1: % \documentclass[12pt,preprint]{aastex}
2: \documentclass{emulateapj}
3: \usepackage{times}
4: %\usepackage[english]{babel}
5: %\usepackage{natbib}
6: %\usepackage{graphicx}
7:
8: % \topmargin=-10mm
9: % \headsep=0mm
10: % \headheight=0mm
11: % \oddsidemargin=-10mm
12: % \evensidemargin=-10mm
13: % \textwidth=18.7cm
14: % \textheight=26.0cm
15: % \columnsep=7mm
16:
17: %-----------------------------------------------------------------------
18: % Authors' Macros
19: %-----------------------------------------------------------------------
20:
21: \newcommand{\avk}[1]{{\bf [#1 -- AVK.]}}
22: \newcommand{\tl}[1]{{\bf [#1 -- TL.]}}
23:
24: \newcommand{\be}{\begin{equation}}
25: \newcommand{\ee}{\end{equation}}
26: \newcommand{\bd}{\begin{displaymath}}
27: \newcommand{\ed}{\end{displaymath}}
28: \newcommand{\bea}{\begin{eqnarray}}
29: \newcommand{\eea}{\end{eqnarray}}
30: \newcommand{\etal}{et al.}
31:
32: \newcommand{\AAp}{\aap}
33: \newcommand{\ApJ}{\apj}
34: \newcommand{\AJ}{\aj}
35: \newcommand{\JGR}{\jgr}
36: \newcommand{\MNRAS}{\mnras}
37: \newcommand{\AdvSR}{\apspr}
38: \newcommand{\PSS}{\planss}
39:
40:
41: \newcommand{\mum}{\,\mu\hbox{m}}
42: \newcommand{\mm}{\,\hbox{mm}}
43: \newcommand{\cm}{\,\hbox{cm}}
44: \newcommand{\m}{\,\hbox{m}}
45: \newcommand{\km}{\,\hbox{km}}
46: \newcommand{\AU}{\,\hbox{AU}}
47: \newcommand{\second}{\,\hbox{s}}
48: \newcommand{\yr}{\,\hbox{yr}}
49: \newcommand{\g}{\,\hbox{g}}
50:
51: \newcommand{\mrm}[1]{\mathrm{#1}}
52: \newcommand{\sbs}[1]{\ensuremath{_\mathrm{#1}}} %subscript
53: \newcommand{\sps}[1]{\ensuremath{^\mathrm{#1}}} %superscript
54: \newcommand{\vect}[1]{\ensuremath{\mathbf{#1}}}
55: \newcommand{\ovl}[1]{\ensuremath{\overline{#1}}}
56: \renewcommand{\d}{\ensuremath{\mathrm{d}}}
57: \newcommand{\textmu}[1]{\micron}
58: % \renewcommand{\clearpage}{{}}
59:
60:
61: \shortauthors{L{\"o}hne et al.}
62: \shorttitle{Decay of Debris Disks}
63: \slugcomment{Submitted to ApJ 1 Aug 2007, accepted 23 Oct 2007.}
64:
65: \begin{document}
66:
67: %------------------------------------------------------------------
68:
69: \title{Long-Term Collisional Evolution of Debris Disks}
70: \author{Torsten L{\"o}hne and Alexander V. Krivov}
71: \affil{Astrophysikalisches Institut und Universit{\"a}tssternwarte,
72: Friedrich Schiller University Jena,
73: Schillerg{\"a}{\ss}chen~ 2--3, 07745 Jena, Germany;
74: tloehne@astro.uni-jena.de, krivov@astro.uni-jena.de}
75:
76: \author{Jens Rodmann}
77: \affil{SCI-SA, Research and Scientific Support Department of ESA,
78: ESTEC, 2200 AG Noordwijk, The Netherlands;
79: jrodmann@rssd.esa.int
80: }
81:
82: \begin{abstract}
83: Infrared surveys indicate that the dust content in debris disks gradually
84: declines with stellar age. We simulated the long-term collisional depletion
85: of debris disks around solar-type (G2~V) stars with our collisional code.
86: The numerical results were supplemented by, and interpreted through, a new
87: analytic model. General scaling rules for the disk evolution are suggested.
88: The timescale of the collisional evolution is inversely proportional to the
89: initial disk mass and scales with radial distance as $r^{4.3}$ and with
90: eccentricities of planetesimals as $e^{-2.3}$. Further, we show that at actual
91: ages of debris disks between 10~Myr and 10~Gyr, the decay laws of the dust mass
92: and the total disk mass are different. The reason is that the
93: collisional lifetime of planetesimals is size-dependent. At any moment,
94: there exists a transitional size, which separates larger objects that still
95: retain the ``primordial'' size distribution set in the growth phase from smaller
96: objects whose size distribution is already set by disruptive collisions. The
97: dust mass and its decay rate evolve as that transition affects objects of
98: ever-larger sizes. Under standard assumptions, the dust mass, fractional
99: luminosity, and thermal fluxes all decrease as $t^\xi$ with $\xi = -0.3$...$-0.4$.
100: Specific decay laws of the total disk mass and the dust mass, including the
101: value of $\xi$, largely depend on a few model parameters, such as the critical
102: fragmentation energy as a function of size, the primordial size distribution
103: of largest planetesimals, as well as the characteristic eccentricity and
104: inclination of their orbits. With standard material prescriptions and a
105: distribution of disk masses and extents, a synthetic population of disks
106: generated with our analytic model agrees quite well with the observed
107: Spitzer/MIPS statistics of 24 and 70 \micron\ fluxes and colors versus age.
108: \end{abstract}
109: \keywords{circumstellar matter --- planetary systems: formation.}
110:
111: %===========================================================================
112: \section{INTRODUCTION}
113: %===========================================================================
114:
115: Since the IRAS discovery of the excess infrared emission around Vega by
116: %the Infrared Astronomical Satellite (IRAS)
117: \citet{aumann-et-al-1984}, subsequent infrared surveys with ISO, Spitzer and other instruments
118: have shown the Vega phenomenon to be common for main-sequence stars.
119: The observed excess is attributed to second-generation circumstellar dust,
120: produced in a collisional cascade from planetesimals and comets down to smallest
121: grains that are blown away by the stellar radiation. While the bulk of such a
122: debris disk's mass is hidden in invisible parent bodies,
123: the observed luminosity is dominated by small particles at dust sizes.
124: Hence the studies of dust emission offer a natural tool to gain insight into
125: the properties of planetesimal populations as well as planets that may shape them and,
126: ultimately, into the evolutionary history of circumstellar planetary systems.
127:
128: In recent years, various photometric surveys of hundreds of nearby stars have
129: been conducted
130: with the Spitzer Space Telescope.
131: These are the GTO survey of FGK stars
132: \citep{beichman-et-al-2005,bryden-et-al-2006,beichman-et-al-2006b},
133: the FEPS Legacy project \citep{meyer-et-al-2004,kim-et-al-2005}, the A star GTO
134: programs
135: \citep{rieke-et-al-2005,su-et-al-2006}, the young cluster programs
136: \citep{gorlova-et-al-2006},
137: and others.
138: These observations were done mostly at
139: 24 and 70~\textmu{m} with the MIPS photometer, but also between 5 and
140: 40~\textmu{m} with the IRS spectrometer
141: \citep{jura-et-al-2004,chen-et-al-2006}.
142: Based on these studies,
143: about 15\% of mature solar-type (F0--K0) stars have been found to harbor cold
144: debris disks at 70~\textmu{m}. For cooler stars, the fraction drops to 0\%--4\%
145: \citep{beichman-et-al-2006b}.
146: For earlier spectral types, the proportion increases to about 33\%
147: \citep{su-et-al-2006}.
148: At 24~\textmu{m}, the fraction
149: of systems with detected excess stays similar for A~stars, but appreciably decreases for FGK ones.
150: Similar results in the sub-millimeter range are expected to become available soon from a survey
151: with SCUBA and SCUBA2 on JCMT \citep{matthews-et-al-2007}.
152: Preliminary SCUBA results for M dwarfs suggest, in particular, that the
153: proportion of debris disks might actually be higher than suggested by Spitzer
154: \citep{lestrade-et-al-2006}.
155:
156: All authors point out a decay of the observed infrared excesses with systems'
157: age.
158: However, the values reported for the slope of the decay, assuming a
159: power-law dependence $t^{-\alpha}$, span a wide range. \citet{greaves-wyatt-2003}
160: suggest $\alpha \la 0.5$, \citet{liu-et-al-2004} give $0.5 < \alpha < 1.0$,
161: \citet{spangler-et-al-2001} report $\alpha \approx 1.8$, and
162: \citet{greaves-2005} and \citet{moor-et-al-2006} derive $\alpha \approx 1.0$.
163: Fits of the upper envelope of the distribution of luminosities over the age yield
164: $\alpha \approx 1.0$ as well \citep{rieke-et-al-2005}. Besides, the dust fractional
165: luminosity exhibits a large dispersion at any given age.
166:
167: In an attempt to gain theoretical understanding of the observed evolution,
168: \citet{dominik-decin-2003} assumed that equally-sized ``comets''
169: produce dust through a cascade of
170: subsequent collisions among ever-smaller objects. If this dust is removed by
171: the same mechanism, the steady-state amount of dust in such a
172: system is proportional to the number of comets. This results in an
173: $M/M_0 \approx \tau/t$
174: dependence for the amount of dust and for the number of
175: comets or the total mass of the disk.
176: Under the assumption of a steady state, this result is valid even for more complex
177: systems with continuous size distributions from planetesimals to dust.
178: Tenuous disks, where the lifetime of dust grains is not
179: limited by collisions but by transport processes like the Poynting-Robertson
180: drag \citep{artymowicz-1997,krivov-et-al-2000,wyatt-2005}, follow $M \propto t^{-2}$
181: rather than $M \propto t^{-1}$.
182:
183: More recently, \citet{wyatt-et-al-2007a} lifted the most severe simplifying
184: assumption
185: of the Dominik-Decin model, that of equal-sized parent bodies, and included them
186: into the collisional cascade. A debris disk they consider is no longer a two-component
187: system ``comets + dust''. Instead, it is a population of solids with a continuous
188: size distribution, from planetesimals down to dust.
189: A key parameter of the description by \citet{dominik-decin-2003} is the
190: collisional lifetime of comets, $\tau$.
191: \citet{wyatt-et-al-2007a} replaced it with the lifetime of the largest
192: planetesimals
193: and worked out the dependencies on this parameter in great detail.
194: Since the collisional timescale is inversely proportional to the amount
195: of material, $\tau \propto 1/M_0$, the asymptotic disk mass becomes independent of its initial mass.
196: Only dynamical quantities, i.e. the disk's radial position and extent, the orbiting objects' eccentricities
197: and inclinations, and material properties, i.e. the critical specific energy and the
198: disruption threshold, as well as the type of the central star determine the very-long-term
199: evolution.
200:
201: Still, there are two important simplifications made in the model by
202: \citet{wyatt-et-al-2007a}:
203: (i) the disk is assumed to be in collisional equilibrium at all sizes, from dust up to the
204: largest planetesimals
205: and (ii)
206: the minimum specific energy needed to disrupt colliding objects
207: is independent of their size. As a consequence of (i) and (ii),
208: the size distribution of solids is a single power-law.
209: To check how reasonable these assumptions are, realistic simulations of the disks with
210: collisional codes are necessary
211: \citep[e.g.,][]{thebault-et-al-2003,krivov-et-al-2005,krivov-et-al-2006,thebault-augereau-2007}.
212:
213: The aim of this paper is two-fold. First, we follow the evolution of debris disks
214: with our elaborate numerical code \citep{krivov-et-al-2005,krivov-et-al-2006}
215: to check the existing analytic models and the assumptions (i) and (ii) they are based upon.
216: Second, in order to make these numerical results easier to use,
217: we develop a new analytic model for the evolution of disk mass and dust mass
218: that relaxes both assumptions (i) and (ii) above.
219:
220: Section~\ref{secNumerics}
221: summarizes the basic ideas and assumptions and describes our numerical model
222: and the runs of the collisional code. In Section~\ref{secScalings} the numerical results
223: are presented and dependences of the collisional timescale on the disk mass, distance to the
224: star, and mean eccentricity of parent bodies are derived.
225: In section~\ref{secAnalytics}, the analytic model
226: for the evolution of disk mass and dust mass is developed.
227: Section~\ref{secEvolLuminosity} analyzes the evolution of dust luminosities.
228: In Section~\ref{secObservations}, we use the analytic model to synthesize representative
229: populations of debris disks and compare them with statistics of debris disks
230: derived from the Spitzer surveys.
231: A summary is given and conclusions are drawn in Section~\ref{secConclusions}.
232: \pagebreak
233:
234: %===========================================================================
235: \section{NUMERICAL MODEL AND DESCRIPTION OF RUNS}
236: %===========================================================================
237: \label{secNumerics}
238: \subsection{Basic Approach}
239:
240: For all numerical runs in this paper, we use a C++-based collisional
241: code (ACE, Analysis of Collisional Evolution).
242: The code numerically solves the Boltzmann-Smoluchowski kinetic equation
243: to evolve a disk of solids in a broad range of sizes (from
244: sub-micrometers to about a hundred of kilometers),
245: orbiting a primary in nearly-Keplerian orbits
246: (gravity + direct radiation pressure + drag forces) and
247: experiencing disruptive collisions.
248: Collisions are simulated with available material- and size-dependent scaling
249: laws for fragmentation and dispersal in both strength and gravity regime.
250: The current version implements a 3-dimensional kinetic model, with masses,
251: semi-major axes, and eccentricities as phase space variables.
252: This approach automatically enables a study of the simultaneous evolution of
253: mass, spatial, and velocity distribution of particles.
254: The code is fast enough to easily follow the evolution of a debris disk over
255: Gyr timescales.
256: A detailed description of our approach, its numerical implementation, and astrophysical
257: applications can be found in our previous papers
258: \citep{krivov-et-al-2000,krivov-et-al-2005,krivov-et-al-2006}.
259:
260: \subsection{Disruption Threshold and Critical Specific Energy}
261: \label{secQD}
262:
263: An object is said to be disrupted in a collision, if the largest fragment is at
264: most
265: half as massive as the original object. If the impactor's relative velocity is so
266: high that the ratio of impact energy and target mass exceeds the
267: target's critical specific energy, $Q\sbs{D}^*$, the target (and the impactor)
268: are disrupted.
269: For small objects, this binding energy is dominated by material
270: strength,
271: and for larger objects, self-gravity takes
272: over.
273: Both regimes are usually described by a sum of two power laws
274: \citep[][Sect.~5.1, and references therein]{krivov-et-al-2005}
275: \begin{equation}
276: Q\sbs{D}^* = A\sbs{s} \left(\frac{s}{1\:\mathrm{m}}\right)^{3b\sbs{s}} +
277: A\sbs{g} \left(\frac{s}{1\:\mathrm{km}}\right)^{3b\sbs{g}},
278: \label{eqQD}
279: \end{equation}
280: where ``s'' and ``g'' stand for strength and gravity, respectively.
281: The reported values of the coefficients $A\sbs{s}$ and $A\sbs{g}$ vary by more
282: than one order of magnitude, and we took $A\sbs{s} = A\sbs{g} = 5 \times 10^6$~erg/g,
283: in agreement with the reference case for
284: basalt given by \citet{benz-asphaug-1999}. The
285: exponents are $3b\sbs{s} = -0.3$ and $3b\sbs{g} = 1.5$ (corresponding to $-0.1$ and $0.5$
286: in the mass scaling). With these
287: parameters, the two power-law components contribute equally at $s \approx
288: 316$~m, and the lowest binding energy, the minimum $Q\sbs{D}^*$, is reached
289: at $s \approx 129$~m. The influence of the choice of parameters on the
290: resulting evolution is discussed in Sect.~\ref{secAnalytics}.
291:
292: For computational reasons, we refrained from including a treatment of cratering
293: collisions in the runs.
294: Note that these were not taken into account in previous studies of the long-term
295: evolution of debris disks \citep[e.g.][]{dominik-decin-2003,wyatt-et-al-2007a}
296: either.
297: \citet{thebault-et-al-2003} and
298: \citet{thebault-augereau-2007}, who focused on shorter time spans, did include this
299: non-disruptive type of collisions that lead to the continuous
300: erosion of a target by small impacting projectiles. They found the effect to
301: be dominant for particles in between 100~\textmu{m} and 1~cm for the case
302: of the inner $\beta$ Pictoris disk, while big, kilometer-sized objects in the
303: gravity regime are mainly lost to disruptive collisions \citep[see Table 4 in][]{thebault-et-al-2003}.
304: However,
305: including cratering
306: can lower the lifetime of large objects, especially when relative
307: velocities are low and disruptive collisions are rare.
308: Another caveat is that cratering collisions alter the shape of the wavy size distribution
309: at the lower end \citep{thebault-augereau-2007}, which
310: affects the observable thermal fluxes.
311:
312: \subsection{Collisional Outcomes}
313:
314: The distribution of sizes and the velocities of fragments in an
315: individual (catastrophic) collision has been subject to studies for decades.
316: Laboratory work was done on high-velocity impacts on scales of millimeters and
317: centimeters
318: \citep[e.g.,][]{fujiwara-et-al-1977,fujiwara-1986,davis-ryan-1990}.
319: Statistics on the mass distributions of observed asteroidal families
320: and geometrical considerations
321: \citep{paolicchi-et-al-1996,tanga-et-al-1999,tedesco-et-al-2005} as well as
322: gravito-hydrodynamic simulations of fragmentation and reaccumulation
323: \citep{michel-et-al-2002} cover the range of larger, kilometer-sized
324: bodies.
325: On small scales, the resulting size distributions show a strong dependence
326: on impact velocity and seem to indicate a turn in the power law at
327: fragment sizes around $\approx 1$~mm (or $\approx 1$\%\ of the size of the
328: used targets). The slope for objects above that size is steeper than the one
329: for smaller objects \citep{davis-ryan-1990}. However, \citet{thebault-et-al-2003}
330: found that the ratio of these two slopes and the size at which
331: the slope changes influence simulation results only slightly.
332: On kilometer and larger scales, the fragmentation is influenced by gravitational
333: reaccumulation of relatively small fragments onto bigger ones. Hence, bigger
334: fragments ($\sim 100$km) will be overabundant, and conversely, smaller fragments
335: ($\sim 1$km) underabundant, compared to the underlying distribution without gravity.
336: The slopes of the size distribution $n(s)\propto s^{-p}$ of kilometer-sized objects
337: are poorly known. A wide range from $p=3.5$ up to $p=9.0$ has been reported.
338: These deviations in the kilometer regime are most probably the severest caveat
339: of the power-law approximation, because they are independent of
340: the actual material and caused only by gravity.
341: Nevertheless, we assume that fragments follow a single
342: power-law distribution $n\sbs{frag} (s) \propto s^{-3.5}$, expecting the
343: influence on the final collisional steady state to be only moderate.
344:
345: \subsection{Commons for All Runs}
346: \label{secCommons}
347:
348: All disk models presented here are set up around a star of solar mass and
349: luminosity. Parameters of the central star affect the disk evolution in
350: various ways. They determine the size limit for grain's blowout by
351: radiation pressure and orbital velocities at a given
352: distance, thereby altering impact velocities and rates.
353: For late-type stars, strong stellar winds may affect the dust dynamics
354: \citep{augereau-beust-2006,strubbe-chiang-2006}.
355: On the observational
356: side, dust temperatures and brightnesses are influenced.
357: Here, we focus on the scalings for a fixed spectral type (G2V),
358: and not on scalings between different types.
359:
360: The disks themselves all share the same material properties and shapes.
361: We adopt the material, described by a bulk density $\rho=2.5$~g/cm$^3$, the
362: radiation pressure efficiency of astronomical silicate
363: \citep{laor-draine-1993}, and a
364: critical fragmentation energy as specified in Sect.~\ref{secQD}.
365: We switched off the Poynting-Robertson effect, which is unimportant for debris disks
366: under study, as well as stellar wind drag, which plays only a minor role
367: around G-type stars.
368: The fragments produced in an individual collision are distributed according
369: to a single power law,
370: $\d N\propto s^{-3.5} \d s \propto m^{-11/6} \d m$.
371: A biggest fragment size is assumed to scale with
372: specific impact energy to the power of $1.24$ \citep[for details,
373: see][]{krivov-et-al-2006}.
374: The initial mass distribution is given
375: by $\d N \propto m^{-q}$, with $q = 1.87$, a value that accounts for the
376: modification of the classical Dohnanyi's (\citeyear{dohnanyi-1969})
377: $q=1.833$ through the size dependence of material
378: strength \citep[see, e.g.,][]{durda-dermott-1997}. The particle masses range
379: from $4.2\times 10^{-15}$~g, corresponding to a radius of 74~nm, to
380: $4.2\times 10^{21}$~g, corresponding to 74~km. The stepping between the 60 mass
381: bins is logarithmic with a factor of $\approx 4$ between neighboring bins.
382: The initial radial profile of the particle density was given by a
383: slope of the normal optical depth of $-1.0$.
384: The initial total mass of each disk was set to
385: 1~$M_\oplus$ (earth mass).
386:
387: \subsection{Specifics of Individual Runs}
388:
389: % \clearpage
390: \notetoeditor{The table should use only one column in the final version.}
391: \begin{deluxetable}{rcc}
392: \tablecaption{Description of numerical runs.\label{tab_runs}}
393: \tablewidth{0pt}
394: \tablehead{\colhead{Run} & \colhead{Distance [AU]} & e\sbs{max}}
395: \startdata
396: \multicolumn{3}{c}{\em Nominal runs}\\
397: ii-0.3 & 7.5--15 & 0.3 \\
398: i-0.3 & 15--30 & 0.3 \\
399: o-0.3 & 30--60 & 0.3 \\
400: oo-0.3 & 60--120 & 0.3 \\
401: \multicolumn{3}{c}{\em Additional runs}\\
402: i-0.1 & 15--30 & 0.1 \\
403: i-0.2 & 15--30 & 0.2 \\
404: i-0.4 & 15--30 & 0.4 \\
405: \enddata
406: \end{deluxetable}
407: % \clearpage
408:
409: We have made four ``nominal'' runs, each of which corresponds to a certain radial
410: part of the disk between 7.5 and 120 AU from the star (Table~\ref{tab_runs}).
411: In these runs we assumed initial eccentricities of planetesimals
412: to be uniformly distributed between $e\sbs{min} = 0.0$
413: and $e\sbs{max} = 0.3$, spanning three bins centered at 0.05, 0.15 and 0.25.
414: In addition, three runs with altered maximum eccentricity of 0.1,
415: 0.2, and 0.4 were made for the $15$--$30\AU$ ring.
416: In all the runs, we assumed that orbital inclinations are distributed between
417: $I\sbs{min} = e\sbs{min}/2$ and $I\sbs{max} = e\sbs{max}/2$
418: in accordance with the energy equipartition relation $I = e/2$.
419:
420: %===========================================================================
421: \section{NUMERICAL RESULTS AND SCALING LAWS}
422: \label{secScalings}
423: %===========================================================================
424:
425: \subsection{Evolution of Disks of Different Masses}
426:
427: A debris disk is said to be in a quasi-steady state or quasi-equilibrium
428: if the amounts of particles with different sizes on different orbits,
429: while changing with time (therefore ``quasi''), stay constant relative to each other.
430: For brevity, we will often omit ``quasi'' and use simply
431: ``steady state'' or ``equilibrium''.
432: To express the condition of a quasi-steady state
433: formally, we can introduce a phase space, in which
434: a dynamical state of each particle is characterized by a vector {\bf p}.
435: That vector may be composed, for instance, of coordinates
436: and velocity components. Alternatively, {\bf p} may represent the
437: set of orbital elements of the object.
438: Let $n({\bf p}, s, t)$ be the number of objects with radii in $[s, s + \d s]$
439: at phase space ``positions'' $[{\bf p}, {\bf p} + \d {\bf p}]$ that the disk contains
440: at the time instant $t$.
441: The assumption of a quasi-steady state can now be expressed as
442: \begin{equation}
443: n({\bf p}, s, t) = \tilde{n}({\bf p}, s) \: f(t) .
444: \label{ntilde}
445: \end{equation}
446: The total disk mass,
447: \begin{equation}
448: M\sbs{disk} (t) = \int\int\nolimits n({\bf p}, s, t) \d {\bf p} \d s,
449: \label{eqMDef}
450: \end{equation}
451: can be rewritten as
452: \begin{equation}
453: M\sbs{disk} (t) = f(t) \int\int\nolimits \tilde{n}({\bf p}, s) \d {\bf p} \d s
454: \end{equation}
455: or, setting $f(0) = 1$,
456: \begin{equation}
457: M\sbs{disk} (t) = f(t) M_0 \label{eqMdisk},
458: \end{equation}
459: where $M_0$ is the initial disk mass.
460: As long as objects are both created and lost in two-particle collisions,
461: their gain and loss rates are given by
462: \begin{eqnarray}
463: \dot{n} ({\bf p}, s, t) &=&
464: \int\int\int\int\nolimits
465: \left[G({\bf p}, s, {\bf p}_1, s_1, {\bf p}_2, s_2)
466: \right.
467: \nonumber\\
468: &&
469: \left.
470: -
471: L({\bf p}_1, s_1, {\bf p}_2, s_2)
472: \delta({\bf p} - {\bf p}_1)
473: \delta(s - s_1)
474: \right]
475: \nonumber\\&\times&
476: \tilde{n}({\bf p}_1, s_1) \: f(t) \:
477: \tilde{n}({\bf p}_2, s_2) \: f(t)
478: \nonumber\\&\times&
479: \d {\bf p}_1 \d s_1 \d {\bf p}_2 \d s_2,
480: \label{eqndot}
481: \end{eqnarray}
482: where
483: the function $G({\bf p}, s, {\bf p}_1, s_1, {\bf p}_2, s_2)$
484: describes the gain in population $\bf p$, $s$ due to collisions
485: between ${\bf p}_1$, $s_1$ and ${\bf p}_2$, $s_2$
486: and the function $L({\bf p}_1, s_1, {\bf p}_2, s_2)$
487: accounts for the loss in population ${\bf p}_1$, $s_1$
488: in collisions with ${\bf p}_2$, $s_2$.
489: The disk mass changes at a rate
490: \begin{equation}
491: \dot{M}\sbs{disk} (t) = \int\int\nolimits \dot{n}({\bf p}, s, t)
492: \d {\bf p} \d s \label{eqMdot1}
493: \end{equation}
494: or
495: \begin{equation}
496: \dot{M}\sbs{disk} (t) = \dot{f}(t) \int\int\nolimits \tilde{n}({\bf p}, s)
497: \d {\bf p} \d s \label{eqMdot2}.
498: \end{equation}
499: From Eqs.~(\ref{eqndot}) and (\ref{eqMdot1}), we find that
500: $\dot{M}\sbs{disk}(t) \propto f^2(t)$, while Eq.~(\ref{eqMdot2}) suggests
501: $\dot{M}\sbs{disk}(t) \propto \dot{f}(t)$. Hence, $\dot{f}(t)\propto f^2(t)$.
502: Integration yields
503: \begin{equation}
504: f = \frac{1}{1+t/\tau} .
505: \label{dd}
506: \end{equation}
507: Using Eq.~(\ref{eqMdisk}) we obtain
508: \begin{equation}
509: M\sbs{disk} (t) = \frac{M_0}{1+t/\tau} \label{eqMOverTime}
510: \end{equation}
511: and
512: \begin{equation}
513: \dot{M} \sbs{disk} (t) = - C M\sbs{disk}^2, \label{eqCdef}
514: \end{equation}
515: where
516: $1/C = M_0 \cdot \tau$, i.e. the product of the initial mass and a
517: characteristic time.
518: This relation is invariant under the transformation
519: $(t,M\sbs{disk})\rightarrow(t\cdot x, M\sbs{disk}/x)$, even if $C$ is not constant.
520: Therefore, the mass scale of a system under collisional evolution is
521: inversely proportional to its timescale. For example, doubling the initial total mass
522: halves the collisional lifetime of the system.
523: All curves in the $M\sbs{disk}(t)$ plots can be shifted along lines of equal $t \cdot
524: M\sbs{disk}$.
525:
526: \citet{dominik-decin-2003} used this approach and equated the characteristic
527: time $\tau$ with the collisional lifetime of their ``comets''.
528: At the initial phase $t\ll\tau$, Eq. (\ref{eqMOverTime}) gives
529: \begin{equation}
530: M\sbs{disk}(t) \approx M_0 \left( 1 - t/\tau \right).
531: \label{eqMOverTimeApprox}
532: \end{equation}
533: If the system is old enough so that $t\gg\tau$, the total mass will be
534: just proportional to $t^{-1}$.
535: Particles whose lifetimes are independent of the total mass are
536: exempt from the asymptotic one-over-$t$ behavior. Examples would be the
537: $\beta$-meteoroids that are blown out and small particles in disks tenuous
538: enough for the Poynting-Robertson effect to be more
539: efficient than collisions. The total mass of such particles is
540: $\propto t^{-2}$ \citep{dominik-decin-2003}.
541:
542: As we have shown, for the systems that undergo a steady-state collisional evolution,
543: the factor $C$ in Eq.~(\ref{eqCdef}) (or $\tau$) should be constant.
544: To check this, we evaluated $C = -\dot M\sbs{disk}/M\sbs{disk}^2$ for every
545: two subsequent time steps of the numerical runs.
546: The results are given in Fig. \ref{figCOverTime}.
547:
548: % \clearpage
549: \begin{figure}[t!]
550: \includegraphics{f1.eps}\\
551: \caption{The coefficient $C$ from Eq. (\ref{eqCdef}) as a function of time for four
552: nominal runs. The total disk mass and
553: time in the runs are scaled according to $M\sbs{disk}\propto t^{-1}$
554: to compensate for the difference in dynamical timescale.
555: Note that the near-constancy of $C$ at the beginning of the evolution is an
556: artefact of the double-logarithmic plotting. The double-linear inset shows
557: that the decrease of $C$ is {\em fastest} at earlier times.
558: \label{figCOverTime}}
559: \end{figure}
560: % \clearpage
561:
562: Instead of being constant at later times, $C$ decreases, roughly following a
563: power law $C\propto t^{-2/3\ldots-4/5}$.
564: The explanation is simple: the systems did not reach an equilibrium
565: where $t\gg\tau$ or at least $t\approx\tau$ during their lifetime.
566: The evolution of the total mass in Fig.~\ref{figMOverTime} demonstrates that as
567: well.
568:
569: % \clearpage
570: \begin{figure}[t!]
571: \includegraphics{f2.eps}\\
572: \caption{The evolution of the total mass in the four nominal runs.
573: Again, the plateau at the beginning of the evolution is an artefact of the
574: logarithmic plotting of time. In fact, the mass decay is strongest at the
575: very beginning (see inset and Eqs.~(\ref{eqMOverTime}), (\ref{eqMOverTimeApprox})).
576: \label{figMOverTime}}
577: \end{figure}
578: % \clearpage
579:
580: \subsection{Dependence on Distance from the Star}
581:
582: Rings of identical mass but at different distances have different
583: collisional timescales. The comparison in Fig. \ref{figCOverTime} shows that
584: doubling the distance requires a 20-fold increase in disk mass to have the
585: same timescale. This corresponds to a power-law dependence
586: \be
587: C \propto r^{-4.3}.\label{eqScalingR}
588: \ee
589: In a thorough analytic approach based on a Dohnanyi-type collisional cascade,
590: \citet{wyatt-et-al-2007a} came up with
591: $C \propto r^{-13/3}$, which is in good agreement with our numerical result.
592: This index is made up of three contributions. First, the density in the rings
593: drops with $r^{-3}$ as their circumference, height, and width increase
594: linearly. Second, the relative velocities have an $r^{-1/2}$ dependence.
595: Third, these impact velocities affect the minimum required mass for a projectile to be
596: disruptive and thereby the total number of such projectiles. That gives
597: another $r^{1-q}$, where $q$ is the slope in the
598: appropriate mass distribution, e.g. $q = 11/6$ for the classical Dohnanyi case.
599: See Sect.~\ref{secCollTime} for details.
600:
601: \subsection{Dependence on Eccentricities of Parent Bodies}
602: \label{secEcc}
603:
604: % \clearpage
605: \begin{figure}[t!]
606: \includegraphics{f3a.eps}\\
607: \includegraphics{f3b.eps}\\
608: \caption{The influence of the average eccentricity of planetesimals
609: on the timescale of disk's collisional evolution.
610: \textsl{Top:} the evolution of the parameter $C$ from
611: Eq.~(\ref{eqCdef}) for four different runs (i-0.1, \ldots, i-0.4).
612: \textsl{Bottom:}
613: four initial $C$ values versus average eccentricity $e = (e\sbs{max} +
614: e\sbs{min})/2$ (pluses) together with the $C \propto e^{9/4}$ fit for those runs
615: (line) and the same for runs with a narrower range of eccentricities,
616: as described in Sect.~\ref{secEcc} (crosses).
617: \label{figCOverEMax}}
618: \end{figure}
619: % \clearpage
620:
621: The intrinsic collisional probability of planetesimals is nearly independent
622: of their eccentricities, as long as they are not too high \citep[see, e.g.][]{krivov-et-al-2006}.
623: Nevertheless, eccentricities determine impact velocities and, through that, the
624: minimum size of a disruptive projectile. Therefore, higher planetesimal eccentricities
625: imply a larger rate of catastrophic collisions and thus a faster collisional
626: evolution.
627: To quantify the dependence, we have made runs with maximum eccentricities
628: of 0.1, 0.2, 0.3, and 0.4 (Table~\ref{tab_runs}) and determined the
629: values of $C$.
630: The results suggest a power law $C \propto e\sbs{max}^{9/4}$ as shown in
631: Fig.~\ref{figCOverEMax}.
632:
633: This result comes as a surprise. \citet{wyatt-et-al-2007a} derive $C \propto
634: e^{5/3}$.
635: The same scaling is inherited by our analytic model, see
636: Eq.~(\ref{eqDependencies}) below.
637: Since this discrepancy can be either due to an incompleteness of the analytic
638: approach or to a non-linear relation between the maximum and the effective
639: eccentricity, we tried to rule out the latter case by
640: performing additional runs with $e$ confined to narrow bins of
641: width $0.1$, centered at $0.05$, $0.15$, $0.25$, and $0.35$.
642: These runs can be well described by the same power law,
643: $C \propto e^{9/4}$ (Fig.~\ref{figCOverEMax}).
644: Therefore, the analytic model fails to reproduce this particular dependence.
645: Nevertheless, it correctly describes many others, as the next sections will
646: show.
647:
648: %===========================================================================
649: \section{ANALYTIC MODEL FOR EVOLUTION OF DISK MASS AND DUST MASS}
650: \label{secAnalytics}
651: %===========================================================================
652:
653: \subsection{Size and Mass Distributions}
654:
655: In what follows, we will analyze size or mass distributions of objects.
656: Different authors use distributions of different physical quantities
657: (number, cross section, mass) with different arguments (particles size or mass)
658: and of different type (differential, cumulative, per size decade, etc.).
659: A standard choice is to use a differential size distribution,
660: $n(s)$, that gives the number of particles per unit size interval:
661: \begin{equation}
662: n(s) \equiv \int\nolimits n({\bf p}, s) \d {\bf p} ,
663: \end{equation}
664: or a differential mass distribution, $n(m)$, that gives the number of particles
665: per unit mass interval.
666: Instead of $n$, it is often convenient to use the mass-per-size-decade
667: distribution,
668: \begin{equation}
669: \frac{\d M\sbs{disk}}{\d \log_{10} s} = \ln (10) \: s\:m(s)\:n(s) .
670: \end{equation}
671: In contrast to $n(s)$, this quantity tells us directly, objects in which size range
672: contribute the most to the mass of the system.
673: Therefore, we will use it when plotting size or mass distributions.
674: %Therefore, we will use it for diagrams.
675:
676: In the case of a power-law size distribution,
677: $n(s) \d s \propto s^{2-3q} \d s$
678: is the number of objects with sizes $[s, s+\d s]$
679: and $n(m) \d m \propto m^{-q} \d m$
680: is the number of objects with masses $[m, m+\d m]$.
681: The mass per size decade is $\propto s^{6-3q} \propto m^{2-q}$.
682: When $q < 2$, the total mass is determined by large bodies,
683: whereas the cross section is dominated by small particles as long as $q > 5/3$.
684:
685: \subsection{Three-Slope Distribution}
686:
687: \label{sec3slope}
688:
689: % \clearpage
690: \begin{figure}[t!]
691: \includegraphics{f4a.eps}\\
692: \includegraphics{f4b.eps}\\
693: \caption{
694: Results of the ii-0.3 run.
695: {\em Top:}
696: Time evolution of mass in individual mass bins, from the
697: largest bodies of $74$~km in radius to the smallest, $74$~nm in
698: radius. The mass ratio between adjacent bins is 4.
699: Each solid line corresponds to one individual bin and gives
700: the mass contained in that bin (see the right axis) as a function of
701: time.
702: The left axis can be used to find the line that corresponds to a
703: given object size.
704: The thick dashed curve corresponds to $\approx 1$~mm radius, i.e. to
705: the largest solids still treated as dust. %$m_d$.
706: The thick dotted curve, which goes roughly through the turning points
707: of the curves, is the transition size $s\sbs{t} (t)$; see
708: Eq.~(\ref{eqEquiMass}).
709: {\em Bottom:}
710: Size/mass distribution at four specific instants of time
711: shown in the top panel with vertical lines:
712: initially,
713: after $5\times 10^5$ years when $s\sbs{t}$ has reached $s\sbs{b}$,
714: and after $5\times 10^7$ and $5\times 10^9$~yr when significant
715: dust depletion has already occurred.
716: \label{figMassDistribution}}
717: \end{figure}
718: % \clearpage
719:
720: The combination of material strength at smaller
721: sizes and self-gravity at larger ones, with a turnover at around 100~m,
722: causes the size distribution in a collisionally evolving system to strongly deviate
723: from a single-slope power law,
724: especially for object sizes of around 1~km.
725: This is illustrated by Fig.~\ref{figMassDistribution}
726: that shows how a disk evolves from the first-guess power
727: law to a more realistic size distribution.
728: The speed of this evolution is determined by the collisional timescales of
729: populations of different-sized particles
730: in the disk. Populations of smaller particles with sufficiently short lifetimes
731: consist mostly of fragments of disruption of larger bodies. They will have
732: reached collisional equilibrium with each other soon, according to their production rate by
733: populations with longer lifetimes. Those latter populations of bigger
734: particles will still be on their way to a steady state. As time goes by,
735: more and more long-lived populations will undergo the transition from
736: primordial to reprocessed material.
737:
738: As this transitional mass moves towards larger objects with time,
739: the smaller particles follow to a new ``intermediate steady state''.
740: The lower panel of Fig.~\ref{figMassDistribution} shows the development of the
741: characteristic wavy shape in the size distribution
742: \citep[e.g.,][]{campo-bagatin-et-al-1994,thebault-et-al-2003,krivov-et-al-2006}
743: at the small-size end near the blowout limit due to radiation pressure.
744: Once established, this shape remains constant. Only the absolute level
745: changes because this distribution at smaller sizes acts as the trail of the
746: distribution at larger sizes.
747: In the upper panel of Fig.~\ref{figMassDistribution}, the number
748: of smaller particles is constant for some time and then
749: goes down,
750: as soon as the distribution in the gravity regime starts to deviate
751: from its primordial one.
752:
753: % \clearpage
754: \begin{figure}[t]
755: \includegraphics{f5.eps}\\
756: \caption{Schematic plot of the three regimes in the mass distribution
757: and its time evolution.
758: The mass $s\sbs{t}$ divides second generation material in collisional
759: equilibrium ($s < s\sbs{t}$) from primordial material ($s > s\sbs{t}$),
760: while $s\sbs{b}$ divides the material strength regime ($s < s\sbs{b}$) from
761: the gravity regime ($s > s\sbs{b}$).
762: \label{figMassDistScheme}}
763: \end{figure}
764: % \clearpage
765: These arguments suggest that an overall size distribution $n(s)$ can be
766: approximated by a combination of
767: three power laws (Fig. \ref{figMassDistScheme}).
768: For particles large enough to be only barely affected by collisions at time $t$,
769: we assume $n$ to follow $s^{2-3q\sbs{p}}$.
770: Here, $q\sbs{p}$ is the ``primordial'' slope determined by the processes in which these
771: planetesimals have formed.
772: Small particles that are in quasi-steady state are separated from bigger
773: primordial objects by a transition zone which we characterize by a
774: time-dependent size $s\sbs{t}(t)$.
775: To distinguish between the strength and gravity regimes,
776: we introduce two more power laws and assume the mass distribution
777: to follow $n\propto s^{2-3q\sbs{g}}$ for gravity-dominated quasi-steady state
778: and $n\propto s^{2-3q\sbs{s}}$ for strength-dominated quasi-steady state.
779: The two regimes are separated by an object size $s\sbs{b}$, which we
780: will call breaking radius. Thus, the waviness is neglected, but
781: the effect of a size-dependent $Q\sbs{D}^*$ is kept.
782:
783: The resulting size distribution is given by
784: \begin{equation}
785: n(s) = n\sbs{max}\left(\frac{s\sbs{max}}{s}\right)^{3q\sbs{p}-2} \label{eqDist0}
786: \end{equation}
787: for $s\sbs{t} \leq s < s\sbs{max}$,
788: \begin{equation}
789: n(s) = n\sbs{max}\left(\frac{s\sbs{max}}{s\sbs{t}}\right)^{3q\sbs{p}-2}
790: \left(\frac{s\sbs{t}}{s}\right)^{3q\sbs{g}-2}
791: \end{equation}
792: for $s\sbs{b} \leq s < s\sbs{t}$, and
793: \begin{equation}
794: n(s) = n\sbs{max}\left(\frac{s\sbs{max}}{s\sbs{t}}\right)^{3q\sbs{p} - 2}
795: \left(\frac{s\sbs{t}}{s\sbs{b}}\right)^{3q\sbs{g} - 2}
796: \left(\frac{s\sbs{b}}{s}\right)^{3q\sbs{s} - 2}\label{eqnm}
797: \end{equation}
798: for $s\sbs{min} < s < s\sbs{b}$,
799: where $n\sbs{max} \equiv n(s\sbs{max})$,
800: with $s\sbs{max}$ being the size of the largest planetesimals.
801: From this distribution, two important quantities can be derived. One is the
802: total {\em disk} mass,
803: \begin{equation}
804: M\sbs{disk} = \int_{s\sbs{min}}^{s\sbs{max}}\limits n(s)
805: \frac{4}{3}\pi\rho s^3 \d s,
806: \label{eqTotalMass}
807: \end{equation}
808: and the other is {\em dust} mass (that determines the infrared luminosity
809: and therefore provides a link to observations),
810: \begin{equation}
811: M\sbs{dust} = \int_{s\sbs{min}}^{s\sbs{d}}\limits n(s)
812: \frac{4}{3}\pi\rho s^3 \d s,
813: \label{eqDustMass}
814: \end{equation}
815: where $s\sbs{min} \leq s\sbs{d} < s\sbs{b}$.
816:
817: \subsection{Collisional Lifetimes of Planetesimals}
818: \label{secCollTime}
819:
820: As seen from Eqs.~(\ref{eqDist0})--(\ref{eqDustMass}),
821: the evolution of $M\sbs{disk}$ and $M\sbs{dust}$ is controlled by
822: $n\sbs{max} (t)$ and $s\sbs{t} (t)$.
823:
824: We start with $n\sbs{max}$
825: and assume, according to Eqs.~(\ref{ntilde}) and~(\ref{dd}):
826: \begin{equation}
827: n\sbs{max}(t) = \frac{n\sbs{max}(0)}{1+t/\tau\sbs{max}},
828: \label{DD}
829: \end{equation}
830: where $\tau\sbs{max}$ is the collisional lifetime of these largest bodies.
831: Equation~(\ref{DD}) closely reproduces the disk evolution as soon as the whole system has
832: reached the quasi-steady state
833: at all sizes or, in other words, as soon as
834: $s\sbs{t}(t)$ has reached $s\sbs{max}$.
835:
836: The second quantity that we need, $s\sbs{t} (t)$, could easily be obtained
837: by inverting the function $\tau(s)$, the collisional lifetime of planetesimals
838: of a given size $s$. To obtain $\tau(s)$, we begin with the lifetime
839: of the largest objects in a disk.
840: Assuming that $q > 5/3$,
841: \citet[][their Eq.~12]{wyatt-et-al-2007a} approximated it as
842: \begin{eqnarray}
843: \tau\sbs{max} &=& \frac{4\pi}{\sigma\sbs{tot}}\cdot
844: \left(\frac{s\sbs{max}}{s\sbs{min}}\right)^{3q\sbs{p}-5}
845: \!\!\!\!\!\!
846: \cdot \frac{r^{5/2}\d r}{\left(\mathsf{G} M_*\right)^{1/2}}
847: \nonumber\\
848: &\times& \frac{I}{f(e,I) G(q, s)}.
849: \label{eqtc}
850: \end{eqnarray}
851: where $e$ and $I$ are the effective orbital eccentricities and
852: inclinations, $\sigma\sbs{tot}$ is the initial cross sectional area of the disk
853: material, $\mathsf{G}$ the gravitational constant, $r$ the radial distance of
854: the ring of parent bodies, and $\d r$ its width. The slope $q$ in their
855: single-power-law approach corresponds to the primordial slope $q\sbs{p}$ in our
856: nomenclature. The functions $f$ and $G$ are given by
857: \begin{eqnarray}
858: f(e,I) &=& \sqrt{\frac{5}{4}e^2 + I^2},\\
859: G(q,s) &=& \left[ X\sbs{c}(s)^{5-3q}
860: - \left(\frac{s\sbs{max}}{s}\right)^{5-3q} \right]\nonumber\\
861: &+& 2\frac{q-5/3}{q-4/3} \left[ X\sbs{c}(s)^{4-3q}
862: - \left(\frac{s\sbs{max}}{s}\right)^{4-3q} \right]
863: \nonumber\\
864: &+& \frac{q-5/3}{q-1}
865: \left[ X\sbs{c}(s)^{3-3q}
866: - \left(\frac{s\sbs{max}}{s}\right)^{3-3q} \right],\label{eqGFull}
867: \end{eqnarray}
868: with
869: \begin{eqnarray}
870: X\sbs{c}(s) &=& \left(
871: \frac{2Q\sbs{D}^*(s) \; rf(e,I)^{-2}}{\mathsf{G}M_*}
872: \right)^{1/3}.\label{eqXcFull}
873: \end{eqnarray}
874: While $f(e,I)$ describes the dependence of the impact velocities on
875: eccentricities and inclinations, the functions $G$ and $X\sbs{c}$ characterize
876: the disruption of planetesimals by smaller projectiles.
877: Namely, $X\sbs{c}(s)$ is the minimum size ratio between the smallest disruptive projectile
878: and the target, and $G(q,s)$ is the number of disruptive projectiles.
879:
880: We need the lifetime of objects of an arbitrary size, $\tau(s < s\sbs{max})$.
881: To derive it, we can simply substitute $s\sbs{max}$ by $s$ in
882: Eq.~(\ref{eqtc}), obtaining
883: \begin{eqnarray}
884: \tau(s) &=& \frac{4\pi}{\sigma\sbs{tot}}\cdot
885: \left(\frac{s}{s\sbs{min}}\right)^{3q\sbs{p}-5}
886: \!\!\!\!\!\!
887: \cdot \frac{r^{5/2}\d r}{\left(\mathsf{G} M_*\right)^{1/2}}
888: \nonumber\\
889: &\times& \frac{I}{f G(q\sbs{p},s)}.
890: \label{eqtausig}
891: \end{eqnarray}
892:
893: In order to replace the dependence on the initial cross
894: sectional area of objects, $\sigma\sbs{tot}$, with their initial total mass,
895: $M_0$, we need to derive both quantities from the initial size distribution in
896: Eq.~(\ref{eqDist0}).
897: The area is given by
898: \begin{equation}
899: \sigma\sbs{tot} = n\sbs{max}(0) \cdot \frac{\pi s\sbs{max}^3}{3q\sbs{p}-5}
900: \left[\left(\frac{s\sbs{max}}{s\sbs{min}} \right)^{3q\sbs{p}-5} - 1\right].
901: \end{equation}
902: Since it is dominated by $s\sbs{min}$ for $q\sbs{p} > 5/3$, we obtain
903: \begin{equation}
904: \sigma\sbs{tot} = n\sbs{max}(0) \cdot \frac{\pi s\sbs{max}^3}{3q\sbs{p}-5}
905: \left(\frac{s\sbs{max}}{s\sbs{min}} \right)^{3q\sbs{p}-5}.\label{eqSigma}
906: \end{equation}
907: The initial total disk mass is
908: \begin{equation}
909: M_0 = n\sbs{max}(0) \cdot \frac{4\pi\rho
910: s\sbs{max}^4}{3(6-3q\sbs{p})}
911: \left[
912: 1 - \left(\frac{s\sbs{min}}{s\sbs{max}} \right)^{6-3q\sbs{p}}
913: \right].
914: \label{eqM0}
915: \end{equation}
916: For $q\sbs{p} < 2$, it is dominated by $s\sbs{max}$. However,
917: since a primordial slope $q\sbs{p} \geq 2$ is not unrealistic (see
918: Sect.~\ref{secParameters}) we refrain from using a further approximation.
919: Then, the area and the mass are related through
920: \begin{eqnarray}
921: \sigma\sbs{tot} &=& M_0 \cdot
922: \frac{3(2-q\sbs{p})}{4(q\sbs{p} - 5/3)}\cdot s\sbs{max}^{-1} \cdot
923: \left(\frac{s\sbs{max}}{s\sbs{min}}\right)^{3q\sbs{p} - 5}
924: \nonumber\\
925: &\times&
926: \left[
927: 1 - \left(\frac{s\sbs{min}}{s\sbs{max}} \right)^{6-3q\sbs{p}}
928: \right]^{-1}.
929: \label{eqM0Sigma}
930: \end{eqnarray}
931: Inserting Eq.~(\ref{eqM0Sigma}) into Eq.~(\ref{eqtausig}) results in
932: \begin{eqnarray}
933: \tau(s) &=&
934: \frac{16 \pi\rho}{3 M_0} \cdot
935: \left(\frac{s}{s\sbs{max}}\right)^{3q\sbs{p}-5}
936: \frac{s\sbs{max} r^{5/2} \d r}{\left(\mathsf{G} M_*\right)^{1/2}}
937: \nonumber\\
938: &\times&
939: \frac{q\sbs{p}-5/3}{2-q\sbs{p}}
940: \left[
941: 1 - \left(\frac{s\sbs{min}}{s\sbs{max}} \right)^{6-3q\sbs{p}}
942: \right]
943: \nonumber\\
944: &\times&
945: \frac{I}{f(e,I)G(q\sbs{p},s)},
946: \label{eqtauM0}
947: \end{eqnarray}
948: which gives the collisional lifetime of an object with radius $s$.
949: Note that
950: \begin{equation}
951: \frac{1}{2-q\sbs{p}}
952: \left[
953: 1 - \left(\frac{s\sbs{min}}{s\sbs{max}} \right)^{6-3q\sbs{p}}
954: \right]
955: \longrightarrow
956: 3\:\ln \frac{s\sbs{max}}{s\sbs{min}}
957: \end{equation}
958: for $q\sbs{p} \rightarrow 2$.
959: %for the initial distribution.
960:
961: If the mean impact velocities in the system are high enough to allow
962: planetesimals of radius $s$ to get disrupted in a collision, i.e.
963: $X\sbs{c}(s) \ll s\sbs{max}/s$, $G(q\sbs{p},s)$ reduces to
964: \begin{equation}
965: G(q\sbs{p}, s) \approx \frac{q\sbs{p}-5/3}{q\sbs{p}-1} \cdot
966: X\sbs{c}(s)^{3-3q\sbs{p}},\label{eqGRed}
967: \end{equation}
968: and $\tau(s)$ to
969: \begin{eqnarray}
970: \tau(s) &=& \frac{16 \pi\rho}{3 M_0} \cdot
971: \left(\frac{s}{s\sbs{max}}\right)^{3q\sbs{p}-5}
972: \cdot r^2 \d r
973: \cdot \left(\frac{r}{\mathsf{G} M_*}\right)^{q\sbs{p} - 1/2}
974: \nonumber\\ &\times&
975: \frac{q\sbs{p}-1}{2-q\sbs{p}}
976: \left[
977: 1 - \left(\frac{s\sbs{min}}{s\sbs{max}} \right)^{6-3q\sbs{p}}
978: \right]
979: \cdot \left(2Q\sbs{D}^*\right)^{q\sbs{p}-1}
980: \nonumber\\ &\times&
981: \frac{I}{f(e,I)^{2q\sbs{p}-1}}
982: \label{eqtaured}
983: \end{eqnarray}
984: Now, we take into account the dependence of $Q\sbs{D}^*$ on the
985: object size $s$, as was done by \citet{o'brien-greenberg-2003}.
986: If we are only interested in the gravity regime, $s > s\sbs{b}$,
987: Eq.~(\ref{eqQD}) is simplified to
988: \begin{equation}
989: Q\sbs{D}^* (s) \approx Q\sbs{D,b}^*\cdot\left(\frac{s}{s\sbs{b}}\right)^{3b\sbs{g}},
990: \end{equation}
991: where $Q\sbs{D,b}^*$ is the critical specific energy at the
992: breaking radius, i.e. around the minimum of $Q\sbs{D}^*(s)$.
993: Assuming, further, that $I \propto e$,
994: we can write down the dependencies of the collisional lifetime,
995: \begin{equation}
996: \tau(s) \propto \sigma\sbs{tot}^{-1} \cdot
997: s^{3q\sbs{p}-5+3(q\sbs{p}-1)b\sbs{g}} \cdot
998: r^{3/2+q\sbs{p}} \cdot \d r \cdot e^{-5/3}.\label{eqDependencies}
999: \end{equation}
1000: \citet{o'brien-greenberg-2003} yield the same size dependence on $s$ in their
1001: Eq.~(11).
1002: %Note that the stellar mass/luminosity, the blowout limit $s\sbs{min}$,
1003: %and the total cross section $\sigma\sbs{tot}$ cannot be varied independently.
1004:
1005: To find $s\sbs{t}(t)$, the object size below which a steady state is reached,
1006: we assume that the populations move from their primordial state to the
1007: quasi-steady state instantaneously when the system age reaches their initial mean
1008: collisional lifetime, $\tau(s\sbs{t}) = t$.
1009: Inverting that, the resulting mass of objects in transition can be retrieved
1010: as a function of system age. Keeping the assumption $X\sbs{c} \ll s\sbs{max}/s$,
1011: the relation is
1012: \begin{eqnarray}
1013: s\sbs{t} (t) &\propto&
1014: t^{1/(3q\sbs{p}-5+3(q\sbs{p}-1)b\sbs{g})}
1015: \label{eqEquiMass}
1016: \end{eqnarray}
1017: for $t > \tau(s\sbs{b}) \equiv \tau\sbs{b}$.
1018: This transitional size is also plotted in Fig.~\ref{figMassDistribution}.
1019:
1020: \citet{pan-sari-2005} followed a similar approach in their study of the Kuiper-belt
1021: size distribution. Describing the propagation of the shock wave
1022: through the target, they introduce a parameter $\beta$
1023: that varies between $3/2$ (if all energy of a projectile goes to the shock wave)
1024: and $3$ (if all its momentum does).
1025: Their $\beta$ equals $1/b\sbs{g}$ in our nomenclature, and $b\sbs{g} = 0.5$
1026: leads to $\beta = 2$.
1027: Additionally, we have to replace their
1028: slope $q_0$ with our $3q\sbs{p} - 2$.
1029: Then, given their Eqs.~(6), (7), and $N_{>s} \propto s^{3-3q\sbs{p}}$,
1030: we yield the same exponent as in our Eq.~(\ref{eqEquiMass}).
1031: Note that what \citet{pan-sari-2005} call ``breaking radius'' is our
1032: ``transition radius'' $s\sbs{t}$, and
1033: their ``radius of equilibrium'' is our ``breaking radius'' $s\sbs{b}$.
1034:
1035: \subsection{Evolution of Disk Mass}
1036:
1037: Now, we derive the full expression for the time-dependent total disk mass.
1038: Using the size distribution given by Eq.~(\ref{eqnm}),
1039: $n\sbs{max}$ from Eq.~(\ref{DD}) and expressing
1040: $n\sbs{max}(0)$ through $M_0$ with the aid of Eq.~(\ref{eqM0}),
1041: we can perform the integration in Eq.~(\ref{eqTotalMass}).
1042: Then, the resulting time-dependent disk mass is
1043: \notetoeditor{Using ``frac'' instead of ``case'' is preferred but problematic
1044: with respect to formula layout. Note that the text refers to individual lines
1045: of the following equation.}
1046: \begin{eqnarray}
1047: M\sbs{disk}(t) &=&
1048: \frac{M_0}{1+t/\tau\sbs{max}}
1049: \left[
1050: 1 - \left(\frac{s\sbs{min}}{s\sbs{max}} \right)^{6-3q\sbs{p}}
1051: \right]^{-1}
1052: \nonumber\\
1053: &\times&
1054: \left[
1055: 1
1056: - \left( \case{s\sbs{t}(t)}{s\sbs{b}} \right)^{6-3q\sbs{p}}
1057: \left( \case{s\sbs{b}}{s\sbs{max}} \right)^{6-3q\sbs{p}}
1058: \left( 1 - \case{2-q\sbs{p}}{2-q\sbs{g}} \right)
1059: \right. \nonumber\\
1060: &+&
1061: \left( \case{s\sbs{t}(t)}{s\sbs{b}} \right)^{3q\sbs{g} - 3q\sbs{p}}
1062: \left( \case{s\sbs{b}}{s\sbs{max}} \right)^{6-3q\sbs{p}}
1063: \left( \case{2-q\sbs{p}}{2-q\sbs{s}} -
1064: \case{2-q\sbs{p}}{2-q\sbs{g}}
1065: \right) \nonumber\\
1066: &-&
1067: \left( \case{s\sbs{t}(t)}{s\sbs{b}} \right)^{3q\sbs{g} - 3q\sbs{p}}
1068: \left( \case{s\sbs{b}}{s\sbs{max}} \right)^{3q\sbs{s} - 3q\sbs{p}}
1069: \left( \case{s\sbs{min}}{s\sbs{max}} \right)^{6-3q\sbs{s}}\nonumber\\
1070: && \left.
1071: \times \left( \case{2-q\sbs{p}}{2-q\sbs{s}} \right)
1072: \right]\label{eqMassContrib}
1073: \end{eqnarray}
1074: for $\tau\sbs{b} < t < \tau\sbs{max}$.
1075: To make Eq.~(\ref{eqMassContrib}) valid for earlier phases,
1076: i.e. for $t < \tau\sbs{b}$, $s\sbs{b}$ should be replaced by
1077: $s\sbs{t}(t)$. The sizes involved are the maximum object size
1078: $s\sbs{max}$, the transition size between the primordial and reprocessed
1079: material $s\sbs{t}$, the breaking radius between the gravity and strength regime
1080: $s\sbs{b}$. The lower limit in the size
1081: distribution, $s\sbs{min}$, is crucial for the
1082: dust emission and it is usually taken to be the radiation pressure blowout
1083: limit. As long as $q\sbs{p} < 2$, it is fairly unimportant for the mass
1084: budget. However, we are interested in $q\sbs{p} \geq 2$
1085: as well. Therefore, we
1086: can safely set $s\sbs{min} = 0$ only in the last line of
1087: Eq.~(\ref{eqMassContrib}), where it enters through $s\sbs{min}/s\sbs{max}$
1088: to the power of $6-3q\sbs{s}$, with $q\sbs{s}\approx 11/6 < 2$.
1089:
1090: % \clearpage
1091: \begin{figure}[t!]
1092: \includegraphics{f6.eps}\\
1093: \caption{%
1094: The contributions of different terms in Eq.~(\ref{eqMassContrib})
1095: (dotted and dashed lines) and their total (solid line).
1096: \label{figMassContribution}}
1097: \end{figure}
1098: % \clearpage
1099:
1100: The relative importance of the terms in Eq. (\ref{eqMassContrib}) is
1101: illustrated in Fig.~\ref{figMassContribution}.
1102: A combination of the classic Dominik-Decin behavior in the first line of
1103: Eq.~(\ref{eqMassContrib}) together with the second line is a reasonably accurate
1104: approximation to $M\sbs{disk}(t)$ for most of the time.
1105: With the aid of Eq.~(\ref{eqEquiMass}),
1106: Eq. (\ref{eqMassContrib}) transforms to
1107: \begin{eqnarray}
1108: M\sbs{disk}(t) &\approx&
1109: \frac{M_0}{1+t/\tau\sbs{max}}
1110: \left[
1111: 1 - \left(\frac{s\sbs{min}}{s\sbs{max}} \right)^{6-3q\sbs{p}}
1112: \right]^{-1}
1113: \nonumber\\
1114: &\times&
1115: \left[
1116: 1 - \left( \frac{s\sbs{b}}{s\sbs{max}} \right)^{6-3q\sbs{p}} \cdot
1117: \left( \frac{t}{\tau\sbs{b}} \right)^
1118: {\frac{2-q\sbs{p}}{q\sbs{p}-5/3+(q\sbs{p}-1)b\sbs{g}}}
1119: \right.\nonumber\\
1120: && \left.\times
1121: \left( 1 - \frac{2-q\sbs{p}}{2-q\sbs{g}} \right)
1122: \right]\label{eqEquiMassShort}
1123: \end{eqnarray}
1124: for $\tau\sbs{b} < t < \tau\sbs{max}$.
1125: At $t \ll \tau\sbs{max}$, and assuming $q\sbs{p} = 1.87$,
1126: a further approximation is
1127: \begin{equation}
1128: M\sbs{disk}(t) \approx M_0
1129: \left(
1130: 1 - {\rm const} \cdot t^{0.2}
1131: \right).
1132: \end{equation}
1133:
1134: % \clearpage
1135: \begin{figure}[t!]
1136: \includegraphics{f7.eps}\\
1137: \caption{Evolution of total masses with (scaled) time,
1138: obtained in four numerical runs and with the analytic model.
1139: \label{figTotalMassAnaNum}}
1140: \end{figure}
1141: % \clearpage
1142:
1143: The evolution of the disk mass, both from the numerical runs and
1144: from the analytic solution (\ref{eqMassContrib}),
1145: is plotted in Fig.~\ref{figTotalMassAnaNum}, showing a good agreement
1146: between analytics and numerics.
1147: A deviation is only seen around $t = \tau\sbs{b}$ where the transition
1148: from primordial to reprocessed state sets on for gravity-dominated objects.
1149: The reason is that, to ease the analytic treatment, we neglect the smooth
1150: natural transition from material strength to self-gravity given by
1151: Eq.~(\ref{eqQD}) and assume a sharp break between the two power laws instead.
1152:
1153: \subsection{Evolution of Disk Mass at Latest Stages}
1154:
1155: As soon as the age of the system has reached the collisional lifetime of the
1156: largest bodies, i.e. at $t > \tau\sbs{max}$, the solids of all sizes in the disk
1157: reach quasi-steady state, and the change in total mass will be dominated
1158: by $1/t$.
1159: At this latest phase,
1160: the projectiles that
1161: can destroy objects of size $s\sbs{max}$ no longer follow a
1162: size distribution with the primordial slope, $2-3q\sbs{p}$.
1163: Instead, they have the slope of a
1164: collisional cascade under gravity regime, $2-3q\sbs{g}$. The slightly
1165: longer collisional lifetime can neither be expressed through
1166: Eq.~(\ref{eqtc}) that uses the \emph{initial} cross section
1167: $\sigma\sbs{tot}$
1168: nor through Eq.~(\ref{eqtauM0}) that contains the initial disk mass $M_0$ and
1169: slope $q\sbs{p}$.
1170: The correct way to evaluate $\tau\sbs{max}$ is to use the initial number density
1171: of biggest objects, $n\sbs{max}(0)$, and the slope $q\sbs{g}$.
1172: Expressing $\sigma\sbs{tot}$ in Eq.~(\ref{eqtc}) through $n\sbs{max}$ with
1173: the help of Eq.~(\ref{eqSigma}) and replacing then $q\sbs{p}$ with $q\sbs{g}$,
1174: we obtain
1175: \begin{eqnarray}
1176: \tau\sbs{max} &=&
1177: \frac{12q\sbs{g}-20}{n\sbs{max}(0)\cdot s\sbs{max}^3} \cdot
1178: \frac{r^{5/2} \d r}{\left(\mathsf{G} M_*\right)^{1/2}}
1179: \nonumber\\
1180: &\times&
1181: \frac{I}{f(e,I)G(q\sbs{g}, s\sbs{max})}.
1182: \end{eqnarray}
1183: Expressing now $n\sbs{max}(0)$ through $M_0$ by virtue of Eq.~(\ref{eqM0})
1184: yields
1185: \begin{eqnarray}
1186: \tau\sbs{max} &=&
1187: \frac{16 \pi\rho}{3 M_0} \cdot
1188: s\sbs{max} \cdot
1189: \frac{r^{5/2} \d r}{\left(\mathsf{G} M_*\right)^{1/2}}
1190: \nonumber\\
1191: &\times&
1192: \frac{q\sbs{g}-5/3}{2-q\sbs{p}}
1193: \left[
1194: 1 - \left(\frac{s\sbs{min}}{s\sbs{max}} \right)^{6-3q\sbs{p}}
1195: \right]^{-1}
1196: \nonumber\\
1197: &\times&
1198: \frac{I}{f(e,I)G(q\sbs{g}, s\sbs{max})}
1199: \label{eqtcEx},
1200: \end{eqnarray}
1201: where both slopes, $q\sbs{p}$ and $q\sbs{g}$, appear
1202: (cf. Eqs.~\ref{eqtc} and \ref{eqtauM0}).
1203:
1204: \subsection{Evolution of Mass in Dynamically ``Cold'' Disks}
1205:
1206: All the treatment above applies to planetesimal belts where relative velocities
1207: are high enough for the biggest objects to be destroyed by
1208: mutual collisions. This might not be the case in dynamically ``cold'' disks
1209: with low eccentricities and inclinations and/or very far from the star.
1210:
1211: Consider again the lifetime of objects $\tau(s)$.
1212: As $s$ increases, $X\sbs{c}(s)$ (Eq.~\ref{eqXcFull})
1213: increases too and at a certain point reaches $s\sbs{max}/s$.
1214: At this point, $G$ (Eq.~\ref{eqGFull}) becomes zero and $\tau(s)$ (Eq.~\ref{eqtauM0})
1215: goes to infinity.
1216: This means that, for a given impact velocity, objects above a certain
1217: critical size cannot be disrupted anymore.
1218: In systems with low relative velocities, that critical size may happen
1219: to be smaller than $s\sbs{max}$. This will affect the mass evolution.
1220: Specifically, when $s\sbs{t}$ reaches that critical size, the overall
1221: mass decay ceases.
1222:
1223: % \clearpage
1224: \begin{figure}[t!]
1225: \includegraphics{f8.eps}\\
1226: \caption{
1227: Influence of the effective eccentricity assumed in the analytic model
1228: for a disk of 1~$M_\oplus$ at $r=10$~AU with a radial extent
1229: $\d r=7.5$~AU. The $I=e/2$ relation between eccentricity and inclination
1230: is assumed.
1231: \label{figQcomp}}
1232: \end{figure}
1233: % \clearpage
1234:
1235: To illustrate such effects,
1236: Fig.~\ref{figQcomp} shows the influence of the effective
1237: $e$ and $I$ on the evolution of the total mass of a disk of initially
1238: 1~$M_\oplus$ at an effective distance of $10$~AU, calculated with our
1239: analytic model.
1240: For colder disks, the curves start to flatten. This happens because the
1241: largest planetesimals (that dominate the total mass) stay intact, which
1242: slows down the mass loss.
1243:
1244: \subsection{Evolution of Dust Mass}
1245:
1246: % \clearpage
1247: \begin{figure}[t!]
1248: \includegraphics{f9.eps}\\
1249: \caption{%
1250: Similar to Fig. \ref{figTotalMassAnaNum} but for dust masses, i.e.
1251: masses in particles with radii below 1~mm.
1252: \label{figDustMassAnaNum}}
1253: \end{figure}
1254: % \clearpage
1255: The dust mass can be evaluated in a similar way as the disk mass.
1256: We use now Eqs.~(\ref{eqnm}), (\ref{eqDustMass}), (\ref{DD}),
1257: (\ref{eqM0}), and (\ref{eqEquiMass}).
1258: Neglecting the minimum mass $s\sbs{min}$ only when
1259: it enters the formula through $s\sbs{min}/s\sbs{max}$, we obtain
1260: \begin{eqnarray}
1261: M\sbs{dust} (t) = \frac{M_0}{1+t/\tau\sbs{max}}\cdot
1262: \left(\frac{t}{\tau\sbs{b}}\right)^
1263: {\frac{q\sbs{g}-q\sbs{p}}{q\sbs{p}-5/3+(q\sbs{p}-1)b\sbs{g}}}\cdot
1264: \frac{2-q\sbs{p}}{2-q\sbs{s}}
1265: \nonumber\\
1266: \times
1267: \left(\frac{s\sbs{b}}{s\sbs{max}}\right)^{2-q\sbs{p}}
1268: \left[\left(\frac{s\sbs{d}}{s\sbs{b}}\right)^{2-q\sbs{s}} -
1269: \left(\frac{s\sbs{min}}{s\sbs{b}}\right)^{2-q\sbs{s}}
1270: \right]^{-1}
1271: \label{eqDustMassEx}
1272: \end{eqnarray}
1273: for $\tau\sbs{b} < t < \tau\sbs{max}$.
1274: Before that, i.e. at $t < \tau\sbs{b}$,
1275: we have $q\sbs{s}$ and $b\sbs{s}$ instead of $q\sbs{g}$ and $b\sbs{g}$,
1276: respectively. If the assumed primordial slope, $q\sbs{p}$, equals the
1277: steady-state slope in the strength regime, $q\sbs{s}$,
1278: the dust mass stays constant, which is the case for the first part of the
1279: numerical integration.
1280: However, as soon as the transitional zone reaches objects large enough to be
1281: influenced by self-gravity,
1282: Eq.~(\ref{eqDustMassEx}) starts to work.
1283: It shows that the evolution of dust mass depends most strongly
1284: on the difference between $q\sbs{p}$ and $q\sbs{g}$.
1285: The dust mass decay, obtained both from the numerical runs and
1286: analytic solution (\ref{eqDustMassEx}),
1287: are shown in Fig.~\ref{figDustMassAnaNum}. For $t > \tau\sbs{b}$,
1288: we roughly have $M\sbs{dust} \propto t^\xi$ with $\xi \approx -0.3$.
1289:
1290: We finally note that Eq.~(\ref{eqDustMassEx}) is valid as long as the
1291: collisional lifetime
1292: of the largest planetesimals is longer than the age of the system.
1293: When $t > \tau\sbs{max}$, $t/\tau\sbs{b}$ in that equation must be replaced by
1294: $\tau\sbs{max}/\tau\sbs{b}$.
1295:
1296: \subsection{The Model Parameters}
1297: \label{secParameters}
1298:
1299: Our analytic model contains several parameters that either
1300: differ from similar parameters in the numerical model (such as $e$) or are absent there
1301: (such as $q\sbs{s}$ and $q\sbs{g}$).
1302: To use the analytic model, we have to specify them.
1303: We now describe how this can be done, explaining, in particular,
1304: the choice of parameters used to plot analytic curves in
1305: Figs.~\ref{figMassContribution}--\ref{figDustMassAnaNum}.
1306:
1307: Two important free parameters of the analytic model are $q\sbs{s}$ and $q\sbs{g}$.
1308: We use the work of \citet{o'brien-greenberg-2003} who
1309: found the slope of the size distribution in a system in a collisional steady
1310: state. With the dependence of the critical specific energy on the object
1311: size given in Eq.~(\ref{eqQD}), they give a power-law index
1312: \begin{eqnarray}
1313: q = \frac{11/6+b}{1+b} \label{eqOBrienq}
1314: \end{eqnarray}
1315: in their Eq.~(24). With $b=b\sbs{s}=-0.1$ for the strength regime we have
1316: $q=q\sbs{s}=1.877$.
1317: Similarly, with $b=b\sbs{g}=0.5$ for the gravity regime,
1318: Eq.~(\ref{eqOBrienq}) can be used to derive
1319: $q\sbs{g} \approx 5/3$.
1320: It is these values that we used in
1321: Eq.~(\ref{eqMassContrib}) to produce Figs.~\ref{figMassContribution}--\ref{figQcomp}
1322: and in Eq.~(\ref{eqDustMassEx}) to plot Fig.~\ref{figDustMassAnaNum}.
1323:
1324: % \clearpage
1325: \begin{figure}[t!]
1326: \includegraphics{f10.eps}\\
1327: \caption{Index $\xi$ of the power-law evolution of the dust mass,
1328: $M\sbs{d}\propto t^\xi$. The horizontal axis gives the dependence on the
1329: slope of the primordial mass distribution, $q\sbs{p}$, for values from
1330: $q\sbs{g} = 1.57$ (bottom) to $q\sbs{g} = 1.77$ (top) for the slope in the
1331: gravity regime. The bold line is for $q\sbs{g} = 1.67 \approx 5/3$.
1332: Vertical lines indicate the mean value and error estimates for
1333: $q\sbs{p}$ from \citet{trujillo-et-al-2001}.
1334: \label{figDustEvoSlope}}
1335: \end{figure}
1336: % \clearpage
1337:
1338: In contrast to $q\sbs{s}$ and $q\sbs{g}$, the primordial slope, $q\sbs{p}$,
1339: is a free parameter not only in the analytic model, but also in the numerical one.
1340: As stated in Sect. ~\ref{secCommons}, in all ``nominal'' runs we assumed
1341: $q\sbs{p} = 1.87$, which corresponds to $p\sbs{p} = 3q\sbs{p} -2 = 3.61$ in the size scaling.
1342: In principle, $q\sbs{p}$ describes the mass
1343: distribution at the onset of the collisional grinding
1344: of the disk and, therefore, represents a link to the planetesimal formation
1345: process. The outcome of the agglomeration phase is the input to the phase of
1346: disruptive collisions.
1347: The Kuiper belt
1348: is the only source for observational constraints to this parameter so far, and
1349: recent surveys suggest a value of $p\sbs{p} = 4.0 \pm 0.5$
1350: \citep[e.g.,][]{trujillo-et-al-2001, bernstein-et-al-2004} or $q\sbs{p} =
1351: 2.00 \pm 0.17$. Simulations by \citet{kenyon-bromley-2004c} yield
1352: $p\sbs{p}=4.0$--$4.5$ or $q\sbs{p}=2.00$--$2.17$. According to
1353: Eq.~(\ref{eqDustMassEx}), where we have $M\sbs{dust} \propto t^\xi$, and
1354: together with $q\sbs{g} \approx 1.67$, this would
1355: change the dust mass evolution from $M\sbs{dust}\propto t^{-0.32}$ for $q\sbs{p}
1356: = 1.87$ to $M\sbs{dust}\propto t^{-0.40}$ for $q\sbs{p} = 2.00$.
1357: Fig.~\ref{figDustEvoSlope} shows the rather moderate dependence of the index
1358: $\xi$ on the two mass distribution slopes, $q\sbs{g}$ and $q\sbs{p}$.
1359:
1360: While the dust size limit, $s\sbs{d}$, has little influence on the
1361: mass budget, the breaking size, $s\sbs{b}$, the maximum size, $s\sbs{max}$,
1362: and the ratio of the two are relevant to the evolution as they define the
1363: lifetime of the largest bodies $\tau\sbs{max}$ relative to $\tau\sbs{b}$.
1364: What is more, the ratio $s\sbs{b}/s\sbs{max}$ determines the rate of the mass
1365: decay in Eq.~(\ref{eqEquiMassShort}).
1366: From Sect.~\ref{secQD} we know the location of the breaking radius to be
1367: $316$~m for the material properties assumed, and the upper size limit
1368: of all the runs was set to $s\sbs{max}=74$~km.
1369:
1370: %As we have seen, the fate of those biggest objects
1371: %largely determines the further evolution on Gyr-timescales.
1372: %If they can get collisionally disrupted, the
1373: %decay will asymptotically follow $1/t$. Otherwise, the decay in our
1374: %analytic model will cease.
1375: Another parameter in the analytic model is the collisional lifetime of objects of breaking radius,
1376: $\tau\sbs{b} = \tau(m\sbs{b})$.
1377: Eq.~(\ref{eqtauM0}) expresses it through other
1378: parameters critical for the efficiency of
1379: collisions: the radial distance to the star $r$, the disk radial extension
1380: $\d r$, and the effective eccentricity $e$
1381: and inclination $I$.
1382: We choose to fix both the effective distance and the disk
1383: extension to be $r = 4/3 \d r = 10$~AU when reproducing analytically the
1384: results of the ii-0.3 run, 20~AU for i-0.3, 40~AU for o-0.3, and 80~AU for oo-0.3.
1385: Further, the inclination can be coupled to eccentricity
1386: by assuming the equilibrium condition $I = e/2$.
1387: Thus, only $e$ remains as a free parameter.
1388: The best fit to, e.g., the ii-0.3 run is achieved if we assume $e\approx
1389: 0.075$ in the analytic model, which is approximately one quarter of
1390: $e\sbs{max}=0.3$.
1391: With these choices, we find $\tau(s\sbs{b}) \approx 4\times 10^5$~years.
1392:
1393: Alternatively, $\tau\sbs{b}$
1394: can be directly retrieved from the break
1395: in the evolution of the dust mass (see Fig.~\ref{figDustMassAnaNum}).
1396: This method gives $\tau(s\sbs{b}) \approx 5\times 10^5$~years,
1397: %$\tau\sbs{b}$ calculated with Eq.~(\ref{eqtauM0}) for $s\sbs{b} = 316$~m is
1398: which is approximately $4/3$ times the value calculated with Eq.~(\ref{eqtauM0}).
1399: This discrepancy is probably a result of the
1400: particle-in-a-box assumptions made by \citet{wyatt-et-al-2007a} in derivation of
1401: Eq.~(\ref{eqtc}).
1402: We prefer this empirical scaling and thus applied the factor of $4/3$ to all analytically
1403: estimated timescales in this paper.
1404:
1405: %===========================================================================
1406: \section{EVOLUTION OF DISK LUMINOSITY}
1407: %===========================================================================
1408: \label{secEvolLuminosity}
1409:
1410: \subsection{Fractional Luminosity for a Given Age}
1411: \label{sec_fmax}
1412:
1413: Following \citet{wyatt-et-al-2007a}, we define the
1414: fractional luminosity of dust as
1415: \begin{equation}
1416: f\sbs{d} \equiv \sigma\sbs{tot} / (4 \pi r ^2) ,
1417: \label{f}
1418: \end{equation}
1419: which assumes that dust grains are black bodies, absorbing
1420: and re-emitting all the radiation they intercept.
1421: \citet[][their Eq.~20]{wyatt-et-al-2007a}
1422: found that there is a maximum possible
1423: fractional luminosity $f\sbs{max}$ for a given age, whose
1424: value is independent of the initial disk mass, but depends on
1425: other model parameters such as
1426: the distance $r$ of the disk center from the star,
1427: its width $\d r$,
1428: size of the largest planetesimals $D\sbs{c}$,
1429: critical fragmentation energy $Q\sbs{D}^*$,
1430: orbital eccentricity of planetesimals $e$
1431: (with their inclination being $I = e/2$),
1432: as well as the stellar mass $M_*$ and luminosity $L_*$.
1433:
1434: We now wish to explore $f_d(t)$ and check whether it has an upper limit
1435: in the framework of our analytic model.
1436: To this end, we used Eq.~(\ref{f}) and calculated $\sigma\sbs{tot}$
1437: with the aid of our Eq.~(\ref{eqDustMassEx}) for the dust mass.
1438: We assumed a solar-type star with $M_* = L_* = 1$ and probed
1439: disks with $M\sbs{disk} = 1, 3, 10$, and $30 M_\oplus$;
1440: $r = 3, 10, 30$, and $100$~AU;
1441: $\d r/r = 1/8, 1/4, 1/2$, and $1$;
1442: $e = 0.05, 0.10, 0.15$, and $0.20$.
1443: The results are presented in Fig.~\ref{fig_fmax} (thick lines).
1444: As a standard case, we adopted
1445: $M\sbs{disk} = 10 M_\oplus$,
1446: $r = 30$~AU, $\d r/r = 1/2$, and $e = 0.10$.
1447: It is shown with a thick solid curve in each of the panels.
1448:
1449: In the same Fig.~\ref{fig_fmax}, we have overplotted with thin lines
1450: the dust luminosity $f\sbs{d}$ computed with Eqs. (14), (19), and (20) of
1451: \citet{wyatt-et-al-2007a}, for comparison.
1452: In that calculation, we assumed $Q\sbs{D}^*=300$~J/kg (constant in their model),
1453: $D\sbs{c} = 60\km$, and the same values
1454: of those parameters that are common in their and our model
1455: ($M_*$, $L_*$, $r$, $\d r/r$, and $e$).
1456:
1457: % \clearpage
1458: \begin{figure*}[t!]
1459: \centering
1460: \includegraphics[width=0.5\linewidth]{f11a.eps}%
1461: \includegraphics[width=0.5\linewidth]{f11b.eps}\\
1462: \includegraphics[width=0.5\linewidth]{f11c.eps}%
1463: \includegraphics[width=0.5\linewidth]{f11d.eps}
1464: \notetoeditor{The "width" parameter should be obsolete in the final two-column version.}
1465: \caption{
1466: Fractional luminosity of dust around a solar-like star as a function of age.
1467: Thick lines: our analytic model; thin lines: $f\sbs{d}$ of \citet{wyatt-et-al-2007a}.
1468: Different panels demonstrate dependence on different parameters:
1469: $M\sbs{disk}$ \textsl{(top left)},
1470: $r$ \textsl{(top right)},
1471: $\d r/r$ \textsl{(bottom left)},
1472: and $e$ \textsl{(bottom right)}.
1473: A standard case with
1474: $M_* = L_* = 1$, $M\sbs{disk} = 10 M_\oplus$,
1475: $r = 30$~AU, $\d r/r = 1/2$, and $e = 0.10$ is shown
1476: with solid lines (common in all panels).
1477: \label{fig_fmax}}
1478: \end{figure*}
1479: % \clearpage
1480:
1481: Analysis of Fig.~\ref{fig_fmax} allows us to make a number of conclusions.
1482: First, as expected, our model yields more gently sloping curves
1483: than that by Wyatt et al.
1484: As discussed above, the $1/t$ law will be asymptotically reached in our model,
1485: too, but this does rarely happen at ages $t < 10$~Gyr. Only the first signs
1486: of the curves' steepening appear at Gyr ages, and that only for the cases when
1487: the collisional evolution is faster (higher masses, closer-in or more confined
1488: dust rings, higher eccentricities).
1489: As a consequence of the slope difference between the two models,
1490: our model places more stringent upper limits of $f\sbs{d}$ at earlier ages,
1491: and conversely, it allows the Gyr-old systems to have a somewhat higher $f\sbs{d}$
1492: than the model by Wyatt et al. does.
1493:
1494: Next, the dependence of $f\sbs{max}$ on the initial disk mass, which cancels out in their
1495: model, is retained in our nominal runs (the top left panel). In fact, the maximum
1496: possible $f\sbs{d}$ is then determined by the maximum initial disk mass that still appears
1497: physically plausible in the framework of theories of planetesimal accretion and planet
1498: formation.
1499:
1500: Another point to mention is that, whereas the dependence on the disk width (bottom right)
1501: and planetesimal eccentricities is relatively weak and monotonic, the dependence on
1502: the disk location (top right) is rather strong and more intricate.
1503: That the dependence is strong is the consequence of Eq.~(\ref{eqScalingR}) that predicts the
1504: timescales to very sensitively depend on the distance from the star, and of Eq.~(\ref{f})
1505: that contains a ``dilution factor'' $r^2$. At the beginning of the evolution, the innermost ring
1506: is always the brightest because the dilution factor $r^2$ in Eq.~(\ref{f}) is the smallest.
1507: At the end of the evolution, the opposite is true: the outermost ring will become the brightest,
1508: because its collisional evolution is the slowest and it retains more mass than
1509: inner disks.
1510: Therefore, all four curves intersect each other at a certain point;
1511: the 30 and 100~AU curve do that after 10~Gyr, i.e. outside that right edge of the plot.
1512: After that, all the curves go parallel to each other in the ``Dominik-Decin regime'',
1513: following a $1/t$ law. Note that inner rings reach
1514: the $1/t$ regime more quickly: already at 10~AU it is established in around 100~Myr for an
1515: initial mass of 10~$M_\oplus$.
1516:
1517: Although the existence of a ``maximum fractional luminosity for a given age'', as suggested by
1518: \citet{wyatt-et-al-2007a}, no longer holds in our model as a robust mathematical
1519: statement, in practice our model still suggests that $f\sbs{d}(t)$ cannot exceed a certain
1520: limit, unless the model parameters take extreme values, incompatible with our understanding
1521: of the planetesimal disks. For instance, we do expect $f\sbs{d} < 10^{-4}$ at $t=10$~Gyr,
1522: provided that the initial disk did not contain more than 30 earth masses of solids and that
1523: the mean orbital eccentricity of planetesimals is not lower than 0.1 (corresponding to
1524: the mean inclination larger than $3^\circ$).
1525: Therefore, plots such as Fig.~\ref{fig_fmax} can be used to check whether or not
1526: $f\sbs{d}$ observed for a certain system with a known age is compatible with a ``smooth'',
1527: unperturbed collisional evolutionary scenario. In case it is not, it will be an indication
1528: that other mechanisms (delayed stirring, recent giant break-ups, non-collisional
1529: dust production etc.) should be thought of to explain the observations.
1530:
1531: \subsection{24 and 70 Micron Fluxes from Partial Rings}
1532: \label{secLuminosity}
1533: In order to produce directly observable quantities from the derived dust masses,
1534: we now concentrate on dust luminosities at particular infrared wavelengths.
1535: We calculated the dust temperature and the thermal emission integrated over the whole
1536: disk with a more accurate, yet sufficiently simple, model, assuming that the
1537: absorption/emission efficiency
1538: is constant up to wavelengths of $2\pi$ times the size of the particles, $s$,
1539: and proportional to $s^{-1}$ beyond that \citep{backman-paresce-1993}.
1540: Then we computed the spectral flux densities
1541: of dust emission $F\sbs{d}$ and of the stellar radiation $F_*$ at a certain wavelength,
1542: as well as their ratio $F\sbs{d}/F_*$.
1543: As the size distribution in the dust regime quickly reaches its steady state,
1544: the luminosity $F\sbs{d}$ is directly proportional to the dust mass. Therefore, the same
1545: initial constancy and subsequent $t^\xi$ decay with $\xi=-0.3 \ldots -0.4$ apply.
1546:
1547: % \clearpage
1548: \begin{figure}[t!]
1549: \includegraphics{f12a.eps}\\
1550: \includegraphics{f12b.eps}\\
1551: \caption{Flux ratio versus time for \textsl{(top)} 24~\textmu{m} and
1552: \textsl{(bottom)} 70~\textmu{m}.
1553: \label{fig2470}}
1554: \end{figure}
1555: % \clearpage
1556:
1557: Fig.~\ref{fig2470} shows the evolution of the excess
1558: emission at the Spitzer/MIPS wavelengths 24 and 70~\textmu{m},
1559: obtained from the four nominal runs.
1560: Since all disks have the same initial total mass ($1$~$M_\oplus$), the disks
1561: closer to the star are brighter and start to decay earlier.
1562: The difference between the excesses at 24 and 70~\textmu{m}, a measure of the
1563: disks' effective temperature, is varying with radial distance as well. Thus, the
1564: convergence of just the 70~\textmu{m} fluxes at later times is only
1565: coincidental. It is a result of the radial dependence of temperature
1566: and the collisional timescale.
1567:
1568: \subsection{Fluxes from Extended Disks}
1569:
1570: %All available data on the evolution of disk luminosities are statistical data
1571: %from an ensemble of different disks of different ages. Disks may vary, for
1572: %example, in mass, radial extent, distribution of eccentricities,
1573: %metallicity, asymmetry, planetary influence, type of the host star, and
1574: %stochasticity of the evolution, with some probable correlations in between these
1575: %properties.
1576: %To explore this parameter space, we follow
1577: %\citet{wyatt-et-al-2007b} and fix the stellar type while varying disk mass and
1578: %radius.
1579: Since resolved debris disks suggest that the parent body
1580: reservoir in the disks is usually confined to a toroidal region
1581: (a planetesimal belt), or is made up of several such tori,
1582: it seems appropriate to simply combine individual rings without taking
1583: into account possible interactions between particles that belong to different rings.
1584: Thus, we summed up the fluxes from the four main runs. Different
1585: radial distributions in the whole disk were simulated by ``weighting''
1586: the individual rings:
1587: \begin{equation}
1588: F\sbs{d} = \sum_{j=1}^{4}\limits F\sbs{d,j} (r_j/r_0)^{\gamma},
1589: \label{F_lambda}
1590: \end{equation}
1591: where $r_j$ are the central distances of the rings and
1592: values of $0$, $1$, $2$, and $3$ were used for the slope $\gamma$. As the
1593: reference runs were made for rings of one earth mass each with volumes
1594: proportional to $r^3$, the corresponding volume density in the extended disk is
1595: proportional to $r^{\gamma - 3}$, while the pole-on surface density and
1596: normal geometrical optical depth follow $\propto r^{\gamma - 2}$.
1597: The distance $r_0$ normalizes the total mass to $1 M_\oplus$.
1598: Therefore, by changing the slope, the mass is only shifted between inner and outer regions.
1599:
1600: % \clearpage
1601: \begin{figure}[t!]
1602: \includegraphics{f13.eps}\\
1603: \caption{Time evolution of the infrared excess of extended disks
1604: %(Eq.~\ref{F_lambda})
1605: with different initial radial distributions
1606: (labels indicate the radial slope of the surface mass density;
1607: the thicker lines, the flatter the profiles)
1608: at 24 \textmu{m} (dashed lines) and 70 \textmu{m} (solid lines).
1609: The total mass is $1 M_\oplus$ in each case.
1610: \label{figSummedFluxes}}
1611: \end{figure}
1612: % \clearpage
1613:
1614: In Fig. \ref{figSummedFluxes} the effect on the 24 and 70~\textmu{m} fluxes is
1615: shown. If the weights are assigned in favor of more distant debris rings, the
1616: resulting fluxes are naturally reduced. The same is true for the speed of
1617: the decay because the timescales get longer. The evolution of the fluxes at the
1618: two Spitzer/MIPS wavelengths 24 and 70~\textmu{m} differs significantly.
1619: At 24~\textmu{m} the decay starts earlier and reaches its maximum speed earlier
1620: because shorter-lived inner regions make the main contribution.
1621:
1622: The models contain a sufficient number of parameters, variation of which
1623: would affect the curves in Fig. \ref{figSummedFluxes} in different ways.
1624: As stated earlier, varying the total mass changes the timescale according to
1625: $\tau \propto M\sbs{disk}^{-1}$. Hence,
1626: the curves can be shifted along the lines of equal $t \cdot M\sbs{disk}$,
1627: i.e. along the top left -- bottom right diagonal.
1628: As seen from Fig.~\ref{figSummedFluxes}, variation
1629: of the radial distribution changes both the absolute level and the tilt of the curves.
1630: Besides, it affects the disk colors, i.e. the separation of the
1631: 24 and 70~\textmu{m} curves in Fig. \ref{figSummedFluxes}.
1632: In addition, the dynamical timescales, and therefore the tilt of the curves,
1633: are affected by eccentricities and inclinations of the parent bodies
1634: that may reflect the presence of planetary perturbers in the disk
1635: (see Sect.~\ref{secEcc}).
1636: Altogether, these degrees of freedom would allow one to reproduce a broad set
1637: of observational data.
1638:
1639:
1640: %===========================================================================
1641: \section{COMPARISON WITH OBSERVATIONAL DATA}
1642: \label{secObservations}
1643:
1644: \subsection{Spitzer Data}
1645:
1646: The advent of the Spitzer Space Observatory has brought a tremendous
1647: increase in the number of main-sequence stars surveyed for the existence
1648: of cold dust emission (see \citet{werner-et-al-2006} for a recent compilation).
1649:
1650: The wealth of data from these debris disk surveys allows us to confront
1651: our models with actual observations. To this end, we searched the literature
1652: for published flux ratios at 24 and/or 70 \textmu{m} (two of the three MIPS
1653: bands) around G-type main-sequence stars.
1654: To qualify as a main-sequence star we applied a lower limit to the stellar age
1655: of 10~Myr. Sources with stellar age estimates younger than this are likely
1656: stars with gas-dominated, protoplanetary disks; these were not taken into
1657: account.
1658:
1659: The bulk of the data taken in the framework of the Legacy program ``Formation
1660: and Evolution of Planetary Systems'' (FEPS)
1661: \citep{meyer-et-al-2004,meyer-et-al-2006} is
1662: public since December 2006. The FEPS archive contains images, spectra,
1663: photometry tables and Kurucz photosphere models and is available at
1664: \url{http://\linebreak[0]data.spitzer.caltech.edu/\linebreak[0]%
1665: popular/\linebreak[0]feps/\linebreak[0]20061223\_%
1666: enhanced\_v1/}.
1667: Age estimates have been published for 46 FEPS G stars
1668: \citep{kim-et-al-2005,stauffer-et-al-2005,silverstone-et-al-2006}.
1669:
1670: The large Guaranteed Time Observer (GTO) survey of FGK stars contains another
1671: 64 stars, where ages are available
1672: \citep{beichman-et-al-2005,beichman-et-al-2006b,bryden-et-al-2006}. Data for ten
1673: more G stars are listed in \citet{chen-et-al-2005a,chen-et-al-2005b}.
1674: In total, 120 G-type main-sequence stars with flux ratios at 24 and/or 70 $\mu$m have
1675: been compiled from the literature for comparison with model flux ratios.
1676:
1677: \subsection{Population Synthesis}
1678:
1679: Based on the analytic prescription presented in Sect.~\ref{secAnalytics} and
1680: motivated by the \citet{wyatt-et-al-2007b} work, we now
1681: build a synthetic set of debris disks around G2 stars. We
1682: generate a set of ring-like disks of width $\d r$
1683: located at distances $r \in [r\sbs{min},r\sbs{max}]$,
1684: with masses $M\sbs{disk}\in [M\sbs{min}, M\sbs{max}]$,
1685: and ages between 10~Myr and 10~Gyr.
1686: The probability to have a disk of initial mass $M_0$ at radius $r$ was assumed
1687: to follow $M_0^{-1} r^{-0.8}$,
1688: where $M_0^{-1}$ corresponds to a log-normal distribution of initial disk
1689: masses and the $r^{-0.8}$ dependence was proposed by \citet{wyatt-et-al-2007b}.
1690: As described in Sect.~\ref{secLuminosity}, the temperatures and the resulting
1691: thermal fluxes are calculated using the modified black-body formulas
1692: by \citet{backman-paresce-1993}, assuming the emitting grains
1693: to have $s=1$~\textmu{m},
1694: in agreement with the size distribution shown in Fig.~\ref{figMassDistribution}.
1695: The other parameters are taken to be: $q\sbs{p} = 2.00$, $q\sbs{g}=1.67$,
1696: $q\sbs{s}=1.877$, $\d r/r=0.5$, $2I = e = 0.15$, $Q\sbs{D}^*(1\:\mathrm{m}) =
1697: Q\sbs{D}^*(1\:\mathrm{km}) = 5\times 10^6$~erg/g, $b\sbs{d}=-0.12$,
1698: $b\sbs{g}=0.47$,
1699: roughly corresponding to basalt in \citet{benz-asphaug-1999}.
1700: %We neglect the influence of the wavy size distribution, of the non-uniform radial
1701: %distribution within each disk, and assume even the innermost massive disks
1702: %to be optically thin.
1703: Due to the small observational sample, our aim was not to perform a multi-parameter
1704: fit to the observations, but rather to cover the range of observed flux densities, which is
1705: defined by the limits of the distributions, not by their slopes.
1706:
1707: % \clearpage
1708: \begin{figure}[t!]
1709: \includegraphics{f14a.eps}\\
1710: \includegraphics{f14b.eps}\\
1711: \caption{Flux ratios versus time for 24 \textmu{m} \textsl{(top)}
1712: and 70 \textmu{m} \textsl{(bottom)}.
1713: The synthesized population (\textsl{small dots}) is compared to the observed
1714: one (\textsl{big dots}).
1715: Individually labeled is the possibly transient system HD 72905, see text.
1716: \label{fig2470obs}}
1717: \end{figure}
1718: % \clearpage
1719:
1720: Varying disk locations and masses easily reproduces the observed
1721: distribution of fluxes at 24~\textmu{m} and 70~\textmu{m}
1722: (Fig.~\ref{fig2470obs}). The synthetic population shown corresponds to
1723: $r\sbs{min}\approx 20$~AU, $r\sbs{max}\approx 120$~AU and $M\sbs{min} <
1724: 0.01$~$M_\oplus$, $M\sbs{max}\approx 30$~$M_\oplus$. Here, the radial range is
1725: needed to cover the range of colors, i.e. the ratios between the excess
1726: emissions at the two wavelengths. The mass range is needed to cover the observed
1727: range of excess, especially for younger disks at 70~\textmu{m}.
1728:
1729: Analyses of Spitzer detections might indicate a statistically significant
1730: increase of both 24 and 70~\textmu{m} fluxes at ages between
1731: a few tens of Myr to a few hundreds of Myr.
1732: (e.g., J.~M. Carpenter et al., in prep.), which can only be marginally seen in
1733: our sample (Fig.~\ref{fig2470obs}).
1734: It is hypothesized that this feature
1735: is caused either by an increased dust production due to delayed stirring by
1736: growing planets or by events
1737: similar to the late heavy bombardment in the solar system.
1738: Such effect could only be studied with an improved version of our analytic
1739: model or with the numerical one.
1740:
1741: % \clearpage
1742: \begin{figure}[t!]
1743: \includegraphics{f15a.eps}\\
1744: \includegraphics{f15b.eps}\\
1745: \caption{Relation between fluxes at 24 and 70~\textmu{m} versus time.
1746: The synthesized population (\textsl{small dots}) is compared to the
1747: observed one (\textsl{big dots and triangles}).
1748: The average photospheric uncertainty for both filters is marked by dashed
1749: lines in the upper panel. Excesses below those limits in either of the
1750: two filters are marked by triangles in the lower panel. In addition, the upper panel
1751: shows lines of equal dust mass and the lower panel gives the ring radii corresponding to the
1752: colors.
1753: \label{fig24/70}}
1754: \end{figure}
1755: % \clearpage
1756: The distribution of disk colors is more difficult to
1757: reproduce. Fig.~\ref{fig24/70} shows a significant abundance of
1758: fainter but warmer disks in an area that is not covered by the synthetic
1759: population. One explanation would be that the upper mass limit is a function of
1760: radial distance, and that the innermost disks tend to be less massive and less
1761: luminous, from the very beginning. In addition, the lower panel of that
1762: Fig.~\ref{fig24/70} shows a trend towards higher effective temperatures for
1763: higher ages, which is difficult to understand. Indeed, as long as faint close-in
1764: disks are observed around older stars, one would expect ever brighter disks, and therefore more
1765: numerous detections of disks at the same distances
1766: around younger stars.
1767: Furthermore, the trend in question contradicts to the results by
1768: \citet{najita-williams-2005}, who found no significant correlation
1769: between the disk radii and ages.
1770: Most likely, the discrepancy is only caused by uncertainties of the
1771: measured excesses at 24~\textmu{m}. \citet{bryden-et-al-2006} report that the average
1772: photometric accuracy in that filter band is only as good as $1 \sigma_{24}
1773: = 6$\%\ due to stellar photosphere fitting errors and flat-field uncertainties.
1774: Therefore, excesses below those 6\%\ of the photospheric emission cannot be
1775: considered as significant. For 70~\textmu{m}, \citet{bryden-et-al-2006}
1776: state $1 \sigma_{70} \approx 15$\%. Both limits are shown in the upper
1777: panel of Fig.~\ref{fig24/70}.
1778:
1779: In Figs.~\ref{fig2470obs} and \ref{fig24/70}, there is one particular system
1780: directly labeled. That system, HD~72905, was observed to show significant excess
1781: emission not only at 24 and 70~\textmu{m}, but also in the spectral
1782: ranges
1783: 8--13 and 30--34~\textmu{m} of the Spitzer/IRS instrument
1784: \citep{beichman-et-al-2006a}. The presence of two dusty regions was
1785: suggested: one exozodiacal at 0.03--0.43~AU and one around 14~AU.
1786: From the excess at 8--13~\textmu{m}, \citet{wyatt-et-al-2007a}
1787: inferred the dust population in HD~72905 to be transient because the observed
1788: fractional luminosity is above the maximum expected for a system of
1789: 300--400~Myr. As long as only 24 and 70~\textmu{m} are considered,
1790: the HD~72905 dust does not seem particularly hot or bright, although
1791: it is among the hotter disks.
1792:
1793: At this point, it is interesting to compare our results to those of
1794: \citet{wyatt-et-al-2007b}. Both analytic approaches aim at explaining and
1795: reproducing the observations. Our model is different from theirs in that
1796: we take into account the size dependence of the critical specific energy
1797: as well as the transition from a ``primordial'' size distribution of planetesimals to the
1798: one set up by a collisional cascade.
1799: The amount of dust in their model is determined, from the very beginning, by the
1800: rather
1801: long collisional timescales of objects of tens of kilometers, so that the
1802: collisional evolution is much slower. This can be seen from the equations: ``1+'' in the
1803: denominator of Eq.~(\ref{eqMOverTime}) causes the mass to stay almost at the initial
1804: level for a long time, before the system reaches the $t^{-1}$
1805: decay. In our model, although the mass decay is asymptotically slower
1806: ($t^\xi$ with $\xi \approx -0.3 \ldots -0.4$),
1807: it sets up very quickly, namely on collisional timescales of objects with minimum binding
1808: energy ($s_b \sim 100$ meters).
1809: Therefore, we would expect the model by \citet{wyatt-et-al-2007b}
1810: to show significantly larger excesses at ages considered, if all other parameters were
1811: comparable. This, however, is not the case.
1812: \citet{wyatt-et-al-2007b} assumed a much
1813: weaker material in their collisional prescription.
1814: Their $Q\sbs{D}^*=300$~J/kg at an object radius of 30~km ($D\sbs{c} = 60$~km)
1815: is by more than two orders of magnitude below the
1816: values we use in Eq.~(\ref{eqQD}). As $\tau \propto Q\sbs{D}^{q\sbs{p}-1}$ in Eq.~(\ref{eqtaured}),
1817: their collisional timescales are shorter and their evolution faster, too.
1818: Besides the material strength, the difference in the assumed effective
1819: eccentricities --- $e=0.05$ in their model against $e\sbs{max}/2 = 0.15$ in
1820: ours~--- causes another factor of roughly 10 in the collisional timescales,
1821: according to Sect.~\ref{secEcc}.
1822: All the differences listed happen to nearly compensate each other.
1823: As a net result, the excesses predicted by Wyatt's et al. and our models
1824: are comparable with each other (see also Fig.~\ref{fig_fmax}), being in reasonable agreement with the observed ones.
1825:
1826: %===========================================================================
1827: \section{SUMMARY AND CONCLUSIONS}
1828: \label{secConclusions}
1829: %===========================================================================
1830:
1831: We investigated the long-term evolution of debris disks around solar-type (G2V) stars.
1832: Firstly, we performed numerical simulations with our collisional code.
1833: Secondly, the numerical results
1834: were supplemented by, and interpreted through, a new analytic model.
1835: The latter is similar to, and builds up on, the model developed earlier by
1836: \citet{wyatt-et-al-2007a}, but extends it in several important directions.
1837: It naturally includes
1838: the transition from the ``primordial'' size distribution of left-over
1839: planetesimals, set up at their agglomeration phase, to the size distribution
1840: established by the collisional cascade. Further, it lifts the assumption
1841: that the critical specific energy needed for disruption is constant
1842: across the full range of sizes, from dust
1843: to the largest planetesimals. With these improvements, a good agreement between the
1844: numerics and analytics is achieved.
1845:
1846: We draw the following conclusions:
1847: \begin{enumerate}
1848: \item
1849: The timescale of the collisional evolution is inversely proportional to the initial disk mass.
1850: For example, halving the total mass doubles all collisional timescales.
1851: This rule is valid for systems where collisions are the only loss mechanism of particles
1852: and only as long as $\beta$-meteoroids are unimportant for the collisional
1853: budget.
1854: \item
1855: Numerics and analytics consistently yield a $\tau\propto r^{4.3}$
1856: dependence of the timescale of the collisional evolution on the radial
1857: distance.
1858: \item
1859: Numerical simulations show that the collisional timescale
1860: varies with the average eccentricity of dust parent
1861: bodies as $\tau \propto e^{-2.3}$. The analytic
1862: approach suggests a somewhat weaker dependence, $\tau\propto e^{-5/3}$.
1863: \item
1864: An evolving three-slope size distribution is proposed to
1865: approximate the numerical results. The biggest objects are still distributed
1866: primordially, with a slope $q\sbs{p}$. The objects below a certain transitional size
1867: are already reprocessed by collisions and thus have a quasi-steady-state
1868: size distribution, determined by their self-gravity (for intermediate-sized
1869: objects, slope $q\sbs{g}$) or by material strength (for smallest objects, slope $q\sbs{s}$).
1870: That transitional size corresponds to the largest objects
1871: for which the collisional lifetime is still shorter than the age of the system.
1872: The transitional size increases with time, meaning that ever-larger planetesimals
1873: get involved into the collisional cascade.
1874: \item
1875: At actual ages of debris disks, $\sim$10~Myr to $\sim$10~Gyr,
1876: the decay of the dust mass and the total disk mass
1877: follow {\em different} laws. The reason is that, in all conceivable debris disks,
1878: the largest planetesimals have longer collisional lifetimes than the system's age,
1879: and therefore did not have enough time to reach collisional equilibrium.
1880: If the system were let to evolve for sufficiently long time,
1881: both dust mass and disk mass would start to follow $t^{-1}$. However, this requires
1882: time spans of much longer than 10~Gyr.
1883: \item
1884: The loss rate of the dust mass, and the decay rate of fractional luminosity,
1885: primarily depend
1886: on the difference between the slope $q\sbs{p}$ of the ``primordial'' size distribution
1887: of largest planetesimals and the slope $q\sbs{g}$ of the size distribution of
1888: somewhat smaller, yet gravity-dominated, planetesimals that already underwent
1889: sufficient collisional evolution.
1890: With ``standard'' values of $q\sbs{p}$ and $q\sbs{g}$, the dust mass and the thermal fluxes
1891: follow approximately $t^\xi$ with $\xi = -0.3\ldots -0.4$.
1892: \item
1893: Specific decay laws of the total disk mass and the dust mass largely
1894: depend on a few model parameters.
1895: Most important are: the critical fragmentation energy $Q\sbs{D}^*$ as a
1896: function of size,
1897: the slope of the ``primordial'' size distribution of planetesimals $q\sbs{p}$
1898: and their maximum size $s\sbs{max}$,
1899: and the characteristic eccentricity $e$ and inclination $I$ of
1900: planetesimals.
1901: \item
1902: The property that the maximum possible dust luminosity for a given age does
1903: not depend on the initial disk mass, established by \citet{wyatt-et-al-2007a},
1904: is only valid in cases of very rapid collisional evolution, i.e. in closer-in
1905: or dynamically very hot disks. For most of the systems
1906: at ages $<10$~Gyr, an increase of the initial disk mass leads to an increase
1907: of the dust luminosity, unless that initial mass is assigned extreme
1908: values, incompatible with our understanding of planetesimal disks.
1909: \item
1910: Assuming standard material prescriptions and disk masses and extents,
1911: a synthetic population of disks generated with our analytic model generally
1912: agrees with the observed statistics of 24 and 70~\textmu{m} fluxes
1913: versus age.
1914: Similarly, the synthetic [24]-[70] colors are consistent with the
1915: observed disk colors.
1916: \end{enumerate}
1917:
1918: As every model, our numerical model makes a number of general simplifying
1919: assumptions; the analytic one imposes further simplifications:
1920: \begin{itemize}
1921: \item The collisional evolution is assumed to be smooth and
1922: unperturbed.
1923: Singular episodes like the aftermath of giant break-ups or special periods of
1924: the dynamical evolution such as the
1925: late heavy bombardment are not included.
1926: \item
1927: Effects of possible perturbing planets are taken into account
1928: only indirectly: through the eccentricities of planetesimals (dynamical excitation)
1929: and confinement of planetesimal belts (truncation of disks).
1930: Further effects such as resonant trapping or ejection of material by planets are neglected.
1931: \item
1932: We only consider disruptive collisions. This is a reasonable approximation
1933: for disks that are sufficiently ``hot'' dynamically. However, cratering collisions
1934: become important when the relative velocities are insufficient for
1935: disruption to occur.
1936: \item
1937: Neither dilute disks under the regime of Poynting-Robertson drag nor
1938: very dense disks with collisional timescales shorter than orbital timescales
1939: and with avalanches \citep{grigorieva-et-al-2007} are covered by the present
1940: work.
1941: \item
1942: Explaining the initial conditions or deriving them from the dynamical
1943: history of the systems at early stages of planetesimal and planetary accretion
1944: was out of the scope of this paper. Correlations between disk masses,
1945: disk radii, and the presence of planets, for example, were not considered,
1946: although they might alter the scalings we found here.
1947: \end{itemize}
1948:
1949: Despite these limitations, our models reproduce, in essential part,
1950: the observed evolution of dust in debris disks. We hope that they may serve
1951: as a starting point for in-depth studies that will certainly be undertaken in the
1952: future, motivated by questions that remain unanswered, as well as by new data
1953: expected from ongoing and planned observational programs.
1954:
1955: \acknowledgements
1956:
1957: We wish to thank Jean-Fran{\c c}ois Lestrade, Philippe Th\'ebault, and Mark Wyatt for fruitful discussions
1958: and Amaya Moro-Mart{\'i}n for useful review comments.
1959: This research has been partly funded by the Deutsche
1960: Forschungsgemeinschaft (DFG), project Kr 2164/5-1.
1961:
1962: %------------------------------------------------------------------
1963:
1964: %\bibliography{References}
1965: %\bibliographystyle{aa}
1966: \input{ms.bbl}
1967:
1968: \end{document}
1969: