1: %\documentclass[12pt,preprint]{aastex}\newcommand{\escl}[1]{\epsscale{#1}}\newcommand{\epsstr}[1][1]{}
2: %\documentclass[12pt,manuscript]{aastex}\newcommand{\escl}[1]{\epsscale{#1}}\newcommand{\epsstr}[1][1]{}
3: \documentclass[apj]{emulateapj}\usepackage{apjfonts}\newcommand{\escl}[1]{}\newcommand{\epsstr}[1][1.15]{\epsscale{#1}}
4: \usepackage{natbib, graphicx}
5: \citestyle{aa}
6: \hyphenation{ana-ly-ti-cal}
7: \bibpunct[, ]{(}{)}{;}{a}{}{,}
8: %
9: %% Some additional LaTeX command definitions
10: \newcommand{\mmbox}[1]{\mbox{$#1$}}
11: \newcommand{\dif}{\mathrm{d}}
12: \newcommand{\diff}[2]{\frac{\dif#1}{\dif#2}}
13: \newcommand{\pma}[2]{{}^{+#1}_{-#2}}
14: \newcommand{\xil}{\xi_\mathrm{L}}
15: \newcommand{\nxi}{\hat{\xi}}
16: \newcommand{\xib}{\xi_\mathrm{bin}}
17: \newcommand{\xis}{\xi_\mathrm{sys}}
18: \newcommand{\pct}{\%}
19: \newcommand{\unit}[1]{\mbox{$\,\mathrm{#1}$}}
20: \newcommand{\abd}{\alpha_\mathrm{BD}}
21: \newcommand{\rbod}{\mathcal{R}}
22: \newcommand{\rpop}{\mathcal{R}_\mathrm{pop}}
23: \newcommand{\rkbd}{\mathcal{R}_\mathrm{HBL}}
24: \newcommand{\nbod}{N_\mathrm{bod}}
25: \newcommand{\nsng}{N_\mathrm{sng}}
26: \newcommand{\nsys}{N_\mathrm{sys}}
27: \newcommand{\nsbd}{N_\mathrm{sys,BD}}
28: \newcommand{\nsmd}{N_\mathrm{sys,MD}}
29: \newcommand{\nsst}{N_\mathrm{sys,star}}
30: \newcommand{\nbny}{N_\mathrm{bin, tot}}
31: \newcommand{\nbbd}{N_\mathrm{BD-BD}}
32: \newcommand{\nbmd}{N_\mathrm{MD-MD}}
33: \newcommand{\nbst}{N_\mathrm{star-star}}
34: \newcommand{\nbsb}{N_\mathrm{star-BD}}
35: \newcommand{\mmin}{m_\mathrm{min}}
36: \newcommand{\mmax}{m_\mathrm{max}}
37: \newcommand{\msys}{m_\mathrm{sys}}
38: \newcommand{\mbin}{m_\mathrm{bin}}
39: \newcommand{\mpri}{m_\mathrm{prim}}
40: \newcommand{\ma}{m_\mathrm{A}}
41: \newcommand{\mb}{m_\mathrm{B}}
42: \newcommand{\msun}{M_{\sun}}
43: \newcommand{\mh}{m_\mathrm{H}}
44: \newcommand{\mlobd}{m_\mathrm{0,BD}}
45: \newcommand{\mlost}{m_\mathrm{0,star}}
46: \newcommand{\mhibd}{m_\mathrm{max,BD}}
47: \newcommand{\ftot}{f_\mathrm{tot}}
48: \newcommand{\fbd}{f_\mathrm{BD-BD}}
49: \newcommand{\fst}{f_\mathrm{star-star}}
50: \newcommand{\fmd}{f_\mathrm{MD-MD}}
51: \newcommand{\fmix}{f_\mathrm{star-BD}}
52: \newcommand{\tmsun}{\mbox{$\msun$}}
53: \newcommand{\simf}{\mbox{$\mbox{IMF}_\mathrm{sys}$}}
54: \newcommand{\bmf}{\mbox{$\mbox{IMF}_\mathrm{bin}$}}
55:
56: \journalinfo{The Astrophysical Journal}
57: \submitted{Accepted 2007 August 10}
58: %%%%
59: %%%%
60: \begin{document}
61: \title{A Discontinuity in the Low-Mass Initial Mass Function}
62: \shorttitle{A discontinuity in the low-mass IMF}
63: \author{Ingo Thies\altaffilmark{1} \& Pavel Kroupa\altaffilmark{1}}
64: \shortauthors{Thies \& Kroupa}
65: \altaffiltext{1}{Argelander-Institut f\"ur Astronomie (Sternwarte), Universit\"at Bonn, Auf dem H\"ugel 71, D-53121 Bonn, Germany}
66: \begin{abstract}
67: The origin of brown dwarfs (BDs) is still an unsolved mystery.
68: While the standard model describes the formation of BDs and stars in a
69: similar way recent data on the multiplicity properties of stars and
70: BDs show them to have different binary distribution functions.
71: Here we show that proper treatment of these uncovers a discontinuity
72: of the multiplicity-corrected mass distribution in the very-low-mass star
73: (VLMS) and BD mass regime.
74: A continuous IMF can be discarded with extremely high confidence.
75: This suggests that VLMSs and BDs on the one hand, and stars on the other,
76: are two correlated but disjoint populations with different dynamical
77: histories.
78: The analysis presented here suggests that about one BD forms per five stars
79: and that the BD-star binary fraction is about 2\pct--3\pct\ among stellar
80: systems.
81: \end{abstract}
82: \keywords{%
83: binaries: general ---
84: open clusters and associations: general ---
85: stars: low-mass, brown dwarfs ---
86: stars: luminosity function, mass function
87: }
88: \maketitle
89: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
90: \section{Introduction}
91: \label{sec:intro}
92: Traditionally, brown dwarfs (BDs) are defined as (sub)stellar bodies with
93: masses below the hydrogen burning mass limit (HBL),
94: $\mh=0.075\,\msun$ for solar composition, and consequently they cool
95: indefinitely after formation \citep{Buetal93,ChaBa00}.
96: Several attempts have been made to
97: explain the formation of BDs by the same mechanisms as for
98: stars, i.e. via fragmentation of a gas cloud and subsequent
99: accretion \citep{AdFa96,PaNo02,PaNo04}.
100: If a gas cloud in a star-forming region fragments there will be a
101: certain number of gas clumps with masses below the HBL.
102: Unless the mass of the fragment is below the local Jeans mass
103: it will contract in essentially the same way as higher mass clumps and
104: finally produce a single or multiple BD.
105: This scenario predicts similar multiplicities and also a substellar
106: initial mass function (IMF) as a continuous extension of the stellar one
107: (e.g. the standard model with BDs in \citealt{Ketal03}).
108:
109: However, recent observations have shown that there
110: is a lack of BD companions to low-mass stars
111: \citep*{2003IAUS..211..279M}. \citet{GreLin06} found a star-BD binary
112: fraction among solar-type primaries of less than 1\pct\ for close
113: companions,
114: the \emph{brown dwarf desert}. This implies two populations of stellar
115: and substellar objects, and that binaries are formed in
116: each population separately (except for pairing due to post-formation
117: dynamical exchanges). Observations e.g. by \citet{Reidetal06}
118: also show that most BD binaries have a primary-to-companion mass ratio
119: of $q>0.8$, in contrast to the mass ratio distribution of stellar binaries
120: which has typically $q<0.4$ \citep{DuqMay91}.
121:
122: There are indications, e.g. \citet{MetHil05}, that the
123: BD desert may not be as dry for larger separations ($>30$~AU) as it is
124: for smaller ones. Since new surveys using adaptive optics or new instruments
125: like the upcoming James Webb Space Telescope might reveal more substellar
126: companions to stars, the fraction of star-BD systems may increase.
127:
128: Apart from the BD desert there are more hints for a separate population.
129: For example, BDs and VLMSs have a relatively low binary fraction
130: of about 15\pct\
131: (\citealt{Bouyetal03,Cloetal03,Maetal03}; \citealt*{Krausetal06};
132: \citealt{LaHoMa07}).
133: By comparison, the stellar binary fraction
134: is close to 100\pct\ for the very young Taurus-Auriga association (TA, about
135: 1~Myr; \citealt{Duchene99,Luetal03a}) and about 40\pct--50\pct\ for other clusters
136: and field stars \citep{Lada2006}.
137: The BD and VLMS binary fraction can be increased to a starlike binary
138: fraction if there are a large fraction of $\la5$~AU binaries,
139: e.g. as deduced by \citet{JefMax05}. But such a semi-major axis distribution
140: would again imply a discontinuity of its form between low-mass stars and
141: VLMSs/BDs and is not supported by the radial-velocity survey of
142: \citet{Joergens2006a}.
143: We therefore do not consider the starlike formation as a major mechanism
144: for BDs.
145:
146: It has also been argued that
147: the low binary fraction of BDs can be understood as a continuous extension
148: of a trend that can already be recognized from G dwarfs to M dwarfs
149: \citep{Lu04b,SteDur03}.
150: Therefore, the binary fraction alone cannot be taken as a strong evidence
151: to introduce a separate population.
152:
153: The most striking evidence for two separate populations is the empirical fact
154: that the distribution of the separations and therefore the binding energies
155: of BD binaries differs significantly from that in the stellar regime
156: \citep{Bouyetal03,Burgetal03,Maetal03,Cloetal03}.
157: This is shown in Figures \ref{sepdist} and \ref{edist}.
158: The solid line is a Gaussian fit to the central
159: peak of the histogram while the dash-dotted one refers to
160: (\citealt{BasRei2006}, a compressed \citealt{FisMar92} fit).
161: BDs and very low-mass stars (VLMS) have a semi-major axis distribution limited
162: to \mmbox{\lesssim15}~AU, whereas M, K, and G dwarfs have a very
163: broad and similar distribution (\emph{long-dashed and short-dashed curves};
164: \citealt{Cloetal03}; \citealt*{LaHoMa07}; \citealt{PPVGoodwinetal}).
165: There is also a dearth of BDs below 1~AU. Recent findings, e.g. by
166: \citet{GuWu03} and \citet{Kenetal05}, suggest a low number of such very close
167: BD/VLMSs binaries. The semi-major axis distribution
168: of BDs/VLMSs binaries based on the data from \citet{Cloetal03} can be
169: modelled with a log~$a$ Gaussian centered at 4.6~AU ($\log a =0.66$) with
170: a half-peak width of $\sigma=0.4$. It corresponds to an overall BD/VLMS
171: binary fraction of $\fbd=0.15$. If data from \citet{Lu04a},
172: \citet{Joergens2006a}, and \citet{Konetal07} are taken as hints to
173: incomplete data between about 0.02 and 1~AU,
174: the compressed \citet{FisMar92} Gaussian from Figure 4 in
175: \citet{BasRei2006} may provide an appropriate envelope. However, for an assumed
176: BD mass of 0.07~\tmsun\ and $\fbd=0.26$ their period distribution corresponds
177: to a semi-major axis distribution with $\sigma\approx0.85$ and is therefore
178: still inconsistent with that of M and G dwarfs.
179:
180: Although \citet{Konetal07} have recently found five VLMS binaries in TA
181: with four of them having separations much larger than 15~AU,
182: the sudden change of the orbital properties
183: remains. In particular, they found two binaries with projected
184: separations slightly above 30~AU and two others with separations
185: between 80 and 90~AU. Possible implications of these discoveries
186: are discussed in \S~\ref{ssec:diskfrag}.
187: However, the truncation near 15~AU cannot be derived from the stellar
188: distribution through downsizing according to Newton's laws
189: \citep{Cloetal03,Bouyetal03}.
190: Not even dynamical encounters in dense stellar environments can invoke such
191: a truncation near 15~AU \citep{Burgetal03}.
192: Using $N$-body simulations \citet{Ketal03} tested the hypothesis that BDs
193: and stars form alike, and mix in pairs, and found that, despite of close
194: dynamical encounters,
195: the distribution of the semimajor axes of BD binaries remains starlike.
196: That is, dynamical encounters even in dense clusters cannot truncate
197: the BD binary distribution near 15~AU.
198: They further found that star-BD binaries would be much more frequent
199: than actually observed. Thus, this classical hypothesis is rejected with
200: high confidence.
201:
202: With this contribution we study the implications of the observed change
203: of binary properties on the underlying single-object initial mass function
204: (IMF). For this purpose, we analyzed the observational mass functions of TA
205: (Luhman et al. 2003a, 2004b), the IC~348 cluster
206: \citep{Luetal03b},
207: the Trapezium cluster \citep{Muetal02}, and the Pleiades cluster based on
208: data by \citet{Doetal02}, \cite{Moetal03}, and the Prosser and Stauffer
209: Open Cluster
210: Database.\footnote{Available at
211: http://www.cfa.harvard.edu/\symbol{126}stauffer/opencl/}
212: For all the systems we analyst, we refer to the MF as the IMF, although
213: strictly speaking this is not correct for the 130~Myr old Pleiades.
214: In \S~\ref{sec:IMF} we shortly review the definition of the IMF
215: and the role of multiplicity on the shape of the
216: individual body IMF compared to the observed system IMF (\simf).
217: We show that the IMF that can be derived from an observed
218: \simf\ does not need to be continuous even if the \simf\ does not show any
219: discontinuity.
220: In \S~\ref{sec:method} we describe the statistical method how to
221: fit a model \simf\ by combining BD and star IMFs
222: as an approximation to the observed \simf.
223: \S~\ref{sec:results} presents the results that indicate a
224: discontinuity close to the HBL. Also the BD to star ratio from the
225: model is calculated there.
226: In \S~\ref{sec:discussion} the results are discussed in the context
227: of four alternative BD formation scenarios, i.e. embryo ejection,
228: disk fragmentation, photoevaporation, and ejection by close stellar
229: encounters.
230:
231: \begin{figure}
232: \begin{center}
233: %\escl{1.0}
234: \epsstr
235: \plotone{f1.eps}
236: %\plotone{../../../../Figures/sepdist.eps}
237: \caption{\label{sepdist}The semi-major axis ($a$) distribution for
238: VLM binaries (VLMBs) with a total mass $\mbin<0.2\,\msun$
239: based on data from Nick Siegler's
240: \emph{Very Low Mass Binaries Archive}
241: (VLMBA, http://paperclip.as.arizona.edu/\symbol{126}nsiegler/VLM\_binaries/,
242: version from June 1, 2007; histogram).
243: The data can be fitted with a $\log$-normal distribution.
244: For the VLMBs (solid curve) the
245: Gaussian parameters are $x_\mathrm{peak}=0.66$ (corresponding to
246: $a=4.6\unit{AU}$) and $\sigma=0.4$ with a
247: normalization factor (= total binary fraction) $\ftot=0.15$,
248: the binary fraction among BDs and VLMSs. For the M0--M4 dwarf
249: binaries by \citet{Cloetal03} $x_\mathrm{peak}=1.6$ and $\sigma=1.26$
250: ($\ftot=0.3$) and for the
251: G dwarf binaries $x_\mathrm{peak}=1.53$ and $\sigma=1.53$ ($\ftot=0.5$;
252: \citealt{FisMar92}, long-dashed and short-dashed curves, respectively).
253: For VLMBs the gap within $-1.5\le\log a/AU\le0$
254: can be interpolated with the compressed Gaussian of \citet{BasRei2006}
255: with $\sigma=0.85$ (thin dot-dashed curve),
256: also enveloping wider BD/VLMS binaries found by \citet{Konetal07}.
257: It corresponds to a total BD/VLMS $\fbd=0.26$.
258: The areas below the histogram and the curves are equal to
259: the corresponding $\ftot$ (eq. \ref{ftot}).}
260: \end{center}
261: \end{figure}
262:
263: \begin{figure}
264: \begin{center}
265: %\escl{1.0}
266: \epsstr
267: \plotone{f2.eps}
268: %\plotone{../../../../Figures/edist.eps}
269: \caption{\label{edist}The distribution of the orbital binding
270: energy $E_\mathrm{bind}=-0.5 G(m_1 m_2/a)$ for VLM binaries and low-mass
271: stellar binaries. Line types and symbols are the same as in Fig.
272: \ref{sepdist}.
273: As for the semi-major axes there is a clearly
274: different distribution for BDs/VLMSs on the one hand side
275: and M, K and G dwarf binaries on the other.}
276: \end{center}
277: \end{figure}
278:
279: \section{The IMF for individual stars and systems}
280: \label{sec:IMF}
281: \subsection{Definition}
282: \label{ssec:IMF:def}
283: The IMF is among the most important properties of
284: a stellar population since it gives hints to the processes that form
285: stars and BDs. Although we can only observe stellar populations at their given
286: age there are data of several very young populations where the mass function
287: is probably still very close to the initial one.
288:
289: In general, the mass distribution of stars and BDs can be approximated by
290: a power law or a combination of several power laws.
291: \citet{Salpeter1955} estimated the relationship between the stellar
292: mass and the relative number of stars of a given mass as a single power law,
293: \begin{equation}\label{powerimf}
294: \xi(m)=\diff{n}{m}=k\,m^{-\alpha}\,,
295: \end{equation}
296: for $0.4\lesssim m/\msun\lesssim 10$ and with $\alpha=2.35$ and a normalization
297: constant $k$.
298: The IMF is often expressed on the logarithmic mass scale and
299: then becomes
300: \begin{equation}\label{powerlog}
301: \xil(\log m)=\diff{n}{\log m}=\left(\ln 10\right)\,m\,\xi(m)=k_\mathrm{L}\,m^{1-\alpha}\,,
302: \end{equation}
303: where $k_\mathrm{L}$ is the corresponding normalization constant for $\xil$.
304: An IMF is said to be \emph{flat} if $\xil(\log m)$ is constant, i.e.
305: $\alpha=1$.
306: More recent work has shown there to be a flattening in the lower-mass regime
307: of the observed mass function and, in the $\xil$ representation,
308: even a turnover near the BD-star transition \citep*{Kr01,RGH02,Cha03}.
309: Here, all objects from BDs to the most massive stars
310: in a cluster are described by a \emph{continuous IMF}, i.e.
311: with a single population containing BDs as well as stars.
312: The universal or ``canonical'' IMF has $\abd\equiv\alpha_0=0.3$
313: ($0.01\le m/\msun\le0.075$), $\alpha_1=1.3$ ($0.075\le m/\msun\le0.5$), and
314: $\alpha_2=2.3$ ($0.5\le m/\msun\le\mmax$), where $\mmax$ is given by the
315: mass of the host cluster \citep{WK06}. The generally accepted wisdom has been
316: that the IMF is continuous from above 0.01~\tmsun\ to $\mmax$
317: (e.g. \citealt{Cha02}).
318:
319: \subsection{Unresolved Binaries}
320: \label{ssec:IMF:bincorr}
321: Star cluster surveys are usually performed with wide-field telescopes with
322: limited resolution that do not resolve most of the binaries. Hence, they yield,
323: as an approximation, system MFs
324: in which unresolved binary systems are counted as one object.
325: The fraction of
326: unresolved multiples can be taken as the total binary fraction of a cluster,
327: \begin{equation}\label{ftot}
328: \ftot=\frac{\nbny}{\nsng+\nbny}\,,
329: \end{equation}
330: where $\nsng$ is the number of singles (or resolved individual bodies)
331: and $\nbny$ the number of (unresolved) binaries.
332: Unresolved binaries increase the number of individual bodies, $\nbod$, in a
333: cluster such that
334: \begin{equation}\label{defnbod}
335: \nbod=\left(1+\ftot\right)\nsys\,,
336: \end{equation}
337: where $\nsys=\nsng+\nbny$ is the total number of systems.
338: Note that a ``system''
339: is either a single body or a binary or a higher order multiple. Here we
340: ignore higher order multiples, because they are rare \citep{GoKr05}.
341: The binary fraction in dependence of the mass of the
342: primary star, $\mpri$, is
343: \begin{equation}\label{fmpri}
344: f(\mpri)=\frac{\nbny(\mpri)}{\nsng(\mpri)+\nbny(\mpri)}\ ,
345: \end{equation}
346: where $\nbny(\mpri)$ and $\nsng(\mpri)$ are, respectively, the number of
347: binaries with primary star mass $\mpri$ and single stars of mass $\mpri$.
348:
349: \subsection{The System Mass Function}
350: \label{ssec:IMF:SIMF}
351: The effect of this \emph{binary error} on the appearance of the IMF can
352: be described as follows.
353: Assume a stellar population with $\ftot<1$ and
354: with stellar masses with a minimum
355: mass $\mmin$ and a maximum mass limit of $\mmax$. The minimum
356: mass of a binary is $2\mmin$, while the mass-dependent binary fraction
357: $f(m)=0$ for $\mmin\le\msys<2\mmin$.
358: A binary closely above $2\mmin$ can only consist of two stars near
359: $\mmin$ making such binaries rare.
360: For higher system masses, where a system can be a multiple or
361: a single star, there are more possible combinations
362: of primary and companion mass, so that the binary fraction
363: increases with the system mass and approaches
364: an upper limit for the most massive objects.
365:
366: Figure \ref{bincorr} shows the general shape of a system IMF for a
367: flat (logarithmic scale) IMF with $\mmin=0.1\,\msun$, $\mmax=1\,\msun$, and
368: $\ftot=0.5$. The \simf\ is flat between $\mmin$ and $2\mmin$
369: ($\log m/\msun\approx-0.7$) and rises above $2\mmin$ to a maximum
370: at $\mmax$.
371: Systems with $\msys<2\mmin$ can only be singles and the \simf\ in this
372: region is just the IMF minus the mass function of objects that are bound
373: to a multiple system.
374: For masses $\msys>\mmax$, on the other hand, only binaries exist,
375: and the \simf\ declines towards zero at $\msys=2\mmax$, the highest
376: mass possible for binaries. The sharp truncation of the
377: IMF at $m=\mmax$ causes the sudden drop at $\msys=\mmax$,
378: while the minor peak at $\msys=\mmax+\mmin$ corresponds to the maximum of
379: the \emph{binary mass function} (\bmf, the IMF of binary system masses).
380: For a binary of this mass the primary and companion mass can be drawn from
381: the whole IMF, and thus the number of possible combinations becomes maximal.
382: It should be noted that natural distributions
383: with smoother boundaries probably do not show such a double peak.
384:
385: Mathematically and in the case of random pairing, which is a reasonable
386: approximation in the stellar regime \citep*{MalZin01,PPVGoodwinetal},
387: the binary mass function
388: is just
389: the integral of the product of the normalized IMF, $\nxi\equiv\xi/\nbod$, of
390: each component times the total number of binaries $\nbny$.
391: Given
392: the masses of the binary components A and B, $\ma$ and $\mb$, the binary
393: mass \mmbox{\mbin=\ma+\mb}. Thus, \mmbox{\mb=\mbin-\ma}. Thus, \bmf\ can
394: now be written as
395: \begin{equation}\label{defbmf}
396: \xib(\mbin)=\nbny\int\limits_{\mmin}^{\mbin-\mmin} \nxi(m)\,\nxi(\mbin-m)\,\dif m\,,
397: \end{equation}
398: where $\mmin$ is the lower mass limit of all individual bodies in the
399: population. The upper limit of the integral, $\mbin-\mmin$, is the maximum
400: mass of the primary component $\ma$ corresponding to a secondary component
401: with $\mb=\mmin$.
402:
403: The other extreme case of assigning the component masses is
404: equal-mass pairing. In that case, equation (\ref{defbmf}) simplifies to the
405: IMF of one of the components and \bmf\ is just the
406: IMF shifted by a factor of 2 in mass and corrected for binarity
407: using equations (\ref{ftot}) and (\ref{defnbod}):
408: \begin{equation}
409: {\xib}_\mathrm{equal}(\mbin=2m)=\frac{\ftot}{1+\ftot}\,\xi(m)\,.
410: \end{equation}
411: This case is more applicable for BD binaries since their mass ratio
412: distribution peaks at a ratio $q=1$ \citep{Reidetal06}.
413: However, due to the low overall binary
414: fraction of BDs the effect of the mass ratio distribution is quite small.
415:
416: The \simf\ is just the sum of \bmf\ and the IMF of the remaining
417: single objects, $(1-\ftot)\xi$:
418: \begin{equation}\label{cimf}
419: \xis(m)=\xib(m)+\left(1-\ftot\right)\,\xi(m)\,.
420: \end{equation}
421:
422: Thus, to obtain the true individual star IMF, $\xi(m)$ has to be extracted
423: from the observed $\xis(m)$ for which a model for $\xib(m)$ is required.
424: That is \simf\ has to be corrected
425: for unresolved binaries in the cluster or population under study.
426: This leads to a significant increase of the numbers of objects at the
427: low-mass end because low-mass objects contribute to both low-mass singles
428: and intermediate-mass binaries and thus more individual objects are
429: required to reproduce the observed \simf\ \citep{KGT91,MalZin01}.
430: In the mass range $\mmin$--$2\mmin$,
431: systems can only be single because the system mass is the sum
432: of the masses of the system members. Increasing the system mass
433: beyond $2\mmin$ causes the binary contribution to
434: rise quickly and then to asymptotically approximate a maximum value.
435: Thus, the fraction of singles among M dwarfs
436: is higher than for G dwarfs \citep{Lada2006},
437: as one would expect for a stellar population ($>0.08\,\msun$), i.e.
438: essentially without BDs.
439:
440: Figure \ref{cartoon} schematically shows the effects of correcting
441: a flat continuous observed \simf\ for unresolved multiples whereby the
442: binary fraction for BDs is smaller than for stars, as is observed to be
443: the case.
444: For a binary fraction
445: of 15\pct\ (i.e. $\fbd=0.15$) among BDs there would thus be
446: \mmbox{N_\mathrm{BD}=(1+\fbd)\nsbd=1.15\nsbd} BDs in total, while
447: for stars with \mmbox{\ftot=0.5} we would have \mmbox{1.5\nsst}.
448: This exemplifies how the change in $f$ leads to a discontinuous \simf.
449:
450: \begin{figure}
451: \begin{center}
452: %\escl{1.0}
453: \epsstr
454: \plotone{f3.eps}
455: %\plotone{../../../../Figures/sketch_bincorr.eps}
456: \caption{\label{bincorr}The system IMF (\simf\, solid line) corresponding
457: to a flat logarithmic IMF (i.e. $\alpha=1$, dashed line) for
458: $\mmin=0.1\le m/\msun\le 1=\mmax$ with an overall binary fraction of 50\pct.
459: The corresponding binary IMF (\bmf\, eq. \ref{defbmf}) is shown by the thin
460: dotted line. The system MF peaks at 1~\tmsun\ and is truncated for higher
461: masses while there is a minor peak at 1.1~\tmsun\ corresponding to the
462: peak of the binary MF at \mmbox{\mmin+\mmax}.}
463: \end{center}
464: \end{figure}
465:
466: \begin{figure}
467: \begin{center}
468: %\escl{1.0}
469: \epsstr
470: \plotone{f4.eps}
471: %\plotone{../../../../Figures/brokenIMF.eps}
472: \caption{\label{cartoon}A population of stars and BDs
473: with different binary fractions can result in a discontinuous IMF,
474: even if the observed (system) IMF appears to be continuous.
475: The binary fraction of the BDs is lower than that of stars and therefore
476: a lower number, $\nbod$, of individual objects is required for the frequency
477: of BD and stellar systems of a given mass being equal (eq. \ref{defnbod}).
478: Note the overlap region
479: indicated by the horizontal dotted lines: some starlike bodies may actually
480: be physical BDs (upper line) while some BD-likes are indeed very-low-mass stars.
481: }
482: \end{center}
483: \end{figure}
484:
485: \section{The Method}\label{sec:method}
486: %
487: \begin{table}
488: \caption{\label{tabparspace}List of the variable parameters that
489: span the parameter space of a two-population IMF studied in this
490: contribution. Note that $\abd$ is not varied for IC~348 and the
491: Pleiades since it is not well-constrained by the data for these
492: clusters.}
493: \begin{center}
494: \begin{tabular}{cccc}\hline
495: Parameter No.&Quantity &varied from&to (approx.)\\\hline
496: 1 &$\abd$ &0 &1\\
497: 2 &$\mhibd$ &$\mlost$ &$\mlost+0.2$\\
498: 3 &$\log\rpop$&-1 &0\\\hline
499: \end{tabular}
500: \end{center}
501: \end{table}
502: %
503: \begin{table}
504: \caption{\label{tabdf}Size of the observed sample, $\nsys$
505: (= the number of observed systems),
506: number of data points (mass bins), $n_\mathrm{data}$,
507: number of fitting parameters, $c$, and the number of degrees of
508: freedom, $\nu=n_\mathrm{data}-1-c$, for the four clusters under study.}
509: \begin{center}
510: \begin{tabular}{c|c|ccc}\hline
511: Cluster &$\nsys$ &$n_\mathrm{data}$&$c$&$\nu$\\\hline
512: Trapezium &1040 &30 &3 &26\\
513: TA & 127 &9 &3 &5\\
514: IC~348 & 194 &9 &2 &6\\
515: Pleiades &$\sim500$&19 &2 &16\\\hline
516: \end{tabular}
517: \end{center}
518: \end{table}
519: %
520: \subsection{The Parameter Space}\label{ssec:parspace}
521: The most straightforward way to calculate the influence of binaries on the
522: stellar/substellar statistics would be the Monte Carlo method
523: \citep{KGT91,KTG93}. However,
524: for better (smoother) results and to reduce the computational efforts we
525: do not use a Monte Carlo approach here but a semianalytical approach in
526: which the binary mass function is calculated via numerical integration of
527: eq. (\ref{defbmf}) for each population. This can be done with a standard
528: quadrature algorithm. This algorithm has been verified
529: with a Monte Carlo simulation of a few million random experiments,
530: each being a random draw from the IMF.
531: The Monte Carlo method is used later to determine the BD-to-star ratio
532: and the total binary fractions
533: within defined mass ranges (\S\S~\ref{ssec:R} and \ref{ssec:f}).
534: It makes use of the \emph{Mersenne Twister} random number generator
535: developed by \citet{mersennetwister}. Since there are only a few
536: runs to be done (in contrast to hundreds of thousands of runs during an
537: iterated parameter scan, see below) the simple Monte Carlo approach with
538: an appropriately large random sample is fully sufficient for this purpose.
539:
540: The lack of star-BD binaries and the truncation of the BD binary separation
541: distribution suggest two disjunct populations where binary components
542: are taken only from the same population. For reasons discussed in
543: \S~\ref{sec:discussion} we call these populations \emph{BD-like} and
544: \emph{starlike}.
545: The binary corrections are
546: therefore applied separately to the stellar and the substellar regime rather
547: than to a combined population.
548: Random pairing over the whole mass regime is not considered further as it
549: leads to too many star-BD binaries \citep{Ketal03}.
550:
551: The approach here
552: requires separate application of normalizations on both populations.
553: For this purpose we define the \emph{population ratio},
554: \begin{equation}
555: \rpop=N_1/N_2
556: \end{equation}
557: where $N_1$ is the number of individual bodies of the BD-like population
558: ($\mlobd\le m\le\mhibd$) and $N_2$ ($\mlost\le m\le\mmax$)
559: that of the starlike population.
560: This must not be mixed up with the \emph{BD-to-star ratio}, $\rbod$,
561: (eq. [\ref{defR}]) which refers to
562: \emph{physical BDs} and stars separated by the HBL.
563:
564: The partial IMF for each population can be
565: described by parameters $\alpha_i$, $m_j$, and a normalization
566: constant. In this work a single power law for BDs ($i$ = ``BD'')
567: and a two-part power law ($i$=1 or 2) for stars is applied.
568: Thus, there is a mass border, $m_{12}$, separating the two power law
569: regimes. The parameters of both populations form a three-dimensional
570: parameter space of the IMF model for each cluster.
571: It has been found that the lower mass limits of the BD-like population,
572: $\mlobd=0.01\,\msun$, and that of the starlike population,
573: $\mlost=0.07\,\msun$, are suitable for all studied clusters.
574: Furthermore, we focus on the canonical stellar IMF \citep{Kr01}
575: with $\alpha_1=1.3$ for $m\le0.5\msun\equiv m_{12}$
576: and $\alpha_2=2.3$ for $m>0.5\msun$.
577: The reason is that this canonical IMF has been verified with
578: high confidence by other observations as well as theoretically.
579: Thus, only the BD-like power law, $\abd$, the normalization of the
580: BD-like population against the stellar one, $\rpop$,
581: and the upper mass border, $\mhibd$, of the BD-like regime are
582: the parameters to be varied.
583: Because of the sparse and probably incomplete data sets
584: for IC~348 and the Pleiades, $\abd$ has also been set to the canonical
585: value (i.e. $\abd=\alpha_0=0.3$) for these clusters, varying only $\rpop$
586: and $\mhibd$.
587:
588: Table \ref{tabparspace} lists the variables and their
589: range of variation for the Trapezium, IC~348, and the Pleiades.
590: As the upper mass limits of stars in the clusters one can either take
591: the maximum observed mass or a theoretical mass limit,
592: as given by \citet{WK06}.
593: Because in our model the IMF is cut sharply rather than declining softly
594: as shown in that work, we here set the mass limit somewhat below the
595: \citet{WK06} limits.
596: For the Trapezium we set an upper limit of 10~\tmsun, 1.5~\tmsun\ for TA,
597: 3~\tmsun\ for IC~348 and 5~\tmsun\ for the Pleiades.
598: Note that the observed most massive star in the Trapezium has a mass of
599: $\approx50\,\msun$, but varying $\mmax$ to this value does not affect
600: the results significantly.
601:
602: Little is known about nonplanetary substellar
603: objects below the standard opacity limit for fragmentation, 0.01\tmsun\,,
604: making this a reasonable lower mass limit for our BD statistics.
605: The lower-mass limit for stars, however, is chosen somewhat
606: arbitrarily but is justified by a smaller test study that finds less
607: agreement with observational data used here if the stellar mass is lowered
608: too much below 0.08\tmsun. As will be shown later, the extreme case
609: of a stellar mass range that includes the BD mass range leads to
610: poor fits to the observational MFs
611: as well as to some inconsistency with the observed binary
612: fraction. This is discussed later in \S\S~\ref{sec:results} and
613: \ref{sec:discussion}.
614: In general, the choice of the mass ranges also influences the
615: mass ratio distribution,
616: since for random-pairing each component is distributed
617: via the IMF. Thus, for a given primary mass out of a chosen population,
618: the companion mass distribution is just equal to $\xi(m)$ within the mass
619: interval \mmbox{\mmin\le m\le\mpri}.
620:
621: The overall binary fraction is taken as a constant for each population.
622: For the stars we adopt an unresolved binary fraction of 80\pct\
623: for TA ($\fst=0.80$) and 40\pct\ for the others ($\fst=0.40$),
624: in accordance to the observational data.
625: For BDs we choose a value of 15\pct\ for all cases ($\fbd=0.15$). Furthermore,
626: we found the need to introduce an overlap of the BD and star regimes between
627: 0.07 and about 0.15--0.2\tmsun. Indeed, there is no reason why the
628: upper mass border of the BD-like population should coincide with the
629: lower mass border of the starlike population.
630: The physical implications of this required overlap region are discussed
631: in \S~\ref{sec:discussion}.
632:
633: \subsection{$\chi^2$ Minimization}
634: \label{ssec:chisq}
635: In \S~\ref{ssec:parspace} a parameter space of three (Trapezium, TA)
636: and two (IC~348, Pleiades) dimensions of individually adjustable
637: quantities has been defined.
638: For each cluster, the quality of a set of fitting parameters
639: is characterized via the $\chi^2$ criterion,
640: \begin{equation}
641: \chi^2=\sum\limits_i^{n_\mathrm{data}} \frac{\left(N_i-N\xis(m_i)\right)^2}{\sigma_i^2}\,,
642: \end{equation}
643: where $\xis(m_i)$ is \simf\ at the midpoint of the $i$th bin of the
644: observational histogram
645: that is taken as the reference and $n_\mathrm{data}$
646: is the number of mass bins in the observational histogram
647: (Tab. \ref{tabdf}).
648: The values
649: $\sigma_i$ represent the error bars of the observational data, or,
650: if none are given, the Poisson errors, and
651: $N_i$ is the number of systems found in the $i$th mass bin.
652: The fitted \simf\ is normalized to the total number of systems in the
653: cluster.
654:
655: That set of $c$ fitting parameters that minimizes the reduced $\chi^2$
656: value, $\chi_\nu^2=\chi^2/\nu$, defines our best-fit model. Here
657: $\nu=n_\mathrm{data}-c-1$ is the number of degrees of freedom.
658: The values for $n_\mathrm{data}$, $c$, $\nu$, and the sample sizes
659: for the studied clusters are listed in Table \ref{tabdf}.
660: The reduction of $\nu$ by 1 is due to the normalization of the
661: total number of stars and BDs against the observational data.
662: For the two-component model the fitting parameters are the power in the
663: BD regime, $\abd$,
664: the upper mass border of the BD population, and the ratio $\rpop$; thus,
665: $c=3$ (2, for IC~348 and the Pleiades).
666: The probability, $P$, that the model has a reduced $\chi^2$,
667: $\chi_\nu^2=\chi^2/\nu$, as large as or larger than the value actually
668: obtained is calculated from the incomplete Gamma function \citep{numres},
669: \begin{equation}
670: Q(\chi_\nu^2|\nu)=P(\tilde{\chi}_\nu^2\le\chi_\nu^2)\,.
671: \end{equation}
672:
673: The logarithmic mass error is not mentioned in the sources, but is given
674: by the photometric measurements and to a larger degree by uncertain
675: theoretical models of stars and BDs.
676: We assumed a value of
677: \mmbox{\Delta\log m=0.05} for the Trapezium and
678: \mmbox{\Delta\log m=0.1} for the
679: others, corresponding to a relative error in the mass estimates of 12\pct\
680: and 26\pct, respectively. This error has been estimated from the width
681: of the logarithmic bins in the observational data for the Trapezium, TA and
682: IC~348. For the Pleiades, for which non-equally spaced data from different
683: sources are given, we assume the same log mass error as for TA, and
684: IC~348.
685: To take this into account the $\log m$ values have
686: been smoothed by a Gaussian convolution corresponding to a lognormal
687: smearing of the masses.
688:
689: After generating the model IMFs for both populations
690: a binary mass function (eq. [\ref{defbmf}]) is derived separately from the
691: stellar and
692: substellar IMF, i.e. there are formally no BD-like companions to starlike
693: primaries, in accordance with the observations.
694: This leads to consistency with the observed binary fraction
695: (\S~\ref{ssec:f}) and the BD desert. Note, though, that as a result of
696: the required overlap both the BD-like and the starlike population
697: contain stars as well as BDs; thus, we do have star-BD
698: pairs in our description (more on this in \S~\ref{sec:discussion}).
699:
700: The addition of the two
701: resulting system IMFs leads to an overall \simf\ for the whole mass range.
702: By adjusting the power law coefficients and the population ratio $\rpop$,
703: the \simf\ is fitted against the observational data such that $\chi^2$
704: is minimized. The prominent substellar
705: peak in the Trapezium cluster below 0.03~\tmsun\ \citep{Muetal02} is ignored
706: here because it is well below the BD--star mass limit and therefore does not
707: interfere with any feature there. Furthermore, it is possibly an artefact
708: of the BD mass-luminosity relation \citep{LaLa03}.
709:
710: \begin{figure}
711: \begin{center}
712: \escl{0.7}
713: \plotone{f5.eps}
714: %\plotone{../../../../Figures/pbTPsstd.eps}
715: \caption{\label{pbsstd}%
716: The marginal probability density distributions of the parameters
717: $\abd$, $\log\rpop$ and $\log\left(\mhibd/\msun\right)$ as an
718: estimate of the errors for the Trapezium cluster fit
719: (see \S~\ref{ssec:errors}).
720: The peak-width engulfing 68\pct\ of the whole
721: parameter sample (thus referring to $\approx 1\sigma$,
722: dashed vertical lines) is taken as the error for each parameter.
723: Because the median (solid vertical line) of the sample
724: does not match the best-fit value (dash-dotted line) in all cases,
725: both are listed in Table \ref{tabalpha}.}
726: \end{center}
727: \end{figure}
728: %
729: \subsection{Error Estimation}
730: \label{ssec:errors}
731: The errors of the parameters are estimated from the marginal
732: distribution of each parameter within the parameter space
733: \citep{STP}. The
734: marginal probability density distribution, $p$, is
735: \begin{equation}
736: \begin{array}{rcl}
737: p(A_i)&=&\dif P(A_i)/\dif {A_i}\\
738: &=&\int\limits_{-\infty}^\infty
739: \int\limits_{-\infty}^\infty \mathcal{L}(A_1,A_2,A_3)\,\dif A_k \dif A_j\,,
740: \end{array}
741: \end{equation}
742: where $A_1, A_2$ and $A_3$ correspond to $\abd$, $\rpop$, $\mhibd$
743: and $i=1,\dots,3$; $j,k\ne i$. The likelihood $\mathcal{L}$
744: is defined as
745: \begin{equation}
746: \mathcal{L}(A_1,A_2,A_3)=\mathcal{N} e^{-\chi^2/2}\,,
747: \end{equation}
748: where the normalization constant $\mathcal{N}$ is such that the
749: integral over the whole range is one.
750: Note that $\chi^2$ instead of $\chi_\nu^2$ is used here.
751: The integral is approximated by summation with equidistant
752: stepping of each parameter.
753: We conservatively rejected all parameter sets with $P<0.27\pct$
754: (corresponding to $\chi^2_\nu>\pm3\,\sigma$; \S~\ref{ssec:chisq}).
755:
756: As an example, the parameter distribution for the Trapezium is
757: illustrated in Figure \ref{pbsstd}.
758: The interval around the median which contains 68\pct\ of the
759: scanned parameter sets is taken as the $\sigma$ measure of the errors.
760: The interval limits and the medians are shown among the best-fit
761: results in Table \ref{tabalpha}. Apparently, the median of the
762: probability density distribution does not always coincide with
763: the best-fit value. The reason for this is that the distribution
764: is slightly
765: asymmetric for most parameters and clusters with the Pleiades
766: being the by far worst case. This results in asymmetric error bars.
767: Note that there are sets of $\abd$, $\rpop$, and $\mhibd$ that are
768: within these error limits but with a $\chi^2>3\sigma$, and thus
769: the error limits may be slightly over-estimated.
770:
771: Another possibility to illustrate the statistical significance of
772: a fit is by its confidence contours within a two-dimensional subspace
773: of the parameter space, as is shown later in \S~\ref{ssec:R}.
774:
775: %%%%
776: \begin{figure}
777: \escl{0.5}
778: \plotone{f6.eps}
779: %\plotone{../../../../Figures/imf4TP.eps}
780: \caption{\label{IMF4TP}%
781: Four different IMF models applied to the Trapezium cluster.
782: Top panel: A classical continuous-IMF fit (dashed line)
783: for the Trapezium cluster
784: with random pairing over the whole BD and stellar mass range.
785: The method of converting the IMF to the \simf\ (solid curve)
786: is essentially the
787: same as for the two-component IMFs discussed in this paper but
788: with no allowance for an overlap or a discontinuity in the BD-star
789: transition region. Although the general shape of the observational
790: histogram (thin steps) is represented
791: by this fit this model leads to an unrealistic high fraction of star-BD
792: binaries and is therefore discarded.
793: Second panel: The Trapezium IMF fitted with two separate IMFs,
794: $\xi_\mathrm{BD}$ (dotted line) and $\xi_\mathrm{star}$ (dashed line),
795: as in Figure \ref{IMF4sstd} (the two-component IMF).
796: Third panel: The same but with the sum IMF of both component IMFs
797: shown here as the possible appearance of the IMF if all binaries could
798: be resolved. Apparently, there is a ``hump'' between
799: about 0.07~\tmsun\ and 0.2~\tmsun\ bracketing the
800: probable overlap region of the BD-like and starlike populations.
801: Bottom panel: Trapezium fitted with two
802: separate IMFs but with no overlap. The binary correction leads to
803: the dip
804: near the hydrogen-burning limit. However, the fit is only slightly worse
805: than that one with an overlap.
806: }
807: \end{figure}
808: %%%%
809: \begin{figure}
810: \escl{0.5}
811: \plotone{f7.eps}
812: %\plotone{../../../../Figures/imf4sstd.eps}
813: \caption{\label{IMF4sstd}%
814: Two-component IMFs of the Trapezium, IC~348, and TA clusters based on
815: observational data by \citet{Lu04b}, \citet{Luetal03b} and \citet{Muetal02}
816: (solid histograms), as well as for the Pleiades cluster using data
817: from \citet{Moetal03} (filled circles), \citet{Doetal02} (open squares)
818: and from the Prosser and Stauffer Open Cluster Database (from which only
819: those for $\log m/\msun\le0.55$ are used for fitting).
820: For all clusters the stellar IMF, shown here as the dashed line,
821: has the canonical or standard shape after \citet{Kr01}, i.e. $\alpha_1=1.3$
822: below 0.5~\tmsun\ and $\alpha_2=2.3$ above 0.5~\tmsun.
823: The Trapezium IMF is fitted without the
824: substellar peak indicated by the faint dotted steps.
825: The dotted lines show the optimized binary-corrected BD-like body
826: IMFs (distribution of individual BD-like bodies).
827: The resulting model system IMFs
828: are represented by the solid curves. Ages are
829: from \citet{Luetal03a}, \citet{Luetal03b},
830: \citet{HiCaFe01} and \citet{BSJ04} for TA,
831: IC~348, Trapezium and the Pleiades, respectively. Note that
832: the BD-like IMF of IC~348 and the Pleiades are set to the canonical
833: slope $\abd=0.3$ since they are not well constrained due to
834: the lack of BD data for these clusters.}
835: \end{figure}
836: %%%%
837: \begin{figure}
838: \begin{center}
839: \escl{1.0}
840: \epsstr
841: \plotone{f8.eps}
842: %\plotone{../../../../Figures/tax.eps}
843: \caption{\label{TAx}%
844: The two-component model with canonical stellar IMF for a slightly
845: modified TA observational histogram (black solid steps). The peak at 1~\tmsun\
846: and the dent around 0.3~\tmsun\ have been slightly flattened without
847: leaving the error bars of the original data (grey steps and bars).
848: This fit has a reasonable confidence of about $P=0.25$, about 12 times
849: larger than for the original histogram ($P=0.02$, see Tab. \ref{tabalpha}).
850: }
851: \end{center}
852: \end{figure}
853: %%%%
854: \begin{figure}
855: \escl{0.5}
856: \plotone{f9.eps}
857: %\plotone{../../../../Figures/cnt4.eps}
858: \caption{\label{cnt4}%
859: Contour plots of 5\pct\ and 1\pct\ significance levels from $\chi^2$
860: fitting (see \S~\ref{sec:method} for details) of
861: $\abd$ and the BD-like to starlike ratio
862: $\rkbd$ at $m=0.075\,\msun$ (eq. \ref{defrkbd})
863: for the Trapezium, TA, IC~348 and the Pleiades.
864: The stellar IMF is canonical, i.e. $\alpha_1=1.3$ and $\alpha_2=2.3$
865: while the upper BD mass limits, $\mhibd$, are the best-fit values
866: from Table \ref{tabalpha}.
867: All fitting parameters outside the
868: solid line are rejected with 95\pct, and those outside the
869: dashed line are rejected with 99\pct\ confidence. The optimum
870: of the Trapezium and TA is marked by the filled circle while
871: the cross marks the standard/canonical configuration with a continuous
872: IMF (i.e. $\log\rkbd=0$) and $\abd=0.3$.
873: As can be seen, it is well outside both levels
874: (even for arbitrary $\abd$). For the Pleiades it is still well outside
875: at least for reasonable choices of $\abd$.
876: Note that the optimum for $\abd$ for IC~348 and the Pleiades does exist
877: in this 2D cross section but not in the full 3D parameter space.}
878: \end{figure}
879: %%%%
880: \begin{figure}
881: \begin{center}
882: \escl{0.50}
883: \plotone{f10.eps}
884: %\plotone{../../../../Figures/bf4.eps}
885: \caption{\label{fbin4}%
886: The binary fraction $f$ (eq. \ref{fmpri})
887: of the four discussed clusters as a function of
888: primary mass, $m_\mathrm{prim}$, for the two-component best-fit
889: IMF model (with canonical stellar IMF) is shown by the solid line.
890: The BD binary fraction appears flat due to the equal-mass pairing used in
891: our algorithm. Since merely very sparse data is available for low-mass BDs,
892: only the higher-mass end of the curve should be taken into account.
893: For comparison,
894: the average binary fractions of G, K, M and L type stars in the solar
895: neighbourhood
896: (average age $\sim 5$~Gyr; filled circles: \citealt{Lada2006}, open squares:
897: \citealt{Ketal03}) have been added to each panel.
898: The shaded region in the top panel indicates the approx. 1~Myr old
899: pre-main sequence data \citep{Duchene99} while the thin solid line
900: represents the initial TA-like dynamical model from \citet{Ketal03}.
901: The single filled square in the top panel at $\log m\approx-1.2$
902: is the VLMS/BD datum inferred by \citet{Krausetal06} for TA.
903: Note that the stars in TA have a binary fraction near
904: 100\pct\ \citep{Duchene99,Ketal03}.}
905: \end{center}
906: \end{figure}
907: %%%%
908:
909: \section{Results}
910: \label{sec:results}
911: \begin{table*}
912: \caption{\label{tabalpha}The best-fit BD-like power law coefficients, $\abd$,
913: population ratios, $\rpop$, and BD-like upper mass limits, $\mhibd$.
914: The uncertainties are derived in \S~\ref{sec:method}.
915: Note that $\abd$ is set to the canonical value for IC~348 and the Pleiades.}
916: \begin{center}
917: \begin{tabular}{|l|ccc|ccc|cc|}\hline
918: &\multicolumn{3}{c|}{Median $\pm1\sigma$}&\multicolumn{3}{c|}{Best fit} &&\\
919: Cluster &$\abd$ &$\log\rpop$ &$\mhibd$ &$\abd$ &$\log\rpop$&$\mhibd$ &$\chi_\nu^2$&$P$\\\hline
920: Trapezium &$0.47\pma{0.18}{0.19} $&$-0.67\pma{0.10}{0.09}$&$0.22\pma{0.04}{0.05}$ & 0.45 &-0.66 &0.22 &0.23&0.99998\\
921: TA &$-0.08\pma{0.63}{1.05}$&$-0.68\pma{0.13}{0.14}$&$0.11\pma{0.03}{0.04}$ & 0.00 &-0.62 &0.12 &2.69&0.020 \\
922: IC~348 &--- &$-0.57\pma{0.23}{0.19}$&$0.22\pma{0.06}{0.08}$ &$0.3\,\mbox{(c)}$ &-0.51 &0.22 &1.01&0.416 \\
923: Pleiades &--- &$-0.82\pma{0.44}{0.11}$&$0.07\pma{0.02}{0.20}$ &$0.3\,\mbox{(c)}$ &-0.80 &0.06 &1.20&0.257 \\\hline
924: \end{tabular}
925: \end{center}
926: \end{table*}
927:
928: \subsection{The IMF for BDs and Stars}
929: \label{ssec:IIMF}
930: Figure \ref{IMF4TP} demonstrates the results for different models
931: (continuous IMF, two-component IMF with and without overlap region) for the
932: Trapezium cluster. The continuous
933: IMF is shown for illustration only. Its underlying model assumes
934: a single population containing BDs as well as stars \citep{Ketal03}.
935: Instead of an overlap
936: as for the two-component model the mass border between the regime of
937: the BD slope $\abd$ and the (canonical) stellar regime is varied. The
938: total binary fraction, $\ftot$, is set to 0.4 with random pairing
939: among the entire population. Our calculations give
940: a best fit with $\abd=-0.4\pm{0.2}$, $\mhibd\equiv\mlost=0.093\pm{0.01}$
941: (and canonical stellar IMF) which is comparable to the canonical
942: IMF for BDs and stars \citep{Kr01}.
943: As already mentioned in \S~\ref{sec:intro}, it leads to a high number
944: of star-BD binaries, $\fmix=15\pct\pm4\pct$ altogether (see \S
945: \ref{ssec:f}),
946: while \citet{Ketal03} found 7\pct--15\pct\ star-BD binaries with
947: $a\ga30\unit{AU}$ in their standard model with stars and BDs.
948: Furthermore, the different orbital
949: properties of BD-BD and star-star binaries (Fig. \ref{sepdist})
950: cannot be explained by a single-population model without further
951: assumptions.
952: Therefore, it is not considered further here.
953:
954: The two-component IMF model (Fig. \ref{IMF4TP}, \emph{second panel})
955: accounts for the empirical binary properties of BDs and stars.
956: Apart from this it also fits the observational data
957: (Fig. \ref{IMF4TP}, \emph{histogram}) slightly better than
958: the continuous model, especially in the BD/VLMS region and the
959: ``plateau'' between 0.1 and 0.5~\tmsun.
960:
961: Objects with equal mass and composition appear as equal in observations,
962: even if they have been formed in different populations. Thus, it is hard
963: to determine their formation history.
964: A high-resolution survey that
965: resolves most of the multiples would yield an overall mass function
966: from the lowest mass BDs to the highest mass stars.
967: However, if such a resolved individual mass function is composed of two
968: overlapping populations, it may be possible to detect its imprints
969: as an excess of objects within the overlap region.
970: To illustrate this ((Fig. \ref{IMF4TP}, \emph{third panel}),
971: we construct such an overall IMF by simply adding the two
972: IMF components from the second panel. BDs and stars are paired separately
973: from their respective mass range.
974: It is clear that the addition of the two IMFs leads to discontinuities,
975: i.e. in the overlap region the continuous IMF increases steeply at the
976: minimum stellar mass and drops again at the upper mass limit of the
977: BD-like regime.
978: Even with a smoother drop at each end of the BD-like and starlike IMF
979: it is likely to be detectable once most of the multiples have
980: been resolved. Thus, a discovery of such a ``hump'' within the
981: resolved IMF would strongly support the two-component model.
982:
983: To illustrate the effect of the overlap region, a two-component IMF
984: without an overlap is shown in the bottom panel of Figure \ref{IMF4TP}.
985: Like the continuous
986: IMF this approach gives a slightly worse fit, especially between
987: 0.1 and 0.2~\tmsun\ ($-1\le\log m/\msun\le-0.7$). However,
988: the ``dip'' in this region is only weak.
989:
990: The results of the two-component IMF model
991: for all clusters are shown in Figure \ref{IMF4sstd}, the power law
992: coefficients being listed in Table \ref{tabalpha}.
993: For the Pleiades and IC~348 the BD IMF slope has been kept
994: standard, i.e. $\abd=0.3$ \citep{Kr01},
995: because the sparse, and in the case of the Pleiades
996: most probably incomplete data, do not allow useful confidence
997: limits to be placed on $\abd$. For the same reason
998: the highly uncertain observational data for $\log(m/\msun)\ge0.65$ have
999: not been used for fitting (but are still plotted for completeness).
1000:
1001: As can be seen the BD/VLMS and stellar IMFs do not meet at the BD-star
1002: boundary. The number density of individual BDs near the BD-star border is
1003: about one third of that of individual low-mass M dwarfs. This discontinuity
1004: cannot be seen directly in the observational data because it is masked by
1005: the different binary fractions for different masses (\S
1006: \ref{ssec:IMF:SIMF}), but the IC~348 data
1007: (histogram in the third panel of Figure \ref{IMF4sstd})
1008: may indeed be showing the discontinuity (compare third panel of Figure
1009: \ref{IMF4TP}).
1010:
1011: %%
1012: Furthermore, there are large uncertainties
1013: in the mass determination of stars in stellar groups such as TA which
1014: may probably lead to even larger error bars of the bins than shown in the
1015: observational data. Even within the given error bars certain variations of
1016: these data provide a much better fit to the canonical stellar IMF
1017: with $P=0.25$ (original: $P=0.02$, Table \ref{tabalpha}). This
1018: has been done in Figure \ref{TAx} by lowering the peak near 1~\tmsun\
1019: and somewhat rising the ``valley'' around 0.3~\tmsun
1020: corresponding to the re-shuffling of 9 stars out of 127
1021: (Tab. \ref{tabparspace})
1022: by 0.1 to 0.4~\tmsun\
1023: (one bin width), e.g. through measurement errors.
1024: This suggests that, in agreement with \citet{Ketal03}, the TA IMF might
1025: not necessarily be inconsistent with the canonical IMF.
1026:
1027: Except
1028: for the Pleiades all clusters show features near the peak in the low-mass
1029: star region that are slightly better fitted with a separate BD population
1030: \emph{and} an overlap region. Although these features alone do not reject
1031: the continuous IMF model, they might be taken as a further support
1032: of the argumentation towards a two-populations model.
1033:
1034: It should be mentioned that the continuous IMF model (Fig. \ref{IMF4TP},
1035: \emph{first panel})
1036: still fits the observed Trapezium and Pleiades MF with high confidence
1037: (while failing for the other clusters) but leads to VLMS binary
1038: properties that are inconsistent with the observed properties as
1039: mentioned in \S~\ref{sec:intro}. A possible extension of our modelling
1040: would thus be to fit both the observed IMF \emph{and} the observed binary
1041: statistics.
1042:
1043: \subsection{BD to Star Ratio}
1044: \label{ssec:R}
1045: We analyzed the BD-to-star ratio $\rbod$ of TA, Trapezium and IC~348,
1046: defined as
1047: \begin{equation}\label{defR}
1048: \rbod=\frac{N(\mbox{0.02--0.075}\,\msun)}{N(\mbox{0.15--1.0}\,\msun)}\,,
1049: \end{equation}
1050: where $N$ is the number of bodies in the respective mass range.
1051: The mass ranges are chosen to match those used in \citet{KB03b} in their
1052: definition of $\rbod$. Since
1053: BDs below 0.02~\tmsun\ are very difficult to observe and since TA and
1054: IC~348 do not host many stars above 1~\tmsun\ (in contrast to Trapezium and the
1055: Pleiades), we restricted the mass ranges to these limits.
1056:
1057: The Pleiades cluster is difficult to handle due to a lack of BD data.
1058: Moreover,
1059: at an age of about 130~Myr \citep*{BSJ04} it has undergone dynamical evolution,
1060: and massive stars have already evolved from the main sequence which
1061: affects the higher mass end of the IMF \citep{Moetal04}.
1062:
1063: \begin{figure}
1064: \begin{center}
1065: \escl{1.0}
1066: \epsstr
1067: \plotone{f11.eps}
1068: %\plotone{../../../../Figures/fb.eps}
1069: \caption{\label{fbuni}%
1070: The binary fraction $f(\mpri)$ of our best 2-component IMF fit for the
1071: Trapezium (solid line) with $\abd=0.5$ in
1072: comparison with the binary fraction, $f_\mathrm{c}(\mpri)$, of the best
1073: continuous IMF model (dotted line, see also top panel in Figure \ref{IMF4TP}).
1074: The 2-component model assumes random pairing of stars but
1075: equal-mass pairing for BD binaries. The continuous model, on the other hand,
1076: assumes random pairing of companions chosen randomly from the canonical IMF
1077: ($\abd=0.3$, $\alpha_1=1.3$ and $\alpha_2=2.3$).
1078: This model is merely shown for illustrative purposes and has already been
1079: discarded (\S~\ref{sec:intro}). The thin dashed line refers to the same
1080: two-component IMF as the solid line, but with random-pairing instead of
1081: equal-mass pairing for BDs.
1082: The visible jumps in $f(\mpri)$ at
1083: $\log(m/\msun)=-1.15$ and -0.7 are due to the truncation
1084: at the upper and lower mass limit of the BD-like and the starlike
1085: population with their different binary properties.
1086: The global shapes of the continuous and the two-component IMF curves
1087: are similar but $f_\mathrm{c}(\mpri)$ is significantly higher
1088: than $f(\mpri)$ in the low-mass star and VLMS regime and outside the
1089: uncertainties of the data by \citet{Lada2006}.}
1090: \end{center}
1091: \end{figure}
1092:
1093: Another relevant quantity is the ratio $\rkbd$ of BD-like to starlike
1094: objects at the HBL, i.e. the classical BD-star border,
1095: \begin{equation}\label{defrkbd}
1096: \rkbd=\frac{\xi_\mathrm{BD}(0.075\,\msun)}{\xi_\mathrm{star}(0.075\,\msun)}\,.
1097: \end{equation}
1098: In the classical continuous IMF approach it is one by definition, because
1099: otherwise the IMF would not be continuous.
1100: In a two-component IMF its value depends on the shapes of the
1101: BD-like and the starlike IMF, as well as on the BD-to-star ratio.
1102: The evaluation of equation (\ref{defrkbd}) yields
1103: $\rkbd=0.17$ for the Trapezium, $\rkbd=0.30$
1104: for TA, $\rkbd=0.22$ for IC~348, and $\rkbd=0.3$ for the Pleiades.
1105: Although $\rkbd$ is not an input parameter here but is calculated from
1106: the two IMFs and their relative normalization, $\rpop$, it
1107: could easily be used as one instead of $\rpop$ while calculating the
1108: latter from $\rkbd$.
1109:
1110: Figure \ref{cnt4} shows the contour plots of the 5\pct\ and 1\pct\
1111: significance ranges in the $\abd-\rkbd$ space for
1112: Trapezium, TA, IC~348, and the Pleiades.
1113: The significance values are calculated from $\chi_\nu^2$ via the incomplete
1114: Gamma function as described in \S~\ref{sec:method}.
1115: The contours mark the regions outside which the
1116: hypothesis of a two-component IMF with a single power law for BDs
1117: and a double power law for stars has to be rejected with 95\pct\
1118: or 99\pct\ confidence.
1119: Also shown (by a cross) is the standard configuration with
1120: $\abd=0.3$ and $\log\rkbd=0$ for the continuous standard IMF.
1121: This point is outside both levels for all three clusters, and at least
1122: for Trapezium and TA it is well outside even for arbitrary
1123: $\abd$.
1124: In other words, \emph{the corresponding hypothesis
1125: of a continuous IMF has to be rejected with at least 99\pct\ confidence.}
1126:
1127: The size of the non-rejection areas can be used for an estimate of the
1128: errors of $\abd$ and $\rkbd$. However, one has to keep in
1129: mind that these are only maximum possible deviations with all the other
1130: parameters kept at the optimum. The non-rejection areas in the
1131: full three-dimensional parameter space are therefore
1132: expected to be somewhat narrower.
1133:
1134: \subsection{Binary Fraction}
1135: \label{ssec:f}
1136: Additionally, the binary fraction for each cluster as well as the
1137: total binary fraction is calculated for the best-fit models.
1138: The fraction of binaries as a
1139: function of the primary mass, \mmbox{f(m_\mathrm{prim})}, among all systems
1140: is shown in Figure \ref{fbin4}.
1141: For stars the binary fraction is a monotonic function of the
1142: primary mass in agreement with the data \citep{Lada2006}, at least
1143: for the Trapezium, IC~348, and the Pleiades.
1144: For BDs, however,
1145: it is flat due to the equal-mass pairing. In the case
1146: of random pairing it would approach zero for very low mass BDs
1147: (Fig. \ref{fbuni}).
1148: The true
1149: mass ratio function grows monotonically with the mass ratio and becomes very
1150: steep near $q=1$, as shown by \citet{Reidetal06}. Thus, for BDs the true binary
1151: fraction is probably closer to the equal-mass case than the random pairing
1152: case.
1153:
1154: For comparison, Figure \ref{fbuni} also shows the binary fraction,
1155: $f_\mathrm{c}(\mpri)$, for the Trapezium that would result from a
1156: continuous IMF. Although the overall shape is very similar to that of the
1157: two-component model the binary fraction near the BD-star transition is
1158: significantly higher for low-mass stars than the observed values while
1159: being approximately equal for stars above 1~\tmsun.
1160: Thus, a continuous IMF cannot fit the observational data as good as a
1161: two-population IMF even if $\ftot$ is reduced.
1162:
1163: The total fractions of BD-BD binaries, $\fbd$, of
1164: star-star binaries, $\fst$, and the fraction of
1165: (very low mass) star-BD binaries, $\fmix$, are of further interest.
1166: Pairs of the latter type consist of
1167: two objects of the BD-\emph{like} or star\emph{like} population (see
1168: \S~\ref{sec:discussion} for the motivation and definition of the
1169: populations) but where the primary object is a star
1170: ($m_\mathrm{prim}\ge0.075\,\msun$)
1171: within the BD-VLMS overlap region between 0.07 and 0.15~\tmsun,
1172: while the companion is a true, physical BD with a mass below
1173: 0.075~\tmsun.
1174:
1175: For each cluster we define
1176: \begin{eqnarray}
1177: \fbd &=&\frac{\nbbd}{\nsbd}\\
1178: \fmd &=&\frac{\nbmd}{\nsmd}\\
1179: \fst &=&\frac{\nbst}{\nsst}\\
1180: \fmix&=&\frac{\nbsb}{\nsst}\,,
1181: \end{eqnarray}
1182: where $\nbbd$ is the number of all BD-BD binaries (i.e. all objects
1183: have masses $\le0.075\,\msun$), $\nbmd$ that of all M dwarf--M dwarf
1184: (MD-MD) binaries, $\nbst$ that of all
1185: star-star binaries, and $\nbsb$ the number of mixed (star-BD) binaries
1186: (composed of a VLMS and a BD).
1187: Furthermore, $\nsbd$ is the
1188: number of all BD systems (including single BDs), $\nsmd$ that of all M dwarf
1189: systems, and $\nsst$
1190: the number of all systems with a star as primary and with a stellar, BD,
1191: or no companion.
1192: As for the BD-to-star ratio we also
1193: applied the (primary) mass ranges from \citet{KB03b}
1194: on the BD and star sample from each cluster but with no gap between the
1195: BD and the stellar regime,
1196: i.e. $0.02\,\msun\le\mpri\le 0.075\,\msun$ for BDs
1197: and $0.075\,\msun\le\mpri\le 1\,\msun$ for stars. In addition, the binary
1198: fraction of M dwarfs is calculated in the same way as the stellar one
1199: but in the mass range $0.075\,\msun\le\mpri\le 0.5\,\msun$.
1200: The uncertainty limits of the binary fractions and BD-to-star ratios
1201: have been derived from those of the IMF slopes by applying the
1202: minimum and maximum slopes.
1203:
1204: The results are shown in Table \ref{tabbinfrac}.
1205: The binary fractions for BDs vary only
1206: slightly between 12\pct\ and 15\pct, while the stellar
1207: binary fraction, $\fst$, is about 70\pct\ for TA and 30\pct--40\pct\
1208: for the others. They are apparently slightly lower than the binary
1209: fractions that are set for the starlike population because of the
1210: nonconstant distribution binary as shown in Figure \ref{fbin4}.
1211: The binary mass function (eq. [\ref{defbmf}]) is smaller in the
1212: mass region below 1~\tmsun\ and thus the binarity is
1213: below average if the focus is set on this region.
1214:
1215: Furthermore, in the overlap region the relatively low binary fraction of
1216: VLMSs from the BD-like population also contributes to $\fst$,
1217: which results in an even lower value of $\fst$. This trend is more
1218: emphasized for the M dwarf binary fraction, $\fmd$, which is about 10\pct\
1219: lower than $\fst$ for each cluster but still much larger than $\fbd$.
1220:
1221: The star-BD binary fraction $\fmix$ is of special interest since it is
1222: the measure for the ``dryness'' of the BD desert.
1223: Note that due to the equal-mass pairing used here for BD-like binaries, the
1224: BD-like population formally does not contribute to $\fmix$.
1225: All star-BD binaries are from the starlike regime which
1226: extends down to $\mlost=0.07\,\msun$, i.e. into the BD mass regime.
1227: For the Trapezium, IC~348, and the Pleiades our two-component models
1228: yield values between 2\pct\ and 2.5\pct, whereas TA shows
1229: $\fmix\approx5\pct$. For comparison, the continuous IMF from the top panel
1230: of Figure \ref{IMF4TP} corresponds to $\fmix=15\pct\pm4\pct$.
1231: For mass ratio distributions other than equal-mass pairing
1232: we expect a higher $\fmix$, consisting mostly of binaries from the BD-like
1233: population with primary masses slightly above the HBL and companion
1234: masses below.
1235: For random pairing of BD-like binaries this increment is between
1236: 0.01 and 0.03. Also the lower mass limit of the starlike population
1237: slightly influences $\ftot$.
1238: It should also be noted that the size of the error limits for $\fmix$
1239: are calculated from the uncertainties of the IMF model
1240: (Tab. \ref{tabalpha}) but do not include the uncertainties of $\mlost$.
1241:
1242: \begin{table*}\small
1243: \caption{\label{tabbinfrac}
1244: Binary fractions $\fbd$ for BD-BD, $\fmd$ for MD-MD,
1245: $\fmix$ for star-BD and $\fst$ for star-star
1246: binaries in the studied clusters. Note that the star-BD binaries
1247: result from the pairing in the overlap region
1248: between about 0.075 and 0.15~\tmsun.
1249: Thus, for example, in IC~348 2.1\pct\ of all stars are expected
1250: to have a physical BD as a companion. In addition, the BD-to-star
1251: ratio between 0.02 and 1.0~\tmsun\, $\rbod$, and at the
1252: HBL, $\rkbd$, is listed. Note that the numerical uncertainties of
1253: $\fmix$ are probably much smaller than true ones since $\fmix$ largely
1254: depends on $\mlost$.
1255: }
1256: \begin{center}
1257: \begin{tabular}{|l@{\quad}|@{\quad}c@{\quad}c@{\quad}c@{\quad}c@{\quad}|c@{\quad}c@{\quad}|}\hline
1258: Cluster &$\fbd $&$\fmd$ &$\fst$ &$\fmix$ &$\rbod$ &$\rkbd$\\\hline
1259: Trapezium&$0.13\pm0.01$&$0.30\pm0.01$&$0.34\pm0.01$&$0.023\pm0.002$&$0.18\pm0.03$&$0.17\pm0.04$\\
1260: TA &$0.15\pm0.01$&$0.64\pm0.06$&$0.69\pm0.05$&$0.046\pm0.005$&$0.27\pm0.09$&$0.30\pm0.11$\\
1261: IC~348 &$0.13\pm0.02$&$0.29\pm0.02$&$0.33\pm0.02$&$0.021\pm0.004$&$0.20\pm0.10$&$0.22\pm0.11$\\
1262: Pleiades &$0.13\pm0.02$&$0.33\pm0.03$&$0.37\pm0.03$&$0.025\pm0.002$&$0.28\pm0.18$&$0.28\pm0.18$\\\hline
1263: \end{tabular}
1264: \end{center}
1265: \end{table*}
1266:
1267: \section{Discussion: Brown dwarfs as a separate population?}
1268: \label{sec:discussion}
1269: \subsection{An Apparent Discontinuity}
1270: By correcting the observed MFs for unresolved multiple systems
1271: a discontinuity in the IMF near the BD/VLMS region emerges.
1272: We have also tried to model continuous
1273: single-body IMFs, but we find this hypothesis of continuity to be
1274: inconsistent with the observed MFs given the observational data on the binary
1275: properties of stars and BDs.
1276: We have
1277: shown that the \emph{empirically determined} difference in the binary
1278: properties between BDs/VLMSs on the one hand side and stars on the other,
1279: and the \emph{empirical finding} that stars and BDs rarely pair, implies a
1280: discontinuity in the IMF near the BD/VLMSs mass.
1281:
1282: Thus, the discontinuity
1283: in binary properties, which has already been interpreted to mean two
1284: separate populations \citep{KB03b}, also implies a discontinuity in the
1285: IMF. This relates the probably different formation mechanism more
1286: clearly to the observational evidence.
1287:
1288: We have performed a parameter survey allowing the IMF
1289: parameters $\abd$, $\rpop$, and $\mhibd$ to vary finding that the canonical
1290: stellar IMF ($\alpha_1=1.3$ and $\alpha_2=2.3$) cannot be discarded even for
1291: TA, and that the BD/VLMS discontinuity is required for all solutions.
1292: The discontinuity uncovered in this way, if measured at the classical
1293: BD-star border, is of a similar magnitude
1294: for the stellar clusters studied ($0.17\le\rkbd\le0.30$),
1295: supporting the concept of a universal IMF, which is, by itself,
1296: rather notable.
1297:
1298: We recommend calling these populations \emph{BD-like} and \emph{starlike}
1299: with respect to their formation history. These populations have probably
1300: overlapping mass ranges since there is no physical reason for the upper
1301: mass limit of the BD-like population to match the lower mass limit
1302: of the starlike one. Furthermore, the best-fit models suggest such an
1303: overlap between 0.07~\tmsun\ and about 0.2~\tmsun.
1304: According to this classification through the formation history,
1305: BD-like objects would include VLMSs,
1306: while starlike ones would include massive BDs.
1307:
1308: The overlap region
1309: implies that BD-like pairs can consist
1310: of a VLMS-BD pair, and that a starlike binary can consist of a
1311: stellar primary with a massive BD as a companion.
1312: As can be seen from Table \ref{tabbinfrac}, the star-BD fractions
1313: are similar (except for the dynamically unevolved TA) and
1314: that the star-BD binary fraction, $\fmix$, is about 2\pct--3\pct.
1315: This is somewhat
1316: higher than the value of $<1\pct$ BD companion fraction \citet{GreLin06}
1317: found for nearby stars but still far below the value expected for a single
1318: population model (at least 7\pct\--15\pct\ in \citealt{Ketal03} and
1319: $15\pct\pm4\pct$ in our calculations). Simulations by \citet{BBB03},
1320: \citet{BaBo05}, and \citet{Bate2005}
1321: predict a fraction of about 2\pct\ star-BD binaries
1322: (actually one M dwarf with a BD companion out of 58 stars formed in three
1323: independent calculations).
1324: Note that, in our model, $\fmix$ is a prediction of
1325: the required overlap region and is sensitive to the overlapping range.
1326: As our model for the Pleiades IMF suggests, the overlapping range might
1327: be considerably smaller than that we have found for the other clusters.
1328: Furthermore, we did not fit the lower mass border, $\mlost$, of the
1329: starlike population but simply assumed a value of 0.07~\tmsun\ which
1330: is well in the BD mass regime (and is the major source of star-BD binaries
1331: in our Pleiades model). A slight increase of $\mlost$ by only
1332: 0.01~\tmsun\ would cause the Pleiades star-BD binary fraction to drop
1333: to nearly zero.
1334:
1335: Several authors doubt the existence of two separate populations.
1336: Most recently,
1337: \citet{EiSt07} summarize that the observational community in general
1338: prefers the model of starlike formation for BDs.
1339: They mention the detection of isolated
1340: proto(sub)stellar ``blobs'' in the Ophiuchus B and D clouds, which
1341: may support the theory of starlike formation for BDs.
1342: It remains unclear, though, how many
1343: of these blobs will actually form BDs instead of dissolving from lack
1344: of gravitational binding energy.
1345: \citet{GoWi07} refer to a private communication with {\AA}. Nordlund
1346: stating that the pure turbulence theory predicts about 20,000
1347: transient cores for every actual pre-stellar core of about 0.1~\tmsun.
1348:
1349: We recall that one of the main reasons for the existence of a separate
1350: population is the semi-major axis distribution of BD binaries (Fig.
1351: \ref{sepdist}). But \citet{Luetal07} also mention a wide binary
1352: BD in Ophiuchus with a separation of approximately 300~AU.
1353: Indeed, a small number of wide BD binaries are known.
1354: However, it can be doubted that the occasional discovery
1355: of a wide BD binary may expand the narrow semi-major axis distribution
1356: (Fig. \ref{sepdist}) to a starlike one. The striking evidence posed
1357: by the lack of BD companions to stars is a strong indication for two
1358: populations. It is usually ignored by the community, though.
1359: We note that even if there actually may be some BDs that formed
1360: starlike they are most probably a minority.
1361:
1362: There is also the interpretation of the BD desert being a
1363: ``low-$q$ desert'' rather than an absolute mass-dependent drop in the
1364: companion mass function. \citet{GreLin06} find a low-mass companion
1365: desert of solar-type primary stars between approximately 0.01 and
1366: 0.06~\tmsun. They find this interval to be dependent on the primary
1367: star mass and therefore predict M dwarfs to have BD companions,
1368: and that M dwarfs ought to have a
1369: companion desert between a few Jupiter masses and the low-mass BD
1370: regime. However, this interpretation does not address the different
1371: orbital properties of BDs and stars as well as the different
1372: $q$ distribution.
1373:
1374: \subsection{Implications for the Formation History}
1375: Can the existence of such a discontinuity, i.e. the formation
1376: of two separate populations, be understood theoretically?
1377: Although BDs and stars appear to be distinct populations the formation
1378: of BDs is likely to be connected to star formation.
1379: Bate et al. (2002, 2003) show that BDs
1380: significantly below 0.07~\tmsun\ cannot form in a classical
1381: way since the minimum mass they need for stellar-type formation would
1382: also lead to progressive accretion and growth to stellar mass unless they
1383: are in regions with very low mass infall rates.
1384: But such regions are very rare, because the prevailing densities
1385: and temperatures cannot achieve the required Jeans masses,
1386: as also stressed by \citet{GoWi07}.
1387: For this reason, \citet{AdFa96} expected BDs to be rare.
1388: Only a small fraction of the BDs, especially those at the high-mass end
1389: of the BD regime, may form this way if the surrounding gas has been
1390: consumed by star formation processes just after the proto-BD has reached
1391: the Jeans mass.
1392: To explain the actually higher BD frequency (per star) in recent
1393: surveys the accretion process has to be terminated or
1394: impeded somehow \citep*{PPVBoetal}.
1395:
1396: Also the above-mentioned
1397: differences in the distribution of the semimajor axes between BDs and stars
1398: cannot be explained by a scaled-down star formation process, because
1399: that would imply a continuous variation and a much broader semi-major axis
1400: distribution for BDs and VLMSs that has not been observed \citep{Ketal03}.
1401: While binary stars show a very broad distribution of their semimajor axes
1402: peaking at about 30~AU the semimajor axes of BDs are
1403: distributed around about 5~AU with a sharp truncation at 10--15~AU
1404: (Fig. \ref{sepdist}). No smooth transition region between both regimes
1405: can be recognized \citep{Cloetal03}. In a high angular resolution
1406: survey \citet{LaHoMa07} found that the orbital radius distribution
1407: of binaries with V-K $<$ 6.5 appears to differ significantly
1408: from that of cooler (and thus lower mass) objects, suggesting a sudden
1409: change of the number of binaries wider than 10~AU at about the M5 spectral
1410: type. This is in agreement with our finding of a possible BD-like population
1411: that extends beyond the hydrogen-burning mass limit into the VLMS
1412: regime.
1413:
1414: In a radial-velocity survey of Chamaeleon I, \citet{Joergens2006a} has
1415: found evidence for a rather low binary fraction below 0.1~AU, while most
1416: companions found in that survey orbit their primaries within a few AU.
1417: For this reason an extreme excess of close BD binaries that cannot be
1418: resolved by imaging surveys appears to be unlikely.
1419: A larger binary fraction than about 15\pct\ would thus not be plausible.
1420: \citet{BasRei2006} suggest an upper limit of $26\pct\pm10\pct$
1421: for the BD binary fraction based on their own results
1422: ($11\pct^{+0.07}_{-0.04}$, for separations below 6~AU) and the survey
1423: by Close et al. (2003; $15\pct\pm7\pct$, for separations greater than 2.6~AU)
1424: by simple addition of the results. This is nearly consistent with
1425: a BD binary fraction of 15\pct, since the survey is neither magnitude-
1426: nor volume-limited. However, they admit that their value may be
1427: over-estimated since
1428: the objects with separations between 2.6 and 6~AU are counted twice.
1429: We note further that even a total BD binary fraction of 25\pct\
1430: ($\fbd=0.25$), although outside the error limits of our best-fit models,
1431: would only lead to a minor change in the fitted IMFs.
1432:
1433: It has been argued \citep{BasRei2006} that the lower binary fraction of BDs
1434: is just the extension of a natural trend from G dwarfs to M dwarfs
1435: (Figs. \ref{fbin4} and \ref{fbuni}).
1436: Our contribution has shown that this trend can be understood by the
1437: simple fact that there are many fewer possibilities to form a binary near
1438: the lower mass end than for higher component masses.
1439: The observational data are in better agreement with a minimum mass, $\mmin$,
1440: near the hydrogen-burning mass limit and a low overall binary fraction
1441: of BD-like objects than with an ``all-in-one'' IMF from the lowest mass BDs
1442: to the upper stellar mass limit, as shown in Figure \ref{fbuni}. This
1443: observed trend thus appears as an additional enforcement of the
1444: two-populations model of BDs and stars.
1445:
1446: Given that the conditions for a starlike formation of BDs are very rare
1447: \citep{BBB03}, four alternative formation scenarios for BDs apart
1448: from starlike formation can be identified, namely
1449: \begin{enumerate}
1450: \item Formation of wide star-BD binaries via fragmentation of a
1451: proto- or circumstellar disk and subsequent disruption
1452: by moderately close encounters.
1453: \item Formation of BDs as unfinished stellar embryos ejected from their
1454: birth system.
1455: \item Removal of the accretion envelopes from low-mass protostars via
1456: photoevaporation.
1457: \item Removal of the accretion envelopes due to extremely close stellar
1458: encounters \citep{PriPod95}.
1459: \end{enumerate}
1460: Scenario 4 can be ruled out as the major BD formation mechanism because the
1461: probability of such close encounters,
1462: with required flyby distances typically below 10~AU \citep{KB03b}
1463: for efficient disruption of accretion envelopes (less than a tenth of
1464: those proposed by Thies et al. [2005] for triggered planet formation),
1465: is far too low for such a scenario being a significant
1466: contribution to BD formation.
1467: The photo evaporation model, as studied by \citet{WiZi04},
1468: also cannot be the major mechanism of BD formation \citep{KB03b}. It
1469: predicts a variation of the IMF with the population number and
1470: density of the host
1471: cluster. In dense starburst clusters (young globular clusters)
1472: with a larger number of
1473: O/B stars or even modest clusters such as the ONC with a dozen O/B stars
1474: compared
1475: to TA, a larger fraction of low-mass stars would have halted in
1476: growth. This would
1477: result in a bias towards M dwarfs, since many of them would be
1478: failed K or G dwarfs.
1479: In contrast to this prediction, \citet{Brietal02} and \citet{Ketal03}
1480: show that the IMFs of TA
1481: and ONC are very similar in the mass range 0.1--1~\tmsun,
1482: while globular clusters likewise have a low-mass MF similar
1483: to the standard form \citep{Kr01}.
1484:
1485: \subsection{Embryo Ejection}
1486: \label{ssec:ejection}
1487: \citet{Rei00} and \citet{ReiCla01} introduced the formation of BDs as
1488: ejected stellar embryos as the alternative scenario 2.
1489: If a forming protostar in a newborn
1490: multiple system is ejected due to dynamical instability
1491: its accretion process is terminated
1492: and the object remains in a protostellar state with only a fraction of
1493: the mass compared to a fully developed star. Since the final mass
1494: is physically independent of the hydrogen fusion mass limit
1495: one would not expect the mass range of ejected
1496: embryos to be truncated at the HBL and thus expect an
1497: \emph{overlap region} between these populations.
1498: This fully agrees with the requirement
1499: of having to introduce such an overlap region in order to fit the
1500: observed \simf\ in \S~\ref{sec:method}.
1501:
1502: This model gives some hints to understand the low BD binary fraction as
1503: well as the truncation of the semimajor axis distribution of BDs.
1504: The decay of a young multiple system
1505: of three or more stellar embryos typically leads to the ejection of
1506: single objects but also to the ejection of a small fraction of close
1507: binaries. In order to survive the ejection, the semi-major axis of
1508: such a binary must be significantly smaller (by a factor of about 3)
1509: than the typical orbital separation within the original multiple system.
1510: A similar explanation is that the orbital velocity of the
1511: BD binary components has to be higher than the typical ejection
1512: velocity in order to keep the interaction cross section of the binary with
1513: other system members small. Indeed, the velocity dispersion of BDs in the
1514: embryo-ejection model shown in \citet{KB03b} is
1515: $\lesssim 2\unit{km\,s^{-1}}$ for the majority of the BDs. This is in
1516: good agreement with the Keplerian
1517: orbital velocity of each BD-binary member of about
1518: 1.5--$2\unit{km\,s^{-1}}$ for an equal-mass binary of 0.05--0.08~\tmsun\
1519: and $a\approx 10\unit{AU}$. The majority of BD binaries have smaller
1520: separations, and, consequently, higher orbital velocities and are bound
1521: tighter. This would set the low binding energy cut near
1522: $E_\mathrm{bind}=0.2\unit{pc^{-2}\,Myr^2}$ in Figure \ref{edist}.
1523:
1524: There have been numerical simulations of star formation and dynamics,
1525: e.g. by \citet{BBB03} and \citet{Umetal05}, in which binaries are
1526: produced via ejection that show remarkably similar properties to the
1527: actually observed ones. \citet{Umetal05} describe the formation of
1528: BDs from decaying triple systems. Their simulations predict a semi-major
1529: axis distribution between about 0.2 and 8~AU (see Fig. 8 in their paper),
1530: peaking at 3~AU.
1531: This is slightly shifted towards closer separations compared with
1532: the results by \citet{Cloetal03} but still in agreement with the
1533: observational data. In contrast to this, \citet{GoWi07} doubt
1534: the frequent formation of close BD binaries via ejection, arguing
1535: that hydrogen-burning stars which formed via ejection were almost
1536: always single.
1537:
1538: In further qualitative support of the embryo ejection
1539: model, \citet{Guetal06} describe a deficit by a factor of 2
1540: of BDs near the
1541: highest density regions of TA relative to the BD abundance in
1542: the less dense regions that can
1543: possibly be explained by dynamical ejection and consequently larger
1544: velocity dispersion of stellar embryos, i.e. BDs.
1545: A starlike fragmentation scenario would
1546: result in an opposite trend since the Jeans mass is smaller for higher
1547: densities, thus allowing gas clumps of lower mass to form (sub)stellar
1548: bodies. Contrary to this, \citet{Lu06} did not find any evidence for a
1549: different spatial distribution of BDs and stars in TA.
1550: Recently, \citet{KumSch07} have found that substellar objects in both
1551: the Trapezium and IC~348 are distributed homogeneously within twice the
1552: cluster core radii while the stellar populations display a clustered
1553: distribution.
1554: They conclude that these distributions are best explained with a higher
1555: initial velocity dispersion of BDs, in accordance with \citet{KB03b},
1556: supporting the embryo-ejection model.
1557:
1558: However, the embryo-ejection model has a challenge to reproduce the
1559: high fraction of significant disks around BDs that have been observed in
1560: young clusters. Several studies, e.g. \citet{Nattaetal04},
1561: reveal a considerable number of BDs with an infrared
1562: excess that indicates the presence of warm circum(sub)stellar material.
1563: While these studies do not show the actual mass of these disks because
1564: a small amount of dust in these disks is sufficient to produce these
1565: excesses, \citet*{SJW06} found 25\pct\ of BDs in TA having disks with
1566: radii $>10$~AU and significant masses (larger than 0.4~$M_\mathrm{J}$).
1567: For the remaining 75\pct\ no disks were detectable.
1568: In simulations, e.g. by \cite{BBB03}, such large disks
1569: survive occasionally, but less frequently (about 5\pct)
1570: than suggested by the observations \citep{SJW06}.
1571: However, \citet{SJW06} point out that their results to not rule out the
1572: embryo-ejection scenario and admit that this mechanism may still be relevant
1573: for some BDs. In general, the embryo-ejection model is in agreement with
1574: at least the existence of low-mass circumsubstellar disks up to about 10~AU
1575: \citep{BBB03,Umetal05,PPVGuetal}. \citet{TKT05} also
1576: show that a thin low-mass disk can survive near-parabolic prograde
1577: coplanar encounters above about
1578: three Hill radii with respect to the disk-hosting BD.
1579: This means that a disk with a radius up to about 5~AU (later
1580: viscously evolved to about 10~AU, see \citealt{BBB03}) can survive
1581: a flyby of an equal-mass embryo within about 15~AU while larger
1582: disks or widely separated binaries would be disrupted.
1583:
1584: Furthermore, \citet{BBB03} suggest from their
1585: simulations that the binary fraction via ejection might by as small
1586: as about 5\pct\ (see also \citealt{WhietalPPV}).
1587:
1588: \subsection{Disk Fragmentation and Binary Disruption}
1589: \label{ssec:diskfrag}
1590: The fragmentation of protobinary disks with subsequent disruption
1591: of a star-BD binary is another promising alternative scenario.
1592: A disk can fragment during the accretion process if it reaches a
1593: critical mass above which the disk becomes gravitationally unstable
1594: against small perturbations. Fragmentation may
1595: be triggered by an external perturbation, i.e. infalling gas clumps
1596: or a passing neighbor star. The latter mechanism may also be capable of
1597: triggering fragmentation in relatively low-mass circumstellar disks
1598: resulting in rapid planet formation \citep{TKT05}.
1599: Thus, BD formation via disk fragmentation
1600: likely plays an important role in the early ages of the cluster where the
1601: frequency of massive disks is highest \citep{HLL2001, WhietalPPV}.
1602:
1603: Following the argumentation of \citet{GoWi07}, disk fragmentation may
1604: explain the observed distribution of BD binary separations at least as
1605: well as the embryo-ejection model. In addition, it may explain the
1606: existence of wide star--BD binaries, since a fraction of the initial
1607: wide binaries can survive without being disrupted. Because the likelihood
1608: of disrupting close encounters depends on the mass and density of the
1609: host cluster, one expects a higher fraction of those wide star-BD
1610: binaries in smaller and less dense clusters and associations like TA.
1611: However, further studies and observations are needed
1612: to test this hypothesis.
1613:
1614: \citet{WhiSta06} show that BDs can actually form as widely
1615: separated companions to low-mass stars with mass $m$ at a sufficiently
1616: large disk radius, $r_\mathrm{disk}$,
1617: \begin{equation}
1618: r_\mathrm{disk} \gtrsim 150\unit{AU}\left(\frac{m}{\msun}\right)\,,
1619: \end{equation}
1620: where the disk is cool enough to allow a substellar clump to undergo
1621: gravitational collapse. For a primary star below 0.2~\tmsun\ this
1622: minimum radius therefore becomes less than 90~AU which is in remarkable
1623: agreement with the two wider VLMS binaries found by \citet{Konetal07}.
1624: Such large distances to the primary star allow the formation of BD-BD
1625: binaries as well as the survival of circumsubstellar disks
1626: up to about 10--30~AU, depending on the total mass of the
1627: presubstellar core and the mass of the primary star.
1628: Furthermore, this scenario explains the existence of wide star-BD binaries.
1629: In addition, such a wide star-BD binary can be disrupted by moderately
1630: close encounters of about 100--200~AU (i.e. a distance similar to the
1631: star-BD orbital radius), the disruption of such systems appears to be
1632: likely in contrast to the disruption of accretion envelopes
1633: as required in the already rejected scenario 4.
1634:
1635: \subsection{Summary}
1636: Both the embryo-ejection model and disk fragmentation with
1637: subsequent wide binary disruption explain the above-mentioned
1638: connection between stars and BDs, since BDs start to form like stars before
1639: their growth is terminated due to their separation from their host system
1640: or from lack of surrounding material in the outer parts of a circumstellar
1641: disk. It is obvious that the formation
1642: rate of these embryos is proportional to the total star formation rate.
1643:
1644: For these reasons these formation mechanisms appear to be the most likely
1645: ones for BDs and some VLMSs and BD/VLMS binaries. It cannot be decided
1646: yet which scenario is the dominant mechanism. This may depend on the
1647: size and the density of the star-forming region.
1648: We expect, on the other hand, the classical starlike formation
1649: scenario to be of some importance only for the most massive BDs.
1650:
1651: The possibility of two different \emph{alternative} BD formation
1652: mechanisms (disk fragmentation and embryo ejection) may
1653: lead to another discontinuity in the intermediate mass BD IMF
1654: since both scenarios correspond to different binary fractions
1655: as mentioned above. The currently available data, however, are far from
1656: being sufficient for a verification of this prediction.
1657:
1658: \section{Conclusions}
1659: \label{sec:conclusions}
1660: The different empirical binary properties of BDs and stars strongly
1661: imply the existence of two separate but mutually related populations.
1662: We have shown that if the IMF of BDs and stars is analyzed
1663: under consideration of their binary properties then
1664: there is a discontinuity in the transition
1665: region between the substellar and stellar regime that is quite
1666: independent of the host cluster.
1667: The discontinuity in the IMF near the HBL
1668: is a strong logical implication of the disjunct binary properties and
1669: suggests splitting up the IMF into two components, the BD-like and the
1670: starlike regime. An alternative but equivalent description would be to
1671: view the stellar IMF as a continuous distribution function ranging from
1672: about 0.07 to 150~\tmsun\ \citep{WK04}, and a causally connected but
1673: disjoint distribution of (probably mostly) separated ultra--low-mass
1674: companions and ejected embryos with masses ranging from
1675: 0.01~\tmsun\ to 0.1--0.2~\tmsun.
1676: While the canonical stellar IMF is consistent with
1677: the observed stellar MFs at least for the Trapezium, IC~348, and the
1678: Pleiades, the sub-stellar IMF of at least the Trapezium and TA has a
1679: power law index that is consistent with the canonical value $\abd=0.3$.
1680: Within the error limits, our analysis does not reject the canonical
1681: power law indices for BDs and stars for any cluster.
1682:
1683: The discontinuity is often masked in the observational data due to a
1684: mass overlap of both populations in the BD-VLMS region
1685: as well as the higher apparent masses of unresolved binaries compared to
1686: single objects in the observed \simf.
1687: The discontinuity in the number density near the
1688: HBL is a step of approximately a factor of 3--5
1689: (Table \ref{tabbinfrac}). This implies
1690: a general dependency between both populations and is, as far as we
1691: can tell, consistent with the
1692: scenario of disrupted wide binaries \citep{GoWi07} as well as with the
1693: truncated-star scenario (e.g. as an ejected stellar embryo,
1694: \citealt{ReiCla01}), since the
1695: number of unready stars is directly correlated with the total amount
1696: of star formation in the host cluster.
1697: Both embryo-ejection and wide binary disruption are also consistent
1698: with the properties of close binary BDs.
1699:
1700: Our results (Tab. \ref{tabalpha})
1701: suggest that about one BD is produced per 4--6 formed stars.
1702: This suggests the necessity
1703: of a distinct description of BDs and stars as well as the connection between
1704: these two populations through their formation process.
1705:
1706: \section*{Acknowledgements}
1707: This work was partially supported by the AIfA and partially by DFG grant
1708: KR1635/12-1.
1709: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1710: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1711: \begin{thebibliography}{}
1712:
1713: \bibitem[\protect\citeauthoryear{{Adams} \& {Fatuzzo}}{{Adams} \&
1714: {Fatuzzo}}{1996}]{AdFa96}
1715: {Adams} F.~C., {Fatuzzo} M., 1996, \apj, 464, 256
1716:
1717: \bibitem[\protect\citeauthoryear{{Barrado y Navascu{\'e}s}, {Stauffer} \&
1718: {Jayawardhana}}{{Barrado y Navascu{\'e}s} et~al.}{2004}]{BSJ04}
1719: {Barrado y Navascu{\'e}s} D., {Stauffer} J.~R., {Jayawardhana} R., 2004,
1720: \apj, 614, 386
1721:
1722: \bibitem[\protect\citeauthoryear{{Basri} \& {Reiners}}{{Basri} \&
1723: {Reiners}}{2006}]{BasRei2006}
1724: {Basri} G., {Reiners} A., 2006, \aj, 132, 663
1725:
1726: \bibitem[\protect\citeauthoryear{{Bate}}{{Bate}}{2005}]{Bate2005}
1727: {Bate} M.~R., 2005, \mnras, 363, 363
1728:
1729: \bibitem[\protect\citeauthoryear{{Bate} \& {Bonnell}}{{Bate} \&
1730: {Bonnell}}{2005}]{BaBo05}
1731: {Bate} M.~R., {Bonnell} I.~A., 2005, \mnras, 356, 1201
1732:
1733: \bibitem[\protect\citeauthoryear{{Bate}, {Bonnell} \& {Bromm}}{{Bate}
1734: et~al.}{2002}]{BBB02}
1735: {Bate} M.~R., {Bonnell} I.~A., {Bromm} V., 2002, \mnras, 332, L65
1736:
1737: \bibitem[\protect\citeauthoryear{{Bate}, {Bonnell} \& {Bromm}}{{Bate}
1738: et~al.}{2003}]{BBB03}
1739: {Bate} M.~R., {Bonnell} I.~A., {Bromm} V., 2003, \mnras, 339, 577
1740:
1741: \bibitem[\protect\citeauthoryear{{Bonnell}, {Larson} \& {Zinnecker}}{{Bonnell}
1742: et~al.}{2007}]{PPVBoetal}
1743: {Bonnell} I.~A., {Larson} R.~B., {Zinnecker} H., 2007, in {Reipurth} B.,
1744: {Jewitt} D., {Keil} K., eds, Protostars and Planets V {The Origin of the
1745: Initial Mass Function}.
1746: pp 149--164
1747:
1748: \bibitem[\protect\citeauthoryear{{Bouy}, {Brandner}, {Mart{\'{\i}}n},
1749: {Delfosse}, {Allard} \& {Basri}}{{Bouy} et~al.}{2003}]{Bouyetal03}
1750: {Bouy} H., {Brandner} W., {Mart{\'{\i}}n} E.~L., {Delfosse} X., {Allard}
1751: F., {Basri} G., 2003, \aj, 126, 1526
1752:
1753: \bibitem[\protect\citeauthoryear{{Brice{\~ n}o}, {Luhman}, {Hartmann},
1754: {Stauffer} \& {Kirkpatrick}}{{Brice{\~ n}o} et~al.}{2002}]{Brietal02}
1755: {Brice{\~ n}o} C., {Luhman} K.~L., {Hartmann} L., {Stauffer} J.~R.,
1756: {Kirkpatrick} J.~D., 2002, \apj, 580, 317
1757:
1758: \bibitem[\protect\citeauthoryear{{Burgasser}, {Kirkpatrick}, {Reid}, {Brown},
1759: {Miskey} \& {Gizis}}{{Burgasser} et~al.}{2003}]{Burgetal03}
1760: {Burgasser} A.~J., {Kirkpatrick} J.~D., {Reid} I.~N., {Brown} M.~E.,
1761: {Miskey} C.~L., {Gizis} J.~E., 2003, \apj, 586, 512
1762:
1763: \bibitem[\protect\citeauthoryear{{Burrows}, {Hubbard}, {Saumon} \&
1764: {Lunine}}{{Burrows} et~al.}{1993}]{Buetal93}
1765: {Burrows} A., {Hubbard} W.~B., {Saumon} D., {Lunine} J.~I., 1993, \apj,
1766: 406, 158
1767:
1768: \bibitem[\protect\citeauthoryear{{Chabrier}}{{Chabrier}}{2002}]{Cha02}
1769: {Chabrier} G., 2002, \apj, 567, 304
1770:
1771: \bibitem[\protect\citeauthoryear{{Chabrier}}{{Chabrier}}{2003}]{Cha03}
1772: {Chabrier} G., 2003, \pasp, 115, 763
1773:
1774: \bibitem[\protect\citeauthoryear{{Chabrier} \& {Baraffe}}{{Chabrier} \&
1775: {Baraffe}}{2000}]{ChaBa00}
1776: {Chabrier} G., {Baraffe} I., 2000, \araa, 38, 337
1777:
1778: \bibitem[\protect\citeauthoryear{{Close}, {Siegler}, {Freed} \&
1779: {Biller}}{{Close} et~al.}{2003}]{Cloetal03}
1780: {Close} L.~M., {Siegler} N., {Freed} M., {Biller} B., 2003, \apj, 587,
1781: 407
1782:
1783: \bibitem[\protect\citeauthoryear{{Dobbie}, {Pinfield}, {Jameson} \&
1784: {Hodgkin}}{{Dobbie} et~al.}{2002}]{Doetal02}
1785: {Dobbie} P.~D., {Pinfield} D.~J., {Jameson} R.~F., {Hodgkin} S.~T., 2002,
1786: \mnras, 335, L79
1787:
1788: \bibitem[\protect\citeauthoryear{{Duch{\^e}ne}}{{Duch{\^e}ne}}{1999}]{Duchene9%
1789: 9}
1790: {Duch{\^e}ne} G., 1999, \aap, 341, 547
1791:
1792: \bibitem[\protect\citeauthoryear{{Duquennoy} \& {Mayor}}{{Duquennoy} \&
1793: {Mayor}}{1991}]{DuqMay91}
1794: {Duquennoy} A., {Mayor} M., 1991, \aap, 248, 485
1795:
1796: \bibitem[\protect\citeauthoryear{{Eisl{\"o}ffel} \&
1797: {Steinacker}}{{Eisl{\"o}ffel} \& {Steinacker}}{2007}]{EiSt07}
1798: {Eisl{\"o}ffel} J., {Steinacker} J., 2007, ArXiv Astrophysics e-prints
1799:
1800: \bibitem[\protect\citeauthoryear{{Fischer} \& {Marcy}}{{Fischer} \&
1801: {Marcy}}{1992}]{FisMar92}
1802: {Fischer} D.~A., {Marcy} G.~W., 1992, \apj, 396, 178
1803:
1804: \bibitem[\protect\citeauthoryear{{Goodwin} \& {Kroupa}}{{Goodwin} \&
1805: {Kroupa}}{2005}]{GoKr05}
1806: {Goodwin} S.~P., {Kroupa} P., 2005, \aap, 439, 565
1807:
1808: \bibitem[\protect\citeauthoryear{{Goodwin}, {Kroupa}, {Goodman} \&
1809: {Burkert}}{{Goodwin} et~al.}{2007}]{PPVGoodwinetal}
1810: {Goodwin} S.~P., {Kroupa} P., {Goodman} A., {Burkert} A., 2007, in
1811: {Reipurth} B., {Jewitt} D., {Keil} K., eds, Protostars and Planets V {The
1812: Fragmentation of Cores and the Initial Binary Population}.
1813: pp 133--147
1814:
1815: \bibitem[\protect\citeauthoryear{{Goodwin} \& {Whitworth}}{{Goodwin} \&
1816: {Whitworth}}{2007}]{GoWi07}
1817: {Goodwin} S.~P., {Whitworth} A., 2007, \aap, 466, 943
1818:
1819: \bibitem[\protect\citeauthoryear{{Grether} \& {Lineweaver}}{{Grether} \&
1820: {Lineweaver}}{2006}]{GreLin06}
1821: {Grether} D., {Lineweaver} C.~H., 2006, \apj, 640, 1051
1822:
1823: \bibitem[\protect\citeauthoryear{{G{\"u}del}, {Padgett} \&
1824: {Dougados}}{{G{\"u}del} et~al.}{2007}]{PPVGuetal}
1825: {G{\"u}del} M., {Padgett} D.~L., {Dougados} C., 2007, in {Reipurth} B.,
1826: {Jewitt} D., {Keil} K., eds, Protostars and Planets V {The Taurus
1827: Molecular Cloud: Multiwavelength Surveys with XMM-Newton, the Spitzer Space
1828: Telescope, and the Canada-France-Hawaii Telescope}.
1829: pp 329--344
1830:
1831: \bibitem[\protect\citeauthoryear{{Guenther} \& {Wuchterl}}{{Guenther} \&
1832: {Wuchterl}}{2003}]{GuWu03}
1833: {Guenther} E.~W., {Wuchterl} G., 2003, \aap, 401, 677
1834:
1835: \bibitem[\protect\citeauthoryear{{Guieu}, {Dougados}, {Monin}, {Magnier} \&
1836: {Mart{\'{\i}}n}}{{Guieu} et~al.}{2006}]{Guetal06}
1837: {Guieu} S., {Dougados} C., {Monin} J.-L., {Magnier} E., {Mart{\'{\i}}n}
1838: E.~L., 2006, \aap, 446, 485
1839:
1840: \bibitem[\protect\citeauthoryear{{Haisch}, {Lada} \& {Lada}}{{Haisch}
1841: et~al.}{2001}]{HLL2001}
1842: {Haisch} K.~E., {Lada} E.~A., {Lada} C.~J., 2001, \apjl, 553, L153
1843:
1844: \bibitem[\protect\citeauthoryear{{Hillenbrand}, {Carpenter} \&
1845: {Feigelson}}{{Hillenbrand} et~al.}{2001}]{HiCaFe01}
1846: {Hillenbrand} L.~A., {Carpenter} J.~M., {Feigelson} E.~D., 2001, in
1847: {Montmerle} T., {Andr{\'e}} P., eds, ASP Conf. Ser. 243: From Darkness to
1848: Light: Origin and Evolution of Young Stellar Clusters {The Orion Star-Forming
1849: Region}.
1850: pp 439--+
1851:
1852: \bibitem[\protect\citeauthoryear{{Jeffries} \& {Maxted}}{{Jeffries} \&
1853: {Maxted}}{2005}]{JefMax05}
1854: {Jeffries} R.~D., {Maxted} P.~F.~L., 2005, Astronomische Nachrichten, 326,
1855: 944
1856:
1857: \bibitem[\protect\citeauthoryear{{Joergens}}{{Joergens}}{2006}]{Joergens2006a}
1858: {Joergens} V., 2006, \aap, 446, 1165
1859:
1860: \bibitem[\protect\citeauthoryear{{Kenyon}, {Jeffries}, {Naylor}, {Oliveira} \&
1861: {Maxted}}{{Kenyon} et~al.}{2005}]{Kenetal05}
1862: {Kenyon} M.~J., {Jeffries} R.~D., {Naylor} T., {Oliveira} J.~M., {Maxted}
1863: P.~F.~L., 2005, \mnras, 356, 89
1864:
1865: \bibitem[\protect\citeauthoryear{{Konopacky}, {Ghez}, {Rice} \&
1866: {Duchene}}{{Konopacky} et~al.}{2007}]{Konetal07}
1867: {Konopacky} Q.~M., {Ghez} A.~M., {Rice} E.~L., {Duchene} G., 2007, ArXiv
1868: Astrophysics e-prints
1869:
1870: \bibitem[\protect\citeauthoryear{{Kraus}, {White} \& {Hillenbrand}}{{Kraus}
1871: et~al.}{2006}]{Krausetal06}
1872: {Kraus} A.~L., {White} R.~J., {Hillenbrand} L.~A., 2006, \apj, 649, 306
1873:
1874: \bibitem[\protect\citeauthoryear{{Kroupa}}{{Kroupa}}{2001}]{Kr01}
1875: {Kroupa} P., 2001, \mnras, 322, 231
1876:
1877: \bibitem[\protect\citeauthoryear{{Kroupa} \& {Bouvier}}{{Kroupa} \&
1878: {Bouvier}}{2003}]{KB03b}
1879: {Kroupa} P., {Bouvier} J., 2003, \mnras, 346, 369
1880:
1881: \bibitem[\protect\citeauthoryear{{Kroupa}, {Bouvier}, {Duch{\^ e}ne} \&
1882: {Moraux}}{{Kroupa} et~al.}{2003}]{Ketal03}
1883: {Kroupa} P., {Bouvier} J., {Duch{\^ e}ne} G., {Moraux} E., 2003, \mnras,
1884: 346, 354
1885:
1886: \bibitem[\protect\citeauthoryear{{Kroupa}, {Gilmore} \& {Tout}}{{Kroupa}
1887: et~al.}{1991}]{KGT91}
1888: {Kroupa} P., {Gilmore} G., {Tout} C.~A., 1991, \mnras, 251, 293
1889:
1890: \bibitem[\protect\citeauthoryear{{Kroupa}, {Tout} \& {Gilmore}}{{Kroupa}
1891: et~al.}{1993}]{KTG93}
1892: {Kroupa} P., {Tout} C.~A., {Gilmore} G., 1993, \mnras, 262, 545
1893:
1894: \bibitem[\protect\citeauthoryear{{Kumar} \& {Schmeja}}{{Kumar} \&
1895: {Schmeja}}{2007}]{KumSch07}
1896: {Kumar} M.~S.~N., {Schmeja} S., 2007, \aap, 471, L33
1897:
1898: \bibitem[\protect\citeauthoryear{{Lada}}{{Lada}}{2006}]{Lada2006}
1899: {Lada} C.~J., 2006, \apjl, 640, L63
1900:
1901: \bibitem[\protect\citeauthoryear{{Lada} \& {Lada}}{{Lada} \&
1902: {Lada}}{2003}]{LaLa03}
1903: {Lada} C.~J., {Lada} E.~A., 2003, \araa, 41, 57
1904:
1905: \bibitem[\protect\citeauthoryear{{Law}, {Hodgkin} \& {Mackay}}{{Law}
1906: et~al.}{2007}]{LaHoMa07}
1907: {Law} N.~M., {Hodgkin} S.~T., {Mackay} C.~D., 2007, ArXiv e-prints, {704
1908: (submitted to MNRAS)}
1909:
1910: \bibitem[\protect\citeauthoryear{{Luhman}}{{Luhman}}{2004a}]{Lu04b}
1911: {Luhman} K.~L., 2004a, \apj, 617, 1216
1912:
1913: \bibitem[\protect\citeauthoryear{{Luhman}}{{Luhman}}{2004b}]{Lu04a}
1914: {Luhman} K.~L., 2004b, \apj, 614, 398
1915:
1916: \bibitem[\protect\citeauthoryear{{Luhman}}{{Luhman}}{2006}]{Lu06}
1917: {Luhman} K.~L., 2006, \apj, 645, 676
1918:
1919: \bibitem[\protect\citeauthoryear{{Luhman}, {Allers}, {Jaffe}, {Cushing},
1920: {Williams}, {Slesnick} \& {Vacca}}{{Luhman} et~al.}{2007}]{Luetal07}
1921: {Luhman} K.~L., {Allers} K.~N., {Jaffe} D.~T., {Cushing} M.~C., {Williams}
1922: K.~A., {Slesnick} C.~L., {Vacca} W.~D., 2007, \apj, 659, 1629
1923:
1924: \bibitem[\protect\citeauthoryear{{Luhman}, {Brice{\~n}o}, {Stauffer},
1925: {Hartmann}, {Barrado y Navascu{\'e}s} \& {Caldwell}}{{Luhman}
1926: et~al.}{2003}]{Luetal03a}
1927: {Luhman} K.~L., {Brice{\~n}o} C., {Stauffer} J.~R., {Hartmann} L., {Barrado
1928: y Navascu{\'e}s} D., {Caldwell} N., 2003, \apj, 590, 348
1929:
1930: \bibitem[\protect\citeauthoryear{{Luhman}, {Stauffer}, {Muench}, {Rieke},
1931: {Lada}, {Bouvier} \& {Lada}}{{Luhman} et~al.}{2003}]{Luetal03b}
1932: {Luhman} K.~L., {Stauffer} J.~R., {Muench} A.~A., {Rieke} G.~H., {Lada}
1933: E.~A., {Bouvier} J., {Lada} C.~J., 2003, \apj, 593, 1093
1934:
1935: \bibitem[\protect\citeauthoryear{Lupton}{Lupton}{1993}]{STP}
1936: Lupton R., 1993, Statistics in Theory and Practice, first edn.
1937: Princeton University Press, Princeton, New Jersey
1938:
1939: \bibitem[\protect\citeauthoryear{{Malkov} \& {Zinnecker}}{{Malkov} \&
1940: {Zinnecker}}{2001}]{MalZin01}
1941: {Malkov} O., {Zinnecker} H., 2001, \mnras, 321, 149
1942:
1943: \bibitem[\protect\citeauthoryear{{Mart{\'{\i}}n}, {Barrado y Navascu{\'e}s},
1944: {Baraffe}, {Bouy} \& {Dahm}}{{Mart{\'{\i}}n} et~al.}{2003}]{Maetal03}
1945: {Mart{\'{\i}}n} E.~L., {Barrado y Navascu{\'e}s} D., {Baraffe} I., {Bouy}
1946: H., {Dahm} S., 2003, \apj, 594, 525
1947:
1948: \bibitem[\protect\citeauthoryear{Matsumoto \& Nishimura}{Matsumoto \&
1949: Nishimura}{1998}]{mersennetwister}
1950: Matsumoto M., Nishimura T., 1998, ACM Trans. Model. Comput. Simul, 8, 3
1951:
1952: \bibitem[\protect\citeauthoryear{{McCarthy}, {Zuckerman} \&
1953: {Becklin}}{{McCarthy} et~al.}{2003}]{2003IAUS..211..279M}
1954: {McCarthy} C., {Zuckerman} B., {Becklin} E.~E., 2003, in {Mart{\'{\i}}n}
1955: E., ed., IAU Symposium {There is a Brown Dwarf Desert of Companions Orbiting
1956: Stars between 75 and 1000 AU}.
1957: pp 279--+
1958:
1959: \bibitem[\protect\citeauthoryear{{Metchev} \& {Hillenbrand}}{{Metchev} \&
1960: {Hillenbrand}}{2005}]{MetHil05}
1961: {Metchev} S., {Hillenbrand} L., 2005, Memorie della Societa Astronomica
1962: Italiana, 76, 404
1963:
1964: \bibitem[\protect\citeauthoryear{{Moraux}, {Bouvier}, {Stauffer} \&
1965: {Cuillandre}}{{Moraux} et~al.}{2003}]{Moetal03}
1966: {Moraux} E., {Bouvier} J., {Stauffer} J.~R., {Cuillandre} J.-C., 2003,
1967: \aap, 400, 891
1968:
1969: \bibitem[\protect\citeauthoryear{{Moraux}, {Kroupa} \& {Bouvier}}{{Moraux}
1970: et~al.}{2004}]{Moetal04}
1971: {Moraux} E., {Kroupa} P., {Bouvier} J., 2004, \aap, 426, 75
1972:
1973: \bibitem[\protect\citeauthoryear{{Muench}, {Lada}, {Lada} \& {Alves}}{{Muench}
1974: et~al.}{2002}]{Muetal02}
1975: {Muench} A.~A., {Lada} E.~A., {Lada} C.~J., {Alves} J., 2002, \apj, 573,
1976: 366
1977:
1978: \bibitem[\protect\citeauthoryear{{Natta}, {Testi}, {Muzerolle}, {Randich},
1979: {Comer{\'o}n} \& {Persi}}{{Natta} et~al.}{2004}]{Nattaetal04}
1980: {Natta} A., {Testi} L., {Muzerolle} J., {Randich} S., {Comer{\'o}n} F.,
1981: {Persi} P., 2004, \aap, 424, 603
1982:
1983: \bibitem[\protect\citeauthoryear{{Padoan} \& {Nordlund}}{{Padoan} \&
1984: {Nordlund}}{2002}]{PaNo02}
1985: {Padoan} P., {Nordlund} {\AA}., 2002, \apj, 576, 870
1986:
1987: \bibitem[\protect\citeauthoryear{{Padoan} \& {Nordlund}}{{Padoan} \&
1988: {Nordlund}}{2004}]{PaNo04}
1989: {Padoan} P., {Nordlund} {\AA}., 2004, \apj, 617, 559
1990:
1991: \bibitem[\protect\citeauthoryear{Press, Teukolsky, Vetterling \&
1992: Flannery}{Press et~al.}{1992}]{numres}
1993: Press W.~H., Teukolsky S.~A., Vetterling W.~T., Flannery B.~P., 1992,
1994: Numerical Recipes in Fortran 77, second edn.
1995: Cambridge University Press, Cambridge
1996:
1997: \bibitem[\protect\citeauthoryear{{Price} \& {Podsiadlowski}}{{Price} \&
1998: {Podsiadlowski}}{1995}]{PriPod95}
1999: {Price} N.~M., {Podsiadlowski} P., 1995, \mnras, 273, 1041
2000:
2001: \bibitem[\protect\citeauthoryear{{Reid}, {Gizis} \& {Hawley}}{{Reid}
2002: et~al.}{2002}]{RGH02}
2003: {Reid} I.~N., {Gizis} J.~E., {Hawley} S.~L., 2002, \aj, 124, 2721
2004:
2005: \bibitem[\protect\citeauthoryear{{Reid}, {Lewitus}, {Allen}, {Cruz} \&
2006: {Burgasser}}{{Reid} et~al.}{2006}]{Reidetal06}
2007: {Reid} I.~N., {Lewitus} E., {Allen} P.~R., {Cruz} K.~L., {Burgasser}
2008: A.~J., 2006, \aj, 132, 891
2009:
2010: \bibitem[\protect\citeauthoryear{{Reipurth}}{{Reipurth}}{2000}]{Rei00}
2011: {Reipurth} B., 2000, \aj, 120, 3177
2012:
2013: \bibitem[\protect\citeauthoryear{{Reipurth} \& {Clarke}}{{Reipurth} \&
2014: {Clarke}}{2001}]{ReiCla01}
2015: {Reipurth} B., {Clarke} C., 2001, \aj, 122, 432
2016:
2017: \bibitem[\protect\citeauthoryear{{Salpeter}}{{Salpeter}}{1955}]{Salpeter1955}
2018: {Salpeter} E.~E., 1955, \apj, 121, 161
2019:
2020: \bibitem[\protect\citeauthoryear{{Scholz}, {Jayawardhana} \& {Wood}}{{Scholz}
2021: et~al.}{2006}]{SJW06}
2022: {Scholz} A., {Jayawardhana} R., {Wood} K., 2006, \apj, 645, 1498
2023:
2024: \bibitem[\protect\citeauthoryear{{Sterzik} \& {Durisen}}{{Sterzik} \&
2025: {Durisen}}{2003}]{SteDur03}
2026: {Sterzik} M.~F., {Durisen} R.~H., 2003, \aap, 400, 1031
2027:
2028: \bibitem[\protect\citeauthoryear{{Thies}, {Kroupa} \& {Theis}}{{Thies}
2029: et~al.}{2005}]{TKT05}
2030: {Thies} I., {Kroupa} P., {Theis} C., 2005, \mnras, 364, 961
2031:
2032: \bibitem[\protect\citeauthoryear{{Umbreit}, {Burkert}, {Henning}, {Mikkola} \&
2033: {Spurzem}}{{Umbreit} et~al.}{2005}]{Umetal05}
2034: {Umbreit} S., {Burkert} A., {Henning} T., {Mikkola} S., {Spurzem} R.,
2035: 2005, \apj, 623, 940
2036:
2037: \bibitem[\protect\citeauthoryear{{Weidner} \& {Kroupa}}{{Weidner} \&
2038: {Kroupa}}{2004}]{WK04}
2039: {Weidner} C., {Kroupa} P., 2004, \mnras, 348, 187
2040:
2041: \bibitem[\protect\citeauthoryear{{Weidner} \& {Kroupa}}{{Weidner} \&
2042: {Kroupa}}{2006}]{WK06}
2043: {Weidner} C., {Kroupa} P., 2006, \mnras, 365, 1333
2044:
2045: \bibitem[\protect\citeauthoryear{{Whitworth}, {Bate}, {Nordlund}, {Reipurth} \&
2046: {Zinnecker}}{{Whitworth} et~al.}{2007}]{WhietalPPV}
2047: {Whitworth} A., {Bate} M.~R., {Nordlund} {\AA}., {Reipurth} B.,
2048: {Zinnecker} H., 2007, in {Reipurth} B., {Jewitt} D., {Keil} K., eds,
2049: Protostars and Planets V {The Formation of Brown Dwarfs: Theory}.
2050: pp 459--476
2051:
2052: \bibitem[\protect\citeauthoryear{{Whitworth} \& {Stamatellos}}{{Whitworth} \&
2053: {Stamatellos}}{2006}]{WhiSta06}
2054: {Whitworth} A.~P., {Stamatellos} D., 2006, \aap, 458, 817
2055:
2056: \bibitem[\protect\citeauthoryear{{Whitworth} \& {Zinnecker}}{{Whitworth} \&
2057: {Zinnecker}}{2004}]{WiZi04}
2058: {Whitworth} A.~P., {Zinnecker} H., 2004, \aap, 427, 299
2059:
2060: \end{thebibliography}
2061: \end{document}
2062: