HST NICMOS Observations of Circumstellar Matter Around CYG X-3

W. Schmutz Institute of Astronomy, ETH Zentrum, CH-8092 Zürich, Switzerland

W. D. Vacca, L. Close, and J. Rayner Institute for Astronomy, Honolulu, Hl 96822

T. R. Geballe Joint Astronomy Centre, Hilo, Hl 96720, USA

H. Schild and R. Walder Institute of Astronomy, Zürich, Switzerland

             

 

Abstract:

Models of the evolution of massive binaries predict that only a few such objects should survive the common envelope phase and result in systems containing a compact object plus a Wolf-Rayet star (c+WR). According to these models, a vast amount of stellar material is lost during the common envelope phase prior to the c+WR phase. We are attempting to test these models by searching for the presence of such circumbinary material around the only known c+WR object: Cyg X-3. Since the large reddening towards Cyg X-3 prevents any investigation in the optical, deep infrared exposures of Cyg X-3 with the NICMOS NIC-2 camera aboard HST and the Pa$\alpha$ narrow band filter have been used to image the system at high spatial resolution.^21.1

The HST images have only recently been acquired and their analysis is currently underway. Here we report the possible detection of emission at a distance of about 03. This result is preliminary and needs confirmation by NIC-1 observations with its better sampling of the stellar profile.

binaries: close - stars: individual: Cyg X-3 - stars: Wolf-Rayet - infrared: stars - circumstellar matter

Introduction

Cyg X-3 is one of the most luminous X-ray sources in the sky. On the basis of its periodic X-ray variability, it has been interpreted as a binary system in which substantial mass transfer occurs from a companion onto a collapsed object (Bonnet-Bidaud & Chardin 1988). Unfortunately, the enormous reddening toward Cyg X-3 (E_B-V>5) effectively prevents any optical studies of this system (V > 24, $I\approx 20$, $K\approx 12$). However, recent infrared observations of Cyg X-3 have provided important information regarding the nature of this object. The first K-band spectrum of the system, obtained by van Kerkwijk et al. (1992), revealed strong and broad emission lines of He I and He II and it was suggested that the mass donor is a Wolf-Rayet (WR) star. Schmutz et al. (1996) found that the K-band emission lines exhibited periodic wavelength shifts, presumably due to the orbital motion of the WR star around the compact companion. The mass function for the system was found to be 2.3 and reasonable estimates of the mass of the WR star yielded a mass of the compact object of $\approx 17$ M$_{\sun}$; thus, the compact object is an excellent black hole candidate. These results strengthen the suggestion that Cyg X-3 is representative of the endpoint of massive binary evolution.

Despite theoretical predictions for the formation of systems composed of a WR star and a compact companion (van den Heuvel & de Loore 1973), observational searches for such objects had yielded negative results (Willis et al. 1989; Moffat 1992; St-Louis et al. 1993) until the observation by van Kerkwijk et al. (1992). The evolutionary steps for the formation of such a system are O+O $\Rightarrow$ RSG+O $\Rightarrow$ WR+O $\Rightarrow$ SN+O $\Rightarrow$ c+O $\Rightarrow$ c+RSG $\Rightarrow$ c+WR, where c denotes a compact object. Numerical calculations (Iben & Livio 1993) indicate that the common envelope phase (c+RSG) is the critical phase in the evolution, and few c+WR systems are expected to exist as a result of the complete disruption of the secondary during this phase. The fact that only one c+WR system has been found to date indicates that it is indeed rare for a system to survive the common envelope phase. Thus Cyg X-3 provides an excellent and completely unique laboratory for testing the models of massive binary evolution.

Prediction of Circumbinary Material

It has been estimated that on the order of a few tens of solar masses of material was lost during the common envelope phase of Cyg X-3 and various models predict that the material should be concentrated in the orbital plane (Livio & Soker 1988; Terman et al. 1994). The current fast wind from the WR star should interact with this slowly expanding circumbinary material. This situation should lead to a ``ring'' of condensed matter around the system, probably broken up into individual knots due to Rayleigh-Taylor and other instabilities (Garcia-Segura & Mac Low 1995), similar to the ring observed around SN1987A. This is demonstrated in Fig. 1, which presents predictions for the density contours resulting from a 2-D numerical simulation in which a fast stellar wind collides with a slowly expanding shell (Folini & Walder 1998, personal communication; see also Walder & Folini 1998).

  

Figure 1: 2-D simulation of the collision of a fast wind with a slowly expanding shell. The developing Rayleigh-Taylor instabilities are traced by density contours. The rectangular grid outlines different levels of the adaptive mesh refinement algorithm. This figure demonstrates that we should expect many small regions with high density rather than a shell with uniform density. The concentration into high density knots is even more pronounced when time-dependent cooling is included in the calculations.

Localized density enhancements such as those seen in this figure would be observed as knots. The formation of high density knots is even more pronounced if the time dependence of the cooling is taken into account, as this leads to further compression and enhancement. The dense regions are relatively cool (with low gas pressure) and the material is ionized by the X-ray emission from the black hole and the strong Lyman continuum flux from the Wolf-Rayet star. Shock heated layers are expected to have fairly low density and therefore contribute little to the emission measure of the the strong nebular recombination lines.

The size of the expected emission region should depend on the velocity of expansion of the shell of material expelled during the common envelope phase and on the time since the common envelope phase. The expansion velocity in turn, is given by the accumulated momentum from the wind that accelerated the shell's mass. (The initial expansion from the shell can probably be neglected.) For a typical duration of the WR phase, $\sim 3\times 10^5$yr, and an assumed average expansion of $\sim 1$kms^-1, we estimate an extension of 0.3pc. At 10kpc, the commonly adopted distance of Cyg X-3 (Dickey 1983), this corresponds to an angular separation of 6. This estimate is highly uncertain, and a search for nebular emission at both larger and smaller separations is also necessary. For example, typical sizes of WR ring nebulae are on the order of a few pc (Tutukov 1982), i.e. on the order of 20 at a distance of Cyg X-3.

The morphology of the emission nebula can provide observational constraints for the predictions of the common envelope mass loss and the nebular emission can be used to place constraints on the initial mass of the system. Contamination by swept up interstellar material can be excluded as the source of any possible nebular emission detected relatively close to Cyg X-3 (i.e. anywhere within the field of view of the NIC-2 camera), because the predicted SN explosion prior to the common envelope phase should have cleared out this region. Detected nebular Pa$\alpha$ emission would therefore allow us to calculate the circumstellar mass and obtain an estimate of the lower limit for the initial mass of the WR star. This assumes that most of the ionized material is density bounded. Alternatively, if the nebula is ionization bounded, then we can probe the far UV radiation field, especially if we detect the nebula in the light of He I, [Fe II], and [S III]. In either case, an unambiguous detection of nebular emission close to Cyg X-3 would confirm the predicted common envelope phase with its dramatic mass loss.

  

Figure: Left panel: H-band image of an 18" x 18" field around Cyg X-3, obtained with the Adaptive Optics system on the CFHT 3.6m telescope on 15 July 1997 (Vacca et al. 1998). North is up and East is to the left. The FWHM is approximately 05; the white cores of sources X and D have diameters approximately equal to the FWHM. Sources are labeled according to Joyce (1990) and Fender & Bell-Burnell (1996); X denotes Cyg X-3. Right panel: The H-band image of the left panel after 300 iterations with the algorithm of Lucy-Richardson, with the PSF made from Z and D. The stellar cores (white) now have a FWHM of about 02.

Previous Observations

There is observational evidence for extended nebular emission associated with the Cyg X-3 system. Spectroscopic observations of Cyg X-3, obtained with $\approx 1$ seeing and slit sizes, reveal the presence of narrow emission lines superimposed on the broad WR emission features. An example of this emission can be seen in Figs. 3 and 4 of Schmutz et al. (1996). These narrow lines are much stronger than any variable emission features often seen on top of the WR emission lines in the spectra of WR binary systems and attributed to the wind-wind collision zones (Moffat et al. 1996). Therefore, the narrow emissions in Cyg X-3 are probably not related to the WR wind but originate at larger distances from the system.

We have recently obtained ground-based near-infrared images of Cyg X-3, using adaptive optics (AO) techniques, which also seem to suggest that extended nebular emission may be present in this system. In Fig. 2 (from Vacca et al. 1998) we present our H-band image, obtained with the University of Hawaii Adaptive Optics system on the CFHT. The total integration time for this image was 600sec. It was constructed by co-adding 20 individual frames, each with an exposure time of 30sec. The FWHM of this image is about 05. Because the only suitable reference star that can be used to make the AO wavefront corrections is 30 away from Cyg X-3 and has a brightness of only V=15 (both values are at the limit of what can be used in current AO systems), our exposure resulted in a PSF which is slightly elongated to the SE, in the direction of this reference star. Nevertheless, this image indicated that we may have detected extended emission to the west of Cyg X-3 in the region of object #2. Unfortunately the detection is marginal because of problems with the flat-fielding and background sky subtraction. But it is sufficiently far from the point sources that we thought it might be real until we saw the HST NICMOS images. In fact, it is an artifact resulting from the flat-fielding.

Recently, Ogley et al. (1997) have reported that Cyg X-3 appears to be slightly extended in their deep K-band images and they suggested that a second point source, a factor of 11 fainter than the primary object, was located within 056 of Cyg X-3. (Dilution by a second stellar object might also explain why the broad WR emission lines seen in the K-band spectra have unusually small equivalent widths for a WR star (van Kerkwijk et al. 1992; van Kerkwijk 1993; Schmutz et al. 1996). However, our AO images (Fig. 2; Vacca et al. 1998), do not confirm this suggestion. Despite the slight elongation of our PSF, a second point source with the claimed separation and brightness would be easily recognizable on our image. (We also do not find a second source in the K-band image). NICMOS images, with their much smaller PSF cores, easily settle this question and place much stricter limits on the brightness and separation of any possible additional source.

NICMOS Observations

Cyg X-3 was observed with HST and NICMOS over 3 orbits on 1998 March 13. We used the NIC-2 camera in combination with the narrow band filters F187N, centered on Pa$\alpha$, and F190N, for the adjacent continuum region. For both filters we obtained 4 dither positions with 2 step size. The total exposure times were 68 min and 34 min for the filters F187, and F190N, respectively. For the analysis presented here we used the mosaiced images as delivered by the standard STScI pipeline. The standard processing leaves a clearly visible offset in the bias levels of the four detector quadrants but otherwise, particularly in terms of image sharpness, the processing resulted in excellent images.

In Fig. 3 we show the resulting image of the Cyg X-3 $19\arcsec\times 19\arcsec$ field.

  

Figure: Mosaiced NIC-2 image of $4\times 17$min exposures through the Pa$\alpha$ filter F187N. The observation angle was $79.4^{\deg}$, i.e. north is approximately to the right and east to the top. The field of view of one NIC-2 exposure is $19\farcs2\times19\farcs2$ and the mosaiced image has an extension of $286\times 286$ pixels or $21\farcs45 \times 21\farcs45$. The strongest source near the center is Cyg X-3 and the next brightest object to the south-east is star Z. Star D is the object to the south of star Z about as bright as Z. The grey scale representation is chosen to enhance faint features just above the background. Due to this representation the different bias offsets in the four detector quadrants are very obvious.

All previously known sources in the field are clearly detected; a comparison with Fig. 2 shows the presence of objects 1, 2, 9, and 10. In addition, there are about half a dozen newly discovered faint objects in the image.

Discussion

With the HST NICMOS observations we were searching for 4 types of emission:

1.

nebular emission close ( $< 1\arcsec$) to Cyg X-3;

2.

nebular emission in the vicinity of Cyg X-3 ( $\approx 5\arcsec$);

3.

extended nebular emission in the general area of Cyg X-3;

4.

continuum emission from a source at about 05 from Cyg X-3.

Each of these items corresponds to a possibility discussed in the Sects. 2 and 3. The easiest item to discuss is item #4. Ogley et al. (1997) suspected an object about 2.6mag fainter and 056 to the west of the main object. In Fig. 4 we show the cores of the stellar images of the

  

Figure 4: As Fig. 3 but the grey scale representation chosen such that the cores of the stellar images are visible.

objects X, Z, and D. The lightest grey level represents emission about 4mag fainter than the central pixel. We conclude that at the time of our HST observation there was no object present with the brightness and distance suggested by Ogley et al. (1997). Thus, we confirm the results of Vacca et al. (1998) based on AO images.

We can also address item #2 relatively easily. Using the intensities of the stars D and Z we scale the image F190N with a factor of 0.97 to that of F187N and we produce a difference image F187N-F190N. An investigation of this difference image does not reveal any extended emission. The only signatures left above the background are strong Pa$\alpha$ emission at the location of Cyg X-3 and imperfect cancelations of the stellar images of star D and Z that leave paired positive and negative intensities. (The negative-positive residuals from stars D and Z integrate to zero intensity, of course, because we have used them to scale the continuum observation to that of the Pa$\alpha$.) The strength of the Pa$\alpha$ excess emission of Cyg X-3 is 40% of the continuum. This value agrees very well with the emission strength estimated from ground-based spectroscopy of Cyg X-3 (see Fig. 4 of Schmutz et al. 1996). In fact, the emission is not from Pa$\alpha$ at all, but rather the broad He II 6-8 stellar wind line.

With the difference image described above we have also addressed, in part, item #3. However, if there is a nebular structure like a ring nebulae around Wolf-Rayet stars then this would be outside the field of the NIC-2 camera. We also have Pa$\alpha$ images of NIC-1 and NIC-3 that are obtained in parallel with the NIC-2 exposures. With the parallel exposures we probe regions about 30 and 80 to the west of Cyg X-3. Although these images are not in focus they would still allow to detect extended nebular emission. In the difference images of NIC-1 and NIC-3 there is also no obvious nebular emission. Of course, with this parallel exposures we are only covering only a small fraction of the total area out to, say, 100. Thus, item #3 cannot be addressed conclusively with our HST NICMOS observations.

The only remaining issue is item # 1, nebular emission close (i.e. $< 1\arcsec$) to Cyg X-3. The detection of such emission requires a deconvolution of the stellar image of Cyg X-3. As can be seen on Fig. 3 the point-spread function (PSF) of NICMOS is quite complicated. But the good news is that it is supposed to be very reproducible, apart for some ``breathing'' effects which change the focus during an orbit.

  

Figure: Deconvolved NIC-2 F187N Pa$\alpha$ image using the image restoration scheme by Lucy and Richardson as implemented in MIDAS. The result shown is after 80 iterations; the results after 40 or 120 iterations look very similar. For the PSF the F190N image of Cyg X-3 was used. Left panel: The $4\arcsec\times 4\arcsec$ region around star D. Right panel: The $4\arcsec\times 4\arcsec$ region around Cyg X-3. This region also includes star 1 (seen at the left).

The best PSF template available for these observations is that of Cyg X-3 itself on the F190N image. The difference in wavelength is small so that the F190N PSF is only about 1.6% more extended than that expected for the F187N image. For the preliminary analysis presented here we disregard this difference. In Fig. 5 we present the results of a deconvolution with the Lucy-Richardson restoration algorithm as implemented in MIDAS. In the left panel we see that our PSF agrees very well with the Pa$\alpha$ image of star D; the deconvolution of this star has produced a diffraction limited point source with no residual wings. On the other hand, the profile of the deconvolved image of Cyg X-3 is more extended than a pure point source. There is an extended structure to the east and a less bright one to the west. Therefore, Cyg X-3 appears to be more extended in Pa$\alpha$ than it is in the continuum image. The number of iterations turned out to be non-critical because the basic result, the extended structures to the east and west, is robust. The intensity in the east feature is comparable to that of star 1.

The extended nebular emission can also be detected in the original image. In Fig. 6 we plot a west-east cut

  

Figure: West-east crosscut, i.e. from bottom to top in Figs. 5, through the NIC-2 F187N image of Cyg X-3 (dashed) and of F190N (dotted). The 5 central columns are averaged and the central intensity of the F190N cut is scaled to that of the F187N. Excess emission in the east wing of Cyg X-3 is visible, at a distance of 02-03 from the center (1 pixel $= 0\farcs075$).

through the profiles of Cyg X-3 in the F187N and F190N images. The lines represents the average of the 5 central columns. The excess emission detected by deconvolution algorithm in Fig. 5 can easily be seen 02-03 east of the main peak; there is also a difference between the two profiles, although less pronounced, to the west. The uncertainties of the profiles at a level of 2 and 3 counts per second is 2% and 1%, respectively. (The total integration time was 4096 sec for the F187N exposure and 2048 sec in the F190N image.) Therefore, the difference between the two curves on the east side is at a level of about 10$\sigma$. Because the F190N image has been used as the PSF template for the deconvolution, we also know that the Cyg X-3 F190N profile agrees well with that of star D and Z in the F187N exposure. Nevertheless, despite the high confidence level, it is still possible that the emission detected close to Cyg X-3 results simply from spatial and/or temporal variations of the PSF, and hence a mismatch between the true PSF for Cyg X-3 and the PSF used to perform the deconvolution.

Conclusions

In view of the importance of Cyg X-3 to the understanding of massive binary evolution -- as the only representative of the c+WR class -- we have searched for nebular emission that would testify to the events during its previous evolutionary phase. We suspect that the previous phase was a common envelope phase which implies that the non-degenerate star has lost a substantial fraction of its mass. This mass is expected to be still around the system, probably in a ring. To date, there have been no clear detections of emission around Cyg X-3. However, on the basis of the first part of our HST observations, we report here the possible detection of matter close to Cyg X-3. Our result is preliminary and at this stage of the analysis it is still not clear whether the features seen are artifacts produced by subtle imperfections in our PSF. Clearly, we need a better sampling of the profiles of the Cyg X-3 image and of the PSF. We hope to achieve this in an upcoming HST observation with NIC-1 camera. If the extended nebular emission is confirmed, we will estimate the mass lost from the donor star during the common envelope phase. This result could have a profound impact on models of binary evolution.

WS is grateful to R. Hook and S. Stolovy for enlightening discussions on the properties of NICMOS PSF. WDV is grateful to Alan Stockton for partial support. Support for this work was also provided by NASA through grant number G)-07838.01-96A from the Space Telescope Science Institute, which is operated by AURA, Inc., under NASA contract NAS5-26555.