1: \documentclass[12pt,preprint]{aastex}
2:
3: \shorttitle{Hyades Mass-[Fe/H] Relation}
4: \shortauthors{Aaron Dotter and Brian Chaboyer}
5:
6: \newcommand{\Ms}{\mathrm{M_{\odot}}}
7: \newcommand{\Me}{\mathrm{M_{\oplus}}}
8:
9: \begin{document}
10:
11: \title{Stellar Pollution and [Fe/H] in the Hyades}
12: \author{Aaron Dotter and Brian Chaboyer}
13: \affil{Department of Physics and Astronomy, Dartmouth College, 6127 Wilder Laboratory,
14: Hanover, NH 03755}
15: \email{Brian.Chaboyer@Dartmouth.edu}
16:
17: \begin{abstract}
18: The Hyades open cluster presents a unique laboratory for planet
19: formation and stellar pollution studies because all of the stars have
20: essentially the same age and were born from the same cloud of gas.
21: Furthermore, with an age of $\sim$650 Myr most of the intermediate and
22: low mass stars are on the main sequence. Given these assumptions, the
23: accretion of metal rich material onto the surface of a star during and
24: shortly after the formation of planetary systems should be evident via
25: the enhanced metallicity of the star. Building on previous work,
26: stellar evolution models which include the effects of stellar
27: pollution are applied to the Hyades. The results of several Monte
28: Carlo simulations, in which the amount of accreted material is drawn
29: at random from a Gaussian distribution with standard deviation equal
30: to half the mean, are presented. An effective temperature--[Fe/H]
31: relation is produced and compared to recent observations. The
32: theoretical predictions presented in this letter will be useful in
33: future searches for evidence of stellar pollution due to planet
34: formation. It is concluded that stellar pollution effects at the mean
35: level of $\geq$2 $\Me$ of iron are ruled out by the current
36: observational data \citep{psc}.
37: \end{abstract}
38:
39: \keywords{open clusters and associations: individual(Hyades) ---
40: planetary systems: protoplanetary disks ---
41: stars: abundances ---
42: stars: evolution }
43:
44: \section{Introduction}
45:
46: The concept of stellar pollution, that as the protoplantery disk
47: around a star forms planets and settles into a stable configuration
48: the star may accrete some of the metal rich material from the disk,
49: has potentially observable implications for the parent star. The
50: magnitude of the observable implications can vary widely, being
51: influenced by the planetary formation process, which depends on the
52: composition and dynamics of the cloud from which the star is born, as
53: well as the depth of the surface mixed layer of the star, which in
54: turn depends on the mass of the star. \citet{mea} compared a sample of
55: $\sim$500 main sequence stars in the Solar neighborhood to stellar
56: evolution models. Their analysis showed that the observations are
57: consistent with the stars having accreted 0.5 $\Me$ of iron on
58: average. The sample consisted of field stars and thus the signal for
59: pollution had to be disentangled from variations in age and bulk
60: metallicity.
61:
62: Based on the proposal by \citet{quillen} that scatter about the main
63: sequence in the Hyades color-magnitude diagram (CMD) of \citet{deB}
64: can be used to limit the level of pollution, \citet[hereafter DC]{dc}
65: applied stellar evolution models which incorporate the effects of
66: stellar pollution to the Hyades. The authors found that pollution on
67: the level of $\sim$1.5 $\Me$ could be ruled out by scatter in the
68: CMD. A more direct test of pollution in the Hyades, presented in this
69: letter, involves comparing polluted stellar evolution models to
70: observations of [Fe/H] in Hyades stars. Such a test is possible with
71: the high precision [Fe/H] data set presented by \citet[hereafter
72: PSC]{psc}. At present, at least three other groups are known to be at
73: work on similar observations in the Hyades \citep{boe,ful,pea}. These
74: observations may have the precision and wide effective temperature
75: range to make a definitive statement about stellar pollution in the
76: Hyades. The work of PSC is, within the quoted uncertainties,
77: consistent with zero dispersion about the mean [Fe/H] of the
78: sample. However, as shown below, it is also in good agreement with the
79: predictions drawn from polluted stellar evolution models.
80:
81:
82: \section{Monte Carlo Simulations}
83:
84: Four distinct Monte Carlo simulations were performed using Chaboyer's
85: stellar evolution program with modifications to deal with stellar
86: pollution. Each simulation uses the same set of 77 stellar models
87: described by DC. These stars lie in the effective temperature range of
88: 3500 K to 8000 K which corresponds to masses in the range of roughly
89: 0.5 to 2 $\Ms$. Each Monte Carlo simulation consists of 1000
90: individual runs through the set of polluted stellar models. For each
91: stellar model in a given run, the amount of polluting material is
92: drawn at random from a Gaussian distribution with standard deviation
93: equal to half the mean. \citet{mea} find that, on average, field stars
94: in the Solar neighborhood have accreted $\sim$0.5 $\Me$ of iron. Three
95: of the four simulations have means of 0.5, 1, and 2 $\Me$ of iron. The
96: fourth simulation begins with a mean of 0.5 $\Me$ but adds in a 5\%
97: probability that a star will have giant planet in a tight orbit
98: leading to an average accretion of 5 $\Me$ of iron \citep{mc}. In
99: order that the simulations have the mean [Fe/H] = 0.13 (within a few
100: hundredths of a dex) in agreement with observations \citep[PSC]{bf} the
101: models have the same initial bulk metallicity, Z=0.024. The only
102: exception is the 2 $\Me$ case for which all models have a lower
103: initial bulk metallicity (Z=0.020) in order to keep the mean [Fe/H]
104: reasonably close to the observed value.
105:
106: The main goal of the simulations is to determine how [Fe/H] varies
107: with stellar mass and pollution. To obtain the mass--[Fe/H] relation
108: from one simulation, the mass--[Fe/H] relation is created for each of
109: the individual runs. A smooth curve is fit to each individual
110: relation using the LOWESS (Locally Weighted Regression) technique
111: \citep{cleve}. Finally, all of the 1000 individual mass--[Fe/H] curves
112: are averaged. The process is analogous to the CMD analysis of DC. Color
113: tables from \citet{kur} are used for the color transformations.
114:
115: To compute overall [Fe/H] statistics for each of the four simulations,
116: the mean [Fe/H] value is computed from the 77 stellar models in each
117: individual run. The overall mean is the average of all 1000 individual
118: means. The [Fe/H] statistics from each simulation are also divided
119: into four mass bins. Within each bin the mean and standard deviation
120: about the mean are determined from all of the stellar models which
121: have masses in the given range.
122:
123:
124: \section{Results}
125:
126: \begin{deluxetable}{ccccccc}
127: \tabletypesize{\footnotesize}
128: %\rotate
129: \tablecolumns{7}
130: \tablewidth{0pc}
131: \tablecaption{Monte Carlo Results\label{tab1}}
132: \tablehead{
133: \colhead{}&\colhead{}&\colhead{}&\multicolumn{4}{c}{$\Delta$[Fe/H]}\\
134:
135: \cline{4-7}\\
136: \colhead{}&\colhead{}&\colhead{}
137: &\colhead{$M_*<$0.6 $\Ms$}&\colhead{0.6--1.0 $\Ms$}
138: &\colhead{1.0--1.4 $\Ms$}&\colhead{1.4--1.8 $\Ms$}\\
139:
140: \colhead{}&\colhead{}&\colhead{}
141: &\colhead{$\mathrm{T_{eff}}<$4000 K}&\colhead{4000--5400 K}&\colhead{5400--6500 K}&\colhead{6500--8000 K}\\
142:
143: \colhead{No.}&\colhead{Accreted Mass ($\Me$ Iron)}&\colhead{Initial Z}
144: &\colhead{B--V$>$1.35}
145: &\colhead{0.74--1.35}&\colhead{0.43--0.74}&\colhead{0.23--0.43}
146: }
147:
148: \startdata
149:
150: 1&0.50$\pm$0.25&0.024&0.000$\pm$0.002&0.001$\pm$0.004&0.017$\pm$0.019&0.086$\pm$0.079\\
151: 2&\#1 (95\%), 5.00$\pm$2.50 (5\%)&0.024&0.000$\pm$0.012&0.002$\pm$0.018&0.026$\pm$0.060&0.107$\pm$0.152\\
152: 3&1.00$\pm$0.50&0.024&0.000$\pm$0.005&0.004$\pm$0.008&0.040$\pm$0.036&0.157$\pm$0.137\\
153: 4&2.00$\pm$1.00&0.020&0.000$\pm$0.012&0.012$\pm$0.022&0.095$\pm$0.079&0.312$\pm$0.248
154:
155: \enddata
156:
157: \end{deluxetable}
158:
159:
160: The primary results of this paper are presented in Table \ref{tab1},
161: which presents theoretical predictions for the mean [Fe/H] and its
162: dispersion as a function of stellar mass. These predictions are
163: presented as relative abundances rather than absolute because relative
164: abundances are immediately obtained from observational data whereas
165: absolute abundances introduce another source of uncertainty, namely
166: the Solar abundances. For this reason, among others, one study may
167: have a different mean value of [Fe/H] than another. For the purposes
168: of testing for the existence of stellar pollution, it is the behavior
169: of the relative abundances that matters most. At each location in the
170: $\Delta$[Fe/H] portion of Table \ref{tab1} the first value is the mean
171: [Fe/H] value for that range of models relative to the mean for the low
172: mass models. The second value (following the ``$\pm$'') is the
173: standard deviation about the mean as described at the end of $\S$2. In
174: a given row, the mean value represents the difference between one mass
175: range and another, and the standard deviation of the mean gives an
176: idea of the dispersion within one mass range. Overall absolute
177: abundances for the four simulations listed in Table \ref{tab1} are:
178: $<$[Fe/H]$>$=0.131$\pm$0.052 (1), 0.142$\pm$0.085 (2), 0.161$\pm$0.092
179: (3), and 0.155$\pm$0.176 (4).
180:
181: These predictions reflect the differences in [Fe/H] exhibited by the
182: four cases outlined in the previous section and differences within
183: each case as a function of stellar mass. It is evident that assuming
184: 5\% of the Hyades stars have giant planets in tight orbits (case 2)
185: increases $<$[Fe/H]$>$ by only 0.011 (8\% relative change) but
186: increases the standard deviation about the mean by 0.033 (63\%) over
187: the entire mass range covered in the simulations. Thus a small
188: probability of having a giant planet in a tight orbit will have
189: relatively little influence on the average metallicity of a stellar
190: population but will have a significant influence on the scatter. The
191: [Fe/H] data is broken into mass bins in order simplify comparisons
192: with observational data. The Monte Carlo simulations indicate that, if
193: stellar pollution is the main factor in metallicity variations, then
194: stars above roughly 6500 K (1.4 $\Ms$) will have a significantly
195: higher mean [Fe/H] and dispersion than stars below 6500 K. The lower
196: the initial bulk metallicity the more pronounced this effect will be.
197:
198: \begin{figure}
199: \plotone{f1.eps}
200: \caption{The mass/$\mathrm{T}_{\mathrm{eff}}$--[Fe/H] relation for the
201: Hyades. The solid line is the 0.5 $\Me$ case (1 in Table
202: \ref{tab1}), the dashed line is the 0.5 $\Me$ with giant planets case
203: (2), the dash-dotted line is the 1.0 $\Me$ case (3), and the
204: dotted line is the 2.0 $\Me$ case (4).
205: \label{mfeh}}
206: \end{figure}
207:
208: The mean mass--[Fe/H] relation is displayed for all four simulations
209: in Figure \ref{mfeh}. The general trend is that [Fe/H] increases with
210: stellar mass which is consistent with the decreasing mass of the
211: surface mixed layer. The decrease at around 1.4 $\Ms$ is due to an
212: artificially increased surface mixed layer, for details see
213: \citet{mea}.
214:
215:
216: \section{Comparison with Observational Data}
217:
218: The hotter temperatures and higher rotational velocities of A and F
219: stars make the task of measuring metal abundances exceedingly
220: difficult. As such, a detailed comparison of observations of A and F
221: stars to the Monte Carlo simulations is not currently
222: possible. However, comments are made regarding two studies of F and A
223: stars in the Hyades. \citet{bf} list effective temperatures and [Fe/H]
224: values for 14 F stars in the Hyades. The uncertainties in [Fe/H] are
225: all of order 0.10 dex and the authors find no [Fe/H] trend with
226: temperature. Nevertheless, a simple comparison suggests that, within
227: the uncertainties, the polluted models are in agreement with their
228: data.
229:
230: \citet{vm} present a study of 29 F and 19 A stars in the Hyades. The
231: authors provide uncertainties in their [Fe/H] values for only about 10
232: of the 48 stars, thus an analysis of the dispersion is not possible.
233: Furthermore, the data varies widely over a range of $\sim$0.6 dex in
234: [Fe/H] so simple averages in several temperature ranges like those in
235: Table \ref{tab1} would favor the outlying points. To avoid giving
236: equal weight to the outliers in the average values, the averages are
237: taken from the LOWESS fit to the data. (LOWESS weights each point
238: based on how close it is to the local average.) In keeping with the
239: type of analysis performed by DC and in the following paragraphs, the
240: absolute values of the relative [Fe/H] values (relative meaning the
241: overall average has been subtracted off) are used. The absolute value
242: of the $\Delta$[Fe/H] data of \citet{vm} has a LOWESS-average value of
243: 0.05 dex between 6000 and 6500 K, and 0.12 dex between 6500 and 8000
244: K. Comparing these data to Table \ref{tab1}, LOWESS-average values in
245: between 6000 and 6500 K are 0.04 dex (1), 0.06 dex (2), and 0.10 dex
246: (3). Between 6500 and 8000 K the simulations have LOWESS-average
247: values of 0.06 dex (1), 0.07 dex (2), and 0.17 dex (3). At the 90\%
248: confidence level, the \citet{vm} data is consistent with stellar
249: pollution at and below the 1 $\Me$ level but is inconsistent with
250: pollution at the 2 $\Me$ level.
251:
252: PSC perform spectroscopic abundance analyses of 55 Hyades stars which
253: range in effective temperature from about 5000 to 6500 K. These
254: authors list relative [Fe/H] data and perform detailed error analysis
255: making it possible to directly compare the data to the predictions of
256: the polluted stellar models. The star chosen as the standard about
257: which the relative abundances are determined (HD 35768) has an [Fe/H]
258: value somewhat below the cluster average so that the $\Delta$[Fe/H]
259: values average to $\sim$0.03 dex. For the purposes of this letter,
260: 0.03 dex has been subtracted from each $\Delta$[Fe/H] value so that
261: $<$$\Delta$[Fe/H]$>$=0. The modified $\Delta$[Fe/H] data of PSC can
262: be compared to the Monte Carlo simulations by constructing
263: color--[Fe/H] curves as described above. The absolute value of the
264: $\Delta$[Fe/H] data is fit by the LOWESS technique for both the
265: observational data and the four simulations. Relative abundances for
266: the simulations are obtained by subtracting from each data point the
267: average [Fe/H] of the run in the same temperature range as the
268: observational data which corresponds to stellar masses between 0.8 and
269: 1.25 $\Ms$. PSC report an average error about the mean of 0.04 dex in
270: $\Delta$[Fe/H].
271:
272: \begin{figure}
273: \plotone{f2.eps}
274: \caption{
275: The absolute values of the $\Delta$[Fe/H] data of PSC and the LOWESS
276: fit. Three outlying points have been omitted from the plot for scaling
277: purposes. See the text for an explanation.
278: \label{psc1}}
279: \end{figure}
280:
281: \begin{figure}
282: \plotone{f3.eps}
283: \caption{Trends in $|\Delta$[Fe/H]$|$ from the Monte Carlo simulations
284: and the PSC data. Lines in each plot represent: the LOWESS fit to the
285: PSC data (dashed line), the Monte Carlo mean (solid line), and Monte
286: Carlo 90\% confidence levels (dotted lines). The number in the upper
287: right of each panel refers to Table \ref{tab1}.
288: \label{psc2}}
289: \end{figure}
290:
291: Figure \ref{psc1} shows the absolute values of the modified
292: $\Delta$[Fe/H] data of PSC along with the LOWESS fit. The locations of
293: three stars have been omitted. The stars are HD 20430, HD 20439, and
294: HD 14127 with (modified) $|\Delta$[Fe/H]$|$ = 0.18, 0.17, and 0.23,
295: respectively. That these stars are indeed members of the Hyades is
296: questionable, see the discussion in PSC. In any case, because of
297: their large deviations these three stars play an insignificant role in
298: the LOWESS fit. It must be noted here that to within the quoted
299: average uncertainty of 0.04 dex the LOWESS fit is consistent with zero
300: scatter in [Fe/H].
301:
302: A comparison of the results of the four Monte Carlo simulations to the
303: observational data is presented in Figure \ref{psc2}. In order to
304: directly compare the Monte Carlo simulations with the observational
305: data an additional step was added. The simulations represent intrinsic
306: values untouched by uncertainties involved in observations. To
307: account for the observational uncertainties, the [Fe/H] value of each
308: star was drawn from a Gaussian distribution with mean given by the
309: simulation and a 0.04 dex standard deviation (in accordance with the
310: average error bar of PSC). The individual graphs are numbered in the
311: upper right hand corners, the numbers refer to Table
312: \ref{tab1}. Stellar pollution at the 0.5 $\Me$ level cannot be ruled
313: out by the observational data, with (2) or without (1) the possibility
314: of having a giant planet in a tight orbit. Likewise with the 1 $\Me$
315: case (3). The 2 $\Me$ case (4), however, is not consistent with the
316: data at the 90\% confidence level. Thus the PSC data set rules out
317: pollution at the 2 $\Me$ level. The divergence of the 90\% confidence
318: levels at the high and low end of each panel is due to the exclusion
319: of the data points outside the limits considered. This behavior is
320: spurious and should not be considered part of the prediction, however,
321: for the purpose of comparing the simulations to the observational data
322: it is necessary to leave it as is. The B--V color is used as the
323: abscissa in the comparison presented here but effective temperature
324: could also have been used, the basic trends and conclusions are the
325: same.
326:
327: The data of PSC cover stars between roughly 0.8 and 1.25 $\Ms$ but, as
328: Figure \ref{mfeh} and Table \ref{tab1} indicate, these stars do not
329: exhibit the greatest dispersion due to pollution effects. A careful
330: analysis of hotter stars will be necessary to determine the level of
331: stellar pollution in the Hyades.
332:
333:
334: \section{Conclusions}
335: The formation of planets around a star may lead to the accreation of
336: metal rich material by the star. Due to variations in the planet
337: formation process and variations of the mass of the surface mixed
338: layer with stellar mass, the effect of such stellar pollution will
339: vary from star to star. In this paper, the effects of stellar
340: pollution on the Hyades is studied. The main result of this study is
341: presented in Table \ref{tab1} where the predicted mean [Fe/H] and
342: dispersion is given as a function of stellar mass. The predictions
343: have been binned by stellar mass in order to be more easily comparable
344: to observational data sets. As stellar mass increases, increases in
345: both the mean [Fe/H] and the dispersion caused by stellar pollution
346: should be evident. A small probability that a star will have a giant
347: planet in a tight orbit has little impact on the average [Fe/H] but
348: does have a noticeable effect on the dispersion. An observational
349: program that can provide conclusive evidence for or against stellar
350: pollution in the Hyades needs precise abundances over a wide range of
351: spectral types. A significant increase in the mean [Fe/H] should arise
352: as the effective temperature increases from below to above 6500 K. In
353: addition, the dispersion is larger at higher temperatures. It will be
354: important to have relative abundances for both cool and hot stars so
355: that a trend can be drawn over a broad range of effective
356: temperatures. The present data sets rule out pollution at the $\ga 2
357: \Me$ level but are consistent with pollution at and below the 1 $\Me$
358: level.
359:
360:
361: \acknowledgments
362:
363: \begin{thebibliography}{}
364: \bibitem[Boesgaard et al.(2002)]{boe} Boesgaard, A. M., Beard, J. L., King, J. R. 2002, AAS, 201, 44.01
365: \bibitem[Boesgaard \& Friel(1990)]{bf} Boesgaard, A. M., \& Friel, E. D. 1990, \apj, 351, 467
366: \bibitem[Buser \& Kurucz(1992)]{kur} Buser, R. \& Kurucz, R. L. 1992, \aap, 264, 557
367: \bibitem[Cleveland(1979)]{cleve} Cleveland, W. S. 1979, J. Am. Stat. Assoc., 74, 829
368: \bibitem[de Bruijne et al.(2001)]{deB} de Bruijne, J. H. J., Hoogerwerf, R., \& de Zeeuw, P. T. 2001, \aap, 367, 111
369: \bibitem[Dotter \& Chaboyer(2003)]{dc} Dotter, A. \& Chaboyer, B. 2003, \apj, in press
370: \bibitem[Fulbright (2002)]{ful} Fulbright, J. P. 2002, AAS, 201, 17.09
371: \bibitem[Murray \& Chaboyer(2002)]{mc} Murray, N., \& Chaboyer, B. 2002, \apj, 493, 222
372: \bibitem[Murray et al.(2001)]{mea} Murray, N., Chaboyer, B., Arras, P., Hansen, B., \& Noyes, R. W. 2001, \apj, 555, 801
373: \bibitem[Paulson et al.(2003)]{psc} Paulson, D. B., Sneden, C., \& Cochran, W. D. 2003, \aj, 125, 3185
374: \bibitem[Primas, et al.(2003)]{pea} Primas, F., Chaboyer, B. Feltzing, S., Ryan, S. 2003, ESO VLT program 70.D-0356(A)
375: \bibitem[Quillen(2002)]{quillen} Quillen, A. C. 2002, \aj, 124, 400
376: \bibitem[Varenne \& Monier(1999)]{vm} Varenne, O. \& Monier, R. 1999, \aap, 351, 247
377: \end{thebibliography}
378:
379: \end{document}
380: