Mon. Not. R. Astron. Soc. 000, 000{000 (0000)
Printed 6 January 1997
(MN LATEX style le v1.4)
Radio and infrared structure of the colliding-wind
Wolf-Rayet system WR 147
P.M. Williams1, S.M. Dougherty2 3, R.J. Davis4, K.A. van der Hucht5,
M.F.
Bode6 & D.Y.A. Setia Gunawan5
1
;
Royal Observatory, Blackford Hill, Edinburgh EH9 3HJ
2 Physics & Astronomy, University of Calgary, 2500 University Drive NW, Calgary, Alberta T2N 1N4, Canada
3 Dominion Radio Astrophysical Observatory, P.O. Box 248, White Lake Rd, Penticton, British Columbia V2A 6K3, Canada
4 University of Manchester Nueld Radio Astronomy Laboratories, Jodrell Bank, Maccleseld, Cheshire, SK11 9DL
5 Space Research Organization Netherlands, Sorbonnelaan 2, 3584 CA Utrecht, The Netherlands
6 Astrophysics Group, School of Electrical Engineering, Electronics and Physics, Liverpool John Moores University,
Byrom St, Liverpool, L3 3AF
Received 1996 Dec 30: v7, Dec 22
ABSTRACT
New, high-resolution infrared and radio images of the X-ray luminous Wolf-Rayet system WR147 (AS 431) are presented. The 5-GHz radio image resolves both components
of this double source. The emission from one component, that to the South and associated with the wind of the WN8 star is 170 253 mas, indicating that the stellar
wind is not spherically symmetric. The second, non-thermal component 0 6 north
of the WN8 star, is extended East-West to 267 mas. The infrared image reveals
the presence of a companion to the WN8 star, close to the non-thermal radio source
but slightly ( 60 mas) more distant from the WN8 star. The companion is 3
mag. fainter than the WN8 star and has a luminosity of a B0.5V star, just suciently
luminous to possess a stellar wind capable of colliding with that of the WN8 star. The
presence of the non-thermal emission between the two stars and much closer to that
with the weaker wind is direct evidence for a colliding-wind origin for the emission.
A signicant portion of the X-ray emission can also be accounted for by the release
of energy in the wind collision. Comparison of the non-thermal ux with those of the
three systems incorporating WC type stars (WR125, WR140 and WR146) shows a
correlation with velocity of the WR wind.
Key words:
radio continuum: stars | Stars: Wolf-Rayet | Stars: individual:
WR 147 | X-rays: stars
00
:
K
1 INTRODUCTION
Wolf-Rayet (WR) stars emit free-free continuum radiation
from their extended ionized stellar wind envelopes. This
emission is readily observable in the infrared and, particularly, radio wavelength regions where the continuum spectrum has the form S / , with spectral index 0:7?0:8;
depending on the conditions in the wind. In addition to this
emission, a number of WR stars exhibit strong, non-thermal
emission which is characterized by zero or negative spectral
index, variable ux density, and typically high brightness
temperature (TB 106 ? 108 K). This emission is thought to
be synchrotron radiation arising from electrons accelerated
to relativistic velocities by rst-order Fermi acceleration in
shocks, arising from either wind instabilities in single stars
(e.g. Chen & White 1994), or the region where the winds of
two components of a massive binary collide (e.g. Eichler &
c 0000 RAS
Usov 1993). A third mechanism, acceleration of the electrons
to relativistic velocities by magnetic eld compression from
the colliding winds, has been proposed by Jardine, Allen &
Pollock (1996).
Four WR stars have been studied as non-thermal emitters: WR 125, WR 140 (HD 193793), WR 146 and WR 147
(Williams 1996). Of these, WR 140 (WC7+O4-5V) has been
demonstrated to be a binary system (Williams et al. 1990a),
while spectroscopic evidence for O-type companions to the
WC stars in WR 125 and WR 146 was found by Williams et
al. (1994) and Dougherty et al. (1996), suggesting that the
non-thermal emission in all three systems arises from the
interaction of the WR winds with their companions.
In the case of WR 140, the non-thermal emission was
shown to vary in phase with the system's 7.94-y period by
Williams et al. (1990a), who proposed a model in which a
2
P.M. Williams et al.
non-thermal source in the wind-collision region was taken
to be much closer to the O4-5 star than the WC star on
account of the higher mass-loss rate of the latter. The nonthermal source therefore moves with the O4-5 star inside the
stellar wind of the WC star during the orbit. As a result,
the free-free opacity in the Wolf-Rayet wind along lines of
sight to the collision region changes throughout the orbit
as the geometry changes. Our ability to observe the nonthermal emission requires the absence of appreciable freefree opacity along our line of sight to the emission region
which, in the case of WR 140, occurs for only a fraction of
each period. This model has been developed quantitatively
by Eichler & Usov (1993). Multi-frequency light curves were
observed and used by Williams, van der Hucht & Spoelstra
(1994) and White & Becker (1995) to study the circumstellar
extinction and intrinsic non-thermal emission. In particular,
White & Becker argued from the variation of the intrinsic
non-thermal ux that the undisturbed wind of the WolfRayet component was not spherical, but attened somewhat
into a disk.
The non-thermal radio emission from WR 125 faded between 1984 and 1988, perhaps due to an increase in circumstellar extinction as the system approached periastron
passage shortly before its dust-formation episode (Williams
et al. 1992). The non-thermal emission had not returned by
1995 (Contreras et al. 1996), indicating that, if the variation
is periodic like that of WR 140, the period is in excess of 13 y.
This is consistent with the minimum 15-y interval between
dust-formation episodes inferred by Williams (1996) from
the infrared light curves.
Dougherty et al. (1996) have recently resolved the radio emission from WR 146 into two components and suggested that the non-thermal component was associated with
a wind-collision region between the WC6 star and a companion star. From optical spectroscopy the companion has been
identied as late O-type, expected to have a much lower
mass-loss rate than the WC6 star. Consequently, the windcollision region and non-thermal source are expected to be
much closer to the O star than the WC star. The wide separation (D cos(i) 140 au) results in a low circumstellar
extinction along our sight-line to the non-thermal source
through the Wolf-Rayet wind.
Strong radio, infrared and X-ray emission from WR 147
(AS 431) was rst reported by Caillault et al. (1985), who
also identied it from a red spectrum as a heavily reddened
WN8 type star. The spectral type was conrmed by Conti
& Massey (1989). From the radio ux density Caillault et
al. derived an unusually high mass-loss rate ( 3:9 10?4
M y?1 ), assuming a distance of 2 kpc and wind terminal
velocity v1 = 2000 km s?1 . At the same time, they recognized the uncertainty of the nature of the radio emission:
its partial resolution suggesting a thermal source while its
negative spectral index that it was non-thermal.
This apparent contradiction was removed by highresolution, 5-GHz MERLIN observations of WR 147 which
demonstrated that the emission arose in two components.
One component was shown to be coincident with the optical image of the WN8 star and identied as its thermal wind
emission and the other, 000 :6 to the north, with the (possible)
source of non-thermal emission (Moran et al. 1989). From
multi-epoch, multi-frequency VLA observations, Churchwell
et al. (1992) conrmed that the northern component was a
non-thermal emitter with a 2{6-cm spectral index of ?0:5,
compared with a spectral index of +0:6 for the southern component, consistent with free-free emission from the
wind of the WN8 star. The two components were observed
to be joined by a bridge of emission at 15 GHz, indicating
that the sources were spatially related rather than being a
chance alignment of the Wolf-Rayet star with an unrelated
non-thermal source.
Churchwell et al. (1992) used a combination of optical and infrared photometry to re-determine the reddening
and distance to WR 147. Their new distance of 630 pc made
WR 147 the second closest WR star to the Sun. From considering the radio ux from the southern, stellar wind component, Churchwell et al. derived a mass-loss rate of 4:2 10?5
M y?1 , similar to those of other WR stars. Contreras et al.
(1996) found a similar value from a 43-GHz observation,
taken to be uncontaminated by non-thermal emission.
In his survey of X-ray emission from WR stars, Pollock (1987) determined a 0.2{4 keV X-ray luminosity of
(47 6) 1032 erg s?1 from EINSTEIN hard-band observations of WR 147, assuming a distance of 1.9 kpc. Re-scaled
to a distance of 630 pc, this becomes (5:2 0:7) 1032 erg
s?1 , or 0:13 0:02 L . The extinction derived by Churchwell
et al. implies a hydrogen column density NH 2:3 1022 ,
20 per cent higher than that adopted by Pollock, so that
derived luminosity will be higher also. The high extinction
may also account for the low ROSAT 0.2{2.4 keV X-ray luminosity of (0:58 0:31) 1032 erg s?1 reported by Pollock,
Haberl & Corcoran (1995), although variability cannot be
ruled out. Churchwell et al. examined several possible mechanisms for the X-ray emission. They found that only the
capture of a slow, dense, equatorially-enhanced wind from
the WN8 star by a neutron star could produce the observed
ux, since a colliding-wind massive-binary model appeared
to be ruled out by the speckle observations of Lortet et al.
(1987), who found no optical companion with m < 3 at
= 7000
A within 1.5 arcsec of the WN8 star. However,
Churchwell et al. could not rule out the possibility that the
X-rays were produced in a collision between the wind of the
WN star and that of an unseen companion.
We re-observed WR 147 in the radio to examine the
structure evident in the maps by Moran et al. and Churchwell et al., exploiting the higher resolution possible with
MERLIN following the commissioning of the Cambridge antenna. A preliminary reduction of these data was presented
by Davis et al. (1995). In addition, we extended the search
for a companion to the infrared given the severe interstellar
extinction suered by WR 147. We also observed the system
at mid-infrared and millimetre wavelengths to form a spectral energy distribution extending from the optical to the
radio wavelength regime.
Our new 2-m map shows the presence of an infrared
companion 3 mag fainter than the WN8 star and a little
beyond the non-thermal radio component. We take this as
evidence that WR 147 is a double star and discuss the system
as a colliding-wind binary. We compare the wind properties
of WR 147 with those of other WR systems showing nonthermal radio emission.
c 0000 RAS, MNRAS 000, 000{000
40 10 38.6
40 10 38.6
38.4
38.4
38.2
38.2
38.0
38.0
DECLINATION (B1950)
DECLINATION (B1950)
Radio and infrared structure of the colliding-wind Wolf-Rayet system WR 147
37.8
37.6
37.8
37.6
37.4
37.4
37.2
37.2
37.0
37.0
36.8
36.8
20 34 53.90
53.88
53.86
53.84
53.82
RIGHT ASCENSION (B1950)
53.80
3
20 34 53.90
53.88
53.86
53.84
53.82
RIGHT ASCENSION (B1950)
53.80
The nal deconvolved MERLIN images at (a) 5 GHz with a synthesized beam FWHM of 58 57 mas and (b) 1.6 GHz with
a beam of 178 172 mas. The beams are shown in the lower right-hand corners of the images. Both images are at the full resolution of
MERLIN (i.e. with no tapering) and are presented at identical scales. The 1 rms noise levels in the images are 89 and 135Jy beam?1
and the contour levels represent ?3; 3; 5; 7; 9; 11, and 3; 6; 9; 12; 15; 18; 20; 22 respectively. The negative contours are denoted by dashed
lines. The crosses in (b) indicate the positions of N and S observed in the 5-GHz image.
Figure 1.
Table 1.
Fluxes of the radio calibrators
Source
5 GHz 1.6 GHz
(Jy)
(Jy)
3C286
7.31a 13.64a
0552+398 6.65
2005+403 2.83
2.36
OQ208
1.08
a From Baars et al. (1977)
2 RADIO OBSERVATIONS
Radio observations were obtained using the MERLIN
array on 1992 June 24 (C band, 5 GHz) and 1993 April 11
(L band, 1.6 GHz) with nominal resolution of 50 mas and
150 mas respectively. The total on-source integration times
were 14.5 and 11 hours at 1.6 and 5 GHz respectively. To
establish the complex antennae gains, the on-source observations were interleaved with frequent observations of the
nearby radio-bright quasar 2005 + 403. The absolute ux
scale was established by observation of 3C286 and the unresolved sources OQ208 (1.6 GHz) and 0552+398 (5 GHz).
The uxes adopted for 3C286 and determined for the other
sources, including the phase-reference source 2005+403, are
given in Table 1.
The data were initially amplitude-calibrated using
MERLIN software, and then transferred to the NRAO aips
package for phase calibration and imaging. The \dirty" im-
c 0000 RAS, MNRAS 000, 000{000
ages were deconvolved using the clean algorithm (see Cornwell & Braun, 1989, and references therein), giving the nal
synthesized images shown in Fig. 1. The images presented
here have a resolution approximately three times that of
the data of Moran et al., which had a synthesized beam
FWHM of 150 130 mas. This is largely due to the fact that
the present observations were made after the Cambridge antenna was added to MERLIN, increasing both baseline coverage and the sensitivity on the longest baselines.
At 5 GHz, WR 147 is resolved into two components, N
and S, as previously observed by Moran et al. (1989) and
Churchwell et al. (1992). At 1.6 GHz, component N is easily
observed but, at the position of S, the radio emission is only
detected at around the 3 level. Moran et al. showed that the
position of WR 147S was coincident with the optical position
of the WR star.
We have convolved the 5-GHz data with a beam similar to that of Moran et al. to enhance low surface-brightness
emission (Fig. 2). It can be seen that there is low surfacebrightness emission extending to the North of N, and between N and S. This latter emission is the \bridge" of emission that Churchwell et al. (1992) observed at 15 GHz.
Our observation conrms the spatial connection between
WR 147N and S and rules out the possibility that N is a
background radio source in a chance alignment with S. As
yet, the nature of the bridge is not known.
Fluxes for the two components were derived by tting
2D-Gaussian functions to the emission features (aips routine
P.M. Williams et al.
Table 2.
LIN maps
Fluxes of the radio emission from WR147 from MER-
S (N)
S (S)
Beam
Method
00
(GHz) (mJy)
(mJy)
5
4:5 1:0 9:2 1:8
0:058 0:057 TVSTAT
5:9 0:2 11:4 0:3 0:15 0:13
JMFIT
6:8 0:1a 11:8 0:1a 0:3 0:3
JMFIT
1.6 11:4 0:6 ?
0:178 0:172
JMFIT
16:3 0:2b
22
JMFIT
a; b: the total ux detected with MERLIN at (a) 5 GHz, and (b)
1.6 GHz.
Positions (equinox B1950, epoch 1992.42) of the 5-GHz
components relative to the extragalactic radio frame, and source
sizes derived from model ts to the visibility data
Table 3.
RA
(0:00:01)
20h34m
53s:853
53s:860
0:00:070
a
Dec
x
ya
(0:00 :01) (mas)
(mas)
41100
3800 :00
267 27 79 10
3700 :43
170 14 253 18
0:00 :57
N
S
a FWHM; b Position angle, moving E from N.
PAb
( )
93 3
65 7
jmfit) in Fig 2. At the full 5-GHz resolution of MERLIN
(Fig. 1) this method gives a poor estimate of ux largely
because the radio emission from neither N nor S is well approximated by a Gaussian function. In addition, it is apparent from the many small patches of low surface brightness
emission in the 5-GHz image that some ux is resolved out
by MERLIN. For comparison, we have determined the uxes
for the features in Fig. 1 by integrating the ux over the area
occupied by each source (aips routine tvstat). In addition,
we also convolved the data with lower resolution beams to
estimate the total ux from WR 147 as detected by MERLIN. The uxes at 1.6 GHz were determined in a similar
fashion. These uxes are given in Table 2.
In Table 3 we give the positions derived from Fig. 2. The
error in the radio positions is dominated by the error in the
position of the radio phase-reference source relative to the
optical reference frame. For the MERLIN calibrator used
in these observations this is approximately 10 mas. The
separation of the two radio components is 575 15mas. This
compares very well with the separations 600 mas derived by
Moran et al. and 580 mas by Churchwell et al.
Both N and S are clearly resolved in our observations,
N being elongated with an apparent E{W orientation, and
S clearly not a symmetric, uniform emission region, as often
assumed a priori in calculations, for example, of mass-loss
rates in hot star winds. Indeed, WR 147S appears to have
a complex intensity distribution. However, since possibly as
much as half of the emission is resolved out by MERLIN
(see below), any detailed analysis of the intensity distribution of S requires the addition of contemporaneous lower
spatial frequency visibility data than observed with MERLIN. Meanwhile, we can estimate the size and aspect ratio
of the 5-GHz emission from WR 147N and S from a simple
analysis of the visibility data presented here. After locating
the peak of S at the phase-tracking centre of the image, we
40 10 38.6
38.4
38.2
DECLINATION (B1950)
4
38.0
37.8
37.6
37.4
37.2
37.0
36.8
20 34 53.90
53.88
53.86
53.84
53.82
RIGHT ASCENSION (B1950)
53.80
The 5-GHz data in Fig. 1 convolved with a 150 130
mas beam, similar to that of Moran et al. (1988) (see lower righthand corner) This increases the sensitivity to extended emission.
A bridge of emission between WR 147S and WR 147N is clearly
observed. The 1 rms of the image is 82Jy beam?1 and the contour levels are ?3; 3; 5; 7; 11; 15; 19; 23; 27; 32. Negative contours
are denoted by dashed lines.
Figure 2.
removed the ux contributed by N from the visibility data
using the aips routine uvsub. The remaining visibility data
were then binned into a 20 20 grid to improve the signal-tonoise ratio. A two-dimensional Gaussian t to the gridded
data then gives an estimate of the source size, ux density
and position angle. A similar procedure was followed for N.
The derived sizes are given in Table 3. From these results,
the aspect ratios for N and S are 3:4 0:5 and 1:5 0:15
respectively.
Also, in a long-term monitoring programme (van der
Hucht et al. 1995), three more 5-GHz observations were
taken with the WSRT in the rst half of 1993 with a 300 :7
synthesized beam and reduced using Westerbork's newstar
software. The ux densities, on the Baars et al. scale, are
consistent within the uncertainties and are averaged for a
\1993.3" ux density in Table 4.
Our MERLIN ux densities (Table 2) are signicantly
lower than our 1993 WSRT data and those derived by
Churchwell et al. and in other studies. These have yielded total 5-Ghz uxes 36 mJy (Table 4), usually from the VLA
in the compact C, C-D or D congurations. In these congurations, WR 147 is largely unresolved and these uxes
closely represent the total ux, from both compact and extended components of WR 147. The higher resolution Aconguration observations of Churchwell et al. gave slightly
lower ux densities, but the dierences are barely significant. Churchwell et al. also reported Westerbork observa-
c 0000 RAS, MNRAS 000, 000{000
Radio and infrared structure of the colliding-wind Wolf-Rayet system WR 147
Table 4.
5
Comparison of previous 5-GHz uxes from WR 147
Date N (mJy)
S (mJy) Total (mJy) Telescope Ref.
1984.3
35:3 0:13 VLA(C)
Abbott et al.
1984.5
38 1
VLA(C-D) Caillault et al.
1984.9 11:9 0:4a 21:4 0:4 33:3 0:6
VLA(A)
Churchwell et al.
1985.1 13:3 0:4a 19:4 0:4 32:7 0:6
VLA(A)
Churchwell et al.
1988.1 8:3 0:3
16:7 0:3 25:0 0:4
MERLIN Moran et al.b
1989.7
36:3 0:4
WSRT
Churchwell et al.
1990.4
33:8 1:2
WSRT
Churchwell et al.
1992.5 6:8 0:1
11:8 0:1 18:6 0:2
MERLIN this paper
1993.3
37:2 0:2
WSRT
this paper
1995.3
38:36 0:09 VLA(D)
Contreras et al.
a These uncertainties are estimates taking the background rms in the synthesized images and multiplying by estimates of the size of
the source relative to the beam size. b Using a ux density of 3.4 Jy for 2005+403 instead of 4.4 Jy used by Moran et al.
tions with a 300 :5-beam. We averaged the 1989 and 1990 6-cm
data to give ux densities for epochs \1989.7" and \1990.4".
Comparison with the total 5-GHz ux of 18:6 0:3
that we derive suggests two possibilities: either MERLIN
detected only half of the total emission of the source or the
source may have been signicantly fainter at the time of
our observations (1992) than at other times. Given that the
stellar wind component, S, is the principal contributor to
the 5-GHz ux from WR 147 (Fig. 3), the variability option
seems unlikely. A 50 per cent fall in the emission by a stellar wind implies a 40 per cent fall in the star's mass-loss
rate over a period of years followed by a recovery in 180
days, little more than the 130-day dynamical time-scale of
the wind. This cannot be ruled out but such a phenomenon
has not been observed from previous radio and extensive
infrared observations of Wolf-Rayet winds.
The alternative to variability is that a signicant fraction of the 5-GHz ux from WR 147 is from a more extended
component than observed by MERLIN, which is only capable of detecting emission on size scales up to about 700
mas at 5 GHz (on account of its shortest antenna spacing
being 6.4 km). Support for this view comes from the fact
that the ux detected by Moran et al. is rather lower than
that cited in their paper because they used too high a ux
for the calibrator, 2005+403. The corrected ux is given
in Table 4 and is also lower than those determined with
the VLA and WSRT, which are mutually consistent and
show no evidence for signicant variability. This is consistent with the small dispersion of the near-infrared magnitudes observed by Churchwell et al. in 1983 and 1990, and
agreement with those from Caillault et al., showing the ux
level from the WR 147 stellar wind to be constant to within
about 5 per cent. We therefore conclude that MERLIN resolves out approximately half the ux, presumably from an
extended component. It is not clear why the uxes from the
two MERLIN studies dier but, as Moran et al. did not give
details of how their uxes were calculated, we cannot take
this further at present. Therefore, we will rely on our MERLIN observations for discussion of structure but use uxes
from Churchwell et al. for the luminosities.
The shape of the non-thermal component, N, will be
discussed in Section 5 but as to that of S, we note that this
is the rst time an asymmetric geometry in a WR star wind
has been observed directly. Previously, the wind of 2 Velorum had been resolved at 5 GHz by Hogg (1985), who found
c 0000 RAS, MNRAS 000, 000{000
Table 5.
Mid-infrared photometry of WR 147.
Band M [8.75] [9.7] [11.6] [12.5] Q
mag. 2.85 2.47 2.77 2.15 1.75 1.31
that the ux density observed at a given VLA antenna separation was independent of the position angle of the baseline, implying a spherical wind. Polarimetric observations
suggest that a few WR stars have attened winds (SchulteLadbeck 1995) and, as noted above, White & Becker (1995)
inferred from the radio light curve of WR 140 that its wind
is attened. Theoretical studies of the winds of rotating
WR stars (Cassinelli, Ignace & Bjorkman 1995) show that
high-density wind-compressed zones (WCZs) can form in the
stars' equatorial planes. The density contrasts between the
equatorial and higher latitudes are signicantly smaller than
those of the wind-compressed disks (WCDs) proposed for Be
stars but can still reach 18 depending on the wind acceleration and stellar rotational velocity (Ignace, Cassinelli &
Bjorkman 1996). As the position angle of WR 147S diers
from that of the WR 147S{N system, it is unlikely that N
lies in the higher density equatorial plane of S unless both
systems are inclined at angles which combine to bring this
about. More probably, WR 147N lies at an intermediate latitude of WR 147S. Accordingly, in our calculations involving
the wind density of WR 147S, we will use the \spherical"
wind mass-loss rate derived conventionally but will note the
possible eects of departures from this.
3 SPECTRAL ENERGY DISTRIBUTION OF
WR 147
Churchwell et al. used optical and near-infrared photometry
to re-determine the reddening and distance to WR 147. We
also observed mid-infrared photometry in 1983 and use these
data, together with data from IRAS and sub-millimetre and
millimetre observations to extend the near-infrared spectralenergy distribution (SED) of WR 147 to longer wavelengths
for comparison with the radio uxes of the two components.
The mid-infrared observations were made using the
United Kingdom Infrared Telescope (UKIRT) on 1983 July
14 (the night after the rst set of near-IR data reported
by Churchwell et al.) with the common-user bolometer pho-
6
P.M. Williams et al.
Table 6.
Date
1991
1992
Sub-millimetre photometry of WR 147.
450m 800m 1.1mm 1.3mm 2.0mm
(Jy)
(mJy) (mJy) (mJy) (mJy)
1.20.2 35770 29215 28030 27530
36840 26330 28730 23050
tometer, UKT7. Both components of WR 147 were included
in the 5-arcsecond focal-plane aperture and we used the intermediate band-pass ( 1m) lters in the 8 { 13 m
region, together with the broad-band 4.8-m M and 20-m
Q lters, to help dene the SED. The magnitudes are given
in Table 5. The relative faintness of the 9:7m magnitude
compared with the other magnitudes in the 8 { 13 m region
is attributable to extinction by the \silicate" feature, consistent with the high reddening (AV = 11:5 mag, Churchwell
et al.) towards WR 147.
The mid-infrared data were strengthened and extended
using IRAS uxes (also observed in 1983) from Cohen
(1995), which were corrected for the spectral shape evident
from the ground-based data following the prescription in the
IRAS Explanatory Supplement (Beichman et al. 1988).
Our sub-millimetre and millimetre observations were
made with the common-user 3 He-cooled bolometer UKT14
(Duncan et al. 1990) on the James Clerk Maxwell Telescope
(JCMT) on 1991 March 21{23 and 1992 May 5{6. The beam
size is dependent on wavelength but was always suciently
large to include both components of WR 147. Flux calibration was by observation of Uranus and secondary standards
(Sandell 1994). For the relative extinctions at dierent wavelengths, we followed Stevens & Robson (1994). The uxes
are given in Table 6.
These optical{IR data were de-reddened by AV =
11:5 mag and are plotted in Fig. 3. It is evident that
the de-reddened mid-IR{mm uxes and the radio uxes of
WR 147S t a power-law SED, S / , characteristic of a
stellar wind. This index of this spectrum ( = 0:66 0:02
by least squares) is slightly greater than that (0.6) predicted
(e.g. Wright & Barlow 1975) for an isothermal, spherically
symmetric, steady-state wind but less than those ( = 0:78)
observed for the WC systems WR 140 (Williams 1996) and
WR 146 (Dougherty et al.). The higher spectral indices are
due to the ion density falling o more rapidly than r2 . This
could be caused by a number of factors: the failure of the
winds to have reached their terminal velocities at the radiocontinuum forming regions, the ionization fraction changing
with radius (cf. V el, Williams et al. 1990b, Leitherer &
Robert 1991) or a geometry that is changing with radius.
The 450-m ux lies about 3 above the power-law wind
spectrum but the observed ux came from one night only
and the high, and variable, atmospheric extinction at this
wavelength makes calibration dicult
The spectrum of WR 147N is very dierent. Plotted in
Fig. 3 is the 5{15-GHz spectrum observed by Churchwell
et al. in 1984, which has a spectral index ?0:4. They
did not resolve the components at 1.45 GHz but, from subtracting the extrapolated wind spectrum from the total ux,
estimated a 1.45-GHz ux for N which suggested a slightly
atter spectrum longward of 5 GHz. From the relatively low
upper limit to the 22-GHz ux found by Churchwell et al.
Spectral energy distribution (SED) of WR 147 from
de-reddened optical (4), ground-based infrared () and IRAS (2)
data. Our JCMT data are marked ?, the 250-GHz ux given by
Altenho, Thum & Wendker (1994) is marked 3 and the 43-GHz
ux from Contreras et al. (1996) is marked | all including both
components of WR 147. Radio data for the separate components
from Churchwell et al. are marked and from this paper (N only)
?. The shift between the uxes of N from Churchwell et al. and
the present study (see text) is evident. The dashed line has the
form S / 0:66 , appropriate for a stellar wind spectrum.
Figure 3.
in 1985, there appears to be a sharp high-frequency cut-o
which needs conrmation and examination. The integrated
spectrum of N gives a non-thermal luminosity of 7 1028
erg s?1 . Also plotted are our 5 and 1.65-GHz uxes of N.
As described above, the dierences between our uxes and
those of Churchwell et al. are most probably attributable to
MERLIN's not detecting extended emission. Consequently,
the agreement between the spectral index ({0.45) indicated
by our 1.6 and 5-GHz observations and that by Churchwell
et al. is fortuitous.
4 THE INFRARED IMAGE AND COMPANION
Infrared observations of WR 147 were obtained in the Service Observing Programme of the United Kingdom Infrared
Telescope (UKIRT) on 1996 May 28 using the IRCAM3 camera and ALICE array control system (Aspin et al. 1997).
The camera was congured to give an image scale of 000 :057
pixel?1 and operated in shift-and-add mode. Each observa-
c 0000 RAS, MNRAS 000, 000{000
Radio and infrared structure of the colliding-wind Wolf-Rayet system WR 147
7
40 10 38.6
38.4
WR 147
(K)
38.2
E
NIR
DECLINATION
38.0
37.8
37.6
3".
65
N
37.4
37.2
Near-infrared (2m, K band) image of WR 147 showing the position of a companion source (NIR) to the WN8 star.
Each pixel is 57 57 mas2 .
37.0
Figure 4.
36.8
20 34 53.90
tion is broken up into a large number of very short integrations. Provided that the eld contains one, dominant point
source, the centroid of this source on the array can be located for each integration and the images shifted in real time
to bring them into alignment. This system is very well suited
to searching for faint companions to bright point sources |
such as the WN8 star in WR 147. Two such observations
through the K lter, each comprising 10 000 integrations of
36 ms duration, were added to provide the \raw" image in
Fig. 4.
In addition to the clearly observed WR star, it is evident from Fig. 4 that there is a faint infrared companion
located slightly to the north. The slight E{W asymmetry in
the halo of the image of the WR star is a consequence of
the alignment of the telescope optics and was observed in
our other images taken during our programme. We used the
image of a single star to dene the telescope point-spread
function and applied the Starlink implementation (mem2d)
of the maximum entropy deconvolution algorithm to our image of WR 147 to give the deconvolved image in Fig. 5,
shown superimposed on the 5-GHz MERLIN observation.
The FWHM of the deconvolved image of the WN8 star is 2
pixel (000 :11) in both coordinates whereas the image of the
companion appears to be slightly broader E-W. As this is
similar to the resolution of the telescope and may well be an
artefact, we do not regard this elongation as signicant and,
for the present, consider the companion to be point like.
The infrared companion is 11:0 0:2 pixels (000 :63 00
0 :01) North and 1:7 0:3 pixels (000 :10 000 :02) West of the
WN8 star, giving a total separation of 000 :64 000 :02, which
corresponds to 403 13 AU at a distance of 630 pc. This
places the companion, hereafter referred to as WR147NIR,
slightly more distant from the WN8 star than the peak of
the non-thermal emission (cf. Table 3) in both coordinates.
If the southern, thermal radio component S is centred on
the infrared image of the WN8 star, the peak of the non-
c 0000 RAS, MNRAS 000, 000{000
53.88
53.86
53.84
RIGHT ASCENSION
53.82
Maximum-entropy reconstruction of the 2:2m image
of WR 147, re-gridded to 15-mas pixels and superimposed on the
5-GHz image from Fig. 1a. The IR image of the WR star has been
aligned with the position of WR 147S. The FWHM of WR 147S
in the IR image is 000:11, the diraction limit of UKIRT at this
wavelength. WR 147NIR is clearly visible to be slightly to the
north of WR 147N. The apparent IR emission to the SE of the
WR star is an artefact; a very similar feature appeared in the
reconstruction of another source observed at the same time.
Figure 5.
thermal emission is not coincident with WR147NIR, but lies
between the two stars (Fig. 5).
To get an idea of what type of star WR147NIR could be,
we estimate its absolute magnitude from the magnitude difference (K = 3:04 0:09) derived from our image and the
absolute magnitude of the WN8 star. We retain the distance
(630 pc) and absolute magnitude determined by Churchwell
et al. based on comparative infrared photometry of WR 147
and WR 105 despite the fact that the latter has recently
been re-classied as a WN9h star by Smith, Shara & Moffat (1996). The only WN8 star which might have a distance
and absolute magnitude determinable from membership of
a stellar association is WR 66 (HD 134877), which lies in the
direction of the association Anon Cir (Lundstrom & Stenholm 1984). The colour-magnitude diagram of this region
(Bassino et al. 1982) shows possible groupings at two different distances and as there is no external evidence that
WR 66 belongs to either, this does not provide a luminosity
calibration for WN8 stars. In the absence of a better calibration, we follow Churchwell et al. in adopting an absolute
magnitude K = ?5:96 for WR 147, which gives an absolute
magnitude of K = ?2:9 for WR147NIR.
We then determined the average absolute K magnitudes of hot stars of dierent spectral type from the MV
given by Underhill & Doazan (1982) and intrinsic (V ? K )
colours from Koornneef (1983). This gives MK = ?3:0 for
8
P.M. Williams et al.
B0V, ?2:9 for B0.5V and ?2:7 for B1V. We therefore assume WR147NIR to have the properties of a B0.5V star.
Of course, this is not a spectroscopic classication and we
recognize that the uncertainty in the absolute magnitude of
the WN8 star implies that if WR147NIR is a main-sequence
star, its type could lie anywhere between late-O and mid-B
type. Given that the visual (V ) magnitude dierence between the primary and secondary is also 3 magnitudes
for a B0.5V companion, Lortet et al. (1987) were unlucky
that the brightness of WR147NIR was at the limit of their
speckle observations.
It is also possible that the MK = ?2:9 star is a latetype star: about K7III if a giant, an earlier type if a luminosity class II star, but not a supergiant (too faint) or
main-sequence star (too bright). If WR147NIR is a K-type
giant, the visible magnitude dierence between the two stars
would be greater than 3 mag. However, if the companion
were coeval with the WN8 star, it is dicult to see how the
secondary could have evolved on to the giant branch in the
lifetime of the WR star primary. Therefore, we follow our
hypothesis that WR147NIR is a B0.5V star.
5 DISCUSSION
5.1 WR 147 as a colliding wind binary
The discovery of an infrared stellar companion to the WN8
star leads us to consider the possibility that WR 147 is
a colliding-wind binary (CWB) system, where the nonthermal emission arises in a wind-collision region, as inferred
in WR 140 and WR 146.
First we address the question whether a B0.5 main sequence star could have a substantial stellar wind. The onset
of mass loss, observable in the UV, by stellar winds from
stars on the main sequence is quite sudden at Te = 27 500
K, log L/L = 4:4 0:1 (Grigsby & Morrison 1995). From
Underhill & Doazan (1982), we see that the eective temperature of a B0.5V star is 29 000 K. From the relation
between bolometric correction and temperature (B:C: =
?0:5 ? 0:08T=103 ) given by Howarth & Prinja (1989), we
estimate a bolometric correction of ?2:8 and hence a bolometric magnitude of ?6:5 for WR147NIR. The corresponding luminosity is log L/L ' 4:5. From this, it is evident
that WR147NIR is just suciently hot and luminous to
have a substantial stellar wind which can collide with that
of the WN8 star. However, we should not lose sight of the
dependence of this result on the adopted luminosity of the
WN8 star: WR147NIR has a luminosity close to the borderline for stellar wind, and a small uncertainty in luminosity
translates into a large uncertainty in mass-loss rate. In any
event, we should expect the mass-loss rate of WR147NIR
to be much lower than that of the WN8 star: Runacres &
Blomme (1996) derive mass-loss rates of 5 ? 6 10?8M y?1
for the B0.5IV and B0.2V stars Lep and Sco, which also
have luminosities log L/L 4:5.
The winds of the WN8 and B0.5 stars are expected
to collide where their momenta balance, forming a contact
discontinuity. This is expected to occur relatively close to
the B0.5 star owing its lower mass-loss rate. The scale of
the wind-interaction region is determined by the distance
rOB of the contact discontinuity from the B0.5 star (Eichler
N
p r OB
B
r OB
D
WN8
Figure 6. Sketch of WN and B star components showing the location of the shocked gas in the wind-interaction region (shaded)
on either side of the contact discontinuity (solid line). The intersection of the contact discontinuity and the line between the stars
is referred to as the stagnation point.
& Usov 1993). The interaction region (Fig. 6) has the form
of a cap of height rOB and diameter rOB centred on the
B0.5 star, facing the WN8 star, and a cone facing away
from the WN8 star. If the non-thermal emission is produced
by electrons accelerated here, we would expect the source to
resemble the interaction region both in its proximity to the
B0.5 companion and in its shape | elongated perpendicular
to the line between the components with an aspect ratio ,
depending on the inclination and extent to which emission
arises \behind" the B0.5 star. Our observations (Table 3 and
Fig. 5) t well in both respects, giving support to the CWB
interpretation of the non-thermal emission.
We therefore estimate the scale of the interaction region by relating the observed width of the source 267 mas
to the cross-section rOB . This would give rOB 85 mas,
an overestimate if the emission arises from \outside" the
contact discontinuity. From the dierence between the IR
and radio separations, we have the projected value of rOB ,
rOB cos(i) = 60 30 mas. This value of rOB is quite uncertain since it comes from the dierence of two much larger
quantities and we do not know which part of WR 147N coincides with the position of the contact discontinuity. It also
depends on the (unknown) inclination but is consistent with
the value of rOB estimated from the width of the source, so
that the 50 per cent uncertainty is pessimistic. For further
discussion of the system, we will adopt rOB = 80 mas and
rOB cos(i) = 60 mas, recognizing their uncertainty.
The scale of the system and position of the contact discontinuity are related to the separation of the two stars, D,
through the ratio of wind momenta, , by
1=2
rOB = 1 + 1=2 D:
(1)
c 0000 RAS, MNRAS 000, 000{000
Radio and infrared structure of the colliding-wind Wolf-Rayet system WR 147
The infrared image gives the projected value of D, Dcos(i) =
635 20 mas, directly. Comparing this with rOB cos(i) =
60 30 mas derived above gives rOB =D 0:09 0:05, and
0:011+016
?009 , a value that is independent of the inclination
angle of the system.
For spherical winds, the wind-momentum ratio is given
_ 1 )OB = (Mv
_ 1 )WN . Here, we will use our obby = (Mv
served value of and the wind momentum of the WN8 star
to estimate that of the B0.5 star. Churchwell et al. derived
a wind velocity of 900 km s?1 for the WN8 star from the
displacement of the absorption component of the 2.058-m
He i line and used this, together with the ux from the southern component and the composition of the wind to derive
a mass-loss rate of 4:2 10?5 M y?1 . Eenens & Williams
(1994) derived a higher wind velocity of 1100 km s?1 by
tting the P Cygni prole of the 1.083-m He i line and
van der Hucht et al. (1996) derived a terminal velocity of
960 km s?1 from the widths of the [Ne III], [S IV] and [Ca
IV] ne-structure lines observed with ISO. We adopt 1000
km s?1 and revise the mass-loss rate of the WN8 star to
4:6 10?5 M y?1 .
To convert the wind momentum of the B0.5 star to a
mass-loss rate, we adopt a terminal velocity of 800 km s?1 ,
slightly greater than the escape velocity (Prinja 1989). Using these parameters and the value of the momentum ratio, we estimate a mass-loss rate of 6 10?7 M y?1 for
the B0.5 star. The formal uncertainty on gives a range of
(1 ? 15) 10?7 but the concurrence of the two estimates
of rOB suggests a narrower range. If WR147NIR and interaction region lie in the equatorial wind-compressed zone
(WCZ) of the WN8 star, the mass-loss rate derived for the
B0.5 star should be scaled up by a factor of a few, depending on the rotational velocity of the WN8 star and the wind
acceleration (cf. Ignace et al. 1996). Similarly, if WR147NIR
lies at a high latitude on the WN8 star, the mass-loss rate
derived for the B0.5 star will be lower. However, as it is
more likely that the interaction region lies at an intermediate latitude on the WN8 star, we will retain the mass-loss
rate derived. It is rather higher than the mass-loss rates for
the B0.2V and B0.5IV stars Sco and Lep (Runacares &
Blomme 1996) and comparable to those of the B0e stars
Cas and o Pup (Waters, Cote & Lamers 1987) determined from IR observations | although the luminosities
given for these B0e stars are about twice that we adopted
for WR147NIR.
5.2 X-rays from the colliding winds
We can now ask how much of the X-ray luminosity observed with EINSTEIN (0:13 0:02 L ) could be provided
by the colliding winds. Certainly, there is ample energy in
the winds. The fraction of the WN8 stellar wind colliding
with the interaction region is 0:006 for a spherically symmetric WN8 wind. This corresponds to 21L . Assuming
about one-third of the B0.5 stellar wind to be involved in
the interaction, we would have an additional 11L of wind
luminosity. These luminositities would be factors of a few
higher or lower if the wind of the WN8 star were attened
and the interaction region lay at equatorial or high latitudes
of the WN8 star respectively.
We consider two mechanisms for the production of Xray photons: thermal bremsstrahlung from the shock-heated
c 0000 RAS, MNRAS 000, 000{000
9
gas and inverse Compton scattering of stellar photons by the
accelerated electrons. Following Usov (1992), the thermal Xray luminosity is calculated from the wind parameters and
source size derived above. The wind from the WN8 star can
be shown to be heated to 4 107 K near the stagnation
point. For the shocked WN8 wind, external to the contact
discontinuity, we get a thermal X-ray luminosity of 1:4 1031
erg s?1 . A similar calculation for the B0.5 wind (the \inner
shock") yields 2:2 1032 erg s?1 : The total thermal X-ray
luminosity is therefore expected to be LX = 0:06L , which
is about half of the observed luminosity. This result is very
sensitive to the mass-loss rates, varying approximately as
(M_ OB )1:5 (M_ WN )0:5 (Usov 1992).
The non-thermal radio luminosity 7 1029 erg s?1 derived in Section 3 may be considered to arise from a source
of radius (=2)rOB mas and lling factor 0.1 (to allow for
its hollow bowl shape). This yields an equipartition magnetic eld strength Beq 9 mG. Though the source size is
uncertain, and Beq / R?6=7 , the minimum magnetic eld
strength has to be 1 mG, otherwise the Razin -Tsytovich
eect (e.g. Pacholczyk 1970) will suppress 1.6-GHz emission
from WR 147 N. For a eld of a few milli-Gauss, the radio
luminosity requires a total energy in relativistic electrons of
5 1039 ergs.
Can these electrons scatter the UV photons of the B0.5
star to produce the observed X-ray ux? The nal energy,
Ef of an inverse-Compton scattered photon is related to its
initial energy, Ei , by Ef = 2 Ei where is the Lorenz factor. At a stellar temperature of 30; 000 K, the peak in the
stellar photon distribution occurs at about 10 eV. Therefore, inverse-Compton X-ray photons at energies near 1 keV
arise primarily from scattering o electrons with = 10, or
with energies of 8 10?6 erg. A relativistic electron with
energy E , created at a distance r from a source of photons
with luminosity L will have a lifetime to inverse-Compton
scattering of
2
t = 4r c ;
(2)
IC
aIC L E
where aIC = 3:97 10?2 . Since WR 147N is at least 8 1014
cm from the B0.5 star, whose luminosity is 1:8 1038
ergs s?1 , electrons with energy 10?5 ergs have tIC 1010
sec. For a total electron energy 5 1039 , the inverseCompton X-ray luminosity is 1030 ergs s?1 . Clearly,
inverse-Compton scattering of non-thermal electrons do not
contribute signicantly to the observed 1-keV X-ray luminosity
The dierence between the observed total X-ray luminosity and the theoretical estimate of that from the collision
region can be accounted for by X-rays from the two stars.
Pollock (1987) found that the mean X-ray luminosity of single WN stars is 0:06L . In addition, the luminosity of the
B0.5 star could be of similar magnitude, as seen by comparison with 0:007L for the B0.5V star Car, and up to
0:07L , depending on the model, for the B0V star Sco
(Cassinelli et al. 1994).
5.3 Comparison with other non-thermal emitting
WR systems
Our conclusion that WR 147 is a CWB non-thermal radio source invites comparison with the three other WR sys-
10
P.M. Williams et al.
Table 7.
Properties of WR systems exhibiting non-thermal emission.
_ 2 )WR Reference
(obs) Size Dist.
L5GHz
M_ WR
v1
(Mv
WR
(mJy)
(au) (kpc) (erg s?1 ) (M y?1 ) kms?1 (erg s?1 )
WC7+O9
1.5
42 4.7 3:9 1022 5.7 10?5 2700 1:3 1038 Williams et al. 1992
WC7+O4-5
22.5
27
1.3 4:6 1022 5.7 10?5 2860 1:5 1038 Williams et al. 1990a
WC6+O9
28.5
139 1.1 4:1 1022 3 10?5
2900 8:0 1037 Dougherty et al. 1996
WN8+B0.5
12
378 0.63 1:0 1022 4.6 10?5 1000 1:5 1037 this paper
a 5-Ghz ux of the non-thermal component; using the data of Churchwell et al. for WR 147
WR Spectra
125
140
146
147
Sa
tems whose non-thermal radio emission has been ascribed
to stellar wind collision. Some properties of the four systems
are collected and compared in Table 7. The spectral types of
the companions to the WR stars are uncertain. Apart from
that of WR 147 inferred from its infrared luminosity above,
they come from composite spectra dominated by the emission lines of the WR stars which mask most of the lines used
for the classication of OB type spectra.
The quantities listed as \sizes" are estimates of the separations of the WR stars and non-thermal sources. The separations come from the observed projected separations of
the sources in WR 146 (Dougherty et al.) and WR 147 (this
paper) and are therefore lower limits. The size of WR 140 is
from the orbit determination of Williams et al. (1990a) and
is the separation at the phase ( 0:7) of radio maximum.
In the case of WR 125, we estimate a lower limit to the orbital size from the inference, from the infrared variations,
that the period of WR 125 is at least twice that of WR 140
(Williams 1996).
For comparison, the radii of the 5-GHz \photospheres",
the wind radii from which the 5-GHz continua of the winds
of the WR stars may be considered to arise (Wright & Barlow 1975) are 36 au for the three WC stars and 75
au for WR 147. For systems smaller than this radius e.g.
WR 140, the 5-GHz non-thermal emission is observable only
if the wind of the WR star is not spherically symmetric and
only when the geometry is favourable. On the other hand,
the non-thermal emission from widely separated systems like
WR 146 and WR 147 is readily observable because the nonthermal emission arises further out in the WR winds, close
to the companions.
The 5-GHz ux densities of the non-thermal components in WR 125 and WR 140 are those at radio maxima,
while those of WR 146 and WR 147 are average values, excluding the MERLIN data for the latter. The most striking
dierence between WR 147 and the other systems in Table
7 is the relative weakness of its non-thermal emission. The
5-GHz luminosities of the three WC systems are similar, and
all four times stronger than that of WR 147.
Given that WR 147 is the most widely separated system and that the projected separation of the components
is about ve times greater than the 5-GHz \photosphere",
we cannot ascribe this dierence to greater circumstellar extinction in the WN8 wind of WR 147. The dierence must
lie in the intrinsic luminosities of the non-thermal sources.
The most signicant dierences in the other properties
of the systems are in the wind terminal velocities and kinetic
energies. The terminal velocities are well determined observationally and that of WR 147 is a factor of 2:8 times lower
than those of the three other systems. The kinetic energy is
a factor 8 lower for a spherically symmetric WR 147 wind,
with the possible modications referred to above if the wind
is attened. We expect the available WR wind luminosity to
scale as the product of wind energy and angular cross-section
of the interaction region. The last scales as (rOB = D)2 , i.e.
_ 1 )OB = (Mv
_ 1 )WN for small (eqn 1).
approximately as (Mv
This suggests that the non-thermal luminosity should scale
approximately as the product of the WR wind velocity and
the OB wind momentum for equal eciencies of conversion
of wind luminosity.
Within the uncertainties, this accounts for the relatively low non-thermal luminosity of WR 147 relative to
WR 125 and WR 146, assuming the wind velocities of the
late-subtype O stars in the latter systems to be in the range
1000 { 2000 km s?1 (Prinja, Barlow & Howarth 1990). The
companion to the WC7 star in WR 140 is of an earlier subtype (O4-5) whose velocity has been measured (3100 km s?1 ,
Fitzpatrick, Savage & Sitko, 1982) to be rather higher than
those expected for the other systems but the non-thermal luminosity of WR 140 is not proportionally higher. This may
be a consequence of the non-spherical geometry of this system (cf. White & Becker 1995) and this source awaits detailed modelling.
It appears from the similarity of the non-thermal luminosities of the three WC-star based systems that the separation of the components is unimportant, which might seem
surprising for interacting systems. The fact that we have
only lower limits to the separations of the more widely separated systems reinforces this result. The explanation lies
in the fact that, for a given wind-momentum ratio, the size
of the interaction region scales with the separation of the
components (eqn 1) and intercepts the same fraction of the
Wolf-Rayet wind and hence the same kinetic energy if the
wind has accelerated to its terminal velocity.
6 CONCLUSIONS
High-resolution infrared and radio images of the X-ray luminous Wolf-Rayet system WR 147 (AS 431) have been
observed. The new radio image resolves both components
of this N{S double source and conrms the presence of a
\bridge" of emission joining them. The emission from the
Southern component, which is associated with the wind of
the WN8 star, is not circular but sized 170 253 mas,
suggesting that the stellar wind is not spherically symmetric. Almost half the ux is resolved out by MERLIN and the
data need augmentation with contemporaneous observations
on shorter baselines to observe the extended emission. The
non-thermal component, 000 :6 north of the WN8 star, is
broadened East-West by 267 mas. The infrared image re-
c 0000 RAS, MNRAS 000, 000{000
Radio and infrared structure of the colliding-wind Wolf-Rayet system WR 147
veals the presence of a companion to the WN8 star, close to
the non-thermal radio source but slightly ( 60 mas) more
distant from the WN8 star. The companion is K 3 mag.
fainter than the WN8 star and has the luminosity of a B0.5V
star, just suciently luminous (log L/L = 4:5) to possess
a stellar wind capable of colliding with that of the WN8
star and allowing us to treat the system as a colliding wind
binary.
The presence of the non-thermal emission between the
two stars, much closer to that with the weaker wind and
having a shape consistent with that of a wind-interaction
region is direct evidence for a colliding-wind origin for the
emission. About half of the X-ray emission can also be accounted for by the release of energy in the wind collision.
Comparison of the non-thermal luminosity with those
of the three non-thermal source systems incorporating WC
type stars (WR 125, WR 140 and WR 146) shows a correlation with the velocity of the WR wind.
ACKNOWLEDGEMENTS
It is a pleasure to thank Mark Casali for taking the UKIRT
images, and Ger de Bruyn, Andy Pollock, Russ Taylor and
Vladimir Usov for stimulating discussions and correspondence. The Multi-Element Radio-Linked Interferometer Network (MERLIN) at Jodrell Bank is operated by the University of Manchester for the Particle Physics and Astronomy Research Council (PPARC). The United Kingdom Infrared Telescope (UKIRT) and James Clerk Maxwell Telescope (JCMT) on Mauna Kea, Hawaii, are operated by the
Joint Astronomy Centre for PPARC. The Westerbork Synthesis Radio Telescope (WSRT) is operated by the Netherlands Foundation for Research in Astronomy with nancial
support from the Netherlands Organization for Scientic Research (NWO).
REFERENCES
Abbott, D.C., Bieging, J.H., Churchwell, E., Torres, A.V., 1986,
ApJ, 303, 239.
Altenho, W.J., Thum, C., Wendker, H.J., 1994, A&A 281, 161
Aspin, C.A., Puxley, P.J.P., Hawarden, T.G., Patterson, M.J.,
Pickup, D.A., 1997, MNRAS in Press
Baars, J.W.M., Genzel, R., Pauliny-Toth, I.I.K., Witzel, A., 1977,
A&A 61, 99
Bassino, L.P., Dessaunet, V.H., Muzzio, J.C., Waldhausen, S.,
1982, MNRAS, 201, 885
Beichman, C.A., Neugebauer, G., Habing, H.J., Clegg, P.E.,
Chester, T.J., 1988, Infrared Astronomical Satellite (IRAS),
Catalogs and Atlases, Explanatory Supplement, NASA RP1190, Vol. 1
Caillault, J-P., Chanan, G.A., Heland, D.J., Patterson, J.,
Nousek, J.A., Takalo, L.O., Bothun, G.O., Becker, R.H., 1985,
Nature, 313, 376
Cassinelli, J.P., Cohen, D.H., MacFarlane, J.J., Sanders, W.T.,
Welsh, B.Y., 1994, ApJ, 421, 705
Cassinelli, J.P., Ignace, R., Bjorkman, J.E., 1995, in van der
Hucht, K.A., Williams, P.M., eds, Wolf-Rayet Stars: Binaries,
Colliding Winds, Evolution. Kluwer, Dordrecht, p. 191
Chen, W., White, R.L., 1994, ApSS, 221, 259
Churchwell, E., Bieging, J.H., van der Hucht, K.A., Williams,
P.M., Spoelstra, T.A.Th., Abbott, D.C., 1992, ApJ, 393, 329
Cohen, M., 1995, ApJS, 100, 413
c 0000 RAS, MNRAS 000, 000{000
11
Conti, P.S., Massey, P., 1989, ApJ, 337, 251
Contreras, M.E., Rodrguez, L.F., Gomez, Y., Velazquez, A.,
1996, ApJ, 469, 329
Cornwell T.J., Brown R, 1989, in Perley, R.A., Schwab, F.R.,
Bridle, A.H., eds, ASP Conference Series 6, p. 167
Davis, R.J., Bode, M.F., van der Hucht, K.A., Williams, P.M.,
1995, in: van der Hucht, K.A., Williams, P.M., eds, WolfRayet Stars: Binaries, Colliding Winds, Evolution. Kluwer,
Dordrecht, p. 547
Dougherty, S.M., Williams, P.M., van der Hucht, K.A., Bode, M.,
Davis, R.J., 1996, MNRAS, 280, 963
Duncan, W.D., Robson, E.I., Ade, P.A.R., Grin, M.J., Sandell,
G., 1990, MNRAS, 243, 126
Eenens, P.R.J., Williams, P.M., 1994, MNRAS, 269, 1082
Eichler, D., Usov, V., 1993, ApJ, 402, 271
Fitzpatrick, E.L., Savage, B.D., Sitko, M.L., 1982, ApJ, 256, 578
Grigsby, J.A., Morrison, N.D. 1995, ApJ, 442, 794
Hogg, D.E., 1985, in Hjellming, R.M., Gibson, D.M., eds, Radio
Stars, D. Reidel, Dordrecht, p. 117
Howarth, I.D., Prinja, R.K. 1989, ApJS, 69, 527
Ignace, R., Cassinelli, J.P., Bjorkman, J.E., 1996, ApJ, 459, 671
Jardine, M., Allen, H.R., Pollock, A.M.T., 1996, A&A, 314, 594
Koornneef, J. 1983, A&A, 128, 84
Leitherer, C., Robert, C., 1991, ApJ, 377, 629
Lortet, M.C., Blazit, A., Bonneau, D., Foy, R., 1987, A&A, 180,
111
Lundstrom, I., Stenholm, B., 1984, A&AS, 58, 163
Moran, J.P., Davis, R.J., Bode, M.F., Taylor, A.R., Spencer,R.E.,
Argue, A.N., Irwin, M.J., Shanklin, J.D., 1989, Nature, 340,
449
Pacholczyk, A.G., 1970, Radio astrophysics, Freeman, San Francisco
Pollock, A.M.T., 1987, ApJ, 320, 283
Pollock, A.M.T., Haberl, F., Corcoran, M.F., 1995, in van der
Hucht, K.A., Williams, P.M., eds, Wolf-Rayet Stars: Binaries,
Colliding Winds, Evolution. Kluwer, Dordrecht, p. 512
Prinja, R.K., 1989, MNRAS, 241, 721
Prinja, R.K., Barlow, M.J., Howarth, I.D., 1990, ApJ, 361, 607
Runacres, M.C., Blomme R., 1996 A&A, 309, 544
Sandell, G., 1994, MNRAS 271, 75
Schulte-Ladbeck, R.E., 1995, in van der Hucht, K.A., Williams,
P.M., eds, Wolf-Rayet Stars: Binaries, Colliding Winds, Evolution. Kluwer, Dordrecht, p. 176
Setia Gunawan, D.Y.A., van der Hucht, K.A., de Bruyn, A.G.,
Williams, P.M., 1996, in Vreux, J.-M., Detal, A., Fraipont, D.,
Gosset, E., Rauw, G., eds, Wolf-Rayet stars in the framework
of stellar evolution, Proceedings of the 33rd Liege International Astrophysical Colloquium, (in Press)
Smith, L.F., Shara, M.M., Moat, A.F.J., 1996, MNRAS, 281,
163
Stevens, J.A., Robson, E.I., 1994, MNRAS, 270, L75
Underhill, A., Doazan, V., 1982, B Stars with and without Emission Lines, NASA SP-456, Washington, NASA
Usov, V.V., 1992, ApJ, 389, 635
van der Hucht, K.A., Williams, P.M., Spoelstra, T.A.Th., Swaanenvelt, J.P., 1995, in van der Hucht, K.A., Williams, P.M.,
eds, Wolf-Rayet Stars: Binaries, Colliding Winds, Evolution.
Kluwer, Dordrecht, p. 559
van der Hucht, K.A., et al., 1996, A&A, 315, L193
Waters L.B.F.M., Cote J., Lamers H.J.G.L.M. 1987, A&A, 185,
206
White, R.L., Becker, R.H., 1995, ApJ, 289, 698
Williams, P.M., 1996, in Vreux, J.-M., Detal, A., Fraipont, D.,
Gosset, E., Rauw, G., eds, Wolf-Rayet stars in the framework
of stellar evolution, Proceedings of the 33rd Liege International Astrophysical Colloquium, (in Press)
Williams, P.M., Eenens, P.R.J., 1989, MNRAS, 240, 445
Williams, P.M., van der Hucht, K.A., Pollock, A.M.T.,
12
P.M. Williams et al.
Florkowski, D.R., van der Woerd, H., Wamsteker, W.M.,
1990a, MNRAS, 243, 662
Williams, P.M., van der Hucht, K.A., Sandell, G., The, P.S.,
1990b, MNRAS, 244, 101
Williams, P.M., van der Hucht, K.A., Bouchet, P., Spoelstra, T.A.Th., Eenens, P.R.J., Geballe, T.R., Kidger, M.R.,
Churchwell, E., 1992, MNRAS 258, 461
Williams, P.M., van der Hucht, K.A., Spoelstra, T.A.Th., 1994,
A&A, 291, 805
Williams, P.M., van der Hucht, K.A., Kidger, M.R., Geballe,
T.R., Bouchet, P., 1994, MNRAS, 266, 247
Wright, A.E., Barlow, M.J., 1975, MNRAS, 170, 41
c 0000 RAS, MNRAS 000, 000{000