ring as well as the curva-ture of the warped disk surface inside of the middle ringR2 ( r .
182 au). The scattered-light morphology showsa strong East-West asymmetry, where the bright Easternarc A3 and the fainter Western arc A4 form together an4 rtist!impression(adopting!derived!3D!geometry) SPHERE!(H−band)
Figure 4: Artist impression of the 3-dimensional geometry of the GW Ori disk (left) and comparison with the SPHEREimage (right). Credit: ESO/Cal¸cada, Exeter/Kraus et al.apparent ellipse with semi-major axis of 90 au and higheccentricity ( e = 0 .
65; Fig. 4). Kraus et al. (2020a) iden-tifies this ellipse as the point where the disk breaks dueto the gravitational torque from the central triple system.The bright arc A3 constitutes the side of the warped disksurface that is facing away from Earth and that appearsbright in scattered light due to the direct illumination fromthe stars. Arc A4 corresponds to the side facing towardsus, where we see only the self-shadowed outer side of thewarped surface (Fig. 3). The shadows from the misalignedring are cast onto this warped surface and appear as shad-ows S1 and S2, while the broad shadows S3 and S4 cor-respond to the regions where the break orbit crosses theplane of the outer disk, which coincides with the directionin the warp with the highest radial column density.
Origin of the disk misalignments
To determine the origin of the extreme disk misalignmentsobserved in GW Ori, two teams recently presented smoothedparticle hydrodynamic simulations. Bi et al. conductedSPH simulations using the ’phantom’ code and concludedthat the gravitational torque of the stars alone cannot ex-plain the observed large misalignment between the dustring. Instead, they propose that an undiscovered compan-ion located between the inner and middle ring that mighthave broken the disk and induced the misalignments. The ’sphNG’ simulation presented in Kraus et al. (2020a),on the other hand, shows the disk tearing effect, where thegravitational torque of the three stars tears the disk apartinto distinct rings that precess independently around thecentral objects. After letting the dust distribution evolvefor a few thousand years, a ring breaks out of the diskplane, whose radius ( ∼
40 au), eccentricity ( e ∼ . α SS = 0 . − .
013 for Bi et al.; 0.01-0.02 for Krauset al.). Therefore, it appears more likely that the differentoutcomes might be related to the setup of the simulation.There are differences concerning the number of stars in-cluded in the simulation (2 stars in Bi et al. simulation;3 stars in Kraus et al. simulation) and the orbit solutionthat is adopted for the simulation (Czekala et al. 2017solution and Kraus et al. 2020a solution, respectively).5
RA z column!density!/!g!cm −2 RA Dec log!(density!/!g!cm ) −3 −17log!(column!density!/!g!cm ) −2 −1 −15 −13 −11 Warp
20 6
RA z100!au
B C
Figure 5: SPH model, computed with the sphNG code developed by Matthew Bate and collaborators. The simulationadopts the triple star orbits shown in Fig. 1 and an initial disk orientation that corresponds to the outer ALMA ringsR1+R2. The snapshot shows the gas density after 9500 years. Panel (A) shows the column density along the line-of-sight seen from Earth; in (B) and (C) the z-axis indicates the direction towards the observer.
Outlook
Over the last few decades several exciting pre-main-sequencemultiple systems have been found and extensively stud-ied, including GG Tau, HD142527, HD98800, and T Tauri.GW Ori stands out with respect to the tight constraintson the full 3-dimensional orbits for all components in thesystem, the dynamical masses, and our knowledge on the3-dimensional geometry of the strongly distorted disk (fora visualization of the deduced disk geometry & orbits,see the interactive 3-dimensional model in Fig. 6). Dueto this unique information, the system could serve as avaluable benchmark for calibrating hydrodynamic mod-els and fundamental parameters under well-defined condi-tions. This could provide the validation & refinement thatis needed before applying the models to the much less-well-constrained planet formation case, where the masses andorbits of the gap-opening bodies are not known in general.The disk-tearing effect that we might witness in GW Ori inaction , constitutes an important new mechanism for mov-ing disk material onto highly oblique or retrograde orbits,even at very wide separations from the star. At the sametime, the observed torn ring seems to be sufficiently mas-sive and might be sufficiently stable for planet formationto occur, potentially giving rise to an yet-undiscoveredpopulation of circum-multiple planets on highly oblique,long-period orbits.An important open question concerns the origin of the out-ermost ring in GW Ori. The high submillimeter brightnessof the inner and middle ring can likely be explained by disk tearing and dust filtration processes near the disk warp re-gion. It is unclear whether the outer-most ring can also beexplained by the dynamical interplay between the centraltriple system and the disk, or whether dust trapping neara planet-induced density gap might be required to explainthe high submillimeter surface brightness in this region.But what caused the misalignment between the disk andthe orbits in the first place? Possibilities include turbulentdisk fragmentation (Offner et al. 2010), perturbation byother stars in a stellar cluster (Clarke & Pringle 1993), thecapture of disk material during a stellar flyby (Clarke &Pringle 1991), or the infall of material with a different an-gular momentum vector from that of the gas that formedthe stars initially (Bate et al. 2010, Bate 2018). To answerthis question, statistical information will be of essence,both on the disk-orbit misalignment in pre-main-sequencemultiples and on the orbital architecture and spin-orbitalignment in main-sequence systems. Obtaining such con-straints on a large sample of stars is a key science objec-tive for the proposed VLTI instrument BIFROST (Krauset al. 2019, 2020b) and could offer important new insightson both the star- and planet-formation processes.
Acknowledgement:
I would like to thank my collaborators on our recent paper on GW Oriand the members of the team around Jiaqing Bi for delightful dis-cussions on this fascinating object. Also I acknowledge support froman ERC Starting Grant (Grant Agreement No. 639889).
References:
Artymowicz & Ludow 1994, ApJ 421, 651Bate, M.R., et al. 2010, MNRAS 401, 1505 ate, M.R., 2018, MNRAS 475, 5618Berger, J.-P., et al. 2011, A&A 529, L1Bi, J., et al. 2020, ApJL 895, L18Bouvier, J. & Bertout, C. 1989, A&A 211, 99Calvet, N., et al. 2004, AJ 128, 1294Clarke, C.J., & Pringle, J.E. 1991, MNRAS 249, 588Clarke, C.J., & Pringle, J.E. 1993, MNRAS 261, 190Czekala, I., et al. 2017, ApJ 851, 132Dˆuchene, G. & Kraus, A. 2013, ARAA, 51, 269Facchini, S., et al. 2013, MNRAS 433, 2142Facchini, S., et al. 2018, MNRAS 473, 4459Fang, M., et al. 2017, A&A 603, 132Fang, M., et al. 2014, A&A 570, 118Folha, D.F.M. & Emerson, J.P., 2001, A&A 365, 90Kraus, S., et al. 2019, zenodo.3356286Kraus, S., et al. 2020a, Science 369, 6508Kraus, S., et al. 2020b, ApJ 897, 8Kounkel, M., et al. 2017, AJ 834, 142Lamzin, S., et al. 1998, Ap&SS, 261, 167Mathieu, R.D., et al. 1991, AJ, 101, 2184Mathieu, R.D., et al. 1995, AJ, 109, 2655Najita, J., et al. 2003, ApJ 589, 931Nixon, C., et al. 2012, ApJ 757, L24Nixon, C., et al. 2013, MNRAS 434, 1946Offner, S., et al. 2010, ApJ 725, 1485Prato, L., et al. 2018, ApJ 852, 38Reipurth, B. et al. 2014, Protostars and Planets V, p. 267Shevchenko, V.S., et al. 1992, IBVS 3746, 1ate, M.R., 2018, MNRAS 475, 5618Berger, J.-P., et al. 2011, A&A 529, L1Bi, J., et al. 2020, ApJL 895, L18Bouvier, J. & Bertout, C. 1989, A&A 211, 99Calvet, N., et al. 2004, AJ 128, 1294Clarke, C.J., & Pringle, J.E. 1991, MNRAS 249, 588Clarke, C.J., & Pringle, J.E. 1993, MNRAS 261, 190Czekala, I., et al. 2017, ApJ 851, 132Dˆuchene, G. & Kraus, A. 2013, ARAA, 51, 269Facchini, S., et al. 2013, MNRAS 433, 2142Facchini, S., et al. 2018, MNRAS 473, 4459Fang, M., et al. 2017, A&A 603, 132Fang, M., et al. 2014, A&A 570, 118Folha, D.F.M. & Emerson, J.P., 2001, A&A 365, 90Kraus, S., et al. 2019, zenodo.3356286Kraus, S., et al. 2020a, Science 369, 6508Kraus, S., et al. 2020b, ApJ 897, 8Kounkel, M., et al. 2017, AJ 834, 142Lamzin, S., et al. 1998, Ap&SS, 261, 167Mathieu, R.D., et al. 1991, AJ, 101, 2184Mathieu, R.D., et al. 1995, AJ, 109, 2655Najita, J., et al. 2003, ApJ 589, 931Nixon, C., et al. 2012, ApJ 757, L24Nixon, C., et al. 2013, MNRAS 434, 1946Offner, S., et al. 2010, ApJ 725, 1485Prato, L., et al. 2018, ApJ 852, 38Reipurth, B. et al. 2014, Protostars and Planets V, p. 267Shevchenko, V.S., et al. 1992, IBVS 3746, 1