The Astrophysical Journal, 646:1203–1208, 2006 August 1

# 2006. The American Astronomical Society. All rights reserved. Printed in U.S.A.

BRINGING VY CANIS MAJORIS DOWN TO SIZE: AN IMPROVED DETERMINATION

OF ITS EFFECTIVE TEMPERATURE
Philip Massey,1, 2 Emily M. Levesque,2, 3 and Bertrand Plez4
Received 2006 March 2; accepted 2006 April 10

ABSTRACT
The star VY CMa is a late-type M supergiant with many peculiarities, mostly related to the intense circumstel-
lar environment due to the star’s high mass-loss rate. Claims have been made that would imply that this star is con-
siderably more luminous (L
5 ; 105 L ) and larger (R
2800 R ) than other Galactic red supergiants (RSGs).
Indeed, such a location in the H-R diagram would be well within the ‘‘Hayashi forbidden zone,’’ where stars cannot
be in hydrostatic equilibrium. These extraordinary properties, however, rest on an assumed effective temperature of
2800–3000 K, far cooler than recent work has shown RSGs to be. To obtain a better estimate, we fit newly obtained
spectrophotometry in the optical and NIR with the same MARCS models used for our recent determination of the
physical properties of other RSGs; we also use V  K and V  J from the literature to derive an effective tem-
perature. We find that the star likely has a temperature of 3650 K, a luminosity L
6 ; 104 L, and a radius of
600 R.
These values are consistent with VY CMa being an ordinary, evolved 15 M RSG and agree well with the Geneva
evolutionary tracks. We find that the circumstellar dust region has a temperature of 760 K, and an effective radius of

130 AU, if spherical geometry is assumed for the latter. What causes this star to have such a high mass loss and
large variations in brightness (but with little change in color) remains a mystery at present, although we speculate
that perhaps this star and NML Cyg are simply normal RSGs caught during an unusually unstable time.
Subject headinggs: stars: evolution — stars: late-type — stars: mass loss — supergiants

1. INTRODUCTION overall spectrum has stayed constant at M3–M5 I for over 40 years,
the spectrum shows emission of low-excitation lines, including
VY CMa is a late-type M supergiant that is remarkable in many Na i and K i, as well as emission in the band heads of TiO and
ways. It has a large IR excess, making it one of the brightest ScO (Wallerstein 1958; Herbig 1970b), which do change over the
objects in the sky at 5–20
m, indicative of a dust shell (or disk) timescales of months. Second, the star’s spectrum has been de-
heated by the star (Herbig 1970a). The inferred mass-loss rate scribed as ‘‘veiled’’ (Herbig 1970b; Humphreys 1974), a term
is huge for a red supergiant (RSG), about 2 ; 104 M yr1 used to denote the general washing out of absorption features seen
(Danchi et al. 1994). Like other late-type mass-losing stars, it in the spectra of some Mira variables (Merrill et al.1962) and other
has strong molecular maser emission; in fact, it was one of the luminous RSGs (Humphreys & Lockwood 1972).
first radio masers discovered (Wilson & Barrett 1968; Eliasson However, our interest in VY CMa was prompted by the
& Bartlett 1969). The star is embedded in an asymmetric dust- extraordinary physical properties often assumed for this star
reflection nebula that extends 800 –1000 from the star and is highly (Le Sidaner & Le Bertre 1996; Smith et al. 2001; Monnier et al.
structured (Monnier et al. 1999; Smith et al. 2001; Smith 2004; 2004; Humphreys et al. 2005), such as a radius of 1800–3000 R
Humphreys et al. 2005). This nebula is so bright that it was (8.3–14 AU ), significantly larger than the
1500 R value
discovered in 1917 with an 18 cm telescope, but was somehow measured for the three largest RSGs in a sample of 74 Galactic
missed by earlier observers with much better equipment, leading stars we have recently analyzed (Levesque et al. 2005, hereafter
to the speculation that the nebula is only 100 years old (Herbig Paper I). The corresponding luminosity (2–5 ; 105 L , or Mbol ¼
1972; Worley 1972). The photometric history of the star extends 8:5 to 9.5) is near or beyond the highest luminosities of other
back to 1801, and shows the star fading visually by 2 mag since known RSGs (Paper I). Usually such problems can be traced to
that time (Robinson 1971), with typical variations of 2 mag an uncertainty in the distance, but in the case of VY CMa, the
(Robinson 1970; A. Henden 2006, private communication) but distance is relatively well determined, as the star is associated with
with little change in color. a molecular cloud at a distance of 1.5 kpc (Lada & Reid 1978).
Spectroscopically, the star is unusual in two regards. First, This distance is consistent with the proper motions measured
although careful long-term monitoring by Wallerstein (1958, from H2O maser features (Richards et al. 1998). Furthermore, the
1977) and Wallerstein & Gonzalez (2001) has shown that the very large diameter appears to be supported by direct interferom-
etry (Monnier et al. 2004).
1 The extreme properties determined for this star rest primarily
Lowell Observatory, 1400 West Mars Hill Road, Flagstaff, AZ 86001;
[email protected]. on the assumed effective temperature of 2800 K, a value that
2
Visiting Astronomer, Cerro Tololo Inter-American Observatory, National comes from Le Sidaner & Le Bertre (1996) who adopted this
Optical Astronomy Observatory, which is operated by the Association of temperature based on the star’s spectral type.5 Were the star at
Universities for Research in Astronomy, Inc., under cooperative agreement with 2800 K with Mbol
9, it would be considerably cooler and more
the National Science Foundation.
3
Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cam-
luminous than current evolutionary models allow, and would be
bridge, MA 02139; [email protected].
4 5
Groupe de Recherche en Astronomie et Astrophysique du Languedoc, Le Sidaner & Le Bertre (1996) cite the effective temperature scale of Dyck
CNRS, UMR 5024, Université de Montpellier II, 34095 Montpellier Cedex 05, et al. (1974), but 2800 K is much cooler than what Dyck et al. (1974) adopt for
France; [email protected]. an M4 star: roughly 3400 K; see their Figure 7.

1203
1204 MASSEY, LEVESQUE, & PLEZ Vol. 646

optical/NIR spectrophotometry and existing IR photometry. Ther-

mal emission from the dust apparently begins to dominate the spec-
tral energy distribution past 2
m (see Fig. 1 of Hyland et al. 1969),
and thus, we expect that the use of V  K will lead to too low
an effective temperature. However, scattering by the dust in the
optical / NIR region may partially fill in spectral absorption features
(i.e., veiling), and fitting the molecular bands using the MARCS
models may lead to too high an effective temperature. The true
effective temperature of the star should be between these limits,
and using V  J may help determine which is closer.
2. DETERMINATION OF PHYSICAL PROPERTIES
We observed VY CMa using the Cerro Tololo Inter-American
Observatory (CTIO) 4 m telescope and RC spectrograph (Loral
CCD in the blue air Schmidt camera) in the red (5300–9000 8)
using a 2 s exposure on 2005 December 20, and in the blue (3400–
6200 8) using a 30 s exposure on 2005 December 21. For both
setups, we used grating KPGL2 (316 lines mm1). The disper-
sion was 2 8 pixel1, and we achieved a spectral resolution of
7.5 8 (3.8 pixels) using a 225
m slit (1B5). For the red obser-
vations, we used a WG-495 filter to block out second-order blue
light. For the blue observations, we used a BG-39 filter to remove
any possibility of scattered red light affecting our observations,
especially in the far blue and near-UV, where the stellar flux is
small (see Massey et al. 2005; Levesque et al. 2006, hereafter
Paper II). The spectrograph was rotated to the parallactic angle
Fig. 1.—Location of VY CMa in the H-R diagram. The three numbered filled
circles correspond to the ‘‘old’’ parameters of VY CMa, where (1) is from Smith for each observation. Multiple observations of spectrophotometric
et al. (2001), (2) is from Monnier et al. (1999), and (3) is from Le Sidaner & Le flux standards were obtained throughout the night, and the agree-
Bertre (1996), where we have adjusted their luminosity to our adopted 1.5 kpc ment between these standards was typically 1%–2% after a gray
distance. In all three cases, the effective temperature has simply been assumed shift. Observations of several ‘‘featureless’’ stars allowed us to
based on the spectral type. All three points lie well into the forbidden zone
(dashed red area) to the right of the Hayashi limit, and so are hydrostatically remove the telluric absorption to first order, following Bessell
unstable. The five-pointed stars correspond to the results of this paper: (4) is (1999). The data were flat-fielded using projector-flat data ob-
based on our solution fitting the NIR TiO bands; (5) is based on our solution tained during the day and were wavelength calibrated using ex-
fitting the blue TiO bands; (6 ) is based on the (V  K )0 photometry (which we posures of a He-Ne-Ar comparison arc obtained immediately before
take to be a lower limit to the temperature and an upper limit to the luminosity); (blue) or after (red) the exposure of VY CMa. Finally, the re-
and (7) is based on the (V  J )0 colors. We denote our preferred values ( point 4)
in purple. The evolutionary tracks are from Meynet & Maeder (2005), with solid duced blue and red spectra were combined with a small adjust-
lines indicating no initial rotation and dashed lines indicating the tracks with ment to bring the flux levels in the overlap region into agreement.
initial rotations of 300 km s1. The diagonal lines at upper right show lines of We show the spectrum of VY CMa in Figure 2a, along with line
constant radii. identifications.
The spectra were fitted using the same MARCS stellar-
located in the ‘‘forbidden zone’’ to the right of the Hayashi track atmosphere models used in Paper I, using a similar technique.
(Fig. 1).6 Oddly, no previous authors have commented on this fact. We compared the synthetic spectra to the observed one, matching
Such objects are not in hydrostatic equilibrium. Although RSGs the line depths of the molecular bands ( primarily TiO) and ad-
would not be found in this region, collapsing protostars might, justing the reddening of the model with a Cardelli et al. (1989)
and indeed Wallerstein (1978) and Herbig (1970b) have long ar- law (RV ¼ 3:1). A satisfactory fit was quickly obtained with a
gued that VY CMa might be a recently formed object rather than TeA ¼ 3650 K model, and AV ¼ 3:20 (Fig. 2b, red curve). Our
an evolved massive star, although based on other reasoning. estimates of the uncertainties are 25 K for Teff and 0.15 for AV .
Understanding the nature of this interesting object rests on This reddening is in good agreement with the E(B  V )
1 value
knowing its effective temperature, as this is also key to deter- suggested by Wallerstein & Gonzalez (2001). We used a log g ¼
mining the star’s bolometric luminosity. We have recently used 0:0 model for the fitting; a model with log g ¼ þ0:5 gives a better
MARCS model atmospheres (Gustafsson et al. 1975; Plez et al. fit to the gentle red slopes of some of the TiO bands in the NIR, but
1992; Plez 2003; Gustafsson et al. 2003) to derive the physical log g ¼ 0:0 is more consistent with the star’s inferred physical prop-
properties of Galactic and Magellanic Cloud RSGs. Here we erties. Use of the higher surface gravity model would increase the
apply the same techniques to VY CMa, using newly acquired reddening negligibly, and not change the temperature. The discrep-
ancy in the UV between the observed and expected fluxes is similar
6
to that seen around other RSGs with large amounts of circumstel-
Note that the Hayashi tracks are analytically of the form log TeA ¼ lar dust, as inferred by excess reddening (Massey et al. 2005), and
0:02Mbol þ 0:2 log M þ C ( Kippenhahn & Weigert 1990; Bohm-Vitense 1992).
For RSGs, log M ¼ 0:5  0:10Mbol (Paper I ), and hence, the edge of the for- is likely due to scattering of light from the star by the dust, pos-
bidden zone must correspond roughly to a constant temperature for massive sibly combined with a distribution of grain sizes that is larger than
stars. The exact value of this temperature depends critically on the treatment of normal.
convection; here we adopt log TeA ¼ 3:55, where we have shifted the results of We list the star’s derived physical properties in Table 1. For
Ezer & Cameron (1967) by 0.05 dex (consistent with a slight lowering of their
adopted value of l/HP from 2) in order to give a better match to the evolutionary this we adopted V ¼ 8:5, a value close to the star’s average bright-
tracks of Meynet & Maeder (2005). See also Henyey et al. (1965) and the ness and similar to the value the star had at the time we observed
discussion in Kippenhahn & Weigert (1990). it spectroscopically, according to estimates by the AAVSO
No. 2, 2006 EFFECTIVE TEMPERATURE OF VY CMa 1205

Fig. 2.—Spectrum of VY CMa. In (a), we compare the spectrum of VY CMa (bottom spectrum) to that of the M2 I star HD 100930 (top spectrum) and the M4 I star
HD 93420 (middle spectrum), where the data for the two standards comes from Paper I and have not been corrected for telluric features. The standards have been shifted
vertically by arbitrary amounts. In (b) we show the model fits. The red curve is our preferred model fit (TeA ¼ 3650 K and AV ¼ 3:20), which does a good job
reproducing the TiO bands in the red (i.e., TiO kk6158, 6658, 7054) but produces TiO bands too weak in the blue (i.e., TiO kk4761, 4954, 5167). The spectrum (and
model fit) have been shifted vertically. The lower spectrum shows the ‘‘fit’’ ( purple) with a cooler model (TeA ¼ 3450 K) and less reddening (AV ¼ 2:00). The TiO bands
in the blue are in better agreement, but the model produces TiO bands that are much too strong in the red. As found for other RSGs with significant circumstellar dust, there
is significant extra flux in the observed stellar continuum in the near-UV, due, we believe, to scattering by the dust (Massey et al. 2005). The models also do not do a good
job of reproducing the atomic Ca i and Ca ii features, as discussed in Paper I.

(A. Henden 2006, private communication; see below). With a primarily on blue (photographic) spectrograms. If instead we had
distance of 1.5 kpc and our values for AV, this determines the used TiO bands further to the blue (such as 4761, 4954, 5167,
absolute visual magnitude (MV). The MARCS models provide and 5448 8), we would in fact have arrived at an M4 I classi-
the bolometric corrections as a function of effective temperature fication (Fig. 2a). We did a fit based on these TiO bands and
(see Paper I), allowing us to determine the star’s bolometric found TeA ¼ 3450 K and a lower reddening (AV ¼ 2:00). One
luminosity L. The effective stellar radius R then follows from can argue that the TiO bands in the blue are less likely to be
4 2
L ¼ 4TeA R . affected by veiling (Humphreys 1974), but alternatively there is
Although the temperature is significantly higher than that clearly excess flux in the blue due to dust scattering (Massey
adopted by other workers, it is in accord with our own recent et al. 2005). In any event, our fit to the TiO bands in the blue was
estimates of the effective temperature scale of RSGs (Paper I), considerably poorer when compared to the overall spectrum
corresponding to a type somewhat later than M2 I. Although the (Fig. 2b, purple curve), and so we are inclined to place more
star’s spectrum has been variously described as M3–M4 I (Joy credence in the NIR fit. We include the derived physical pa-
1942), M5 I (Wallerstein 1958), and M4–M5 I (Humphreys rameters from the blue fit in Table 1 for comparison.
1974), we would actually classify it as M2.5 I, based on the TiO We can also estimate the star’s effective temperature using
bands we normally use, i.e., 6158, 6658, and 7054 8 (see also the star’s V  K color. There are two difficulties with this. The
Jaschek & Jaschek 1990). However, our results are not really first is that the star is a notorious variable (Robinson 1970, 1971).
in conflict with earlier studies, which based their classifications In Figure 3, we show the variations in the visual brightness of

TABLE 1
Physical Properties of VY CMa

Teff log g R
Method ( K) AV (cgs) MV Mbol (R)

TiO (red / NIR bands)a ........... 3650  25 3.20 0.1 5.6 7.2 605
TiO (blue/yellow bands) ........ 3450  25 2.00 0.1 4.4 6.9 595
V  K ..................................... >3475  35 3.20b 0.2 5.6b >8.0 <955
V  J ...................................... 3705  90 3.20b 0.2 5.6b 7.0 545
a
Preferred; see text.
b
Adopted.
1206 MASSEY, LEVESQUE, & PLEZ Vol. 646

of 35 K. We include in Table 1 the physical parameters derived

from this value.
The second difficulty with using V  K to estimate the effec-
tive temperature of the star is that this will provide only a lower
limit, given that thermal emission significantly contaminates the
K-band (Hyland et al. 1969) photometry. A better estimate then
might come from the V  J colors. For both the Hyland et al.
(1969) and 2MASS data sets, we obtain V  J ¼ 5:43 and 5.45,
respectively. We can deredden these by AV ¼ 3:20 and AJ ¼ 0:90,
where we have assumed AJ ¼ 0:28 AV , again following Schlegel
et al. (1998). We then obtain (V  J )0 ¼ 3:14. Using the MARCS
models for Galactic metallicity, we find that we expect

TeA ¼ 7260:0  2073:74(V  J )0

þ 371:600(V  J )20  22:802(V  J )30 K ð1Þ

for log g ¼ 0:0 and 4300 K  TeA  3000 K. Thus, for VY

CMa, the V  J colors imply a temperature of 3705 K. An
uncertainty of 0.2 mag corresponds to an error of 90 K. This is
in good agreement with the 3650 K temperature we obtained
from fitting the red / NIR TiO bands.
How reliable are the temperatures derived from the MARCS
models? In Paper I, we showed that the temperatures and lu-
minosities derived using these models brought Galactic RSGs
into accordance with the predictions of modern stellar evolution-
Fig. 3.—Visual photometry of VY CMa. The plus signs are visual estimates
(typically 0.2 mag) from the AAVSO (A. Henden 2006, private communication). ary theory; in Paper II, we showed that this was also true at the
The colored filled circles are photoelectric V measurements: the blue point is that lower metallicities of the LMC and SMC. We consider it unlikely
quoted by Hyland et al. (1969); the green points are from Cousins & Lagerweij that this agreement is coincidental, and instead believe this find-
(1971); and the red point is that quoted by Wallerstein (1971). These indicate good ing provides a powerful endorsement of each. Further, we have
agreement with the visual data. We indicate the times of the IR photometry (Hyland
et al.1969; 2MASS) and of our spectrophotometry. The solid line at V ¼ 8:5 shows
now shown (Paper II) that the MARCS models are self-consis-
the median value throughout the AAVSO data set. tent in that the temperatures derived from broadband (V  R)0
are similar (to 30 K) with those derived from fitting the TiO lines.
We do find (Paper II) that the broadband (V  K)0 colors of the
the star collected by the AAVSO over the years (A. Henden 2006, models yield effective temperatures that are
100 K warmer than
private communication), along with any published photoelectric V those derived by the other means. We attribute this to the limi-
values. Oddly, the star has stayed remarkably constant in color tations of static one-dimensional models, as spectra of RSGs in
during these shifts; for instance, the star faded by 0.4 mag in V dur- the optical and IR may reflect different atmospheric conditions
ing the photoelectric monitoring by Cousins & Lagerweij (1971) due to the large surface granulation present in these stars.
but remained essentially constant in B  V and U  B.7 There are
two IR measurements in the literature: K ¼ 0:62 by Hyland 3. DISCUSSION
et al. (1969) from 1968 November and K ¼ þ0:34 (after trans- We have brought the same techniques to bear on VY CMa as
formation to the standard system of Bessel & Brett [1988], we have on other RSGs. Contrary to previous claims, we find
following Carpenter [2001]) from 1999 November. Despite the that its effective temperature is typical of RSGs (3450–3700 K)
0.9 mag difference, the J  H and J  K colors are nearly iden- and that its luminosity, while high (Mbol ¼ 7:0 to 8.0, or L ¼
tical: J  H ¼ 1:49  0:10 versus 1:35  0:47 and J  K ¼ 5 ; 104 to 1:3 ; 105 L ), is not extraordinary. Our preferred value
2:63  0:2 versus 2:26  0:50, where the large errors correspond comes from fitting the red / NIR TiO bands, which yield TeA ¼
to the 2MASS colors and are due to the photometry being 3650 K and Mbol ¼ 7:2 (L ¼ 6:0 ; 104 L ); this is in good
derived from fitting the wings of the stellar profile.8 More agreement with that found from the V  J photometry.
remarkably, if we take the V ¼ 7:44 magnitude reported by Monnier et al. (2004) used 2
m interferometry to obtain a
Hyland et al. (1969) (which is presumably contemporaneous diameter of VY CMa of 18.7 mas; at a distance of 1.5 kpc, this
with their K value) and the V ¼ 8:40 magnitude that corre- corresponds to a radius of 3000 R, in seeming contradiction to
sponds to the average of the AAVSO estimates near the time of the results obtained here. Adopting this large size leads to a tem-
the 2MASS observation, we find fortuitously identical colors, perature of 2700 K, in (un)fortuitous agreement with the value
V  K ¼ 8:06. We can deredden this using AV ¼ 3:20 and proposed by Le Sidaner & Le Bertre (1996). J. Monnier (2006,
AK ¼ 0:35, where we have assumed that AK ¼ 0:11AV , follow- private communication) was kind enough to offer two possible
ing Schlegel et al. (1998), and obtain (V  K)0 ¼ 5:21. Using explanations. First, their analysis assumed that the dust shell did
our calibration from Paper I, this corresponds to an effective tem- not have significant structure on the scale of the stellar diam-
perature of TeA ¼ 3475 K. An uncertainty of 0.2 mag in the color eter; if that were not the case, then the angular measurement
(not unreasonable) would correspond to an error in the temperature might have included part of the dust shell. Second, it could be
that VY CMa has a molecular water layer around the star that
7
Note that this conclusion differs from that of Robinson (1970), who finds
emits at 2
m, similar to what is seen around some cool asymp-
that the star reddens as it fades. totic giant branch (AGB) stars. This would cause the continuum
8
See http://www.astro.virginia.edu/
mfs4n/2mass/allsky/bs_allsky.html. diameter to be larger than the photosphere. Higher resolution
No. 2, 2006 EFFECTIVE TEMPERATURE OF VY CMa 1207

studies are now possible, and are being planned, to resolve this excitation emission lines) are related to the rich circumstellar en-
possible discrepancy. vironment caused by the star’s very high mass-loss rate. We have
Our improved estimate of the temperature of the star has little recently shown (Massey et al. 2005) that, in general, the dust-
impact on the inferred properties of the circumstellar dust shell. production rate of RSGs is proportional to the bolometric lumi-
If we adopt TeA ¼ 3650 K for the star and assume that the J-band nosity of the star, but Danchi et al. (1994), in particular, have
photometry is uncontaminated by the thermal emission (as is emphasized that the dust production is quite sporadic, with time-
implied by the good agreement with the V  J colors and that scales on the order of a few decades. Could the mass-loss rates of
found by fitting the red / NIR TiO bands), then we can compute RSGs vary so significantly that VY CMa is simply a normal RSG
both the temperature Tcs and effective area Acs of the circumstel- going through a short period of intense mass loss that is normal?
lar material by comparing the dereddened IR colors with those of We have called attention here to a newly noted peculiarity, namely,
the stellar model. The dereddened 2MASS observations yield J0 ¼ that the star’s brightness seems to change by large amounts with
2:07, H0 ¼ 1:03, and K0 ¼ 0:02. The colors corresponding to a little change in colors, even the V  K colors. This is likely not
MARCS TeA ¼ 3650 K, log g ¼ 0:0 model are (J  H)mod ¼ due to dust (because of the grayness) and suggests a luminosity
0:88 and (J  K)mod ¼ 1:11. We assume that the observed J0 is change at nearly constant effective temperature. We note that since
just the same as that of the model plus a normalization constant, RSGs are fully convective, they do (by definition) lie on the edge
J0 ¼ Jmod þ C, but that H0 and K0 contain a flux component due of the Hayashi forbidden zone. If some instability caused the star
to the circumstellar shell (Fcs): H0 ¼ 2:5 log (FH; þ Acs FH;cs ) to venture slightly into this zone, we would expect the star to un-
(with a similar equation for K0), where Acs is the emitting area dergo a very short, unstable period. We speculate that this might
of the circumstellar shell compared to that of the star and FH, is be responsible for variations in the star’s luminosity and in the
the flux from the star in H, namely, 10(Hmod þC)/2:5 . Then it fol- dust production rate. Possibly VY CMa, and the more extreme-
lows that Acs FH;cs /FH; ¼ 10(Hmod þCH0 )/2:5  1. We find that IR object NML Cyg, are examples of normal RSGs that we have
Acs FH;CS /FH; ¼ 0:15 and Acs FK;cs /FK; ¼ 1:45. The ratio FK;cs / simply caught during an unusual time.
FH;cs ¼ 9:6FK; /FH; ¼11:9, which corresponds to a blackbody
temperature of 760 K. This is close to the 850 K temperature de-
duced from mid-IR photometry by Le Sidaner & Le Bertre (1996), Our attention was originally called to this interesting star by
given the uncertainties in JHK and the small leverage we have John Monnier and Roberta Humphreys. We are also grateful for
here compared to the mid-IR. It is also consistent with the temper- correspondence with George Herbig and George Wallerstein,
ature range estimated by Wallerstein (1958) for the low-excitation whose seminal papers on this star made for interesting and en-
emission lines, and the 600 K excitation temperature estimated for joyable reading. We thank Knut Olsen for useful comments on
the ScO emission by Herbig (1974). We can go one step further, the manuscript, and for his continued collaboration on the overall
however, and compute the area based on the calculated fluxes per project of which this is a small part. This paper made use of data
unit area of the MARCS model and a 760 K blackbody in the H products from the Two Micron All Sky Survey, which is a joint proj-
and K bands. We find Acs ¼ 2130 from H, and Acs ¼ 2155 from ect of the University of Massachusetts and the Infrared Processing
K. Thus, the radius of the emitting region must be about 46 times and Analysis Center/California Institute of Technology, funded
that of the stellar radius, if it is a spherical surface. If the stellar by the National Aeronautics and Space Administration and the
radius is
600 R, this circumstellar material would have an effec- National Science Foundation. We acknowledge with thanks the
tive radius of
130 AU. variable star observations from the AAVSO International Data-
We find that the location of VY CMa in the H-R diagram is base contributed by observers worldwide and used in this research.
consistent with that of an evolved 15 M star. Why then does An anonymous referee made useful comments on the paper, lead-
VY CMa have so many peculiarities, as noted above? All of ing to improvements in the final version. This work was partially
these phenomena ( photometric fading by 2 mag over 2 centu- supported by the National Science Foundation through grant
ries, ‘‘veiling’’ of the optical spectra, intense IR emission, low- AST 0604569.

REFERENCES
Bessell, M. S. 1999, PASP, 111, 1426 Humphreys, R. M., Davidson, K., Ruch, G., & Wallerstein, G. 2005, AJ, 129,
Bessell, M. S., & Brett, J. M. 1988, PASP, 100, 1134 492
Bohm-Vitense, E. 1992, Introduction to Stellar Astrophysics, Vol. 3 (Cam- Humphreys, R. M., & Lockwood, G. W. 1972, ApJ, 172, L59
bridge: Cambridge Univ. Press) Hyland, A. R., Becklin, E. E., Neugebauer, G., & Wallerstein, G. 1969, ApJ,
Cardelli, J., Clayton,G. D., & Mathis, J. S. 1989, ApJ, 345, 245 158, 619
Carpenter, J. M. 2001, AJ, 121, 2851 Jaschek, C., & Jaschek, M. 1990, The Classification of Stars (Cambridge:
Cousins, A. W. J., & Lagerweij, H. C. 1971, Mon. Notes Astron. Soc. South Cambridge Univ. Press)
Africa, 30, 12 Joy, A. H. 1942, ApJ, 96, 344
Danchi, W. C., Bester, M., Degiacomi, C. G., Greenhill, L. J., & Townes, C. H. Kippenhahn, R., & Weigert, A. 1990, Stellar Structure and Evoution ( Berlin:
1994, AJ, 107, 1469 Springer)
Dyck, H. M., Lockwood, G. W., & Capps, R. W. 1974, ApJ, 189, 89 Lada, C. J., & Reid, M. J. 1978, ApJ, 219, 95
Eliasson, B., & Bartlett, J. F. 1969, ApJ, 155, L79 Le Sidaner, P., & Le Bertre, T. 1996, A&A, 314, 896
Ezer, D., & Cameron, A. G. W. 1967, Canadian J. Phys., 45, 3429 Levesque, E. M., Massey, P., Olsen, K. A. G., Plez, B., Josselin, E., Maeder, A.,
Gustafsson, B., Bell, R. A., Eriksson, K., & Nordlund, 8 1975, A&A, 42, 407 & Meynet, G. 2005, ApJ, 628, 973 ( Paper I )
Gustafsson, B., Edvardsson, B., Eriksson, K., Mizumo-Wiedner, M., Jorgensen, Levesque, E. M., Massey, P., Olsen, K. A. G., Plez, B., Meynet, G., & Maeder,
U. G., & Plez, B. 2003, in ASP Conf. Ser. 288, Stellar Atmosphere Mod- A. 2006, ApJ, 645, 1102 ( Paper II )
eling, ed. I. Hubeny, D., Mihalas, & K. Werner (San Francisco: ASP), 331 Massey, P. Plez, B., Levesque, E. M., Olsen, K. A. G., Clayton, G. C., &
Henyey, L. G., Vardya, M. S., & Bodenheimer, P. L. 1965, ApJ, 142, 841 Josselin, E. 2005, ApJ, 634, 1286
Herbig, G. H. 1970a, ApJ, 162, 557 Merrill, P. W., Deutsch, A. J., & Keenan, P. C. 1962, ApJ, 136, 21
———. 1970b, Mem. Soc. R. Sci. Liège, 19, 13 Meynet, G., & Maeder, A. 2005, A&A, 429, 581
———. 1972, ApJ, 172, 375 Monnier, J. D., Tuthill, P. G., Lopez, B., Cruzalebes, P., Danchi, W. C., &
———. 1974, ApJ, 188, 533 Haniff, C. A. 1999, ApJ, 512, 351
Humphreys, R. M. 1974, ApJ, 188, 75 Monnier, J. D., et al. 2004, ApJ, 605, 436
1208 MASSEY, LEVESQUE, & PLEZ
Plez, B. 2003, in ASP Conf. Ser. 298, GAIA Spectroscopy: Science and Wallerstein, G. 1958, PASP, 70, 479
Technology, ed. U. Munari (San Francisco: ASP), 189 ———. 1971, ApJ, 169, 195
Plez, B., Brett, J. M, & Nordlund, 8 1992, A&A, 256, 551 ———. 1977, ApJ, 211, 170
Richards, A. M. S., Yates, J. A., & Cohen, R. J. 1998, MNRAS, 299, 319 ———. 1978, Observatory, 98, 224
Robinson, L. J. 1970, Inf. Bull. Variable Stars, 465, 1 Wallerstein, G., & Gonzalez, G. 2001, PASP, 113, 954
———. 1971, Inf. Bull. Variable Stars, 599, 1 Wilson, W. J., & Barrett, A. H. 1968, Science, 161, 778
Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525 Worley, C. E. 1972, ApJ, 175, L93
Smith, N. 2004, MNRAS, 349, L31
Smith, N., Humphreys, R. M., Davidison, K., Gehrz, R. D., Schuster, M. T., &
Krautter, J. 2001, AJ, 121, 1111