Crystal Field Theory; Trace Element Behavior During Melting

1. Crystal Field Theory;Trace Element Behavior During Melting Lecture 26

2. Crystal Field Theory

3. Electron Orbits and Shells

4. Electron Shells

5. First Transition Metals +2 +3 +4 +5 +6 (+2) +3 +4 +5 +3 +4 +2 +4 +2 +3 +1 +2 +3 +2 +2 +2 In transition metals, filling the3 d shell follows filling the 4s shell. In transition metal ions the 4s electrons will have been removed, leaving the 3d electrons as outermost (except Sc3+, where the single 3d electron is not present).

6. Crystal Field Theory The directional d electrons are outermost in transition metal ions. Each of the 5 d orbitals can accept 2 electrons, one with spin up, the other with spin down. Normally, additional electrons are places in empty orbits before a second electron is placed in an occupied orbit. Because of the orientation of the orbits and the coordinating negatively charge ligands (usually O or S), the next electron may go into an occupied shell rather than an unoccupied one (this is known as degeneracy). Either choice has an energy cost: so there is a trade off in energy costs between placing 2 electrons in 1 orbit and placing them in geometrically unfavorable orbits for a given lattice site geometry. The sites available to transition metals in both minerals and melts are surrounded by either 4 (tetrahedral) or 6 (octahedral) ligands. The site preference energy, is the difference between the Crystal Field Stabilization Energy(CFSE) of octahedral and tetrahedral sites.

7. Crystal Field Theory ligands Energy lowered by avoiding x2y2 and z2 orbitals. Energy lowered by avoiding xy,xz, and yx orbitals. Transition metal partitioning depends on the distribution of octahedral and tetrahedral sites in the mineral and the liquid (and well as charge and size) Thus partitioning is more complex.

8. Color & Transition Metals Electrons can be excited into a higher energy orbit by absorption of a photon. They then emit the photon when the jump back to a lower energy orbit. It turns out the energy difference between the crystal field-influenced orbitals corresponds to photons of visible light. Consequently, even minor amounts of transition metals in minerals can produce bright colors.

9. Ni partitioning Crystal Field theory helps us understand transition metal partitioning. For example, in a pure Mg2SiO4–Ni2SiO4 system, Ni partitions preferentially into the melt. In natural silicate melts, Ni partitions strongly into the olivine. The reason for this is that Ni prefers octahedral sites and in the pure olivine system, only octahedral sites exist in both the melt and the solid. But basaltic melts have both tetrahedral and octahedral sites, while olivine has only octahedral sites (available to Ni). The greater availability of octahedral sites in the solid provides an added ‘incentive’ for Ni to partition into olivine relative to basaltic liquid.

10. Trace Element Behavior During Melting

11. Thinking About Melting • Melting in the Earth is essentially always incomplete (impacts would be the exception). Thus we always deal with partial melting. We can imagine two simple models: • In batch melting, a finite amount of melt, for example 5 or 10%, is produced and equilibrates completely with the solid residue. • In fractional melting, the melt is extracted as soon as it is created, and only an infinitesimal increment of melt will be in equilibrium with the solid residue at any given time. • Choosing between alternative models of partial melting requires a knowledge of how melting and melt extraction actually occurs. Unfortunately, melting and melt extraction in the Earth remain poorly understood because we are unable to observed them directly.

12. Batch Melting Model • From mass balance we may write: • where i is the element of interest, C° is the original concentration in the solid phase (and the concentration in the whole system), Cl is the concentration in the liquid, Cs is the concentration remaining in the solid and F is the melt fraction (i.e., mass of melt/mass of system). Since D = Cs/Cl, and rearranging: • or • We can understand this equation by thinking about 2 end-member possibilities. • First, where D ≈ 0 and D<<F (the case of a highly incompatible element), the Cl/Co= 1/F. • Second where F ≈ 0, then Cl/Co = 1/D. Thus the maximum enrichment of an incompatible element (or maximum depletion of a compatible one) in the melt is the inverse of the partition coefficient.

13. Fractional Melting Model • Now consider the case where only an infinitesimal amount of melt equilibrates with the solid residue before being removed. If is is the mass of element i in the solid phase being melted, S the mass of the solid phase, L the mass of the liquid phase, il the mass of i in the liquid, S° the original mass of the solid (and mass of the system), andi° the original mass of i in the solid (and system), then: • and • Integrating and letting F=L/S˚: • The maximum enrichment is still 1/D, but our 1/F relation does not hold.

14. Aggregate and Continuous Melts Infinitesimal melt fractions will not erupt alone, but will mix before erupting. The enrichment in an aggregate of melts produced corresponding to a total melt fraction F will be: Because of surface free energy, there will always be a little bit of melt left in the solid (like the water left in sand) occupying pore space ϕ. If we imagine that a fraction of melt ϕ is always present and all excess melt is extracted, the enrichment in this continuous melt will be And its aggregate will be:

15. Aggregate and Continuous Melts The f in book should be ϕ One take-away: trace element concentrations in aggregate fractional or continuous melts vary as a function of F in much the same way as do batch melts.

16. Multiphase Solids These equations are relevant when the solid undergoing melting is homogenous. If it consists of several phases, we need to replace D with a bulk distribution coefficient, which is simply the weighted mean of the individual solid/liquid partition coefficients: Minerals do not enter the melt in the proportion in which they exist in a rock. Thus a realistic melting model requires that we modify our equations to account for this. We need to define a new parameter P, which is simply the average individual partition coefficients weighted according to the proportions in which the minerals enter the melt: The equations for batch and fractional melting in this non-modal melting case become: and

17. Summarizing • The concentration of a trace element in a melt is a function of: • (1) the concentration of the element in the original solid (i.e., C˚) • (2)the nature of the solid phases (i.e., minerals) present in the system before and during melting because that governs the: • (3) the partition coefficient of the element • (4) the extent of melting (i.e., F) • (5) the manner of melting (e.g., fractional vs. batch) • The first of these is generally only weakly constrained and often what we want to know. We can calculate those values is we know everything else. • Xenoliths provide some evidence of the mineral assemblage and partition coefficients can be determined experimentally (and placed in a theoretical context). • Can we constrain the last two from thermodynamics?

18. Why the Mantle Melts The main cause of melting in the mantle is decompression. Rising mantle material will cool along an adiabat: The solidus temperature also decreases according to the Clapeyron Equation: The slope of the former is steeper than the latter, so provided the mantle is initially hot enough, it will eventually cross the solidus and melt. As it continues to rise, it cools more rapidly as energy is being consumed by ∆Hm. Nevertheless, melt fraction increases as it continues to rise. We can derive the following for isoentropicmelting:

19. How much does the mantle melt? Bottom line: if a rising packet of mantle intersects the solidus at 100 km depth (≈ 3 GPa), upon reaching a depth of 30 km (~1 GPa) that packet would have undergone about 24% melting. (In reality, things are more complex.) Our equation was: So the extent of melting depends on T and thermodynamic properties and (∂F/∂T)P. Experiments provide constraints on this. The coefficient of thermal expansion, α, is about 3 x10–5 K–1, V is about 0.3175 cm3/g (= 0.3175 JMPa‑1g–1), and Cp is about 1.15 JK–1g-1. Thus the adiabatic gradient at 1673 K is about 12 K/GPa. (∂T/∂F)Pis about = 350.88 K. ∆Vm is about 0.0434 cm3/g (0.0434 JMPa-1g-1) and ∆Sm is about 0.362 JK-1 g-1, which corresponds to a slope of the solidus of about 120 K/GPa. From this we calculate a value for (∂F/∂P)S of about -0.12 GPa–1, or about -1.2%/kbar (it is negative because the extent of melt increases as pressure decreases).

20. How is Melt Extracted? • Batch or Fractional? • Again, our physical chemistry can provide some clues. • Whether the melting process approximates the batch (equilibrium) model or the fractional model depends upon the permeability of the source region: whether the melt is confined to pockets or is interconnected. This depends upon the crystal–liquid interfacial energy. • When melt is present, there will also be an interfacial energy between the grains and the melt, σsm. If θis the angle formed by a melt pocket at a grain triple junction, the balance of forces may be described as: • Rearranging:

21. How is Melt Extracted? • We had: • The lower the solid–liquid interfacial energy, the more extensively melt will interconnect (the more extensively grains will “wet”) and the more readily melt will flow. • Depending on the value of θ, the melt can distribute itself in number of ways. • The first case occurs when the solid–solid interfacial energy is twice that of the solid–melt interfacial energy; if so, then cosθ/2 = 1 and θ≈ 0. In this case, melt will form a thin film that coats all grain boundaries. • The second case is where the solid–solid interfacial energy is more than 1.73 times but less than twice that of the solid–melt interfacial energy (2σsl> sss > 1.73σsm), which corresponds to 0 < θ< 60˚. In this case, the melt will form interconnected channels along grain triple junctions. • The third case corresponds to σss< 1.73σsland θ> 60° and melt forms isolated pockets at junctions between 4 or more grains and but is absent elsewhere. These pockets become connected only at relatively high melt fractions (several percent). • Permeability will be high for the first two cases where melt forms films or channels that allow melt to flow, but low for the last case of isolated melt pockets.

22. Melt Connectivity Experiments in which basaltic melt is allowed to come to textural equilibrium with olivine indicate that θis characteristically between 25° and 50°. Since the upper mantle consists of over 60% olivine, however, it is likely that melt forms an interconnected network. This means melt is likely to be extracted fairly quickly after it is created, and that very small melt fractions, perhaps as low as 0.1%, can segregate from the mantle. Thus the fractional melting model may more closely approximate melting in the mantle. Experiments with granitic-like melts indicates a much lower dihedral angle, 12-18˚, but this still means interconnected channels. But because the viscosity granitic magmas is four orders of magnitude greater than that of basalt, segregation of granitic melt requires a higher melt. Nevertheless, melt fractions as low as 10% should segregate on time-scales of 105yrs.

23. Melting Beneath Mid-Ocean Ridges The most common kind of melting on Earth is at mid-ocean ridges. Once the parcel of mantle has risen above the initial melting depth, melt from below will continually stream through it and will react with the solid in an attempt to reach equilibrium with it. Despite the complexity of the melting process, the batch melting equation gives a reasonably good approximation of incompatible element concentrations in the melt as a function of the average degree of melting. Beneath mid-ocean ridges, the average degree of melting will be less than the maximum degree of melting, because different parcels of mantle follow different paths. Only mantle directly beneath the ridge is able to rise the maximum amount, and hence melt the maximum amount. In the simple case, the average extent of melting is one half the maximum extent. Other ratios are possible for other models of mantle flow.