A circumstellar disk associated with a massiveprotostellar objectZhibo Jiang1, Motohide Tamura2, Misato Fukagawa2, Jim Hough3, Phil Lucas3, Hiroshi Suto2, Miki Ishii2

& Ji Yang1

The formation process for stars with masses several times that ofthe Sun is still unclear. The two main theories are mergers ofseveral low-mass young stellar objects1, which requires a highstellar density, or mass accretion from circumstellar disks in thesame way as low-mass stars are formed2, accompanied by outflowsduring the process of gravitational infall. Although a number ofdisks have been discovered around low- and intermediate-massyoung stellar objects3,4, the presence of disks around massiveyoung stellar objects is still uncertain and the mass of the disksystem detected around one such object5, M17, is disputed6. Herewe report near-infrared imaging polarimetry that reveals an out-flow/disk system around the Becklin–Neugebauer protostellarobject, which has a mass of at least seven solar masses (M().This strongly supports the theory that stars with masses of at least7M( form in the same way as lower mass stars.

The Becklin–Neugebauer object (BN; ref. 7) is a famous candidateprotostar located within the central part of the Orion molecularcloud (OMC-1). With a distance of ,1,500 light-years from the Sun(,1.39 £ 1019 m; ref. 8), this object has a luminosity of ,2,500 solarluminosities (L(), but may be up to ,104L( taking its associatedluminosity into account9. A few studies suggest a mass of ,7M(

(refs 10, 11), but it may be as much as 20M( (ref. 11). The object isred in the near-infrared with [H 2 K] < 3.8 mag (lH ¼ 1.65 mm,lK ¼ 2.2 mm). Even after correction for a line-of-sight visual extinc-tion of AV < 17 (ref. 9), it has a significantly large colour excess,suggesting the presence of circumstellar materials.

To further investigate the properties of BN, we report here H-bandand K-band high-resolution (full-width at half-maximum,FWHM ¼ 0.1 00) polarimetric imaging. Figure 1 shows the imagesof polarization degrees superposed by corresponding brightnesscontours and polarization vectors in the H- and K-bands in thevicinity of BN. A butterfly-shaped bright pattern and a flare-shapeddark lane running from the southeast to the northwest at aposition angle of ,1268 (PA, measured from north to east) areclearly seen in the H-band (Fig. 1a), but are less clear in the K-band.This structure can be interpreted as an outflow/disk system aroundthe star.

There are two main processes for producing polarization: differ-ential absorption by non-spherical dust grains with the short axispreferentially aligned along the local magnetic field, referred to asdichroic extinction12, and by scattering. Dichroic absorption alone isunlikely to give the observed polarization morphology, because thiswould require a sandwich-like dust distribution with the opticaldepth being lower in the central lane than in the two wings13,14, andthe shape of the bright wings is quite regularly parabolic, which isunlikely to arise from extinction. The highly polarized bipolar featurein Fig. 1a is most easily interpreted as the cavity walls formed by abipolar outflow from BN. Our unpublished polarimetric data of long

exposure but lower spatial resolution in the K-band indicate reflec-tion nebulosities ,5 00 away to the northeast and southwest, which areilluminated by BN. Other evidence for a bipolar outflow comes fromthe presence of local H2 emission to the northeast of BN15,16. The darklane, which is nearly perpendicular to the bipolar outflow, almostcertainly represents the circumstellar disk around the object. Sup-porting evidence can be found from the 12.5-mm image in ref. 17, inwhich the mid-infrared emission of BN is elongated in the samedirection, although the disk is not resolved. In view of the roughlysymmetric polarization morphology, the disk axis should be lyingnearly in the plane of the sky. This kind of configuration has beenobserved around low-mass young stellar objects such as HL Tau18,and Herbig Ae/Be stars5.

To show that a disk/bipolar outflow system can create such apolarization structure, we use a Monte Carlo code for dust scatteringto simulate the emerging photons from the system. The resultantpolarization image from the scattering model is shown in Fig. 2a. Thepolarization image clearly shows a butterfly structure with a dark lanerunning between two wings. This structure is quite similar to theobserved features in the H-band, thus confirming the presence of acircumstellar disk around the object.

Despite the overall similarity between the observations and themodel, there are some obvious differences. Instead of a centro-symmetric pattern, the observed polarization vectors appear parallel.This is different to low-mass young stellar objects, where centro-symmetric polarization patterns dominate the reflection nebulae.Here, dichroic absorption may be producing significant polarization.Observations have shown that the foreground magnetic field isoriented at PA < 1208 (refs 14, 15, 19), and non-spherical dustgrains will produce absorptive polarization vectors parallel to thefield. The overall effect will be to produce a pattern of essentiallyparallel polarization vectors, with degrees of polarization higher inthe H-band than in the K-band. To explore this possibility, we use avirtual partial polarizer, with its main axis aligned parallel to the diskplane, in front of the scattering model. Assuming the polarizer canproduce 20% polarized light, we successfully modified the centro-symmetric pattern, as shown in Fig. 2b. It is therefore possible tointerpret the observed polarization in terms of the combined effect ofdust scattering and dichroic extinction.

Another interesting feature is that the polarization degrees in thetwo wings are quite different. The polarization degrees in the north-east are larger than those in the southwest by ,20%. Figure 3 showsthe polarization degrees changing with the distance from the centralobject along the axis (PA < 368) and disk plane for the H-band.Monte Carlo simulations with different disk inclinations eitherchange the polarization morphology significantly or do not yieldsuch large differences in polarization degree. Therefore it is not likelythat the difference comes from the viewing angle of the disk. The

LETTERS

1Purple Mountain Observatory, Chinese Academic of Sciences, Nanjing 210008, China. 2National Astronomical Observatories of Japan, Osawa 2-21-1, Mitaka, Tokyo 181-8588,Japan. 3Department of Physics, Astronomy & Mathematics, University of Hertfordshire, College Lane, Hatfield AL10 9AB, UK.

Vol 437|1 September 2005|doi:10.1038/nature04012

112© 2005 Nature Publishing Group

most likely interpretation is differential extinction or different dustdensities between the two wings.

The orientation of the system is interesting: the axis of the disk andthe outflow cavity is along PA < 368, which is roughly perpendicularto the large-scale magnetic field in OMC-1. It is believed that themagnetic field plays an important role when a star is forming throughgravitational collapse, with the system axis aligning with the localmagnetic field20–22. The case of BN is quite different. In addition, thedisk orientation rules out BN as the powering source for the large-scale energetic H2 outburst16 in the region, whose orientation is

roughly orthogonal. The misalignment between the system and thelocal magnetic field, and the survival of the disk in such a turbulentregion suggest that BN is not associated with other embeddedprotostars within OMC-1, such as IRc2.

Two competitive scenarios have been proposed to account formassive star formation. One proposes a process similar to that forlow-mass star formation with the massive stars formed throughgravitational collapse of the molecular clouds and subsequent accre-tion from the circumstellar disk2. An argument against this model isthat the enormous radiation produced by more massive stars wouldhalt the mass accretion, preventing very massive stars from beingformed. Although ways of overcoming this have been proposed, forexample, by changing the dust properties and the accretion rate23,24 orby outflows25, it is essential to demonstrate that massive stars can beproduced through accretion by observing an outflow/disk system.The mass of BN is estimated to be ,7M( or higher, and so ourobservations provide evidence that stars with masses of up to at least,7M( can be formed through gravitational collapse and mass

Figure 1 | Polarimetric result of the observation. a, Polarization vectors(red) and brightness contours (white) superposed on polarization degreeimage (blue) in the H-band. The contours start from 2.4 £ 1024 mJy pixel21

with an increasing factor of 101/2. The peak intensity is at right ascension,RA ¼ 05 h 35 min 14.12 s, and declination, dec. ¼ 258 22’ 23.2 00 (J2000).Polarization vectors are binned within 5 £ 5 pixels to avoid crowding. Abutterfly pattern with a dark lane (lower polarization degrees) passingthrough the centre of the object is clearly seen. The dark lane isapproximately 500 AU (1 AU ¼ 1.5 £ 1011 m) long and 100 AU wide at thenarrowest section. The polarization degrees are typically 20% in the darklane, rising to as much as ,35% in the two wings of the butterfly. The linearand polarization scales are shown at the lower-right corner. North is up andeast is to the left. b, Same as a, but in the K-band. The contours start from0.21 mJy pixel21 with an increase factor the same as that of the H-band. Toimprove the signal to noise ratio, the polarization image has been processedwith median filtering within a box size of 3 £ 3 pixels.

Figure 2 | Results ofMonte Carlo simulations. The polarization vectors andbrightness contours are superposed on the polarization degree image as inFig. 1. The image as well as the polarization vectors is rotated to align thedirection of Fig. 1. The scales of polarization degrees and linear size of 100 AU

are indicated at the lower-right corner of the figure. a, Result of the purescattering model; b, Result of the scattering model plus dichroic extinction(producing 20% polarization).

NATURE|Vol 437|1 September 2005 LETTERS

113© 2005 Nature Publishing Group

accretion. Recent near-infrared spectroscopy shows some accretionsignatures from massive young stars26, but high-resolution near-infrared imaging polarimetry, as reported here, can provide the mostdirect evidence for circumstellar disks around the most massive stars.

METHODSImaging polarimetry. The polarimetric data were obtained on 3 January 2003with the CIAO (Coronagraphic Imager with Adaptive Optics; ref. 27) and itspolarimeter28 mounted on the Subaru telescope. The pixel scale was 0.021 00 . Noocculting masks were used in these observations. The polarimeter consists of astepped half-wave plate (images taken at waveplate angles of 0, 22.5, 45 and 67.5degrees), upstream of the adaptive optics system, and a cooled wiregrid polarizerinside the CIAO cryostat. The integration time per waveplate position is 240 s atH-band and 3 s at K-band. A nearby optical star was used for wavefront sensingof the adaptive optics. The sky was clear, and the natural seeing size was 0.5 00 .The spatial resolution achieved with the adaptive optics system was 0.10 00

(FWHM). After image calibrations in the standard manner, including darksubtraction, flat-fielding with dome-flats, bad-pixel substitution, and sky sub-traction, using the data reduction package Image Reduction and Analysis Facility(IRAF), the Stokes parameter images can be obtained using Q ¼ I0 2 I45,U ¼ I22.5 2 I67.5, and I ¼ (I0 þ I22.5 þ I45 þ I67.5)/2. The polarization degreeimage is then derived by Ip ¼ sqrt(Q2 þ U2)/I and polarization angle image byI v ¼ 1/2arctan(U,Q). The polarization angles are calibrated by BN itself, notedin ref. 14. Aperture polarimetry towards BN with an aperture size of,0.5

00shows

that the overall polarization degrees, 14.5% at K and 26.3% at H, agree well withthe values in ref. 14.Modelling. The modelling is based on the work of ref. 18. The detailed

description of the code and of the parameters can be found there. We used theforced scattering code to ensure calculation efficiency. The density profile of thedisk is in the form:

rðr;hÞ ¼ r0ðr=r0Þ2aexp{2 1=2ðz=hÞ2}

where h is the disk scale height at radius r: h(r) ¼ h0(r/r0)b, and r0 is the innerradius of the disk. The density profile of the envelope is in the form:

rðR;hÞ ¼ C=R1:5½1=ððjzj=RÞ0:5 þ 0:05Þ� for Rcav , r, Rsys

and rðR;hÞ ¼ constant for r, Rcav

where R ¼ sqrt(r2 þ z2), C is a free parameter governing the optical depth, andR cav and R sys are the radii of the cavity and system envelope, respectively. A dustmodel of silicate coated with ice has been used in the simulation. The maximumsize of the particles in the dust mixture is 1.0mm. The simulation is in theH-band. Assuming an edge-on viewing angle, virtually all parameter sets give apolarization image similar to that observed, with the best-fit parameters given inTable 1. The parameter set results in an optical depth of about 8.5 at H. With thedust configuration we roughly estimate the mass of the envelope and in the cavityto be 1.7 £ 1023M( and 1.0 £ 1023M(, respectively. The disk mass cannot beestimated from the data.

We simulate the dichroic extinction by introducing a virtual partial polarizer,producing p% polarization. The emerging Stokes parameters can be obtained by:

I0¼ 0:5 P2

1 þP22

� �Iþ P2

2 2P21

� �Q

� �

Q0¼ 0:5 P2

2 2P21

� �Iþ P2

1 þP22

� �Q

� �

U0¼2P1P2U

where P1, P2 are the transparencies in the x and y directions. To ensurethat the polarization degree of the emerging light is p%, P1 and P2 are setto P 1 ¼ [(100 þ p)/(100 2 p)]1/4exp(2t/2), and P2 ¼ ½ð1002 pÞ=ð100þpÞ�1=4expð2t=2Þ where t is the optical depth along the line of sight. In thesimulation the t value does affect the polarization.

Received 21 May; accepted 29 June 2005.

1. Bonnell, I. A., Bate, M. R. & Zinnecker, H. On the formation of massive stars.Mon. Not. R. Astron. Soc. 298, 93–-102 (1998).

2. Shu, F. H., Adams, F. C. & Lizano, S. Star formation in molecular clouds—Observation and theory. Annu. Rev. Astron. Astrophys. 25, 23–-81 (1987).

3. O’Dell, C. R. & Wen, Z. Postrefurbishment mission Hubble Space Telescopeimages of the core of the Orion Nebula: Proplyds, Herbig-Haro objects, andmeasurements of a circumstellar disk. Astrophys. J. 435, 194–-202 (1994).

4. Perrin, M. D. et al. Laser guide star adaptive optics imaging polarimetry ofHerbig Ae/Be stars. Science 303, 1345–-1348 (2004).

5. Chini, R. et al. The formation of a massive protostar through the disk accretionof gas. Nature 429, 155–-157 (2004).

6. Sako, S. et al. No high-mass protostars in the young stellar object M17–-S01.Nature 434, 995–-998 (2005).

7. Becklin, E. E. & Neugebauer, G. Observations of an infrared star in the Orionnebula. Astrophys. J. 147, 799–-802 (1967).

8. Genzel, R. & Stutzki, J. The Orion molecular cloud and star-forming region.Annu. Rev. Astron. Astrophys. 27, 41–-85 (1989).

9. Gezari, D. Y., Backman, D. E. & Werner, M. W. Mid-infrared imaging of OrionBN/KL. II. Luminosity sources, extinction distribution, and the nature of IRc2.Astrophys. J. 509, 283–-298 (1998).

10. Tan, J. C. The Becklin-Neugegauer object as a runaway B star, ejected 4000years ago from the v1 Orionis C system. Astrophys. J. 607, L47–-L50 (2004).

11. Hillenbrand, L. A., Carpenter, J. M. & Skrutskie, M. F. Periodic photometricvariability in the Becklin-Neugegauer object. Astrophys. J. 547, L53–-L56(2001).

12. Davis, L. & Greenstein, H. L. The polarization of starlight by aligned dust grains.Astrophys. J. 114, 206–-240 (1951).

13. Jones, T. J. Infrared polarimetry and the interstellar magnetic field. Astrophys. J.346, 728–-734 (1989).

14. Minchin, N. R. et al. Near-infrared imaging polarimetry of bipolar nebulae—I.The BN-KL region of OMC-1. Mon. Not. R. Astron. Soc. 248, 715–-729 (1991).

15. Chrysostomou, A., Hough, J. H., Burton, M. G. & Tamura, M. Twisting magneticfields in the core region of OMC-1. Mon. Not. R. Astron. Soc. 268, 325–-334(1994).

16. Schultz, A. S. B. et al. NICMOS narrowband images of OMC-1. Astrophys. J. 511,282–-288 (1999).

17. Shuping, R. Y. & Morris, M. A new mid-infrared map of the BN/KL region usingthe Keck telescope. Astron. J. 128, 363–-374 (2004).

18. Lucas, P. W. et al. High-resolution imaging polarimetry of HL Tau and magneticfield structure. Mon. Not. R. Astron. Soc. 352, 1347–-1364 (2004).

19. Hough, J. H. et al. Infrared polarization in OMC-1—discovery of a molecularhydrogen reflection nebula. Mon. Not. R. Astron. Soc. 222, 629–-644 (1986).

Table 1 | Parameters used in Monte Carlo simulation

Number of photons 10,000,000Inner radius of accretion disk, r0 (AU) 0.80Outer radius of accretion disk (AU) 800Disk radial density power law index, a 1.875Disk flaring power law index, b 1.125Scale height of accretion disk at r ¼ r0 (AU) 0.33Albedo of dust grain mixture 0.50Polar viewing angle (degrees) 85Adopted dust model Coated silicates18

Inner radius of cavity (AU) 40Conical cavity opening angle, axis to edge (degrees) 75Cavity optical depth 0.3System radius (AU) 1,500

Figure 3 | Polarization degrees as a function of the distance R from thecentral object in the H-band. Blue dots present the polarization variancecutting along the outflow axis (PA ¼ 368), while red dots are along the disk(PA ¼ 1268). For the blue dots, negativeR is to the southwest and positive tothe northeast. For red plots negative R is to the northwest and positive to thesoutheast. The error bars indicate the length of ^1j (standard deviation).The asymmetric polarization along the axis, within 400 AU, is significant.Although the errors are larger beyond 400 AU, the average polarizationdegrees in the northeastern lobe are still higher than those in thesouthwestern one. There is also some asymmetry in the polarization alongthe disk.

LETTERS NATURE|Vol 437|1 September 2005

114© 2005 Nature Publishing Group

20. Cohen, R. J., Rowland, P. R. & Blair, M. M. The source of bipolar outflow inCepheus A. Mon. Not. R. Astron. Soc. 210, 425–-438 (1984).

21. Tamura, M. et al. Infrared polarimetry of dark clouds—III. The relationshipbetween the magnetic field and star formation in the NGC 1333 region. Mon.Not. R. Astron. Soc 231, 445–-453 (1988).

22. Matsumoto, T. & Tomisaka, K. Directions of outflows, disks, magnetic fields,and rotation of young stellar objects in collapsing molecular cloud cores.Astrophys. J. 616, 266–-282 (2004).

23. Wolfire, M. & Cassinelli, J. Conditions for the formation of massive stars.Astrophys. J. 319, 850–-867 (1987).

24. Yorke, H. W. & Sonnhalter, C. On the formation of massive stars. Astrophys. J.569, 846–-862 (2002).

25. Krumholz, M. R., McKee, C. F. & Klein, R. I. How protostellar outflows helpmassive stars form. Astrophys. J. 618, L33–-L36 (2005).

26. Blum, R. D. et al. Accretion signatures from massive young stellar objects.Astrophys. J. 617, 1167–-1176 (2004).

27. Tamura, M. et al. Coronagraph imager with adaptive optics (CIAO):

description and first results. Proc. SPIE 4008, 1153–-1161 (2000).

28. Tamura, M. et al. Near-infrared polarimeter for the Subaru telescope. Proc. SPIE

4843, 190–-195 (2003).

Acknowledgements This paper is based on data collected at Subaru Telescope,which is operated by the National Astronomical Observatory of Japan. Thiswork is supported by a Grant-in-Aid from MEXT, Japan, and NSFC of China.

Author Contributions M.T., M.F., H.S. and M.I. collected the data. P.L. and M.F.did the modelling. M.T., J.H., P.L. and J.Y. contributed to the scientific discussion.Z.J. conducted data reduction and wrote the paper with help from allco-authors.

Author Information Reprints and permissions information is available atnpg.nature.com/reprintsandpermissions. The authors declare no competingfinancial interests. Correspondence and requests for materials should beaddressed to Z.J. (zbjiang@pmo.ac.cn).

NATURE|Vol 437|1 September 2005 LETTERS

115© 2005 Nature Publishing Group