CANDIDATE CORONAGRAPHIC DETECTIONS OF PROTOPLANETARY DISKS AROUND FOUR YOUNG STARS

The observations presented here are part of a large coronagraphic survey of young stars with the Subaru telescope. This survey had two goals: discovering low luminosity companions to young stars, and directly resolving scattered light from disks in the NIR. NIR stellar light scattered from the surface of a circumstellar disk can be a powerful tracer of the morphology and extent of the circumstellar disk, although the interpretation of the results is challenging. Unlike molecular line or submillimeter continuum observations, scattered light observations are sensitive to the surface of the disk only, for an optically thick disk, rather than the total amount of material. Consequently, these observations are potentially sensitive to relatively low amounts of material. Conversely, information about the interior of the disk is not probed. An additional complication in interpretation is caused by the fact that scattered light can also be observed from the interior walls of the outflow cavity of embedded young stars, and structures in the envelope itself, leading to confusion regarding the source of the emission.

All the imaging data were taken with the Subaru Coronagraphic Imager with AO (Tamura et al. 2000; Murakawa et al. 2004) and were retrieved from the public Subaru archive. The original, full sample of targets was composed of roughly 100 targets selected for the likelihood of observing a disk. The sample was not chosen as a statistically uniform one; rather, it consists of a selection of sources of various masses and spectral types which show strong evidence for the presence of a circumstellar disk in a variety tracers. Evidence for a circumstellar disk can be inferred from excess emission in the near- to far-infrared (Adams et al. 1988), resulting from the absorption and re-radiation of stellar light by dust, from submillimeter and millimeter line and continuum emission emitted by the molecular gas and dust, infrared spectra showing dust features, and from X-ray emission (particularly for weak-lined T Tauri stars), Hα and lithium line emission indicative of active accretion, which imply the presence of an associated disk.

A large number of individual sources in the original sample have been analyzed by a variety of researchers, particularly for the higher mass Herbig Ae/Be stars and classical T Tauri stars with optically thick, more readily observable disks, and those sources which clearly show evidence of extended structure in scattered light (e.g., Fukagawa et al. 2004; Mayama et al. 2006; Hioki et al. 2007; Itoh et al. 2008; Kudo et al. 2008). The remaining sources, consisting primarily of lower mass K and M stars and more evolved weak lined T Tauri stars, do not immediately show evidence for scattered light from disks. As a result, careful, detailed data reduction and analysis are required in order to detect extended emission from scattered light. The existence and nature of circumstellar disks around WTTS are still a subject of much debate. This makes the detection of disks around WTTS a less likely proposition; at the same time it makes the detection of such a disk scientifically very significant, and of great interest in furthering the understanding of disk evolution. Therefore, the study of these objects is of great interest in spite of the small amount of material.

In light of this, we have reduced and analyzed 45 sources from the above survey in order to detect extended emission on small physical scales (<200 AU). Table 1 lists all the sources in this sample. This sample is heavily weighted toward lower mass stars, particularly K and M stars, and WTTS. The sample is composed of 12 Classical T Tauri stars (CTTS), 31 WTTS, one Herbig Be star, and one YSO. The spectral types are weighted toward lower mass stars: eight M stars, 24 K stars, seven G stars, one F star, and three stars of unknown spectral type. Most of the sources are located in Taurus, at a distance of 140 pc, while a few are located in other regions. The complete 45 object sample, with detection limits and companions, will be presented in a subsequent paper.

NIR coronagraphic observations in the H band were performed on 13 nights between 2003 November and 2005 November. The observing times and number of frames used vary from source to source and are listed in Table 2. Even with the use of a coronagraphic mask, there is still significant remnant flux in the wings of the point spread function (PSF). Consequently, reference PSFs from non-binary, non-disk sources were observed for each target in order to produce the near-complete post-observation subtraction of the PSF. Photometric standards from the UKIRT faint spectral standard catalog (Leggett et al. 2006) were observed on each night.

The data were reduced using the standard IRAF routines: dark frame subtraction, flat fielding, bad pixel masking, cosmic-ray reduction, and sky subtraction. The position of the source in the image can shift slightly due to minor variations in the AO performance during the night. The partial transparency of the CIAO mask (≈2%) means that the central source is visible through the mask. Consequently, the position of the central source can be used to center each frame exactly. The target frames were checked for AO performance, using a measure of the AO quality which compares the peak flux to the flux in the wings, measuring the sharpness of the image. Each target and reference frame was visually inspected to check for bad frames, misaligned PSFs or other issues.

For each target source, the 10 frames with the best AO were selected. Ten PSF reference frames were then selected to best match the quality of the AO in the target source. The reference frames were rotated to match the instrumental rotation of each target frame, aligning the spider patterns on the image. The reference frames were then co-added to create a reference PSF for each individual target frame.

There are several possible methods for determining the best parameters for the subtraction of the reference frame, which may need to be shifted slightly in position (due to the centering of the source within the mask), and/or scaled to match the intensity of the target frame. We have adopted a two-phase approach; an automated routine, followed by a manual fine tuning. In the first phase, the averaged intensity in the wings of the PSF was used to estimate the scaling factor. The shift in the position of the central pixel was determined by calculating the noise in the wings of the PSF and minimizing the difference between the noise at four positions located away from the diffraction spikes. In the second phase, we interactively checked the parameters derived above to test for the quality of the fit and the robustness of the parameters. It should be noted that the automated algorithm will work best in the absence of any extended emission, and will result in oversubtraction for symmetric extended emission. At this point, corrections to the position and scaling were generally minor, typically less than a few pixels in position and 10% in scaling. The final subtracted images were then co-added to produce a final image.

PSF subtraction for binary sources is a more complex process. In this case, a reference PSF was created by adding two PSFs, each scaled and shifted to match the relative fluxes and positions of the binary components, before subtraction.

Potential detections were selected through a multi-stage process. Only one source, Haro 6-10, showed clear extended emission in the original coronagraphic image, as is the typical case in coronagraphic detections of extended emission from disks and outflows. Haro 6-10 is a known embedded binary protostar, and the detected emission is not likely to be disk emission, but rather scattered light off the interior of the outflow cavity (Koresko et al. 1999). For the remaining sources, potential detections were limited to the region within the radius of the PSF halo. As a consequence, the quality of the final subtraction is in large part determined by the fine details of the AO quality in the images, and the matching of the target and reference PSFs prior to subtraction. Subtraction artifacts are therefore of major concern.

The detectability of scattered light from the surface of disks depends on the brightness of the scattered light, the orientation and morphology of the disk, and the distance of the source. The coronagraphic observations are also highly dependent on the technical details of the observations. The coronagraphic mask size determines the minimum angular scale at which scattered light can be detected; this is in practice larger than the mask size. The quality of the data is also highly sensitive to the performance of the AO system and the subtraction of the residual PSF effects. Even with the selection of the best-quality image frames there are noticeable variations in the size of the PSF. Poor AO performance and mismatched PSFs can either mask existing emission or create false detections. For sources which show extended emission outside of the area dominated by the PSF halo, as in most previously reported cases, this makes the selection of the best PSF subtraction parameters problematic. For extended emission that is contained entirely within the halo-dominated region, as in the sample presented here, the uncertainty comes in at the detection level, rather than the interpretation level.

A source was rejected if it showed a good but negative PSF subtraction, in which there was no residual extended emission or remaining artifacts. A source was also rejected if it showed a poor but negative PSF subtraction, with no remnant extended emission but significant artifacts from the subtraction process. Figure 1 shows an example of a rejected subtraction; DF Tau is the result of a null result from a moderately bright source. The inner region of the subtraction, immediately outside the mask, is dominated by subtraction artifacts and is useless for the purposes of data analysis, even in example of clear detections. In practice, subtraction artifacts from the diffraction spikes remain even in the best subtractions, and for bright sources in particular, these are impossible to remove completely. In Figure 2, we show the pre- and post-PSF subtraction profiles for the same source, more quantitatively illustrating the magnitude of the halo subtraction effect; compare this with the detection profiles discussed in Section 4. If a source showed evidence of extended emission, it was rejected unless the extended emission was present with the same morphology and direction in all 10 of the individual frames.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We performed a series of tests on potential detections to attempt to rule out instrumental or subtraction artifacts. We checked the variation of the PSF from frame to frame by eye; a visible variation in PSF size was noticeable in most observation runs, but a distortion in the shape of the PSF was not readily apparent. A rotation of the reference source by 90° before subtraction was in general inconclusive; small-scale emission was dominated by the effects of the diffraction spikes. When possible, we tested variations in the individual reductions to check the robustness of the subtraction.

In addition, we checked for systematic effects in the observational data and processing which could explain the detected sources. We compared the relative rotations of the target and reference, the brightness of the sources, and the quality of the PSF matching between the target and reference source. The potential detections did not show any systematic biases when compared to the rest of the sample, in relative rotation angle, flux of the source, quality of the PSF matching or night observed.

The profile of the extended emission was calculated, and compared to the profile from known bad subtraction artifacts. This was also relatively inconclusive due to the small angular scale over which extended emission was observed.

CI Tau is a classical T Tauri star of type K7, located in the Taurus region at a distance of 140 pc (Kenyon & Hartmann 1995). The post-subtraction image of CI Tau in Figure 3 shows a nebulosity extending from NE to SW. To the southwest there is some contamination from a ghost image in the target frame, which has been labeled. The physical extent of the nebulosity is approximately 190 AU, and the position angle 145° east of north.

There have been previous resolved detections of the circumstellar disk around CI Tau. Dutrey et al. (1996), in 2.7 mm continuum observations with the Plateau de Bure interferometer, made a marginal resolution of the disk, with an extent of 2.0 by 0.8 arcseconds and a position angle of 40°. Later interferometric submillimeter observations performed by Andrews & Williams (2007) at 340 GHz with the Submillimeter Array (SMA) resolved emission. They fit a disk with a radius of 225 AU, a mass of 0.04 M_☉, an inclination of 46° and a position angle of 131°, nearly perpendicular to the previous detection. It should be noted that these observations were barely resolved. The latter angle is consistent with the major axis of our NIR-scattered disk, in which case the scattered light images might show the light from the surface of the disk detected with SMA.

DI Cepheus is a CTTS with a spectral type of G8, located at a distance of 300 pc (Gameiro et al. 2006). We detect a nebulosity, shown in the second two panels of Figure 3, that is similar in morphology to CI Tau, although the greater distance gives a physical scale of 370 AU for DI Cep. The position angle is 95°. A companion candidate can be seen to the southeast of the image.

There have been no high-resolution observations at the scales probed by the Subaru observations. The strong infrared excess in the SED (discussed further in Section 6.1) is strongly indicative of extensive circumstellar material, while the Hα line width (Hessman & Guenther 1997) implies accretion, which assumes the presence of a disk. Mid-infrared spectroscopy by Bowey et al. (2003) detected strong silicate emission from circumstellar dust; they fit a disk with a 10⁴ AU radius to the line profile. Gameiro et al. (2006) performed high-resolution optical spectroscopy of DI Cep and argued that its disk is observed nearly edge-on. The morphology of the scattered light images indicates a disk that is seen at an angle from face on, although not likely directly edge-on.

LkCa 14 (aka V1115 Tau) is a weak-lined T-Tauri source with a spectral type of M0V, also located in the Taurus region (Kenyon & Hartmann 1995). The extended emission shown in the third set of panels in Figure 3 is of smaller scale than the other sources, and as a result it is the weakest of our detections. The emission is again elongated, with a physical radius of 120 AU and a position angle of 135°. The corresponding profiles are shown in the third panel of Figure 4.

There has been no previous, resolved detection of the disk around LkCa 14. Andrews & Williams (2005) place an upper limit on the mass of the disk of 0.004 M_☉, based on single-dish submillimeter observations at 340 GHz. Hα is seen in emission, which is an indication of accretion, and indirectly of the presence of circumstellar material. It should be noted that coronagraphic observations probe scattered light off the surface of the disk, and consequently a relatively small amount of material is needed to produce a scattered light image.

RX J0338 is a WTTS of type G9, also located in Taurus at 140 pc (Magazzu et al. 1997). It shows irregular extended emission, as seen in the last two panels of Figure 3, with a radius of 190 AU. Very little is known about this X-ray-selected source.

A team associated with the Spitzer GLIMPSE Legacy Survey (Robitaille et al. 2006, 2007) has developed a powerful online tool for fitting SEDs to photometric observations of YSOs, making it possible to efficiently analyze SEDs in the context of physical models of YSO systems. Fitting detailed YSO models to an SED using radiative transfer simulations is a non-trivial task. The group takes a novel approach; first generating a suit of 20,000 three-dimensional radiative transfer models (Whitney et al. 2003a, 2003b) for a grid of two-dimensional YSO models, using a Monte Carlo technique which traces the path of individual photons. The models cover a wide range of stellar masses and ages, masses and geometries for disks, envelopes and outflow cavities, and a set of 10 inclinations for each model. Observed SEDs, from the optical to millimeter, are then fitted to the grid of models, with the addition of distance and interstellar extinction effects.

There are several important caveats to this method of fitting. The most important is the degeneracy of the fits. With the large number of physical parameters input to the radiative transfer model, the fits are highly non-unique, with multiple combinations of parameters fitting a given SED. Consequently, care must be taken not to overinterpret the physical significance of the results. However, the fits can be used to interpret the plausibility of a given set of models, and to constrain the general best fit of given a parameter. This works better for some parameters than others; the disk inclination, in particular, is poorly constrained by the SED fits.

Figure 5 shows the SEDs and fits for the tentative detections. In each case, we have used the best literature photometry available (optical, near- to far-infrared and sub-mm/mm); the photometry is summarized in Table 4. The distance and interstellar extinction have been constrained to measured values in the initial fit. From the suite of models that result, we have then selected the best model which matches the stellar mass of our source.

Zoom In Zoom Out Reset image size

CI Tau and DI Cep both show strong infrared excesses in the mid-infrared and beyond. CI Tau is fit with a 0.4 M_☉ star with a 0.03 M_☉ disk, similar to observational estimates of the disk mass, while DI Cep is fit with a 1.1 M_☉ central star, with a disk mass of 0.03 M_☉.

LkCa 14 and RXJ 0338, on the other hand, are adequately fitted by an extincted stellar spectrum in the range of the available photometry. This is consistent with their classifications as WTTS, where little, if any, infrared excess is generally seen. They do not, therefore, have a large disk, but the SEDs do not rule out the possibility of a low-mass remnant disk, or a cooler disk at larger radii with a central gap. The latter would show an excess only at longer wavelengths. The model for LkCa 14 shown is for a 0.4 M_☉ star with a disk of 2.1 ×  10⁻⁴ M_☉, within the observational limits, while RXJ shows a photospheric SED for a central star of 5600 K, the best-fit stellar spectrum.

For further analysis of the morphology of the NIR emission, the scattered light properties of the above SED models can be determined. The radiative transfer codes used for generating the suite of models may be downloaded by the user and run directly. This generates multi-wavelength images that can be directly compared with high-resolution observations, after convolution with the relevant instrument parameters.

For the purposes of this paper, we have used the SEDs fit using this technique, as discussed previously in Section 6.1, as input to the radiative transfer code to generate high-resolution images. It is important to note that this is not a fit to the images; rather this allows us to explore how probable our observed images are given the source SED and reasonable system geometries. Several appropriate disk models were tested to explore the range of emission parameters for the given SED. The inclinations, which are poorly constrained by the SEDs but strongly affect the morphology of the scattered light emission, were tuned to match the morphology of the extended emission.

The results for CI Tau are shown in Figure 6 for a model with a total disk mass of 0.03 M_☉, outer radius of 133 AU and a disk inclination of 63°, matching the observed extinction and distance of the source. To compare with the observed data, a mask has been manually added to the images where the coronagraphic mask and diffraction spikes would dominate the image in real data, and the central few pixels (containing the stellar flux) have been removed. The simulated image has then been convolved with an approximation of the Subaru PSF, measured within the coronagraphic mask, and background noise added to match the final co-added image. The extent, morphology, and surface brightness of the observed extended emission can be reproduced with a physical disk/star model that is consistent with the observed SED. The flux levels match approximately, although the model levels are lower (by 50%) than the observed values. The physical scales are similar; however, the emission in the model drops off abruptly when the edge of the disk is reached, while the observed emission tapers off more gradually.

Zoom In Zoom Out Reset image size

For the WTTS, the situation is more difficult, as the observed SED only covers the stellar photosphere portion of the spectrum, with no observed excess emission. Therefore, fitting the disk portion of the SED is not possible. However, there are some limits on the parameters of the disk, as an infrared excess is not seen at NIR wavelengths. For RXJ 0338, we ran the SED fitter to produce a suite of consistent models; typically low-mass disks with a central hole. We selected several SED models which matched the central source, and calculated the corresponding scattered light images. Figure 7 shows an image for a disk which is consistent with the observed SED. The model has a disk mass of 2.8 ×  10⁻³ M_☉, with an inner disk gap of 20 AU and an inclination of 41°.

Zoom In Zoom Out Reset image size

One area of uncertainty in the calculation of the scattered light images is the dust emission characteristics, as the radiative transfer code does not accurately model the dust properties expected for a highly evolved disk. The original models used in the SED fits were generated using dust emission properties derived from the interstellar medium (Whitney et al. 2003b). Calculating the scattered light properties with the same disk geometry and mass, but using the alternate dust models provided with the code (Cotera et al. 2001; Wood et al. 2002; Whittet et al. 2001) does produce a noticeable difference in the scattered light properties, with a variation of about 50% in the size of the disk, measured at a given surface brightness level.