The Three Dimensional Morphology of the Circumstellar Ejecta of IRC

Transcription

1 The Three Dimensional Morphology of the Circumstellar Ejecta of IRC A DISSERTATION SUBMITTED TO THE FACULTY OF THE GRADUATE SCHOOL OF THE UNIVERSITY OF MINNESOTA BY Chelsea Lynn Tiffany IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF Master Of Science Dr. Roberta M. Humphreys, Advisor July, 2010

2 c Chelsea Lynn Tiffany 2010 ALL RIGHTS RESERVED

3 Acknowledgements There are many people that I would like to thank for their assistance and support. My advisor, Dr. Roberta Humphreys, who has been helpful and supportive of my work in graduate school. Also, the Astronomy department for allowing me to teach. The other graduate students for their support and for making graduate school fun. L. Andrew Helton for support in data reduction and IRAF knowledge. Andrea Mehner and Jennifer Delgado for Friday lunches. Thanks as well to my commitee of Dr. Roberta Humphreys, Dr. Terry J. Jones and Dr. Alex Heger. And lastly, my family for their support through graduate school. This work was supported by NASA through grant GO from the Space Telescope Science Institute. This research has made use of SAOImage DS9, developed by Smithsonian Astrophysical Observatory as well as Image Reduction and Analysis Facility (IRAF). IRAF is distributed by the National Optical Astronomy Observatories (NOAO), which are operated by the Association of Universities for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation. Additional software packages for IRAF designed by the STScI were also used in this research. The journal article with this published work is available online from the Astronomical Journal [Tiffany et al. (2010)] and at i

4 Dedication This is dedicated to my mother Dr. Sally J. Pyle and grandmother Betty M. Pyle, and in memory of my grandmother Dr. Lois H. Tiffany. ii

5 Abstract This thesis explores the 3 dimensional behavior of the outer ejecta from the post-red supergiant IRC This star inhabits a region of the Hertzsprung- Russell diagram just below the Humphreys-Davidson instability limit. It also appears to be evolving from a cooler temperature to a hotter one, making it a very unique star. It is the only star of its type of have observed nebulosity. Images of this star show the nebulosity, or complex ejecta that is evidence of episodic mass loss. The transverse motions of the various features seen in the inner ejecta are explored. This information is obtained from two sets of images taken with the HST/WFPC2 in 1996 and Additional information from a previous paper also gives the radial velocity for several features, allowing the determination of the total motion for a number of the features. Combining the radial and transverse velocity information into a time since ejection, this gives a picture of the mass loss history, the time between ejection epochs and the outflow from the star. Additionally, the radial velocity measurements prove that these features are spatially distinct and that are expanding away from the star. The motion of these features is predominantly dominated by their transverse motion and they appear to be moving within a few degrees of the plane of the sky. The connection between this and the magnetic fields are discussed. This leads to the conclusion that IRC has approximate rotational symmetry and is viewed almost pole-on, with the equatorial plane nearly in the plane of the sky. iii

6 Contents Acknowledgements Dedication Abstract List of Tables List of Figures i ii iii vi vii 1 Introduction 1 2 Observations, Data Processing and Measurement Procedure Observations Data Reduction PSF Removal Measurement Procedure South East Outer Feature cross-correlation An Explanation of the Headings for Table Radial Velocity Measurements and Total Velocity Kinematics of the Ejecta East Jet and North East Knot G South East Jets iv

7 3.3 South East Knots South East Outer Feature South West Square & South West Jet South West Triangle & South West Knot A South West Fan The Semi-Circular Arcs The Arcs as Expanding Bubbles Arc Arc Arc Summary Geometry of the Ejecta 31 5 Discussion, Conclusions and Future Work Discussion Mass Loss History Concluding Remarks & Additional Work References 40 Appendix A. Data Table 44 A.1 Appendix Data v

8 List of Tables 2.1 HST / WFPC2 Filter Information Observations and Combined Images Postion, Velocity and Direction of Motion of the Features D Postion, Velocity and Direction of Motion Arc Center Point Fit Information The 3D Motion of Arc 1 Relative to its Center Values Relative to the Arc Center Point A.1 Mesurements (in milliarcseconds) vi

9 List of Figures 1.1 Slit positions from Humphreys et al. (2002) Image 547l1 with general features labeled TinyTim PSF generated for the 467 filter long before and after reduction Inner and outer features labeled Position Orientations Example of motion for two features Image 547l1 with arcs labeled Image 675ex scaled to show faint nebulosity Image 547l2 with maser emission marked vii

10 Chapter 1 Introduction IRC is one of the most important stars in the upper HR diagram for understanding the final stages of massive star evolution. With its high luminosity (L L ) and A-F type spectrum it is one of the stars that defines the empirical upper luminosity boundary for evolved stars in the HR diagram (Humphreys & Davidson (1994)). IRC is also a strong OH maser, one of the warmest known, and one of the brightest µm IR sources in the sky with an extraordinary mass loss rate ( M yr 1 ) found in [Knapp & Morris (1985)], [Oudmaijer et al. (1996)] & [Humphreys et al. (1997)]. It has been variously described in the literature as either a true supergiant [Humphreys et al. (1973)], [Giguere et al. (1976)] and [Mutel et al. (1979)] or a proto-planetary/post-agb star [Habing et al. (1989)], [Hrivnak et al. (1989)], & [Bowers & Knapp (1989)], depending on distance estimates that ranged from 1.5 to 7 kpc. [Jones et al. (1993)] (Paper I) combined multi-wavelength spectroscopy, photometry, and polarimetry to confirm a large distance of 4 6 kpc and the resulting high luminosity mentioned above. This conclusion was supported by [Oudmaijer et al. (1996)] who demonstrated from CO data that IRC has to be much more luminous than the AGB limit. All the work in this thesis assumes a distance of 5 kpc, based on these findings. 1

11 2 Images from the Wide Field Planetary Camera 2 on the Hubble Space Telescope (HST/ WFPC2) in 1996 (Humphreys et al. 1997, Paper II) revealed a complex circumstellar environment with numerous small condensations or knots, raylike features, and intriguing semi circular arcs or loops. A few other intermediatetemperature hypergiants such as ρ Cas and HR 8752 occupy the same region in the HR diagram, but IRC is the only one with apparent circumstellar nebulosity [Schuster et al. (2006)], making it the best candidate for a star in transition from a red supergiant to an S Doradus type variable (LBV), a Wolf-Rayet star, or a pre supernova state [Humphreys et al. (2002)]. IRC has shown some significant changes during the past century. It brightened by a magnitude or more in the 50 years prior to 1970 (Gottleib & Liller 1978, Paper I) and its apparent spectral type changed from a late F to an A-type supergiant in just the past 30 years [Oudmaijer et al. (1996)] and [Oudmaijer (1998)], although Humphreys et al. (2002), hereafter Paper III, demonstrated that IRC s wind is optically thick. Consequently, the observed variations in apparent spectral type and inferred temperature are more likely due to changes in its wind, and not to interior evolution on such a short timescale. The HST images of IRC show no obvious axis of symmetry, although the complex ejecta provide evidence for more than one high mass loss episode. The outermost reflection arcs at 5 arcseconds were ejected about 3000 years ago, possibly when the star was still a red supergiant, while the very complex structures closer to the star correspond to much more recent asymmetric mass loss events. Surface photometry of the optical and near-infrared images (Paper II) showed that IRC experienced a high mass loss episode during the past 600 years, shedding about 1 M with a mass loss rate 10 4 M yr 1. [Blöcker et al. (1999)] suggested that a high mass loss episode ended years ago which interestingly corresponds to the apparent onset of its brightening. The numerous arcs, knots, and jetlike structures are suggestive of localized ejection events in apparently random directions which may be due to large scale active regions on the star (Paper III).

12 3 Long-slit spectroscopy with the optical spectrometer on the Hubble Telescope (STIS) of the reflection nebula allowed the star to be effectively viewed from different directions (Humphreys et al. 2002). The positions of the slits in relation to the star is shown in Figure 1.1. The extracted spectrum at each location is essentially that of a reflection or scattering nebula. Figure 1.1: Slit positions from Humphreys et al. (2002) Measurements of the absorption minimum in the strong double-peaked Hα emission profile showed that the reflection nebula is expanding in a spherical outflow at 60 km s 1. Where a slit crossed one of the semi-circular arcs the velocities deviate from the expansion of the surrounding nebulosity indicating that the arc

13 4 is a kinematically separate feature. The shape of the Hα emission profile was remarkably uniform throughout the ejecta, contrary to what we would expect from previous models with an equatorial disk (Paper I) or bipolar outflow (Oudmaijer et al 1994, Paper II). More recent observations from integral-field spectroscopy [Davies et al. (2007)] provide additional evidence for a bipolar outflow, although the signature is not strong close to the star. Recent near-infrared interferometry [de Wit et al. (2008)] reveals an elongated emitting region on the milli-arcsecond scale about twice the size of the star. Their results are not conclusive, however, as to whether it is an edge-on disk or bipolar outflow. One possible explanation for the apparent conflicting evidence for the outflow and structure of IRC s circumstellar ejecta may be due its orientation with respect to the line of sight. IRC may be aligned nearly pole on to the Earth s line of sight. Additional second epoch images of IRC (with the WFPC2) were obtained to measure the transverse motions of the discrete structures in the ejecta. In combination with radial velocities from the previous long-slit spectroscopy, the time since ejection for the discrete structures can be determined along with a history of IRC s episodic mass loss and the morphology of its ejecta based on direct observations. The orientation and vector motions of these features will be important for understanding the origin of the ejecta and the responsible mass loss mechanism. In Chapter 2 the observations, data processing and measurement procedure for the transverse motions are described. Chapter 3 describes the kinematics of the ejecta, while Chapter 4 describes the geometry. Finally, the results are discussed in Chapter 5. Figure 1.2 gives a general outline of IRC with the major features and areas marked

14 Figure 1.2: Image 547l1 with general features labeled. 5

15 Chapter 2 Observations, Data Processing and Measurement Procedure In order to measure the transverse velocity of the ejecta around IRC , a minimum of two sets of observations were needed. With the assumed distance of 5 kpc and expansion velocities on the order of 40 km s 1 as seen in VY Canis Majoris (Humphreys et al. 2007), a minimum of 4.5 years was needed between observations to see movement of 0.33 pixels, which is the optimum resolution for the data in this thesis. A longer baseline was used both to allow objects with velocities less than 40 km s 1 to be seen and the account for a possibility of less than optimum resolution. After the observations were made (outlined in Chapter 2.1), the data were reduced together (Chapter 2.2) and these velocities were measured (Chapter 2.3). 2.1 Observations Images of IRC were taken on April 5, 1996 (Epoch 1) and on March 9, 2008 (Epoch 2) with the Planetary Camera 2 on the Hubble Telescope, giving two sets of observations with a baseline of years between them. The images from Epoch 1 were taken with the 467M and 547M filters; the numbers signify 6

16 7 the central wavelength of the filters (in nanometers) and the M indicates it was a medium filter (Table 2.1 for the details). The images from Epoch 2 were in both these filters as well as two additional filters (Table 2.1 and Table 2.2). The filter information is from the WFPC2 Handbook [McMasters et al. (2008)]. Table 2.1: HST / WFPC2 Filter Information Filter Central λ (Å) Filter width (Å) F467M F547M F675W F1042M Because of the wide dynamic range of the nebulosity associated with IRC , a range of exposure times was necessary to sample the complex ejecta. For ease of comparison, similar exposure times were used for F467M and F547M. The complete list of filters, exposure times and the subsequently combined images is given in Table 2.2. A number of images were taken using the Dither mode of the Hubble Telescope. This allows images with matching filters and exposure times (during the same epoch) to be taken with a 2.5 pixel (short exposure) or 5 pixel (long exposure) offset, which helps identify bad pixels on the chip as well as giving better image detail in the final combined images. The dithered images are in Table 2.2 with exposure times listed with 2 or 3. As noted above, the exposure times were similar between the two epochs. In the 547 filter, the second epoch had one long exposure (300 seconds) instead of two shorter exposures (140 seconds) as in Epoch 1. The Epoch 2 exposure had a large number of cosmic rays due to its long exposure time. In order to help remove these cosmic rays, the long exposure was combined with the next longest time in the filter (40 seconds). In order to match the times between the two epochs as closely as possible, the 40 second image from Epoch 1 was also combined with the

17 8 two 140 second exposures, even though cosmic ray removal was possible with only the two images. See Table 2.2 for the exact time combinations. In each epoch, the images were combined into short (s) and long (l) exposures. Table 2.2: Observations and Combined Images Date Filter Exposure Time (sec.) Combined Images April 5, 1996 F467M 12 s, 30 s 2 F467s1 (72 s) (Epoch 1) F467M 140 s 2 F467l1 (280 s) F547M 0.5 s, 3 s, 10 s F547s1 (13.5 s) F547M 40 s, 140 s 2 1 F547l1 (320 s) March 9, 2008 F467M 12 s, 30 s 2 F467s2 (72 s) (Epoch 2) F467M 140 s 2 F467l2 (280 s) F547M 0.5 s, 3 s, 10 s 2 F547s2 (23.5 s) F547M 40 s, 300 s F547l2 (340 s) F675W 0.5 s, 5 s F675s (5.5 s) F675W 30 s 3 F675l (90 s) F675W 600 s F675ex (600 s) F1042M 0.5 s, 5 s F1042s (5.5 s) F1042M 30 s, 100 s F1042l (130 s) 2.2 Data Reduction Both sets of data were reduced in tandem, using IRAF packages, to assure consistency. The routines in the HST Dither Handbook [Koekomoer et al. (2002)] for a diffuse source imaged with the Planetary Camera 2 (PC2) were used as the basis for the reduction. Cosmic rays were flagged with the task precor, and the shift between images in the same epoch were measured with crossdriz and shiftfind. 1 The 140 second exposures for filter 547 in Epoch 1 were not dithered.

18 9 The Space Telescope Science Institute (STScI) provides a list of the bad pixels at any given time in the detector (this changes between epochs). This information was combined with the flagged cosmic rays to create bad pixel masks for each image. The final images were created using these masks and the shifts between images with the task drizzle to combine the images together. The images were subsampled by a factor of two during the drizzle routine for a final resolution of arcseconds/pix. Images from the same epoch were then aligned and combined as listed in Table 2.2. With this resolution, and assuming a distance of 5 kpc, we expect to be able to measure motions as small as 15 km s 1 over the 11.9 year baseline. This is well below the expected average value of 40 km s 1 seen in VY CMa [Humphreys et al. (2007)] PSF Removal The next step was to remove the Point Spread Function (PSF). The procedure described in [Humphreys et al. (1997)] was used. TinyTim was used to create simulated HST specific PSFs to match the WFPC2 observations. 1 However, TinyTim has several limitations that affect its application to IRC To produce an accurate PSF for the PC chip, the PSF was sub-sampled by two, but this decreased the size of the PSF which was already too small for IRC For IRC an input size of 18 arcseconds is needed. Furthermore, the TinyTim PSF is color dependent and the maximum allowed B V color is below the value for IRC (B V of 2.7 mag), which is highly reddened by interstellar extinction; consequently the input color table was adjusted to match the spectral energy distribution of IRC as closely as possible [Jones et al. (1993)]. The TinyTim PSFs were then smoothed to match the images. Figure 2.1 shows one of the simulated PSF created by TinyTim. 1 The main page, along with the user guide v.6.3 which was used for this reduction, can be found at

19 Figure 2.1: TinyTim PSF generated for the 467 filter 10

20 11 The IRAF task lucy was used to deconvolve the images. Even with the reddened spectra, matching and removing the diffraction spikes (Figure 2.1 and Figure 2.2) for IRC was only partially successful. When the bright central area was fully subtracted, the diffraction spikes in the extended nebulosity were oversubtracted, creating artifacts. These were easy to recognize because the artifacts were aligned with the diffraction spikes, which were not aligned between the two epochs. To avoid creating artifacts number of iterations of lucy was reduced to ten. To more accurately remove the diffraction spikes in the regions of the knots and arcs that would be measuring, the innermost region was undersubtracted. Even with this care taken, the diffraction spikes were not completely removed (Figure 2.2). In the 547M long exposure combinations (547l1 and 547l2), this central area was completely saturated. It was replaced with the center of the corresponding short image (547s1 and 547s2 respectively), scaled to match the long exposure time, to simulate the correct profile for the deconvolution routine to remove the diffraction spikes. Figure 2.2 shows the oversaturated central region for filter 547 (left hand side - Before) and compares it to the reduced and replaced central region (right hand side - After). While the very center of the image was replaced and can t be used to measure motion, the outer areas were deconvolved and are much clearer in the final version of the image. A precision alignment was then done with the aid of field stars on the edges of the images. 2.3 Measurement Procedure The combined short and long images in each filter pair were then blinked using DS9 to identify features common to both epochs. Not all of the features were equally visible in both filters and in both epochs. The images could not simply be cross-correlated due to the non-uniform background and the residuals from the PSF subtraction. Each individual knot, condensation or arc would have required

21 12 Figure 2.2: 547 long before and after reduction. its own cross-correlation which would have been difficult with the variable background. Therefore the aligned images were blinked to find observable motion. The center point and position angle of each feature was then measured relative to the central star. This procedure was repeated for each filter combination where the feature was present. With short and long exposure in two filters, there were four images in both epochs. These measurements were repeated three times each, cycling through the four filter combinations in a different order. The measurements of each set were separated by several days to avoid measurement bias. Because of the complications described above with the PSF subtraction, no features near the residual diffraction spikes were measured, or any features within a radius of 0.33 arcseconds from the star to avoid confusion with artifacts created in the deconvolution of the star. The diffraction spikes were rotated by 16 between the two epochs, so none of the measured features were artifacts from the diffraction spikes in both epochs. These can be seen in Figure 1.2. The positional offsets in x and y between the two epochs were then determined for each feature. The angular offsets with the number of measurements are in

22 13 Appendix A, Table A.1. An example of the motion seen is shown in Figure 2.5 (image shows 0.36 arcseconds across); it was created by subtracting Epoch 2 image from the Epoch 1 image. The complex inner two arcsec region is shown in Figure 2.3 (Panels (a) and (b)) with individual knots and condensations identified. Figure 2.3: Inner and outer features labeled South East Outer Feature cross-correlation The Outer SE Feature (Figure 1.2, South East corner) was the one exception to the measurement procedure described above. This large, diffuse feature has no well defined bright center or condensation, so a cross-correlation was used to determine the offset. This worked for this filamentary feature because it is far from the central star and the diffraction spikes and has no significant background. It also has a distinct V shape that is easily identifiable between epochs and can be matched using cross-correlation. This was done twice with different sized regions ( 1.5 arcseconds and 2.5 arcseconds) around the feature to verify the cross-correlation. No other features in the outer regions are prominent enough to be measured this way.

23 2.3.2 An Explanation of the Headings for Table Table 2.3 lists the results of the basic measurements procedures described in section 2.3. The distance listed (Dist.) is the distance from the central star (in arcseconds) as measured in the Epoch 1 images. The position (Pos.) is the position angle relative to the central star. Values 0 to +180 describe the direction from North going counter-clockwise to South (through East in the images). Values 0 to -180 are the clockwise direction from North (through West) to the South, with the central point being the star (Figure 2.4(a)). Table 2.3 lists the transverse velocity only, which is based only on the measurements using these images. The value for φ is the direction of motion of the feature. The numerical values following the same rules as the Position Angle, except the center point is now the feature s position in Epoch 1. So the first two columns describe the exact position of the feature in Epoch 1, and the second two columns describe the motion observed between the epochs. The expansion age (Expan. Age) is the extrapolated time since the feature was at the position of the star based on its position and transverse velocity. All velocities are calculated assuming a distance of 5 kpc, and the errors quoted are 1 sigma errors from the position measurements. The features listed in bold italics are the features where there is also a radial velocity available, and they will be listed on subsequent tables. Table 2.3: Postion, Velocity and Direction of Motion of the Features Dist. Pos. V Trans φ Expan. Feature ID (arc) (deg) (km s 1 ) (deg) Age (y) SE Jet(1), Knot A ± 3 17 ± ± SE Jet(1), Knot B ± 3 27 ± ± SE Jet(2), Knot A ± ± ± SE Knot A ± 3 89 ± ± Continued on next page

24 Table 2.3 continued from previous page Dist. Pos. V Trans φ Expan. Feature ID (arc) (deg) (km s 1 ) (deg) Age (y) SE Knot B ± 2 42 ± ± SE Knot C ± ± ± SE Knot D ± 3 60 ± ± SE Knot E ± 2 54 ± ± SE Knot F ± 2 69 ± ± SE Knot G ± 2 98 ± ± SE Outer Knot ± 2 49 ± ± SW Jet, Knot B ± ± ± SW Square, Knot A ± ± ± SW Square, Knot B ± ± ± SW Square, Knot C ± 2 96 ± ± SW Square, Knot D ± 2 89 ± ± SW Knot A ± 3 41 ± 6-80 ± SW Triangle, Knot A ± ± ± SW Triangle, Knot B ± 2 15 ± 9-6 ± SW Triangle, Knot C ± 3 50 ± ± SW Fan, Knot A ± ± ± SW Fan, Knot B ± ± ± E Jet, Knot B ± ± ± E Jet, Knot C ± ± ± NE Knot G ± 2 94 ± ± Arc 1, Knot A ± 2 50 ± ± Arc 1, Knot B ± ± ± Arc 1, Knot C ± ± ± Arc 1, Knot D ± ± ± Continued on next page 15

25 Table 2.3 continued from previous page Dist. Pos. V Trans φ Expan. Feature ID (arc) (deg) (km s 1 ) (deg) Age (y) Arc 1, Knot E ± 2 69 ± ± Arc 1, Knot G ± ± ± Arc 2, Knot A ± ± ± Arc 2, Knot B ± 2 62 ± ± Arc 2, Knot C ± 2 87 ± ± Arc 2, Knot D ± 2 88 ± ± Arc 2, Knot E ± ± ± Arc 2, Knot F ± 2 97 ± ± Arc 3, Knot A ± 2 23 ± ± Arc 3, Knot B ± ± ± Arc 3, Knot C ± ± ± Radial Velocity Measurements and Total Velocity As previously mentioned in Chapter 1, radial velocities were available for some positions around IRC (Figure 1.1). The slit positions were matched to the WFPC2 images and then any overlapping features were picked out. Since the radial velocities were taken in 1999 [Humphreys et al. (2002)], the positions were matched to the Epoch 1 images (obtained in 1996). The sample size of the spectra extracted was 0.1 arcseconds 0.15 arcseconds. Since the movement between the two epochs ranged from 0.01 to 0.10 arcseconds over 11.9 years between epochs (Table A.1), the positions found in Epoch 1 were used to determine the position in age. 1 The transverse velocity is too small to find a meaningful expansion age. 2 This knot is very close to a diffraction spike which causes the high error the velocity and

26 17 the spectra. This assumes that the features had a roughly constant velocity over this time. Once the radial velocities were found for the corresponding identified features, the total velocity for the feature was calculated. These total velocities are listed in Table 2.4 along with the position relative to the plane of the sky (θ). Values of θ range from 0 (motion in the plane of the sky) to +90 (motion away from us) or to -90 (motion towards us). See Figure 2.4(b) for an example. The errors for θ come from the errors for the total velocity, which are dominated by the errors in the transverse motion measurement. The errors for the radial velocity were taken to be on the order of 5 km s 1. The subsequent expansion ages are calculated using the total velocity and shown in Table 2.4; there was very little change from the ages calculated with only the transverse velocity. All the velocities listed in Table 2.4 are in km s 1. Table 2.4: 3D Postion, Velocity and Direction of Motion V Trans V R V Total θ Expan. Feature ID (km s 1 ) ( ) ( ) (deg) Age (y) East Jet, Knot C ± ± SE Jet (1), Knot A ± ± SE Jet (2), Knot A ± ± SW Jet, Knot B ± ± SW Sq, Knot C ± ± SW Fan, Knot B ± ± Arc 1, Knot B ± 16 8 ± Arc 1, Knot C ± ± Arc 1, Knot G ± ± Arc 2, Knot F ± ±

27 18 (a) Orientation of φ (b) Orientation of θ Figure 2.4: (a), (b)

28 Figure 2.5: Example of motion for two features. 19

29 Chapter 3 Kinematics of the Ejecta Radial velocities, measured from the absorption minimum in the very strong Hα line are available for numerous positions along two separate slits from the long-slit spectra obtained with HST/STIS in 1999 (Figure 1.1 and from Paper III). The slits cross several of the brighter knots and arcs near the star, including one of the semicircular arcs. The observed spectrum is that of an expanding reflection nebula and the apparent Doppler velocity at each reflective condensation is due to the velocity of the star when it is observed directly, the expansion velocity of the nebula, and the relative velocity component along the line of sight of the condensation. Figure 9 in Paper III shows the measured Doppler velocity increasing with increasing distance from the star. The relation between the measured velocities and the three dimensional position can be fit by an expansion of 50 km s 1 (Paper III), consistent with the outflow velocities from the CO and OH observations and the double-peaked hydrogen and Ca II emission. Adopting this model for the expansion of the nebula, an estimate of the relative motions of the knots and arcs along the line of sight can be made. The relative radial component is combined with the measured transverse velocity to determine the total space motion and the combined direction of motion relative to the plane of the sky for individual knots. The results are included in Table 2.3. In the subsequent discussions, all of the expansion ages 20

30 21 or time since ejection are determined assuming a constant speed since ejection. For those objects with only a transverse motion, the expansion age is an upper limit. The resulting ages are identified and listed separately to differentiate them from the ages found from the total velocity. The features are discussed below, starting with the north east and continuing in a counter-clockwise direction (with the exception of the arcs). The individual knots can be seen listed in Figure East Jet and North East Knot G The East Jet (Figure 2.3, Panel b) is one of the more visible features in the ejecta. It appears to be a row of several knots aligned almost directly east of the star. In the first epoch image the knots are quite close to a diffraction spike, so the knots closest to the star were not measurable between the two epochs. Knots B and C have a transverse velocities of 165 km s 1 and 136 km s 1. Their respective times since ejection are 100 and 150 years, indicating that they are from a common recent ejection event, similar to SE Jet(2) Knot A. Knot C also has a small radial motion (-5.9 km s 1 ). It is thus moving at 136 km s 1 essentially in the plane of the sky at an angle of only 2.5 towards us. North East Knot G (Figure 2.3, Panel b) is the only knot that was measured north of the star. While there appear to be many condensations in this region, they are diffuse and faint, making positive identification between the epochs difficult. NE Knot G has a transverse velocity of 90 km s 1 and resulting expansion age of 200 years. 3.2 South East Jets SE Jet(1) (Figure 2.3, Panel b) appears to be moving relatively slowly, with transverse velocities of only 17 km s 1 and 27 km s 1 for Knots A and B, respectively. Knot A also has a small radial motion of -3.6 km s 1, relative to the star, for a total velocity of only 17 km s 1 and a time since ejection of 800 years. While

31 22 there is no radial velocity measurement for Knot B, with its close proximity to A, it is unlikely to have a significantly different radial velocity. Knots A and B are thus examples of slow moving ejecta. SE Jet(2) (Figure 2.3, Panel b) appears to be made up of multiple knots. However, their proximity to the central star makes measurement difficult. Knot A is the only one that was consistently recognizable as a distinct feature between images and is the closest feature to the star that was measured. Knot A also has a radial velocity that is much smaller than its transverse velocity. Its total velocity of 126 km s 1 gives an expansion age or time since ejection of only 80 years. 3.3 South East Knots With the exception of Knot A, the South East Knots (Figure 2.3, Panel a) are all at arcseconds from the star with position angles 137 to 152. They appear to be two kinematic groups. Knots C and G have transverse velocities of km s 1 with resulting expansion ages of years, while Knots B, D, E and F have significantly lower transverse velocities (42-69 km s 1 ) and a time since ejection of years. None of these knots have corresponding radial motions, so it is uncertain if they are in fact from a common event or from two different ejection events (at different times) and are spatially distinct but appear connected along our line of sight. South East Knot A (Figure 2.3, Panel b), much closer to the star (0.57 arcsec.), has a transverse velocity of 89 km s 1 and a resulting expansion age of 160 years. 3.4 South East Outer Feature The distant South East Outer Feature is 4 arcseconds away from the star (Figure 1.2). This feature s transverse motion was measured using cross-correlation (Chapter 2.3.1). Its transverse velocity of 49 km s 1 is consistent with the outflow and wind speed from the emission lines of km s 1 [Humphreys et al. (1997)]

32 23 and the expansion velocity of 40 km s 1 from the OH maser and CO observations [Bowers (1984), Knapp & Morris (1985)]. The corresponding expansion age is 2000 years. Nebulous material in one or more nearly spherical shells at about 5 arcseconds from the star is easily seen in the long exposures (Figure 4.1). Assuming the above expansion velocities they are probably due to material ejected about 3000 years ago. However, this was the only feature in the outer ejecta for which a transverse motion could be measured; the others were too diffuse and faint. 3.5 South West Square & South West Jet The South West Square is a grouping of 4 knots that are arcseconds from the star (Figure 2.3, Panel a) with position angles from -125 to Knots B, C and D have similar velocities ( km s 1 ) as well as similar directions of motion (from -160 to -174 ) which are more or less radially away from the star. Their expansion ages from their transverse motions range from years. Knot C also has a radial velocity of km s 1 which gives a total velocity of 98 km s 1 and an expansion age of 320 years. These results suggest a common ejection event for these 3 knots. Knot A has a velocity of 131 km s 1, similar to Knots B-D, but its apparent direction of motion (φ = 133 ) is peculiar. It is unclear if this unexpected direction is due to a projection effect or interaction with ejecta. The South West Jet only has one measurable knot. It is moving radially outward at 218 km s 1 with an expansion age of only 70 years. Its small associated radial velocity of -5.6 km s 1 does not significantly alter its space motion or time since ejection. 3.6 South West Triangle & South West Knot A The South West Triangle is formed by three knots (Figure 2.3, Panel a); however, their different kinematics suggest that this is not a physically associated grouping.

33 24 Knot A has a transverse velocity of 214 km s 1 and is moving nearly radially away from the star (φ = -145 ) with a corresponding expansion age of 140 years while Knot C is only moving at 50 km s 1 at an angle of -109 and an ejection time of 700 years. Knot B s transverse velocity is comparable to the error in the measurements so no expansion time is assigned. It is possible that these features could be expanding from a common center, similar to the semi-circular arcs discussed in Chapter 3.8. South West Knot A (Figure 2.3, Panel a), located between the SW Triangle and the SW Square, has a transverse velocity of 41 km s 1. Its corresponding time since ejection of 800 years suggests it may be associated with the SW Triangle (if it is an expanding feature as mentioned above) or at least with Knot C if they are not a physically associated group. 3.7 South West Fan The scalloped or rippled pattern to the SW of the star that (called the fan) is very prominent in the images of IRC (Figure 1.2). The complexity of this region makes it hard to compare the diffuse condensations between the two epochs with a high degree of confidence, however. Knots A and B (Figure 2.3, Panel a) are sufficiently distinct and bright enough that they can be identified in the images from both epochs. The knots have essentially the same transverse velocities ( 100 km s 1 ) and a corresponding expansion age of years. Both knots are moving nearly radially away from the star. Knot B also has a radial motion of km s 1 yielding a total velocity of 104 km s 1 and an ejection time of 450 years. The fact that these knots have such similar properties, despite being separated by 1.5 arcsec, suggests that the fan is a physically distinct feature from a single mass loss event

34 3.8 The Semi-Circular Arcs 25 Three very intriguing nearly circular arcs can be easily seen to the east of the star (Figure 1.2). The arcs appear to be made up of several small knots, identified in Figure 3.1, that form larger arcs or loops. Given their appearance, it is possible that these knots may have been part of three initially much more compact features, and the nearly circular arcs are actually expanding bubbles or loops. This is explored in more detail and discussed in Chapter Initially, the motions of the individual knots were measured with respect to the star as was done for the other condensations discussed in this chapter and in Table 2.3. The six knots in Arc 1 have a mean transverse velocity of 97 km s 1 and a corresponding mean time since ejection of 400 years. Given its location, Knot A may be a separate condensation simply projected onto Arc 1. Excluding it gives 103 km s 1 and 370 years. Radial velocities are available for three of the knots (Table 2.4) giving a mean total velocity of 111 km s 1 and an expansion age of 320 years. The three knots are consistent with a mean orientation that places them essentially in the plane of the sky, although there may be a slight tilt to the arc. In Arc 2, Knot A appears to be on the inside edge of the Arc making it difficult to measure and placing some uncertainty on its association. Excluding Knot A, the other five knots give a mean transverse velocity of 91 km s 1 and a corresponding mean time since ejection of 500 years. Knot F has an associated radial velocity and a corresponding expansion age of 400 years. Arc 3 is the least well-defined in the images, and only three knots were measured. Knots B and C in Arc 3 give consistent results with a mean transverse velocity of 150 km s 1 and an expansion age of 300 years (Table 2.4). The arcs are all about the same radial distance from the star, and the results for the individual knots suggest that they may all have been ejected at about the same time, years ago.

35 26 Figure 3.1: Image 547l1 with arcs labeled The Arcs as Expanding Bubbles To investigate whether the arcs are expanding bubbles, the motions of the knots with respect to the center of their respective arcs was measured. The center of each arc was found by fitting an ellipse to the positions of the knots in DS9. Then, different ellipticities were experimented with and the best fit to the measured postions of the knots that also best mapped the inter-knot diffuse nebulosity was adopted. This was done independently for each epoch. The percentage difference between the positions of the knots and the elliptical fit was used to determine the quality of the fit. The center of the best-fit ellipse was then adopted as center

36 of each arc. summarized in Table The corresponding transverse motion between the two epochs is 27 While a best fit was found, the adopted error for the velocity of each arc center is from the range in the transverse velocity derived from the different fits. These values are listed in Table 3.1. Radial velocities are available for three knots in Arc 1. Consequently a radial velocity of the arc center was calculated assuming that the velocities of the knots, relative to the star, depend on the radial velocity of the center and their x and y positions relative to the arc center. The resulting radial velocity for the center of Arc 1 is -0.4 km s 1 (which is essentially zero km s 1 ), from both Epochs 1 and 2, giving a total velocity of 51 km s 1 and a time since ejection of 700 years. This puts the arc center basically in the plane of the sky as well. Table 3.1: Arc Center Point Fit Information Position V Trans φ Expansion Arc (deg) (km s 1 ) (deg) Age (yr) 2 Arc 1 64 ± 5 51 ± ± Arc 2 84 ± ± ± Arc ± ± ± Arc 1 With this information for Arc 1, a transverse motion was determined for the individual knots relative to the arc center. For knots B, C, and G, which had associated radial velocities, the total motion and orientation or tilt relative to the arc center was found. These results are summarized in Table 3.3. The results confirm that Arc 1 is slightly tilted with respect to the line of sight. Also, the 1 In Arc 2, Knot A was not used to determine the best-fit position and was not included in the percentage difference calculations. 2 This expansion age is for the arc center point from the star, and not for individual knots of the arc.

37 28 expansion age of the arc was found to be 100 years; this time is significantly less than the time since ejection from the star whether the 700 years for the arc center (Table 3.1) or the 320 years determined from the three arcs with total motions (Table 2.4) is used. Table 3.2: The 3D Motion of Arc 1 Relative to its Center Arc 1 Pos. V Trans φ V R θ 1 V Total ID (deg) (km s 1 ) (deg) (km s 1 ) (deg) (km s 1 ) Knot A -132 ± 2 23 ± ± Knot B -66 ± 2 79 ± ± ± 3 81 ± 13 Knot C -4 ± 2 99 ± ± ± 3 99 ± 14 Knot D 69 ± 3 44 ± 12-20± Knot E 88 ± 2 24 ± ± Knot G 169 ± 3 74 ± ± ± 3 74 ± Arc 2 The transverse motion for the Arc 2 center gives a time since ejection of years. Only Knot F in Arc 2 has a measured radial velocity; consequently, there is no radial velocity estimate for the arc center. The transverse motions for the individual knots give an average expansion time for the arc of 200 years. Although there is a large spread in the results for the knots (Table 3.3), the results for the expansion of the arc itself and the ejection times for the knots are consistent within the measurement uncertainties. 1 These errors are based solely on the measurement values and don t include errors from the measurement of the arc center.

38 3.8.4 Arc 3 29 Arc 3 has an expansion age with respect to the star of years from its transverse motion. No radial velocity information is available for this arc. The expansion time of 50 years for the knots relative to the arc center (Table 3.3) is significantly less than the time since ejection from the star (Chapter 3.8.2), although motions are available for only three knots. Except for knot A, the expansion times relative to the star from the transverse motion of the other two knots (Table 2.3) are generally consistent with the result for the arc center Summary Thus, for the semi-circular arcs there is evidence that the expansion time for the arc itself is less than the time since its ejection from the star (seen for both Arc 1 in Chapter and Arc 3 in Chapter 3.8.4). The possible causes of these different expansion times are discussed in Chapter 5. So in summary, the features (knots, arcs, etc.) are kinematically distinct from the general expansion of the ejecta. The mean total space motion for the 10 knots in Table 2.4 is 113 ± 14 km s 1 compared to the km s 1 velocity of expansion inferred from the double-peaked profiles of the hydrogen and Ca II emission and the maser emission. Their mean space motion is due almost entirely to the mean transverse velocity of 112 ± 15 km s 1 compared to only -5 ± 3 km s 1 for the radial motion. Similarly, the mean transverse velocity for the remaining features in Table 2.3 is 89 ± 9 km s 1. In almost all cases with a radial velocity, the transverse motions are significantly larger than the radial component. The motions of the discrete features are thus dominated by their transverse motions.

39 Table 3.3: Values Relative to the Arc Center Point 30 Dist. Pos. V Trans φ Expan. Feature ID (arcs.) (deg) (km s 1 ) (deg) Age (yr) Arc 1, Knot A ± 2 23 ± ± Arc 1, Knot B ± 2 79 ± ± Arc 1, Knot C ± 2 99 ± ± Arc 1, Knot D ± 2 44 ± ± Arc 1, Knot E ± 2 24 ± ± Arc 1, Knot G ± 3 74 ± ± Arc 2, Knot A ± 3 21 ± ± Arc 2, Knot B ± 3 83 ± ± Arc 2, Knot C ± 3 31 ± ± Arc 2, Knot D ± 3 20 ± 12-5 ± Arc 2, Knot E ± 3 42 ± ± Arc 2, Knot F ± ± ± Arc 3, Knot A ± ± ± Arc 3, Knot B ± 3 85 ± ± Arc 3, Knot C ± 3 68 ± ±

40 Chapter 4 Geometry of the Ejecta The morphology of IRC s circumstellar ejecta has eluded previous studies. While the outer rings are consistent with a basically spherical outflow from several thousand years ago, perhaps when the star was in a different evolutionary state such as a red supergiant, the inner ejecta is very complex and indicative of localized ejection events. Suggestions for the orientation and geometry of the ejecta have ranged from an equatorial disk viewed at an angle to our line of sight (Paper I), a bipolar outflow [Oudmaijer et al. (1994), Humphreys et al. (2002), Davies et al. (2007)]) to an essentially spherical outflow (Paper III). Similarly, the various interpretations based on the strong OH maser emission include a spherical outflow [Bowers (1984)], a bipolar outflow with a disk-like structure viewed edgeon [Diamond (1983)], and a weakly bipolar, slightly oblate outflow with clumping [Nedoluha & Bowers (1992)]. The results for the total space motions and resulting orientation for several of the discrete knots and condensations (Table 2.4) interestingly show that while they have a range of ejection velocities, and expansion ages they are all moving very close to the plane of the sky or towards us by at most about With an 1 The transverse velocity, expansion age, and orientation depend on the adopted distance of 5 kpc. A possible distance range of 4 6 kpc corresponds to an uncertainty of up to 20% in the transverse velocity. The corresponding uncertainty in the time since ejection and the angle θ relative to the plane of the sky is small and significantly less than that due to the measurement 31

41 optical thickness greater than one for IRC s inner ejecta (Paper II), it is not surprising that there is no feature found moving away. The similar orientation of most of these features, at different position angles and ejected at different times over several hundred years, suggests that our view of IRC is nearly pole-on and therefore, we are looking nearly directly down onto its equatorial plane. The SW fan, which extends over an arc covering approximately 70 projected onto the sky, is a likely candidate to represent the equatorial plane which would then be tilted by only about 8 out of the plane of the sky with the southwest side of the ejecta towards us. The other features, the various knots and the semi-circular arcs, would then all lie within 10 of the equatorial plane. Furthermore as noted above, the motions of the various knots etc are dominated by their transverse velocities. They have little radial motion, supporting the interpretation that we are viewing these features essentially face-on. [Davies et al. (2007)] have presented evidence in support of a bipolar outflow based on the velocities of the reflected Hα and Fe II emission across the nebula and argue for a preferred axis of symmetry at 45 on the basis of the optical and infrared images. The Hα and Fe II emission however show different kinematic patterns. The Fe II does not show a clear bipolar pattern; it has radial gradient with lower velocities near the center and higher in the outer parts. Hα shows the strongest evidence for bipolarity with lower velocities to the SW and higher velocities to the NE with a typical velocity difference of 10 km s 1 within the inner region; the strongest evidence for an outflow is beyond the inner 2 arcseconds The observations were checked for any evidence for an axis of symmetry in the motion of the knots especially between the SW and SE/NE quadrants of the nebula. 2 For those few knots with radial motions, there is a small difference of 12 km s 1, with the SW quadrant having a larger component of motion towards us. Given the apparent orientation of the ejecta, this is not surprising and may be due as much to the tilt of the SW side towards our line of sight as to a bipolar outflow. errors. 2 The semi-circular arcs were not included because of the possible expansion. 32

42 33 Given the nearly pole-on geometry of the inner ejecta, the motions and orientation of the various arcs and knots within about 2 arcseconds do not provide any direct information on an axis of symmetry or bipolar outflow, and the apparent velocity difference in the Hα emission with position in the inner nebula may also be due as much to its geometry as to an actual bipolar outflow. In previous papers on the extreme red supergiant VY CMa [Smith et al. (2001), Humphreys et al. (2005), Humphreys et al. (2007)] it has been emphasized that the presence of prominent arcs, knots and large loop-like structures in its visible ejecta are all evidence for localized ejections which could be due to large-scale surface activity and magnetic fields. There are both similarities and important differences between the circumstellar environments of VY CMa and IRC , even though they are seen from different perspectives. 3 IRC s ejecta is apparently concentrated to the equatorial plane and it does not show the large prominence-like loops seen in VY CMa; however, the semi-circular arcs may be evidence for related structures. The Northwest arc seen in VY CMa [Humphreys et al. (2007)] would be at least 2-3 times larger than the arcs of IRC if VY were at the same distance of 5 kpc. 3 VY CMa may be tilted to the line of sight.

43 Figure 4.1: Image 675ex scaled to show faint nebulosity. 34

44 Chapter 5 Discussion, Conclusions and Future Work 5.1 Discussion Mass Loss History The circumstellar ejecta of IRC separates into the outer approximately spherical shells 5-6 arcseconds (Figure 4.1) from the star and the complex inner ejecta within 2 arcseconds of the star (Figure 1.2). Adopting the nominal expansion velocity of km sec 1, the outer shells were ejected about 3000 years ago. The transverse motion for the outer nebulosity about 4 arcseconds away confirms an expansion age of 2000 years. There is also evidence for more distant associated ejecta 8-9 arcseconds away (Figure 4.1). This visual nebulosity very likely corresponds to the distant arc reported by [Kastner et al. (1995)] from coronagraphic imaging in the near-infrared. Assuming that this material was expelled with a similar velocity, it would have been ejected years ago. These outer shells are similar to the shells or ejecta associated with many post-agb stars and very likely result from pulsational mass loss as a red supergiant in the case of IRC as seen in models [Heger et al. (1997), Yoon & Cantiello (2010)]. In contrast, the inner circumstellar material was apparently ejected at different times and in different directions beginning about years ago up to fairly 35

45 36 recently ending perhaps less than 100 years ago. This result is consistent with other studies suggesting that a high mass loss period ended recently (Blöcker et al. (1999), Paper III). Ejecta younger than years would be within the PSF dominated region and would not be included in these measurements, so they neither support or disprove a recent termination of the current mass loss episode. Thus, there are at least two major epochs of mass loss separated by 1000 years. Furthermore, the results for the inner ejecta suggest that the more recent period was punctuated by times of increased activity. Many of the features have expansion ages corresponding to years ago and again from about years ago. This supports the idea that this recent episode of mass loss has a different mechanism driving it. As discussed in Chapter 3.8, the semi-circular arcs may be expanding as well as traveling away from the star. An ejection time of years is consistent within the errors for all three arcs, although it may be as high as 700 years for Arc 1. The results for Arcs 1 and 3 also indicate expansion times much less than the time since ejection from the star. This could be due to the significant uncertainty in these measurements, although the results for the knots in Arc 1 are internally consistent. As discussed in the next section, magnetic fields may play a role in the origin of IRC s episodic mass loss. They may also provide an explanation for the much shorter measured expansion times for the arcs. The strength of the magnetic field is estimated from the circular polarization of the OH masers on the inside edge of the maser shell at about 7000 AU from the star [Nedoluha & Bowers (1992)], the average distance of the arcs (Figure 5.1 the inner ring of OH at 1.5 arcsec.). Suzuki (2007) has demonstrated that it is possible for hot bubbles ejected from red supergiants to last much longer than their cool down times due to extended magnetic field lines from the star which constrain the bubbles of gas as they travel away from the star. While the models do not include stars as warm as IRC , magnetic fields may restrict the expansion of a bubble or loop and account for the delayed start to the internal expansion of the arcs. Given the apparent tilt of Arc 1, it also seems likely that

46 the arcs are actually expanding loops similar to the much larger structures in the ejecta of VY CMa Concluding Remarks & Additional Work The morphology of IRC s inner ejecta shows a recent mass loss history dominated by localized ejection events in random directions (near the plane of the sky) and at different times. Mass loss and the winds of cool evolved stars including the AGB stars, red supergiant and red giants are normally attributed to global pulsation combined with radiation pressure on the dust which further drives the mass loss. These mechanisms can account for relatively uniform, essentially spherical ejecta like the outer shells of IRC (Figure 4.1). But the complex and episodic mass loss evident in the inner regions require a non-uniform mass loss mechanism such as large-scale surface activity. There is now an increasing number of observations of starspots, large surface asymmetries, and outflows associated with red supergiants, giants and AGB stars consistent with a convective origin [Tuthill et al. (1997), Monnier et al. (2004), Kiss et al. (2009)]. Non-radial pulsations may be an alternative but are not consistent with the narrow looplike structures observed, in the ejecta of VY CMa or IRC Furthermore, magnetic fields associated with the maser emission are now confirmed in the ejecta of several of these stars including the strong OH/IR sources, VX Sgr, S Per, NML Cyg, and VY CMa [Vlemmings et al. (2002), Vlemmings et al. (2004)]. Magnetic field strengths of from 0.17 to 15 mg have been reported in the ejecta of IRC from the observed circular polarization of the OH maser emission at 1.5 arcseconds [Reid et al. (1979), Cohen et al. (1987)] and [Nedoluha & Bowers (1992)]. Figure 5.1 shows the distribution of the various maser sources superimposed on an image of IRC The inner OH emission at 1.5 arcseconds is the location of the circularly polarized emission and is coincident with the inner ejecta. Adopting a conventional extrapolation for the relationship between the magnetic field (B) and the distance from the star (r)

47 38 of B r 2, a 1 mg field at r AU would give B 3 kg at the stellar surface, high for a global field, and it would also exceed other local energy densities (Paper III). A field proportional to r 1, however, would give a surface magnetic energy density comparable to the thermal energy density. Also noted is that in the case of IRC , the knots and arcs appear to be concentrated in the equatorial region. As a star evolves to warmer temperatures, the increased dynamical instability will be enhanced in the equatorial region as the star s rotation also increases. Thus for IRC this may be observing the combined effects of turbulence/convection and increased rotation, leading to an equatorial preference for the recent mass loss. The case for the role of convection and magnetic fields on the mass lass history and mechanism in evolved massive stars is of course strongest in the red supergiant stage. IRC is a post-red supergiant and the inner ejecta were formed long after it had been a red supergiant. However, it is also close to the upper luminosity boundary for evolved stars in the H-R diagram and to the critical temperature regime, K (de Jager 1998), where dynamical instabilities become important for stars evolving to warmer temperatures. In summary, the evidence for episodic mass loss events associated with convective/magnetic activity is visible in the resolved circumstellar ejecta of objects like VY CMa and IRC with their numerous knots and arcs ejected at different times from separate regions on the star s surface. The role of these dynamical instabilities on the mass loss histories of the most luminous evolved stars must be considered as a probable source of their high mass loss episodes, especially in their final stages. To determine a more complete 3 dimensional morphology of the ejecta around IRC , high resolution spectroscopy is needed over a larger region of the inner 5 arcseconds. This would complement the transverse velocities found here, and give a more complete description of the motion of the ejecta.

48 Figure 5.1: Image 547l2 with maser emission marked 39