1: \documentclass[preprint]{aastex}
2: \newcommand{\vdag}{(v)^\dagger}
3: \newcommand{\myemail}{seager@mit.edu}
4: \slugcomment{PASP, in press}
5: %\shorttitle{Computational tool for planet interiors}
6: %\shortauthors{Zeng and Seager 2007}
7:
8: \bibliographystyle{apj}
9:
10: \begin{document}
11:
12:
13: \title{A Computational Tool to Interpret the Bulk Composition of Solid Exoplanets based on Mass and Radius Measurements}
14:
15: \author{Li Zeng\footnote{Department of Physics, Massachusetts Institute of Technology, 77 Massachusetts Ave., Cambridge, MA 02139 }, S. Seager$^{1,}$\footnote{Department of Earth, Atmospheric
16: and Planetary Sciences, Massachusetts Institute of Technology, 77 Massachusetts Ave., Cambridge, MA 02139 }}
17: %\affil instead of footnote but takes up more space
18:
19:
20: \begin{abstract}
21: The prospects for finding transiting exoplanets in the range of a few
22: to 20 $M_{\oplus}$ is growing rapidly with both ground-based and
23: spaced-based efforts. We describe a publically availalble computer
24: code to compute and quantify the compositional ambiguities for
25: differentiated solid exoplanets with a measured mass and radius,
26: including the mass and radius
27: uncertainties.
28: \end{abstract}
29:
30: \section{Introduction}
31:
32: Over 250 extrasolar planets are known to orbit nearby main sequence
33: stars. Among
34: these include over a dozen exoplanets with minimum masses below 22
35: $M_{\oplus}$ and several with minimum masses less than 10 $M_{\oplus}$. Of
36: key interest are transiting planets with measured masses and radii,
37: which can be used to constrain the planet's interior bulk composition.
38: The relationship between mass and radius for solid exoplanets has
39: hence received much attention in the last few years \citep{vale2006,
40: fort2007, seag2007, sels2007, soti2007}. The recent
41: activity builds on much earlier work
42: \citep{zs1969, stev1982}, with improvements on the equations of state
43: and treatment of different mantle and core compositions to varying
44: degrees of complexity.
45:
46: Unlike for the solar system planets, we have no access to the
47: gravitational moments of exoplanets. Hence the density distribution in
48: the interior is unknown and this leads to an ambiguity, or degeneracy,
49: in the interior composition for an exoplanet of a fixed mass and
50: radius. One way to capture the degeneracies of exoplanet interior
51: composition is using ternary diagrams (introduced to exoplanet
52: interiors by \citet{vale2007}).
53:
54: We adopt the idea of using ternary diagrams to quantify the
55: compositional uncertainty in exoplanets. The planet mass and radius
56: are the observed quantities and therefore we focus solely on on
57: ternary diagrams for a planet of fixed mass and fixed radius
58: \citep[c.f.][]{vale2007}. We compute ternary diagrams for solid
59: exoplanets ranging in mass from 0.5 to 20 $M_{\oplus}$. We
60: furthermore explain the behavior of the mass-radius curves in two and three
61: dimensions. We also present a description of our publically available
62: computer code to compute fixed mass-radius ternary diagrams, including
63: the observational uncertainties.
64:
65: \section{Computer Model}
66:
67: \subsection{Background and Equations}
68: \label{sec-eqns}
69: We begin by assuming the major components of a solid exoplanet are
70: limited to an iron core, a silicate mantle, and a water ice outer layer. In other words, we
71: assume the interior of the planet is differentiated with the
72: denser materials interior to the less dense materials. We further assume
73: each layer to be homogeneous in its composition.
74:
75: We can then model the interior of a solid exoplanet by using:
76:
77: (1) the equation for mass of a spherical shell
78: \begin{equation}
79: \frac{dm(r)}{dr}
80: =4 \pi r^2 \rho(r),
81: \end{equation}
82: where $m(r)$ is the mass included in radius $r$, $\rho(r)$ is the density at radius $r$;
83:
84: (2) the equation for hydrostatic equilibrium
85: \begin{equation}
86: \frac{dP(r)}{dr}=
87: \frac{-Gm(r) \rho(r)}{r^2},
88: \end{equation}
89: where $P(r)$ is the pressure at radius $r$;
90: and
91:
92: (3) the equation of state (EOS) that relates $P$ and $\rho$.
93:
94: The EOS is different for each different material. We used Fe ($\epsilon$)
95: for the planet core, MgSiO$_3$ perovskite for the silicate mantle,
96: and water ice VII, VIII, and X for the water-ice outer layer.
97: See \citet{seag2007} for a detailed discussion of the EOSs
98: including their source. The temperature has
99: little effect on the EOS especially in the high pressure regime \citep{seag2007};
100: we ignore the temperature
101: dependence of the EOS. This simplifies the equations and their solution,
102: while enabling a relatively accurate analysis.
103:
104: In this problem, we have five variables:
105: \begin{enumerate}
106: \item the iron mass fraction ($\alpha$);
107: \item the silicate mass fraction ($\beta$);
108: \item the water-ice mass fraction ($\gamma$);
109: \item the total mass of the planet ($M_p$);
110: \item and the mean radius of the planet ($R_p$).
111: \end{enumerate}
112: The variables $\alpha$, $\beta$, and $\gamma$ are not
113: independent of each other. Based on the assumption
114: that the planet only consists of iron, silicate, and water we have:
115: $\alpha + \beta + \gamma = 1$, which can also be expressed as $\gamma
116: = 1 - \alpha - \beta$. We therefore have four variables:
117: $\alpha, \beta, M_p$ and $R_p$. Given any three of
118: these variables, we can determine the fourth variable uniquely.
119: We can also see that given an $M_p$ and $R_p$, there is a
120: relationship between $\alpha$ and $\beta$. There is not a single value
121: of $\alpha$ and $\beta$ that produces a given $M_p$ and
122: $R_p$. Instead, there are infinite pairs of $\alpha$ and $\beta$
123: which give the same $M_p$ and $R_p$, and we call this a
124: degeneracy in the interior composition.
125:
126: \subsection{Algorithm for Solving the Differential Equations}
127: \label{sec-alg1}
128: Our program integrates from the surface $r = R_p$ inward to the center
129: of the planet. The outer boundary condition is $m(R_p)=M_p$ and
130: $P(R_p)=0$. That is, at the surface, the mass is the specified total
131: planet mass and the pressure is approximately 0.
132:
133: We aim to interpret observations of a planet of a given mass and
134: radius. We therefore choose to intergrate inwards instead of outwards
135: based upon the known parameters of the planet ($M_p$, $R_p$, and
136: $P(R_p)$). The independent variable is $r$, decreasing from $r=R_p$
137: to $r=0$. The interior boundary condition is $m(r) = 0$ at $r=0$.
138: Typically, when integrating $m(r)$, $m$ does not equal zero at
139: $r=0$. We therefore must iterate, tuning the mass fraction of each
140: layer until $m=0$ and $r=0$ is reached.
141:
142: Given a single $M_p$ and $R_p$ the computer program finds all possible
143: combinations of ($\alpha$, $\beta$, $\gamma(=1-\alpha-\beta)$) that
144: give the same planet mass and radius. The program begins with a chosen
145: value of $\alpha$, and then takes a guess of $\beta$. Next, the
146: computer program integrates the differential equations (1) and (2)
147: using the above boundary conditions to find the value of the planet
148: radius. By comparing this radius to the desired $R_p$, the computer
149: program tunes the value for $\beta$, by using the bisection
150: method. This process is repeated several times, until $\beta$ is found
151: to a satisfactory accuracy of 1/1000. By varying $\alpha$ within the
152: range of 0 to 1, we can get all possible combinations of $\alpha$ and
153: $\beta$ which produce a specific $M_p$ and $R_p$.
154:
155: \subsection{Algorithm for Generating a Database of $M_p = 0.5$Ð-$20 M_{\oplus}$ }
156: \label{sec-alg2}
157: $M_p$ and $R_p$ are observed parameters and one usually wants to find
158: the corresponding allowed $\alpha$ and $\beta$. For a range of $M_p$
159: and $R_p$---corresponding to observational uncertainties---it can be
160: very time consuming to use the first algorithm described in
161: \S\ref{sec-alg1}. We therefore generate a database that is a discrete
162: representation of the relation $R_p = R_p(\alpha, \beta,
163: M_p)$. Figure~\ref{fig:3D} illustrates the 4D database.
164:
165: This database is a 3-D array which contains the data of $R_p$
166: corresponding to each combination of $\alpha$ and $\beta$ (0 to 1 with
167: 1\% spacing) and $M_{\oplus}$ (ranging from 0.5 to 20 $M_{\oplus}$
168: with $0.25 M_{\oplus}$ spacing). The database can be used via linear
169: interpolation to find $R_p$ for any given $M_p$, $\alpha$ and
170: $\beta$. A conservative estimate of the fractional error in the
171: database interpolation is 1/1000. For the same range of $M_p$ and
172: $R_p$, interpolation in the database is about 45 times faster than
173: solving the differential equations.
174:
175: To generate a database of all values of $\alpha$, $\beta$, and $R_p$
176: for $M_p$ ranging from 0.5 to 20 $M_{\oplus}$. Given $M_p$, $\alpha$,
177: and $\beta$, this algorithm also integrates from the surface inward to
178: $r=0$ and $m(r)=0$ to find $R_p$. In contrast to the first algorithm
179: (which solves for a given $M_p$ and $R_p$), a single integration in
180: radius results in the desired solution of $R_p$, for a given $M_p$,
181: $\alpha$, and $\beta$. In other words is there is no iteration
182: required, making this algorithm much more efficient.
183: %feed back to tune the input parameters $\alpha$ and
184: %$\beta$ making this algorithm much more efficient.
185:
186: \subsection{Instructions for Downloading and Using the Code}
187:
188: The code is based in MATLAB and can be downloaded from
189: http://web.mit.edu/zengli/www/ under ``Research Field'' or from
190: http://seagerexoplanets.mit.edu/research/interiors.html If using this
191: computer code please cite this paper and also \citet{seag2007}.
192:
193: We have made two different codes available. The codes have the same
194: output, but the first is based on a differential equation solver
195: (\S\ref{sec-alg1}) and the second code is based on interpolation of
196: the large database (\S\ref{sec-alg2}). For the codes, the planet mass
197: must be in the range 0.5--20~$M_{\oplus}$. The inputs to the codes
198: are: the planet mass in Earth masses ($M_p$), the planet mass
199: uncertainty in Earth masses ($\sigma_{M_p}$), the planet radius in
200: Earth radii ($R_p$), and the planet radius uncertainty in Earth radii
201: ($\sigma_{R_p}$). The values $\sigma_{M_p} = 0$ and $\sigma_{R_p}$ = 0
202: are allowed. If the combination of input values $M_p$ and $R_p$ are
203: unphysical, the code will return an error.
204:
205: {\bf ExoterDE($M_p$, $\sigma_{M_p}$, $R_p$, $\sigma_{R_p}$).} This
206: code solves the two differential equations described in
207: \S\ref{sec-eqns}. This code consists of three subroutines (each of
208: which must be downloaded) that are automatically called by the above
209: command. The first subroutine is the differential equation solver,
210: which also reads the equations of state. The second subroutine
211: contains the actual differential equations. The third subroutine plots
212: the ternary diagrams; this subroutine calls a ternary diagram plotting
213: routine \citep{the2005}\footnote{
214: http://www.mathworks.com/matlabcentral/\\
215: fileexchange/loadFile.do?objectId=7210\&objectType=file} which plots a
216: single line for each of the 1-, 2-, and 3-$\sigma$ contour lines. See
217: \S \ref{sec-uncertainty}. An example from this code is shown in
218: Figure~\ref{fig:ternunc}.
219:
220: {\bf ExoterDB($M_p$, $\sigma_{M_p}$, $R_p$, $\sigma_{R_p}$)}. This
221: code reads in the database of $M_p$, $R_p$ and fractional composition
222: ($\alpha$ and $\beta$). The output is a ternary diagram, shaded
223: throughout the 1-, 2-, and 3-$\sigma$ contour curves. This subroutine
224: uses the same ternary diagram plotting routine as described above.
225:
226: The differential equation solver ExoterDE is much slower than the
227: database extracter ExoterDB. In principle, ExoterDB is more accurate
228: than ExoterDB.
229:
230:
231: \section{Data Display}
232:
233: \subsection{2-D Cartesian Diagram}
234: For a given $M_p$ and $R_p$ we want to know the interior composition
235: of the relative mass fraction of the three components. There are three
236: variables we have solved for ($\alpha, \beta, \gamma$), but only two
237: of them are independent (since $\gamma=1-\alpha-\beta$). Therefore
238: points on a 2-D diagram can describe all the possible combinations of
239: $\alpha, \beta, \gamma$ for a given $M_p$ and $R_p$. We show such a
240: solution in Figure~\ref{fig:cart1}. We note that $0 \leq \alpha
241: \leq1$, $0 \leq \beta \leq 1$, $\alpha + \beta \leq 1$, and therefore
242: not every point in the 2-D plane will correspond to a set of $\alpha,
243: \beta, \gamma$. Only the points which are in a right-angled triangular
244: region will respresent the set of allowed solutions.
245:
246: \subsection{Ternary Diagram}
247:
248: Ternary diagrams to describe the interior composition of exoplanets
249: were introduced by \citet{vale2007}. In a ternary diagram, $\alpha$,
250: $\beta$, and $\gamma$ are each one axis of an equilateral
251: triangle. Although $\gamma$ is extraneous, the ternary diagram is
252: useful because it is more intuitive to see the three components of the
253: planet interior (in a symmetric way) compared to a 2D Cartesian
254: diagram with only two of the components. Figure~\ref{fig:terndef}
255: shows how to read a ternary diagram.
256:
257: \subsection{Relationship Between the 3D and 2D Cartesian Diagrams and the Ternary Diagram}
258: \label{sec-convert}
259: To explain the full origin of a curve on the ternary diagram we start
260: with the 3D Cartesian diagram with all solutions of $M_p$, $R_p$ and
261: composition (in terms of $\alpha$ and $\beta$), as shown in
262: Figure~\ref{fig:3D}. We take an isoradius and isomass surface as
263: shown in Figure~\ref{fig:isorisom}. As an example, in
264: Figure~\ref{fig:isorisom} the red surface is the iso-radius surface of
265: $R_p = 1.7 R_{\oplus}$ and one of the blue colored planess is the
266: iso-mass surface of $M_p = 4 M_{\oplus}$. These two surfaces intersect
267: each other and result in a curve. This curve can be projected
268: vertically to the $x$-$y$ plane which is the iso-mass plane. We
269: therefore have a (iso-radius and isomass) curve on the isomass
270: plane. This curve is shown in Figure~\ref{fig:cartproj} in a Cartesian
271: diagram.
272:
273: The Cartesian and ternary diagrams are two different ways to represent
274: the same information. There exists a linear coordinate transformation
275: between the two. That means if a function is a straight line appearing
276: in the 2-D Cartesian diagram, it will still be a straight line in the
277: ternary diagram.
278:
279: The transformation from 2-D Cartesian coordinates to the ternary
280: diagram coordinates is
281: \begin{equation}
282: \label{eq:coordconversiona}
283: x_{ternary}= \frac{1}{2} (1+x-y)
284: \end{equation}
285: \begin{equation}
286: y_{ternary}=\frac{\sqrt{3}}{2} (1-x-y).
287: \label{eq:coordconversionb}
288: \end{equation}
289: Here $x$ and $y$ are the coordinates of a point in a 2D Cartesian
290: diagram, the $x_{ternary}$ and $y_{ternary}$ are the coordinates of the point
291: in ternary diagram in the Cartesian grid variables.
292: Figure~\ref{fig:ternproj} shows the same $R_p = 1.7 R_{\oplus}$, $M_p = 4
293: M_{\oplus}$ curve represented by a ternary diagram.
294:
295:
296: %**********************************************************************************************
297: \section{Results and Discussion}
298: %**********************************************************************************************
299: \subsection{Observational Uncertainties}
300: \label{sec-uncertainty}
301: Real planet mass and radius measurements have uncertainties. The
302: planet mass and radius uncertainties are typically 5 to 10 percent
303: \citep[e.g.,][]{sels2007}, and even smaller for the most favorable targets. We
304: now present examples of ternary diagrams that include the mass and
305: radius uncertainties.
306:
307: We consider uncertainties of 1, 2, and 3 standard deviations
308: ($\sigma$) from the measured value. See Figures~\ref{fig:ternunc} and
309: \ref{fig:ternmult}. In more detail, the uncertainty in composition on
310: the ternary diagram is
311: \begin{equation}
312: \sigma_{comp}=\sqrt{(\sigma_{compM})^2+(\sigma_{compR})^2},
313: \end{equation}
314: where $comp$ refers to composition and $compM$ and $compR$ refer to
315: composition uncertainties caused by the planet mass and radius
316: uncertainty respectively. Here we have assumed that the uncertainties
317: in mass and radius are independent from each other and have assumed
318: the linearity of the superposition of small uncertainties.
319:
320: Figure~\ref{fig:contour} shows a planet with $M_p = 10 \pm 0.5
321: M_{\oplus} $ and $R_p = 2 \pm 0.1 M_{\oplus}$. We see that taking the
322: 3-$\sigma$ limit, almost the entire ternary diagram is filled. In
323: other words, for a 5\% 3-$\sigma$ (i.e., 15\%) uncertainty on the
324: planet mass and radius, the interior composition in terms of
325: fractional composition of iron, silicates and water cannot be
326: determined. The reason this example fills the almost the whole
327: ternary diagram is that a planet with $10 M_{\oplus}$ and $2
328: R_{\oplus}$ has an average density in between two extreme cases
329: (purely iron or purely water). Therefore, a large variety of different
330: combinations of iron, silicate and water can result in a similar $M_p$
331: and $R_p$. Even taking a 1-$\sigma$ uncertainty of the planet mass
332: and radius, the uncertainty in internal composition is large.
333:
334: We note that an uncertainty in $R_p$ has more of an effect on the uncertainty in the interior composition than an uncertainty in $M_p$. This is because the planet's average density
335: $\rho \sim M_p/R^3$. Considering error propagation, the uncertainty in radius
336: has a three times larger effect on the uncertainty in average density than does
337: the mass uncertainty.
338:
339: We show ternary diagrams for planets with various masses, radii, and
340: 5\% fractional uncertainty in Figure~\ref{fig:ternmult}. Only
341: solutions in part of the ternary diagram are allowed, despite
342: considering the 3-$\sigma$ range. In Figure~\ref{fig:ternmult}b, the
343: upper 3-$\sigma$ boundary is absent because it goes below the lowest
344: allowed density of the $2M_{\oplus}$ and $1.5 R_{\oplus}$ planet and
345: is thus unphysical. In Figure~\ref{fig:ternmult}c we see the opposite
346: case, where the lower 3-$\sigma$ boundary is absent because it goes
347: above the highest planet density allowed and is thus unphysical.
348:
349: A ternary diagram for a fixed planet mass and radius that includes
350: observational uncertainties is one of the primary outcomes of this
351: paper.
352:
353:
354: %**********************************************************************************************
355: \subsection{Model Uncertainties}
356:
357: The model and computer code we present assumes a differentiated planet
358: composed of an iron core, a silicate mantle, and a water ice outer layer.
359: The division into three major materials is
360: based on the point that the densities of iron, silicate,
361: and water are much more different from each other than any
362: minor compositional variant of each individual material.
363: The model neglects phase variation and temperatures which, as argued
364: in \citet{seag2007}, have little effect on the total
365: planet radius (to an uncertainty of about
366: $\sim$1 to 3 percent uncertainty in planet radius, decreasing with
367: increasing planet mass.).
368:
369: Low-pressure phase changes (at $< 10$ GPa) are not important for a
370: planet's radius because for plausible planet compositions most of the
371: mass is at high pressure. For high pressure phase changes we expect
372: the associated correction to the equation of state (and hence derived
373: planet radii) to be small because at high pressure the importance of
374: chemical bonding patterns to the equation of state drops.
375:
376: Regarding temperature, at low pressures ($\lesssim$ 10 GPa) in the outer planetary layers,
377: the crystal lattice structure dominates the material's density and the
378: thermal vibration contribution to the density are small in comparison.
379: At high pressures the thermal pressure contribution to the EOS is
380: small because the close-packed nature of the materials prevents structural
381: changes from thermal pressure contributions.
382:
383: Although the code can model radii for planets in the mass range
384: 0.5 to 20~$M_{\oplus}$, the model is more accurate for planets
385: above a few Earth masses \cite{seag2007}.
386:
387:
388: The model also neglects variation in composition, such as a light element
389: in the iron core as Earth and Mercury are believed to have. The model also
390: omits other impurities in the mantle and water layer, including
391: iron in the mantle. Molten cores have
392: also been omitted. At the present time, these model uncertainties are expected
393: to have an effect on the planet radius much less than the $\sim$5 percent
394: radius observational uncertainty.
395:
396: For all of the above reasons, we therefore argue that for the present
397: time the observational uncertainties dominate the model
398: uncertainties; the model
399: presented here is adequate for an estimate of planet bulk composition.
400: In any event, the main results of our work described in the following subsections
401: despite any model uncertainties.
402:
403: It is possible to rule out parts of the ternary diagram as being
404: physically unplausible \citep[e.g.,][]{vale2007}. This is based on the initial composition of the
405: protoplanetary nebula and on planet differentiation.
406: For example, a pure iron planet is unlikely to exist, because removing all
407: of the mantle would be difficult. A pure water planet is also unlikely to exist.
408: Where water ice forms, so do silicate-rich and iron-rich materials, making planet accretion
409: of pure water unlikely. We prefer to leave
410: the omission of parts of the ternary diagram to users of the code, because
411: in exoplanets surprising exceptions to the "rules" of planet characteristics are not uncommon.
412:
413:
414:
415: \subsection{Spacing, Shape, Direction, and Rotation of Curves on
416: the Ternary Diagram.}
417:
418: We now turn to a discussion of the spacing, direction, and shape of
419: the curves in the ternary diagrams. A quantitative and qualitative
420: description of these is a main point of this paper. We emphasize that
421: each curve shown in our Figure~\ref{fig:ternmult} diagrams represents
422: a different mass and radius. The curves to the lower right are more
423: dense, as they have a higher mass and lower radius than the curves
424: moving to the upper left.
425:
426: All behavior results from the equations of state of the materials,
427: and, in some cases, how they behave differently under pressure.
428:
429: We begin with an equation that we use repeatedly in this section. We
430: consider the simplified case that the planet core has a uniform
431: density, where $ {\bar{\rho}_{Fe}}$ is the average density of Fe in
432: the core, ${\bar {\rho}_{MgSiO_3}}$ is the average density of silicate
433: in the mantle, and ${\bar {\rho}_{H_2O}}$ is the average density of
434: water in the outer water layer. We then have
435: \begin{equation}
436: \label{eq:basic}
437: \frac{4}{3} \pi R_p^3 = \left[
438: \frac{\alpha M_p}{\bar {\rho}_{Fe}} +
439: \frac{\beta M_p}{\bar{\rho}_{MgSiO_3}} +
440: \frac{\gamma M_p}{\bar{\rho}_{H_2O}}
441: \right],
442: \end{equation}
443: where $ \alpha + \beta + \gamma = 1$.
444:
445:
446: \subsubsection{Spacing}
447:
448: The iso-mass iso-radius curves for adjacent
449: curves with equal differences of mass and radius have uneven spacing on a ternary
450: diagram. This spacing is generally smaller in the lower right part of the ternary diagram (high iron fraction region) than in the upper left part (high silicate or water fraction region).
451: The curves in the lower right part of the diagram have a higher density (higher mass
452: and smaller radius) than the curves found on the upper left part of the diagram. The density is not a linear function of both mass and radius hence we do not expect equal
453: spacing on the ternary diagram. We can, however, give both a quantitative and qualitative
454: explanation of the uneven spacing.
455:
456: We can provide a quantitative description, beginning with equation~(\ref{eq:basic}),
457: but using $\alpha = 1 - \beta - \gamma$ to get
458: \begin{equation}
459: \frac{4}{3} \pi R_p^3 = M_p \left[
460: \frac{(1 - \beta - \gamma) }{\bar {\rho}_{Fe}} +
461: \frac{\beta }{\bar{\rho}_{MgSiO_3}} +
462: \frac{\gamma}{\bar{\rho}_{H_2O}}
463: \right]
464: \end{equation}
465:
466: We use the fact that the curves on the ternary diagram
467: are almost perpendicular to the water side of the ternary diagram,
468: and therefore set the silicate mass fraction $\beta = 0$ for our discussion. In other words, the
469: distance (separation) between the points
470: produced by the intersection of the isomass-isoradius curves and the water axis
471: is a good representation
472: of the spacing between the curves throughout the ternary diagram,
473: \begin{equation}
474: \label{eq:basic2}
475: \frac{4}{3} \pi R_p^3 = M_p
476: \left[ \frac{1}{\bar {\rho}_{Fe}} +
477: \gamma \left(
478: \frac{1}{\bar {\rho}_{H_2O}} -
479: \frac{1}{\bar{\rho}_{Fe}} \right)
480: \right].
481: \end{equation}
482:
483: We define the following constants:
484: \begin{equation}
485: A_1 = \frac{1}{\bar {\rho}_{Fe}}
486: \end{equation}
487: and
488: \begin{equation}
489: A_2 = \frac{1}{\bar {\rho}_{H_2O}} - \frac{1}{\bar {\rho}_{Fe}}.
490: \end{equation}
491: Equation~(\ref{eq:basic2}) then becomes
492: \begin{equation}
493: \gamma A_2 = \frac{4/3 \pi R_p^3}{M_p} - A_1.
494: \label{eq:basic3}
495: \end{equation}
496: We note that $A_1>0$ and $A_2 > 0$.
497:
498: Now we proceed to take the derivative of
499: equation~({\ref{eq:basic3}) with respect to the
500: change of mass ($dM_p$) and change of radius ($dR_p$).
501: \begin{equation}
502: A_2 d \gamma = \frac{4 \pi}{3} \left[
503: \frac{3 R_p^2}{M_p} dR_p
504: - \frac{R_p^3}{M_p^2}dM_p
505: \right].
506: \end{equation}
507: We can also rewrite equation~({\ref{eq:basic3}) in terms
508: of the overall average density of the planet ($\bar{\rho}$),
509: \begin{equation}
510: \gamma A_2 = \frac{1}{\bar \rho} - A_1,
511: \end{equation}
512: and the corresponding derivative relative to the overall average density ($\bar{\rho}$)
513: \begin{equation}
514: A_2 d\gamma = -\frac{1}{\bar \rho^2} d {\bar \rho}.
515: \label{eq:basic4}
516: \end{equation}
517:
518: This leads to our quantitative understanding, where we first recall
519: that $d \gamma$ is the water fraction spacing on the ternary diagram.
520: The lower right part (Fe-rich) of the ternary diagram is where the
521: average density ($\bar \rho$) of a planet is high. This implies (for
522: the same $d \bar \rho$) $d \gamma$ is small (since $\bar \rho$ is in
523: the denominator). In the upper left region of the ternary diagram
524: (water-rich) the average planet density is smaller than a planet
525: located in the lower right part of the diagram, and therefore $d
526: \gamma$ is larger.
527:
528: Qualitatively, to have wider spacing an increasing water fraction is
529: needed. In other words, towards the upper right part of the ternary
530: diagram, for the same density difference more water than iron must be
531: replaced.
532:
533: We note that the spacing is predominantly the
534: result of the nonlinearity of equations~(\ref{eq:basic3}) and (\ref{eq:basic4})
535: and the average density of each compositional layer,
536: not of any $P$-$\rho$ properties of the EOS (i.e.,
537: how materials condense under high pressure). This statement is correct
538: under the assumption of a single material for each layer in the planet
539: (in our case for iron, silicate, and water). The assumption
540: that the average density within each layer does not change significantly from curve to curve (for example, from the 1-$\sigma$ curve to the 3-$\sigma$ curve for the case $M_p=2 M_\oplus$ and $R_p=1.5 R_\oplus$) is reasonable for a ternary diagram that spans only a small
541: mass and radius range.
542:
543: %**********************************************************************************************
544: \subsubsection{Shape, Direction, and Rotation}
545:
546: To explain the shape and direction of the curves in the ternary diagrams
547: we start by explaining the slope of the curves in the Cartesian diagram.
548: In other words, we are aiming for an expression of $d \alpha / d \beta$.
549:
550: We start with a different form of equation~(\ref{eq:basic}),
551: \begin{eqnarray}
552: \alpha \left[
553: \frac{1}{\bar \rho_{H_2O}} - \frac{1}{\bar \rho_{Fe}} \right]
554: +
555: \beta \left[
556: \frac{1}{\bar \rho_{H_2O}} - \frac{1}{\bar \rho_{MgSiO_3}} \right]
557: = \nonumber \\
558: \frac{1}{\bar \rho_{H_2O}} - \frac{4 \pi R_p^3}{3 M_p}.
559: \end{eqnarray}
560: We now differentiate this equation with respect to $\alpha$ and $\beta$ to find
561: \begin{equation}
562: d\alpha \left[
563: \frac{1}{\bar \rho_{H_2O}} - \frac{1}{\bar \rho_{Fe}} \right]
564: +
565: d\beta \left[
566: \frac{1}{\bar \rho_{H_2O}} - \frac{1}{\bar \rho_{MgSiO_3}} \right]
567: \approx 0,
568: \end{equation}
569: based on the assumption that the average density is changing slowly
570: with respect to the change in composition.
571:
572: We write the slope of the curves in the Cartesian diagram
573: \begin{equation}
574: \label{eq:slopecartesian}
575: \left | \frac{d \beta}{d \alpha} \right | =
576: \frac{ \left[ 1 - \frac{\bar \rho_{H_2O}}{\bar \rho_{Fe}} \right]}
577: { \left [1 - \frac{\bar \rho_{H_2O}}{\overline \rho_{MgSiO_3}} \right]}
578: \end{equation}
579: We can see that
580: \begin{equation}
581: \left | \frac{d \beta}{d \alpha} \right | > 1,
582: \end{equation}
583: assuming that ${\bar \rho_{H_2O}} < {\bar \rho_{MgSiO_3}} < {\bar \rho_{Fe}}$.
584:
585: %#############################################################
586:
587: It can be shown that the slope in the the ternary diagram
588: has a positive connection to the slope in the Cartesian diagram.
589: Using the the equations that convert the Cartesian coordinates
590: to coordinates on the ternary diagram (equations~(\ref{eq:coordconversiona}) and
591: (\ref{eq:coordconversionb})) we find that
592: \begin{equation}
593: \label{eq:slopeternary}
594: \left | \frac{d \beta_{ternary}}{d \alpha_{ternary}} \right | =
595: \sqrt{3}
596: \frac{\left[ \left | \frac{d \beta}{d \alpha} \right | - 1 \right]}
597: {\left[ \left | \frac{d \beta}{d \alpha} \right | + 1\right]}.
598: \end{equation}
599:
600: We can now go on to describe the direction, shape, and rotation in the
601: Cartesian diagram based on equation~(\ref{eq:slopecartesian}), with
602: the knowledge that the same qualitative behavior will appear in the
603: ternary diagrams. We first emphasize that the slope of an isomass,
604: isoradius curve on the ternary diagram describes adding and removing
605: mass of the different species.
606:
607: We begin with a qualitative explanation of the direction of the curves
608: on the ternary diagram---why each curve goes from the lower left to
609: the upper right. This is largely a coincidence in the different
610: values of ${\bar \rho_{H_2O}}$, ${\bar \rho_{MgSiO_3}}$, and ${\bar
611: \rho_{Fe}}$. The
612: coincidence lies in the fact that iron is more dense and water less dense
613: than silicate---and for zero-pressure densities, an equal
614: mass of iron and water combined densities are roughly
615: similar to the silicate density. At the lower left of the ternary diagram in
616: Figure~\ref{fig:ternproj}, the silicate mass fraction is 80 percent,
617: the iron mass fraction is zero and the water mass fraction is 20
618: percent. As the silicate fraction decreases, a combination of equal parts iron and
619: water must be added to maintain the same overall planet mass and
620: radius. This description is consistent with the direction
621: of the curves in the ternary diagrams.
622:
623: More quantitatively, from equation~(\ref{eq:slopecartesian}),
624: we can take the zero-pressure densities of Fe, H$_2$O, and MgSiO$_3$
625: to find the slope of the curve on the cartesian diagram
626: \begin{equation}
627: \label{eq:slopecart2}
628: \left | \frac{d \beta}{d \alpha} \right | = 2,
629: \end{equation}
630: and from equations~(\ref{eq:coordconversiona}) and
631: (\ref{eq:coordconversionb}),
632: \begin{equation}
633: \left | \frac{d \beta_{ternary}}{d \alpha_{ternary}} \right | = \frac{1}{\sqrt{3}}.
634: \end{equation}
635: We can consider removing a fixed amount
636: of iron mass, e.g., 1 gram. According to equation~(\ref{eq:slopecart2}),
637: 2 grams of silicate must be added.
638: For mass balance (because each curve on the ternary
639: diagram represents a planet of fixed mass and fixed radius), 1 gram of water must be removed.
640: The direction of the curves on the ternary diagram do correspond
641: to removing roughly equal masses of iron and water for
642: every mass of silicate added.
643:
644: {\it Rotation}
645:
646: In Figure~\ref{fig:ternmult}a--d (where a through d are in
647: order of increasing mass), we see that for more massive
648: planets, the isomass-isoradius curves are rotated. In other words, the
649: slope of the curve on the ternary diagram increases for increasing mass.
650:
651: We again return to equation~(\ref{eq:slopecartesian}).
652: With increasing planet mass, each of
653: ${\bar \rho_{H_2O}}$, ${\bar \rho_{MgSiO_3}}$, and $ {\bar \rho_{Fe}}$
654: changes. Because iron is in the core, it suffers
655: more compression than the silicate mantle or water icy layer.
656: In other words, $ {\bar \rho_{Fe}}$
657: must increase more than ${\bar \rho_{H_2O}}$ and ${\bar \rho_{MgSiO_3}}$
658: as the planet mass increases.
659: Therefore, the numerator of equation~(\ref{eq:slopecartesian})
660: gets larger and thus the slope of the curve
661: increases. The rotation is counter clockwise with increase of mass.
662:
663: {\it Slope of the Curves}
664:
665: We now turn to discuss the slope of each isomass-isoradius curve.
666: We see from Figures~\ref{fig:ternunc} and \ref{fig:ternmult} that the
667: slope of each isomass-isoradius curve for a fixed mass and radius is greater in the lower
668: left part of the curve than in the upper
669: right part of the curve.
670:
671: The slope is due to the
672: differential compression of water, silicate and iron under pressure.
673: This slope is again explained by
674: equation~(\ref{eq:slopecartesian}). At the lower left the slope is
675: smaller; this is the silicate-rich region of the ternary
676: diagram. There is more silicate and less iron and water. The ${\bar
677: \rho_{MgSiO_3}}$ will increase as it gets compressed in the inner part
678: of the planet. This causes the slope in
679: equation~(\ref{eq:slopecartesian}) to get smaller. In contrast, at
680: the upper right, there is little silicate, but more water and more
681: iron. The silicate and water are less compressed, but the iron is
682: more compressed, making the slope increase.
683:
684: For a conceptual explanation, first recall the idea described above that
685: removing 2 grams of silicate can be compensated by adding approximately
686: 1 gram of water and 1 gram of iron.
687: In a silicate-rich planet (lower left of the ternary diagram),
688: silicate is compressed. For a massive planet, this compression makes
689: the silicate density closer to iron's density than in the uncompressed case.
690: Therefore, removing a fixed mass of silicate requires much more iron than water to be added, for a fixed
691: mass and radius.
692:
693: In contrast, along the upper right part of the ternary diagram, the
694: planet is iron-rich or water-rich. The slope of an isomass-isoradius
695: curve is steeper than a curve in the lower left part of the ternary
696: diagram. For an isomass-isoradius curve in the upper part of the
697: ternary diagram, the planet has more iron, the iron is very compressed
698: (There is less silicate and the silicate is overall less compressed
699: compared to a planet in the lower left of the ternary diagram). If a
700: fixed mass of silicate is removed, more water than iron must be added
701: to compensate for the density of compressed iron. The density of water
702: does not change much, because water is always the outer layer and thus
703: the least compressed.
704:
705: %We also see from Figure~\ref{fig:ternmult} that a given
706: %isomass/isoradius curve in the lower right (Fe-rich) region of the
707: %diagram are each more curved than a given iso-mass/iso-radius curve in
708: %the upper left part of the diagram. This is the iron-rich region of
709: %the diagram and the iron gets more compressed. This is because the
710: %above explanation for the slope along a curve gets exaggerated as
711: %there is more iron and it gets more compressed in the core.
712:
713: As a qualitative explanation, if the average density of each of the
714: three layers remains constant, then an isomass-isoradius curve should
715: always be a straight line in either the Cartesian or ternary
716: representation. The curvature in a given isomass-isoradius curve
717: appears because of the compression of material under pressure, which
718: is a property of the EOS.
719:
720: %**********************************************************************************************
721: \subsection{Shape and Direction of an Isoradius Surface in our 3D Representation}
722:
723: We return to the 3D representation of the relationship between mass,
724: radius and composition shown in Figure~\ref{fig:isorisom}. An
725: isoradius surface (red oblique surface) is shown for $R_p = 1.7
726: R_\oplus$. An isomass surface (one of the blue flat planes) is shown for
727: $M_p=4 M_\oplus$. At the bottom tip of the isoradius surface, the
728: silicate mass fraction = iron mass fraction = 0, and the planet is
729: composed of 100\% water. This is the minimum mass for this
730: radius. The isoradius surface also has a maximum mass, reached by a
731: composition of 100 percent iron.
732:
733: For the same radius, if either the silicate or iron mass fraction is
734: increased, the planet mass must also increase. This is because iron
735: and silicate are denser than water ice. We further note that for
736: an increase in iron mass fraction, the mass of the planet must
737: increase more steeply than for an increase in the silicate mass
738: fraction. This is seen by the different length and shape "edges" of
739: the isomass surface in the iron-mass plane and silicate-mass plane
740: in Figure~\ref{fig:isorisom}.
741:
742: We now show that the shape of the isoradius surface is concave.
743: The isoradius surface can be considered as the isovolume surface, where the volume is the sum of
744: the core, the silicate mantle, and the water crust. To calculate the total mass of a point on the isoradius curve, we use equation~(\ref{eq:basic}), and since $R_p$ is constant on this surface we can rewrite this equation (with $C$ a constant) as,
745: \begin{equation}
746: M_p = C
747: \left[
748: \frac{\alpha}{\bar{\rho_{Fe}}} +
749: \frac{\beta}{\bar{\rho_{MgSiO_3}}} +
750: \frac{(1 - \alpha - \beta) }{\bar{\rho_{H_2O}}}
751: \right]^{-1}.
752: \end{equation}
753: The term in brackets is a linear function and its inverse is
754: hyperbolic.
755: For example, if we let $\alpha = 0$ (no iron), then
756: we have
757: \begin{equation}
758: M_p = C \left[
759: \frac{1}{\bar{\rho_{H_2O}}}
760: - \beta \left( \frac{1}{\bar{\rho_{H_2O}}} -\frac{1} {\bar{\rho_{MgSiO_3}}} \right)
761: \right]^{-1},
762: \end{equation}
763: where
764: \begin{equation}
765: C = \frac{4}{3} \pi R_p^3.
766: \end{equation}
767: Both terms in the square brackets are positive and
768: $ 0 \leq \beta \leq 1$.
769: This is in the form
770: \begin{equation}
771: M_p = \frac{C}{ C_1 - C_2 \beta},
772: \end{equation}
773: where
774: \begin{equation}
775: C_1 = \frac{1}{{\bar \rho_{H_2O}}}
776: \end{equation}
777: and
778: \begin{equation}
779: C_2 = \frac{1}{{\bar \rho_{H_2O}}} - \frac{1}{{\bar \rho_{MgSiO_3}}}
780: \end{equation}
781: We have $ C_1 > C_2 > 0$.
782: This is the form of a concave hyperbola,
783: because, taking the first derivative we find
784: \begin{equation}
785: \frac{dM_p}{d\beta} = \frac{C C_2}{\left ( C_1 - C_2 \beta \right)^2}
786: \end{equation}
787: The slope increases as $\beta$ increases because
788: $C_1 - C_2 \beta$ decreases but is always greater than zero.
789:
790: Although we have made the simplification
791: that the average densities of each layer remain constant,
792: our qualitative description holds because
793: $ {\bar \rho}_{H_2O} < {\bar \rho}_{MgSiO_3} < {\bar \rho}_{Fe}$
794: as long as the planet is differentiated into layers of increasing
795: density towards the planet center.
796:
797:
798: A similar argument shows that the total iron fraction
799: vs. total mass is also a hyperbola, making the whole
800: isoradius surface concave.
801:
802:
803:
804: %**********************************************************************************************
805: \section{Summary and Conclusion}
806: %**********************************************************************************************
807:
808: An ambiguity in an exoplanet interior composition remains for
809: any planet with a measured mass and radius, no matter how
810: precisely measured. We can accept this ambiguity and quantify it with the aid
811: of ternary diagrams \citep{vale2007}. We have presented ternary diagrams
812: for a single planet of fixed mass and radius, for a planet composed of
813: an iron core, a silicate mantle, and a water ice outer layer. Our ternary
814: diagram presentation includes observational uncertainties.
815: We have provided a publically available computer code to generate a ternary
816: diagram for a given input mass, radius, and observational uncertainties.
817:
818: In addition to presenting ternary diagrams for fixed mass and radius,
819: we showed their origin from a 4D database ($M_p$, $R_p$, iron mass-fraction $\alpha$,
820: and the silicate mass fraction $\beta$; recall that the water-ice mass fraction
821: $\gamma = 1 - \alpha - \beta$). We further described the shape and direction of
822: the composition curves on a ternary diagram .
823:
824: We conclude with the sentiment that in order to fully
825: understand the interior structure of an exoplanet,
826: a third measurement beyond planet mass and radius is required.
827:
828:
829: \acknowledgements{We thank the MIT John Reed Fund for supporting an
830: undergraduate research opportunity for L. Z.}
831:
832: %---------------------------------------------------------------------------
833: \bibliography{planets}
834: %---------------------------------------------------------------------------
835:
836: \begin{figure}
837: \plotone{f1.eps}
838: \caption{The 3D representation of planet mass, planet radius, and
839: composition expressed as mass fractions of iron ($\alpha$) and
840: silicate ($\beta$). The $x$-axis is the iron mass fraction and the
841: $y$-axis is the silicate mass fraction. The water mass fraction comes
842: from $\alpha + \beta + \gamma = 1$. The $z$-axis is the total planet
843: mass in Earth masses. The planet radius is not indicated, except by
844: three separate surfaces of constant radius. An isomass and isoradius
845: surface intersect with a curve in a 2D iron mass-fraction vs. silicate
846: mass fraction Cartesian diagram (Figure~\ref{fig:cartproj}).}
847: \label{fig:3D}
848: \label{fig:isorisom}
849: \end{figure}
850:
851: \begin{figure}
852: \plotone{f2.eps}
853: \caption{The Cartesian diagram for silicate mass fraction ($\beta$)
854: vs. iron mass fraction ($\alpha$) for a planet with $M_p = 4
855: M_{\oplus}$, $R_p = 1.7 R_{\oplus}$ (Note that the water mass fraction
856: $\gamma = 0.4$ since $\gamma = 1 - \alpha - \beta.$) A planet of a
857: fixed internal composition is represented as a point in the grey
858: shaded region (only compositions in the grey shaded region are
859: allowed.) With a given planet mass and radius there is not a unique
860: interior composition; the allowed compositions are described by the
861: curve. The curve originates from the intersection of the isomass
862: surface and the isoradius surface in the 3D representation
863: (Figure~\ref{fig:isorisom}).}
864: \label{fig:cart1}
865: \label{fig:cartproj}
866: \end{figure}
867:
868: \begin{figure}
869: \plotone{f3.eps}
870: \caption{A ternary diagram for a planet of the same mass and radius as
871: shown in the corresponding Cartesian diagram in
872: Figure~\ref{fig:cart1}. The light grey $x$ and $y$ axes are shown to
873: illustrate how this ternary diagram relates to a Cartesian diagram;
874: $x_{ternary}$ and $y_{ternary}$ are the coordinates transformed from
875: the Cartesian coordinates. (See \S~\ref{sec-convert} for the conversion
876: equations.) Any point on the curve is a possible combination of iron,
877: silicate, and water that will result in a planet with with $M_p = 4
878: M_{\oplus}$ and $R_p = 1.7 R_{\oplus}$. This Figure also illustrates
879: how to read a ternary diagram. Consider a triangle oriented such that
880: the value 1 (of a given material) is at the triangle's apex. The mass
881: fraction of that given material can be read off of a horizontal line
882: that is perpendicular to the line connecting the triangle's apex with
883: the triangle base.}
884: \label{fig:terndef}
885: \label{fig:ternproj}
886: \end{figure}
887:
888: \begin{figure}
889: \plotone{f4.eps}
890: \caption{ A ternary diagram including the mass and radius
891: uncertainties for a planet of a fixed mass and radius. This example is
892: for a planet with $M_p = 10 \pm 0.5 M_{\oplus} $ and $R_p = 2 \pm 0.1
893: M_{\oplus}$, showing the 1-, 2- and 3-$\sigma$ uncertainty curves.
894: This Figure shows the continuous distribution of possible combinations
895: of iron, silicate, and water with increasing uncertainties according
896: to the color bar. Notice that considering the 3-$\sigma$ uncertainties
897: almost the entire ternary diagram is covered---in other words there is
898: no constraint on the planet internal composition. See text for a
899: discussion of the direction and spacing of the curves. }
900: \label{fig:contour}
901: \label{fig:ternunc}
902: \end{figure}
903:
904: \begin{figure}
905: \plotone{f5.eps}
906: \caption{Ternary diagrams including 5\% mass and radius uncertainties
907: for planets of fixed mass and radius. Panel a) $M_p = 1 \pm 0.05 M_{\oplus}$
908: and $R_p = 1 \pm 0.05 M_{\oplus}$. Panel b) $M_p = 2 \pm 0.1 M_{\oplus}$
909: and $R_p = 1.5 \pm 0.075 M_{\oplus}$. Panel c) $M_p = 8 \pm 0.4 M_{\oplus}$
910: and $R_p = 1.5 \pm 0.075 M_{\oplus}$. Panel d) $M_p = 16 \pm 0.8 M_{\oplus}$
911: and $R_p = 2.5 \pm 0.125 M_{\oplus}$. See text for a discussion of the
912: direction and spacing of the curves. }
913: \label{fig:ternmult}
914: \end{figure}
915:
916: \end{document}
917: