On the alignment between the circumstellar disks and orbital planes of Herbig Ae/Be binary systems
aa r X i v : . [ a s t r o - ph . S R ] J un Astronomy&Astrophysicsmanuscript no. 16996 c (cid:13)
ESO 2017December 12, 2017
On the alignment between the circumstellar disks and orbitalplanes of Herbig Ae/Be binary systems
H.E. Wheelwright , J.S. Vink , R.D. Oudmaijer , and J.E. Drew School of Physics and Astronomy, University of Leeds, Leeds LS2 9JT, UKe-mail:
[email protected] Armagh Observatory, College Hill, Armagh BT61 9DG, Northern Ireland Centre for Astrophysics Research, STRI, University of Hertfordshire, College Lane Campus, Hatfield AL10 9AB, UKReceived March 31, 2011; accepted June 09, 2011
ABSTRACT
Context.
The majority of the intermediate mass, pre-main-sequence Herbig Ae / Be stars reside in binary systems. As these systemsare young, their properties may contain an imprint of the star formation process at intermediate masses (2-15 M ⊙ ). However, thesesystems are generally spatially unresolved, making it di ffi cult to probe their circumstellar environment to search for manifestations oftheir formation process, such as accretion disks. Aims.
Here we investigate the formation mechanism of Herbig Ae / Be (HAe / Be) binary systems by studying the relative orientationof their binary orbits and circumstellar disks.
Methods.
We present linear spectropolarimetric observations of HAe / Be stars over the H α line, which are used to determine theorientation of their circumstellar disks. In conjunction with data from the literature, we obtain a sample of 20 binaries with knowndisk position angles (PAs). We subsequently compare our disk PA data to a model to investigate whether HAe / Be binary systems andtheir disks are co-planar. Moreover, in the light of a relatively recent suggestion that some HAe / Be star spectropolarimetric signaturesmay not necessarily be related to circumstellar disks, we re-assess the relationship between spectropolarimetric signatures and diskPAs. We do this by comparing spectropolarimetric and high spatial resolution observations of young stellar objects (both HAe / Be andT Tauri stars).
Results.
We find that spectropolarimetric observations of pre-main-sequence stars do indeed trace circumstellar disks. This findingis significant above the 3 σ level. In addition, our data are entirely consistent with the situation in which HAe / Be binary systems andcircumstellar disks are co-planar, while random orientations can be rejected at the 2.2 σ level. Conclusions.
The conclusive alignment (at more than 3 σ ) between the disk PAs derived from linear spectropolarimetry and highspatial resolution observations indicates that linear spectropolarimetry traces disks. This in turn allows us to conclude that the orbitalplanes of HAe / Be binary systems and the disks around the primaries are likely to be co-planar, which is consistent with the notionthat these systems form via monolithic collapse and subsequent disk fragmentation.
Key words.
Techniques: polarimetric – Stars: emission-line – formation – binaries – pre-main-sequence – variables: T Tauri, HerbigAe / Be
1. Introduction
Herbig Ae / Be stars are intermediate mass, 2 − M ⊙ , pre-main-sequence stars (see e.g. Herbig 1960; Hillenbrand et al. 1992;Waters & Waelkens 1998; Hern´andez et al. 2004). They are themost massive objects to experience an optically visible pre-main-sequence (PMS) phase. Therefore, they provide a key op-portunity to study the early evolution of stars more massive thanthe lower mass T Tauri stars at these wavelengths. Herbig Ae / Be(HAe / Be) stars span the transition between low and high massobjects. As a result, it has been postulated that there is a changein accretion mechanism within the HAe / Be mass range (see e.g.Vink et al. 2002; Eisner et al. 2004; Monnier 2005). If this is thecase, the more massive Herbig Ae / Be stars may form in the samefashion as the most massive stars. Consequently, the study ofHerbig Ae / Be stars may be used to constrain the process of mas-sive star formation, which is still not fully understood (see e.g.Zinnecker & Yorke 2007).Baines et al. (2006) used spectroastrometry to detect unre-solved HAe / Be binary systems and they compared the posi-tion angle (PA) of the systems to the orientation of their cir-cumprimary disks. They reported that the binary and circumpri- mary disk PAs of the systems in their sample are preferentiallyaligned. This indicates that the circumstellar disks lie in the sameplane as the binary orbit. In turn, this suggests that these systemsformed via a scenario that features fragmentation, as opposed toalternatives involving the capture of binary companions.This is consistent with one model of massive star formation.Using 3D radiation-hydrodynamic simulations, Krumholz et al.(2009) show that stars of even 40 M ⊙ can form via monolithiccollapse and disk accretion. A prediction of these models is thatdisk fragmentation leads to binary systems. The resultant bi-nary system is aligned with the original fragmented disk struc-ture. Therefore, because of angular momentum considerations,the binary system is also aligned with the inner accretion diskthat reaches onto the primary star’s surface. In addition, the sec-ondary, at a distance of approximately 1000 AU, has a relativelyhigh mass. It has also been proposed that massive stars formvia competitive accretion. In this scenario, a star gains mate-rial from its parental cluster. In the process, binary systems areformed via stellar capture. These binaries start as solar mass bi-naries at separations of the order of 1000 AU and evolve intoclose high mass binaries at approximately 1 AU (Bonnell & Bate / Be binaries and disks / Be binary systems. They found thatthe mass-ratio distribution of HAe / Be binary systems favourscomparable masses, further indicating that these systems formedvia disk fragmentation. However, the sample of Baines et al.(2006) which initially suggested this was small (6 objects), whilea comparison with model simulations of co-planar binary or-bits and disks was lacking. Therefore, the relative alignment ofHAe / Be binary orbits and circumstellar disks is still uncertain.We aim to extend the work of Baines et al. (2006) and addressthis uncertainty by enlarging the sample of binary HAe / Be starswith known disk PAs.To achieve this, we require a large sample of HAe / Be bi-nary systems with known disk orientations. We generate a sam-ple of 20 such systems using linear spectropolarimetry and highspatial resolution data from the literature. Spectropolarimetryo ff ers a unique opportunity to probe the circumstellar envi-ronment of young stellar objects on scales small enough tostudy accretion disks. If a star is surrounded by a hot disk,free electrons in the disk can polarise the light of the cen-tral star. Emission-line photons emanating from the disk do notpass through the polarising medium, unlike the continuum light.Consequently, emission lines are depolarised with respect to thecontinuum. Therefore, a depolarisation signature over an emis-sion line can be used to infer the presence of a small-scale,otherwise undetected disk (Clarke & McLean 1974; Poeckert1975; Poeckert & Marlborough 1976). Alternatively, if accretionshocks produce emission lines close to the star, the presenceof a disk may be manifest by scattering polarisation of a com-pact source of line emission (Vink et al. 2002, hereafter V2002).Furthermore, spectropolarimetric signatures can also be used toconstrain the geometry of such disks (e.g. Vink et al. 2005b).Linear spectropolarimetry is well established as a tech-nique to study spatially unresolved disks around HAe / Be stars(see e.g. Oudmaijer & Drew 1999; Vink et al. 2002, 2005a;Mottram et al. 2007). However, the use of spectropolarimetry tostudy disks has recently been revisited by Harrington & Kuhn(2007). The authors report observations of spectropolarimetricsignatures across absorption components of H α emission lines.It is suggested such signatures are due to a process other thanscattering in a disk. Instead, Kuhn et al. (2007) propose thatoptical pumping and absorption in winds may be responsible.Optical pumping, i.e. a preferential population of di ff erent mag-netic sub-states of the lower level of an atomic transition, maybe caused by an anisotropic radiation field. If a gas is opticallypumped, its opacity will depend upon the orientation of the elec-tric field of incident radiation. As a result, the absorption of lightin such gas can result in linear polarisation e ff ects across ab-sorption features (see Kuhn et al. 2007, 2010). We note that allspectropolarimetric signatures require a flattened, asymmetricgeometry otherwise the polarisation vectors cancel. This is in-dependent of the polarising mechanism. Nonetheless, if the po-larisation signatures of HAe / Be stars originate in significantlyasymmetric winds, this does not automatically demand an in-ward coplanar extension of any large-scale circumstellar disk. Here we investigate whether there is a clear relationship betweenthe angle of polarisation and disk orientation which, at minimumsignals alignment between the unresolved and resolved spatialscales. Such an outcome is required if indeed the bulk of theobserved linear polarisation of these objects is due to a circum-stellar disk.This paper is structured as follows. In Section 2 we evaluatethe alignment of HAe / Be binary systems and circumstellar disksusing linear spectropolarimetry over H α to determine the PAsof disks in HAe / Be binary systems. To justify the use of spec-tropolarimetry to trace circumstellar disks, we reassess the rela-tionship between spectropolarimetric signatures and disk PAs inSection 3. Finally, we discuss the results and conclude the paperin Section 4.
2. The relative orientation of HAe/Be binarysystems and circumstellar disks.
The key to determining whether HAe / Be binary planes and cir-cumstellar disks are preferentially aligned is to use a large sam-ple of binary systems with known disk PAs. There are severalstudies of HAe / Be star binary systems (see e.g. Baines et al.2006; Wheelwright et al. 2010). However, there are few mea-surements of the orientation of the circumstellar disks in thesesystems. Linear spectropolarimetry is the favoured technique toprobe the orientation of HAe / Be star disks since a sample of ∼ α spectropolarimetric observations; taken linear spectropolari-metric results from previous work; and assimilated additionalconstraints on HAe / Be star disk orientations from the literature.We first present the additional spectropolarimetric observationswe undertook (Section 2.1). We then use the assembled sampleof disk PAs, in conjunction with a simple model (described inSection 2.3), to evaluate the hypothesis that circumstellar disksin HAe / Be binary systems lie in the orbital plane (Section 2.4).
We present linear spectropolarimetric observations of HAe / Bestars conducted in the R band and centred on H α . Here we de-scribe the observing procedure and the data reduction steps be-fore presenting the observed spectropolarimetric signatures.The sample was predominately chosen from the catalogueof Th´e et al. (1994). Targets were selected based on their visualmagnitude ( V ≤ / Be stars that had not been observed with spectropolarimetrypreviously, and these were primarily drawn from Baines et al.(2006) and Thomas et al. (2007). Several objects with existingspectropolarimetric data were observed to check the results areconsistent with previous observations.The linear spectropolarimetric data were obtained using theWilliam Herschel Telescope (WHT) from 08-11-2008 to 10-11-2008. Clouds were present for the majority of the three nights,preventing observation for some, but not all, of the time. Theseeing was typically fair ( ∼ ′′ ), although on occasions it becamepoor (2–2.5 ′′ ). The observations were conducted with the ISISspectrograph which was equipped with polarising optics com-prising of a calcite block and a rotating half-wave plate. TheR1200R grating was used and the central wavelength was setto 6560 Å. Several slit widths were used, ranging from 1 to 1.8 / Be binaries and disks arcsec, and the minimum spectral resolution was found to be ∼ − .The calcite block separated the incident light into two per-pendicular rays: the ordinary (O) and the extraordinary (E) rays.Each observation comprised of both the O and the E ray spec-trum of the science target, and a corresponding set of sky spec-tra. The polarisation at PAs of 0 ◦ , . ◦ , ◦ and 67 . ◦ was mea-sured by rotating the half-wave plate. Multiple polarisation stan-dard stars were observed to characterise the instrumental polari-sation and calibrate the polarisation angle. A log of the observa-tions is presented in Table 1.Data reduction was conducted using the Image Reductionand Analysis Facility ( iraf ) , in conjunction with routines writ-ten in Interactive Data Language ( idl ). The data reduction pro-cess for each observation consisted of trimming, bias subtrac-tion, flat-field division and cosmic-ray removal. Following theabove, the target O and E spectra, and those of the sky if theywere present, were extracted from each frame. Sky spectra,which were not always detected, were typically a few percentof the stellar spectra. We note that polarisation signatures overH α , which are the focus of the paper, are una ff ected by con-taminant sky polarisation. Therefore, the sky polarisation has noinfluence on our final results. Wavelength calibration was per-formed using CuNe and CuAr arc spectra, which were obtainedperiodically during the observing run.Once the O and E ray spectra had been extracted, the Stokesparameters for each data set were calculated using a routine writ-ten in idl . The method used is that outlined in the ISIS polarisa-tion manual by Jaap Tinbergen and Ren´e Rutten . For each setof polarisation data, i.e. data obtained with the half-wave plateat 0 ◦ and 45 ◦ or 22 . ◦ and 67 . ◦ , the ratio of the O and E raysin each spectrum was calculated. To obtain the degree of polar-isation, the data obtained at a given PA were averaged and thefollowing equations were used: R = I O , ◦ / I E , ◦ I O , ◦ / I E , ◦ (1) q = R − R + q = Q / I , I is the total flux input and I O , ang and I E , ang are the fluxes of the O and E rays at the stated half-waveplate PAs. This procedure was repeated for the other set of po-larisation data, i.e. data obtained with half-wave-plate PAs of22 . ◦ and 67 . ◦ , to calculate u .To calculate the total polarisation and the polarisation anglethe data were combined using the following equations: P = q q + u (3) θ =
12 tan − uq ! (4)where P represents the total polarisation, and θ is the polari-sation angle.Instrumental polarisation was not corrected for. The standardobservations indicate that the instrumental polarisation is of the http: // iraf.noao.edu / , see Tody (1993) http: // / Astronomy / observing / manuals / html manuals / wht instr / isispol / isispol.html order 0.1 per cent, and is not the dominant source of continuumpolarisation. Interstellar polarisation is not corrected for either,as such corrections are typically subject to significant uncertain-ties (see e.g. Jensen et al. 2004). Contaminant polarisation sim-ply adds a wavelength independent vector to the Stokes Q and U parameters. Therefore, plotting the spectrally dispersed q against u allows the intrinsic angle of polarisation, and hence the polar-ising media, to be established. We present an example of the spectropolarimetric signatures ob-served in Figure 1. Data around H α for all the stars in the sam-ple are presented in Appendix A in Figure A.1, while continuumpolarisation measurements are included in Table 1. The signa-tures of the objects previously observed with spectropolarimetryare generally consistent with published results (e.g. Vink et al.2002, 2005a; Mottram et al. 2007). This provides an importantcheck on the data reduction process. In general, the data areof slightly inferior quality to previous observations. This is at-tributed to the poor weather conditions throughout the observ-ing. Consequently a coarser binning is used than is typical forsuch data. Five objects exhibit a change in both linear polar-isation and the polarisation angle over H α (HD 179218, HKOri, MWC 1080, V586 Ori & MWC 147). MWC 1080 andMWC 147 were also observed by Mottram et al. (2007, hereafterM2007). While the data are broadly consistent with the previousobservations, the line e ff ects are not obvious in the QU diagram.This is probably a result of the coarse binning. Less coarse bin-ning does not reveal any signatures as the scatter increases con-siderably. Therefore, the results of M2007 are used rather thanthese new data. Of the three remaining objects that exhibit linee ff ects, HK Ori and V586 Ori exhibit an excursion in QU spaceand HD 179218 exhibits a clump of data surrounding the contin-uum value, with a slight extension along the U axis.To arrive at the disk position angle from these spectropo-larimetric signatures, one should have some knowledge of thepolarising mechanism. In general, if the polarimetric signatureis the result of simple depolarisation, the polarisation vector in QU space should be measured from the line to the continuum. Incase of intrinsic line polarisation, the reverse is true, e ff ectivelyresulting in a di ff erence of 90 ◦ from the former mechanism. Todi ff erentiate between the two, the width of the line-e ff ect is oftenused as a proxy (see V2002). The intrinsic polarisation angles ofHK Ori and V586 Ori are determined via linear fits to their QU excursions, assuming the signature is due to depolarisation (asthe width of the spectropolarimetric signatures is comparable tothe emission line width, see Vink et al. 2002). The polarisationangle of HD 179218 is calculated assuming the excursion is onlyin the U direction and that the signature is due to intrinsic polar-isation (since the signature is narrower than the emission line).The uncertainties in the polarisation angles calculated are ap-proximately 10 ◦ .The continuum polarisation presented in Table 1 is broadlyconsistent with literature values (see e.g. Maheswar et al. 2002).However, there are exceptions. For example, we note that thecontinuum polarisation of HK Ori and MWC 758 di ff ers fromliterature values. This may be due to intrinsic variability asboth objects are known to exhibit variable polarisation (seeBaines et al. 2004; Beskrovnaya et al. 1999). / Be binaries and disks
Table 1.
The log of observations. The continuum polarisation was measured in the wavelength region 6520–6600 Å, excludingchanges over the H α line. The uncertainty in the continuum polarisation is typically 0.1 per cent and the uncertainty in the continuumpolarisation angle is of the order of 1 ◦ . The polarisation angles are in the equatorial frame. Object Alt. Name RA Dec Spec. Type V T exp Slit Seeing Date P cont θ cont (J2000) (J2000) (mags) (mins) ( ′′ ) ( ′′ ) (%) ( ◦ )XY Per HD 275877 03 49 36.3 +
38 58 55.5 A2IIv 9.4 120.0 1.2 0.7 10-11-2008 1 . +
25 19 57.1 A5IVe 8.3 60.0 1.5 1.0 08-11-2008 0 . +
12 09 10.2 A4pev 11.7 133.3 1.5 1.5 08-11-2008 1 . −
06 09 16.4 A2V 9.8 146.7 1.5 0.9 09-11-2008 0 . −
06 35 0.6 A5II-IIIev 10.3 133.3 1.5 0.8 09-11-2008 0 . −
09 42 11.1 A1 11.5 120.0 1.5 0.6 10-11-2008 1 . +
10 19 20.0 B6pe 8.8 53.4 1.5 0.9 09-11-2008 1 . −
11 18 3.3 B2Vne 6.6 15.3 1.5 1.3 08-11-2008 0 . +
15 47 15.6 A0IVe 7.2 60.0 1.5 1.2 08-11-2008 0 . +
47 13 43.6 B9.5Ve 10.2 110.0 1.5 1.0 09-11-2008 1 . +
62 08 45.0 B2IV-Vne 9.3 86.7 1.5 1.3 08-11-2008 4 . +
60 50 43.6 B0eq 11.6 96.7 1.5 1.0 09-11-2008 1 . Fig. 1.
An example of the spectropolarimetric signatures of the sample. The top half of the figure presents the spectropolarimetricPA, the percentage polarisation, and the Stokes intensity spectra centred upon H α . The data are binned to a constant polarisationerror, which is stated in the plots. The solid red line is the un-binned line profile. The QU diagrams of the signatures are displayedin the lower half of the figure. The solid lines mark the direction of the intrinsic polarisation. To supplement the spectropolarimetrically observed sample, in-trinsic polarisation angles (determined using spectropolarime-try) and binary PAs were taken from the literature. The resultantsample of objects for which both spectropolarimetric and binaryPAs are available is presented in Table 2. As the original studiesfrom which the data are taken were not primarily concerned withdisk orientations, we re-assessed all the polarisation angles takenfrom the literature to ensure they are calculated consistently. The QU data associated with the PAs listed in Table 2 were employedto determine approximate angles. If these di ff ered by more than90 ◦ from the literature values, the reported angles were rotatedby 90 ◦ (see the discussion above on calculating disk PAs fromspectropolarimetry). Table 2 presents the resultant sample. The sample was then further supplemented by adding diskPAs determined from direct imaging and multi-baseline interfer-ometry. The additional sample is also presented in Table 2. In order to compare disk and binary PAs in a meaningful way,a model is required to predict the di ff erence in the two anglesexpected for various scenarios. Baines et al. (2006) simply com-pare disk and binary PAs. However, binary PAs do not necessar-ily relate to the binary orbital plane. For example, even when abinary system and its circumprimary disk lie in the same plane,if the system is seen face-on, the binary PA is unrelated to the PAof the disk. In the case of more edge-on systems, the binary PA / Be binaries and disks
Table 2.
Binary systems for which a measurement of the intrinsic polarisation angle (from linear spectropolarimetry) or a directconstraint on the orientation of the circumprimary disk and the PA of the binary system is available. Column 3 denotes the spectraltype of the system primary, taken from SIMBAD unless otherwise stated. Column 4 lists the binary PA, column 5 lists the intrinsicpolarisation angle and column 6 contains the disk PA. Given the typical errors, the angles are presented to the closest degree. Object Alt. Name Type Bin. PA Pol. PA Disk PA( ◦ ) ( ◦ ) ( ◦ )Only spectropolarimetryMWC 166 HD 53367 B0 298 HD 58647 BD − ◦ MWC 158 HD 50138 B9 30 MWC 120 HD 37806 A2 34 V586 Ori HD 37258 A2 217 T Ori MWC 763 A3 107 HK Ori MWC 497 A4 47 Spectropolarimetry and imaging / interferometryHD 200775 MWC 361 B2 164 MWC 147 V700 Mon B6 82 HD 45677 FS CMa B2 BD + ◦ MWC 1080 V628 Cas B0 269 CQ Tau HD 36910 F3 56 HD 179218 MWC 614 A0 141 ∼ Only imaging / interferometryMWC 758 HD 36112 A5 311 V892 Tau HBC 373 B8 23 R Mon MWC 151 B0 287 ∼ MWC 297 NZ Ser B0 313 HR 5999 V856 Sco A7 ∼ ∼ HD 101412 PDS 57 B9.5 226 References.
1: From the data presented in Wheelwright et al. (2010), 2: M2007, 3: Baines et al. (2006), 4: Vink et al. (2005a), 5: These data,see Section 2.1.1, 6: Pirzkal et al. (1997), 7: Okamoto et al. (2009), 8: Kraus et al. (2008), 9: Cidale et al. (2001), 10: Patel et al. (2006), 11:Monnier et al. (2006), 12: Thomas et al. (2007), 13: Eisner et al. (2004), 14: Doucet et al. (2006), 15: Fedele et al. (2008), 16: Monnier et al.(2008), 17: Weigelt et al. (2002), 18: Fuente et al. (2006), 19: Vink et al. (2005c), 20: Manoj et al. (2007), 21: Stecklum et al. (1995), 22:Preibisch et al. (2006). is likely to be aligned with the disk PA. Nonetheless, at certainphases of the orbit, the binary PA will be quite di ff erent to thecircumprimary disk PA. It is only in the extreme case of an edge-on, co-planar system that the disk and binary PAs are constantlyaligned. Here we employ a new model to predict the averagedistribution in the di ff erence between disk and binary PA whencircumstellar disks and binary orbits lie in the same plane.The model is characterised by a random orbital phase; in-clination; semi-major axis and PA of the line about which thesystem is inclined. The eccentricity of the system was kept con-stant. Although a constant eccentricity is not truly representativeof the eccentricity distribution of PMS binary systems (see e.g.Goodwin et al. 2007), neither is a random eccentricity distribu-tion. Since the eccentricity can a ff ect the results, the input ec-centricity is left as a constant and is treated as a free parameter.This allows us to fit the observed distribution in di ff erences be-tween binary and disk PAs. The masses of the components werekept constant at 6 & 1 M ⊙ . The component masses have littleinfluence on the final distribution and are mentioned only forcompleteness.We used 10 000 random systems to determine the distribu-tion in the di ff erence between the disk and binary PAs ( ∆ PA).Spectropolarimetry is insensitive to face on systems as the pro-jected polarisation vectors cancel one-another out. Therefore,systems with low inclinations are discarded. The value of thecut-o ff inclination does a ff ect the final distribution. However, solong as large values (e.g. > ◦ ) are not chosen, the di ff erencesbetween the final distributions are relatively small. Since we also use imaging observations which are less sensitive to inclination,a low cut-o ff value (10 ◦ ) is used. The imaging and spectropolari-metric samples have di ff erent cut-o ff values. However, neithersample is large enough to be used in isolation, and thus a singlecut-o ff value is required. Changing the cut-o ff value has a similare ff ect to changing the eccentricity. Since we use the eccentricityas a free parameter, the use of a single cut-o ff value for the sam-ple of both imaging and spectropolarimetric data will not preventa fit to the data. By varying the eccentricity to fit the data, the un-certainty in the cut-o ff value is essentially incorporated into thebest-fitting eccentricity as a systematic uncertainty.Figure 2 presents a typical distribution. We are only con-cerned with the magnitude of the di ff erence between disk andbinary PAs. Therefore, we convert all di ff erences between theseangles to be in the range 0 − ◦ . For example, a di ff erence of170 ◦ is 10 ◦ away from alignment, and is thus classified as a dif-ference of 10 ◦ . An alignment between disk and binary PAs cor-responds to an o ff set of 0 ◦ . It can be seen that the distributiontends towards intrinsic alignment, and appears noticeably di ff er-ent to the distribution expected if the two angles are not related.Specifically, the distribution does not exhibit a direct correlationbetween binary and disk PA, but does demonstrate an excess ofaligned angles over the random distribution. This is to be ex-pected based on the previous discussion. To reiterate, co-planardisks and orbits are more likely to be observed to be aligned thannot, since at high inclinations the disk and binary PAs will be thesame. This is not the case for non-coplanar systems. Therefore,the co-planar distribution exhibits a preference for an alignment / Be binaries and disks
Fig. 2.
The expected distribution in the di ff erence between po-sition angles of disks and binary systems in a co-planar model(blue solid line) with an eccentricity of e = .
5. The black dottedline represents the expected distribution due to a random orien-tation of binary and disk PAs.between disk and binary position angles compared to the randomdistribution.We note that comparing this model to observations assumesthat the orientation of the circumstellar disks in the sample isprimordial. This need not be the case. Even if the circumstellardisks in HAe / Be binary systems originally lie in the plane of thebinary orbits, gravitational interactions with another star havethe potential to alter their orientation (see e.g. the discussion inBate et al. 2000). This would weaken an intrinsic correlation be-tween binary and disk PA. Therefore, any observed correlationbetween disk and binary PA may be more significant than it ini-tially appears. However, quantifying this is beyond the scope ofthe paper.
The final sample of HAe / Be binary systems with disk PAs con-tains 20 objects, and is thus more than three times greater thanthat of Baines et al. (2006). The di ff erences between disk andbinary PAs in the sample are compared to the co-planar modeland a random distribution. In Figure 3 we show the cumulativedistribution of the di ff erences in PA between the binary systemsand those derived for the disks. Also in the figure are the distri-butions predicted by the co-planar model and a completely ran-dom association of disk and binary PAs. A first glance indicatesthat the aligned scenario provides a better fit to the data thanthe random distribution. This is corroborated by a Kolmogorov-Smirnov (KS) test; while the co-planar model is found to be con-sistent with the data (it can only be ruled out at the 0.2 σ level),the random hypothesis is barely consistent with the data, and canbe rejected at a level of 2.2 σ .This 2 σ level of rejection may seem low, but we note thatboth models are more similar in their cumulative distributionsthan one might naively expect. As mentioned earlier, this is be-cause inclined binary systems do not necessarily have the sameobserved orientation as their circumprimary disks. This may ex-plain why our significance, with an improved data-set, is of thesame order as Baines et al. (2006) found with a smaller data-setand a more simplistic model. Fig. 3.
The di ff erence in disk and binary PAs presented in Table2 (red dashed). This is compared to the co-planar model with aconstant eccentricity of 0.1 (blue solid) and a distribution gen-erated by assuming the two angles are randomly aligned (blackdotted). Note that ∆ PA refers to the di ff erence between binaryand disk position angle (i.e. accounting for the 90 ◦ o ff set be-tween spectropolarimetric signatures and disks).Improving the current statistics requires further enlargingthe sample. We find that if the data followed the model distri-bution exactly, di ff erentiating between the two scenarios at a3 σ level would require a sample of approximately 50 objects.Nonetheless, the current data are consistent with the co-planarhypothesis and furthermore, this scenario is favoured over ran-dom alignments.
3. Does spectropolarimetry really tracecircumstellar disks?
Our use of spectropolarimetry assumes that the spectropolari-metric signatures of HAe / Be stars can be used to trace the ori-entations of their circumstellar disks. Linear polarisation hasbeen used for some time to probe the orientation of circumstel-lar disks (see e.g. Bastien & Menard 1990). However, the pro-posal of Harrington & Kuhn (2007) and Kuhn et al. (2007) thatthe spectropolarimetric signatures of HAe / Be stars are due to op-tical pumping and not scattering in a disk implies that these sig-natures do not have to trace disks. We address this issue here anddo this by comparing spectropolarimetric observations of youngstellar objects with independent measurements of the orientationof their disks. To increase the sample we include observations ofboth HAe / Be and T Tauri stars.Table 3 presents the sample of young stellar objects, bothlow mass T Tauri objects and Herbig Ae / Be stars, for which thecombination of spectropolarimetric observations and indepen-dent observations of their circumstellar disks is available. Wefirst concentrate on the HAe / Be sample and then assess whetherthere is a general trend within the combined sample of youngstellar objects. The majority of HAe / Be disks are thought to beoptically thin (see e.g. Natta et al. 2001). If the observed polar-isation signatures are due to single scattering in these disks, thepolarisation angles will be perpendicular to the disks’ PA on thesky. Therefore, if the spectropolarimetric signatures of HAe / Bestars are due to disks, the di ff erence between the disk and po-larisation angles for this sample will be 90 ◦ . In testing this hy-pothesis, we allow for an uncertainty in the di ff erence between / Be binaries and disks polarisation and disk PAs of ∼ ◦ , which is typical for the val-ues used. In some cases, there is a systematic uncertainty dueto di ff erent PAs being reported by di ff erent authors. For exam-ple, in the case of AB Aur, Mannings & Sargent (1997) report adisk PA of 79 ◦ while Corder et al. (2005) report various possi-ble angles ranging from 26 ◦ to 85 ◦ . However, such cases are notcommon, and are unlikely to influence conclusions regarding thewhole sample. In the case of the HAe / Be stars in Table 3, the majority of the ob-jects (7 out of 9) have polarisation angles approximately perpen-dicular to their imaged disks. This might be expected if the polar-isation signatures are due to single scattering in disks. Therefore,this appears to validate the use of spectropolarimetry to tracedisks. Here we quantify this. We use the HAe / Be star data inTable 3 to test the hypothesis that the intrinsic polarisation angleis always perpendicular to the disk PA (within the errors). Thecumulative distribution of the di ff erence in disk PA and polarisa-tion PA is shown in Figure 4. We find that the hypothesis that thedisk and polarisation angles are unrelated to each other and thusrandomly oriented can be discounted at a significant level (at3.1 σ according to the KS test). In contrast, the hypothesis thatthe spectropolarimetric signatures of the HAe / Be stars are ori-ented perpendicularly to their disks cannot be rejected beyondthe 1 σ level, and is thus consistent with the data. Therefore, wefind that spectropolarimetric signatures of HAe / Be stars do tracethe orientation of their circumstellar disks.We note that, although the majority of HAe / Be objects ex-hibit a di ff erence in disk and polarisation angle that is close to90 ◦ , two objects have disk and polarisation angles that are es-sentially aligned. This is contrary to expectations based on sin-gle scattering occurring in an optically thin disk. In their smallersample, Vink et al. (2005a) also note several objects where thisis the case. They suggest that while the spectropolarimetric sig-natures of all young stellar objects are due to circumstellar disks,the angle of the resulting polarisation vector is dependent uponthe properties of the inner disk. If the inner disk is optically thin,single scattering dominates and the resulting polarisation vectoris perpendicular to the disk PA. In contrast, if the inner disk isoptically thick, the polarisation vector is parallel to the disk PAdue to multiple scattering.Many of the T Tauri stars in Table 3 also have disk and po-larimetric PAs that are aligned. Therefore, if this argument iscorrect, it would appear that the majority of T Tauri star disksare optically thick in the inner regions. Although the previous test provides a statistically significant re-sult, it could be argued that this is dependent upon prior knowl-edge of the disks’ optical depth. Assuming HAe / Be star disksare optically thin is in part justified. The vast majority of theHAe / Be stars considered have an o ff set between disk and polari-sation angle that is close to 90 ◦ , as expected for single scatteringin optically thin disks. However, it could be argued that this is acircular argument. Moreover, in the case of the T Tauri objects,the disk and polarisation angles are generally aligned. A moreappropriate hypothesis for the total sample may be a combina-tion of the two scenarios, i.e. that polarisation angles are relatedto the PAs of disks, but that the polarisation angles can be either aligned or perpendicular to the disk. In general, we do not haveprior knowledge of the optical depth of the disks. Therefore, wetest the relationship between polarisation and disk angle withoutmaking a priori assumptions.In the scenarios mentioned above, the di ff erence between thedisk and spectropolarimetric angle is either 0 or 90 ◦ , dependingupon the optical depth of the inner disk. Consequently, it can beexpected that the o ff set from 45 ◦ to the di ff erence between diskand spectropolarimetric angle (henceforth ∆Ψ ) is always 45 ◦ .We note that this is the case regardless of whether the signatureis interpreted as being due to line polarisation or depolarisation,or whether the disk is optically thin or thick. Therefore, this testis even more robust than the previous which assumes opticallythin scattering and is subject to the polarisation signature beinginterpreted correctly. Here we compare this hypothesis to thesample presented in Table 3. The disk and spectropolarimetricangles in Table 3 are used to calculate ∆Ψ . This is then com-pared to the hypothesis that ∆Ψ is 45 ◦ by calculating the averageof 10 000 equally sized samples in which ∆Ψ is 45 ◦ but with anadditional random error contribution with a maximum value of15 ◦ (see Figure 4).The hypothesis that ∆Ψ is random can be discounted at asignificant level (above 3 σ ). This leaves the hypothesis that ∆Ψ is always 45 ◦ , which cannot be rejected at greater than a 1 σ level and is thus consistent with the data. This confirms the ear-lier finding that spectropolarimetric signatures trace disks anddemonstrates that this is not dependent on spectral type and clas-sification (i.e. T Tauri star or HAe / Be star).To summarise, we show that the spectropolarimetric datapresented in Table 3 do appear to trace circumstellar disks. Thiswould be expected if the polarisation is due to scattering in thesedisks. However, Harrington & Kuhn (2009) claim many of theHAe / Be stars in our sample have polarisation signatures whichrequire optical pumping and absorption in outflows (e.g. AB Aurand MWC 480). We show here that regardless of the polarisingmechanism, spectropolarimetry can be employed to trace the ori-entation of circumstellar disks.
4. Discussion and Conclusion
The results presented in this paper indicate a direct correla-tion between the spectropolarimetric signatures of pre-main-sequence stars and the orientations of their circumstellar disks.This is significant above the 3 σ level and appears indepen-dent of the classification of the young stellar objects. Therefore,we conclude that spectropolarimetric signatures of young stel-lar objects do indeed trace the orientation of their circumstel-lar disks. This is expected in the case of polarisation of stellarand accretion shock photons by disks (McLean & Clarke 1979;Vink et al. 2002). In the case of polarisation via optical pump-ing and absorption, the relationship between spectropolarimetricsignatures and disks is less clear.We note that Kuhn et al. (2010) suggest that the polarisa-tion signature of the Herbig Be star HD 200775 is due to op-tical pumping and that the signature does trace the orienta-tion of an imaged disk. However, the spectropolarimetric sig-nature of this object is observed across a double-peaked emis-sion line profile, and might therefore be the result of depolari-sation after all. Nevertheless, many Herbig Ae / Be stars, several / Be binaries and disks
Table 3.
Young stellar objects (column 1) for which spectropolarimetric observations and a direct measurement of their disk PA areavailable. The disk and adopted polarisation angles are presented in columns 4 and 5 and the di ff erence between them is listed incolumn 6. Finally, column 7 indicates objects where the di ff erence is close to 90 ◦ ( ⊥ ) or 0 ◦ ( k ). Object Alt. Name Type Disk PA Pol. PA ∆ PA( ◦ ) ( ◦ ) ( ◦ ) HAe / Be HD 200775 MWC 361 B2 7 ⊥ MWC 147 V700 Mon B6 80 ⊥ HD 45677 FS CMa B2 ⊥ BD + ◦ ⊥ MWC 1080 V628 Cas B0 55 k CQ Tau HD 36910 F3 120 ⊥ MWC 480 HD 31648 A3 150 ⊥ AB Aur HD 31293 A0 79 ⊥ HD 179218 MWC 614 A0IVe 23 ∼ k T Tauri
RY Tau HD 283571 F8 62 ⊥ SU Aur HD 282624 G2 127 k FU Ori HBC 186 G3 47 k GW Ori HD 244138 G5 56 (60) k DR Tau HBC 74 K5 128 k References.
1: Okamoto et al. (2009), 2: M2007, 3: Kraus et al. (2008), 4: Cidale et al. (2001), 5: Monnier et al. (2006), 6: Patel et al. (2006), 7:Eisner et al. (2004), 8: Doucet et al. (2006), 9: Vink et al. (2005a), 10: Mannings & Sargent (1997), 11: Fedele et al. (2008), 12: these data, 13:Akeson et al. (2003), 14: Akeson et al. (2002), 15: Malbet et al. (2005), 16: Mathieu et al. (1995), 17: Kitamura et al. (2002).
Fig. 4.
The distribution in the di ff erence between spectropolarimetrically predicted disk PA and observed disk PA for the samplepresented in Table 3 (blue dashed). This is compared to a random distribution (black short dotted). On the left we show a distributionwhere polarisation signatures are always orientated perpendicularly to circumstellar disks and on the right we present the distributionfor a scenario where the spectropolarimetric signatures can be either perpendicular or parallel to disks (see the text for more detail).Both model distributions have a maximum error of 15 ◦ . In both cases, a random orientation of disk and polarisation position anglescan be discarded at the 3 σ level.of which are in our sample, exhibit polarisation signatures as-sociated with P Cygni line profiles, i.e. outflowing gas (see e.g.Harrington & Kuhn 2009). Here we show that the signatures stillappear to trace the orientation of circumstellar disks. This im-plies that, if the spectropolarimetric signatures are due to opticalpumping and absorption in a wind, the wind geometry essen-tially mirrors that of the disk, at least in the regions where thepolarisation occurs.This is partly substantiated by recent observations ofthe H α emission of the Herbig Ae star AB Aur with theCHARA array by Rousselet-Perraut et al. (2010). This objecthas been proposed to exhibit polarisation due to optical pump-ing since it displays polarisation across the P Cygni absorp-tion component of its H α emission (Harrington & Kuhn 2007). Rousselet-Perraut et al. (2010) resolved the H α emitting regionaround AB Aur and found that it could be modelled as the baseof a wind represented by a flattened torus encompassing a cir-cumstellar disk. Provided the inclination and the angle betweenthe wind surface and disk mid-plane is low (20 ◦ and 35 ◦ in thecase of the disk-wind model of Rousselet-Perraut et al. 2010),such a flattened torus might well appear to have a similar mor-phology to the disk. To investigate the relative alignment of HAe / Be binary systemsand circumstellar disks we have used spectropolarimetry and / Be binaries and disks high spatial resolution data to determine the orientation of cir-cumstellar disks around the primary components of such sys-tems. We then combined these disk angles with binary param-eters to assess whether HAe / Be circumstellar disks and binarysystems are co-planar. Studies of lower mass T Tauri stars havefound that the circumstellar disks in T Tauri star binary sys-tems tend to be aligned, suggesting that such systems may formvia fragmentation (see e.g. Wolf et al. 2001; Jensen et al. 2004;Monin et al. 2006). Here we investigate whether this is also thecase for HAe / Be systems. We note that Maheswar et al. (2002)also compared HAe / Be binary and polarisation angles (althoughthese were calculated via broadband polarimetry and thus sub-ject to uncertainties in the correction for the interstellar polari-sation). These authors find that their data are inconsistent witha random association of disk and binary position angles with asignificance of 84 per cent. This is not very conclusive and theauthors note that they do not account for projection e ff ects andthus the actual correlation may be stronger. We do account forprojection e ff ects and demonstrate that, in principle, co-planarand randomly orientated disks and binaries can be di ff erentiated.We show that the data are best fit with a model in whichthe binary orbit and circumprimary disk are co-planar. This isconsistent with the suggestion that these systems formed via themonolithic collapse of a core and subsequent disk fragmenta-tion, which is how massive binary systems are thought to form(see Krumholz et al. 2009). However, as of yet, a random asso-ciation of disk and binary planes can only be excluded at a 2 σ level. A sample of approximately 50 objects is required to rejectthe random hypothesis and thus to di ff erentiate between the twoscenarios at 3 σ or higher. To summarise, we find that spectropolarimetric signatures ofyoung stellar objects do trace the orientation of their circum-stellar disks. This is independent of a specific mechanism forthe linear polarisation. In itself, this finding is insensitive to thepolarising mechanism as all mechanisms require some form ofasymmetric geometry. We note that scattering in a disk appearsa plausible polarisation mechanism as it naturally explains therelationship between disk and polarisation angle. Furthermore,assuming the polarisation of T Tauri stars is due to multiple scat-tering in optically thick disks, it is consistent with the observa-tion that T Tauri star polarisation is generally parallel to imageddisks while the reverse is true for most HAe / Be stars. It is notclear how optical pumping and polarisation via absorption canreproduce the di ff erent polarimetric behaviour of HAe / Be andT Tauri stars. Further modelling is required to investigate thisissue.We conclude that our results are entirely consistent withthe disks and orbits of HAe / Be binaries being co-planar, andthus with the scenario of binary formation via disk fragmen-tation. Further spectropolarimetric observations, e.g. providedby SALT, in conjunction with additional high resolution data,e.g. provided by NIR interferometry, are required to increase thesample and conclusively di ff erentiate between aligned and ran-dom orientations of disks and binaries. References
Akeson, R. L., Ciardi, D., & van Belle, G. T. 2003, in SPIE Conf. Ser., ed.W. A. Traub, Vol. 4838, 1037–1042Akeson, R. L., Ciardi, D. R., van Belle, G. T., & Creech-Eakman, M. J. 2002,ApJ, 566, 1124 Baines, D., Oudmaijer, R. D., Mora, A., et al. 2004, MNRAS, 353, 697Baines, D., Oudmaijer, R. D., Porter, J. M., & Pozzo, M. 2006, MNRAS, 367,737Bastien, P. & Menard, F. 1990, ApJ, 364, 232Bate, M. R., Bonnell, I. A., Clarke, C. J., et al. 2000, MNRAS, 317, 773Beskrovnaya, N. G., Pogodin, M. A., Miroshnichenko, A. S., et al. 1999, A&A,343, 163Bonnell, I. A. & Bate, M. R. 2005, MNRAS, 362, 915Cidale, L., Zorec, J., & Tringaniello, L. 2001, A&A, 368, 160Clarke, D. & McLean, I. S. 1974, MNRAS, 167, 27PCorder, S., Eisner, J., & Sargent, A. 2005, ApJL, 622, L133Doucet, C., Pantin, E., Lagage, P. O., & Dullemond, C. P. 2006, A&A, 460, 117Eisner, J. A., Lane, B. F., Hillenbrand, L. A., Akeson, R. L., & Sargent, A. I.2004, ApJ, 613, 1049Fedele, D., van den Ancker, M. E., Acke, B., et al. 2008, A&A, 491, 809Fuente, A., Alonso-Albi, T., Bachiller, R., et al. 2006, ApJ, 649, L119Goodwin, S. P., Kroupa, P., Goodman, A., & Burkert, A. 2007, in Protostars andPlanets V, ed. B. Reipurth, D. Jewitt, & K. Keil, 133–147Harrington, D. M. & Kuhn, J. R. 2007, ApJL, 667, L89Harrington, D. M. & Kuhn, J. R. 2009, ApJS, 180, 138Herbig, G. H. 1960, ApJ Supplement, 4, 337Hern´andez, J., Calvet, N., Brice˜no, C., Hartmann, L., & Berlind, P. 2004, AJ,127, 1682Hillenbrand, L. A., Strom, S. E., Vrba, F. J., & Keene, J. 1992, AstrophysicalJournal, 397, 613Jensen, E. L. N., Mathieu, R. D., Donar, A. X., & Dullighan, A. 2004, ApJ, 600,789Kitamura, Y., Momose, M., Yokogawa, S., et al. 2002, ApJ, 581, 357Kraus, S., Preibisch, T., & Ohnaka, K. 2008, ApJ, 676, 490Krumholz, M. R., Klein, R. I., McKee, C. F., O ff ner, S. S. R., & Cunningham,A. J. 2009, Sci, 323, 754Kuhn, J. R., Berdyugina, S. V., Fluri, D. M., Harrington, D. M., & Stenflo, J. O.2007, ApJL, 668, L63Kuhn, J. R., Geiss, B., & Harrington, D. M. 2010, ArXiv e-printsMaheswar, G., Manoj, P., & Bhatt, H. C. 2002, A&A, 387, 1003Malbet, F., Lachaume, R., Berger, J., et al. 2005, A&A, 437, 627Mannings, V. & Sargent, A. I. 1997, ApJ, 490, 792Manoj, P., Ho, P. T. P., Ohashi, N., et al. 2007, ApJL, 667, L187Mathieu, R. D., Adams, F. C., Fuller, G. A., et al. 1995, AJ, 109, 2655McLean, I. S. & Clarke, D. 1979, MNRAS, 186, 245Moeckel, N. & Bally, J. 2007, ApJ, 656, 275Monin, J.-L., M´enard, F., & Peretto, N. 2006, A&A, 446, 201Monnier, J. D., Berger, J.-P., Millan-Gabet, R., et al. 2006, ApJ, 647, 444Monnier, J. D., Tannirkulam, A., Tuthill, P. G., et al. 2008, ApJL, 681, L97Monnier, J. D. e. a. 2005, ApJ, 624, 832Mottram, J. C., Vink, J. S., Oudmaijer, R. D., & Patel, M. 2007, MNRAS, 377,1363Natta, A., Prusti, T., Neri, R., et al. 2001, A&A, 371, 186Okamoto, Y. K., Kataza, H., Honda, M., et al. 2009, ApJ, 706, 665Oudmaijer, R. D. & Drew, J. E. 1999, MNRAS, 305, 166Patel, M., Oudmaijer, R. D., Vink, J. S., Mottram, J. C., & Davies, B. 2006,MNRAS, 373, 1641Pirzkal, N., Spillar, E. J., & Dyck, H. M. 1997, ApJ, 481, 392Poeckert, R. 1975, ApJ, 196, 777Poeckert, R. & Marlborough, J. M. 1976, ApJ, 206, 182Preibisch, T., Kraus, S., Driebe, T., van Boekel, R., & Weigelt, G. 2006, A&A,458, 235Rousselet-Perraut, K., Benisty, M., Mourard, D., et al. 2010, A&A, 516, L1Stecklum, B., Eckart, A., Henning, T., & Loewe, M. 1995, A&A, 296, 463Th´e, P. S., de Winter, D., & Perez, M. R. 1994, A&AS, 104, 315Thomas, S. J., van der Bliek, N. S., Rodgers, B., Doppmann, G., & Bouvier, J.2007, in IAU Symp, Vol. 240, IAU Symp., ed. W. I. Hartkopf, E. F. Guinan,& P. Harmanec, 250–253Tody, D. 1993, in A.S.P. Conf.Ser., Vol. 52, Astronomical Data AnalysisSoftware and Systems II, ed. R. J. Hanisch, R. J. V. Brissenden, & J. Barnes,173–183Vink, J. S., Drew, J. E., Harries, T. J., & Oudmaijer, R. D. 2002, MNRAS, 337,356Vink, J. S., Drew, J. E., Harries, T. J., Oudmaijer, R. D., & Unruh, Y. 2005a,MNRAS, 359, 1049Vink, J. S., Harries, T. J., & Drew, J. E. 2005b, A&A, 430, 213Vink, J. S., O’Neill, P. M., Els, S. G., & Drew, J. E. 2005c, A&A, 438, L21Waters, L. B. F. M. & Waelkens, C. 1998, ARA&A, 36, 233Weigelt, G., Balega, Y. Y., Hofmann, K.-H., & Preibisch, T. 2002, A&A, 392,937Wheelwright, H. E., Oudmaijer, R. D., & Goodwin, S. P. 2010, MNRAS, 401,1199Wolf, S., Stecklum, B., & Henning, T. 2001, in IAU Symp., Vol. 200, The / Be binaries and disks
Formation of Binary Stars, ed. H. Zinnecker & R. Mathieu, 295–304Zinnecker, H. & Yorke, H. 2007, ARA&A, 45, 481
Appendix A: Observed spectropolarimetricsignatures
Here we present the spectropolarimetric signatures of the entiresample over H α . Figure A.1 presents the signatures in terms ofthe total flux, the amount of polarisation and the polarisation an-gle over the H α line. The QU diagrams of MWC 1080 and MWC147, objects which display a spectropolarimetric signature thatwas not presented in the main body of the paper, are presentedin Figure A.2. / Be binaries and disks
Fig. A.1.
The spectropolarimetric signatures of the sample. For each object the spectropolarimetric PA, the percentage polarisation,and the Stokes intensity spectra are presented centred upon H α . The data are binned to a constant polarisation error, which is statedin the plots. The solid red line is the un-binned line profile. Fig. A.2.
The QU diagrams of MWC 1080 and MWC 147. The data are binned to a constant polarisation error, which is stated inthe plots.diagrams of MWC 1080 and MWC 147. The data are binned to a constant polarisation error, which is stated inthe plots.