Академический Документы
Профессиональный Документы
Культура Документы
of Exoplanets
Johnson Liu
7 April 2019
Contents
1 Some variables and constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
4 Integration Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
4.1 Euler Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
4.2 Euler-Cromer Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
4.3 4th Order Runge-Kutta Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . 4
4.4 Integration Profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.4.1 RK4 Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.4.2 Euler Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.5 Comparing the Euler and RK4 Algorithms . . . . . . . . . . . . . . . . . . . . . 7
5 Optimization Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.1 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.2 Optimizing Pi and Rc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
5.3 Interpolation of Surface Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
5.4 Finding Best Values for the Independent Variables . . . . . . . . . . . . . . . . . 10
5.4.1 Pi and Rc Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
5.4.2 Linear Interpolation and Extrapolation to Find Better Estimates for Pi
and Rc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
6 Earth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
6.1 Three-Layer System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
6.2 Two-Layer System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
6.2.1 Solid Core . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
6.2.2 Liquid Core . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
6.2.3 Comparing Solid Core and Liquid Core Simulations . . . . . . . . . . . . 13
6.2.4 Optimization Scheme for Simulating Earth-like Planets . . . . . . . . . . 14
7 Mass-Radius Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1
Modeling the Interior of Exoplanets Johnson Liu
1 Some variables and con- The change in pressure of the planet’s interior
is given by
stants
Variables ρ(r)GM
Variable Symbol Value Units ∆P = − ∆r , (2)
Mass of
r2
MMercury 3.302e26 g
Mercury
Radius of where ρ(r) is the density of the planet’s in-
RMercury 2.439e8 cm
Mercury
terior at a distance r from the center of the
Mass of Earth MEarth 5.972e27 g
Radius of Earth REarth 6.3781e8 cm planet and G is the gravitational constant.
Gravitational Given proper initial conditions, Equa-
G 6.673e-8 cm3 g−1 s−2
Constant tions (1) and (2) can be used to simulate the
Density of Rock radius and the total mass of a given planet.
ρrock
0 3.98 gcm−3
at zero pressure
Since the pressure on the surface of the planet
Bulk Modulus
Krock
0 206 GPa is zero, we can use Equation (2) to make incre-
of Rock
Density of Solid mental changes to the initial pressure at the
Iron at zero ρsolid
0
iron
7.85 gcm−3 center of the planet until we reach a radius
pressure where the simulated pressure reaches zero.
Bulk Modulus
Ksolid
0
iron
255 GPa
of Solid Iron
Density of
Liquid Iron at ρliquid
0
iron
7.05 gcm−3
zero pressure 2.1 Equation of State
Bulk Modulus
Kliquid
0
iron
201 GPa The density ρ(r) of the planet at a radius r
of Liquid Iron
Pressure at from the center needs to be computed for each
center of the Pi — GPa shell. ρ(r) is found using a chosen equation of
planet state. In this project, the equation of state
Radius of the
core
Rc — cm used is
! 37 ! 53
3K0 ρ(r) ρ(r)
P = − , (3)
2 Equations of Planetary 2 ρ0 ρ0
Structure
where K0 is the bulk modulus of a given ma-
Give the pressure at the center of a planet and
terial and ρ0 is the density of the material at
the radii of the planet’s layers, the mass and
zero pressure [ZSJ16]. Equation (3) can be
radius of the planet can be computed using
written as
equations of planetary structure. In the sim-
ulation, thin spherical shells are successively
added to a central sphere. When each shell a = x7/3 − x5/3 , (4)
is added to the simulated planet, an incre-
mental change in mass is added to the overall
planetary mass. The change in pressure of the where
planet’s interior is also computed at each time
step. The incremental change in mass of the 2P
a = (5)
planet at each time step is given by 3K0
ρ(r)
∆M = 4πρ(r)r2 ∆r . (1) x = . (6)
ρ0
2
Modeling the Interior of Exoplanets Johnson Liu
Mi = ρVi
4πρ 3
2.2 Shooting Method for Esti- = r . (7)
3 i
mating the Density of the
Planet Interior
In our simulations, we have chosen ri = ∆r.
Given the pressure at some radius r, the den- Once we have the initial pressure and ini-
sity at r can be computed from Equation (4) tial mass, we integrate outwards towards the
using the shooting method. planet surface using our choice for ∆r and
To find the solution for x, we make a guess Equations (1) and (2).
for the value of x, compute the right-hand side For a planet with multiple layers, we must
of Equation (4), and compare this value to change the values for K0 and ρ0 when the
a. If our guess for x undershoots the value simulation reaches the radius where one layer
of a, we increase the value of our guess. If we meets another. For example, the interior of
overshoot the value of a, we decrease the value Mercury can be approximated with a solid
of our guess. If we undershoot in one iteration iron core surrounded by a layer of rock.
and overshoot in the next, or if we overshoot For the simulation of Mercury, the simula-
and then undershoot, we decrease the amount tion starts with K0 = K0solid iron and ρsolid
0
iron
by which we change our guess for x. With this [ZSJ16]. When the simulation reaches the
approach, we home in on a and we get better core radius (Rc ), the values for the bulk
and better estimates for x. When we compute modulus and the density under zero pressure
a value for a to within the desired accuracy, switches to K0 = K0rock and ρ0 = ρrock 0 , re-
we stop the iteration. We then multiply x by spectively. This process can be generalized
ρ0 to retrieve a value for ρ(r) for the current to other cases where there are more than two
shell. layers.
3
Modeling the Interior of Exoplanets Johnson Liu
ρ(rn )GMn
Pn+1 = Pn −
!
∆r . (13) h k2 l2
rn2 k3 = hf xn + , yn + , zn + (24)
2 2 2
4.2 Euler-Cromer Algorithm h k2 l2
!
l3 = hg xn + , yn + , zn + (25)
The Euler-Cromer algorithm is a modification 2 2 2
of the set of equations given in Equations (11)
and (11):
k4 = hf xn , yn + k3 , zn + l3 (26)
vn+1 = vn + g(xn , tn )∆t (14)
l4 = hg xn , yn + k3 , zn + l3 (27)
4
Modeling the Interior of Exoplanets Johnson Liu
where h is the step size. Applying this to our The simulated density, pressure, and mass for
equations of state, we get Mercury at each time step for different ap-
! proximations for ρ are shown in Figures (2),
GMn (3), and (4), respectively.
k1 = − ρ(rn ) 2 ∆r (30)
rn
l1 = 4πρ(rn )rn2 ∆r (31)
!
∆r G Mn + l21
k2 = − ρ rn + 2
2 rn + ∆r
rn
(32)
!
∆r ∆r 2
l2 = 4πρ rn + rn + ∆r
2 2
(33)
!
∆r G Mn + l22 Figure 2: The simulated density of Mercury using
k3 = − ρ rn +
linear and quadratic extrapolation for ρ(rn + ∆r
2
2 rn + ∆r
rn 2 )
and ρ(rn + ∆r). The results from making
the
(34) ∆r
! approximation ρ rn + 2 ≈ ρ rn + ∆r ≈ ρn
∆r ∆r 2 is also shown.
l3 = 4πρ rn + rn + ∆r
2 2
(35)
!
G(Mn + l3
k4 = − ρ(rn + ∆r) ∆r
(rn + ∆r)2
(36)
l4 = 4πρ(rn + ∆r)(rn + ∆r)2 ∆r (37)
1
k = k1 + 2k2 + 2k3 + k4
6
1
l = l1 + 2l2 + 2l3 + l4 .
6
Since the change in mass does not explicitly
depend on mass or the pressure, l2 and l3 end
up being the same term.
ρ(rn + ∆r
2
) and ρ(rn + ∆r) can be esti-
mated through linear or quadratic extrapola-
tion. These values can also be approximated Figure 3: The simulated pressure of Mercury us-
by substituting them with ρ(rn ): ing linear and quadratic extrapolation for ρ(rn +
∆r
2 ) and ρ(rn +∆r).
The results from making
the
∆r
∆r
ρ rn + ≈ ρ rn + ∆r ≈ ρ(rn ) . (38) approximation ρ rn + 2 ≈ ρ rn + ∆r ≈ ρn
2 is also shown.
5
Modeling the Interior of Exoplanets Johnson Liu
Figure 4: The simulated mass of Mercury using Figure 6: A heatmap of the simulated mass for
linear and quadratic extrapolation for ρ(rn + ∆r
2 ) Mercury using the RK4 algorithm.
and ρ(rn + ∆r). The results from making Figures (7) and (8) show the simulated ra-
the
dius and simulated mass, respectively, from
∆r
approximation ρ rn + 2 ≈ ρ rn + ∆r ≈ ρn
using the RK4 algorithm as a function of
is also shown.
step size. These simulations used Pi =
40.1415507194 and Rcfrac = 0.675567506625.
These are the optimal values of Pi and Rc to
4.4 Integration Profile produce a simulated radius and a simulated
mass close to that of Mercury’s. According
4.4.1 RK4 Algorithm to the the figures, we begin to lose accuracy
when the step size becomes larger than 1e4
Figures (5) and (6) show heatmaps of the cm.
simulated radius and simulated mass, respec-
tively, of Mercury as functions of Pi and Rc
using the RK4 alogrithm. The color on the
heatmaps show the absolute value of 1 mi-
nus the ratio of the simulated values (mass
and radius) to the actual mass and radius of
Mercury. Values on the heatmap close to 0
(shown as dark red in the figures) show good
agreement between the simulated mass and
simulated radius relative to the actual mass
and radius of Mercury. The two heatmaps
Figure 7: Final radius (shown as a fraction of
are plotted on with the same colorscale. Both the actual radius of Mercury) as a function of
heatmaps agree on the general area where the log10 (step size) using the RK4 algorithm.
optimal values for Pi and Rc are found.
Figure 14: The simulated pressure of Mercury as Figure 17: The difference in the pressure between
a function of the distance from the center of the the RK4 and Euler algorithms.
planet.
• Run trajectory
9
Modeling the Interior of Exoplanets Johnson Liu
10
Modeling the Interior of Exoplanets Johnson Liu
keeping Pi constant. We repeat this cycle un- we use the point (x0 , y0 ) for xp and yp . Oth-
til the simulated radius and mass are within erwise, we use the point (x1 , y1 ). Once we
a desired error threshold. have a value for xn , we use it in our simu-
An example optimization scheme is: lations to produce a value for yn . We then
repeat the process using the points (xn , yn )
1. Optimize Pi using planet mass as a mea- and (xp , yp ) to solve for the next guess for xn .
sure of best fit When our value for yn is close enough to our
desired value yd :
2. Optimize Rc using planet radius as a
measure of best fit |yd − yn | < threshold , (47)
we keep the value for the optimized indepen-
3. Optimize Pi using planet mass as a mea-
dent variable fixed and we begin optimizing
sure of best fit
the other independent variable.
4. Optimize Rc using planet radius as a
measure of best fit
6 Earth
5. etc ...
6.1 Three-Layer System
5.4.2 Linear Interpolation and Extrap- The interior of the Earth can be modeled as
olation to Find Better Estimates a three-layer system with a solid iron core, a
for Pi and Rc liquid iron middle layer, and a rocky mantle.
Given two initial guesses for one of the in- For a three-layer system, we can either make
dependent variables, the simulated value for a guess for the radius of the inner layer and
one of the dependent variables can be found. the radius of the middle layer or we can make
These points can then be used to interpolate, a guess for the radius of the middle layer and
or extrapolate, for a better guess for the in- make a choice for the pressure at the inter-
dependent variable. face between the inner layer and the middle
Given two points (x0 , y0 ) and (x1 , y1 ), the layer. For the latter choice, the simulation
slope of the line between the points is given switches from the inner layer to the middle
by layer when the interface pressure is reached.
In the following figures, simulations were per-
∆y formed using 330 GPa as the pressure where
y0 =
∆x the inner solid iron layer of the Earth meets
the middle liquid iron layer.
y1 − y0
= . (44) Figures (23) and (24) show heatmaps of the
x1 − x0 simulated radius and simulated mass, respec-
In order to get an estimate for the indepen- tively, of a three-layer Earth system.
dent variable xn that will give us a simulated
value for yn that is close to the desired value
yd , we use the slope in Equation (44) to make
a change in our value for x. We solve for xn
using a previous guess xp :
yd − yp
xn = xp + . (45)
y0
If
|yd − y0 | < |yd − y1 | , (46) Figure 23: A heatmap of the simulated radius for
a three-layer Earth using the RK4 algorithm.
11
Modeling the Interior of Exoplanets Johnson Liu
Figure 24: A heatmap of the simulated mass for Figure 26: The simulated pressure of a three-
a three-layer Earth using the RK4 algorithm. layer Earth as a function of the distance from
the center of the planet.
Figure 28: A heatmap of the simulated radius for Figure 30: A heatmap of the simulated radius for
a two-layer Earth with a solid iron core using the a two-layer Earth with a liquid iron core using the
RK4 algorithm. RK4 algorithm.
13
Modeling the Interior of Exoplanets Johnson Liu
R = (0.0592rmf + 0.0975)(log M )2
+ (0.2337rmf + 0.4938) log M
+ (0.3102rmf + 0.7932) , (48)
Figure 33: The difference between the simulated where R is the radius of the planet in units
density of a two-layer Earth from using a solid
of the radius of Earth, M is the mass of the
core and a liquid core.
planet in units of the mass of the Earth, and
rmf is the rock mass fraction of the planet
[FMB07].
Figure (35) shows a mass-radius diagram
for planets made entirely out of rock and plan-
ets made entirely out of solid iron. The equa-
tion of state used for the simulations of the
planets in the diagram is the one in Equa-
tion (3). Figure (36) shows the same plot as
in Figure (35), but on a log-log scale.
Figure 36: The mass-radius diagram from Fig- Figure 37: The Zeng, Vinet, and BME equations
ure (35) in log-log scale. of state.
8 Comparing Different
Equations of State
In addition to the Zeng equation of state
from Equation (3), the Vinet equation of state
[Sea+07] is
of state for x ≥ 1.
3K0 7/3
x − x5/3
P =
2
3 0 2/3
1 + (K0 − 4) x − 1 . (50)
4
9 Mass-Radius Diagram
of Planets with an
Earth-like Composition
16