Planet Formation

lecture by Roy van Boekel (MPIA)

suggested reading:
“LECTURE NOTES ON THE FORMATION AND EARLY
EVOLUTION OF PLANETARY SYSTEMS”

by Philip J. Armitage
http://arxiv.org/abs/astro-ph/0701485

with material from

Kees Dullemond, Christoph Mordasini, Til Birnstiel

1
 the scales
from ~13,000 km

to

~140,000 km
~1 µm

in or

~30,000,000,000 km (~100 AU) 2

star formation vs. planet formation

• ~spherical collapse, “own • stellar gravity usually

gravity” dominant dominant

• rotation, shed very much • Keplerian shear, much less

angular momentum net angular momentum loss

• cooling • cooling!

• (elemental) composition • composition can (locally) be

highly enriched in heavy
~interstellar/solar
elements/“dust”
• scales: ~0.1 pc —> ~1 RSun
• scales: <1 to ~several AU —>
~1 REarth to ~1 RJupiter
3
 ler terrestrial planets
m Mercury, all of the iron) in each planet, and augment this mass with

Minimum Mass Solar Nebula

d orbital inclinations.
oximately, but not ex-
enough hydrogen and helium to bring the mixture
to Solar composition. This is a mild augmentation
otation axis (the mis- for Jupiter, but a lot more for the Earth.
“What is the minimum amount
2. Divide theofSolar
disk material
System to make
into the solar
annuli, with one
system
nd terrestrial planets?” Basic
planets idea:
planet per annulus. Distribute the augmented mass
adius, with the inner for each planet uniformly across the annuli, to
(1)separated
anets being Consider the disk yield
region from which each
a characteristic planet (given
gas surface density Σ (units
by the maincurrent
asteroidmass / location)
g cm−2would
) at theaccrete
locationmaterial
of each planet.
planets coincide with
(2)have
ry disk to Consider The result
been the amount is that between
of refractory elementsVenus and Neptune
in each planet (and
−3/2
(add ignoring
volatiles/ices
present. This is a sig- the
beyondasteroid
iceline)belt) Σ ∝ r . The precise nor-
theory of giant planet malization is mostly a matter of convention, but if one
for giant diskneeds
(3) assume
e time scale a specific number
bulk composition the most
is solar; common
add H+He gasvalue used is
accordingly that due to Hayashi (1981),
e mass of condensible
pect faster growth to # r $−3/2
Result (Hayashi
e protoplanetary disk. 1981): Σ = 1.7 × 103 g cm−2 . (4)
AU
Σ is gas surface density (~total dens., H+He gas ~99% of mass)
This estimate relies on several assumptions and is only an
approximate result
1981IAUS...9
Minimum Mass Solar Nebula
Hayashi (1981)

5
Minimum Mass Solar Nebula
more recent rendition, ~equivalent to previous plot

Ruden (2000)
(Giant) Planet Formation

Two main theories of planet formation:

(1) Core accretion: formation of solid core (terrestrial planet),

if core sufficiently massive (~8 M⊕) subsequent accretion
of gas (gas giant planet).

(2) Gravitational Instability: direct collapse from gas phase

(only gas giant planets, relatively far out in the disk)

7
 Gravitation Instability (GI)
Main idea:

(1) instability causes initial over-

density, subject to self-gravity

(2) if these “clumps” can get rid

of potential energy faster
than pressure and differential
rotation smooth them out
again, they can collapse

(3) fast process; planet

composition ~ (local) disk
bulk composition W.K.M. Rice, P.J. Armitage, M.R. Bate &
I.A. Bonnell, MNRAS, 339, 1025 (2003)
8
 Gravitation Instability (GI)
We have already derived the necessary conditions for angu
gravitationalCriterion
Safronov-Toomre instability to occur.
for disk We against
stability need the Toomre
isothermal dens
Q in
collapse parameter
kepleriantodisk:
be low enough, specifically, 1990
Gi
cs Ω Gam
Q≡ > Qcrit ≃ 1
< (213)
πGΣ num
contr
where cs is the sound speed in a gas disk of local sur-
Q is “Toomre parameter”, unst
face density Σ andcs the
is sound
disk speed
mass is[cm s-1], Ω istoorbital
assumed be
frequency we ca
small[rad s-1], Gthat
enough = gravitational constant
the distinction [cm3 gthe
between -1 s-2], Σ is
orbital
surfaceand
density -2].
[g cmfrequencies scale
epicyclic is of little import. If we con-
sider a disk with h/r = 0.05 at 10 AU around a Solar
Disk will stable
mass star,against fragmentation
then the relation h/rwhere Q >yields
= cs /v Qcrit. Typical
a sound
φ
value Qcrit ≈ 1. Having Q < Q is necessary but not sufficient
speed cs ≃ 0.5 kms−1 . crit To attain Q = 1, we then require
condition for fragmentation;
a surface density, wher
unit
Σ ≈ 1.59× 103 gcm2 . (214) with
 Gravitation Instability (GI)
Example:
h/r = 0.05 at 10 AU around solar-type star;
h/r = cs/vφ —> sound speed cs≃0.5 km s−1. (vφ is orbital velocity)

To get Q < 1 we require Σ > 1500 g cm-2.

Compare to “minimum mass solar nebula”: Σ≃54 g cm-2 @ 10 AU

(using normalization of Hayashi (1981))

—> GI works only for very massive disks

10
 fo
unit surfa
much larger Σ than estimates
≈ 1.5 × 103 gcm2 .
based, for example,
(214) with γ =
minimum mass Solar Nebula, from which we con-
his is much
bustly GI: resulting planets
that larger than estimates
gravitational based, forisexample,
instability most likely
for fragm
n the minimum mass Solar Nebula, from which we con- • tcoo
rude
at robustly
an early
spatial epoch
scale most when
unstable the
to disk
collapse: mass
that gravitational instability is most likely
is still high.
ngoccur
thatat the characteristic
an early epoch when thewavelength forhigh.
disk mass is still gravita- • tcoo
nstability is λ =
ecalling that the characteristic 2c 2
/(GΣ), we find that the hea
crit s wavelength for gravita- lenc
2
onal instability
objects formed crit is λ = 2c /(GΣ), we find
if suchs a disk fragmented would that the
ass ofresulting
objectsplanet
formed if such a disk fragmented would
T
mass if such a disk region would collapse (for This our con
e, example with Σ = 1500 g cm-2, at 10 AU with cs=0.5 km s-1): sulting su fr
4 within w
th
22 ∼ 4πc
4πc 4
ss nitude n Ω
Mp ∼p πΣλcrit
M ∼ πΣλ crit∼
G 22Σ
∼∼6
5M5MJ J (215) (215)
G Σ shorter shth
here MJ is the mass of Jupiter. These order of magni- sion can
M is
—>the
GI mass
produces of Jupiter.
very massive These
planets. order
Works of
better inmagni-
outer disk
pressure si
ude J estimates suffice to indicate that gravitational insta-
imates suffice
lity followed to indicate could
by fragmentation that form
gravitational
gas giants. insta- rived p
loca
11
it provide
 1990).
Given these consideration, when will a disk fragment?
13) Importance of cooling
Gammie (2001) used both analytic arguments and local
numerical simulations to identify the cooling time as the
(additional requirement on top of Toomre criterion)
control parameter determining whether a gravitationally
ur-
unstable disk will fragment. For an annulus of the disk
be Collapsing fragment —> release
we can define the equivalent of the of gravitational energy,
Kelvin-Helmholtz time needs to
tal be radiated away sufficiently quickly for collapse to proceed.
scale for a star,
on- Cooling time:
lar U
nd tcool = 4 (217)
2σTdisk
ire
where U
where U isis the
thethermal
thermalenergy
energycontent
contentofofthe
thedisk
disk
perper
unit surface
unit surface area. Then for an ideal gas equation of state
14) area. For an ideal gas EOS (𝞬 = 5/3) we get:
with γ = 5/3 Gammie (2001) found that the boundary
for fragmentation is:
ple, • tcool︎ ≳ 3Ω−1 — the disk fragments.
−1
on- • tcool︎ ≲ 3Ω — disk—reaches
• t cool !
−1 3Ω the diska fragments.
steady state in which heating due
ely to dissipation
gh. • tcool " 3Ωof −1gravitational turbulence balances cooling.
— disk reaches a steady state in which
ta- heating due to dissipation of gravitational turbu-
the lence balances cooling. 12
 Is this (GI) how it goes?
(1) vast majority of planets (solar system, exoplanets
discovered mainly with transit & radial-velocity
techniques) are less massive than GI predicts

(2) In solar system, gas/ice giants are enriched in heavy

elements; hard to reconcile with GI.

(3) some “direct imaging” observations

found (young) planets at large
distances. Possibly these formed
through GI (“jury is still out”).

13
 core-accretion (CA)
Basic idea:

(1) solid “refractory” material comes together to form larger

bodies. Beyond “snowline” this includes (water-) ice.

(2) bodies grow by low-velocity collisions, until they get so

large that their mutual gravity becomes important for the
dynamics.

(3) gravity-assisted growth, initially in a “run-away” fashion,

later in an “oligarchic” fashion, to rocky/icy planets

(4) if rocky/icy core reaches sufficient mass (surface gravity)

to retain H+He gas (5-10 Mearth) while the disk is still gas-
rich, then gas accretion ensues. If ~30 Mearth is reached,
run-away gas accretion forms a ~Jupiter mass planet
 rough timeline
(~100,000 km)

(~10,000 km)

C. Mordasini
 Formation of gas giants
(if sufficient gas is present)

(~100,000 km)

(~10,000 km)

C. Mordasini
 The long road from dust to planets
First growth phase Final phase

Observable
Gravity Gas is
Aggregation keeps/pulls accreted
(=coagulation) bodies
together

1µm 1mm 1m 1km 1000km

Covers 13 orders of magnitude in size = 40 (!!) orders of magnitude in mass

C. Dullemond
 initial growth
Assumed initial situation: small dust particles, e.g. 0.1 μm,
mixed homogeneously with gas. (ignores dust growth in
cloud/core/collapse phase).

(1) small dust grains well coupled to gas; small relative

motions (brownian motion dominant until ~1 μm, Δv≈100
μm/s; turbulence-driven motions dominant for larger
grains, with Δv up to several 10 m/s for ~cm particles

(2) “touch and stick” at low Δv (≲1 m/s for pure silicates, ≲10
m/s if particles are icy), destruction / erosion at high Δv

(3) vertical component of gravity, “dust settling”, accelerates

growth. Also radial drift —> problem!
 vertical
Sedimentation above.
Early phases
where a subscript of 0 denotes mid-plane values. Close to the mid-plane, thi

settling
profile indeed approaches a Gaussian profile with the scale height derive

(Safronov 1969)

A. Johansen

The basic picture of the early stage of planet formation (growth f

planetesimals) is the following:
Dust grains coagulate and gradually decouple from
•DustSediment
grains condense,
to form coagulate and gradually
a thin mid-plane layerdecouple fro
in the dis
•The dust grains settle into a thin mid-plane layer in the disk.
Planetesimals
•Planetesimals form form by continued
by continued coagulation
coagulation (two bodyorcoll
se
combination)
gravitational inofdense
instability the dust
T. Birnstiel
mid-plane
(or a layer
combination of the tw
Fig. 3 Trajectory of a settling particle at 1 AU. Figure and parameters after Dullemond &
plane layer. Dominik (2005). The initial particle size is one micrometer and the initial position is 5 H
 gas-dust coupling
tstop T. Birnstiel
“ tstop ¨ ⌦K ” St (Stokes number)
torb

St << 1 i.e. ⌧fric ⌧ ⌧orb

St ~ 1 i.e. ⌧fric ' ⌧orb

St >> 1 i.e. ⌧fric ⌧orb

It‘s (mostly) not size that matters - it‘s the Stokes number!
 gas-dust coupling
Gas rotational velocity
“Epstein” drag regime, particle radius a << λ, where
eal gas, we can write the pressure in terms of a power law as well:
λ is mean free path of gas molecule.
⇥ ( +⇥)
k k r dp ( + ⇥)k
⇤T = ⇤0 T 0 ⇥ =4⇡ 2
⇤T r 1
mH µmH r0 Fdrag “dr´ ⇢gµm a Hvth v
3
essure gradient in the gas centrifugal equilibrium yields:
ρg is gas density, v is the velocity of the particle
2 2
g v kT relative to the
1 bulk2motion 2 of kTthe gas, vth is the
= k
( + ⇥)r ⇥ vg = vk ( + ⇥)
r µmHthermal velocity of the gas, given µmH by :
8kT
he thermal velocity is given by: 2
vtherm =
µmH

e rotational velocity
whereof # the gasmean
is the as: molecular weight in AMU (typically
⇥2 atom.
#=2.3) and mH is the mass of a hydrogen
⌅ vtherm
2
= v (1
2
2⇤(r)) ; ⇤(r) = ( + ⇥)
 radial drift problem
Fgravity Fcentrifugal dust particle
“feels” no pressure
towards star
Fgravity Fcentrifugal gas molecule
“feels” pressure
Fpressure

(1) Gas pressure radially decreasing (in continuous disk).

(2) Pressure gradient force supports gas disk (but not the
dust)

(3) Force Equilibrium leads to sub-Keplerian gas rotation

(4) Head-wind removes dust angular momentum

(5) Orbital decay;

 relative particle velocities
• very small particles ~completely coupled to gas (they go
wherever the gas goes), Stokes << 1; small Δv.

• very large particles ~completely de-coupled from gas (they

“don’t care” about the gas), Stokes >> 1; small Δv (all
~keplerian), unless gravitational stirring becomes important
(later on)

• Intermediate range where particles are semi-coupled (gas

affects velocities differently for different sizes), Stokes ~ 1;
large Δv. Relative velocity several 10 m/s.
 “meter-sized barrier”
(note: more “barriers” exist; e.g. “bouncing barrier”)

two-fold “barrier” for reaching “boulder” sizes (large enough

to ~de-couple from gas)

(1) drift: In a MMSN, the decay time for particle with

maximum drift velocity (“Stokes parameter” ~1) from:
1 AU : ~200 yrs (~3 m particle)
5 AU: ~1,400 yr (~1 m particle)
100 AU: ~30,000 yrs (~10 cm particle)

(note: numbers are for specific set of assumptions, only

indicative of order of magnitude).

(2) fragmentation relative velocities up to ~several 10 m/s

around St~1, destructive collisions.
24
overcome m-sized barrier
(1) Gravitational Instability planetesimal formation: if
dust settles in very thin disk (Hdust/Hgas~10-4) that is also
nearly perfectly free of turbulence (relative velocities
<~10 cm/s at 1 AU), then dust disk may fragment into
clumps that collapse under own gravity (“Goldreich &
Ward mechanism”). Considered unlikely (tubulence
prohibits these circumstances to be reached).

(2) gravo-turbulent planetesimal formation: the

1995A&A...295L...1B

turbulence itself causes local enhancements of dust

density, vortices can “trap" dust

25
H
why dust piles up in regions
of high (gas) density

26
why dust piles up in regions
of high (gas) density

27
why dust piles up in regions
of high (gas) density

28
why dust piles up in regions
of high (gas) density

29
 gravo-turbulent
planetesimal formation
(1) MRI tubulence

(2) particles gather in over-

densities, local
enhancements in solids;

(3) gravity of concentrated

dust sustains over-
density, attract more
solids

(4) simulations produce

~few to ~35 Ceres mass
bodies in just 13 orbits
Johansen, Klahr, Henning
30
 further growth
(1) particles are >>1 m, all move on ~keplerian orbits.

(2) velocities are “damped" by gas —> low Δv

(3) growth by low velocity collisions, “mergers”. Slow,

“orderly” growth Collis
Collision
Collisionc
Gravitational focussing
v
In a billiard game, the collisional cross sections of two bodies is simply given by
vv
cross sections
Collision cross-s
(4) initially gravity of growing bodies is not important, but r r ! ! r1
1 1
!

once bodies reach ~km sizes, “gravitational focusing" m

v
The attracting nature of gravity leads for planet growth
!
m
to m
1
an 1
increase
1
of the co
starts to become important and growth goesr1faster.
section over the geometrical one. This is called gravitational focussing. Let us bc
collisional cross section for two arbitrary sized, gravitating (spherical) bodies, n
influence of the sun (twoCollision
body approximation) 1 m
cross-section: 2 body
r1
v! b
m1
b Conservation of ene
Conservation of of
Conservation energy:
energy:
1 12 2Er 1= 112µv22 =m11m µv m
122
E= E =µv1µv= = µv µv 1G G2
2 2 1m2 22 v r1 2+
r1 r+
C. Mordasini
Conservation ofConservation
energy: of energy: 2
2
at ! at clo
1 2 1 m1 m2 at ! at ! ✓ at closest approach
at◆closest approach
= µv 2
11 1 1b2m=22(r1 + r12 )m
E = µv1 G 2
Gm v m
2 2 r1 + r222 2
12m
1 + 2esc
 further growth (II)
(5) gravity-assisted, accelerated growth. For “dynamically
cold” disk, growth rate dM/dt ∝M4/3 —> largest bodies
grow fastest, run-away growth where “winner takes all”

(6) largest body accretes most planetesimals within its

gravitational sphere of influence.

(7) largest bodies start to gravitationally disturb (“excite”)

smaller ones, planetesimal disk gets “dynamically hot” and
gravitational focusing is less effective.

(8) “oligarchic growth” where dozens of ≳0.1 MEarth bodies

dominate their local environment and slowly grow further

(9) complex N-body interactions; many “oligarchs” merge in

“giant impacts” (last one in our case is thought to have
resulted in Earth-Moon system.
 On to final planets

(~100,000 km)

(~10,000 km)

C. Mordasini
 Formation of a Gas Giant Planet
Total
Gas
Solids

Original: Pollack et al. 1996;

Here: Mordasini, Alibert, Klahr &
Henning 2012

C. Dullemond
 Formation of a Gas Giant Planet
Total
Growth by accretion of Gas
planetesimals until the Solids
local supply runs out
(isolation mass).

Original: Pollack et al. 1996;

Here: Mordasini, Alibert, Klahr &
Henning 2012

C. Dullemond
 Formation of a Gas Giant Planet
Slow accretion of gas Total
(slow, because the Gas
gas must radiatively Solids
cool, before new gas
can be added). Speed
is limited by opacities. If planet migrates, it
can sweep up more
solids, accellerating
this phase.

The added gas

increases the mass,
and thereby the size
of the feeding zone.
Hence: New solids are
accreted.
Original: Pollack et al. 1996;
Here: Mordasini, Alibert, Klahr &
Henning 2012

C. Dullemond
 Formation of a Gas Giant Planet
Total
Once Mgas > Msolid, the Gas
core instability sets in: Solids
accelerating accretion of
more and more gas

Original: Pollack et al. 1996;

Here: Mordasini, Alibert, Klahr &
Henning 2012

C. Dullemond
 Formation of a Gas Giant Planet
Total
A hydrostatic envelope Gas
smoothly connecting core Solids
with disk no longer
exists. Planet envelope
detaches from the disk.

Original: Pollack et al. 1996;

Here: Mordasini, Alibert, Klahr &
Henning 2012

C. Dullemond
 Formation of a Gas Giant Planet
Total
Something ends the gas Gas
accretion phase, for Solids
example: strong gap
opening. „Normal“ planet
evolution starts.

Original: Pollack et al. 1996;

Here: Mordasini, Alibert, Klahr &
Henning 2012

C. Dullemond
Is this (CA) how it goes?
CA is a complex, “multi stage” problem, still only partially
understood theoretically and many phases poorly/not constrained
observationally. But: huge progress in modeling; “barriers”
resulting from over-simplification go away if better physical
models are applied.

(1) CA can produce rocky planets

(2) CA can yield strongly enriched planets (GI much less so)

(3) CA can yield ice giants (GI cannot do this)

Core Accretion is the currently favored scenario. Not much

doubt that the basic idea is right. Unclear whether GI, in
addition, is at work in outer regions of massive disks
40
 migration
(forming) planets interact gravitationally with the
disk (and other planets), and may move from Type I
where they form(ed), sometimes a lot

(1) type I migration: relatively low-mass planets

(e.g. ~1 Mearth) do not significantly alter
surface density profile Σ(R) but material
concentrates asymmetrically in resonances
and exerts torque causing migration
Type II
(2) type II migration: high-mass planets (~1 Mjup)
open gaps and launch strong spiral arms that
exert torque.

(3) Planet-planet interaction can significantly alter

orbits of planets on timescales of >>1 orbit
P. Armitage