1: %\documentclass[12pt,preprint]{aastex}
2:
3: %% manuscript produces a one-column, double-spaced document:
4: %\documentclass[manuscript]{aastex}
5:
6: %% preprint2 produces a double-column, single-spaced document:
7: \documentclass[preprint2]{aastex}
8:
9: %% Sometimes a paper's abstract is too long to fit on the
10: %% title page in preprint2 mode. When that is the case,
11: %% use the longabstract style option.
12: %% \documentclass[preprint2,longabstract]{aastex}
13:
14:
15: \newcommand{\vdag}{(v)^\dagger}
16: \newcommand{\myemail}{mcswain@lehigh.edu}
17:
18: \slugcomment{Accepted to ApJ}
19:
20: \shorttitle{NGC 3766}
21: \shortauthors{McSwain et al.}
22:
23: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
24:
25: \begin{document}
26:
27: \title{The B and Be Star Population of NGC 3766}
28:
29: \author{M.\ Virginia McSwain\altaffilmark{1,2,3}}
30: \affil{Department of Astronomy, Yale University, P.O.\ Box 208101, New
31: Haven, CT 06520-8101; mcswain@lehigh.edu}
32:
33: \author{Wenjin Huang}
34: \affil{Department of Astronomy, California Institute of Technology,
35: MC 105-24, Pasadena, CA 91125; wenjin@astro.caltech.edu}
36:
37: \author{Douglas R.\ Gies, Erika D.\ Grundstrom\altaffilmark{4}}
38: \affil{Center for High Angular Resolution Astronomy, Department of Physics and Astronomy, Georgia State University, P.O.\ Box 4106, Atlanta, GA 30302-4106; gies@chara.gsu.edu, erika@chara.gsu.edu}
39:
40: \author{Richard H.\ D.\ Townsend}
41: \affil{Bartol Research Institute, University of Delaware, Newark, DE 19716; rhdt@bartol.udel.edu}
42:
43: \altaffiltext{1}{Visiting Astronomer, Cerro Tololo Inter-American Observatory.
44: CTIO is operated by AURA, Inc.\ under contract to the National Science
45: Foundation.}
46: \altaffiltext{2}{NSF Astronomy and Astrophysics Postdoctoral Fellow}
47: \altaffiltext{3}{Current address: Department of Physics, Lehigh University, 16 Memorial Drive East, Bethlehem, PA 18015}
48: \altaffiltext{4}{Current address: Physics and Astronomy Department, Vanderbilt University, 1807 Station B, Nashville, TN 37235}
49:
50:
51: \begin{abstract}
52:
53: We present multiple epochs of H$\alpha$ spectroscopy for 47 members of the open cluster NGC 3766 to investigate the long term variability of its Be stars. Sixteen of the stars in this sample are Be stars, including one new discovery. Of these, we observe an unprecedented 11 Be stars that undergo disk appearances and/or near disappearances in our H$\alpha$ spectra, making this the most variable population of Be stars known to date. NGC 3766 is therefore an excellent location to study the formation mechanism of Be star disks. From blue optical spectra of 38 cluster members and existing Str\"omgren photometry of the cluster, we also measure rotational velocities, effective temperatures, and polar surface gravities to investigate the physical and evolutionary factors that may contribute to the Be phenomenon. Our analysis also provides improvements to the reddening and distance of NGC 3766, and we find $E(B-V) = 0.22 \pm 0.03$ and $(V-M_{\rm V})_0 = 11.6 \pm 0.2$, respectively. The Be stars are not associated with a particular stage of main-sequence evolution, but they are a population of rapidly rotating stars with a velocity distribution generally consistent with rotation at $70-80$\% of the critical velocity, although systematic effects probably underestimate the true rotational velocities so that the rotation is much closer to critical. Our measurements of the changing disk sizes are consistent with the idea that transitory, nonradial pulsations contribute to the formation of these highly variable disks.
54:
55: \end{abstract}
56:
57: \keywords{stars: emission-line, Be --- open clusters and associations:
58: individual (\object{NGC 3766})}
59:
60:
61: \section{Introduction}
62:
63: \setcounter{footnote}{3}
64: NGC 3766 is a rich, young open cluster in the Carina spiral arm that is
65: well known for its high content of Be stars \citep{slettebak1985}, and
66: many previous studies of this cluster have focused on the
67: characteristics of these stars to identify their evolutionary status.
68: The cluster has been the target of numerous photometric studies
69: \citep{ahmed1962, yilmaz1976, shobbrook1985, shobbrook1987,
70: moitinho1997, piatti1998, tadross2001, mcswain2005b}. But despite these
71: intensive investigations, the cluster's age and distance remain somewhat
72: uncertain; measurements of its age range from 14.5 to 25 Myr
73: (WEBDA\footnote{The WEBDA database is maintained by E.\ Paunzen and is
74: available online at http://www.univie.ac.at/webda/navigation.html.};
75: \citealt{lynga1987, moitinho1997, tadross2001}), and its distance is
76: between 1.5 and 2.2 kpc. The reddening $E$($B$-$V$) is between 0.16 and
77: 0.22 (see the discussion of \citealt{moitinho1997}).
78:
79: Spectroscopic investigations of NGC 3766 have targeted a limited sample
80: of cluster members, focusing primarily on the Be star and supergiant
81: populations (\citealt{harris1976}; \citealt{mermilliod1982} and
82: references therein; \citealt{slettebak1985, levesque2005}). Even the
83: eclipsing double-lined spectroscopic binary BF Centauri (= HD 100915), a
84: member of NGC 3766, has been largely neglected by modern spectroscopic
85: observations (\citealt{clausen2007} and references therein). For most
86: cluster members, no detailed information about their physical
87: characteristics such as temperature, gravity, rotation, and metallicity
88: are known.
89:
90: In this work, we present red and blue optical spectra for both normal
91: B-type and Be stars in the cluster. Like many prior studies of NGC
92: 3766, our primary goal is to investigate the Be star population; but
93: unlike other works, we achieve a more complete understanding of this
94: subset of B stars by comparing these emission-line objects to their
95: non-emission counterparts. Therefore we present measurements of the
96: effective temperature, $T_{\rm eff}$, surface gravity, $\log g$, and in
97: most cases the projected rotational velocity, $V \sin i$, for 26 normal
98: B stars and 16 Be stars in NGC 3766. We use these results to improve
99: the known reddening and distance to the cluster. From multiple epochs
100: of H$\alpha$ spectroscopy, we also investigate the variability of the
101: circumstellar disks and estimate the disk mass loss/gain rates for 11 Be
102: stars. Finally, we use the observed disk masses and angular momenta to
103: show that nonradial pulsations are a possible origin for the disks, and
104: they probably fill during short-lived bursts of mass flow from the
105: stellar surface.
106:
107:
108: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
109:
110: \section{Observations}
111:
112: We obtained spectra of NGC 3766 during multiple observing runs in 2003
113: March, 2005 February, 2006 May, and 2007 January$-$July using the CTIO
114: Blanco 4-m telescope with the Hydra multi-fiber spectrograph and the
115: CTIO 1.5-m telescope with the Cassegrain spectrograph, operated by the
116: SMARTS Consortium. The details of all runs are summarized in Table
117: \ref{spectroscopy}. Most of the runs targeted the H$\alpha$ emission
118: line profile with low spectral resolution to characterize the Be stars'
119: emission; however, during one run we observed the blue optical region
120: with higher resolving power to observe numerous other H Balmer and
121: \ion{He}{1} line profiles and measure the physical parameters of the
122: cluster members.
123:
124: We selected the targets for each run by giving highest priority to the
125: known Be stars in this cluster (save WEBDA No.\ 232, which was saturated
126: in our photometric study). We then selected other B-type stars in the
127: cluster by ranking them according to their $y$--H$\alpha$ color to
128: preferentially select any weak emission stars that were not detected in
129: our photometry \citep{mcswain2005a, mcswain2005b}. All observations
130: were performed by M.\ V.\ McSwain except the 2007 CTIO 1.5-m runs, which
131: were taken in service mode by a SMARTS observer. For the Hydra
132: observations, we generally began by taking short exposures and then
133: parking the fibers used for the brightest stars to avoid saturation in
134: the longer exposures. Not all of the known Be stars could be observed
135: in one fiber configuration, so we took three to four exposures each of
136: two configurations to observe all of the targets. Therefore up to eight
137: exposures of each star were obtained with the Hydra runs. We also
138: observed a HeNeAr comparison lamp source just before and after the set
139: of cluster observations for wavelength calibrations. For the CTIO 1.5-m
140: observations, we alternated each stellar observation with a Ne
141: comparison lamp spectrum.
142:
143: The CTIO 1.5-m spectra were reduced and rectified to a unit continuum
144: using standard routines for slit spectra in IRAF. All of the Hydra
145: spectra were zero corrected using standard routines in IRAF, and they
146: were flat fielded, wavelength calibrated, and sky subtracted in IRAF
147: using the \textit{dohydra} routine. In comparing the slit spectra and
148: fiber spectra for many of the same objects in our data set, we find no
149: evidence of systematic differences in the background subtraction due to
150: cluster nebulosity. For each Hydra spectral configuration, we
151: transformed the observations to a common heliocentric wavelength grid
152: and co-added them to achieve good S/N for each star.
153:
154: The complete sample of stars presented in this work is listed in Table
155: \ref{eqwidth}. Column 1 gives each star's identification number based
156: on the assigned number in \citet{mcswain2005b}; the corresponding WEBDA
157: numbers are given in column 2, where available. We obtained H$\alpha$
158: spectra for each of these stars during at least three epochs in most
159: cases, and these are shown in Figures \ref{specvar1}--\ref{specvar3}.
160:
161:
162: \placetable{eqwidth}
163: \placefigure{specvar1}
164: \placefigure{specvar2}
165: \placefigure{specvar3}
166:
167: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
168:
169: \section{Physical Parameters from Spectral Models}
170:
171: \subsection{$V \sin i$ Measurements}
172:
173: We began our investigation of each star's physical parameters by
174: generating a grid of synthetic, plane-parallel, local thermodynamic
175: equilibrium (LTE) atmospheric models using the Kurucz ATLAS9 code
176: \citep{kurucz1994}. We adopted solar abundances and a microturbulent
177: velocity of 2 km~s$^{-1}$ for these stars, which corresponds to the mean
178: microturbulence observed among late-type, main-sequence (MS) B stars
179: \citep*{lyubimkov2004}. Each atmospheric model was then used to
180: calculate a grid of model spectra using SYNSPEC \citep{lanz2003}.
181:
182: For the 38 stars with available blue spectra, we made a preliminary
183: estimate of their effective temperature and gravity, $T_{\rm eff}$ and
184: $\log g$ respectively, by comparing the observed H$\gamma$, H$\delta$,
185: \ion{He}{1} $\lambda4143$, and \ion{He}{1} $\lambda4471$+\ion{Mg}{2}
186: $\lambda4481$ line profiles to our grid of Kurucz spectral models. To
187: measure $V \sin i$, we compared the observed \ion{He}{1} line profiles
188: to the model profiles convolved with a limb-darkened, rotational
189: broadening function and a Gaussian instrumental broadening function. We
190: determined the best fit over a grid of values, spaced 2~km~s$^{-1}$
191: apart, minimizing the mean square of the deviations, rms$^2$. The
192: formal error, $\Delta V \sin i$, is the offset from the best-fit value
193: that increases the rms$^2$ by $2.7 \, \rm rms^2$/$N$, where $N$ is the
194: number of wavelength points within the fit region. Our measured $V \sin
195: i$ and $\Delta V \sin i$ are listed in columns 3--4 of Table
196: \ref{params}.
197:
198: Even the \ion{He}{1} lines may contain some weak emission in Be stars,
199: partially filling and narrowing their line profiles. Furthermore, a
200: number of the Be stars show evidence of narrow ``shell'' line components
201: (formed in the outer disk), and the presence of a shell component may
202: make the profile appear too narrow in some cases. Therefore we consider
203: our $V \sin i$ measurements for the Be stars to be lower limits.
204: However, we note that the \ion{He}{1} lines do not exhibit obvious signs
205: of emission among most of the Be stars, and these lines are much less
206: susceptible to emission than the H$\gamma$ or H$\delta$ lines. For the
207: case of No.\ 154, a shell star with strong emission and contamination
208: present in the \ion {He}{1} lines, we used the \ion{Mg}{2} $\lambda4481$
209: line to measure $V \sin i$.
210:
211: There are few previous measurements of $V \sin i$ for members of NGC
212: 3766 in the literature, but we found that 10 stars in our sample were
213: also measured by \citet{slettebak1985}. Our $V \sin i$ measurements
214: generally agree well, with the exception of No.\ 154. Slettebak found
215: $V \sin i = 220$~km~s$^{-1}$ for that star, nearly double our measured
216: value. We emphasize that the exceptionally strong He and metal lines of
217: this shell star amplify the difficulty of measuring its $V \sin i$.
218:
219: \subsection{$T_{\rm eff}$ and $\log g$ Measurements of B stars}
220:
221: For the B stars with $T_{\rm eff} < 15,000$ K, we used the ``virtual
222: star'' method of \citet{huang2006b} to improve our $T_{\rm eff}$ and
223: $\log g$ measurements. (Their virtual star is a spherically symmetric
224: star with constant $T_{\rm eff}$ and $\log g$ across the stellar
225: surface.) They generated detailed H$\gamma$ line profiles using
226: line-blanketed, LTE Kurucz ATLAS9 and SYNSPEC codes. Huang \& Gies show
227: that the H$\gamma$ line strength and equivalent width can be used as
228: starting parameters in a line profile fit to obtain unique values of
229: $T_{\rm eff}$, $\log g$, and their corresponding errors. We used their
230: procedure to measure these quantities from our observed H$\gamma$ line
231: profiles.
232:
233: Among the hotter B-type stars, non-LTE effects alter the equivalent
234: width of the H$\gamma$ line, and thus the LTE Kurucz model line profiles
235: systematically underestimate $T_{\rm eff}$. Therefore we used the new
236: TLUSTY BSTAR2006 grid of metal line-blanketed, non-LTE, plane-parallel,
237: hydrostatic model spectra \citep{lanz2007} to measure $T_{\rm eff}$ and
238: $\log g$ for those stars with $T_{\rm eff} > 15,000$ K. We used their
239: models with solar metallicity and helium abundance and a microturbulent
240: velocity of 2 km~s$^{-1}$. The grid includes $T_{\rm eff}$ from
241: 15,000--30,000 K in increments of 1,000 K and $\log g$ from 1.75--4.75
242: in increments of 0.25 dex. For these hot stars we measured $T_{\rm
243: eff}$ and $\log g$ by comparing the H$\gamma$ line profile to the
244: rotationally and instrumentally broadened model spectral line profiles
245: at each value in the grid, minimizing rms$^2$ across the line region.
246: We then refined our measurements to a higher precision using a linear
247: interpolation between the available line profiles in the grid. Finally,
248: we determined the errors, $\Delta T_{\rm eff}$ and $\Delta \log g$, from
249: the values which produce a $\rm rms^2$ no more than $2.7 \, \rm
250: rms^2$/$N$ greater than the minimum rms$^2$. Our measurements of
251: $T_{\rm eff}$ and $\log g$, with their corresponding errors, are listed
252: in columns 5--8 of Table \ref{params}.
253:
254: Many of the B stars in our sample are rapidly rotating with $V \sin i >
255: 200$ km~s$^{-1}$. Such a rapidly rotating star will experience strong
256: centrifugal forces that distort the star into an oblate spheroidal
257: shape, as recently found for the star Regulus \citep{mcalister2005}.
258: The surface gravity at the equator can therefore be much lower than at
259: the poles. For such a rapid rotator, our measured $T_{\rm eff}$ and
260: $\log g$ represent the average across the visible stellar hemisphere and
261: are therefore biased toward lower values, causing the star to appear
262: more evolved. Because the polar regions are not distorted, the surface
263: gravity at the poles is a better indicator of the evolutionary state of
264: the star. \citet{huang2006b} performed detailed spectroscopic modeling
265: of such distorted rotating stars to determine a statistical correction
266: factor for $\log g$, averaged over all possible $i$, for a variety of
267: stellar models. We made a bilinear interpolation between their models
268: to convert our measured $\log g$ to $\log g_{\rm polar}$ for a more
269: accurate comparison between slow and rapid rotators.
270:
271: For each B star, we also measured its mass, $M_\star$, and radius,
272: $R_\star$, by interpolating between the evolutionary tracks for
273: non-rotating stars from \citet{schaller1992}. The errors $\Delta
274: M_\star$ and $\Delta R_\star$ correspond to our measured $\Delta T_{\rm
275: eff}$ and $\Delta \log g$. Our results for $\log g_{\rm polar}$,
276: $M_\star$, $\Delta M_\star$, $R_\star$, and $\Delta R_\star$ are also
277: listed in Table \ref{params}, columns 9--13.
278:
279:
280: \subsection{$T_{\rm eff}$ and $\log g$ Measurements of Be stars}
281:
282: Star No.\ 196 did not show any sign of Be emission until our most recent
283: observations, hence we included it among the normal B stars and measured
284: $T_{\rm eff}$ and $\log g$ from the H$\gamma$ line in its 2006 blue
285: spectrum. For other Be stars in our blue spectra, the above method to
286: measure $T_{\rm eff}$ and $\log g$ was not useful because the equivalent
287: width of the H$\gamma$ line may be decreased by emission in the line,
288: even if the line profile does not exhibit obvious signs of emission.
289:
290: Our first attempt to measure the Be stars' $T_{\rm eff}$ and $\log g$
291: relied upon the \ion{He}{1} $\lambda\lambda$ 4143, 4388, 4471 lines. We
292: fit these three lines using TLUSTY model spectra using the same
293: procedure described above for the H$\gamma$ line. However, these
294: \ion{He}{1} line strengths are less sensitive to $T_{\rm eff}$, and the
295: line wings show only a very small dependence on $\log g$. Therefore
296: these line fits resulted in very large errors for both parameters in
297: many cases. Furthermore, the Be disks contribute continuum flux that
298: dilutes the apparent strength of the \ion{He}{1} lines, so $\log g$
299: values measured from these lines do not always agree well.
300:
301: To improve our $T_{\rm eff}$ measurements for the Be stars, we turned to
302: available Str\"omgren $m_1$, $c_1$, and $\beta$ indices for our targets
303: (\citealt{shobbrook1985, shobbrook1987}; WEBDA). Several temperature
304: relations for Str\"omgren indices are available in the literature (see
305: \citealt{napiwotzki1993} and references therein), so we began by using
306: our normal B-type stars with $T_{\rm eff} > 15000$~K as calibrators to
307: determine the best relation for our data. We adopted the reddening
308: value of $E(b-y) = 0.15$ (corresponding to $E(B-V) = 0.2$) for NGC 3766
309: \citep{shobbrook1985, moitinho1997}. Eight stars in our sample (Nos.\
310: 16, 42, 49, 54, 57, 161, 170, and 196) have available Str\"omgren
311: indices as well as $T_{\rm eff}$ measured from our $H\gamma$ line fits,
312: and Napiwotzki et al.\ provide eight calibrators with well-known $T_{\rm
313: eff} > 15000$~K (measured from their absolute integrated stellar flux)
314: and available Str\"omgren data. For these 16 B stars, we found the best
315: overall agreement using the temperature relation from
316: \citet{balona1984}, shown in Figure \ref{tempcalibration}. However, we
317: found that $T_{\rm Balona}$ systematically underestimated $T_{\rm eff}$,
318: and a correction factor was necessary to improve their agreement. We
319: performed a linear fit to the data and found the relationship
320: \begin{equation}
321: T_{\rm eff, fit} = 1.052 \: T_{\rm Balona} - 359.636 \; \rm K
322: \end{equation}
323: (excluding one outlying point from \citealt{napiwotzki1993}). The slope
324: of this correction is virtually identical to the values found by both
325: \citet{gies1992} and \citet{cunha1994}. After applying this correction,
326: we found a mean scatter of 264 K from our two independent measurements
327: of $T_{\rm eff}$ for the B stars in NGC 3766.
328:
329: \placefigure{tempcalibration}
330:
331: $T_{\rm Balona}$ relies upon the dereddened $c_0$ index as well as the
332: narrow-band $\beta$ magnitude, so determining an accurate temperature
333: for the Be stars also requires confidence in $\beta$. However, the Be
334: stars' emission makes the $\beta$ magnitude highly unreliable.
335: Therefore we used the B star calibrators from our sample (listed above)
336: to investigate several ($c_0, \beta$) relations in the literature
337: \citep{crawford1978, balona1984b}. The ($c_0, \beta$) diagram is
338: essentially a H-R diagram that reveals temperature and evolutionary
339: trends in a population. In Figure \ref{c0_Hbeta}, we show that the
340: values of $c_0$ and $\beta$ for these B-type calibrators generally agree
341: with the relations for luminosity class III and V stars from
342: \citet{balona1984b}. However, the calibration stars have $3.43 < \log g
343: < 3.98$ ($3.49 < \log g_{\rm polar} < 4.17$) since these hot stars are
344: evolving along the MS, and neither ($c_0, \beta$) relation can be
345: applied to the entire population. Therefore we performed a linear fit
346: to account for the range in evolution, and we found the relationship
347: \begin{equation}
348: \beta_{\rm fit} = 0.417 \: c_0 + 2.545
349: \end{equation}
350: among the B stars in NGC 3766 with $T_{\rm eff} > 15000$~K. The mean
351: scatter between $\beta_{\rm fit}$ and the measured $\beta$ is 0.010,
352: which implies an additional error of 144 K in $T_{\rm eff}$ using the
353: corrected $T_{\rm Balona}$ relation above.
354:
355: \placefigure{c0_Hbeta}
356:
357: \citet{shobbrook1985, shobbrook1987} presented Str\"omgren photometry
358: for nine Be stars that have accompanying blue spectra in this work
359: (excluding No.\ 196, which we used as a calibrator) and four Be stars
360: with only red spectra in this work. Thus we measured $T_{\rm eff}$ for
361: all 13 of these Be stars using the adopted $\beta_{\rm fit}$ and the
362: corrected $T_{\rm Balona}$ as described above. We adopt a total error
363: of $(264^2 + 144^2)^{0.5} = 301$~K for $T_{\rm eff}$ measured with this
364: method, and the results are listed in columns 5--6 of Table
365: \ref{params}.
366:
367: Using this technique, only the ``shell'' Be star No.\ 154 results in
368: $T_{\rm eff} < 15000$~K . The comparisons between $T_{\rm Balona}$ and
369: our independently measured $T_{\rm eff}$ are less reliable below this
370: temperature threshold, hence we used only B star calibrators with
371: $T_{\rm eff} > 15000$~K. However, spectroscopy is even less likely to
372: produce accurate measurements for No.\ 154 since its spectrum is
373: contaminated by \ion{He}{1} emission and the metal lines are
374: exceptionally strong due to the disk's edge-on orientation. Therefore
375: we include in Table \ref{params} its $T_{\rm eff}$ from the corrected
376: relations of \citet{balona1984}, but the errors are somewhat higher than
377: for the other Be stars.
378:
379: The values of $\log g$ are more strongly dependent on $\beta$, so we
380: were reluctant to use our $\beta_{\rm fit}$ with the Str\"omgren
381: relation for $\log g$ given by \citet{balona1984}. Instead, we used the
382: calculated $T_{\rm eff}$ to determine the Be stars' bolometric
383: corrections, BC, from \citet{lanz2007}, at first assuming $\log g=4.0$.
384: For the cooler star No.\ 154, we interpolated the BC from the values for
385: MS stars given by \citet{cox2000}. We calculated each stellar radius,
386: $R_\star$, and luminosity, $L_\star$, using the measured $T_{\rm eff}$,
387: BC, $V$ magnitude (\citealt{shobbrook1985, shobbrook1987}; WEBDA),
388: distance modulus $(V-M_V)_0 = 11.73 \pm 0.33$ \citep{moitinho1997}, and
389: $E(B-V) = 0.2 \pm 0.1$ \citep{shobbrook1985, moitinho1997}. We measured
390: the stellar mass, $M_\star$, from the computed $T_{\rm eff}$ and
391: $L_\star$ by interpolating between the evolutionary tracks of
392: \citet{schaller1992}. Finally, we obtained a preliminary value of $\log
393: g$ from $M_\star$ and $R_\star$. Since the BC is weakly dependent on
394: $\log g$, we improved the BC from the initial estimate and iterated to
395: compute the final $\log g$. We adopt a formal error in $\log g$
396: computed from $\Delta T_{\rm eff}$ and the quoted errors in $(V-M_V)_0$
397: and $E(B-V)$.
398:
399: We note that the scatter between the calibrators' $\log g_{\rm polar}$
400: and their Str\"omgren $\log g$ is identical to the formal error. Based
401: on this good agreement, we do not perform any further correction to
402: obtain $\log g_{\rm polar}$ for the Be stars measured with this
403: technique. However, we show below that the Be stars are more rapidly
404: rotating than the normal B stars, which may make them appear slightly
405: more evolved and artificially brightened. Thus these values of $\log
406: g_{\rm polar}$ are lower limits. The final parameters for these Be
407: stars are listed in Table \ref{params}.
408:
409: Two Be stars in our sample of blue spectra have no available Str\"omgren
410: data in the literature. By coincidence, these stars (Nos.\ 31 and 127)
411: have two of the best $T_{\rm eff}$ and $\log g$ measurements from our
412: preliminary \ion{He}{1} line fits. Therefore we adopt the mean
413: measurements from the \ion{He}{1} $\lambda\lambda$ 4143, 4388, 4471
414: lines, with a formal error determined from the values which produce a
415: $\rm rms^2$ no more than $2.7 \, \rm rms^2$/$N$ greater than the minimum
416: rms$^2$. A contour plot of the errors in $T_{\rm eff}$ and $\log g$ for
417: No.\ 31 is shown in Figure \ref{contour}, and the final values are
418: listed in Table \ref{params}. We determined their $\log g_{\rm polar}$
419: using the same method as the B stars.
420:
421: \placefigure{contour}
422:
423:
424: \subsection{Discussion of Physical Parameters}
425:
426:
427: The resulting values of $T_{\rm eff}$ and $\log g_{\rm polar}$ are
428: plotted in Figure \ref{tlogg} with the corresponding evolutionary tracks
429: for non-rotating stars from \citet{schaller1992}. The symbol sizes are
430: proportional to $V \sin i$ to investigate the relation between rotation
431: and the apparent evolutionary state of each star, but no trends are
432: observed. Also in Figure \ref{tlogg}, we plot the isochrones for 25--50
433: Myr populations from \citet{lejeune2001}. Our distribution of $T_{\rm
434: eff}$ and $\log g_{\rm polar}$ are generally consistent with a
435: population within this range, indicating an age slightly greater than
436: previous estimates for NGC 3766.
437:
438: The masses of the rapidly rotating stars may be overestimated since the
439: evolutionary tracks do not account for rotation. We minimize this
440: effect by using the corrected $\log g_{\rm polar}$ for all stars, but
441: slight mass differences may still be present since rapid rotation is
442: expected to alter the evolution \citep{heger2000, meynet2000}.
443: \citet{mcalister2005} found a small, 15\% mass discrepancy for the rapid
444: rotator Regulus when comparing its mass derived from non-rotating
445: evolutionary tracks and its true mass from a detailed spectroscopic and
446: interferometric analysis.
447:
448: We find that the Be stars are generally among the hotter stars in our
449: sample, consistent with our earlier findings from a photometric
450: investigation of 48 open clusters that Be stars are preferentially found
451: among the more luminous cluster members \citep{mcswain2005b}. There are
452: no systematic differences between the $\log g_{\rm polar}$ of the Be
453: star and normal B star populations of the cluster, indicating that the
454: Be stars in NGC 3766 are distributed across a range of $\log g_{\rm
455: polar}$ and are not associated with any particular stage of the MS
456: evolution of B-type stars. This is also consistent with our earlier
457: results \citep{mcswain2005b} and with \citet{zorec2005}, who performed
458: an evolutionary study of field Be stars. On the other hand,
459: \citet{levenhagen2006} found that field Be stars are preferentially
460: found at later stages of the MS evolution. They did not perform any
461: corrections for gravity darkening among these rapid rotators, so their
462: Be stars may be found closer to the zero-age MS than their results
463: suggest.
464:
465: \placetable{params}
466: \placefigure{tlogg}
467:
468:
469: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
470:
471: \section{Rotational Velocities}
472:
473: Be stars are often described as a population of rapidly rotating B-type
474: stars with a true rotational velocity $V_{\rm rot}$ comparable to the
475: critical velocity $V_{\rm crit}$. (Note that $V_{\rm rot}$ should not
476: be confused with $V$ in our measured $V \sin i$, which may be subject to
477: systematic errors such as gravitational darkening or weak emission in
478: the \ion{He}{1} lines.) However, a recent study by \citet{cranmer2005}
479: has cast some doubt on their fast rotation. He compared available $V
480: \sin i$ measurements in the \citet{yudin2001} database of Oe, Be, and Ae
481: stars to the predicted distribution of $V \sin i$ accounting for gravity
482: darkening, limb darkening, and observational effects. He found that the
483: Be stars of the Yudin database have intrinsic rotations between
484: 40--100\% critical, with more early-type Be stars having significantly
485: subcritical rotation. However, the Yudin database is a compilation of
486: measurements from many different authors and instrumental setups, and
487: therefore the available $V \sin i$ may contain significant systematic
488: differences. Our measurements of $V \sin i$ rely upon data of identical
489: origin and measurements of the same spectral lines in nearly every case.
490: Therefore we have greatly reduced the systematic differences among our
491: Be star measurements, and our measurements are also a reliable
492: comparison of Be stars relative to normal B stars.
493:
494: A rotationally distorted star has an equatorial radius $R_{\rm e} = 1.5
495: R_{\rm p}$ in the Roche approximation; here, $R_{\rm p}$ is the star's
496: polar radius. For simplicity, we assume $R_{\rm p}$ is equal to the
497: radius $R_\star$ derived from the position in Figure \ref{tlogg} and
498: given in Table \ref{params}. The resulting critical velocity is
499: \begin{equation}
500: V_{\rm crit} = \sqrt{\frac{G M_{\star}}{R_{\rm e}}}
501: \end{equation}
502: and is included in Table \ref{params}. A small mass discrepancy, as
503: found for Regulus \citep{mcalister2005}, will not affect $V_{\rm crit}$
504: significantly since a 15\% error in mass produces only a 4\% error in
505: $V_{\rm crit}$. Here we investigate the rotational properties of
506: several edge-on Be stars in our sample and compare the distribution of
507: Be star velocities with the normal B stars to compare the two
508: populations.
509:
510: Two Be stars in our sample, Nos.\ 92 and 139, show H$\alpha$ emission
511: only in the line wings, with a deep absorption profile, suggesting the
512: disks are observed nearly edge-on. This is probably a good assumption
513: for No.\ 139 since it has $V/V_{\rm crit} \geq 0.7$. However, No.\ 92
514: may not be edge on since its H$\alpha$ profiles between 1985$-$1990
515: \citep{balona1991} do not resemble the profiles we observe. Either the
516: disk is precessing or other structural changes have occurred. A third
517: star, No.\ 154, exhibits a shell spectrum with strong metal lines that
518: also suggests an edge-on orientation.
519:
520: If we assume that all of the Be stars have $V/V_{\rm crit} = 0.95$, then
521: our measured $V \sin i$ indicate that stars 92, 139, and 154 have $i =
522: 34^\circ$, $48^\circ$, and $15^\circ$, respectively. Assuming a slower,
523: $V/V_{\rm crit} = 0.70$ increases the derived values to $i = 49^\circ$,
524: $90^\circ$, and $20^\circ$ respectively, but the low $i$ for Nos.\ 92
525: and 154 are still inconsistent with the observed line profiles.
526: However, if we fix $i = 80^\circ$ for all three stars, we find that
527: $V/V_{\rm crit} = 0.53$ for No.\ 92, $V/V_{\rm crit} = 0.71$ for No.\
528: 139, and $V/V_{\rm crit} = 0.24$ for No.\ 154. \citet{townsend2004}
529: show that measured $V \sin i$ may be too low due to gravitational
530: darkening in highly distorted, rapidly rotating stars. But even
531: accounting for a 20--33\% underestimate in $V \sin i$ for Nos.\ 92 and
532: 154, we do not find that these two Be stars are near critical rotation.
533:
534: The distributions of $V \sin i$ for both the Be stars and normal B stars
535: are ploted in Figure \ref{vsini}. Although this sample of 38 stars is
536: small, this study is the first to measure $V \sin i$ for both groups
537: using a consistent method and a uniform data set. Our measured $V \sin
538: i$ for the Be stars may be only lower limits, but we clearly find that
539: the Be stars of NGC 3766 are more rapidly rotating than the normal B
540: star population. In Figure \ref{vsini}, we also include theoretical
541: distributions of two uniform, rapidly rotating populations with $V = 0.7
542: \; V_{\rm crit}$ and $V = 0.8 \; V_{\rm crit}$, assuming $V_{\rm crit} =
543: 386$~km~s$^{-1}$ (the mean value among the Be stars, excluding No.\ 154
544: due to its large measurement errors). The distribution of Be stars is
545: consistent with $0.7 \; V_{\rm crit} < V < 0.8 \; V_{\rm crit}$ in most
546: cases. However, if each measured $V \sin i$ is underestimated by
547: 20--33\% due to gravitational darkening, as claimed by
548: \citet{townsend2004}, the distribution of Be star velocities is
549: consistent with $V_{\rm rot} \geq 0.84 \; V_{\rm crit}$. Nos.\ 92 and
550: 154 represent significant exceptions to this rule, as discussed above.
551:
552: \placefigure{vsini}
553:
554:
555:
556: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
557:
558: \section{Other Results from Spectra}
559:
560: We do not wish to broaden the scope of this paper to discuss the radial
561: velocities and helium abundances observed in our spectra. However, we
562: found several instances that suggest these topics are worth exploring in
563: future studies of NGC 3766.
564:
565: Figure \ref{specvar1} includes the variable H$\alpha$ profile of No.\
566: 61. The absorption line is deep and narrow in 2005, yet wide and
567: shallow in 2006 and 2007. These line profile changes are not consistent
568: with a Be star, and they probably indicate line blending in a
569: double-lined spectroscopic binary (SB2). Likewise, No.\ 197 may be an
570: SB2. We observed double \ion{Mg}{2} 4481 lines in 2006, and Figure 3
571: shows a large wavelength shift in its H$\alpha$ line profile between
572: 2005$-$2006. There is possible line blending visible in the 2005 and
573: 2007 H$\alpha$ profiles.
574:
575: We measured the physical parameters of the B stars using spectral models
576: that have solar helium abundances, and most were consistent with a solar
577: abundance. However, we noticed a large number that have abnormal He
578: abundances; Nos.\ 41, 55, 94, 118, and 178 appear to be He strong, while
579: Nos.\ 126, 129, 170, 173, and 197 appear He weak. Since all cluster
580: members are expected to have the same He abundance, this may indicate
581: widespread magnetic fields in the cluster which can produce nonuniform
582: distributions of He across the stellar photospheres (see discussion of
583: \citealt{huang2006b}). The He weak stars may also be binaries with
584: cooler companions adding flux but diluting the appearance of the
585: \ion{He}{1} lines.
586:
587: Finally, we used our spectra to reinvestigate the reddening and distance
588: to NGC 3766. For the nonemission B stars with $\log g_{\rm polar} >
589: 3.9$, we used the MS relation of \citet{harmanec1988} to assign a
590: spectral type to each star based on its measured $T_{\rm eff}$. We then
591: used the MS relation of \citet{wegner1994} to find the stars' intrinsic
592: colors, $(B-V)_0$. Observed $B-V$ colors were obtained from WEBDA. For
593: the 17 B-type, MS stars in our sample, we found $E(B-V) = 0.223 \pm
594: 0.030$. We then used the B stars' $V$ magnitude (WEBDA), BC
595: (\citealt{lanz2007}, \citealt{malagnini1986}), and our measured
596: $R_\star$ to compute the distance modulus, $(V-M_{\rm V})_0$. The
597: resulting $(V-M_{\rm V})_0 = 11.42 \pm 0.15$. To determine the mean
598: $(V-M_{\rm V})_0$, we excluded No.\ 45 with a somewhat lower $(V-M_{\rm
599: V})_0$ of 10.3, probably a foreground star.
600:
601: While the B stars of NGC 3766 are rotating more slowly than the Be
602: stars, they have a mean $V \sin i/V_{\rm crit} = 0.5$. Assuming a
603: random distribution of $i$, this corresponds to a mean $V/V_{\rm crit} =
604: 0.7$. Such rapidly rotating stars are probably rotationally distorted
605: with surface areas $\approx 1.2$ times larger than spherical star with
606: $R_\star = R_{\rm polar}$, causing them to appear overluminous. The
607: distance modulus is then underestimated by a factor of $2.5 \log 1.2 =
608: 0.21$. With a realistic distribution of $V$ and $i$, this correction
609: may not be appropriate for all stars, and the range in $\Delta (V-M_{\rm
610: V})_0$ is $0.02-0.30$ for $0.28 \leq V/V_{\rm crit} \leq 0.83$.
611: We apply this correction to find a final $(V-M_{\rm V})_0 = 11.6 \pm
612: 0.2$ for NGC 3766, corresponding to a distance of $1.9-2.3$ kpc. Our
613: new values of $E(B-V)$ and $(V-M_{\rm V})_0$ are highly consistent with
614: previous results for this cluster.
615:
616:
617: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
618:
619: \section{Be Star Variability \label{var}}
620:
621: While the long term H$\alpha$ variability of Be stars has often been
622: noted in the literature (e.g.\ \citealt{porter2003}), few studies have
623: attempted long term monitoring of a large sample of Be stars to quantify
624: their variability. \citet{hubert1998} investigated 273 bright Be stars
625: with $V < 7.5$ that were observed by Hipparcos between 1989 August and
626: 1993 August. They found that early type Be stars exhibit a very high
627: degree of variability while most late Be stars maintained a constant
628: magnitude during the duration of the Hipparcos mission. They also
629: identified 14 Be stars with recurrent short-lived outbursts ($0.06 \leq
630: \Delta Hp \leq 0.3$ over timescales of 50$-$500 days) and 8 with
631: long-lived outbursts ($\Delta Hp \geq 0.12$ over timescales of $> 500$
632: days) in the sample. However, the Hipparcos magnitude, $Hp$, covers a
633: broad waveband ranging from 3400 to 8500 \AA, and it traces large scale
634: variations in the continuum from scattered light associated with the
635: disk. H$\alpha$ spectroscopy is better suited to measuring variations
636: over a larger dynamic range in the disks of Be stars of all spectral
637: types.
638:
639: Figures \ref{specvar1}$-$\ref{specvar3} show the H$\alpha$ line profile
640: variations of all 47 stars in our spectroscopic sample. We were
641: intrigued to discover a new Be star (No.\ 130) in NGC 3766 based on our
642: comparison of its H$\alpha$ line profiles. In 2005 February, the
643: absorption line was broad and shallow, with some bumps that suggest a
644: weak emission disk, nonradial pulsations, or simply noise. By 2006 May,
645: however, the depth of the line had increased while its width remained
646: nearly the same, but a slight asymmetry in the line suggested that
647: perhaps a very weak disk may still have been present. The line remained
648: in absorption during our most recent 2007 observations. While an
649: inverse relation between the line width and depth is associated with
650: line blending in an SB2, the observed variations in No.\ 130 cannot be
651: attributed to such line blending. Instead, the three sets of
652: observations indicate the presence of a weak circumstellar disk in 2005
653: that largely disappeared by 2006.
654:
655: In fact, we were startled to find that a total of 11 Be stars in our
656: sample exhibit significant changes in their disk state between
657: 2003--2007. Stars 25, 31, 73, 83, 92, 98, 119, 130, 133, 139, and 196
658: each show H$\alpha$ in absorption in at least one of the four epochs
659: available, yet each star also experiences at least one epoch when the
660: H$\alpha$ line is partially or fully filled with emission. These
661: observations suggest that the mechanism responsible for the disk
662: formation is unstable over timescales of only a few years, consistent
663: with the results of \citet{hubert1998}. The disk growth of No.\ 31
664: during 2007 reveals that the disk formation can be very rapid, only
665: requiring a few days or weeks, as observed in other Be stars (e.g.
666: \citealt{grundstrom2007}). However, the high fraction of Be stars that
667: show significant variability in a single cluster is unprecedented, and
668: the cluster should be monitored with increased frequency to measure an
669: accurate timescale of the disk state changes.
670:
671: Five additional Be stars (47, 127, 154, 198, and 200) show H$\alpha$
672: emission in every observation, although the emission strength is usually
673: variable. Such variations in H$\alpha$ emission strength are typical
674: among Be stars.
675:
676: To quantify the observed changes in the Be star disks, we measured the
677: equivalent width, $W_\lambda$, of the H$\alpha$ line in each of our red
678: spectra by normalizing each spectrum to a unit continuum and integrating
679: over the line profile. Columns 3$-$6 of Table \ref{eqwidth} give the
680: measured values of $W_{\lambda}$ for each spectrum. The error in each
681: measurement is about 10\% due to noise in the continuum region.
682:
683: With multiple epochs of H$\alpha$ observations available for so many
684: variable Be stars, it is worthwhile to estimate the changing size of the
685: circumstellar disks. \citet{grundstrom2006} describe a simple model to
686: measure the ratio of the projected effective disk radius to the stellar
687: radius, $R_{\rm disk}$/$R_{\star}$, and the density at the base of the
688: disk, $\rho_0$, using $W_{\lambda}$ and $T_{\rm eff}$ as input
689: parameters. We determined the disk inclination $i$ by assuming that
690: each star is rotating with $V$ at 70\% of $V_{\rm crit}$ in most cases.
691: We also assumed a disk truncation radius of 100 $R_{\star}$, the nominal
692: value unless a close binary companion is present. The disk temperature
693: is assumed to be constant at $0.6 \; T_{\rm eff}$.
694:
695: To estimate the total masses of the disks, we used an axisymmetric,
696: isothermal density distribution,
697: \begin{equation} \label{densityeqn}
698: \rho(r,z) = \rho_0 \left (\frac{R_\star}{r} \right )^n \exp \left[-\frac{1}{2}
699: \left(\frac{z}{H(r)}\right)^2 \right]
700: \end{equation}
701: \citep{carciofi2006} and a radial density exponent $n = 3$, typical of
702: other Be star disks \citep{gies2007}. The scale height of the disk is
703: \begin{equation}
704: H(r) = H_0 \left( \frac{r}{R_{\star}}\right)^\beta,
705: \end{equation}
706: where
707: \begin{equation}
708: H_0 = \frac{a}{V_{\rm crit}} R_{\star},
709: \end{equation}
710: \begin{equation}
711: a = \sqrt{\frac{kT}{\mu m_{\rm H}}},
712: \end{equation}
713: and $\beta = 1.5$ for an isothermal disk \citep{bjorkman2005, carciofi2006}.
714:
715: The resulting disk properties are listed in Table \ref{bevar} for each
716: Be star with an available $T_{\rm eff}$ measurement. The time of each
717: observation is listed in column 2, and the dates for the CTIO 4m+Hydra
718: observations are less precise since these $W_{\rm H\alpha}$ were
719: measured from co-added spectra rather than individual spectra obtained
720: with the CTIO 1.5m telescope. Our determined values for $i$, $\rho_0$,
721: $R_{\rm disk}/R_{\star}$, and $M_{\rm disk}$ are listed in columns 4--7.
722:
723: Certainly, these disk measurements should be viewed with caution since
724: the true density profiles may be quite different from the assumed
725: distribution.
726: The disk density exponent may be different from our assumed value of
727: $n=3$, and the resulting mass estimates may differ from our results by
728: orders of magnitude. Contemporaneous optical and infrared spectra have
729: also revealed evidence of density waves that alter the azimuthal disk
730: structure \citep{wisniewski2007}. Finally, the disks are likely not
731: isothermal as we assumed, and the true thermal structure may be very
732: complex \citep{carciofi2006}. Therefore the processes associated with
733: the H$\alpha$ emission profile are much more complex than assumed in the
734: model of \citet{grundstrom2006}, but by using their simple model we
735: obtain relative estimates of the variations in disk base density that
736: are sensible provided the disk density exponent is assumed constant in
737: time.
738:
739: \placetable{bevar}
740:
741: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
742:
743:
744: \section{Disk Formation by Nonradial Pulsations}
745:
746: The estimates for $\rho_0$ and $M_{\rm disk}$ above are not sensitive to
747: $i$ and offer a consistent method to measure changes in the size of the
748: disks. Thus we can estimate the disk growth rate (or dissipation rate),
749: $\Delta M_{\rm disk}/\Delta t$. Our generally sparse observations
750: cannot determine whether $\Delta M_{\rm disk}/\Delta t$ is constant or
751: highly variable over timescales less than one year; however there is an
752: indication from the two closely spaced observations of Nos.\ 25 and 31
753: that the disk size can change rapidly. We provide our measured $\Delta
754: M_{\rm disk}/\Delta t$ between each pair of observations in Table
755: \ref{bevar}. The long-term buildup rates are comparable for every Be
756: star in our sample, hinting that all of the disks might be formed by the
757: same mechanism. Finally, the buildup and dissipation rates are also
758: comparable in magnitude.
759:
760: Many Be stars are known to exhibit nonradial pulsations (NRP) with the
761: mode $\ell = 2$, $m = \pm2$ \citep{rivinius2003}, and such pulsations
762: are commonly proposed as the source of kinetic energy to inject material
763: into the disk \citep{porter2003}. The $m = +2$ pulsational mode is
764: retrograde, and naively it will counteract the rotational velocity of
765: the star and hinder disk formation. However, \citet{townsend2005} shows
766: that retrograde mixed modes behave in a way that could contribute to Be
767: disk formation. Although their phase velocity is retrograde, their
768: group velocity is prograde. The density enhancements occur when the
769: pulsational velocity perturbation is in the same direction as the
770: rotation -- a configuration favorable to mass ejection. He also found
771: the mixed mode instability strip likely overlaps with the temperatures
772: and spectral types of known Be stars if their rotation is nearly
773: critical.
774:
775: Observations of known $\beta$ Cephei pulsators and slowly pulsating B
776: stars indicate that their pulsation modes have velocity amplitudes on
777: the order of 10 km~s$^{-1}$ (combining the radial, azimuthal, and
778: longitudinal components of the total velocity vector;
779: \citealt{deridder2001}). \citet{owocki2005} shows that the velocity
780: needed for ejection into the disk is $\Delta V_{\rm orb} = V_{\rm crit}
781: - V_{\rm rot}$. For a weak atmospheric process such as NRP, the
782: atmospheric sound speed $a \sim 12$ km~s$^{-1}$ must be comparable to
783: $\Delta V_{\rm orb}$. If we assume that the stars have $V_{\rm rot} =
784: 0.95 \; V_{\rm crit}$, then $\Delta V_{\rm orb} \sim 20$ ~km~s$^{-1}$,
785: highly comparable to $a$ and the observed pulsation velocities in other
786: NRP stars \citep{deridder2001}. As we show above, this assumption of
787: near critical rotation is reasonable for most of the Be stars in NGC
788: 3766.
789:
790: To compare the NRP velocity with the outflow velocity required to fill
791: the Be disks, we use the equation of mass continuity for an equatorial
792: outflow with velocity $v_{\rm r, eq}$:
793: \begin{equation}
794: \sigma (r) = \frac{\Delta M_{\rm disk}/\Delta t}{2 \pi r \: v_{\rm r, eq}}
795: \end{equation}
796: where the surface density of the disk is
797: \begin{equation}
798: \sigma (r) = \int^{\infty}_{-\infty} \rho (r,z) \: dz.
799: \end{equation}
800: Since we remove the vertical dependence of the disk density by using its
801: surface density, this method effectively assumes that all of the
802: material leaves the star through a cylindrical surface of arbitrary
803: height. For a disk reaching down to the stellar surface, the mass must
804: flow across this surface, so we use the surface density at the stellar
805: surface, $\sigma(R_{\rm eq})$, in the continuity relation. Since the
806: $\ell = 2$, $m = \pm 2$ modes have the greatest pulsational amplitude at
807: the stellar equator, where the effective surface gravity is also a
808: minimum, it is reasonable to expect the majority of the disk material to
809: be ejected from the equator in the NRP model. The observed disk
810: formation rates are generally $\Delta M_{\rm disk}/\Delta t = 10^{-12} -
811: 10^{-11}$ ~$M_\odot$ ~yr$^{-1}$ with $\rho_0 = 10^{-12}$ ~g~cm$^{-3}$
812: from Table \ref{bevar}. This implies a very slow surface flow on
813: average, $v_{\rm r, eq} \sim .01 - 0.1$ ~km~s$^{-1}$. However, a few
814: observations reveal that the surface flow can be an order of magnitude
815: faster, but this is still consistent with NRP.
816:
817: Likewise, we use our observed disk densities to demonstrate that NRP are
818: a capable source of angular momentum, $L$, for disks. The rotational
819: inertia, $I$, of a thin disk shell is given by
820: \begin{equation}
821: I(r) = 2 \pi r^3 \sigma(r) \: dr
822: \end{equation}
823: and the total angular momentum of the disk, extending from the stellar
824: surface to $100 R_\star$, is
825: \begin{equation}
826: L = \int^{100 R_\star}_{R_{\rm eq}} I(r) \: \omega(r) \: dr.
827: \end{equation}
828: where $\omega(r) = (GM_\star)^{0.5} \: r^{-1.5}$ is the angular velocity
829: of a Keplerian disk. We find that $L \sim
830: 10^{43}-10^{44}$~g~cm$^2$~s$^{-1}$ for the observed disks, with $\Delta
831: L/\Delta t \sim 10^{32}-10^{36}$~g~cm$^2$~s$^{-2}$. \citet{osaki1986}
832: shows that the angular momentum flux in the equatorial plane due to NRP
833: with $\ell = 2$, $m = \pm2$ is given by
834: \begin{equation}
835: \frac{dL}{dt} = 2 \pi R_{\rm eq}^2 \: \sigma(R_{\rm eq}) \: A^2 k \: \sin \delta.
836: \end{equation}
837: Here, $A$ and $2Ak$ are the radial and azimuthal amplitudes of
838: pulsation, respectively, and $\delta$ is the phase shift between the two
839: velocity components. Thus we find that a small azimuthal pulsation with
840: $2Ak \sim 0.01 - 0.1$~km~s$^{-1}$ can, in principle, provide enough
841: angular momentum to eject material into the observed disks, and both the
842: radial and azimuthal amplitudes are highly consistent with the observed
843: pulsation amplitudes of other B stars with NRP \citep{deridder2001}.
844:
845: Our results suggest that the NRP may be a transitory phenomenon for the
846: Be stars, and the disks may fill substantially during short periods of
847: surface activity. Evidence for changing pulsational modes among several
848: stars in NGC 3766 has been observed by \citet{balona1991},
849: \citet{vanvuuren1988}, and \citet{balona1986}. They found periodic
850: light curve variations in several Be stars in NGC 3766 (our Nos.\ 47,
851: 92, 133, 154, and 200). Additional Be stars, our Nos.\ 98, 130, 198
852: also exhibit light curve variations that are possibly periodic. In
853: several cases, the shapes and possibly periods of the light curves
854: change with timescales of only a few weeks. \citet{balona1991} argue
855: that this is due to magnetic fields, but the light curves are not
856: qualitatively similar to variations observed in other magnetic stars,
857: such as the prototype $\sigma$ Ori E \citep{oksala2007}. The changing
858: light curves among NGC 3766 members are probably more consistent with
859: changes in active pulsation modes over short timescales.
860:
861:
862:
863: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
864:
865:
866: \section{Conclusions}
867:
868: Our spectroscopic analysis of NGC 3766 has revealed that Be stars may be
869: much more common than we originally thought. In our photometric study
870: of NGC 3766 \citep{mcswain2005b}, we found up to 13 Be stars (5
871: definite, 8 uncertain) out of an expected 191 B-type stars, not counting
872: the one Be star that saturated our photometry. The new total of 16 Be
873: stars is 23\% greater. Among these 16 Be stars, 2--5 of them appear to
874: have almost no disk at any given time, and an additional 2--4 have
875: extremely subtle emission in their H$\alpha$ line profile that could
876: easily be mistaken for other phenomena (such as NRP manifesting
877: themselves as bumps moving across the line or SB2 line blending).
878: Therefore 25--50\% of the Be stars may go undetected in a single
879: spectroscopic observation, and photometric snapshots are even less
880: likely to discern such weak emitters. We note four stars (Nos.\ 27, 45,
881: 49, and 77) that were found to be possible or likely Be stars in the
882: photometric study by \citet{shobbrook1985, shobbrook1987}, but they
883: never showed emission during our observations and thus remain
884: unconfirmed. The existence of transitory, weak disks (especially Nos.\
885: 130 and 196) could mean that many more Be stars are waiting to be
886: discovered.
887:
888: For our total sample of 48 Southern open clusters in our photometric
889: survey, we found a low Be fraction of $2-7$\% \citep{mcswain2005b}.
890: Considering the very weak disks that are observed in NGC 3766 and the
891: exceptionally high variability among the cluster's Be population, the
892: total fraction of Be stars could be much greater. We are currently
893: performing a similar spectroscopic study of several other clusters from
894: our survey, and we will address those results in a future paper.
895:
896: While the Be stars of NGC 3766 are not distinguishable from normal
897: B-type stars by their evolutionary states, they do form a population of
898: rapidly rotating stars. With two exceptions, their measured velocities
899: are consistent with a uniform population of rapid rotators having $V =
900: 0.7-0.8 \; V_{\rm crit}$. Gravitational darkening and weak emission in
901: the \ion{He}{1} lines may mean that these velocities are underestimated
902: by as much as 33\% \citep{townsend2004}, so the true $V_{\rm rot}$ is
903: probably at least $0.84 \; V_{\rm crit}$. From the measured changes in
904: the disks' masses and angular momenta, NRP are a capable source for the
905: mass flow into the equatorial plane. The pulsations may be a transitory
906: phenomenon, however, and the variable nature of the Be stars probably
907: reflects dramatic changes in the surface activity.
908:
909:
910: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
911:
912: \acknowledgments
913: We thank the referee, Phil Massey, for providing comments that improved
914: this work. Also, we are grateful to Giovanni Carraro, Mark
915: Pinsonneault, and Swetlana Hubrig for helpful discussions that
916: contributed to this work. We are grateful to Yale University and the
917: SMARTS Consortium for providing observing time at the CTIO 1.5m
918: telescope. This research has made use of the WEBDA database, operated
919: at the Institute for Astronomy of the University of Vienna. MVM was
920: supported by an NSF Astronomy and Astrophysics Postdoctoral Fellowship
921: under award AST-0401460. This work was also supported by the National
922: Science Foundation under grant AST-0606861 (DRG) and by NASA under grant
923: LTSA/NNG05GC36G (RHDT).
924:
925: Facilities: \facility{CTIO}.
926: \input{ms.bbl}
927:
928:
929: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
930:
931: \clearpage
932: \begin{figure}
933: \includegraphics[angle=0,scale=.7]{f1.ps}
934: \caption{H$\alpha$ profiles, labeled by MG ID number. Spectra from 2003, where available, are shown with \textit{dotted lines}, spectra from 2005 are shown with \textit{dot-dashed lines}, spectra from 2006 are shown with \textit{dashed lines}, and spectra from 2007 are shown with \textit{solid lines}.
935: \label{specvar1}
936: }
937: \end{figure}
938:
939: \clearpage
940: \begin{figure}
941: \includegraphics[angle=0,scale=.7]{f2.ps}
942: \caption{H$\alpha$ profiles in the same format as Fig.\ \ref{specvar1}.
943: \label{specvar2}
944: }
945: \end{figure}
946:
947: \clearpage
948: \begin{figure}
949: \includegraphics[angle=0,scale=.7]{f3.ps}
950: \caption{H$\alpha$ profiles in the same format as Fig.\ \ref{specvar1}.
951: \label{specvar3}
952: }
953: \end{figure}
954:
955: \clearpage
956: \begin{figure}
957: \includegraphics[angle=90,scale=0.3]{f4.ps}
958: \caption{$T_{\rm eff}$ measured for the B-type temperature calibration stars from this work (\textit{diamonds}) and from Napiwotzki (1993; \textit{triangles}) compared to their calculated $T_{\rm Balona}$ \citep{balona1984}. A linear fit to the two temperature scales (\textit{solid line}) and the 1:1 agreement (\textit{dotted line}) are also shown.
959: \label{tempcalibration}
960: }
961: \end{figure}
962:
963: \begin{figure}
964: \includegraphics[angle=90,scale=0.3]{f5.ps}
965: \caption{Str\"omgren $c_0$ index plotted against the $\beta$ magnitude for the B-type temperature calibration stars from this work (\textit{diamonds}). Be stars are also plotted (\textit{filled diamonds}) to illustrate the contamination in $\beta$ due to their disk emission. We also show the ($c_0, \beta$) relations for luminosity class V and III stars (\citealt{balona1984b}; \textit{dotted lines}) and our linear fit to this evolving population (\textit{solid line}).
966: \label{c0_Hbeta}
967: }
968: \end{figure}
969:
970: \begin{figure}
971: \includegraphics[angle=90,scale=0.3]{f6.ps}
972: \caption{A contour plot of the $1\sigma$, $2\sigma$, and $3\sigma$ errors in the \ion{He}{1} $\lambda4388$ line fit for No.\ 31. We adopted the center of the $1\sigma$ error region as the best fit value, and the extent of this region indicates the sizes of the error bars for $T_{\rm eff}$ and $\log g$.
973: \label{contour}
974: }
975: \end{figure}
976:
977: \clearpage
978: \begin{figure}
979: \includegraphics[angle=90,scale=0.3]{f7.ps}
980: \caption{$T_{\rm eff}$ and $\log g_{\rm polar}$ are plotted with the evolutionary tracks of \citet{schaller1992} (\textit{solid lines}) and isochrones of \citet{lejeune2001} (\textit{dotted lines}). The zero-age MS mass of each evolutionary track is labeled along the bottom, and the age of each isochrone is labeled along the top. Normal B-type stars are shown as open diamonds while Be stars are filled diamonds, and each symbol size is proportional to the star's $V \sin i$. \label{tlogg}
981: }
982: \end{figure}
983:
984: \begin{figure}
985: \includegraphics[angle=90,scale=0.3]{f8.ps}
986: \caption{Cumulative distribution function of $V \sin i$ for the Be stars (\textit{dashed line}) and the normal B-type stars (\textit{dotted line}). Theoretical distributions of rapidly rotating stars with random orientation and $V = 0.7 \; V_{\rm crit}$ and $0.8 \; V_{\rm crit}$ are also shown (\textit{solid lines}; assumes the mean value for the Be stars, $V_{\rm crit} = 386$~km~s$^{-1}$).
987: \label{vsini}
988: }
989: \end{figure}
990:
991: \input{tab1.tex}
992: \input{tab2.tex}
993: \input{tab3.tex}
994: \input{tab4.tex}
995:
996: \end{document}
997: