1: \documentclass[12pt,preprint]{aastex}
2:
3: \input psfig.sty
4:
5: \slugcomment{ApJ, in preparation}
6:
7: \shorttitle{}
8: \shortauthors{Rice \& Armitage}
9:
10: \begin{document}
11:
12: \title{Quantifying orbital migration from exoplanet
13: statistics \\ and host metallicities}
14:
15: \author{W.K.M. Rice\altaffilmark{1} and Philip J. Armitage\altaffilmark{2,3}}
16: \altaffiltext{1}{Institute of Geophysics and Planetary Physics and Department
17: of Earth Sciences, University of California, Riverside, CA 92521;
18: ken.rice@ucr.edu}
19: \altaffiltext{2}{JILA, Campus Box 440, University of Colorado, Boulder CO 80309;
20: pja@jilau1.colorado.edu}
21: \altaffiltext{3}{Department of Astrophysical and Planetary Sciences, University of Colorado, Boulder CO 80309}
22:
23: \begin{abstract}
24: We investigate how the statistical distribution of extrasolar planets
25: may be combined with knowledge of the host stars' metallicity to yield
26: constraints on the migration histories of gas giant planets. At any radius,
27: planets that barely manage to form around the lowest metallicity stars accrete
28: their envelopes just as the gas disk is being dissipated, so the lower
29: envelope of planets in a plot of metallicity vs semi-major axis defines
30: a sample of non-migratory planets that will have suffered less than
31: average migration subsequent to gap opening. Under the assumption
32: that metallicity largely controls the initial surface density of
33: planetesimals, we use simplified core accretion models to calculate
34: how the minimum metallicity needed for planet formation varies as a
35: function of semi-major axis. Models that do not include core migration
36: prior to gap opening (Type I migration) predict that the critical metallicity
37: is largely flat between the snow line and $a \approx 6$~AU, with a weak
38: dependence on the initial surface density profile of planetesimals. When
39: slow Type~I migration is included, the critical metallicity is found
40: to increase steadily from 1-10~AU. Large planet samples, that include
41: planets at modestly greater orbital radii than present surveys, therefore
42: have the potential to quantify the extent of migration in both Type~I and
43: Type~II regimes.
44: \end{abstract}
45:
46: \keywords{solar system: formation --- planets and satellites: formation ---
47: planetary systems: formation}
48:
49: \section{Introduction}
50: The discovery of 51 Pegasi \citep{mayor95} provided evidence for the
51: potential importance of orbital migration \citep{goldreich80,lin86} in
52: determining the structure of extrasolar planetary systems. For 51~Peg,
53: and for other members of the class of hot Jupiters, in situ formation
54: scenarios face obvious difficulties due to the predicted high temperature
55: ($T > 10^3 \ {\rm K}$) of the protoplanetary disk \citep{bell97} at
56: the radii -- within 0.1~AU of the star -- where the planets now orbit.
57: This conclusion is largely borne out by detailed models of giant planet
58: formation, which confirm that planet formation via core accretion is
59: unlikely at such small distances from the star \citep{bodenheimer00}.
60: For the much larger population of extrasolar giant planets which orbit
61: at $a > 1$AU, however, the situation is less clear. These planets,
62: which are still orbiting their stars at much smaller radii than
63: Jupiter's 5.2~AU, {\em could} potentially form in situ via core accretion
64: \citep{bodenheimer00}, especially when recent downward revisions to
65: both the radial location \citep{sasselov00} and importance \citep{lodders03}
66: of the snow line are considered. Alternatively, the typical formation
67: radius for giant planets around roughly Solar mass stars could fall
68: beyond the radius of any observed extrasolar planet, implying a
69: dominant role for migration in presently known systems.
70:
71: Observationally, the statistical distributions of basic extrasolar planet
72: properties (mass $M_p \sin (i)$, semi-major axis $a$, and eccentricity $e$)
73: provide only conflicting clues as to the importance of migration. The low
74: masses of some close-in planets point to relatively short residence
75: times within the gaseous protoplanetary disk, since numerical simulations
76: show that Saturn-mass planets accrete gas across gaps and grow to
77: larger masses rather promptly \citep{lubow99,bate03}. If mass growth
78: is ignored, however, the increasing fraction of known extrasolar planets
79: with orbital radius is consistent with numerical models that assume that
80: these planets all formed further out, and migrated inward due to
81: planet-disk interactions \citep{armitage02}. Theoretically, it is well
82: known that the time scale for core accretion to form planets at large radii
83: $a \sim 20$AU is long \citep{pollack96} -- at least in the simplest
84: versions of the theory -- but where in the inner disk planet formation
85: is most favored is uncertain. Theoretical work has demonstrated that the
86: time scale and outcome of core accretion can be modified by Type~I
87: migration \citep{ward97,tanaka02} of giant planet cores \citep{hourigan84,alibert04},
88: by gravitational interaction of cores with turbulent fluctuations in the
89: disk \citep{rice03}, or by competition for planetesimals between several
90: growing cores \citep{hubickyj04}. That this is by no means an exhaustive
91: list of the possibilities illustrates that additional observational
92: clues as to where and when giant planets form would be valuable.
93:
94: In this paper, we investigate how models of giant planet formation via
95: core accretion could be constrained by adding observational knowledge of the
96: metallicity of the host star. The fraction of roughly Solar-type stars
97: that host known extrasolar planetary systems rises rapidly with the
98: stellar metallicity \citep{gonzalez98,santos01,santos04,fischer04,fischer05},
99: with several lines of evidence pointing to the high [Fe/H] being the
100: cause, rather than the consequence, of planet formation \citep{kornet05}. The strikingly
101: strong scaling of planet frequency with metallicity -- which rises from
102: almost zero frequency at ${\rm [Fe/H]} < -0.5$ to
103: $\approx$20\% at ${\rm [Fe/H]} \approx +0.5$ -- suggests that metallicity
104: is a more important parameter in determining the probability of giant
105: planet formation than intrinsic dispersion in either the gas disk mass
106: or disk lifetime. If true, then the epoch of giant planet formation at
107: a given radius in the disk (defined as the moment when the core accretes
108: a significant gaseous envelope via runaway accretion) should correlate
109: with the metallicity. High metallicity implies a larger surface density
110: of planetesimals, and much shorter planet formation time scales \citep{pollack96}.
111: In particular, as illustrated in Figure 1, at any
112: radius there should be a mimumum threshold or critical metallicity,
113: below which core accretion fails to reach runaway within the lifetime
114: of the gas disk. These `failed' gas giants potentially constitute
115: a large population of $\sim 10 M_\oplus$ planets much closer in than
116: Uranus and Neptune in the Solar System \citep{ida04}. More importantly
117: for our purposes, planets just {\em above} the critical metallicity
118: form just as the gas is being dissipated. There is therefore little or
119: no opportunity for these planets to migrate via Type~II gravitational
120: interactions with the protoplanetary disk. Moreover, if the critical
121: metallicity increases monotonically with increasing semi-major axis
122: (and migration is inward, as is very probable at small radii), then
123: there is no way to populate the region of parameter space below the
124: threshold curve. Observational definition of the lowest metallicity
125: stars that host planets at different radii can therefore probe the
126: relative radial efficiency of the planet formation process,
127: independent of the separate uncertainties that attend the migration
128: process.
129:
130: The outline of this paper is as follows. In \S2, we describe
131: simplified models for gas giant planet formation that allow us
132: to calculate the time scale for a single core to reach runaway
133: gas accretion in the protoplanetary disk. In \S3, we compute
134: the threshold metallicity as a function of orbital radius, and
135: discuss how it depends upon the surface density profile of
136: planetesimals and on the extent of Type~I migration of giant
137: planet cores prior to the accretion of the envelope. The
138: most significant differences between the models occur at
139: relatively large radii which, although poorly sampled today,
140: should be accessible to future astrometric and direct
141: imaging surveys. In \S4, we analyze existing statistics of
142: extrasolar planetary systems within the context of our
143: theoretical model. Although useful constraints on planet
144: formation are not possible with existing data, we demonstrate
145: that one of the basic theoretical premises -- that planets
146: observed to lie near the threshold metallicity curve should
147: have formed on average at later epochs -- is consistent with
148: the observed distribution of planetary masses around stars
149: with different [Fe/H].
150:
151: \section{Giant planet formation model}
152: We assume that the massive planets observed in extrasolar planetary
153: systems formed via core accretion, and that the dominant factor
154: controlling whether planets form at a given radius prior to the
155: dispersal of the gas disk is the metallicity of the gas that
156: formed the star and disk. Observations suggest that in most
157: cases the current photospheric stellar metallicity reflects the
158: primordial value \citep{santos04b}, and hence we assume that the
159: ratio of the surface densities of the planetesimals to that of
160: the gas is linearly related to the stellar metallicity. Our
161: goal is to compute as a function of radius the minimum metallicity
162: that allows massive planets -- defined as those with
163: $M_{pl} > 100 M_\oplus$ -- to form within the lifetime of the gas
164: disk. We describe here a simplified model for core accretion, similar
165: to that developed by \citet{ida04}, that accomplishes this objective.
166:
167: \subsection{The gas disk}
168: Within core accretion models of giant planet
169: formation \citep{bodenheimer86, pollack96}, the time
170: dependent evolution of the gaseous component of the protoplanetary disk
171: primarily affects the final masses of planets and the rate of Type~II
172: migration subsequent to gap formation. The {\em threshold} for planet
173: formation, i.e. whether a gas giant can form at all in a given disk,
174: does depend upon the properties of the gas at late epochs (via the
175: pressure and temperature of the disk, which provide boundary conditions
176: for the planetary envelope), but this dependence is weak compared to
177: the effect of changes in the planetesimal surface density. A minimal
178: description of the gaseous disk therefore requires only specification
179: of the surface density profile and lifetime. We adopt for the gaseous
180: surface density a power-law over the range of radii ($1 \ AU < a < 10 \ AU$)
181: of interest,
182: \begin{equation}
183: \Sigma_g = 7.2 \times 10^3 \left( {a \over {1 \ {\rm AU}}} \right)^{-\beta}
184: \ {\rm g \ cm^{-2}},
185: \end{equation}
186: where $a$ is the orbital radius and $\beta$ is a parameter that specifies
187: the mass distribution in the
188: disk (note that we fix the gas surface density at a radius of 1~AU). We
189: consider values of $\beta$ in the range $1 \leq \beta \leq 2$, which
190: includes the most commonly considered possibilities \citep{weidenschilling77,
191: bell97,kuchner04}. This surface density profile is strictly the
192: profile at the time when planetesimals form from the dust within the gas -- since
193: this is likely to occur at an early time it will differ little from the
194: initial profile. We ignore star-to-star variations in the initial disk
195: mass \citep{armitage03}. Provided that these are of less importance
196: than (and uncorrelated with) variations in gas metallicity they should
197: only introduce scatter in the minimum metallicity required for planet
198: formation.
199:
200: For the lifetime of the gas disk, we take $\tau = 5 \times 10^6 \ {\rm yr}$.
201: This value is similar to observational estimates \citep{strom89,haisch01}.
202: We assume that once $t > \tau$, the gas disk is promptly dispersed at
203: all radii and the potential for further gas giant formation is
204: quenched. Observations support the idea that dispersal of the gas
205: occurs rapidly \citep{wolk96}.
206:
207: \subsection{The planetesimal disk}
208: We assume that the surface density of planetesimals, $\Sigma_d$, has a power-law
209: distribution such that
210: \begin{equation}
211: \Sigma_d = 10 f_{dust} \eta_{ice}
212: \left(\frac{a}{1 \mathrm{AU}}\right)^{-\beta} \
213: \mathrm{g \ cm^{-2}},
214: \label{sigma_dust}
215: \end{equation}
216: where $\beta$ has the {\em same} value for the planetesimals as for the
217: gas\footnote{We note that although this is the simplest assumption, it
218: might not be correct, in particular if there is significant radial
219: migration of solids prior to planetesimal formation. Recent models
220: of planetesimal formation via gravitational instability
221: \citep{goldreich73,youdin02,youdin04} require such migration.}.
222: We consider $\beta$ values of
223: $1$, $1.5$, and $2$. $f_{dust}$ is a parameter that we vary to change
224: the normalization of the planetesimal surface density (and stellar
225: metallicity), and $\eta_{ice}$ is a step-function that represents the ice
226: condensation/sublimation across the snow line which occurs, around a star
227: of mass $M_*$, at $a_{ice} = 2.7 (M_*/M_\odot)^2$ AU. Since we will
228: be considering the formation of planets around solar-like
229: stars, we assume $M_* = M_\odot$ and therefore $a_{ice} = 2.7$ AU.
230: Following \citet{ida04} we take $\eta_{ice} = 1$ when $a < a_{ice}$,
231: and $\eta_{ice} = 4.2$ when $a > a_{ice}$.
232:
233: To convert between our parameter $f_{dust}$ and the stellar [Fe/H], we
234: note that if $f_{dust} = 3$, then the metallicity is solar. If $f_{dust}$
235: is greater/less than $3$ the metallicity is greater/less than solar. The
236: opacity of the gas is also important, since there is a weak dependence
237: of the critical core mass, and a stronger dependence of the envelope's
238: Kelvin-Helmholtz time, on $\kappa$. We assume that,
239: \begin{equation}
240: \kappa = {f_{dust} \over 3} \ {\rm cm^2 g^{-1}}
241: \end{equation}
242: so that for solar metallicity the opacity is $1 \ {\rm cm^2 g^{-1}}$.
243:
244: \subsection{Core accretion model}
245: To calculate the growth of a giant planet core within the above disk
246: model, and the subsequent accretion of a gaseous envelope, we employ
247: a modified version of the scheme developed by \citet{ida04}. For
248: each run, we start our calculation with a solid core of mass
249: $M_{core} = 10^{-3} M_\oplus$, density of $3.2 \mathrm{\ g \ cm^{-3}}$,
250: and radius $a$. The core grows at the rate $\dot{M}_{core} = \pi R_c^2 \Sigma_d \Omega F_g$,
251: where $R_c$ is the radius of the core, $\Omega$ is the angular frequency,
252: and $F_g$ is the gravitational enhancement factor \citep{greenzweig92}. The
253: core accretes planetesimals from a annular region around it known as the
254: `feeding zone' that has a full width of $\Delta a_c = 8 r_H$, where
255: \begin{equation}
256: r_H =\left(\frac{M_{pl}}{3 M_*}\right)^{1/3} a
257: \label{rH}
258: \end{equation}
259: is the Hill radius for a planet with mass $M_{pl}$, around a star with mass
260: $M_*$.
261:
262: As the core grows we adjust, accordingly, the planetesimal surface density
263: in the feeding zone. In this way the core self-consistently stops growing
264: once it reaches its isolation mass. In our initial calculations we assume
265: that the core's orbital radius remains fixed, though subsequently we
266: relax this restriction and allow for orbital migration.
267:
268: When the core's mass exceeds a critical value, $M_{crit}$, it is no longer able to
269: support a gaseous envelope in quasi-hydrostatic equilibrium, and
270: runaway gas accretion occurs \citep{mizuno80, bodenheimer86,papaloizou99}. This
271: critical core mass depends on the planetesimal accretion rate, $\dot{M}_{core}$,
272: and on the opacity, $\kappa$, associated with the disk gas. We adopt a
273: representative estimate for the critical core mass calculated by
274: \citet{ikoma00}
275: \begin{equation}
276: M_{crit} = 10 \left(\frac{\dot{M}_{core}}{10^{-6} M_\oplus \mathrm{yr}^{-1}}\right)^
277: {0.25}\left(\frac{\kappa}{1 \mathrm{\ cm^2 \ g^{-1}}}\right)^{0.25} M_\oplus.
278: \label{Mcrit}
279: \end{equation}
280: Once the core exceeds the critical core mass, the gaseous
281: envelope contracts on a Kelvin-Helmholtz timescale, $\tau_{KH}$.
282: \citet{ikoma00} show that the Kelvin-Helmholtz timescale can be written as
283: \begin{equation}
284: \tau_{KH}=10^b\left(\frac{M_{pl}}{M_\oplus}\right)^{-c}\left(\frac{\kappa}{1 \mathrm{\ g \ cm^{-2}}}\right) \mathrm{yr}
285: \label{tKH}
286: \end{equation}
287: where the exact values of $b$ and $c$ depend
288: on the choice of opacity table. \citet{ikoma00} found that
289: $b \simeq 8$ and $c \simeq 2.5$, while \citet{bryden00} obtained a
290: fit to the results of \citet{pollack96} with $b \simeq 10$ and $c \simeq 3$.
291: We follow \citet{ida04} and use $b = 9$ and $c = 3$. Once the critical core
292: mass has been exceeded, we allow gas accretion, at a rate,
293: \begin{equation}
294: \frac{dM_{pl}}{dt} = \frac{M_{pl}}{\tau_{KH}},
295: \label{dMpl}
296: \end{equation}
297: where $M_{pl}$ includes the mass of both the solid core and the gaseous envelope.
298: It should be noted that in this model the core can continue to grow while gas is being added.
299:
300: Equations (\ref{tKH}) and (\ref{dMpl}) define a model for growth of a giant planet's
301: envelope that is entirely demand-driven -- there is no dependence whatsoever on the
302: properties of the gas disk. This is reasonable for low mass planets, but will fail
303: at higher masses when either the supply of gas becomes limited or the planet opens
304: a gap. Since we are interested in planets that are forming just
305: as the disk gas is dissipating, we assume that there is sufficient supply
306: for these planets to reach masses of $\sim 100 M_\oplus$. If runaway growth does occur in
307: our model, we stop the calculations when $M_{pl} > 100 M_\oplus$. This mass
308: threshold (which corresponds to 0.3 Jupiter masses) defines `success' in
309: forming a giant planet.
310:
311: The procedure for modeling planet growth is then as follows. At $t=0$ we start
312: with a $10^{-3} M_\oplus$ core located at a radius $a$. We calculate $\dot{M}_{core}$
313: which we use to determine the core mass at the next timestep. Simultaneously we determine $M_{crit}$.
314: If $M_{core} > M_{crit}$ then gas is accreted with the contraction timescale given by $\tau_{KH}$.
315: We stop the calculation once $t > 5 \times 10^6 \ {\rm yr}$ or once $M_{pl} > 100 M_\oplus$.
316: For a given $\beta$ and a given $a$ we then determine the value of $f_{dust}$ (which within
317: our assumptions is a measure of the metallicity) for which a gas giant planet ($M_{pl} > 100 M_\oplus$)
318: will form in exactly $5 \times 10^6$ yrs. This value of $f_{dust}$ ($f_{dust,min}$) is the minimum
319: value for which a gas giant planet can form prior to the dissipation of the gas disk, and
320: defines a group of planets that should undergo very little Type II migration.
321:
322: \section{Results}
323: For a protoplanetary disk with a specified value of $\beta$, we use the model
324: described above to calculate the mimimum or threshold metallicity required to
325: form a planet at radius $a$ prior to the dissipation of the gas disk.
326: We consider radii between $1$ and $10$ AU and assume, initially, that the core
327: and growing planet suffer negligible Type~I migration during the formation
328: process. Illustrative runs for a planet growing at 5~AU in a disk with
329: $\beta = 1.5$ are shown in Figure~\ref{planprofile}. As the metallicity is
330: increased (i.e. larger values of $f_{\rm dust}$) growth of the planet
331: toward the critical core mass is accelerated. In this case values of $f_{dust} = 2.05$
332: and $f_{dust} = 2.35$ fail to yield a fully-formed giant planet by our
333: definition, as the planet mass after $5$ Myrs is less than $100 M_\oplus$.
334: Slightly higher metallicity ($f_{\rm dust} = 2.45$) however results in
335: runaway runaway growth and produces a $100 M_\oplus$ gas giant planet
336: within 5 Myr.
337:
338: By interpolating from a series of such runs in which the planet formation
339: time scale brackets the assumed disk lifetime, we determine $f_{dust,min}$
340: as a function of orbital radius for each disk model. We consider surface
341: density profiles of $\beta = 1$, $\beta = 1.5$, and $\beta = 2$, in each case
342: normalizing the value of the planetesimal surface density at an
343: arbitrary radius of 1~AU.
344:
345: Figure \ref{nomig} shows the derived values of $f_{dust,min}$ against radius
346: for the three planetesimal surface density profiles that we consider. The most
347: obvious feature in the plots is the sharp rise in the predicted threshold
348: metallicity interior to the radius of the snow line. Although the {\em absolute}
349: values of the threshold metallicity obviously scale with the assumed
350: normalization (and hence the values agree for all disk models at 1~AU), for
351: all models there is a jump of around an order of magnitude across the
352: snow line. Since observationally the frequency of planets around stars
353: significantly more metal-poor than the Sun is very low \citep{gonzalez98,
354: santos01,santos04,fischer04,fischer05}, this implies that gaseous planets
355: cannot form, {\em in situ}, within the snowline (here at 2.7~AU) unless
356: the metallicity is significantly higher than solar. This result is largely
357: consistent with \citet{ida04}, who found that gaseous planets would only
358: form within the snowline for large planetesimal surface densities. Although
359: there are hints of a correlation between planet orbital period and host
360: [Fe/H] \citep{sozzetti04}, it is clear that the observed distribution of
361: extrasolar planets in the $a$-[Fe/H] plane does not resemble Figure~\ref{nomig}.
362: There are a number of planets with semi-major axes $a < 1$~AU around metal-poor
363: stars, and no clear sign of a jump at any plausible snow line radius. At
364: relatively small orbital radii, then, migration appears to be necessary
365: in order to explain the observed statistics of extrasolar planets.
366:
367: Figure \ref{nomig} also shows the behavior of the threshold metallicity with
368: orbital radius beyond the snow line. The minimum metallicity required for
369: gas giant formation depends only weakly on radius out to $\sim 6$~AU for
370: all of the surface density profiles that we have considered, and for the
371: flattest profile ($\beta = 1$) there is little dependence out to larger
372: radii of around 10~AU. This is consistent with the model developed by
373: \citet{pollack96}, which predicts that Jupiter forms at $5.2$ AU in $8$ Myr
374: with a local planetesimal surface density of $10 \mathrm{ \ g \ cm^{-2}}$
375: while Saturn forms within a comparable time scale ($10$~Myr) if the
376: local planetesimal surface density at 9.5~AU is $3 \mathrm{ \ g \ cm^{-2}}$.
377: There is a weak trend for planet formation to be favored at smaller radii
378: around lower metallicity stars \citep{pinotti05} if $\beta = 2$, but in
379: general we would expect that there should not be a strong radial dependence
380: over the range of orbital radii currently accessible to radial velocity
381: surveys (i.e. the threshold metallicity for planet formation at $3$~AU
382: should be similar to that at $6$~AU). At larger radii, however, there is
383: a significant dependence which varies between models. The predicted threshold
384: metallicity rises by a factor of $\approx 2$ between 6~AU and 10~AU if
385: $\beta = 1$, whereas for a steeper $\beta = 2$ profile the rise is closer
386: to a factor of 4. Within the context of core accretion models, measurement
387: of the lower envelope of detected planets in the $a$-[Fe/H] plane at fairly
388: large radii can therefore constrain some inputs to the formation model.
389: Astrometric surveys appear to offer the best possibilities for
390: assembling large enough planet samples at the desired radii.
391:
392: \subsection{Core migration}
393: The above calculations, like the baseline core accretion models presented by
394: \citet{pollack96}, assume that the core grows at fixed orbital radius. If the
395: core remains embedded within a gaseous disk, however, analytic calculations
396: \citep{ward97} show that differential torques arising from the core's
397: gravitational interaction with the gas should induce rapid Type~I migration.
398: The influence of torques from turbulent fluctuations in the disk surface
399: density may also drive orbital drift \citep{nelson04, laughlin04}. Migration
400: in either regime can accelerate gaseous planet formation \citep{hourigan84,
401: alibert04,rice03}, but can also prevent planet formation if the cores
402: migrate into the central star prior to the accretion of the envelope and
403: subsequent gap opening. Here, we consider the possible influence of
404: Type~I core migration on the expected distribution of massive planets in
405: the orbital radius-host metallicity plot.
406:
407: Preliminary calculations showed, as expected, that the survival prospects
408: for cores allowed to migrate at the analytic Type~I rate \citep{ward97} are
409: slim. To allow the core to survive, we therefore assume that the migration,
410: on average, must be slower than the canonical Type I migration rate by about
411: an order of magnitude. This assumption is similar to that made by \citet{alibert04},
412: and may be justified, in part, by numerical simulations that suggest that
413: the true Type~I migration rate may be significantly slower than previously
414: assumed \citep{miyoshi99,jangcondell05}. Specifically, we consider three
415: different core migration rates,
416: \begin{eqnarray}
417: \dot{a} &=& 4 \times 10^{-7} \ {\rm AU \ yr^{-1}} \\
418: \dot{a} &=& 8 \times 10^{-7} \ {\rm AU \ yr^{-1}} \\
419: \dot{a} &=& 1 \times 10^{-7} \left({M_{pl} \over M_\oplus}\right)
420: \left({a \over \mathrm{AU}}\right)^{1/2} \ {\rm AU \ yr^{-1}}.
421: \label{eq_rates}
422: \end{eqnarray}
423: These prescriptions are intended only to sample a range of migratory behavior
424: that allows giant planets to form without being consumed as cores by the star.
425: We do not suggest that any of these rates is representative of the actual core
426: migration rate, although the latter rate has a mass dependence
427: that matches the standard Type~I migration rate. Since we adjust the
428: planetesimal surface density in the feeding zone as the core grows, it is
429: straightforward to add core migration to the model described in section \S2.
430: As before, we vary $f_{dust}$ to determine the threshold metallicity required
431: to form a planet at {\em final} radius $a_{final}$ in $5 \times 10^6$ yr. This
432: requires only an additional iteration to determine the (initially unknown)
433: value of $a_{initial}$ that yields such a planet. Henceforth we consider
434: only surface density profiles of $\beta = 1.5$.
435:
436: Figure \ref{mig} shows $f_{dust,min}$ against $a_{final}$ for the three migration rates
437: that we considered (\ref{eq_rates}). The results for all three migration prescriptions
438: are qualitatively similar. The prominent jump in threshold metallicity at the snow line,
439: seen in the calculations with a static core, is erased in the case of a migrating core
440: which can accrete planetesimals across a much wider range of orbital radii. These
441: smaller threshold values interior to the snow line appear to be in better accord
442: with observations. Moreover, at larger radii there is now a steady increase in the
443: predicted threshold metallicity with radius, even at radii accessible to ongoing
444: radial velocity surveys for extrasolar planets. If Type~I migration is implicated
445: in the formation of gas giants, we would therefore expect to see a steady rise
446: in the minimum host metallicity as the orbital radius of the planet increases.
447:
448: Figure \ref{afin} shows $a_{initial}$ against $a_{final}$ for the non-constant migration rate,
449: and illustrates the amount of migration that has taken place.
450: For planets that remain beyond the snowline ($a_{final} > 2.7$ AU), the
451: change in semi-major axis is largely independent of radius ($\Delta a \sim 3$ AU). The radial
452: dependence of the migration rate appears to be balanced by the radial dependence of planet's
453: growth rate. For planets that end up within the snowline ($a_{final} < 2.7$ AU), $a_{initial}$
454: appears to be almost constant, with $\Delta a \sim 4$ AU for $a_{final} = 1$ AU, and $\Delta
455: a \sim 3$ AU for $a_{final} = 2$ AU. This suggests that the surface density discontinuity at the
456: snowline can produce a large change in $a_{final}$ for a small change in $a_{initial}$. Although
457: our chosen migration rates are somewhat arbitrary, these results suggest that migration rates of
458: between $2$ and $4$ AU in $5 \times 10^6$ yrs can produce planets within the snowline around stars with
459: reasonably low metallicities, and can result in an increase in the threshold metallicity with
460: radius.
461:
462: By focusing on the shape of the critical metallicity curve, we have deliberately
463: avoided detailed discussion of the {\em relative number} of planets expected to populate
464: different regions of the [Fe/H]-$a$-$M_{pl}$ space. This depends upon the fraction
465: of plausible initial conditions that yield each specific outcome \citep{armitage02,ida04}.
466: We note, however, that such statistical considerations may yield additional
467: evidence for Type~I migration of gas giant cores. In particular, \citet{ida04} predict
468: a dearth of planets with masses between $10$ and $100 M_\oplus$ within the snowline,
469: a region they term the ``planet desert''. In their model, a planet can only fall
470: within the desert if its growth is halted during the rapid envelope growth phase.
471: This is an unlikely outcome since this phase lasts for such a short time. Equivalently,
472: in the absence of core migration the range of metallicity values $\Delta f_{dust}$
473: that populate the desert region is small compared to $f_{dust,min}$, and as a
474: consequence few planets are predicted in the desert.
475:
476: When core migration is included, this prediction of an unoccupied planet desert can
477: be modified, though the results do depend on the details of the migration history. As
478: shown in Figure \ref{mig}, $f_{dust,min}$ at small orbital radii is greatly reduced
479: in the presence of migration. Compared to this threshold value, we find a reasonably
480: large range of $f_{dust}$ values that result in planet masses that fall within the
481: ``planet desert''. For the non-constant migration rate,
482: $f_{dust} = 1.3$ yields a $100 M_\oplus$ planet at $1$ AU within $5$ Myr, while
483: $f_{dust} = 1.1$ produces a $48 M_\oplus$ planet after $5$ Myr. Roughly 20\% of
484: stars with metallicities close to the threshold value would then be candidates for
485: forming planets in the otherwise (in the absence of migration) unoccupied region
486: of phase space. Unfortunately, the range of $f_{dust}$ values that resulted in
487: planets with masses within the desert region was smaller for the constant migration
488: rates than for the non-constant migration rate. This suggests that for planets to lie
489: within the desert, the core migration rate must be such as to maintain a large
490: planetesimal accretion rate, $\dot{M}_{\rm core}$, resulting in a large critical core
491: mass (see equation \ref{Mcrit}). Therefore, although the existence of planets
492: within the desert region may point to the role of Type~I migration, more
493: detailed modeling will be required to establish this clearly.
494:
495: \section{Application to current data}
496: To compare these expectations with current observations of extrasolar
497: planetary systems, we use the metallicity data published by \cite{santos04}.
498: The \cite{santos04} catalog includes [Fe/H] measurements, derived using
499: uniform analysis methods, for 98 planet host stars. For our purposes we
500: include all planets in multiple planet systems, and take the average value
501: of [Fe/H] for those stars with more than one derived abundance. HD47536b and
502: its host star are excluded due to the large uncertainty in the mass of that
503: planet, leaving 110 massive planets for which we have the host star
504: metallicity, together with the planet mass $m_p \sin(i)$, semi-major axis and
505: eccentricity. Fig.~\ref{FeHvsa} shows the distribution of these planets
506: in the $a$-[Fe/H] plane. Although there are some hints that this distribution
507: is not merely a scatter plot (for example there are no observed extrasolar
508: planets at $a > 1 \ {\rm AU}$ around stars with [Fe/H]~$<$-0.4), the lower
509: envelope of the distribution which has been the focus of the theoretical
510: discussion is obviously poorly defined in current data due to the small
511: number of planets observed that are unambiguously outside the snow line.
512: Nevertheless, we can use the data for the more limited purpose of testing
513: whether some of our basic assumptions are consistent with observations.
514:
515: Theoretically, we expect that planets observed near the critical metallicity
516: line formed late in the
517: lifetime of the gaseous protoplanetary disk. We would expect them to have
518: lower masses -- since there was less time available to accrete gas before
519: the disk was dissipated -- and possibly different eccentricity. To test
520: whether these expectations are consistent with current data, we divide
521: the planets with semi-major axis $a > 0.5$~AU (i.e. excluding the hot
522: Jupiters, but probably including some planets that are now seen interior
523: to the snow line) into high and low host star metallicity samples. Some
524: care is necessary, because the existing sample of planets discovered via
525: radial velocity is biased against the discovery of low mass planets at
526: large radii. Specifically, for a fixed number of radial velocity
527: measurements with a given noise level (and some implicit assumptions as
528: to the time sampling of the survey), the minimum detectable planet
529: mass will scale with orbital radius roughly as $[m_p \sin(i)]_{\rm min}
530: \propto a^{1/2}$. This detection bias means that simply dividing
531: the sample of planets into two -- those with host stars above and below some
532: fixed value of [Fe/H] -- risks mixing a variation of host metallicity with
533: planet orbital radius into a spurious trend of planet mass with host
534: metallicity.
535:
536: To avoid such potential biases, we construct matched samples of planets
537: around high and low metallicity host stars. We first bin planets
538: according to their semi-major axis (in 0.5~AU increments between
539: 0.5~AU and 3~AU, plus one final bin from 3-4~AU). Within each radial bin,
540: the minimum detectable planet mass in a radial velocity survey
541: varies by much less than the intrinsic dispersion in observed planet masses.
542: We then split the planets {\em within each radial bin} into
543: subsamples of high and low metallicity host stars, with (as far as is
544: possible) equal numbers of planets in each. Our final sample of planets
545: around relatively low (high) metallicity stars then consists of all the
546: planets below (above) the dashed lines shown in Fig.~\ref{FeHvsa}.
547:
548: The mass distribution of the planets in the high and low metallicity
549: samples, defined as above, is shown in Fig.~\ref{fig_KS}. The two
550: distributions are consistent with the expectation that the planets
551: around lower metallicity stars have, on average, lower masses. Formally,
552: a Kolmogorov-Smirnov test shows that the probability that the two
553: distributions are drawn from the same parent distribution is
554: $P_{\rm KS} = 6 \times 10^{-3}$, so the statistical evidence in support
555: of the existence of the hypothesized `no-migration' curve in the $a$-[Fe/H]
556: plane is currently suggestive rather than overwhelming. However, the fact
557: that some indications of a signal are present in current data does suggest
558: that the larger planet samples that will plausibly be accumulated in the
559: near future should suffice to test some of the ideas outlined in the
560: preceding Section.
561:
562: Mindful of the fact that in some theoretical models orbital migration
563: is associated with concurrent eccentricity growth \citep{papaloizou01,
564: murray02,goldreich03,ogilvie03}, we have tested whether the eccentricity
565: distribution of the high and low metallicity samples is significantly
566: different. Unlike in the case of the $m_p \sin(i)$ distributions, the
567: eccentricity distributions of the two samples are statistically
568: indistinguishable. Within the context of our model there
569: is no evidence that relatively early planet formation, followed by
570: significant radial migration, promotes growth of the final eccentricity.
571:
572: \section{Summary}
573: In this paper we have investigated how, with the addition of knowledge
574: of the host stars' metallicity, the statistics of extrasolar planets
575: can be used to constrain models for giant planet formation. Our main
576: results are:
577: \begin{itemize}
578: \item[1]
579: By studying planets at a particular radius around the lowest metallicity
580: host stars, it is possible to isolate a subsample that will have suffered,
581: on average, less Type~II migration than the typical planet at that
582: radius. This separation will only be clean outside the snow line, and
583: only if metallicity is the most important random variable affecting the
584: time scale for core formation. Comparison of such a non-migratory
585: sample with planets around metal-rich stars could constrain the amount
586: of mass accreted onto the planet during Type~II migration.
587: \item[2]
588: If giant planet cores grow in place, it is difficult to explain the
589: presence of massive extrasolar planets around relatively low metallicity
590: hosts within the snow line. Static core growth models predict a threshold
591: metallicity for planet formation that is roughly flat within 6~AU, but which
592: rises to larger radius. The functional dependence varies according to the slope
593: of the planetesimal surface density distribution.
594: \item[3]
595: If giant planet cores suffer Type~I migration as they grow, the
596: threshold metallicity rises smoothly beyond $a \approx 2$~AU. The
597: details of the migration are relatively unimportant, provided that
598: the overall rate is slow compared to the canonical analytic
599: predictions. Under limited conditions, it is possible to populate
600: what would otherwise be a desert in the distribution of planets
601: with masses $10 \ M_\oplus < M_{pl} < 100 \ M_\oplus$ at small
602: orbital radii \citep{ida04}.
603: \end{itemize}
604: Existing observations appear to be consistent with the basic theoretical
605: premise of this paper---that host metallicity is the dominant factor
606: controlling the time scale for massive planet formation. This hypothesis
607: is motivated by the observed dependence of planet frequency on stellar
608: [Fe/H] \citep{gonzalez98,santos01,santos04,fischer04,fischer05}, and
609: implies a planet mass - metallicity correlation that is seen, albeit
610: at low significance, in the data. This leaves us hopeful that with
611: larger samples of massive extrasolar planets useful constraints on
612: planet formation models will be attainable.
613:
614: \acknowledgements
615:
616: This work was supported by NASA under grants NAG5-13207 and NNG04GL01G
617: from the Origins of Solar Systems and Astrophysics Theory Programs, and
618: by the NSF under grant AST~0407040. The hospitality of the Aspen Center for
619: Physics, where part of this paper was completed, is gratefully acknowledged.
620:
621: \begin{thebibliography}{}
622:
623: \bibitem[Alibert, Mordasini \& Benz(2004)]{alibert04}
624: Alibert, Y., Mordasini, C., \& Benz, W. 2004, A\&A, 417, L25
625:
626: \bibitem[Armitage, Clarke \& Palla(2003)]{armitage03}
627: Armitage, P. J., Clarke, C. J., \& Palla, F. 2003, MNRAS, 342, 1139
628:
629: \bibitem[Armitage et al.(2002)]{armitage02}
630: Armitage, P. J., Livio, M., Lubow, S. H., \& Pringle, J. E. 2002,
631: MNRAS, 334, 248
632:
633: \bibitem[Bate et al.(2003)]{bate03}
634: Bate, M. R., Lubow, S. H., Ogilvie, G. I., \& Miller, K. A. 2003,
635: MNRAS, 341, 213
636:
637: \bibitem[Bell et al.(1997)]{bell97}
638: Bell, K. R., Cassen, P. M., Klahr, H. H., \& Henning, Th. 1997, ApJ, 486, 372
639:
640: \bibitem[Bodenheimer, Hubickyj \& Lissauer(2000)]{bodenheimer00}
641: Bodenheimer, P., Hubickyj, O. \& Lissauer, J. J. 2000, Icarus, 143, 2
642:
643: \bibitem[Bodenheimer \& Pollack(1986)]{bodenheimer86}Bodenheimer, P.,
644: \& Pollack, J.B. 1986, Icarus, 67, 391
645:
646: \bibitem[Bryden, Lin \& Ida(2000)]{bryden00}
647: Bryden, G., Lin, D.N.C., \& Ida, S. 2000, \apj, 544, 481
648:
649: \bibitem[Fischer, Valenti \& Marcy(2004)]{fischer04}
650: Fischer, D., Valenti, J. A., \& Marcy, G. 2004, in ``Stars as Suns: Activity,
651: Evolution \& Planets'', IAU Symposium 219, ed. A.K. Dupree, ASP Conf. Ser., in press
652:
653: \bibitem[Fischer \& Valenti(2005)]{fischer05}
654: Fischer, D. A., \& Valenti, J. 2005, ApJ, 622, 1102
655:
656: \bibitem[Goldreich \& Sari(2003)]{goldreich03}
657: Goldreich, P., \& Sari, R. 2003, ApJ, 585, 1024
658:
659: \bibitem[Goldreich \& Tremaine(1980)]{goldreich80}
660: Goldreich, P., \& Tremaine, S. 1980, ApJ, 241, 425
661:
662: \bibitem[Goldreich \& Ward(1973)]{goldreich73}
663: Goldreich, P.; Ward, W. R. 1973, \apj, 183, 1051
664:
665: \bibitem[Gonzalez(1998)]{gonzalez98}
666: Gonzalez, G. 1998, A\&A, 334, 221
667:
668: \bibitem[Haisch, Lada \& Lada(2001)]{haisch01}
669: Haisch, K. E., Lada, E. A., Lada, C. J. 2001, \apj, 553, L153
670:
671: \bibitem[Greenzweig \& Lissauer(1992)]{greenzweig92}Greenzweig, Y., \&
672: Lissauer, J.J. 1992, Icarus, 100, 440
673:
674: \bibitem[Hourigan \& Ward(1984)]{hourigan84}
675: Hourigan, K., \& Ward, W. R. 1984, Icarus, 60, 29
676:
677: \bibitem[Hubickyj, Bodenheimer \& Lissauer(2004)]{hubickyj04}
678: Hubickyj, O., Bodenheimer, P., \& Lissauer, J. J. 2004, in
679: "Gravitational Collapse: From Massive Stars to Planets", eds G. García-Segura,
680: G. Tenorio-Tagle, J. Franco, \& H. W. Yorke, Rev. Mex. Astron. Astrophys.
681: (Conf. Ser.) 22, p.~83
682:
683: \bibitem[Ida \& Lin(2004)]{ida04}
684: Ida, S., \& Lin, D. N. C. 2004, ApJ, 604, 388
685:
686: \bibitem[Ikoma et al.(2000)]{ikoma00}
687: Ikoma, M., Nakazawa, K., \& Emori, H. 2000, \apj, 537, 1013
688:
689: \bibitem[Kornet et al.(2005)]{kornet05}
690: Kornet, K., Bodenheimer, P., Rozyczka, M., \& Stepinski, T.F. 2005, A\&A, 430, 1133
691:
692: \bibitem[Kuchner(2004)]{kuchner04}
693: Kuchner, M. 2004, \apj, 612, 1147
694:
695: \bibitem[Jang-Condell \& Sasselov(2005)]{jangcondell05} Jang-Condell, H., \&
696: Sasselov, D.D. 2005, \apj, 619, 1123
697:
698: \bibitem[Laughlin, Steinacker, \& Adams(2004)]{laughlin04} Laughlin, G.,
699: Steinacker, A., \& Adams, F.C. 2004, \apj, 608, 489
700:
701: \bibitem[Lin \& Papaloizou(1986)]{lin86}
702: Lin, D. N. C., \& Papaloizou, J. 1986, ApJ, 309, 846
703:
704: \bibitem[Lodders(2003)]{lodders03}
705: Lodders, K. 2003, ApJ, 591, 1220
706:
707: \bibitem[Lubow, Seibert \& Artymowicz(1999)]{lubow99}
708: Lubow, S. H., Seibert, M., \& Artymowicz, P. 1999, ApJ, 526, 1001
709:
710: \bibitem[Mayor \& Queloz(1995)]{mayor95}
711: Mayor, M., \& Queloz, D. 1995, Nature, 378, 355
712:
713: \bibitem[Miyoshi et al.(1999)]{miyoshi99}
714: Miyoshi, K., Takeuchi, T., Tanaka, H., \& Ida, S. 1999, \apj, 516, 451
715:
716: \bibitem[Murray, Paskowitz \& Holman(2002)]{murray02}
717: Murray, N., Paskowitz, M., \& Holman, M. 2002, ApJ, 565, 608
718:
719: \bibitem[Mizuno(1980)]{mizuno80} Mizuno, H. 1980, Prog. Theor. Phys., 64, 544
720:
721: \bibitem[Nelson \& Papaloizou(2004)]{nelson04} Nelson, R.P., \& Papaloizou, J.C.B. 2004,
722: \mnras, 350, 849
723:
724: \bibitem[Papaloizou, Nelson \& Masset(2001)]{papaloizou01}
725: Papaloizou, J.C.B., Nelson, R.P., \& Masset, F. 2001, A\&A, 366, 263
726:
727: \bibitem[Tanaka, Takeuchi \& Ward(2002)]{tanaka02}
728: Tanaka, H., Takeuchi, T., \& Ward, W. R. 2002, ApJ, 565, 1257
729:
730: \bibitem[Papaloizou \& Terquem(1999)]{papaloizou99} Papaloizou, J.C.B., \&
731: Terquem, C. 1999, \apj, 521, 823
732:
733: \bibitem[Pinotti et al.(2005)]{pinotti05} Pinotti, R., Arany-Prado, L.,
734: Lyra, W., \& Porto de Mello, G.F. 2005, \mnras, submitted
735:
736: \bibitem[Pollack et al.(1996)]{pollack96}
737: Pollack, J. B., Hubickyj, O., Bodenheimer, P., Lissauer, J. J.,
738: Podolak, M., \& Greenzweig, Y. 1996, Icarus, 124, 62
739:
740: \bibitem[Ogilvie \& Lubow(2003)]{ogilvie03}
741: Ogilvie, G.I., \& Lubow, S.H. 2003, \apj, 587, 398
742:
743: \bibitem[Rice \& Armitage(2003)]{rice03}
744: Rice, W. K. M., \& Armitage, P. J. 2003, ApJ, 598, L55
745:
746: \bibitem[Santos, Israelian \& Mayor(2001)]{santos01}
747: Santos, N. C., Israelian, G., \& Mayor, M. 2001, A\&A, 373, 1019
748:
749: \bibitem[Santos et al.(2003)]{santos03} Santos, N.C., Israelian, G., Mayor, M.,
750: \& Rebolo, R. 2003, A\&A, 398, 363
751:
752: \bibitem[Santos et al.(2004)]{santos04b}
753: Santos, N. C., Israelian, G., Garc\'ia L\'opez, R. J., Mayor, M., Rebolo, R.,
754: Randich, S., Ecuvillon, A., \& Dom\'inguez Cerde\~na, C. 2004, A\&A, 427, 1085
755:
756: \bibitem[Santos, Israelian \& Mayor(2004)]{santos04}
757: Santos, N. C., Israelian, G., \& Mayor, M. 2004, A\&A, 415, 1153
758:
759: \bibitem[Sasselov \& Lecar(2000)]{sasselov00}
760: Sasselov, D. D., \& Lecar, M. 2000, ApJ, 528, 995
761:
762: \bibitem[Sozzetti(2004)]{sozzetti04}
763: Sozzetti, A. 2004, \mnras, 354, 1194
764:
765: \bibitem[Strom et al.(1989)]{strom89}
766: Strom, K. M., Strom, S. E., Edwards, S., Cabrit, S., \& Skrutskie, M. F. 1989,
767: \aj, 97, 1451
768:
769: \bibitem[Ward(1997)]{ward97}
770: Ward, W. R. 1997, Icarus, 126, 621
771:
772: \bibitem[Weidenschilling(1977)]{weidenschilling77}
773: Weidenschilling, S. J. 1977, Astrophysics and Space Science, 51, 153
774:
775: \bibitem[Wolk \& Walter(1996)]{wolk96}
776: Wolk, S. J., \& Walter, F. M. 1996, \aj, 111, 2066
777:
778: \bibitem[Youdin \& Chiang(2004)]{youdin04}
779: Youdin, A. N., \& Chiang E. I. 2004, \apj, 601, 1109
780:
781: \bibitem[Youdin \& Shu(2002)]{youdin02}
782: Youdin, A. N., \& Shu, F. H. 2002, \apj, 580, 494
783:
784: \end{thebibliography}
785:
786: \newpage
787:
788: \begin{figure}
789: \plotone{f1.eps}
790: \figcaption{Illustration of the expected influence of core formation
791: and orbital migration on the final distribution of extrasolar planets.
792: In an idealized model, in which the initial mass and lifetime of the gas
793: disk have a narrow range of values, giant planet formation will occur at
794: a given radius provided that the host star's metallicity exceeds a threshold
795: value. At this threshold value, runaway occurs and the gaseous envelope is
796: accreted just as the protoplanetary disk is about to be dissipated, allowing
797: no opportunity for Type~II migration. For higher host metallicities, Type~II
798: migration is expected to become increasingly important. If the threshold
799: metallicity is an increasing function of radius -- in practice at radii
800: outside the snow line -- then inward migration {\em cannot} populate the
801: unshaded region below the critical curve. The shape of this curve then defines
802: a threshold for giant planet formation that is independent of the effects
803: of orbital migration.}
804: \label{fig_schematic}
805: \end{figure}
806:
807: \begin{figure}
808: \plotone{f2.ps}
809: \caption{Growth curves for a core located at $a = 5$ AU and for three
810: different planetesimal surface densities, as defined by $f_{dust}$. The different
811: $f_{dust}$ values are represented by different line thicknesses, and in each case
812: the solid line shows the core mass while the dashed line is the total planet mass (core + envelope).
813: For $f_{dust} = 2.45$
814: a giant planet forms within $5$ Myr, while for lower values of $f_{dust}$ gaseous planet
815: formation does not occur within the disk lifetime.}
816: \label{planprofile}
817: \end{figure}
818:
819: \begin{figure}
820: \plotone{f3.ps}
821: \caption{The minimum value of $f_{dust}$ required to form a gaseous planet, {\em in situ}, within $5$ Myr.
822: This illustrates the difficulty in forming gaseous planets within the snowline ($\sim 2.7$ AU), since the
823: required planetesimal surface density is extremely high, and shows that even for reasonably steep surface density profiles,
824: the radial dependence beyond the snowline is relatively weak out to $\sim 7$ AU. }
825: \label{nomig}
826: \end{figure}
827:
828: \begin{figure}
829: \plotone{f4.ps}
830: \caption{The minimum value of $f_{dust}$ required to form a gaseous planet if the core is assumed to
831: migrate to a final radius $a_{final}$. We consider three different migration rates and find that
832: in all cases gaseous planets can have final semi-major axes within the snowline for $f_{dust}$ values that
833: are significantly smaller than that required when core migration is ignored. This figure also suggests that
834: if core migration plays a role in gaseous planet formation the metallicity required may increase with increasing
835: $a_{final}$.}
836: \label{mig}
837: \end{figure}
838:
839: \begin{figure}
840: \plotone{f5.ps}
841: \caption{Initial semi-major axis ($a_{initial}$) against final semi-major axis ($a_{final}$) for
842: the non-constant migration rate. For planets that remain beyond the snowline the change in semi-major axis
843: is almost constant despite the radial dependence of the migration rate. For planets that end up
844: within the snowline, $a_{initial}$ appears almost constant suggesting that the surface density
845: discontinuity at the snowline can produce large changes in $a_{final}$ for small changes in $a_{initial}$.}
846: \label{afin}
847: \end{figure}
848:
849: \begin{figure}
850: \plotone{f6.eps}
851: \caption{Distribution of extrasolar planets in $a$-[Fe/H] plane. Although there
852: is insufficient data to strongly constrain the relationship between semi-major axis and
853: metallicity, there is a hint that the metallicity required for giant planet formation
854: increases with increasing radius. The divisions within the figure are used to divide the
855: planets within each radius bin into high- and low-metallicity samples.}
856: \label{FeHvsa}
857: \end{figure}
858:
859: \begin{figure}
860: \plotone{f7.eps}
861: \caption{Mass distribution of the planets in the high and low metallicity samples defined by
862: the divisions in Figure \ref{FeHvsa}. The two distributions are consistent with the expectation that
863: the planets around lower metallicity stars (dashed line) have, on average, lower masses. This is
864: consistent with the idea that planets around lower metallicty stars are likely to have formation times
865: comparable to the disk lifetimes, and are unlikely to undergo significant Type II migration.}
866: \label{fig_KS}
867: \end{figure}
868:
869:
870: \end{document}
871:
872: