Skip to content

Core and mantle split

How much of the H, C, N and S of a planet goes into its core when the core forms?

The short answer. The split divides the bulk mass of each element between the core (the metal) and the mantle (the silicate). No mass is lost, and the ratio of the two mass fractions equals the partition coefficient \(D\) of the element. \(D\) says how strongly the element prefers the metal. CALLIOPE takes \(D\) from a law at one pressure, one temperature and one redox state. \(D\) depends on the metal composition, so the code repeats the calculation until the metal stops changing. In the tutorial planet the core takes 0.871 to 0.992 of the S, depending on the redox state. The split runs on its own: PROTEUS does not call it.

To run the split, follow the core split tutorial. Every argument, output flag and error is in the partition file reference.

What the split does

When a planet forms its core, liquid iron sinks through the magma ocean and takes part of the light elements with it. calliope.core_split.split_core_mantle divides the bulk mass of H, C, N and S between the core (the metal) and the mantle (the silicate). It does this once, at one pressure, one temperature and one redox state. The metal reservoir answers another question: how a melt and a metal share the elements at one moment of a run, together with the atmosphere.

The mass balance

The mass balance of each element \(i\) uses 4 quantities. The split gets the first 3 as inputs and computes \(D_i\) from the law of the element:

  • \(M_i\), the bulk mass of the element, in kg;
  • \(M_\mathrm{metal}\), the mass of the metal that equilibrates with the silicate, in kg;
  • \(M_\mathrm{silicate}\), the mass of the silicate, in kg;
  • \(D_i\), the partition coefficient on the mass basis: the mass fraction of \(i\) in the metal divided by its mass fraction in the silicate.

Each element is either in the metal or in the silicate. Let \(w_i\) be its mass fraction in the silicate. The metal then holds the mass fraction \(D_i w_i\), and the bulk is \(M_i = M_\mathrm{silicate} w_i + M_\mathrm{metal} D_i w_i\). Solved for \(w_i\):

\[ w_i = \frac{M_i}{M_\mathrm{silicate} + D_i\,M_\mathrm{metal}}. \]

The mantle holds \(M_\mathrm{silicate} w_i\) and the core holds \(D_i M_\mathrm{metal} w_i\). With \(x = M_\mathrm{metal}/M_\mathrm{silicate}\), the core takes the fraction \(D_i x/(1 + D_i x)\) of the bulk.

Worked example. The Earth has \(M_\mathrm{metal} = 1.94 \times 10^{24}\) kg and \(M_\mathrm{silicate} = 4.03 \times 10^{24}\) kg, so \(x = 0.481\). An element with a constant \(D = 100\) puts \(100 \times 0.481 / (1 + 100 \times 0.481) = 0.9796\) of its bulk into the core (Step 4 of the core split tutorial). With the S law of the code, \(D_\mathrm{S}\) depends on the conditions, so the core fraction does too; see the redox state.

After each split, the code checks for each element that core plus mantle equals the bulk, to \(10^{-12}\) of the bulk. It raises RuntimeError when this element ledger does not close.

Why the split must iterate

\(D_i\) is not a constant. It depends on what else is in the metal: the \(\varepsilon_j\) terms of a law (one fitted coefficient for each other light element \(j\) in the metal, see The partition coefficient) change \(D_i\) with the amount of the other light elements, and a law on the mole basis is converted to the mass basis with the moles per kg of the metal. The metal composition in turn depends on how much of each element the split puts into the metal. So the split repeats 3 steps until the metal stops changing: compute \(D_i\) for the present metal, split the bulk, and compute the new metal from the split. One round of these 3 steps is one pass. This is a fixed point on the metal composition: a loop in which each pass uses the result of the pass before, until the result stops changing. The metal reservoir uses the same loop.

The inputs

The split takes 5 kinds of input besides the masses.

The equilibration point

The equilibration point \((P_\mathrm{eq}, T_\mathrm{eq})\) is the pressure and temperature at which the metal and the silicate equilibrate. The default is 60 GPa and 4000 K, the deep magma ocean column of Suer et al. (2023)1, Table 1. The S law is inside its pressure and temperature range there. The core-mantle boundary is not a good default. The Earth's core-mantle boundary is at about 136 GPa, above the data of every packaged law (at most 91 GPa), and larger planets have higher pressures there.

The redox state

\(\Delta\mathrm{IW}\) is the oxygen fugacity relative to the iron-wΓΌstite buffer in log10 units. It has no default, and it is the one redox input of the split. The H, C and N laws read it through their \(\Delta\mathrm{IW}\) term. The S law has no such term. Its redox dependence is the FeO of the melt: S enters the metal as FeS in exchange for the FeO of the melt, so \(D_\mathrm{S}\) rises with the FeO. The split computes the FeO mole fraction of the melt, \(X_\mathrm{FeO}\), from \(\Delta\mathrm{IW}\) and the iron of the metal in each pass. It uses the definition of Suer et al. (2017, p. 90)2 for ideal mixing, \(\Delta\mathrm{IW} = 2\log_{10}(X_\mathrm{FeO}/X_\mathrm{Fe})\):

\[ X_\mathrm{FeO} = x'_\mathrm{Fe}\,10^{\Delta\mathrm{IW}/2}, \]

with \(x'_\mathrm{Fe}\) the Fe mole fraction over the metal without H and C. A more oxidised equilibrium therefore puts more S into the core. For the tutorial planet (the Earth's core and mantle masses, 2 wt% O and 3 wt% Si in the metal) the core takes 0.871, 0.974 and 0.992 of the S at \(\Delta\mathrm{IW}\) = -3, -1.5 and -0.5.

\(X_\mathrm{FeO}\) cannot be larger than 1. Above \(\Delta\mathrm{IW} = -2\log_{10} x'_\mathrm{Fe}\) no melt fits, and a law with an FeO term raises ValueError that names \(\Delta\mathrm{IW}\) and this limit. The first pass uses the metal without the split elements, which has the most Fe. The limit of the split is therefore the limit of the start metal w_fixed: +0.06 for 2 wt% O, and +0.11 for the tutorial metal with 2 wt% O and 3 wt% Si.

The fixed metal elements

w_fixed holds the metal mass fractions of O and Si. The split does not change them (see Limits).

The melt

oxides and nbo_t describe the melt. A law on the mole basis gives \(D\) as a ratio of mole fractions. To turn it into a mass-basis \(D\), the split multiplies by the moles of metal per kg and divides by the moles of melt per kg. The moles of melt per kg are one mole for each mole of cations of the oxides (the single-cation basis), plus the dissolved H and C. The melt composition oxides sets this number. oxides is also the melt that a law written as a Python function gets; it has no other role in the split. nbo_t is the NBO/T of the melt, the number of non-bridging O atoms per tetrahedral cation, a measure of how polymerised the melt is: it enters the NBO/T term of a law (the C law has one), and a function law gets it too. A function law that needs the melt FeO computes it from \(\Delta\mathrm{IW}\) and the metal.

The laws

The split takes one law per element, either by name from calliope.metal.get_law or as a loaded law entry. The defaults are tagawa2021_eq2 (H), fischer2020_eq1 (C), suer2023_line (N) and suer2017_eq7 (S). The S law of the metal reservoir, suer2023_line, is a line in \(\Delta\mathrm{IW}\) at one pressure and one temperature. The split uses the law of Suer et al. (2017)2 instead, because it has a pressure and a temperature term. To add your own law, see Add a partition law to the core split.

What the output flags mean

The split evaluates each law at the given point. It does not move the point to the edge of the range of the law, because that would hide the temperature dependence that dominates \(D_\mathrm{S}\), as Suer et al. (2017, p. 91)2 show. It flags the law instead, and a law outside its range logs one warning per process. flags[e] holds the flags of element e, and flags['metal'] holds those of the metal:

Flag Meaning Unit
flags[e]['out_of_range'] 1 when any of the 3 distances below is not 0, else 0 none
flags[e]['P_GPa'], ['T_K'], ['dIW'] distance of the point outside each range of the law; 0 inside the range, NaN for an unknown range (calliope.metal.range_distance) GPa, K, log10 units
flags[e]['D_residual'] relative difference between the metal mass fraction of e and \(D_e\) times its silicate mass fraction, at the returned split none
flags['metal']['x_light'], ['x_S'] mole fraction of all light elements, and of S, over the whole metal none
flags['metal']['S_over_Fe'] S to Fe mole ratio of the metal none
flags['metal']['beyond_FeS'] 1 when the metal holds more S than Fe in moles, past FeS; with other light elements in the metal this happens below \(x_\mathrm{S} = 0.5\). A flag only: the split does not clamp the metal none
flags['metal']['root_rule'] 1 when the iron-rich root rule (next section) replaced the first root, else 0 none

How the solver finds the split

The split starts from the metal of w_fixed and runs one pass after another until the metal stops changing. If the passes do not settle, it repeats them with smaller steps.

Some S-rich, iron-poor metals have 2 roots: 2 metal compositions that both close the mass balance, one with more Fe than the other. When the smaller steps end on a metal with more S than Fe, the split searches again with very small steps from the start metal and keeps the root with more Fe; root_rule is 1 when this changes the result.

The Core split solver page lists the steps, their tolerances and the errors they raise.

How reliable it is

A sweep of 5500 random splits, from \(10^{20}\) to \(3 \times 10^{24}\) kg of metal, converges in 5074 cases and stops with a named error in the others. Where the metal keeps more than 1 % of Fe by mass, the returned metal meets each law to \(2.6 \times 10^{-10}\). The sweep of random inputs gives the counts per input range and the known limits.

Limits

Each limit below says what the code does when a value is outside it.

  • \(\Delta\mathrm{IW}\) above \(-2\log_{10} x'_\mathrm{Fe}\). A law with an FeO term raises ValueError (see the redox state).
  • A law outside the range of its data. The split flags it and logs one warning per process. It does not move the point.
  • A metal with no Fe left, or passes that do not settle. The split raises ValueError when the iron runs out and RuntimeError otherwise. In the sweep of random inputs, 416 of 5500 splits raise ValueError and 10 raise RuntimeError.
  • An element ledger that does not close to \(10^{-12}\) of the bulk. The split raises RuntimeError.
  • O and Si are not split. Their bulk is the O and Si of the rock, and CALLIOPE does not hold the rock. split_core_mantle refuses O and Si in its bulk and takes their metal mass fractions as fixed values (w_fixed). The code has no split of O and Si by their exchange \(K_D\).
  • The melt basis of the S law is an assumption. Suer et al. (2017)2 do not state the mole basis of the melt for \(X_\mathrm{FeO}\) and \(C_\mathrm{S}\). CALLIOPE takes \(X_\mathrm{FeO}\) from \(\Delta\mathrm{IW}\), uses \(\log_{10} C_\mathrm{S} = -5.372\) as given and assumes the single-cation basis. The Suer et al. (2017) S runs case shows how well this reproduces their Table 1.
  • One \(C_\mathrm{S}\) for every melt. The S law uses \(\log_{10} C_\mathrm{S} = -5.372\) of a pyrolite melt, from Fig. 6 of Suer et al. (2017)2, for every melt composition.

Where this is checked. The Core and mantle split case checks the mass balance against its closed form, the Suer et al. (2017) S runs case checks the S law, and the sweep of random inputs shows where the solver converges.


  1. T.-A. Suer, C. Jackson, D. S. Grewal, C. Dalou, T. Lichtenberg, The distribution of volatile elements during rocky planet formation, Frontiers in Earth Science, 11, 1159412, 2023. SciX. ↩

  2. T.-A. Suer, J. Siebert, L. Remusat, N. Menguy, G. Fiquet, A sulfur-poor terrestrial core inferred from metal-silicate partitioning experiments, Earth and Planetary Science Letters, 469, 84-97, 2017. SciX. ↩↩↩↩↩