The Hidden Clumps in VY CMa Uncovered by the Atacama Large Millimeter/submillimeter Array

The red hypergiant VY CMa is famous for its very visible record of high-mass-loss events. Recent CO observations with the Atacama Large Millimeter/submillimeter Array (ALMA) revealed three previously unknown large-scale outflows (Singh et al). In this paper, we use the CO maps to investigate the motions of a cluster of four clumps close to the star, not visible in the optical or infrared images. We present their proper motions measured from two epochs of ALMA images and determine the line-of-sight velocities of the gas in emission at the clumps. We estimate their masses and ages, or time since ejection, and conclude that all four were ejected during VY CMa’s active period in the early 20th century. Together with two additional knots observed with the Hubble Space Telescope, VY CMa experienced at least six massive outflows during a 30 yr period, with a total mass lost ≥0.07 M ⊙. The position–velocity map of the 12CO emission reveals previously unnoticed attributes of the older outer ejecta. In a very narrow range of Doppler velocities, 12CO absorption and emission causes some of this outer material to be quite opaque. At those frequencies the inner structure is hidden and we see only emission from an extended outer region. This fact produces a conspicuous but illusory dark spot if one attempts to subtract the continuum in a normal way.

Mass loss is observed across the upper Hertzsprung-Russell (H-R) diagram and alters the evolution of the most massive stars.It may be slow and continuous or occur in more dramatic high-mass-loss events, which may be irregular or single giant eruptions.In the red and yellow supergiants, these mass-loss events may determine their final fate, whether as supernovae or direct collapse to a black hole.
The red hypergiant VY CMa is one of the most important evolved massive stars for understanding the role of high-massloss episodes on the final stages of the majority of massive stars which pass through the red supergiant (RSG) stage.Its massloss history is highly visible in optical and infrared images, with prominent arcs and filaments and clumps of knots throughout its extended diffuse ejecta.Doppler velocities of K I emission, together with proper motions of the discrete ejecta (Humphreys et al. 2005(Humphreys et al. , 2007(Humphreys et al. , 2019) ) demonstrated that the arcs and clumps of knots are spatially and kinematically distinct from the diffuse ejecta and were expelled in separate events over several hundred years, with several episodes in the last century (Humphreys et al. 2021).
New high-resolution images with the Atacama Large Millimeter/submillimeter Array (ALMA; Singh et al. 2023, hereafter Paper I) reveal a more extended diffuse ejecta with at least three previously unknown prominent large arcs or outflows observed in CO and HCN emission.The addition of three major outflows expands our record of its high-mass-loss episodes and frequency, and presents additional challenges to possible models for their origin in VY CMa and similar luminous RSGs.
ALMA Science Verification submillimeter and continuum emission images of VY CMa at 321, 325, and 658 GHz (Richards et al. 2014;O'Gorman et al. 2015) earlier revealed an additional, large, potentially massive clump or knot close to the star.It is not visible in the optical Hubble Space Telescope (HST) images.Designated "Clump C" by Richards et al. (2014), it is optically thick at these frequencies, with an estimated dust mass of 2.5 × 10 −4 M e , or 5 × 10 −2 M e assuming a gas-to-dust ratio of 200.Using the Science Verification data for Band 5 at 163-211 GHz, Vlemmings & Khouri (2017) concluded that the dust in Clump C was optically thick even at 178 GHz, and estimated a dust mass of > .3× 10 −3 M e or a total mass (dust + gas) of more than 0.1 M e , making it potentially the most massive ejecta in VY CMa.
It is not surprising that Clump C has added to questions about VY CMa's mass-loss history and its mass-loss mechanism.Using additional ALMA observations in Bands 6 and 7 with a higher angular resolution of 0 1, Kaminski (2019) identified three additional features or smaller clumps clustered near Clump C and the star; see Figure 3 in Kaminski (2019) and our Figure 1.All are dusty, and he comments that the dust mass for Clump C may be even higher than suggested by Vlemmings & Khouri (2017), but notes that porosity would allow a lower dust mass.For example, with higher-resolution ALMA images, Asaki et al. (2020) show that Clump C is resolved into many small condensations with a range of sizes and intensities.
Despite its significance, there is much we do not know about Clump C and associated knots such as their motions, spatial orientation with respect to VY CMa, and when they were ejected.Do they represent a single massive eruption or are they from separate events over a period of time?In this paper, we report on the proper motions of Clump C and its neighbors measured from two epochs of continuum images observed with ALMA, and the motions of the CO gas in emission at the clumps.Hereafter, to distinguish them from the numerous knots, clumps, etc., observed in the optical images, we refer to them in this paper as the ALMA clumps.
In Section 2, we describe our observations and propermotion measurements.In Section 3, we discuss the line-ofsight velocities of the CO emission at the four ALMA clumps.Section 4 summarizes their total motion, orientation, and ages or time since ejection as well as the relation to VY CMa's other recent mass-loss events.In Section 5, we estimate the masses of the clumps and review the energies involved in these highmass-loss events.The outer ejecta revealed by the CO emission and its obscuration at Clump C is discussed in Section 6.The final section gives a review of questions about VY CMa's evolutionary state.

Data and Proper-motion Measurements
Our high-resolution observations, obtained with ALMA in Band 6 covering frequencies 216-270 GHz, are described in Paper I. 6 In this paper, we use the continuum image at 249 GHz and a CO image cube with a spatial resolution of 0 2 and a velocity resolution of 1.25 km s −1 .The observational parameters are described in Table A1 in Appendix A. Our continuum image at 249 GHz is shown in Figure 1. Figure 2 shows the 12 CO emission at the clumps in two representative channels.The feature labeled B¢ is an outflow not mentioned in previous papers, and is not present in the continuum images.7 Also note that the brightest 12 CO emission is not centered on the position of VY CMa.This apparent offset to the east is observed at all the frequencies in our image cubes with bright CO emission from the star's circumstellar ejecta.The peak of the 12 CO emission-line profile at VY CMa, shown in Appendix B, has a blueshift with respect to the star's expected LSR velocity of 22 km s −1 .Thus, a cloud of CO gas (and dust) may be asymmetric and probably expanding relative to VY CMa.
The data-reduction steps and calibration are described in Paper I and its Appendix.A descriptive summary is given in Appendix A.

Proper Motions of the Clumps
To measure the proper motion of Clump C and the other knots, we compare their positions relative to VY CMa in our continuum image from 2021 December 9 with an earlier image.The closest data set to ours in terms of frequency and resolution with a decent timescale is from Kaminski (2019), project 2013.1.00156.S from 2015 September 27, giving a 6.2 yr baseline.The continuum images used for the proper motions, the measurement method, and uncertainties are described in the Appendix B.
The separate measurements for 2015 and 2021 are given in Table 1 for the four clumps.Their proper motions, positions, and derived parameters are summarized in Table 2.For example, for Clump C, the continuum images yield separations of 314 mas in 2015 and 341 mas in 2021 with respect to VY CMa.The increase in angular separation of 23 ± 2 mas at a distance of 1.15 kpc is 26.5 ± 2.8 au.8In 6.2 yr, the proper motion is 3.7 mas per year, and the transverse velocity in the plane of the sky is thus 20.2 ± 2.3 km s −1 .The angular change in decl.and R.A. between 2015 and 2021 gives the direction of motion (f) for Clump C, 102°, measured north through east.The same parameters are included in Table 2 for the other clumps.
Figure 3 is a map showing the measured proper motions and the direction the clumps are moving.Clumps C and B have a direction of motion from their proper motions that do not point radially back to the star.The uncertainties in f, the direction of motion in Table 2, however, are within 3σ of the position angle for Clumps C and B. Other factors may influence the direction the clumps appear to be moving such as the overall asymmetry of the circumstellar ejecta near VY CMa and the role of magnetic fields; see Section 5.

Line-of-sight Velocities toward the ALMA Clumps
At 200 mas resolution, the CO emission in this region is partially resolved into small condensations or features of a size similar to or slightly larger than the beam.Figure 2 shows that Clumps C and B are not well separated from the bright 12 CO emission from VY CMa's circumstellar ejecta.Clump D, however, appears as a small emission peak or spot and is spatially separate from the circumstellar 12 CO emission in several channels.Clump A is also visible in several blueshifted channels as a relatively bright CO emission spot marginally separate from VY CMa.We find that C and D are optically thick in the continuum but A and B are optically thin (Section 5).As discussed below, Clumps C and D together mark an interesting continuous structure or outflow.
In Table 3, we summarize the range of Doppler velocities measured toward each clump in the 12 CO and 13 CO image cubes.The velocities relative to the LSR are measured with respect to their respective rest frequencies.VY CMa has a LSR velocity of 22 ± 1 km s −1 based on OH and SiO maser emission from numerous sources.The velocities toward the ALMA clumps demonstrate that clumps C and D are redshifted, moving away from and projected behind VY CMa, while clump A is blueshifted relative to the star.Clump B is more uncertain.Kaminski (2019) reported a wider range of velocities from molecular emission from the clumps, which can be seen in the line profiles in Appendix C, but Figure 5, discussed below, shows that the higher positive and negative velocities are from extended and diffuse separate features.
Although the CO emission at Clump D appears to be spatially separate from the emission associated with Clump C and the stellar envelope in Figure 2, the continuum image suggests an outflow to the southeast of the star with D at its outer extremity.Figure 4 shows the summed 12 CO image over 28 frequency channels covering a range in V LSR from +49 to +15 km s −1 .It clearly illustrates the continuous emission structure to the SE of the star, although the separate channels show that C and D represent different parts of this elongated feature.
In each outflow event there is usually a strong correlation between the Doppler velocity and distance from the star.Hence, a position-velocity slice at an oblique angle in the image cube can reveal the local morphology discussed here.Figure 4 illustrates the procedure.The relevant plane in the image cube is defined by position angle 120°, which samples both C and D with our spatial resolution.At each frequency channel, we measure the intensity of the signal at positions along a line or slit at 120°± 0.05 arcsec.The extractions for each frequency are stacked to produce a two-dimensional map where one axis is spatial and the other is Doppler velocity.The resulting position-velocity map is shown in Figure 5 for the LSR velocity range −20 to +80 km s −1 and reveals the morphology of 12 CO emission associated with Clumps D and C.
Figure 5 also reveals bands of whispy or "cirrus-like" emission stretching horizontally across the lower half and top of the figure.Based on their brightness and size scale the "cirrus" bands represent older ejecta from VY CMa moving at ≈± 30-40 km s −1 relative to the star.A prominent emission feature is also visible nearly aligned in position with the brighter star at ≈+30 to +40 km s −1 , at r  400 au.The dark band at about −3 km s −1 that appears to cut off the extended CO emission is due to optically thick outer ejecta, discussed in Section 6.

Outflow Velocities of the Clumps
The prominent diffuse emission feature from V LSR +45 to about +25 km s −1 in Figure 5 extends from Clump C to D. Clump C is visible as the bright spot at the top of the arc marked in Figure 5 and, based on its position, Clump D is near the tip of the arc. Figure 5 shows an arc-like diffuse feature with increasing velocity with increasing distance from VY CMa, presumably due to an outflow from the star.Thus, we suggest that there is an apparent velocity gradient in the 12 CO emission toward Clump D with distance from VY CMa.
Figure 6 shows the Doppler velocities relative to the star, V rel* , of the small separate 12 CO emision spots viewed toward Clump D (see Figure 2) as a function of projected distance from the star.Each point is the centroid position of the local emission in a particular frequency channel.Figure 6 illustrates what we see in Figure 5 projected onto the plane of the sky.These points fall along a remarkably smooth curve.The high internal consistency is not due to subjective factors, since we employed objective algorithms and the chosen data points were determined by the frequency channels (see Appendix C).
Together, Figures 5 and 6 demonstrate an outflow of gas with a consistent gradient of increasing velocity with distance from the star that is not linear, suggesting an arc-like outflow with a maximum distance from the star, with no substantial 12 CO emission at farther distances.Figure 6 also shows what may be a hook or loop at the very tip of the arc, albeit traced by the more diffuse fainter emission plotted as open circles.Thus, the C-D feature may be part of a loop, resembling VY CMa's well-known large stuctures, but smaller and younger.The 12 CO emission may be analogous to the K I emission observed behind the leading edges of the large expanding arcs (Humphreys et al. 2005).
Clump D is near the tip of the arc and is optically thick in the continuum (see Section 5), so we do not expect to see emission directly beyond it.It may be part of the outermost loop segment, which is nearly tangential to our line of sight.The maximum velocity of CO emission relative to the star at this position is approximately 22 km s −1 .This velocity coincides with Clump D's projected angular distance at 0 73 in Tables 1  and 2, marked by an arrow in Figure 6.Consequently, we suggest that this is the Doppler velocity of the 12 CO gas at or nearest to Clump D and adopt it for Clump D's line-of-sight velocity V rel* , with an uncertainty of ±3 km s −1 based on the range in velocities near Clump D's position.
Clump C is also optically thick.It is located at the top of the diffuse arc in Figure 5, near the position of the star at LSR velocities +25 to +30 km s −1 and at ≈0 4 from VY CMa corresponding to its position in the continuum image (Table 1).The line profile (Appendix C) for Clump C shows a small emission peak at a LSR velocity +26 km s −1 , which we adopt for its line-of-sight velocity corresponding to +4 km s −1 , V rel* , with an uncertainty of approximately ± 3 km s −1 based on its position in Figure 5 and the width of the emission peak in the line profile.
Clumps A and B are apparently in an outflow extending to the northeast of VY CMa, similar to the diffuse emission arc including Clumps C and D to the southeast (Figure 1).Based on its position angle and projected angular distance from VY CMa, Clump A is very likely associated with the northern extension reported by Richards et al. (2014) andO'Gorman et al. (2015).We observe blueshifted CO emission close to the star beginning at a LSR velocity +19 km s −1 with a position angle of about 25°, which we associate with this outflow, and identify Clump A with the 12 CO emission marginally resolved from the star with a LSR velocity range of +15 to −4 km s −1 .The blue edge of its emission profile in Figure 10 is cut off by the absorbing ejecta (Section 6) at LSR velocities -4 km s −1 .The 12 CO emission in the image cubes agrees with the position of Clump A for the LSR velocity range ≈+8.6 to −1.6 km s −1 .Adopting the middle of this range at +3.5 km s −1 , the V rel* is −18 ± 5.0 km s −1 .
Clump B is more complicated.It is not resolved from the extended circumstellar emission surrounding the star (Figure 2).It can be seen as an extension on the stellar envelope coinciding with its position in the continuum over a small range of redshifted velocities.Its line profile in Figure 10, however, shows a very broad emission envelope with two peaks on either side of the star's velocity plus a peak near it and the velocity of Clump C. Clump B is not distinguishable at velocities blueshifted relative to the star.The emission peak at ≈+8 km s −1 overlaps with emission from the northeast outflow/Clump A at the same velocity.Clump B cannot be separated from the emission from the circumstellar ejecta coinciding with the second peak near the stellar velocity.Thus, we tentatively adopt the LSR velocity of the redshifted emission peak at +35 km s −1 .The V rel* is +13 km s −1 ± 7 with the error from the range in the velocities in Table 3.We consider this to be very uncertain.

Morphology and Ages of the ALMA Clumps
The velocities are summarized in Table 4 with the total or expansion velocity relative to VY CMa, the orientation or projection angle (θ) with respect to the plane of the sky, the distance from VY CMa corrected for the projection, and an estimate of the age or time since ejection measured from the proper motions.The reference epoch for the proper motions, The age estimates assume no acceleration or deceleration.We discussed the effect of acceleration or deceleration in Humphreys et al. (2021) and showed that the effect was minimal, on the order of 5%-10%, less than the errors in the estimated ages.We find that Clump C is nearly in the plane of the sky, as suggested by O' Gorman et al. (2015) and Kaminski (2019).The transverse velocities for Clumps D and A are rather high for knots and condensations close to the star, which have typical velocities of 20-30 km s −1 (Humphreys et al. 2019(Humphreys et al. , 2021)), but both have large uncertainties.These uncertainties from the proper motions dominate the error in the age estimates, the expansion velocities, and subsequent derived parameters.For reference, the escape velocity for VY CMa at the stellar surface is ≈70 km s −1 (Humphreys et al. 2021).
VY CMa's light curve for the first half of the 20th century is shown in Figure 7 with the probable ejection dates and range.Notes.
a The projected distance from VY CMa, in milliarcseconds, is the average of the separation from Kaminski (2019) and for 2015 in Table 1.b The increase in angular separation or distance from VY CMa from 2015 to 2021. Note.
a VY CMa has a V LSR of +22 ± 1 km s −1 .
Figure 4. Sketch illustrating how Figure 5 was produced.The upper panel is a sum of the 12 CO images at Doppler velocities from +15 to +49 km s −1 relative to the LSR, and shows a virtual sampling slit oriented at position angle 120°to include Clumps C and D (Figure 2).The lower panel shows its spatial parameters.The intensity along that virtual slit was measured for each frequency channel, and the resulting one-dimensional data sets were stacked to produce the two-dimensional Figure 5.
The clumps were all ejected in the early 20th century during VY CMa's very active period, 1920-1950.There are seven episodes or deep minima during this period, each lasting 2-3 yr, which may correspond to the high-mass-loss events and accompanying obscuration by dust, similar to what was observed in Betelgeuse but on a larger scale.Based on the continuum image (Figure 1), it may be tempting to speculate that the four clumps have a common origin in the same highmass-loss episode.Within the uncertainties, the clumps could all have been ejected during the same minimum, but this seems unlikely with their different projection angles and directions they are moving.The largest and brightest, Clump C with the highest-quality measurements, has a relatively well-determined ejection time circa 1930.9Although Clump B's Doppler velocity and spatial orientation are uncertain, Clumps B and C are the closest to VY CMa, with similar proper motions.This active period might also correspond with the ejections of knots W1 A and W1 B just to the west of VY CMa (Humphreys et al. 2019).W1 B has an estimated ejection date of circa 1932.If so, their different directions, on opposite sides of VY CMa, would be evidence that surface activity may occur over much of the star during the same short period, yielding separate outflows.
Six outflows or ejecta are now identified within VY CMa's ≈25 yr active period beginning about 100 yr ago.Seven minima are observed, thus there were very likely additional ejecta which are obscured or occurred out of our current line of sight.In the next section, we discuss the mass lost in these ejections and the energy required.

Mass Lost and the Mass-loss Mechanism in VY CMa
Measured and derived parameters including the integrated flux density at 249 GHz and the dust mass for each clump are summarized in Table 5.The expected grain or dust temperature, T gr , is estimated using the standard approximation T R r T * , where the efficiency factor ζ is close to unity for most normal types of grains above T gr ∼ 200 K (see, e.g., Draine 2011).r gr is the distance of the clump from the star corrected for the projection angle in Table 4, with an adopted temperature of 3500 K (Wittkowski et al. 2012) for VY CMa and radius, R * , ≈7 au.A more precise approximation requires information about the grain properties that is not readily available.This approximation assumes that the dust is optically thin.The optical depth, τ, is estimated from the standard relation between the observed and dust temperatures. 10We determined the dust mass following O'Gorman et al. (2015)  with the same input parameters.The suggested gas-to-dust ratio for RSGs ranges from 100 to a high of 500 for VY CMa (Decin et al. 2006).For comparison with previous work on VY CMa, we adopt a gas-to-dust ratio of 200 (Mauron & Josselin 2011) for the total mass lost.The mass estimates, especially for Clumps C and D, are lower bounds since the calculation uses the dust temperature for an optically thin case.
Clumps C and A have our highest mass estimates.Our result for Clump C is comparable to that from O' Gorman et al. (2015), but is less than the value from Vlemmings & Khouri (2017).The masses for the other two clumps are typical for other knots in VY CMa with dust masses based primarily on their infrared fluxes; see Table 1 in Humphreys & Jones (2022).Clumps C and A may be representative of the more massive outflows or ejecta in VY CMa.The energies required, though, are quite modest for VY CMa, with a luminosity of 3 × 10 5 L e .For example, Clump C's kinetic energy is a little more than 10 44 erg, assuming its outflow velocity of 21 km s −1 , equivalent to VY CMa's luminosity in less than a day.
The total mass shed during this active period by the four ALMA knots is 0.05 M e .Assuming 10 −2 M e for knots W1 A and W1 B, the estimated mass lost from the observed knots and clumps is at least 0.07 M e , yielding an effective mass-loss rate of ≈10 −3 M e yr −1 or more in 25-30 yr.The mass lost in these discrete episodes not only dominates VY CMa's recent mass-loss history (Humphreys et al. 2021;Humphreys & Jones 2022), but explains its measured high mass-loss rate of 4-6 × 10 −4 M e yr −1 (Danchi et al. 1994;Shenoy et al. 2016).The mass-loss estimate from the recent dimming of Betelgeuse (Montargès et al. 2021) together with the mass in its circumstellar condensations (Kervella et al. 2009(Kervella et al. , 2011) also contribute significantly to its measured mass-loss rates (Humphreys & Jones 2022).Betelgeuse and the similar recent dimming of the high-luminosity K-type supergiant RW Cep (Anugu et al. 2023;Jones et al. 2023) are increasing evidence that high-mass surface outflows are more common in RSGs and a major contributor to their mass loss.
The primary questions concern the mass-loss mechanism for the high-mass-loss episodes.The standard methods for mass loss that work well for red giants and asymptotic giant branch stars, e.g., radiation pressure on grains, pulsation and convection, are not adequate for RSGs and cannot account for the elevation of the material to the dust formation zones (Arroyo-Torres et al. 2013, 2015;Climent et al. 2020) in the very dusty, high-luminosity RSGs like VY CMa.
The dynamical timescale for VY CMa is about 3 yr, comparable to the timescales for a nonradial instability or a magnetic/convective event similar to coronal mass ejections (CMEs).The presence of magnetic fields in VY CMa is supported by the circular polarization and Zeeman effect   observed in the OH, H 2 O, and SiO masers in its ejecta (Vlemmings et al. 2002;Richter et al. 2016;Vlemmings & Khouri 2017).Vlemmings & Khouri (2017) reported polarized dust emission from magnetically aligned grains on subarcsecond scales close to the star.Humphreys & Jones (2022) estimated a surface field of 500 G based on the magnetic field strengths from the maser emission in the ejecta.They also showed that scaling the most energetic solar CMEs to the kinetic energies of the knots in VY CMa would require a 1000 G field, similar to the above estimate.Fundamental clues to VY CMa's mass loss can be seen in the morphology of its ejecta.In this regard, stellar outflows can be categorized by their acceleration mechanisms.(i) Supernova outflows are driven by blast waves.They generally produce numerous filaments, often with roughly spheroidal symmetry.Nova remnants may also fit into this category.(ii) Giant eruptions are driven by continuum radiation pressure (Davidson 2020).The only well-resolved example, η Car, exhibits fossil turbulence cells that visually resemble volcano eruptions and some other terrestrial-scale explosions.(iii) Planetary nebulae are driven by dynamical instabilities.Their filamentary systems often include small mass condensations and large-scale bipolar or multipolar symmetries.(iv) Solar or stellar activity is driven by interactions of convection and plasma physics with magnetic fields.The outbursts are spatially localized rather than global, with loops and arcs.
The morphology of VY CMa's ejecta is most similar to stellar activity.A lack of global symmetry is obvious, and each loop or arc (see Humphreys et al. 2005Humphreys et al. , 2007, , and Figure 9 below) may be a fossilized remnant of initial magnetic confinement of outflowing material.The same process may also create localized mass concentrations like the knots or clumps.The major question is whether MHD-like processes can liberate enough energy for the outburst, in addition to confining it.

Unexpected Phenomena in the Outer Ejecta
Figure 5 in Section 3 was designed to illustrate the Clump C-D structure, but it also reveals remarkable larger-scale details that were not recognized earlier.Unlike most other figures in this paper, Figure 8(a) includes the total observed 12 CO emission plus continuum; continuum has not been subtracted.Clump C, the strongest continuum source, produces a conspicuous vertical feature about 0 4 to the left of the star.The bright vertical structure near the center of the figure is mostly 12 CO emission in the inner material with r < 600 au, ejected less than 200 yr ago.Horizontal cirrus-like features around V rel* ∼ ± 30 km s −1 , on the other hand, represent older ejecta at r ∼ 3000-8000 au, with ages 800 yr.This molecular emission was reported earlier by Ziurys et al. (2007) and in Paper I, while Shenoy et al. (2016) described infrared emission from the associated dust. 12CO images at various frequencies confirm that the pattern of inner   versus outer ejecta is broadly similar toward other position angles.
The upper half of Figure 8(a) reveals a fact that was not recognized earlier: The outer ejecta are opaque in a very narrow range of frequencies corresponding to 12 CO V rel* ≈ −28 km s −1 .Pictorially, the cirrus-like layer forms a conspicuous ceiling to the inner-ejecta emission, and the latter is undetectable in several of the frequency channels.At those frequencies the outer material is optically thick, with brightness temperatures in the range 20-40 K along the lines of sight that we discuss here.This is remarkable, and arguably surprising, for several reasons.(i) In most cases of old, strongly inhomogeneous stellar ejecta, some radiation escapes through interstices or gaps between the filaments or condensations even if each condensation is optically thick.But that is evidently untrue here.(ii) The transition is remarkably abrupt in frequency space; for instance, continuum from Clump C is detectable at V rel* = −22 km s −1 but undetectable at −25 km s −1 .(iii) The velocity range of opaque material is only about 8 km s −1 , even though the outer ejecta are extremely nonspherical.We cannot explore the physical implications of a fully opaque molecular shell here, because that is beyond the scope of this paper.At present, the main point is that one must be aware that it affects our view of the inner ejecta.
Moreover, this case provides a dramatic example of a strange illusion generated by the standard data reduction described in Paper I and here in Appendix A. Thoughout this paper, we have attempted to remove the continuum emission in order to focus on 12 CO emission.Generally, a renormalized 249 GHz continuum image (Figure 1) is subtracted from the data image for each frequency channel.This procedure is satisfactory at frequencies where the intervening material is transparent.However, it becomes invalid near V rel* ≈ −28 km s −1 because the inner continuum sources are entirely hidden at those frequencies by 12 CO absorption and emission in the outer ejecta, as described above.Moreover, Clump C dominates in the subtracted continuum image.Figure 9 shows the peculiar result.Figure 9 is the image after continuum subtraction; it shows a very conspicuous dark spot, which is merely a negative image of Clump C. Kaminski (2019) assumed that this is a real feature, which would require strange attributes (e.g., a mismatch between size and internal velocity dispersion).In fact, as stated above, it is an artifact caused by unsuitable continuum subtraction.It can also be seen in Figure 8(b), marked by an arrow.

Comments on the Evolutionary State of VY CMa
VY CMa's observed properties suggest a very evolved RSG that may be near the end of the RSG stage.It is one of most luminous known RSGs, with a corresponding initial mass of ≈30-35 M e , significantly above the 18 M e upper mass for the progenitors of the Type IIP supernovae (Smartt 2015).Thus, its final fate may be to either directly collapse to a black hole or evolve back to warmer temperatures before the terminal explosion.Stellar structure models (Eggenberger et al. 2021) show that the latter requires sufficient mass loss to increase the ratio of the He/C core relative to the total mass to send the star on a "blue loop" across the H-R diagram.
Our observations of the CO emission from the outer ejecta, discussed in the previous section, reveal diffuse filamentary ejecta expanding at 30-40 km s −1 relative to the star at distances of 5″-6″ from VY CMa (Ziurys et al. 2007;Paper I).At about 7000 au from the star, the expansion age of the outflow is 800-1100 yr.The CO emission appears to surround the more complex inner ejecta including the large arcs in the HST images.The prominent Arc 1 is the oldest, with an expansion age of about 800 yr (Humphreys et al. 2007).In an independent study based on a radiative transfer model of the CO emission, Decin et al. (2006) concluded that a high-mass- loss phase in VY CMa began about 1100 yr ago.Furthermore, the 37 μm image of the cold dust (Shenoy et al. 2016) shows that there is no cold dust beyond a radius of 10″, corresponding to an ejection age of about 1400 yr.We thus conclude that VY CMa entered its presently observed active phase with relatively frequent massive outflows about 1200 yr ago.
We have no explanation for the onset of its enhanced surface activity, which may have been stimulated by a change in the interior or more likely a change in the structure of the convective layers.Of course, VY CMa is not alone.Surface outflows have been observed in Betelgeuse and RW Cep, although the record for VY CMa is remakable.
VY CMa also has a unique rich and peculiar chemistry.Twenty-five molecules have been identified in its ejecta, including molecules of carbon and silicon.The only comparable star is the similar luminous RSG, NML Cyg (Singh et al. 2022), with 21 molecules in common with VY CMa.In Paper I, we reported on preliminary 12 C/ 13 C ratios of 22-38 in various structures in the ejecta.These ratios are significantly higher than measured in oxygen-rich red giants and supergiants and may be indicators of additional dredge-up, perhaps related to VY CMa's surface activity.Our future work will include abundance measurements of the additional molecules observed in the program (Paper I).Their association with separate outflows, arcs, and clumps at different locations and with expansion ages may provide more clues to VY CMa's current state and fate.Another possibility, not often discusssed, is whether VY CMa may be a second-generation RSG.Similar to lower-mass stars, it may have evolved back to warmer temperatures, has now returned, and is in a very short highmass-loss state prior to core collapse to a black hole.

Figure 1 .
Figure 1.The continuum image at 249 GHz.The contours showing the structure and outlines of the clumps are in multiples of (1, 4, 16, 64) × I 1 , where I 1 ≈ 0.6 mJy beam −1 ≈8 mJy arcsec −2 .The brightness temperature at the outer contour is about 0.2 K.The synthesized beam (FWHM) is shown as an ellipse in the lower-left corner.

Figure 2 .
Figure 2. 12 CO images of a 6″ × 6″ region centered on VY CMa.The tick marks are 1″.Clumps C and D are redshifted relative to VY CMa and observed over the same range of Doppler velocities, while Clump A is blueshifted.Clump B is not resolved from the bright CO emission from VY CMa's circumstellar ejecta.(a) 12 CO emission at V LSR 43 km s −1 toward Clumps C and D, and new outflow B¢.The position of Clump B is marked. (b) Toward Clump A at V LSR +6 km s −1 .The position of VY CMa is shown as a + in each panel; its Doppler velocity is +22 km s −1 relative to the LSR.The brightness scales differ between panels.The images are continuum subtracted.

Figure 3 .
Figure 3. Map of the proper motions showing the direction of motion for the clumps relative to their positions in 2015.The 2015 positions are marked by solid circles and the 2021 positions by hollow squares.The vectors and error bars have been multiplied by 3 to make them more visible.
Figure 1 also shows two major outflows: to the southeast with Clumps C and D and to the northeast with Clumps A and B. With their relatively high transverse velocities, Clumps D and A may represent the initial ejection and leading edge of a major eruption that in the case of the southeast outflow includes Clump C, but their different projections imply different active regions for D and C. Based on their motions, we suggest that Clumps A and D are separate events from C.

Figure 5 .
Figure 5. Position-velocity map of the continuum-subtracted 12 CO emission at position angle 120°; see Figure 4.The vertical coordinate is the Dopper velocity, while the horizontal scale is the spatial position along a line oriented at position angle 120°(roughly ESE).The location of the star is marked "+," and the approximate position of Clump C by a small circle.To some extent the Doppler velocities are correlated with location along our line of sight, with the "near side" at the top of the figure.However, at least two sets of ejecta with different ages and size scales are superimposed, e.g., at velocities +30 to +50 km −1 .See also Figure 8.

Figure 6 .
Figure 6.The observed line-of-sight velocity of the 12 CO emission relative to VY CMa toward Clump D vs. the corresponding projected distance from the star in milliarcseconds and au for 15 channels representing the diffuse arc in Figure 5.The open circles show the more diffuse fainter 12 CO emission at the end of the arc in Figure 5.Each measurement used a sampling parameter of 5 pixels (0 2).The error bar represents the typical uncertainty in the centroid position.The main trend agrees with the diffuse arc in Figure 5.
Notes a The 12 CO line-of-sight velocity relative to VY CMa, V rel* .b The reference epoch is 2015 for the ages measured from the proper motions and the projected distance from the star in arcseconds.c Clump B's Doppler velocity is uncertain because of doubtful identification of the associated 12 CO emission source; see text.

Figure 7 .
Figure 7.The historic light curve for VY CMa from Robinson (1970) with the estimated ejection dates and possible range of dates from the uncertainty in the expansion velocities for the four clumps.
Figure5in Section 3 was designed to illustrate the Clump C-D structure, but it also reveals remarkable larger-scale details that were not recognized earlier.Figure8is similar to Figure5, but has brightness and contrast parameters that show the outer ejecta more clearly.Again, the horizontal scale represents projected spatial location along a line with position angle 120°and −60°(roughly ESE and WNW), centered at the star.The vertical scale is frequency expressed as 12 CO Doppler velocity relative to the star, e.g., negative velocities at the top of the figure represent ejecta moving toward us.Unlike most other figures in this paper, Figure8(a) includes the total observed 12 CO emission plus continuum; continuum has not been subtracted.Clump C, the strongest continuum source, produces a conspicuous vertical feature about 0 4 to the left of the star.The bright vertical structure near the center of the figure is mostly 12 CO emission in the inner material with r < 600 au, ejected less than 200 yr ago.Horizontal cirrus-like features around V rel* ∼ ± 30 km s −1 , on the other hand, represent older ejecta at r ∼ 3000-8000 au, with ages 800 yr.This molecular emission was reported earlier byZiurys et al. (2007) and in Paper I, whileShenoy et al. (2016) described infrared emission from the associated dust. 12CO images at various frequencies confirm that the pattern of inner at 249 GHz, mostly thermal emission by dust grains.b Based on the continuum flux density and apparent projected size of each clump (FWHM).May be underestimated because the clumps are not well resolved.

Figure 8 .
Figure 8.An oblique plane in the image cube, similar to Figure 5.The vertical scale is frequency, represented by the 12 CO Doppler velocity relative to the star.The horizontal axis is spatial position along a projected line oriented at position angle 120°; see Figure 4.The star's position is marked by "+."(a) Total observed intensity, 12 CO emission plus continuum.The prominent vertical line just left of center is continuum emission from Clump C. (b) Same image after the formal continuum-subtraction process.An arrow marks the invalid dark spot shown in Figure 9 and explained in the text.
(a), like Figure 8(a), represents the total observed data without continuum subtraction; it is almost entirely 12 CO emission with T b ∼ 30 K produced in the opaque layer, with no perceptible contribution by Clump C. Figure 9(b)

Figure 9 .
Figure 9. 12 CO spatial images at Doppler velocity V rel* −27 km s −1 .At this frequency the outer ejecta are optically thick, hiding the inner ejecta.(a) Observed intensity, almost entirely 12 CO emission.(b) Result of the formal continuum-subtraction procedure, which is valid at most other frequencies.The prominent central dark spot is an illusory negative image of continuum from Clump C. In both images the mottled mesh pattern is an artifact of the interferometry reduction process.

Table 2
Measured Positions and Proper Motions

Table 3 V
LSR Range a for the CO Emission Observed at the Clumps

Table 4
Velocities, Distance, and Time since Ejection

Table 5
Derived Parameters for the ALMA Clumps