Examensarbete vid Institution för geovetenskaper ISSN 1650-6553 Nr 136
On the geothermal gradient and heat production in the inner core
Peter Schmidt
Abstract
In this thesis I have investigated the upper bounds on the possible presence of radiogenic heat sources in the inner core of the Earth, using both steady state and transient models, built during this work. Necessary theory for this and model descriptions are collected into appendices at the end of this work. In addition, the published literature is reviewed for various formation scenarios, modelling of the inner core, theoretical and experimental values of relevant thermodynamic parameters. A general expression for the upper thermodynamical bounds on the initial heat source abundance at onset of inner core soldification is derived, which in the range of the published values of the thermodynamical parameter space yields upper bounds of 0.20 ± 0.15 wt% initial abundance of 40K, the most favoured radiogenic candidate in the inner core. Alternatively the expression can be used to set an upper limit to the age of the inner core given that we know the present abundance of heatsources and thermal parameters. E.g. assuming a heat transfer coefficient of k = 80 W m−1 K−1, a melting temperature of iron of 5500 K at the inner core boundary, and a value of the thermodynamical Gruneisen¨ parameter of γth,ICB = 1.5, it is found that if the core is older than 0.9 Gyr the inner core 40K abundance has to be lower than 0.142 wt% (the constraint set by cosmochemical arguments) and if the inner core is older than 2.52 Gyr the upper bound is less than 0.058 wt% (upper limit as set by high pressure experiments). Several geotherms for the inner core in subspaces of the parameter space are also presented. A comparison between the steady state and transient models is also performed, with the result that steady state models generally underestimates the temperatures and are not suitable for the inner core geotherm, mainly due to the transient nature of inner core formation and evolution. Finally the nickel-silicide/georeactor inner core model, as proposed by Herndon is investigated. It is found that this would generate a large molten region at the centre of the inner core, which has not been observed today. Hence it is concluded that a georeactor can not be operational at the centre of the Earth today. Sammanfattning
I detta examensarbete har jag studerat de ovre¨ granserna¨ for¨ forekomsten¨ av radioaktiva nukleider i Jordens inre karna¨ med hjalp¨ av sav˚ al¨ jamvikts¨ som evolutionara¨ modeller, vilka har utvecklats i detta arbete. Den teoretiska bakgrunden sav˚ al¨ som modell beskrivningar finns samlade i olika appendix i slutet av detta dokument. Utover¨ detta revideras tillanglig¨ publicerad litteratur over¨ olika formations scenarier, modellering av jordens karna,¨ teoretiska och experimentella varden¨ pa˚ for¨ problemet relevanta termodynamiska parametrar. Ett generellt analytiskt uttryck for¨ den termodynamiskt begransade¨ ovre¨ gransen¨ for¨ koncentrationer av varmek¨ allor¨ vid tiden for¨ den innersta karnans¨ initiala stelning ar¨ harlett,¨ baserat pa˚ detta ar¨ det funnet att inom publicerade parametervarden¨ ligger den ovre¨ gransen¨ pa˚ 0.20 ± 0.15 wt% 40K, vilken ar¨ den troligaste radionukliden i den inre karnan.¨ Alternativt kan uttrycket anvandas¨ for¨ att satta¨ en ovre¨ grans¨ pa˚ aldern˚ av den inre karnan,¨ givet att dagens koncentration av varme¨ kallor¨ och termala parametrar ar¨ kanda.¨ Till exempel, antagande en varmelednings¨ koefficient pa˚ k = 80 W m−1 K−1, en smalt¨ temperatur for¨ jarn¨ pa˚ 5500 K vid gransen¨ till den inre karnan,¨ och ett varde¨ pa˚ den termodynamiska Gruneisen¨ parametern pa˚ γth,ICB = 1.5, sa˚ ar¨ det funnet at om den inre karnan¨ ar¨ aldre¨ an¨ 0.9 miljarder maste˚ koncentrationen av 40K vara lagre¨ an¨ 0.142 wt% (ovre¨ gransen¨ baserad pa˚ kosmokemiska argument) och om den inre karnan¨ ar¨ aldre¨ an¨ 2.52 miljarder ar˚ maste˚ koncentrationen vara lagre¨ an¨ 0.058 wt% (ovre¨ gransen¨ satt av hortrycks¨ experiment). Ett flertal mojliga¨ geotermer for¨ den inre karnan¨ ar¨ presenterad i olika subrum av parameterrumet. Vidare gors¨ en jamf¨ orelse¨ mellan jamvikts¨ och de evolutionara¨ modellerna, vilken ger vid handen att jamvikts¨ modeller overlag¨ tenderar att ge for¨ laga˚ temperaturer och darf¨ or¨ inte kan anses vara anvandbara¨ vid studier av geotermer i jordens inre karna.¨ Orsaken till detta star˚ huvudsakligen att finna i den transienta naturen av den inre karnans¨ tillblivelse och utveckling. Slutligen har en modell foreslagen¨ av Herndon, i vilken den inre karnan¨ bestar˚ av nickel-silicid med en central georeaktor, blivit studerad. Det ar¨ funnet att modellen skulle resultera i en relativt stor uppsmalt¨ region i centrum av den inre karnan,¨ vilket inte har blivit observerat. Slutsatsen ar¨ dragen att en dylik georeactor inte kan vara aktiv i karnan¨ idag. Contents
1 Introduction 1
2 Formation of the Solar system and the Earth 1 2.1 Star and circum stellar disk formation ...... 2 2.2 Gravitational instability model ...... 3 2.3 Core accretion model ...... 3 2.4 Formation of terrestrial planets and core-mantle differentiation ...... 4
3 The inner structure of the Earth 5 3.1 The Mantle ...... 6 3.2 The core ...... 6 3.3 The PREM model ...... 7
4 Probing the Earth’s interior 8 4.1 Mass and density distribution ...... 9 4.2 Elastic properties ...... 9 4.3 Composition ...... 9 4.4 Heat ...... 10 4.5 Laboratory measurements ...... 10 4.6 Theoretical ...... 11
5 Fe at high Pressure and Temperature 11 5.1 Fe phase diagram and melting curve ...... 12 5.2 Thermal and elastic properties ...... 13
6 Models of the Earth’s Core 15 6.1 Thermal models ...... 15 6.2 Earlier core models ...... 16 6.3 Inner core solidification ...... 17 6.4 Radionuclides in the core ...... 18 6.5 Alternative core models ...... 19
7 Constraints on the inner core geotherm and heat production 20 7.1 Inner core heat sources ...... 20 7.2 Inner core geotherm ...... 22 7.3 Constraining the inner core temperature profile ...... 28 7.4 Transient models ...... 32
8 Summary and Discussion 37
9 Acknowledgements 40
A PREM 45
B The heat equation 47 B.1 Spherical symmetry ...... 48 B.2 Solving the heat equation in spherical coordinates ...... 49
C Thermodynamics and Mechanics 51 C.1 Thermodynamic fundamentals ...... 51 C.2 Thermoelastic coupling ...... 52 C.3 Lattice vibrations and the Debye approximation ...... 53 C.4 Gruneisen¨ parameters ...... 54 C.5 Adiabatic temperature ...... 55 C.6 Melting curves ...... 57 D COMSOL Multiphysics 59 D.1 Spherical symmetry using the Heat module ...... 59 D.2 Constants, Expressions, Variables, and Functions ...... 63 D.3 Meshing and Solvers ...... 66 D.4 Evaluating the output of COMSOL ...... 67
E Transient heat transfer model of the inner core 69 E.1 Transient solution to the heat equation ...... 69 E.2 Inner core growth model ...... 70 E.3 Numerical issues ...... 71 E.4 Testing ...... 76 E.5 CoreT.m ...... 77 1 Introduction
The Earth is truly a remarkable place, where ever you look at its surface, from the deepest oceans to the highest mountains, from the coldest polar regions to the hottest desserts, you will find life. And so far, our home planet is the only place in the universe where we know for sure that life once evolved and continues to exist today. So what makes our planet so suitable for life? For one thing, the composition and evolution of the Earth has offered the proper conditions for life. Thus, an understanding of life as we know it demands knowledge of the geology of the Earth, as well as its geological history. A second criterion is the presence of liquid water at the surface, which is made possible not only by the Earth’s distance to the sun, but also by the atmosphere. It needs to be realized that had it not been for the (so commonly referred to) greenhouse effect, the surface temperature of the Earth would have varied by more than hundred degrees during the course of a day, with a mean temperature below the freezing point of water. Now, the relatively close proximity to the sun is not only beneficial, as a result the Earth is under a constant heavy bombardment of charged high energy particles, emanating from the sun. Exposed to such flux levels, most life forms on our planet would not stand a chance to survive, neither would the atmosphere. Fortunately the Earth is shielded from the lethal stream by its relatively strong magnetic field, originating in the liquid outer core. However, as we know form our geological archives, the magnetic field is not static in time, neither in strength nor in direction. Even if life has shown to be able to survive the numerous polar reversal and field variations in our past, the high technological society of mankind that has evolved over the last decades is highly sensitive to such events. Therefore it is of greatest importance that we understand the operation of the geodynamo. This in turn demands knowledge of the energy supplies available in the core, in which the inner core plays an essential part, as well as the possible presence of radionuclides in the core.
In this thesis work I have investigated the upper bounds for the presence of radionuclides in the inner core and the associated thermal profiles (geotherms) as constrained by the fact that the inner core is solid. In order to do this I have reviewed the literature (mainly of the last decade) for our present knowledge of the inner core. However, as it will turn out, the presences of radionuclides in the core is dependent on how the Earth once formed. Therefore this report will start with a review of our present understanding of the formation of the solar system, leading up to the formation and differentiation of the Earth in chapter 2. I will then present the interior structure of the Earth in chater 3 as well as a seismic model that will be used later. In chapter 4 various methods availible for measuring the Earth’s interior will be presented, whilst chapter 5 will present the high pressure, high temperature behaviour of iron, believed to be themain constituent of the core. Chapter 6 will then present different models of the Earths core, including sections on the inner core solidification and the presence of radionuclides in the core. My own work will be presented in chapter 7, followed by a discussion of the method and summary in chapter 8. In order to increase the readability of this thesis, I have tried to compile the relevant physics and mathematical derrivations in appendices at the end of this report. In addition, an introduction to the COMSOL multiphysics software used for modelling as well as a presentation of my models can be found in one of the appendices.
2 Formation of the Solar system and the Earth
The foundations for today’s scientific models of the Solar system formation was laid in 1775 by the German philosopher Immanuel Kant (1724-1804)1. In what is known as the Nebula Hypothesis, Kant had come to realize that stars are born in the gravitational collapse of slowly rotating giant clouds. During the collapse the rotation of the cloud increases, due to conservation of angular moment, giving rise to an increasing centrifugal force. The gravitational force is spherically symmetric and directed toward the centre of mass, whilst the centrifugal force is cylindrically symmetric about the rotational axis, directed outwards. Thus the cloud will contract toward a disk shaped structure with a central bulge. In the central bulge, the gravitational force will out win the centrifugal force
1This has been contested by some historians, claiming that the idea originated from the Swedish scientist, philosopher, seer and theologian Emanuel Swedenborg (1688-1772) who had published it in his book Arcana Caelestia (1749). It is said that Kant who was one of only three persons who purchased this expensive work during Swedenborgs lifetime. Arcana Caelestia, which was Swedenborgs attempt to interpret spiritual meaning of every verse of the bible, covers Genesis and part of Exodus. Swedenborg is said to have claimed that the work had been revealed to him by angels, a provenance that according to some creationists disproves today’s scientific view. It should also be noted that a similar model was independently proposed in 1796 by the French mathematician and astronomer Pierre-Simon Laplace (1749-1827).
1 and the contraction will continue. However, due to the contraction, the temperature and gas-pressure will increase, eventually enough to balance the gravitational force and halting the infall, a central star has been born. In the disk, instabilities will cause local regions to start to gravitationally contract into the planets.
The main features of the Nebula Hypothesis have survived into the modern day theory of stellar formation which is summarized in section 2.1. Sections 2.2 and 2.3 will then present two rivalling scenarios for the formation of giant gas planets2, whilst the formation of terrestrial planets and two different core formation models will be presented in section 2.4
2.1 Star and circum stellar disk formation
Today we believe that stars are born in giant interstellar molecular clouds (e.g. the Orion nebula). Due to perturba- tions, local regions in the cloud becomes gravitationally instable3 and starts to contract. In addition the contracting region is likely to fragment into individually contracting regions during the collapse [87, 112]. Hence most stars are born in groups and preferentially as double or multiple stellar systems.
When the cloud contracts the temperature in the cloud will rise, which increases the gas pressure. In fact, unless energy is lost during the contraction the gas pressure will reach values high enough to stop the gravitational col- lapse long before any star can form. This is the reason why molecules are needed for star formation today4, as the dissociation of molecules provides a means for removing energy from the cloud. In addition to gas and molecules, the clouds also contains dust, typically gas to dust ratios are of the order of 1/100 [44], and the vaporisation of this dust also removes some energy from the cloud. Measured by astronomical standards, the time from onset of collapse to ignition of the first fusion reactions inside the proto-star (< 1 My) is very short, and once the star has lightened up the radiation pressure from the star will start to clear its surrounding disk from gas and dust. Observations of T Tauri stars (young stellar objects of approximately solar mass) has yielded that ∼ 30 % of the objects has cleared their disks within 1 My, while the rest has accomplished this within 10 My. Thus if planets are formed around these objects, planetary formation has to be a fairly rapid process.
During the formation of the disk structure, temperatures might reach values high enough to fully vaporise the dust content in the inner parts of the disk. But, as the disk has a large surface to volume ratio coupled to the low release of gravitational energy, the disk quickly cools by radiation. Minerals and metals can therefore re-condense rather quickly into small grains and dust particles, starting with high temperature condensates such as hibbonite (CaAl12O19), corundum (Al2O3) and grossite (CaAl4O7).
As the disk evolves so will the dust content of the disk, by coagulation the dust size will increase from sub micron up to mm sized as evident from observations [115]. Larger grains will settle toward the mid plane of the disk, thus creating a vertical stratification of grain sizes [91]. Observations has yielded that a significant part of the dust is of crystalline form, compared to the dust found in molecular clouds and the Inter Stellar Medium (ISM). This is generally attributed to a thermal annealing of the amorphous material in the very inner regions of the disk, from where it later on can be redistributed to the outer parts of the disk by radial mixing [139, 71, 43]. Observations of 5 the dust chemistry has indicated the presence of silicates such as amorphous olivine ((Mg,Fe)2SiO4), forsterite (Mg2SiO4), enstatite (MgSiO3), pyroxenes (both amorphous and crystalline, Mg0.8−0.9Fe0.2−0.1SiO3) and quartz (SiO4) [65, 48], as well as Polycyclic Aromatic Hydrocarbons (PAH, [50]).
Further out from the star, beyond the so called frost line, temperatures are low enough for ices like water, ammonia
2Although this is really of no importance for this work, these sections has been included for the completeness of this work 3The criterion for gravitational instability is really a local condition stating that the gravitational force is locally larger then the force exerted by the gas pressure. Consider then a giant molecular cloud where gas is being moved into new regions by e.g. the stellar wind from a newly born star or an old star going of as a supernova. This might cause nearby regions to become gravitationally unstable. 4 The first stars formed did not have access to molecules (except possibly H2) since in principle no elements heavier than He where formed in the Big Bang. As a reason for this the first generation stars are believed to have been extremely massive stars, but their formation is still today not fully understood. 5 Note that olivine is the name of a series of minerals where forsterite is one of the end members (the other being fayalite, Mg2SiO4). The reason for dividing between amorphous olivine and forsterite here lies entirely in the degree of crystal structure, giving rise to different spectra.
2 and methane (H20, NH3, CH4) [104] to condense. Increasing the mass fraction of solids by a factor of 2-3 [21]. Today two rivalling models for planetary formation in circum stellar disks exists, the core accretion model and the gravitational instability model. However these models are really concerned with the formation of gas giants, and the formation for terrestrial planets6 are in principle the same in both models.
2.2 Gravitational instability model
In the gravitational instability model, first proposed by Kuiper 1957 [75] and later discussed by Cameron [35] and Boss [26, 25], giant gas planets are formed from local gravitational instabilities in a circum stellar disk. As a result of this, the formation time of a planet is short (∼ 100 - 1000 years). Modelling has shown that planets are likely to form with on eccentric orbits, as observed in extra solar planetary systems.
As a planet starts to accumulate mass, non-negligible torques and drag forces are exerted on it from the surrounding disk, forcing the planet to migrate inwards (type I migration). However, once the planet has become large enough, it will clear a gap in the disk7 reducing the inwards migration (type II migration). In general type I migration occurs on timescales shorter than the disk lifetime, whereas type II migration is a slower process. As the planets of the gravitational instability scenario quickly grows large enough to clear gaps in their host disk, type I migration does not exist. Therefore the model can explain the presence of giant planets close to their stars.
However, the gravitational instability scenario demands very massive disks, approaching the upper end of what has been observed. In addition, gravitationally instability does not necessary mean that the disk fragments into gas giants. Planets formed are also very massive (approximately 10 Jupiter masses), making it hard to explain the gas giants of the Solar system as well as most extra solar planets observed. A further obstacle is that the model has problems explaining the presence of solid cores in gas giants.
2.3 Core accretion model
In the core accretion model, originally proposed by Safronov 1969 [118], the formation of gas giants can be split into three stages:
1. Accretion of solids and dust to form a core of a few Earth masses with a thin gas envelope 2. Continued accretion of dust and gas, but with higher accumulation rate of gas, resulting in a faster mass increase of the gas envelope. This proceeds until the mass of the core and the gas envelope approximately equals. 3. Runaway gas accretion.
As pointed out above, larger grains will settle toward the mid-plane of the disk. It is believed that during their settling the grains will also coagulate into even larger boulders (∼ 0.1 - 1 m). It is not fully understood today how these boulders then accumulate further into planetesimals, but various suggestions involves gravitational instabili- ties [54], secular instabilities due to gas drag [55] and ”sticky” particle collisions [39]. Once the planetesimals has reached sizes of ∼ 1-10 km they decouple from the gas and move in keplerian orbits. Continued accretion of dust, gas and collisions leads to the formation of the terrestrial planets and solid cores of gas giants. As the mass of the planetary embryos increases so does their feeding zone. Gas capture becomes an increasingly important mecha- nism, and a gas envelope builds up around the cores. During this stage an energy balance is established between the release of gravitational energy from accreting solids, and emergent radiation from the proto planet. However, once the mass of the gas envelope becomes approximately equal to the mass of the solid core, the balance can not
6By terrestrial planets we refer to solid planets, as opposed to the gas giants. 7A criterion for this can be set as when the Roche lobe of the planet grows larger than the scale height of the disk
3 be supported by the accumulation of solids only. The gas envelope starts to contract, increasing the rate of gas accumulation, which in turn increases the rate of contraction and so on. Eventually a runaway gas accretion onto the proto-planet occurs, resulting in the build up of a gas giant.
Two mechanisms has been discussed for the termination of the runaway gas accretion phase, disk dispersal and the opening of a gap in the disk (see footnote 7). Where the first might be important in order to explain the presence and structure of Neptune and Uranus, the second scenario seems critical in order to explain Jupiter mass planets and above. Finally follows a stage of further contraction until a hydrodynamical equilibrium of the planet is established.
Among the strength’s of the core accretion model are the mass range of planets that can be formed as well as the initial disk masses needed. The presence of solid cores in the gas giants are also naturally explained. However, the formation times are rather long due to the relatively slow second (and first) phase. A process that might well exceed the disk lifetime, also the orbits of the planets formed generally have low eccentricities. A final obstacle is that type I migration will be important, so unless the gas giants can form close to their stars it is hard to explain observed presence of hot Jupiter’s.
2.4 Formation of terrestrial planets and core-mantle differentiation
In principal the formation schemes for terrestrial planets are the same in the gravitational instability model and the core accretion model. The difference being that the formation of planetesimals precedes the formation of gas giants in the core accretion model, whilst they are assumed to form simultaneously or even after in the gravitational instability model. The following description of the formation of terrestrial planets assumes the core accretion model, but could as well be used for the gravitational instability model.
Not all planetesimals would have been accumulated into the giant gas planets, nor accumulated enough mass to trigger runaway gas accretion. Hence, at the time the proto-sun started to clear its nebula from gas, a swarm of planetary embryos would be present. This would be a highly unstable situation due to numerous close encounters and perturbations from the newly formed outer giant planets. Collisions would be frequent, ultimately leading to the final accumulation of the inner terrestrial planets [140]. It is estimated that the time of formation of the terrestrial planets is of the order of 40-200 My [102, 142], as compared to accumulation times of the gaseous giants8
From gravity and moments of inertia considerations it is known today that the terrestrial planets have a metallic core surrounded by a silicate mantle. Thus, during the formation or early evolution of the terrestrial planets a core-mantle differentiation has to have occurred. Radiometric dating (Pb-isotopes) of carbonaceous chondrites9 10 +0.002 and achondrites has yielded a time span from the formation of the Solar nebula (4.566−0.001 Ga) to the presence of magmatic activity and partial melting on planetesimals of only 1-8 My [9, 15]. While U/Pb ratios of the continental crust and 129Xe excess in the Earth’s atmosphere has yielded an age, at end of Earth accretion and early differentiation, approximately 100 My younger than the formation of the Solar nebula [9]. It is therefore likely that silicate/metal differentiation occurred early on in small planetesimals, so that the earth accumulated from already differentiated bodies. This is also supported by recent high precision W isotopic data from iron meteorites [81, 123], indicating the presence of differentiated bodies at very early ages. In fact some studies even
8As noted above, disk life times are expected to be below 10 My, thus the accumulation of gaseous giants has to occur within this time. 9Chondrites are one of two major types of stony meteorites, composed mainly of iron and magnesium bearing minerals. As chondrites originates from bodies that has never undergone melting or differentiation (asteroids) their overall elemental abundance (except for the H and He) is believed to reflect the original composition of the Solar nebula, at the site they formed. Thus chondrites can be used to infer the chemical stratification of the Solar nebula. In addition, nearly all chondrites contains chondrules, little round droplets of olivine and pyroxene that condensed during the cooling of the newly formed Solar nebula (before planetary formation), making them suitable for age determination of the Solar nebula. The chondrites is usually subdivided into subgroups of common properties, among these are the carbonaceous chondrites which are believed to be the most pristine of the chondrites. 10Achondrites are the second major type of stony meteorites and are made of rock that crystallised from a melt. Mainly composed of one or more of the minerals plagioclase, pyroxene and olivine, most achondrites are chemically similar to basalts, and are believed to have originated as melts on large celestial bodies that had either undergone or where in the process of differentiation.
4 suggest that the formation of differentiated bodies occurred earlier then the formation of the parental bodies of chondrites [72].
For differentiation to have occurred early in the relatively weak gravitational fields of the planetesimals rather high temperatures are necessary (> 1000 K [119]). Even though impacts will contribute with some energy this heat source would not be sufficient to sustain the temperatures needed given the short timescales. Instead suggestions has been raised that the heating was due to the decay of short-lived radio nuclides injected into the young Solar nebula by a nearby supernova11. In particular the isotopes 26Al and 60Fe have been discussed and modelling has yielded that differentiation is likely to occur given an early formation of planetesimals and reasonable estimates of isotope fractions [145, 92].
The formation of the Earth’s core from the planetesimals could then have followed two extremes. On one side the colliding planetesimals results in a relatively homogeneous mixing of the material. Accumulation of material onto the proto-earth proceeded by collisions with increasingly larger bodies, as also the planetesimals grew. A significant fraction of the gravitational energy released in these collisions was converted into heat. When the radius of the proto-Earth had grown to ∼ 2000-3000 km melting is likely to have occurred. Models have indicated that a completely molten magma ocean could have formed at shallow depths (< 300 km) while a partly molten magma ocean could have continued almost all the way down to the centre [70]. In the melt, hydrodynamically stable droplets of Fe (< 1 cm) would have been gravitationally transported toward the centre. Additionally, lakes of molten Fe could have accumulated on rheological boundaries, resulting in a gravitational overturn, upon which the Fe would have been transported deeper into the Earth and eventually all the way to the centre.
In the other extreme, as suggested by Saxena et al. [121], the collisions of differentiated bodies could have stripped of the outer layers of the planetesimals, leaving a primitive metallic proto-core (Fe-Ni-FeS-C). Whereas the proto- mantle could have accreted later on, partly by infall of the stripped material but also by accumulation of less dense chondritic planetesimals. As the proto-earth continued to accumulate mass, temperatures and pressures increased in the inner parts, with the result that silicates broke down into oxides. At pressures above 80 GPa perovskite transforms into a magnesiowustite¨ and a silica phase. This reaction depleted the lower mantle from silica, and the magnesiowustite¨ could dissociate into a MgO-rich and a FeO-rich phase. The core could then grow by reactions between FeO and the proto-core metal, accumulating even more Fe into the core along with O. The main-stream scenario today is rather the initially homogeneous Earth model than the proto-core model.
3 The inner structure of the Earth
The interior of the Earth is not homogeneous, therefore it is customary to radially divide the Earth into different regions (spherical shells) where bulk properties are alike. This can be done in two different ways (at least), one originating from a seismic velocity point of view (crust, mantle and core) and one originating from a mechanical point of view (Lithosphere, asthenosphere, mesosphere and core). However, the radial differences between the regions of the two schemes are small, and mainly important when considering the outermost part of the earth (lithosphere vs. crust). So as we are mainly interested in the inner parts we need not worry about this duality. Traditionally, when dealing with energy transport, the seismic vocabulary has been employed (mantle and core) and so shall also I.
In this chapter I will give a brief presentation of the mantle in section 3.1, and of the core in section 3.2, as well as presenting a seismological model of the Earth’s interior in section 3.3.
11Possibly also triggering the onset of the Solar nebula formation.
5 3.1 The Mantle
The mantle comprises the region from the Core Mantle Boundary (CMB, about 3480 km from the centre of the Earth) up to the thin crust (some tens of km thick). Its main constituent is believed to be magnesian silicates, in the forms of olivine, spinel and perovskite (ranging from lower to higher pressure). The mantle can further be divided into two sub regions, the upper and the lower mantle, separated at a depth of 660-700 km below the surface of the earth. This depth corresponds to the transition between the spinel and perovskite structure and can easily be seen in S-wave velocity profiles. Thus the olivine and spinel phases belong to the upper mantle, whilst the perovskite phase belongs to the pressures and temperatures of the lower mantle.
Over geological time the mantle behaves as a high viscous fluid. Calculations based postglacial uplifts has yielded viscosity values of the order of 1021 Pa s for the upper mantle and 1021 − 1023 Pa s for the lower mantle [49]. With a critical Rayleigh number12 of the order of 103 in the mantle, and an estimated value of the mantle Rayleigh number above 106 [136], it is clear that the mantle is convecting. This implies that the temperature profile of the mantle is close to adiabatic. Whether or not this convection occurs separately in the upper and lower mantle or throughout the full mantle is still debated though.
At the lower most 200 km of the mantle a sharp increase of S-wave velocity of 3-4% indicates a chemically peculiar layer, also refer to as the D”-layer. The chemistry and physics of the D”-layer are still poorly understood [63], but recent findings of a post-perovskite phase transition [93] could explain some of the features observed [14]. The temperature profile of the D”-layer is assumed to be very step, estimates ranges in the interval ∆T = 1000 - 1800 K over the layer [31, 130], indicating that conduction is the major transport mechanism of heat over the D”-layer.
3.2 The core
As in the case of the mantle, also the core can be divided into two sub regions, the outer and the inner core. The liquid outer core stretches from the Inner Core Boundary (ICB, at a radius of 1221 ± 1 km) out to the CMB, just below the D”-layer. Equivalently this corresponds to about 15.6 % of the Earth’s volume or 30.8 % of the Earth’s mass due to the high density of the core. It is generally believed that the outer core mainly consists of a Fe-Ni mixture13 (8 ± 7 wt% Ni [68]), alloyed with small amounts of lighter elements [19]. Based on cosmochemical abundances, various suggestions for the lighter elements covers S, Si, Mg, O, C and H [3, 20, 113, 137], although today S, Si and O are considered to be the top candidates. In addition it has been claimed that the core contains more than 90% of the Earth’s entire inventory of Highly Siderofile Elements (HSE: Ru, Re, Rh, Pd, Os, Ir, Pt and Au) [108].
Estimates of the viscosity of the outer core spans over the order of 14 magnitudes [125], but with an upper limit of the order of 1012 Pa s, the viscosity is still much lower than in the mantle. Hence the outer core is vigorously convecting, generating the Earth’s magnetic field that shields life on the surface from a lethal bombardment of charged particles from space. Due to the convection, the outer core is generally considered to be adiabatic (isen- tropic) and well mixed, with a hydrostatic pressure gradient. It is not known at what time the georeactor became active. However, the earliest records of the geomagnetic activity dates at least 3.5 Ga [90]. Thus, a sufficiently large portion of the core must have been liquid by that time.
In contrast to the outer core, the inner core is a solid body, comprising approximately 0.7 % of the Earth’s volume or equivalently 1.7 % of the mass. The consensus is that the present inner core has solidified over time from the
12The Rayleigh number is a dimensionless measure of a fluids stability to thermal convection. As the quantity is not normalised and depends on the fluids thermal and thermomechanical properties, as well as geometry and thermal gradient, the critical Rayleigh number marks the boundary between stable conditions (no convection for lower Rayleigh numbers) and unstable conditions. 13An alternative proposal for the composition of the Earth’s core was put forward in 1979 by Herndon [60]. Based on the observation of oxygen poor enstatite chondrites Herndon came to the conclusion that the core should consist of nickel-silicide (Ni2−3Si) instead. However, as Ni is about 5 % as abundant as Fe [86] this would lead to an apparent lack of Fe and over-abundance of Ni in the Earth, as compared to Solar system abundances.
6 liquid outer core, and that the growth continues today as the inner core cools14. By density considerations, it is believed that its composition is similar to that of the outer core, but contains even smaller amounts of lighter (than Fe, Ni) elements.
The inner core shows some degree of anisotropy. It has been known since the early eighties [107] that P-waves travelling through the core parallel the rotational axis of the Earth has a shorter travel time than P-waves travelling in the equatorial plane. More surprising is then the recent observation that there exists an anisotropy between the eastern and western hemisphere. It appears that P-waves travelling parallel to the rotational axis in the western hemisphere arrive about 4 s faster than P-waves travelling in the equatorial plane, while P-waves travelling in the eastern hemisphere only shows an anomaly of about 1 s. Several explanations has been put forward to the directional anisotropy ranging from solid-state convection, magnetic alignment to anisotropic growth (for a review see [135]).
Another peculiarity of the inner core is that the rotation of the inner core seems to be somewhat faster, than the rotation of the mantle [135, 149]. This is consistent with predictions from numerical models of the geodynamo, even though the measured super-rotation is somewhat slower than the predicted one. Numerical values ranges from about 1 deg/year to about 0.2 deg/year.
3.3 The PREM model
The most widely used model of the Earth’s present interior structure in terms of density, pressure, seismic veloci- ties and mechanical properties, is the Preliminary Reference Earth Model [46] (PREM). PREM is an azimuthally averaged model, based on an extensive Earth data set15 where the Earth is radially divided into sub regions, separated by seismological discontinuities. In each region the parameters of the model are either given as low- order polynomials or tabulated (see appendix A). It should be noted that even though the Earth is recognised as anisotropic, tabulated values are given for the ”equivalent” isotropic Earth. Meaning that the tabulated model has approximately the same bulk and shear modulus as the anisotropic model, not that it provides an equivalent, or satisfactory, fit to the data16. It should also be noticed that the Earth is dispersive, i.e. seismic wave speeds are frequency/period dependent, therefore the PREM model are given for two reference periods, 1 and 200 s. For other periods, T, the velocities must be modified according to the equations:
lnT V (T ) = V (T = 1s) 1 − Q−1 s s π µ (3.1) lnT V (T ) = V (T = 1s) 1 − (1 − E)Q−1 + EQ−1 p p π K µ where 4 V 2 E = s 3 Vp
And QK and Qµ are the bulk and shear quality factor.
14Note however that in the proto-core scenario, the present inner core could be a remnant (partly or fully) of the initial primitive proto-core. In the extreme scenario we could also have a situation where the inner core even diminish in size due to chemical reactions at the ICB. This is not a likely scenario though. 15including total mass, moments of inertia, free oscillations/normal modes, normal mode Q-values, long period surface waves and body waves 16The anisotropy is recognised in the upper mantle, in the region between 24.4 km to 220 km depth. As the model is azimuthally averaged, it is effectively the spherical equivalent of transverse anisotropy, with the symmetry axis along the vertical (radial) direction.
7 (a)
(b)
Figure 1: Some of the tabulated PREM model parameters for a reference period of 1 s
4 Probing the Earth’s interior
In order to understand the past, present and future of our planet we need to know its interior. However, the interior of the Earth is inaccessible to us for in situ explorations and measurements. Today, the deepest we have penetrated into the Earth is about 12 km (The Kola Superdeep Borehole, TKSB), which in comparison with the Earth’s radius of about 6371 km means that we have only scratched the surface, and this is not likely to change in the future. Thus we need to have methods by which we can measure the interior from a distance. In this chapter I will briefly present different methods used to investigate the interior of the earth. But it should be recognized that the methods presented here does not present a full coverage of the methods available.
8 4.1 Mass and density distribution
From Newton’s law of gravity we know that we can measure that mass, M, of the Earth by measuring its gravity which today can be done by careful laboratory experiments, alternatively we can infer the Earth’s mass from the orbits of satellites or the acceleration of spacecrafts. What is actually measured is not the Earth mass though, but rather the product GM, where G is Newton’s gravitational constant. And so the major uncertainty in today’s value of the Earth’s mass, 5.977 x 1024 kg, resides in the determination of G, which is much harder to measure17. However, this will not give us any information on the density distribution in the deep interior of the Earth. A measure of the density distribution can be found from the moment of inertia of the Earth (about 8.0 x 1037 [116]), which can be estimated from the astronomical precession constant and the second degree zonal harmonics of the geo potential. However the uncertainty in the numerical value is large, resulting in a large uncertainty of the Earth interior.
The best methods available are instead seismological. As wave speeds through a material are inversely pro- portional to the density of the material we here have an opportunity to get detailed information on the density distribution of the Earth’s interior. To be able map the interior of the Earth, the waves naturally needs to pass through the interior, hence a sufficiently powerful source is needed in order to generate these waves. In general the seismic sources of the deepest parts of the Earth are naturally occurring Earthquakes (although underground tests of nuclear weapons have also served as sources.). As a rule of thumb the longer wavelength the deeper penetration of the wave is possible. As we are mainly interested of the inner core in this study the wavelength of the waves needed, are of the order of 10-100 km. This sets an upper limit to the resolution that can be achieved in the inner core, since the waves will not be sensitive to the existence of lateral heterogeneities on scales much below its wavelength. However, due to the fact that the sources cannot be controlled by the scientists, the actual resolution is generally worse. In addition to seismology, studies of the Earth’s eigen (normal) modes, triggered by large earthquakes, yields further constraints on the density distribution [59, 135].
4.2 Elastic properties
In addition to the density, both seismic waves and normal modes are dependent on the elastic properties of the Earth’s interior. Seismic waves can be divided into S- and P-waves depending on the particle motion18. Whereas the velocity of P-waves are dependent on both the bulk modulus and the shear modulus of the material it propagates through, the S-wave velocity only depends on the shear modulus, and so as the shear modulus vanishes for a liquid, S-waves can not propagate through liquids. Due to this S-waves can not travel through the outer core. However, when a seismic wave impinges on a boundary at an angle other than 90 degrees, a phenomenon known as mode conversion occurs, resulting in at least two waves, one S-wave and one P-wave. Therefore S-waves will exist in the inner core. Now, S- generally have a smaller propagation speed than P-waves, hence the S-waves will arrive to the observer after the P-wave. A seismometer measuring waves that have travelled through the Earth will therefore detect several wave trains stemming from the same event. This way seismologists have the possibility to measure the properties of the inner core.
4.3 Composition
The composition of the crust and mantle can be studied geologically on the surface of the Earth, as different geological processes has transported material from these regions up to the surface. Unfortunately this is not the case for the core. Instead the composition of the core has to be inferred from mass constraints and elemental abundances of meteorites, believed to be the ancestors of our home planet. It is generally believed today that the bulk composition is close to that of chondrites (see footnote 9)
17The value recommended by the Committee on Data Science and Technology (CODATA) from 1998 has an uncertainty of 0.15 %, however more recent measurements [124] yields uncertainties on the ppm level. 18For the P-wave the particle motion is parallel to the propagation of the wave
9 4.4 Heat
In investigating the geotherm of the Earth it is important to know the heat flow of the Earth. At the surface this can simply be measured, yielding a current value of 44.2 TW [106]. The heat flow over internal surfaces of the Earth are not known, but reasonable estimates of the heat flow at the CMB ranges in between 5-10 TW [38, 97, 100], whilst the heat flow at the ICB is generally believed to be well below 1 TW. However these numbers are highly sensitive to the presence of radiogenic isotopes in the core. Now, knowing the heat fluxes in the Earths interior yields values of the energy available for driving various motions (e.g. mantle convection giving rise to plate tectonics, or outer core convection driving the geodynamo). It is therefore of great importance to know the distribution of heat sources throughout the Earth. One way to do this could be by detecting the neutrinos produced in the decay [110] as we already today posses detectors appropriate for this [45, 52, 110], in fact the first reports on successful detections of 28 geo-neutrino has already been published [13].
4.5 Laboratory measurements
As we move further into the Earth’s interior, both pressure and temperatures increases, hence a study of the interior of the Earth is a study of the behaviour of material properties under high pressures and temperatures. In order for us to be able to relate the results from the techniques mentioned above to each other, and to the geology of the Earth, we need to now this behaviour in advance. E.g. where we to assume that the density of solids does not change as pressure and temperatures increase, we might be tempted to believe that the inner core (ρmean ≈ 12.9 g/cm3) consisted mainly of Hf (ρ ≈ 13.31 g/cm3) or some even heavier element mixed with some lighter element. Even if we somehow knew that the major constituent should be Fe, our assumption would force us to mix in at least 34 % of elements denser19 than Hf, which would be very hard to accept from an elemental abundance point of view.
It is therefore of great importance that we study the behaviour of different materials under high pressures and temperatures. There are two different techniques used today to achieve this, shock wave (SW) experiments and Diamond Anvil cell’s (DAC)
4.5.1 Shock wave experiments
In SW experiments, the pressure in the studied target is almost instantaneously increased to high values, generating a shock wave that propagates through the target in fractions of a second. Several methods of achieving this exists, including, contact explosives, high speed impact of solid bodies, and rapid deposition of energy by high power laser irradiation [143]. The quantities measured during the experiments are, shock wave velocity, particle velocity, and temperature and pressure, which can then be related to other mechanical and thermodynamical quantities of the studied sample (for a review of different measurement techniques see Ahrens [2]). The peak shock states of several SW experiments, starting from equal initial conditions and reaching different pressures, are embodied in Hugoniot curves, or Hugoniots. It is of importance to recognize that these curves does not represent thermodynamical paths followed by the material under the influence of increasing pressure. The actual thermodynamical path followed by the material under specific experiment is instead a straight line from the initial state to the final state, referred to as a Rayleigh line. Likewise the Hugoniots are neither isotherms20 nor isentropes (or adiabats21), but isotherms and isentropes can be derived from the Hugoniot.
The main advantage of shock wave experiments over static pressure experiments is the pressure range available. In SW experiments pressures beyond TPa (1012 Pa) have been reached (e.g. Ragan [109] used neutrons from an
19There are 17 naturally occurring elements heavier than the mean density of the inner core, Hf, Ta, W, Re, Os, Ir, Pt, Au, Hg, Pa, U, Np, Pu, Am, Cm, Bk, and Cf. The densest of these, Os, would demand a presence of 34 %. 20Higher temperatures are reached for higher pressure experiments. 21Some heat is always dissipated through viscosity, plastic deformation and other dissipative processes.
10 underground nuclear explosion to investigate the behaviour of molybdenum at 2 TPa). However, disadvantages include difficulties in measuring the temperature, as well as the question whether the material behaves equally under shock conditions as under static conditions.
4.5.2 Diamond Anvil Cells
In a DAC the sample is uni-axially compressed between two diamonds. By keeping the flat culet22 surface of the diamonds small (0.05 to 1 mm depending on the pressure range of interest) high pressures are easily achieved. The sample is then placed in an even smaller hole (∼ 0.3 - 0.5 µm) of a metal gasket, placed in between the diamonds culets, where after the pressure is increased. It is possible to reach pressures of the order of 350 GPa using DAC.
In addition to the obvious advantages of being able to keep the sample under high pressure for an extended period of time, diamonds are transparent to a large portion of the electromagnetic spectrum. This means that we can directly observed our sample, as well as examine it using techniques such as mossbauer¨ spectroscopy and various diffraction methods. It is also possible to alter the temperature of the sample using directed laser beams or electrical resistive heating. The uni-axial pressure can be transformed into a hydrostatic pressure by surrounding the sample by some fluid pressure medium, or into a quasi-hydrostatic stress state by the use of a soft solid pressure-transmitting medium23.
4.6 Theoretical
The theoretical work relating to the interior of the Earth can coarsely be divided into three parts
• Macroscopic behaviour of the Earth. Including geotherms, convection, and the geodynamo. The results of such work relating to this work will mainly be dealt with in sect.6 • Behaviour of the Earths constituents under high pressures and temperatures. In addition to laboratory measurements theoretical studies of materials behaviour under high pressure, has yielded new insights and constraints on the interior of our planet. The results from such studies regarding Fe will be presented in sect. 5. • Extrapolation of material properties. As the measured high pressure properties only covers a limited pres- sure range, an extrapolation of the values to higher pressures might be needed for the deeper parts of the Earth, e.g. melting curves. A solid theoretical basis, i.e. an Equation Of State (EOS) is then necessary for appropriate extrapolation. For a collection of various EOS used for extrapolation of different quantities see Poirier [105]
5 Fe at high Pressure and Temperature
As the Earth’s core is believed to mainly consist of Fe a lot of effort has been put into examining the high pressure and temperature behaviour of elemental Fe. Using DAC, SW experiments, and theoretical modelling the phase diagram, thermal and elastic properties of Fe has been revised several times over the last two decades, although large discrepancies still exists. In this chapter we will review the results from such studies relevant for this work.
22The culet on a diamond is the tiny flat facet that is formed by polishing off the tip at the bottom, where all the facets of the pavilion (bottom part of the diamond, below the girdle) otherwise come to a point. 23All known liquid pressure media freeze at pressures above ∼ 16 GPa at room temperature [42], hence at higher pressures a hydrostatic pressure is not achievable.
11 5.1 Fe phase diagram and melting curve
At ambient conditions Fe have a body centred cubic (bcc) structure, also known as α-Fe. Increasing the tempera- ture, this transforms into a face centred (fcc) structure at 911 oC, also known as γ-Fe, and at even higher temper- atures again to a bcc structure (δ-Fe) at 1394 oC, before melting at 1808 oC. Increasing the pressure transforms the Fe-structure to hexagonal close-packed (hcp), also known as -Fe. It has been suggested that at increasing temperature, the -Fe undergoes a transformation to a β-phase with a double hexagonal close-packed structure [120, 40] or an orthorombicly distorted hcp structure [12]. However, the existence of the β-phase has been de- bated [74, 111, 126] and its presence is still not settled. In addition it has been indicated from SW experiments that an additional phase, possibly a bcc structure, develops at pressures above 200 GPa [29]. Hence a large uncertainty is still associated with the phase diagram of Fe (see figure 2)
(a) (b)
Figure 2: Phase diagram of Fe. Left figure from Boehler [24], triangle and dots indicate individual measurements of Fe melting temperatures (see original paper for details). Right figure modified from Nguyen & Holmes [99], triangles, diamonds and dots indicate Fe-melting data from SW experiments, (see original paper for references), dotted lines for the location of the CMB and ICB has been added to the original figure and references in the figure has been removed as they do not agree with the numbering in this work.
Neglecting the solid phases of Fe and concentrating on the melting temperature only, we find a scatter at the ICB of about 1700 degrees between 5400 (± 400, [80]) K to 7100 K [18] (see figure 3). In general it can be said that SW [30, 141, 144]24 experiments yields somewhat higher melting temperatures than DAC measurements [23, 122, 126], whilst theoretical experiments scatter all over the temperature range [5, 18, 80] (in fact both the upper and lower limit are set by theoretical considerations). It needs to be remembered though that not all experiments has reached ICB-pressures (this especially holds for DAC measurements) and the actual ICB melting temperatures presented are relatively sensitive to the extrapolation performed.
As pointed out above the composition of the core is not pure Fe, but rather an Fe-Ni alloy with some lighter elements, Si, S, and O being the most favoured candidates. This will also affect the melting temperature of the core material. It is generally considered that the inclusion of Ni will not have a major effect on the melting temperatures, which can not be said for the lighter elements. Using ab initio simulations of chemical potentials at core conditions, constrained by core densities, Alfe et al. [4, 6] came to the conclusion that the outer core contains 8.5 ± 2.5 molar % S and/or Si plus 10.0 ± 2.5 molar % O, while the inner core contains 0.2 ± 0.1 molar % S and/or Si plus 8.0 ± 2.5 molar % O, resulting in a melting temperature depression of -700 ± 100 K. Whereas Andersson [10] considers several mixtures of S, Si and O, yielding ICB melting point depressions in the range -700 to -2271 K. However, DAC measurements of Boehler [22, 23] indicates that the melting depression observed in the Fe-FeO-FeS systems diminishes at increasing pressure, which could possibly affect the theoretical results.
24Note that the Williams [141] used a combination of SW and DAC
12 Figure 3: Comparison of Fe melting curves from theoretical (heavy solid, long dashed, dotted, and light solid curves as well as filled circles)and experimental results, including both DAC measure- ments (chained and short dashed curves, and open diamonds and stars) and SW experiments (open squares, open circle and full diamond). reprinted from Alfe [8] (see original paper for references).
5.2 Thermal and elastic properties
Table 5.2 summarizes values found in the literature of the last decade, for a number of thermal and elastic pa- rameters of hcp-Fe at inner core pressures. Although some of the values are based on results from high pressure experiments, it should be remembered that most of the quantities are not directly measurable, but rather rely on the measurement of some other quantity which they can be related to. In addition, most of the parameters are both pressure and temperature dependent [7], and so an interpolation of experimental values to inner core pressures and temperatures is necessary. Thus, a number of possible sources of uncertainty exists in the presented numerical values, including:
• Data in all measured data there is always a minimum level of uncertainty. This will naturally set the minimum uncertainty of any parameter derived from that data. Unfortunately not all values are presented with their estimated uncertainties as they really should be. • Theory The relation between a measured quantity and a derived quantity goes via some theory, hence the derived quantity will only be as good as the theory. • Interpolation I Today, only SW experiments has the capability to ”easily” reach inner core pressures and beyond25. But they do so on Hugoniots, so unless the initial conditions are precisely the right, the tem- peratures in SW experiments at inner core pressures will differ from inner core temperatures. In general, interpolations are needed both with respect to pressure and temperature. Now, to do an interpolation one needs to have an EOS relating the interpolated quantity to the variables. In the geological high pressure community a number of such EOS flourish (for a recent review and discussion see [132, 130]). The use of two different EOS will generally give some variations in the derived quantity. Now, the EOS is derived from some theory, however, this is not always the same theory used to relate the derived quantity to the measured quantity, or possibly a very simplified version, which is the main reason why we split theory and interpolation here.
25Although DAC experiments is closing in, e.g. [88, 41]
13 • Interpolation II To do an interpolation we need at least one known data point with known coordinates26 as well as the coordinates of the desired data point. Now as discussed in sect. 5.1, we do not know the inner core temperatures very precisely. In addition we might not know the coordinates of known data points to a high precision either. So unless the interpolated quantity is relatively insensitive to variations in the coordinate values, this could seriously affect the presented data. It is therefore important to check what assumptions that has gone into the interpolation.
Anderson [10] uses the Wiederman-Franz law and the electrical conductivity found by Matassov [89], to find the −1 −1 electronic thermal conductivity, ke, for the core to be 39 W m K . Adjusting for the lattice contribution, kl, (assumed to be about 4 W m−1 K−1), he presents a value of k of 43 W m−1 K−1. Although, it is recognized that values of Matassov might need correction that could lead to a value of k of the order of 60-70 W m−1 K−1. Stacey [131] also uses the data of Matassov but imposes corrections on them, interpolates to core values and compares the values to those measured by Brigman [27], resulting in a value of k of 79 W m−1 K−1 in the inner core. It should be noted though that Stacey considers an Fe-Ni-Si alloy, and Anderson considers a Fe-Si alloy, both papers finds k(140 GPa) to be 43 W m−1 K−1.
−1 −1 −1 −1 −5 −1 k [W m K ] γ Cp [J kg K ] α [10 K ] KT [GPa] T [K] Ref. - 1.45 - 1.2 - 6000I [7] - 1.27 - - - 5100II [10] - 1.52 - - - 5709 [11] - - - - - 7100 [18] - 1.28III - - - - [41] 60 1.5 715 1.95 - 1.02IV - 5500 [57] - - - - - 5400 [80] 79V - - - - 4971V [131] - 1.38VI 860VI 1.32VI - 4500VI [133] - 1.27VII 826VII 1VII 1194VII 4500VIII [129] - 1.5 - 1.2 900-1177IX 6000X [138]
Table 1: Various values of thermal and elastic properties for Fe at inner core pressures. The column T refers to the temperature at which the tabulated value is given. A - sign indicates that no values are given for this parameter. I Values for T = 4000 and 2000 K can be found in the original paper. II It is assumed that γ is independent of T, otherwise this temperature should be revised. III Value given at a radius of 1400 km. IV Values given from ICB-pressures to centre of Earth. V Values given for Fe-Ni-Si alloy matching inner core density (10 % Ni). VI Tabulated values are given at P = 243 GPa (mid core) VII Values given at 1400 km radius VIII Temperature given at 2400 km radius IX values tabulated for densities 12.37 and 13.31 g cm−3 of hcp-Fe, approximately equal to expected hcp-Fe densities at ICB and centre of Earth. X Values for T = 4000 and 2000 K can be found in the original paper.
Anderson [10] also finds the pressure derivative of the bulk modulus at zero pressure, K’0, from PV of hcp-Fe data of Mao et al. [88] and the Birch-Murnagahan EOS. Using a thermodynamical relation for hcp-Fe, derived by Stacey [129] he then finds γ at zero pressure. He finally interpolates to ICB pressure by assuming27 that γ ∝ ρ−0.7, yielding a value of 1.27. Where as in a later paper Anderson [11] relates γ to the vibrational and electrical Specific heat at constant pressure, CV,vib and CV,el. Using data by Stixrude [134] he then finds a value of γ of 1.53. Other examples of experimentally determined quantities comes from Dubrovinsky et al. [41], who used high-quality powder X-ray diffraction data, collected in DAC’s up 285GPa, to determine the γ to be 1.28 at a radius of 1400 km (i.e. about 180 km above the ICB)28
26I.e. numerical values of the variables, e.g. temperature and pressure in our case 27the value 0.7 is constrained from earlier theoretical estimates of γ by Stacey, [129] and 28Unfortunately the derived value of γ rests on the use of a thermal equation of state of hcp-Fe, from an unpublished manuscript by Dubrovinsky et al. I have not been able to find out whether this paper has been published afterward.
14 In addition to experimental measurements, the advancements in both theory and computer technology has made it possible to perform thermodynamical ab initio simulations of the high pressure and temperature behaviour of materials. E.g. Gubbins et al. [57] and Alfe et al. [7] both computes Helmholtz free energy of hcp-Fe using Density Functional Theory (DFT), upon which a number of parameter values are derived. An alternative route is taken by. Stacey [129] who combines results from molecular dynamical calculation by Barton and Stacey [17] with thermodynamical relations and derivatives to find numerical values for a number of parameters.
6 Models of the Earth’s Core
Any model describing the Earth’s core demands a knowledge of the temperature profile, as this relates to both material properties and the energy present. Now, in a self consistent model the temperature should therefore always be a parameter solved for, even if the primary target is not the temperature profile. An alternative approach would be to use a known temperature profile, e.g. resulting from a self consistent modelling. The question is then whether we can construct a simple self consistent model that will return only the temperature profile. To answer this question let us have a look at the requirements for a thermal model, starting with the inner core only (as this is the region of interest of this thesis).
6.1 Thermal models
As it turns out, building a thermal model of the inner core is not a trivial problem. A closer look at the problem yields that we are facing the problem of solving the non-homogeneous transient heat equation in spherical coor- dinates. Non-homogeneous, since that even if we do not have any radiogenic isotopes in the inner core, we still some contributions to the heat budget due to the release of gravitational energy when the inner core contracts upon cooling. Transient since the nature of the possible inner core heat sources are transient by nature. However, we know that the inner core is solid, i.e. we need not worry about convection. To simplify the problem even further, we can assume the inner core to be spherically symmetric, hence we can settle with a 1D solution over the radius of the inner core. So what are then the boundary conditions? At the centre of the core our assumption of spherical symmetry implies that the heat flux should be equal to zero.
∂T k = 0 (6.1) ∂r r=0 However, at the ICB we have a moving boundary, so we are facing a Stefan problem, but even worse, the boundary condition at the ICB is dependent of the position of the ICB, which in turn is dependent on the heat flow through the outer core. Thus we also need to solve the heat equation in the outer core, but as this is fluid and vigorously convecting we can no longer confine ourselves to a 1D solution. In addition, as convection means conversion of heat into kinetic energy, that might be lost due to dissipation (e.g. the generation of a magnetic field), we can no longer only solve the heat equation, the problem has become a multi-physics problem.
And so it continues as we realize that we cannot solve the heat equation in the outer core without knowledge off the heat flow over the CMB, which is dependent on the heat transported through the mantle etc. In fact we have to continue all the way to the surface of the Earth (or even the upper layers of the atmosphere) before we can find a rigid boundary condition. It should be evident by now that no analytical solution can be found to the problem, hence we are left with numerical schemes. Now, building a self consistent full Earth model is not a feasible task. What we must do is to make some physical assumptions of the boundary conditions at a proper radius, and construct a model using these assumptions.
15 6.2 Earlier core models
Several core models has been found in the literature. Labrosse et al. [78, 79] considered models constrained by the CMB heat flux, Labrosse and Macouin [77] and Labrosse [76] constructed models constrained by ohmic dissipation. Gubbins two models compares different core composition (one including compositional convection [56] and one without [57]) to investigate the heat fluxes needed to drive the geodynamo. Where as Butler [33], Costin [38], Nakagawa [97], Nimmo [100], and Yukutake [148] included mantle models to constrain the heat flow at the CMB. The most striking result from the different models is the young age of the inner core. With the exception of Gubbins models, these all fall in the range of about 1-2 Gyr (e.g. see figure 4(a)). Likewise, the models solving for the CMB heat flow all displays values in the range 8-9.5 TW, and ICB heat flows in the range of 0.2-0.5 TW (e.g. see figure 4(b)). Another interesting feature is that many of the papers present several models with different contents of radiogenic isotopes, but this shall be discussed further in section 6.4 below.
Figure 4: Time evolution of the inner core radius after onset of inner core solidification, and various contributions to the heat balance of the outer core. Reprinted from Labrosse et al. [78], for model parameters see original paper.
In the models presented above, it is generally assumed that the inner core geotherm is close to isothermal. I have only found one paper, by Yukutake [147], that actually considers the inner core geotherm. Unfortunately this paper was found in a very late stage of my work, when all my work was done and this report almost finished. The aim of the paper is to investigate the claim by Jeanloz and Wenk [69] that the inner core should poses thermal convection. One criterion for thermal convection to occur, is that the gradient of geotherm should be steeper than the adiabatic gradient, and so in order to investigate the claim, Yukutake sets up an evolutionary model of the inner core geotherm. The model assumes pressure induced freezing of the inner core (see section 6.3) and an adiabatic temperature profile of the outer core. As initial condition the adiabatic temperature equals the melting temperature at the centre of the Earth, where after the adiabatic temperature is assumed to decrease linearly with time.
For every time step, δt, the radius of the inner core is initially evaluated where after the temperature profile is evolved, using an analytical expression29 given by Carlslaw and Jaeger [37]30 for the transient heat equation with a varying boundary temperature. As a minimum age of the inner core 1.5 Gyr used and a maximum presence of 100 ppm 40K is considered. For all models it is found that the inner core geotherm gradient is below the adiabatic gradient, or equivalently a maximum temperature difference today between the ICB and centre amounts to 129
29The analytical expression is found using Green’s functions (see eq.(B.23) in appendix B), resulting in an infinite series solution. 30Reference taken from paper [147], as I could not gain access to the referred book.
16 degrees, which is less than the adiabatic temperature difference of 143 degrees. Hence the conclusion is drawn that the inner core geotherm is always sub-adiabatic.
6.3 Inner core solidification
Inside the Earth the pressure monotonically increases with depth, reaching a maximum value of about 364 GPa at the centre. This will increase the melting temperature of any material. We know that the geodynamo has been operational for at least 3.5 Gyr [90] which implies that the liquid outer core has been convecting throughout this period, which in turn implies that the geotherm of the outer core has been close to adiabatic throughout most of Earth’s lifetime. It can be shown that for reasonable values of the Gruneisen¨ parameter, the gradient of the core adiabat will be less step than the melting temperature of the inner core. Consider an initially molten core, the minimum difference between the core geotherm and the melting temperature of the core will then be at the centre of the Earth. As time elapse, the Earth cools and eventually the core geotherm will be equal to the meting temperature of the core. The inner core then starts to solidify. As time goes on the Earth cools further and the intersection between the core geotherm and the melting curve will progress toward larger radii, causing the inner core to grow. At every instant having a temperature at the ICB equal to the melting temperature of the core material (see left panel in figure 5). I.e. the inner core solidifies from inside out. From now on we shall refer to this scenario as pressure induced freezing.
Even though pressure induced freezing is the most frequently encountered model for inner core growth, this is a relatively simplified picture. The reason being that the core is not of elemental composition, but rather a mixture. When a mixture freeze (unless of eutectic composition) it will not have a clearly defined melting temperature, but rather a melting temperature interval (see figure 23(b) in section C.6), under which the composition of the solid phase will differ from the composition of the liquid phase. In the case of the outer/inner core this is evident from the increased density of the inner core, which can not be explained from a volume change upon solidification only. Thus the interpretation is that the fraction of lighter elements is lower in the inner core than in the outer core. And so the temperature of the ICB is bounded in the freezing interval of the core material, with upper limit set by the melting temperature of the inner core composition (see right panel in figure 5). This has certain implications for the solidification of the inner core. To start with the solidification of the inner core will not only take place at the ICB, but rather in an extended region, stretching out into the outer core. This will enrich the solidification region in lighter elements, giving rise to a compositional convection in the region, transporting lighter elements out into the outer core, and accumulating more material onto the inner core. In addition it is likely that the ICB is not a distinct interface but rather a mixture of solid and liquid material, with an inwards increasing fraction of solids, i.e. a mushy layer will develop on top of the core. In fact, some indications for the existence of such a layer has emerged from recent seismological studies of the inner core [36, 73, 146], for one thing, instabilities in such a mushy layer could explain the growth of the ICB by 1.37 ± 0.38 km between 1993 to 2003 as observed by Lianxing [85]. Note that the enrichment of lighter elements in the solidification zone will alter the melting temperature interval of the liquid fraction, possibly leading to a state of super cooling, however, a recent study by Shimizu [128] has indicated that the cooling rate of the Earth, and hence growth rate of the inner core is far to low for this to happen.
It should be recognized though, that the expected magnitude of the difference in the melting temperatures of the inner and outer core alloys is small, at maximum a few tenth’s of degrees. Given then that the adiabatic gradient at the ICB is of the order of 0.37 K km−1 (see figure 22) the region of inner core solidification will be at maximum about 100 km wide.
Finally we shall mention a third effect that will influence the solidification of the inner core. As recognized by Gubbins et al. [57] the solidification of the inner core is associated with a small volume decrease, which will cause a small contraction of the core and hence an small increase of pressure. This in turn will increase the melting temperature, and therefore also increase the size of the inner core. Even though one would naively consider the effect to be small, the model constructed by Gubbins et al. indicates that this could account for as much as 10-20 % of the inner core growth. However, it should be said that the model presented by Gubbins et al. displays several peculiarities, like an inner core age of only a few hundred Myr and heat fluxes of the order of 30-50 % of the
17 Earth’s current heat flux at the surface. In addition it is not fully transparent (at least not for me) exactly how Gubbins et al. evaluates the effect.
Figure 5: Models of inner core solidification. Blue line indicates the core geotherm, red line indicates the melting temper- ature of the core. Left panel displays the pressure induced freezing scenario and right panel displays the scenario with different melting temperature of the inner and outer core alloy.
6.4 Radionuclides in the core
The partitioning of K into the core was suggested on geochemical grounds in the early seventies [53, 58, 84], especially if S was the main light constituent of the core. In the beginning of the nineties this idea was opposed by results from simulations and high pressure experiments [67, 127]. But, as time went on and experiments was conducted at even higher pressures, evidence for the possibility of K partitioning into the core accumulated [51, 95, 103], culminating with the estimate by Lee and Jeanloz [82] that up to 0.7 wt.% of K, could be incorporated into hcp-Fe at core pressures, yielding a present day abundance of about 580 ppm. Assuming the initial composition of the Earth to be chondritic, and that no K was lost during its accumulation and evolution, an upper limit is set to about 0.142 wt.% in the core today [114]. On the other hand, given the age of the Earth and the half-life of 40K (see table 2), we have that this equals approximately 1/12 of the initial 40K. Since 0.142 wt.% 40K in the core would produce a total of about 9.56 TW today (see table 3), this would imply a very high heat production in the early days, which could be hard to defend.
Other indications for the presence of radionuclides in the core comes from different models of the core and the operation of the geodynamo [33, 38, 97]. The reason is that most core models indicate a relatively young age of the inner core (see section 6.2), as well as cooling of the core being an insufficient energy source for the geodynamo. Thus these models more or less demands the presence of radionuclides for the geodynamo to be operational over geological time. Estimates on the amount of radionuclides necessary ranges from 100 to 600 ppm.
In addition to K, also U and Th has been considered as possible candidates in the core [127, 94]. Even though the general opinion seems to be against U and Th in the core, recent experiments has indicated that small amounts, of the order of a few tenth’s ppb, could be present in the core [16, 96]. Table 2 lists the suggested radionuclides for the core including natural abundances, half life’s and heat production rates for pure substances.
The question is then whether or not radionuclides could be present in the inner core as well. Unfortunately the high pressure experiments performed so far can not answer this question as the pressure range of interest has been restricted to pressures relevant for different core-formation scenarios (i.e. well below 50 GPa). However if the presence of radionuclides in the core is associated with the presence of lighter elements, and especially S, the deficiency of lighter elements in the inner core (relative to the outer core) could indicate a partitioning coefficient between liquid/solid core in favour of the outer core. On the other hand, if, as indicated by Parker et al. [103], the presence of K is associated with the presence of Ni in the core, the same argument would lead to a partitioning coefficient in favour for the inner core. A further consideration could be the atomic weight of the radionuclides
18 considered, both U and Th are very dense atoms that could possibly have tendency of gravitational migration toward the centre of the Earth, making them likely constituents of the inner core. Where as in the case of K, being an alkali metal and therefore having a relatively low density, the situation could be the reversed depending on the exact configuration of (elemental or molecular) the core K content.
Finally it needs to be pointed out that the presence of radionuclides in the core, as indicated by high pressure experiments is highly sensitive to the exact formation scenario of the inner core. It should be remembered that most of the experiments has investigated the partitioning coefficient between various Fe-alloys (mainly iron-sulfides) and Silicates. And, as indicated by the experiments, the partitioning coefficient increases in favour of the Fe-alloy with increasing pressure. Hence the results can only be transferred into core values if the Earth’s core formed at chemical equilibrium at depth in an initially homogeneous Earth, or at least in the interior of a relatively large proto-planetary body.
Nuclide Abundancea Half lifeb Heat productionb [%] [Gyr] [W kg−1] 40K 0.0117 1.2511 ± 0.002 2.966 x 10−5 232Th 100 14.01 ± 0.08 2.556 x 10−5 235U 0.7200 0.7038 ± 0.00048 5.749 x 10−4 238U 99.2745 4.4683 ± 0.0024 9.166 x 10−5
Table 2: Possible candidates for core radionuclides. Abundances in present day natural abundance. Heat production assumes the decay chain to have reached steady state, and values are given for pure substances, i.e. 100 % of the specified isotope. a from Nordling & Osterman¨ [101] b from Rybach [117]
6.5 Alternative core models
Even though the consensus in the scientific community is that the core is composed of an Fe-Ni alloy with the addition of some lighter elements, other alternatives have been proposed. In a series of paper by Herndon [60, 61, 62, 64] it has been claimed that composition of the inner core should be nickel-silicide31 rather than an Fe-Ni alloy. In addition it is proposed that the inner core should poses an actinide sub-core of a few km radius at its centre, surrounded by a shell of decay and fission products (see figure 6).
In Herndon’s model most of the inner parts of the Earth are in a highly reduced state and the radius of the inner core is constant (although it is allowed to initially have grown rapidly from a liquid state). Based on the natural occurrence of Ni2Si in oxygen-poor enstatite chondrites it is then concluded that the inner core composition should be that of nickel-silicide. The model makes the assumption that the melting temperature of nickel-silicide at core conditions is much higher than the inner core temperatures. Thus making it possible to have a fixed size inner core throughout the lifetime of the Earth. Also found in highly reduced enstatite chondrites are traces of uranium, U, and thorium, Th, associated with the nickel-silicide. It is therefore argued that U and Th will be present in the inner core as well. As a consequence of the higher densities of U and Th the model postulates the precipitation of a high-density, high-temperature sub core within the inner core, in which the concentration of actinides is high enough for self-sustaining nuclear fission reactions to occur. This is referred to as the georeactor, producing an estimated effect of 3 TW presently. Taking into account also the natural decay of 235U and 238U would increase the effect to a present day value of 3.5 TW.
Over time, as the georeactor operates, a build-up of decay products will occur, halting the fission reactions thus lowering the effect. However, the decay products are lighter, and will over time be removed by gravitationally driven diffusion, and so the effect of the georeactor will again increase. This behaviour could then explain the variability and reversibility of the geomagnetic field, as it is dependent on the power supply in the core. Another evidence of the georeactor comes from 3He/4He ratios released to the oceans at the mid oceanic ridges, which
31 nickel-silicide is a group name of several Ni-Si compounds, e.g. Ni3Si, Ni2Si, NiSi
19 is about eight times greater then in the atmosphere. 4He is naturally produced by the decay of U and Th in the mantle, but so far no mechanism deep within the Earth that can account for a substantial 3He production has been known. Instead the deep-Earth 3He has been assumed to be of primordial origin. Simulations by Herndon has indicated that the observed levels could well be explained by the presence of a georeactor.
Figure 6: The model of the Earths inner core structure as proposed by Herndon. Reprinted from [61]
7 Constraints on the inner core geotherm and heat production
In this chapter I have studied temperature profiles of the inner core using COMSOL multiphysics. Before we go in to this let’s quickly recall our knowledge of the core from previous chapters (for a discussion of each claim, see the preceding chapters). The core of the Earth has an outer radius of 3480 ± 5 km, and comprises a total of about 16.3 % of the Earth’s volume. We can divide the core into two regions, the solid inner core with a radius of 1221 ± 1 km, and the liquid outer core where the Earth’s magnetic field is generated. The core consist mainly of a Fe-Ni alloy mixed with some lighter elements, O, S and Si being the top candidates. It is further recognised that the amount of lighter elements is higher in the outer core than in the inner core. In addition it is also possible that the core contains fractions of long lived radionuclides, in particular 40K has been considered. Since the composition of the inner core is different from the composition of the outer core, it is possible that the melting temperatures of the inner and outer core differs by a few tenth of degrees. The melting temperature of elemental Fe at ICB pressures is estimated to lie within the interval 5400 to 7100 K, which relates to the core alloys via a melting depression of a few hundred up to 2000 K. For the inner core it is believed that k ∈ [60; 80] W m−1 K−1, γ ∈ −1 −5 −1 [1.27; 1.53], Cp ∈ [715; 800] J kg K, α ∈ [1; 2] x 10 K .
7.1 Inner core heat sources
There are only two possible heat sources in the inner core, release of gravitational energy and decay of radionu- clides. However, it has been argued [32] that the release of gravitational energy, due to shrinking of the cooling inner core, is fully converted into compressional (strain) energy in the inner core and therefore does not contribute to the heat equation. In any case, if radionuclides are present in the inner core, the gravitational contribution will be small in comparison and can be neglected.
20 It is most likley that the core has been melted throughout at some stage of the Earth’s evolution. Given the timescales involved in inner core formation and the fact that the outer core is convecting it should then be safe to assume that the inner core formed at chemical equilibrium with the outer core. Therefore, any distribution of radionuclides in the inner core is initially set by the partition coefficients of the radionuclides present. Although a redistribution may have occurred with time, due to gravitational migration. In any case, from the processes involved we should expect any distribution to be a spherically symmetric smooth function (i.e. continuously differentiable).
From a physical point of view we know that the function describing the distribution also has to be non-singular32 inside the inner core. It follows then that any source distribution can be described by a power series, i.e. a polynomial in only positive powers of the radius, r.
X n Q = cnr (7.1) n
Where cn are some constants. Now if radionuclides are present in the core they have to be present in both the inner and outer core, with relative abundances initially set by the partition coefficient, Dsolid/melt, of each speciem. Unless the partition coefficient is heavily biased toward partitioning into the solid inner core we should expect any reasonable function, describing the distribution of radionuclides in the inner core, to be dominated by its zeroth order constant, i.e. by c0 in equation (7.1). The reason being that the inner core comprises only about 4.3 % of the total volume (or equivalently about 5% of the total mass) of the core, and so the levels of radionuclides in the outer core can, to a first order approximation, be considered to be independent of the radius of the inner core.
Even if the initial distribution is not constants in the inner core, the decay of the radionuclides will drive the distribution towards an evenly dispersed (i.e. constant) with time (see figure 7.1 below). Of the radionuclides considered as possible candidates for the inner core, only 40K have a half-life of the order of the estimated age of the inner core33. On the other hand 40K is the top candidate of the suggested core radionulides and therefore also for the inner core. Hence any initial deviation from a constant distribution should have diminished to about half its value over the life time of the inner core.
Figure 7: Time evolution of the inner core distribution of radionuclides due to partitioning coefficient and decay. A = abundance, r = Distance from the Earth’s centre. Solid line indicates distribution due to the behaviour of Dsolid/melt, left panel displays a increasing trend with increasing pressure/density, middle panel displays constant value and right panel an decreasing trend with increasing pressure/density. Dotted line adds the effect of decay of radionuclides in outer core, dashed line adds the effect of decay also in the inner core. Note that the decay tends to level out any non constant behaviour of the distribution. For simplicity the distribution are displayed as linear profiles, although it should be recognized that they in reality are not (e.g. decay is inversely exponential) this does not affect the final conclusion.
So what about a redistribution due to gravitational migration. First of all we need to realize that the gravity of the inner core is low, reaching a maximum value of about 4.36 m s−2 at the ICB (see figure 7.1 below), and so any gravitational redistribution will be slow. In addition, entropy driven diffusion34 will oppose any tendency away from a constant value of the heat source distribution, including gravitational migration.
32I.e. it can not go to infinity 33This is strictly not true since the half life of 235U is about half that of 40K. On the other hand 235U has a natural abundance of only 0.72 % today, in comparison to 99.27 % for 238U. And there is really no reason why this should not also be the case for the inner core. Hence the Half life of 235U is not important in this discussion 34Diffusion driven by a non-zero gradient of the concentration.
21 Figure 8: Gravitational acceleration force inside the inner core.
In light of these arguments we can think about the feasibility of an actinide core as proposed by Herndon (see sect. 6.5 above). In the discussion above it has been assumed that the core initially was completely liquid and well mixed. If this holds true the formation of an actinide core could only have proceeded via a partitioning coefficient heavily biased toward the solid inner core composition. Otherwise the low gravitational field and entropy diffusion would have effectively hindered the formation of an actinide core. Relaxing the well mixed assumption in the innermost parts of the core, does give some possibilities for the accumulation of radionuclides before the formation of the core, however being liquid, entropy driven diffusion will be more effective, thus the actinide core scenario still seems rather unfeasible. In Herndon’s view the inner core need not have been liquid throughout the life time of the Earth, hence the only way for an actinide core to have formed, would have been if it formed elsewhere than in the centre of the Earth and then became accumulated into the inner core upon the formation of the Earth. However this only transfers the problem to another location. Thus the formation of an actinide core seems a highly unlikely scenario.
Several radio nuclides has been discussed as constituents of the core (see table 2 of sect. 6.4). In order to reduce the complexity of the inner core geotherm, I shall in what follows speak in terms of integrated inner core heat production, rather than abundance of a particular element, unless the discussion is dependent on the exact choice of heat sources. Any presented heat production can the be related back to a particular abundance of a particular element via table 3 below, e.g. for an inner core heat production of 0.5 TW we would need a K abundance of about 0.147 % and assuming this abundance to hold throughout the whole core we would have a heat production in the core of about 9.88 TW.
Element Abundance Volumetric Inner core Core heat heat generation heat generation generation [10−9 W m−3] [TW] [TW] K 0.1 % 44.80 0.34076 6.7307 Th 0.1 ppm 33.00 0.25102 4.9581 U 0.1 ppm 122.81 0.93413 18.451
Table 3: Examples of heat production in the inner core and the whole core for a given elemental abun- dance. ppm = 10−6
7.2 Inner core geotherm
To estimate the geotherm of the inner core for various distributions of radionuclides I have used steady state models, assuming that the age of the inner core is great enough for this to be a valid assumption. Using a steady state model means that we can find an analytical solution for the heat equation (see eq.(B.21) of appendix B.2).
1 1 Z Z A T (r) = Qr2dr − Qrdr − + B (7.2) k r kr
22 Where Q is heat sources and sinks per volume35, [W m−3], k is the thermal conductivity, [W m−1 K−1], and A and B are constants determined by the boundary conditions. Note that in the derivation it has been assumed that k is constant. If not so we have to use the form Z 1 Z A T (r) = B − Qr2dr dr − (7.3) kr2 kr
Given the boundary conditions of the inner core we find that A = 0, as dT/dr|r=0 = 0, and that B is related to the temperature at some radius36 of the inner core37. Hence, we see that the temperature profile is linearly dependent on the temperature at the reference radius. Changing the reference temperature by an amount ∆T will change the whole temperature profile by an amount ∆T . This means that we can work with relative temperatures rather than absolute, i.e. we do not need to know the temperature at any point of the core to find a solution38. In what follows I shall denote the relative temperature by ∆T , rather than T , to indicate that the presented geotherms can be scaled to any temperature of the chosen reference radius by simple addition. For convenience I shall always use the ICB as the reference radius, i.e. ∆Ticb = 0 in what follows.
It is very instructive to study the behaviour of the geotherm and its gradient, dT/dr, due to the source distribution and thermal conductivity under steady state conditions. Therefore we shall do so before I present any geotherms for specific combinations of Q and k, with the aim that the discussion of any presented geotherm can be related this study. It needs to be stressed that this study assumes steady state conditions, hence conclusions drawn are not necessarily valid under non-steady state conditions
First we shall consider the quantity Q itself. Let’s start with the sign. According to our definition above (see footnote 35) a positive sign means that heat is being produced, a source, and a negative sign means that heat are being consumed, a sink. Now, inside a simply connected volume39 the only way heat can be consumed is by conversion into some other energy form or by a decrease of the temperature. However, in steady state the temperature can not change by definition, and if heat is directly transformed into some other energy form it must not enter into the heat equation. Thus we can not have any internal sinks inside the inner core.
To investigate the geotherms behaviour due to Q we shall start by assuming k to be constant. Consider a sphere of radius r0 with no internal heat sources. This means that the temperature profile has to be isothermal (i.e. constant). Now add an external layer of thickness ∆r, containing some heat source distribution Q, to the surface of the sphere. Since integrals are linear operators, we can then split eq.(7.2) into two integrals over the continuously connected regions, 0 to r0 and r0 to r0 + ∆r. The inner region (i.e. the integral over 0 to r0) will not contain any heat sources, hence the temperature will still be isothermal here. In the outer region, we will have a thermal gradient < 0. Let’s add one more layer to our sphere, this time with no internal heat sources. From energy conservation we know that the heat produced inside of this layer has to diffuse outwards through the layer. Denote the integrated heat produced inside of r0 + ∆r by P (r0 + ∆r), we then have that the heat flow per unit surface, q, at a radius r > r0 + ∆r is P (r0 + ∆r) q(r) = 4πr2 and so from Fourier’s law in spherical coordinates we have that
0 dT q P (r + ∆r) = − = − 2 (7.4) dr r>r0+∆r k 4πkr
35I will refer to this term as heat sources in what follows, equating heat sinks as negative heat sources 36referred to as reference radius here after 37This can easily be realized by considering T (r = r0) = B + C(r’), where C(r’) becomes a numerical value that might be 6= 0 depending on the distribution of heat sources. 38Note that the above stated not necessarily holds true if k is a function of temperature. On the other hand, if that is the case, the given solutions (eq.(7.2) and eq.(7.3)) will not be applicable anymore. We shall therefore in what follows neglect any temperature dependence of k 39Simply connected means that we can continuously shrink the volume to a single point, i.e. no ”holes” are allowed inside the volume. Thus we can consider the inner core to be continuously connected.
23 To summarise we have that
0 0 0 ≤ r ≤ r If Q(r) = 6= 0 r0 ≤ r ≥ r0 + ∆r 0 ≥ r0 + ∆r ≤ r 0 0 ≤ r ≤ r0 dT Then = < 0 r ≥ r0 dr ∝ −r−2 ≥ r0 + ∆r ≤ r
Thus the gradient of the geotherm at a given radius, will not depend on heat sources outside of that radius. The more heat sources inside, the steeper gradient. Note that if Q is constant inside of some radius r0 then q(r < r0) is proportional to r, hence the gradient of the geotherm will then be proportional to −r inside of r0. To relate the gradient to the geotherm we need to choose some reference radius where we keep the temperature fix. By our convention stated above we choose then the ICB. We see then that concentrating the heat sources toward the centre of the core will increase the central temperatures, whilst moving them further toward the ICB will flatten the geotherm.
We shall note another important aspect that follows from the linear behaviour of an integral with respect to its integrand. Given a complex source distribution that can be described as a sum of several simple source distribu- tions, i.e. Q = Q1 + Q2 + Q3 + ... we can always split the integrals of eq.(7.2) into several integrals. Define the operator I(x) as 1 1 Z Z I(x) = xr2dr − xrdr (7.5) k r We can then write eq.(7.2) as X T (r) = B + I(Q) = B + I (Qn) (7.6) n and so the geotherm can be expressed as the sum of several simpler geotherms. Since we have argued for that any source distribution of the inner core has to be continuously differentiable and that it can not posses any singularities, we already know that we can express any inner core source distribution in the desired way. In addition we have argued that any source distribution function will be dominated by its zeroth term coefficient. Using eq.(7.1) above we have that
X n T (r) = B + I(c0) + I (cnr ) ≈ B + I(c0) (7.7) n>0
And so the steady state gradient of the inner core geotherm in the presence of internal heat sources are to a first order approximation proportional to −r, i.e. the geotherm is proportional to −r2.
We then move on to the response of the inner core geotherm to k, given some arbitrary source distribution. Again we denote the integrated heat sources inside of some radius r by P (r). Even if k is not a constant function of the radius of the sphere we shall assume it to spherically symmetric. Inspecting the Fourier’s law in spherical coordinates, we see that for a given source distribution, the thermal gradient at any radius is proportional to k−1 at that radius. Thus, assume that we change k at some radius, r0 of the sphere, this will only affect the geotherm gradient locally, or equivalently geotherm inside of r0 if the reference temperature is outside of r0 and vice versa. Hence, whilst the geotherm at r is affected by all heat sources inside of r it is only affected by k at r.
To conclude, since we have set the reference radius to the ICB, we need to know the source distribution over the whole inner core. But this we have argued for to be constant to a first order approximation. If we are only interested in the temperature at some specific radius inside the core we only need to know k in between this radius and the ICB.
Figure 9 displays various relative geotherms for the inner core assuming evenly distributed heat sources and a constant heat transfer coefficient. Note that the shape of the profiles are proportional −r2 as discussed above, likewise we see the k−1 behaviour of the geotherms. Now we could expect k to increase somewhat with pressure and temperature, i.e. inwards. Hence from our study above, the given profiles should be considered to be upper
24 limits. If we assume the inner core contains 40K as only radionuclide present, with an abundance equal to the maximum allowed on cosmochemical grounds (0.142 wt%) the present inner core heat productions amounts to about 0.5 TW. Using the upper estimate of the thermal conductivity (80 W m−1 K−1) we find a maximum tem- perature difference over the inner core of about 204 K, where as a thermal conductivity of 60 W m−1 K−1 would yield a maximum temperature difference of about 272 K.
(a) (b)
(c) (d)
Figure 9: Relative geotherms of the inner core for evenly distributed heat sources with an integrated effect of P , and constant heat transfer coefficient, k. Figure 9(a) displays profiles for k = 80 W m−1 K−1 with Q varying in the interval [0 2] TW , figure 9(b) displays profiles for Q = 0.5 TW with k varying in the interval [20 100] W m−1 K−1, figure 9(c) shows the central relative temperatures for a number of models in P − k space, and figure 9(d) shows the temperatures at a radius of 500 km in P − k space. White solid lines superimposed on the surface indicate isotherms for every 100 K for the upper two figures, for every 150 K for the lower left figure and for every 50 K for the lower right figure.
The above used term evenly distributed heat sources refers to a constant amount of radionuclides per volume, i.e. the same amount of heat is generated in every cm3 of the entire core. However, what is commonly used in the literature is weight percentage, i.e. evenly distributed in that sense should then be constant per mass, i.e. the same amount of heat is generated in every kg of the entire core. We know that the density of the inner core increases inwards and so defining a constant weight percentage will mean an increased number of radionuclides per unit volume toward the centre of the core. And from our study above we know that this in turn implies a higher central temperature. On the other hand, as the volume of a spherical shell is proportional to the radius squared, this effect will be small unless the density changes drastically over the radius. Anyway, it can be instructive to investigate the difference between the two scenarios. Assuming the PREM model (see table 4 of appendix A) the density over the inner core increases by less than 2.6 percent. Using a heat transfer coefficient of 80 W m−1 K−1 and the PREM model density distribution we find that the temperature increase at the centre (where the effect is largest, see figure 10) is about 1.54 K or 0.03 % (assuming a temperature at the ICB of 5500 K) for the constant weight percentage scenario with an integrated inner core heat production of 0.5 TW. Increasing the heat production to 2 TW yields an increase at the centre of 6.16 K or 0.1 %. Given that the magnitude of the temperature change between the two models is so small, we conclude that our use of constant heat production per unit volume is acceptable.
25 (a) (b)
Figure 10: Temperature difference over the inner core between models assuming constant internal heat generation per unit volume and unit mass for k = 80 W m−1 K−1. Figure 10(a) shows the difference in degrees K whilst figure 10(b) shows the difference in percentage assuming the ICB temperature to be fixed at 5500 K. Note that the difference goes to zero just before the ICB, this means that both models has the same thermal gradient at the ICB which should be the case since both models generate an equal amount of heat inside of the ICB. White solid lines super imposed onto surface indicate isotherms for every 1 K in figure 10(a) and for every 0.02 % in figure 10(b)
In sect.6.5 I reviewed an alternative model of the core proposed by Herndon [60, 62], in which a georeactor of a few km radius would be operating at the centre of the core. I have shown above this would imply a very high temperature at the centre of the core, thus it is now time to investigate the thermal profiles of Herndon’s model.
Figure 11 displays some geotherms over the inner core for the Herndon georeactor model, as well as temperatures at the centre and at a radius of 500 km. The models assumes that all heat is being evenly produced inside a radius of 12 km of the inner core. Note how the temperature profiles outside of the georeactor drops off like r−2 as proven above and that the central temperatures are proportional to P/k. Clearly the resulting central temperatures are ridiculous. E.g. for the proposed georeactor output of 3.5 TW the central temperatures reach as high as 423 500 K for a thermal coefficient of 80 W m−1 K−1. Even if we increase the thermal coefficient to 300 W m−1 K−1 we still reaches a value of 112 930 K. Clearly something must be wrong.
We do not really know the thermal coefficient of nickel-silicide at core pressures and temperatures, but at room temperature and pressure it is about40 1/8 to 1/4 of that of Fe [98]. Let’s estimate the magnitude of the thermal conductivity needed to reach some reasonable central temperatures. From the known behaviour of the geotherm we find that the relative central temperatures are given by a function on the form ∆Tc(P, k) = aP/k. Using the −6 results of my modelling we find that a = 9.68 x 10 , hence assuming P = 3.5 TW and ∆Tc = 10 000 K, we find that we need to have a constant value of k of 3388 W m−1 K−1 through out the entire inner core. This is about 42 times the estimated thermal conductivity of the Fe at core pressures. It seems very unlikely that this should be the case. Now at high temperatures we should expect the thermal conductivity to increase, especially at the extreme temperatures found in Herndon’s model. But from our discussion above it should be clear that if we tried to lower the thermal conductivity to reasonable values near the ICB, the central temperatures would increase even further. To compensate for this we would then have to increase the thermal conductivity in the inner parts to values even higher than 3388 W m−1 K−1. Thus increasing the thermal conductivity in the inner core does not solve the problem.
At the temperatures found we should expect the centre of the inner core to be liquid and convective. Now, we did argue in section 7.1 that the gravitation is low in the inner core (see figure 7.1), and this will especially be true at the centre of the Earth. Therefore buoyancy forces would be small. Still, the tremendous temperatures would lead to convection in the liquid region. The convection has to be relatively fast and extend out to a radius of at least a few tenths (or even a few hundreds depending on model, see figure 11), of km from the core. Otherwise the temperature of the inner regions will still reach very high temperatures41. But this implies that the inner most
40As nickel-silicide is a group name for several Ni-Si compounds the exact value of the thermal conductivity of the inner core will be set by the exact composition. E.g. k(Ni3Si) = 18.2 W/m/K, k(NiSi) = 10.3 W/m/K 41This is naturally dependent on the heat capacity of nickel-silicide (45 J K−1 mol−1 at room temperature [1]), but as this is about the same
26 region of the core should be low viscous to allowed for an effective enough convection. This is on the other hand compatible with high temperatures. The big issue is then how the georeactor could have survived over time, and not have become distributed out over the convective region?
(a) (b)
(c) (d)
Figure 11: Relative geotherms of the inner core assuming Herndon’s georeactor model. figure11(a) shows profile for various effects of the georeactor assuming k = 80 W m−1 K−1, figure 11(b) shows various models for different values of k, assuming a constant georeactor effect of 3.5 TW, figure 11(c) shows the central relative temperatures for a range of models in P − k space, and figure 11(d) shows temperatures at a radius of 500 km in P − k space. Isotherms are indicated by white solid lines at every 100 000 K for the upper two figures, extra isotherms has been added for the relative temperatures 500, 1000, 2000, 5000, and 10 000 K, for the lower left figure isotherms are displayed at an interval of 100 000 K with extra lines at 50 000 and 25 000 K, and for the lower right figure isotherms are displayed at an interval of 1000 K with extra lines at 500 and 250 K.
Unfortunately the central convection that must arise in Herndon’s model makes it impossible for us to find the central temperature from our steady state solid models, we can only conclude that they will be high. But what about the temperatures further out. Since we know the source distribution over the entire core, we can find geotherms for the outer part where temperatures are low enough for us to give a reasonable estimate of the thermal conductivity. It seems reasonable to believe that if the high pressure thermal conductivity of nickel-silicide is larger that of Fe, then it is at maximum a few times (say 3) larger. Let us then assume that we have a temperature span of a few thousands of degrees for ∆T with only negligible changes in the thermal conductivity, at what distance from the would we then find a relative temperature of say 2000 K, or maybe 5000 K?
Figure 12 displays Isothermal surfaces in the P − k space for T = 2000 K and T = 5000 K. We can see that even if we increase the thermal conductivity to 300 W m−1 K−1 and lower the georeactor output to 0.5 TW, the temperature in the core would be 5000 K at a radius of 26.8 km and 2000 K at a radius of 64.7 km , where as in Herndon’s scenario (for k = W m−1 K−1) we would have a temperature of 5000 K at about 162.3 km radius and 2000 K at about 336.7 km radius. So at what temperatures would then nickel-silicide melt in the inner core? At ambient pressures Ni2Si has about the same melting temperature [60] as the Fe-alloys that has been suggested for the core composition. Given that Ni and Fe have relatively similar properties and that Si also is a likely constituent
as for Fe, although this is not an assurance that so is the case also at high pressures and temperatures I assume this to be the case.
27 of the outer core, one should not expect the melting temperatures of nickel-silicide to differ much from that of the Fe-alloy of the outer core42, hence an inner core of nickel-silicide should be melted in regions where the temperature was more than 2000 K above Ticb. But this would then give an central liquid region of radius 380 km today. It seems very unlikely that this would still be undetected. Even if we assume that the inner core can survive in solid form to temperatures of 5000 K above Ticb the liquid region would still be very large and likely to be detected by seismics. Now even a georeactor output of 0.5 TW only, would still produce relatively large liquid regions.
(a) (b)
Figure 12: Isothermal surfaces in the P −k space for the georeactor model of Herndon with k = 80 W m−1 K−1. Left panel shows iso-surfaces at T = 2000 K and right panel at T = 5000 K. White solid lines indicates the distance to the centre for every hundred km, extra lines has been added for the distances 50 and 25 km.
Having said this two things needs to be realised. First of all the central parts of the inner core are not easily accessed. The only reliable data with sensitivity to these regions are absolute travel times from antipodal distances, which ammounts to about 3000 unevenly spread measurments. Secondly the central parts of the inner core does display some peculiarities. In a study by Ishii and Dziewonski [66], using the availible data set, indications was found of seismically distinct region at the center of the inner core, with a radius of 300 km. ”although higher quality data are required to draw a firm conclusion.” Note however, that this is a solid region, although if one whished to support the Herndon model one could interpret this as evidence of a different chemistry, or perhaps the remnants of a melt that solidified inside the inner core.
Turning our attention back to the geotherms of the mainstream models we wish to find ways to constrain these. As in the case of Herndon’s models we shall try to do so again by looking at the melting temperature of the inner core material.
7.3 Constraining the inner core temperature profile
From eq.(7.3) above we can compute the relative temperature profile of the inner core, however we wish to somehow constrain the possible set of temperature profiles of the inner core. If we assume that the inner core is solid throughout we can use the melting temperature in the inner core to put an upper limit on the possible temperature profiles. At the high temperatures and pressures present of the core we can assume Debye theory to hold43. We can then estimate the melting temperature over the inner core from the Lindemann law of melting (where we for simplicity have assumed q = 1, see eq.(C.45) of appendix C.6)
ρ 2/3 ρ T (r) = T ICB exp 2γ 1 − ICB (7.8) m m,ICB ρ(r) th,ICB ρ(r)
42It is known from high pressure experiments [47] that at least up to 60 GPa the melting temperature of elemental Ni is lower than that of elemental Fe. 43The molar volume of hcp-Fe is expected to be less than 4.8 cm3/mol at inner core pressures [10], which implies that debye theory holds, see appendix C.3
28 The subscript ICB indicates quantities measured at the ICB, and γth is the thermodynamical gruneisen¨ parameter. It should be recognized that the Lindemann law of melting is derived for elemental substances and should be applicable for Fe at core pressures. The inner core is not pure Fe but rather a Fe-Ni mixture including some lighter elements, and this will affect the melting temperature toward lower temperatures. Still experiments has shown that the Lindemann law is applicable if the molecule complexity is low [105], and so as the core predominately consists of Fe with Ni and some lighter elements as minor constituents, we will assume that the Lindemann law is valid.
As we do not know TICB with acceptable accuracy we would like to work with relative temperatures as we did with the temperature profiles. However note from eq.(7.8) that we can not express the Lindemann law in terms of relative temperatures. We could express it as a ratio, but this is on the other hand not possible with the temperature distribution (see eq.(7.3)). In addition we have that the melting law is dependent on the gruneisen¨ parameter. Let us therefore start by examining the behaviour of the Lindemann law. For the density profile we shall use the
PREM model (see sect.3.3 above). To generalise we form the dimensionless melting temperature ratio QTm
Tm(r) Qtm (r) = (7.9) Tm,ICB
As can be seen in figure 13, for a reasonable range of γth,ICB values, QTm (r) can be assumed to display a linear behaviour with respect to γth,ICB with slope given by figure 14. Also note that the curves could be relatively well 44 2 fitted by a function QTm = a − br , where a and b are linear functions of γth,ICB.
(a) (b)
Figure 13: Melting temperature ratios, QTm (r) = Tm(r)/Tm,ICB , for various values of γth, left panel shows profiles over the inner core for γth = 1.2 to 1.8 in steps of 0.1, right panel shows the ratio between melting temperatures at the ICB and at the centre of the Earth, as a function of γth. The model assumes PREM densities.
Returning back to the inner core geotherm we see that if the geotherm at some radius, r0, coincides with the 0 0 melting temperature, i.e. TIC (r ) = Tm(r ), the temperature gradient locally has to be proportional to the radial coordinate to the power of at least 1, i.e.
0 0 0 dTIC (r ) n if TIC (r ) = Tm(r ) ⇒ ∝ r ; n ≥ 1 dr r=r0±
where is some small number. The reason being that dTm/dr ∝ r, and so if the condition above is not fulfilled 0 0 TIC (r + ∆r) > TIC (r + ∆r) and the inner core will re-melt. Note that the criterion covers nearby regions on both sides of r0. As argued above, the likely inner core distribution of radionuclides should be constant, which was shown to imply a thermal gradient proportional to r, therefore fulfilling our criterion. Consider then if the geotherm coincides with the melting temperature at some radius. We then have two possibilities, since both the
44 2 This is so since we have used the PREM density, which can be described by an analytical function on the form ρ(r) = ρc − dr (see 2 2 table 5) QTm = a − br of appendix A). Thus QTm = a − br is simply the first terms in a Taylor series expansion. Using figure 13 and −8 2 −4 14 one can find the fit to be QTm = 0.9806 + 0.0525γth,ICB − 10 (−1.3035 + 3.5273γth,ICB )r , which is accurate to within 10 for γth,ICB ∈ [1.2 1.8]. N.B. r is in units of km.
29 geotherm gradient and the melting temperature gradient are proportional to r the temperatures either can coincide at one radius only or at all radius. If they coincides at one radius only this can only be at the ICB, otherwise we can not have a solid inner core, or alternatively the whole core has to be solid. Since we are interested in constraining the possible distributions of heat sources in the inner core, we are interested in the case where the geotherm coincides with the melting temperature at the centre. Recognizing this, we can then use figure 13(b) and 9(c) to constrain the possible total heat generation of the inner core.
Figure 14: dQTm /dγth as a function of radius in the region γth ∈ [1.21.8]
(a) (b)
(c) (d)
Figure 15: Maximum inner core heat production. ∆T is the difference between the geotherm and the inner core melting temperature at the ICB. Upper left figure displays maximum inner core heat production assuming evenly distributed heat sources, TICB = 5500 K, and γth,ICB = 1.5. Upper right figure displays the change in % if we lower γth,ICB by 0.1. Lower left figure displays the percentage change if we lower TICB by 500 K, and lower right figure displays the percentage change if we lower TICB by 500 K and γth,ICB by 0.1.
Figure 15(a) displays the maximum inner core heat production as constrained by the Lindemann melting law, −1 −1 assuming evenly distributed heat sources, TICB = 5500 K, and γth,ICB = 1.5. We see that for k = 80 W m K , the pressure induced freezing model has an upper limit of about 0.80 TW, or about 105.3 W m−3, whilst allowing
30 for difference of 100 K between the geotherm and the inner core melting temperature at the ICB, increases this to about 0.98 TW. From the other figures of figure 15 we also see that lowering TICB by 500 K, or γth,ICB by 0.1 units, lowers the possible inner heat production by about 9.5 % for the pressure induce freezing model, whilst the effect is about 7 % if the difference between the geotherm and the inner core melting temperature at the ICB is 100 K.
However, we need to realize that the inner core heat productions presented in figure 15 are primordial in the sense that they lack the time aspect of inner core formation. Only if the inner core were very young, the current day heat generation could correspond to that constrained by the melting temperature, other wise more energy would have been produced at earlier times, increasing the temperature above the melting temperature. As we discussed in sect. 7.1 above we need to compensate for the decay of the radionuclides present. We also came to the conclusion that the effect of decay would be to smooth the source distribution toward an evenly distributed (see figure 7.1). If we integrate equation (7.2) above we find that the temperature difference between the ICB and the centre at any time t is equal to Qe−λt ∆T = R2 (t) (7.10) 6k ICB Where we have taken into account the effect of decay of the radio nuclides via the decay constant, λ, and the time dependence of the radius of the inner core. The melting temperature can within a reasonable interval of the gruneisen¨ parameter be approximated by (see footnote 44)