Distribution Coefficients, Nucleation and Diffusion
CSIR NET Earth Sciences geochemistry: partition and bulk distribution coefficients, batch and Rayleigh models, nucleation, Fick's laws, closure temperature.
- 18 min
On this page · 13 sections
Prerequisites
- Compatible and incompatible elements, LILE and HFSE: Atomic Structure, the Periodic Table and Ionic Substitution.
- Henry’s law, activity and the Gibbs energy of a phase: Phase Rule and Thermodynamics of Solutions and Fluids.
- The Arrhenius equation from chemical kinetics: Rate Laws and Temperature Dependence.
The Hook: A Lava Flow With Two Textures
Distribution Coefficients
Nernst partition coefficient
At equilibrium between a mineral and a melt (or fluid), a trace element \( i \) divides according to
$$\boxed{D_i = \frac{C_i^{\text{mineral}}}{C_i^{\text{melt}}}}$$(concentrations by weight). Because trace elements obey Henry’s law at low concentration, \( D_i \) is independent of \( C_i \) over a wide range, but it depends on temperature, pressure, melt composition (polymerisation), oxygen fugacity (for multivalent elements such as Eu, V, Ce) and the mineral’s composition.
Exchange coefficient. For a major-element pair the ratio of ratios is more stable than either \( D \):
$$K_D = \frac{(X_{\text{Fe}}/X_{\text{Mg}})^{\text{olivine}}}{(X_{\text{Fe}}/X_{\text{Mg}})^{\text{melt}}} \approx 0.30$$for olivine–basaltic melt (Roeder and Emslie, 1970). It tests whether an olivine crystal was in equilibrium with the lava that carries it, and lets you estimate the forsterite content of the olivine that a given melt would crystallise.
Bulk distribution coefficient
A rock or residue is a mixture of minerals with weight fractions \( x_j \) (summing to 1):
$$\boxed{D_i^{\text{bulk}} = \sum_j x_j D_i^{j}}$$| Element | Olivine | Orthopyroxene | Clinopyroxene | Garnet | Plagioclase | Behaviour in mantle melting |
|---|---|---|---|---|---|---|
| Ni | high (≫ 1) | > 1 | > 1 | low | very low | strongly compatible |
| Cr | about 1 | > 1 | > 1 | > 1 | very low | compatible (spinel, pyroxene) |
| Rb, Ba, Th, Nb | ≪ 1 | ≪ 1 | ≪ 1 | ≪ 1 | Ba, Rb small but larger in plagioclase | highly incompatible |
| Sr | ≪ 1 | ≪ 1 | < 1 | ≪ 1 | > 1 | incompatible in mantle, compatible where plagioclase crystallises |
| HREE (Yb, Lu) | ≪ 1 | < 1 | < 1 | ≫ 1 | ≪ 1 | compatible only if garnet is present |
Values are qualitative: published \( D \) values vary with \( T \), \( P \) and melt composition. Use the values a question gives.
What controls \( D \): the lattice-strain model
For a given site, \( D \) plotted against ionic radius for ions of one charge forms a near-parabola (an Onuma diagram): the peak sits at the ideal radius \( r_0 \) of the site, and \( D \) falls on either side because a misfit ion strains the lattice. Blundy and Wood (1994) expressed this as an elastic strain energy that depends on \( (r_i - r_0)^2 \) and on the site’s Young’s modulus: a stiff site (high modulus, high charge) gives a narrow parabola, so it discriminates sharply between ions. This is the quantitative form of Goldschmidt’s size rule.
Melting and Crystallisation Models
Notation: \( C_0 \) = concentration in the original source (or parent melt), \( C_L \) = in the liquid, \( F \) = melt fraction (for melting) or fraction of liquid remaining (for crystallisation), \( D \) = bulk distribution coefficient of the residue or the crystallising assemblage.
Batch (equilibrium) melting
The melt stays in equilibrium with the residue until it is extracted all at once. Mass balance \( C_0 = FC_L + (1-F)DC_L \) gives
$$\boxed{\frac{C_L}{C_0} = \frac{1}{D + F(1 - D)}}$$Limits worth memorising:
- \( D \to 0 \): \( C_L/C_0 \to 1/F \). A perfectly incompatible element is concentrated by \( 1/F \): 10% melting gives a tenfold enrichment, and no more.
- \( F \to 0 \): \( C_L/C_0 \to 1/D \). The first drop of melt has the largest possible enrichment.
- \( D = 1 \): \( C_L = C_0 \) at all \( F \).
- Ratios of two highly incompatible elements (for example La/Nb, Ba/Th) barely change during melting, because both enrichments are close to \( 1/F \). Such ratios reveal the source. Ratios of elements with different \( D \) (La/Yb with garnet in the residue) change strongly with \( F \).
The same equation describes equilibrium crystallisation, with \( F \) the fraction of liquid remaining.
Fractional (Rayleigh) crystallisation
Crystals are removed as they form, so the liquid continuously changes:
$$\boxed{\frac{C_L}{C_0} = F^{(D-1)}}$$- Incompatible (\( D < 1 \)): \( C_L \) rises steeply as \( F \to 0 \), without the \( 1/D \) ceiling of batch processes.
- Compatible (\( D > 1 \)): \( C_L \) falls very fast. Ni and Cr in basalts drop by orders of magnitude after modest olivine and spinel removal, which is why Ni (with MgO) is the best index of how primitive a basalt is.
Fractional melting
Each increment of melt is removed as it forms. The instantaneous melt and the pooled (aggregated) melt are:
$$\frac{C_L}{C_0} = \frac{1}{D}(1 - F)^{(1/D - 1)} \qquad \frac{\bar{C}_L}{C_0} = \frac{1 - (1 - F)^{1/D}}{F}$$For incompatible elements the instantaneous melt is extremely enriched at first, then becomes depleted almost at once; the pooled melt behaves much like a batch melt. Mid-ocean ridge melting is often modelled as near-fractional melting.
| Model | Equation | Limit for \( D \to 0 \) |
|---|---|---|
| Batch melting / equilibrium crystallisation | \( 1/[D + F(1 - D)] \) | \( 1/F \) |
| Rayleigh fractional crystallisation | \( F^{D-1} \) | \( 1/F \) |
| Fractional melting, instantaneous | \( (1/D)(1 - F)^{1/D - 1} \) | very large at \( F \to 0 \), then near 0 |
| Fractional melting, aggregated | \( [1 - (1 - F)^{1/D}]/F \) | \( 1/F \) |
Nucleation and Crystal Growth
Classical nucleation theory
A crystal cannot grow until a stable nucleus exists. Forming a spherical cluster of radius \( r \) from an undercooled melt releases volume free energy but costs surface energy:
$$\Delta G(r) = \frac{4}{3}\pi r^3\,\Delta G_v + 4\pi r^2\gamma$$where \( \Delta G_v < 0 \) is the free-energy change per unit volume of crystal (it grows with undercooling \( \Delta T \) below the liquidus, roughly \( \Delta G_v \approx -\Delta S_v\Delta T \)) and \( \gamma \) is the crystal–melt interfacial energy per unit area. Setting \( d\Delta G/dr = 0 \):
$$\boxed{r^* = -\frac{2\gamma}{\Delta G_v}} \qquad \boxed{\Delta G^* = \frac{16\pi\gamma^3}{3\,\Delta G_v^2}}$$- Clusters smaller than \( r^* \) tend to dissolve (embryos); larger ones grow (nuclei).
- Because \( \Delta G_v \propto \Delta T \): \( r^* \propto 1/\Delta T \) and \( \Delta G^* \propto 1/\Delta T^2 \). Doubling the undercooling halves the critical radius and quarters the barrier.
Nucleation rate. \( I = I_0\exp(-\Delta G^*/kT)\exp(-Q/kT) \): the first term grows with undercooling (smaller barrier), the second falls (slower atomic transport at lower \( T \)). So \( I \) is zero at the liquidus, rises to a maximum, and falls again at large undercooling.
Heterogeneous nucleation. On an existing surface (a crystal, a bubble, a wall) with wetting angle \( \theta \), the barrier is reduced by
$$\Delta G^*_{\text{het}} = \Delta G^*_{\text{hom}}\,f(\theta) \qquad f(\theta) = \frac{(2 + \cos\theta)(1 - \cos\theta)^2}{4}$$with \( f = 0.5 \) at \( \theta = 90^\circ \), about 0.16 at 60° and about 0.013 at 30°. In nature most nucleation is heterogeneous: crystals nucleate on other crystals (glomerocrysts, epitaxial overgrowths) and on vesicle walls.
Growth and texture
Growth rate also passes through a maximum with undercooling, but its peak lies at a smaller undercooling than the nucleation peak.
| Cooling regime | Undercooling | Nucleation vs growth | Texture |
|---|---|---|---|
| Very slow (plutons) | small | few nuclei, fast growth | coarse, phaneritic, equigranular |
| Moderate | moderate | both | fine-grained |
| Fast (chilled margin, lava surface) | large | many nuclei, little growth | aphanitic, microlites, spherulites |
| Quench (pillow rims, ash) | very large | nucleation suppressed too | glass |
| Two stages (slow then fast) | small then large | — | porphyritic: phenocrysts in a fine groundmass |
At large undercooling growth is limited by diffusion of components to the crystal, and crystals become skeletal, hopper, dendritic or spherulitic. Ostwald ripening (textural coarsening) takes the opposite route: small crystals, with higher surface energy per volume, dissolve and feed larger ones at nearly constant temperature.
Diffusion
Fick’s laws
First law (flux down a concentration gradient):
$$\boxed{J = -D\frac{\partial C}{\partial x}}$$Second law (how a profile evolves with time, for constant \( D \)):
$$\boxed{\frac{\partial C}{\partial t} = D\frac{\partial^2 C}{\partial x^2}}$$Here \( D \) is the diffusion coefficient (diffusivity), in m\(^2\) s\(^{-1}\); do not confuse it with the partition coefficient. For a step in concentration (a diffusion couple), the solution is an error function, \( C(x,t) = \frac{C_1 + C_2}{2} + \frac{C_1 - C_2}{2}\,\text{erf}\!\left(\frac{x}{2\sqrt{Dt}}\right) \) with the step at \( x = 0 \), concentration \( C_1 \) initially on the \( x > 0 \) side and \( C_2 \) on the \( x < 0 \) side.
Characteristic length and time
$$\boxed{x \sim \sqrt{Dt}} \qquad \boxed{t \sim \frac{x^2}{D}}$$(Some texts use \( \sqrt{2Dt} \) or \( \sqrt{4Dt} \); the factor changes the number, not the order of magnitude. Use the form the question gives.) Diffusion distance grows with the square root of time: 100 times longer gives only 10 times farther.
| Medium | Typical \( D \) (order of magnitude, m\(^2\) s\(^{-1}\)) |
|---|---|
| Ions in water at 25 °C | \( 10^{-9} \) |
| Cations in basaltic melt near 1200 °C | \( 10^{-11} \)–\( 10^{-12} \) |
| Cations in silicate minerals at magmatic \( T \) | \( 10^{-16} \)–\( 10^{-20} \) |
| Same minerals at crustal metamorphic \( T \) | far smaller |
These are rough orders; actual values depend strongly on the species, mineral and temperature.
Temperature dependence
Diffusion is thermally activated:
$$\boxed{D = D_0\exp\left(-\frac{E_a}{RT}\right)}$$A plot of \( \ln D \) against \( 1/T \) is a straight line of slope \( -E_a/R \). With typical activation energies of a few hundred kJ mol\(^{-1}\), cooling by 200 °C slows diffusion by two or more orders of magnitude.
Mechanisms and pathways
- Vacancy mechanism (most cations in silicates) and interstitial mechanism (small atoms such as H, He, Li).
- Volume (lattice) diffusion is slowest; grain-boundary and surface diffusion are much faster. Fluids along grain boundaries speed up metamorphic reactions greatly.
- In multi-component systems, interdiffusion (Fe–Mg exchange in olivine or garnet) couples the fluxes so that charge is conserved.
Closure temperature (Dodson, 1973)
A cooling mineral exchanges a daughter isotope (or an element) with its surroundings while diffusion is fast; it stops when diffusion becomes too slow over the grain’s size. Dodson’s closure temperature \( T_c \) is the temperature at the apparent age of the mineral:
$$\frac{E_a}{RT_c} = \ln\left(\frac{A\,\tau\,D_0}{a^2}\right) \qquad \tau = \frac{RT_c^2}{E_a\,(dT/dt)}$$where \( a \) is the effective diffusion radius (grain size) and \( A \) is a geometry factor (55 for a sphere, 27 for a cylinder, 8.7 for a plane sheet). The equation is solved by iteration. What matters for Part C is the direction of each effect:
| Factor | Effect on \( T_c \) |
|---|---|
| Larger grain size \( a \) | higher \( T_c \) |
| Faster cooling rate | higher \( T_c \) |
| Higher activation energy \( E_a \) | higher \( T_c \) (for comparable \( D \) at high \( T \)) |
| Larger \( D_0 \) | lower \( T_c \) |
This is why zircon (U–Pb, very slow Pb diffusion) closes above about 900 °C and biotite (Ar) near 300 °C, and why the ages of several minerals in one rock record its cooling path (see Radiogenic Isotopes and Geochronology).
Diffusion chronometry
A zoning profile that started sharp and has been smoothed records \( Dt \). If \( D \) is known at the relevant \( T \), \( t \approx x^2/D \) gives the time the crystal spent hot: days to years between magma mixing and eruption, from Fe–Mg profiles in olivine rims, for example.
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| Partition coefficient \( D \) (dimensionless ratio) | Diffusion coefficient \( D \) (m\(^2\) s\(^{-1}\)) | |
| Nernst \( D_i \): one element, one mineral | \( K_D \): exchange ratio of two elements (Fe/Mg) | |
| Batch: melt stays with residue, ceiling \( 1/D \) | Fractional: melt removed continuously | |
| Homogeneous nucleation: within the melt | Heterogeneous: on a surface, lower barrier | |
| Nucleation peak at large undercooling | Growth peak at small undercooling | |
| Volume diffusion (slow) | Grain-boundary diffusion (fast) |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper. Take \( R = 8.314 \) J mol\(^{-1}\) K\(^{-1}\) and 1 yr \( = 3.156\times 10^{7} \) s.
Level 1: Foundation
- Define the Nernst partition coefficient.
- What is the maximum enrichment of a perfectly incompatible element in a 5% batch melt?
- How does the critical nucleus radius change when undercooling doubles?
Answers
- \( D = C_{\text{mineral}}/C_{\text{melt}} \) at equilibrium, for a trace element.
- 20 times (\( 1/F = 1/0.05 \)).
- It is halved (\( r^* \propto 1/\Delta T \)).
Level 2: Part B Style
Practice Q1. A mantle residue contains 60% olivine, 30% orthopyroxene and 10% clinopyroxene by weight. Partition coefficients for Ni are 10, 4 and 2. The bulk \( D_{\text{Ni}} \) is:
(a) 5.3 (b) 6.0 (c) 7.4 (d) 16.0
Answer: (c) 7.4
\( D = 0.6\times 10 + 0.3\times 4 + 0.1\times 2 = 6.0 + 1.2 + 0.2 = 7.4 \). Option (a) is the unweighted mean of 10, 4 and 2; (d) is the plain sum.
Practice Q2. A basaltic melt with 300 ppm Ni undergoes Rayleigh fractional crystallisation of an assemblage with bulk \( D_{\text{Ni}} = 7.4 \). After 20% crystallisation (\( F = 0.8 \)), the Ni content of the liquid is closest to:
(a) 72 ppm (b) 150 ppm (c) 240 ppm (d) 375 ppm
Answer: (a) 72 ppm
\( C_L = C_0F^{D-1} = 300\times 0.8^{6.4} \). \( \ln 0.8 = -0.2231 \); \( 6.4\times(-0.2231) = -1.428 \); \( e^{-1.428} = 0.240 \). \( C_L = 72 \) ppm. Option (c) is 300 × 0.8, which ignores \( D \).
Practice Q3. Which statement about heterogeneous nucleation is correct?
(a) It needs a larger undercooling than homogeneous nucleation (b) Its energy barrier is lower, by a factor that decreases as the wetting angle decreases (c) Its critical radius is independent of undercooling (d) It cannot occur in silicate melts
Answer: (b)
\( \Delta G^*_{\text{het}} = \Delta G^*_{\text{hom}}f(\theta) \), and \( f(\theta) \) falls from 1 at 180° to 0 at 0°. A well-wetted substrate (small \( \theta \)) lowers the barrier sharply, so nucleation starts at smaller undercooling.
Practice Q4. With \( D = 10^{-20} \) m\(^2\) s\(^{-1}\), the characteristic diffusion distance \( \sqrt{Dt} \) in 1 Myr is about:
(a) 0.6 µm (b) 56 µm (c) 0.56 mm (d) 5.6 cm
Answer: (c) 0.56 mm
\( t = 10^{6}\times 3.156\times 10^{7} = 3.16\times 10^{13} \) s. \( Dt = 3.16\times 10^{-7} \) m\(^2\); \( \sqrt{Dt} = 5.6\times 10^{-4} \) m = 0.56 mm.
Level 3: Part C Style
Practice Q5. A peridotite source has 1.0 ppm La and 0.40 ppm Yb. During batch melting with residual garnet, \( D_{\text{La}} = 0.01 \) and \( D_{\text{Yb}} = 2.0 \). Which statements are correct?
- At \( F = 0.05 \), La in the melt is about 16.8 ppm.
- At \( F = 0.05 \), Yb in the melt is about 0.21 ppm.
- La/Yb in the melt at \( F = 0.05 \) is about 80, much higher than in the source (2.5).
- Raising \( F \) to 0.20 would increase La/Yb in the melt.
(a) 1, 2 and 3 only (b) 1 and 4 only (c) 2 and 3 only (d) 1, 2, 3 and 4
Answer: (a) 1, 2 and 3 only
- 1 true: \( C_L = 1.0/[0.01 + 0.05\times 0.99] = 1/0.0595 = 16.8 \) ppm.
- 2 true: \( C_L = 0.40/[2.0 + 0.05\times(1 - 2.0)] = 0.40/1.95 = 0.205 \) ppm.
- 3 true: \( 16.8/0.205 = 82 \), about 80, against 2.5 in the source.
- 4 false: at \( F = 0.20 \), La = \( 1/[0.01 + 0.2\times 0.99] = 4.81 \) ppm; Yb = \( 0.40/[2.0 - 0.2] = 0.222 \) ppm; La/Yb = 21.6, lower. Steep REE patterns need small melt fractions with garnet in the residue.
Practice Q6. For Fe–Mg interdiffusion in a mineral with \( E_a = 300 \) kJ mol\(^{-1}\), a crystal is held for the same time at 1000 °C (case A) and at 800 °C (case B). Which conclusion is correct?
(a) \( D_A/D_B \approx 1.19 \); zoning profiles are nearly equal (b) \( D_A/D_B \approx 200 \); the profile in A is about 14 times wider (c) \( D_A/D_B \approx 200 \); the profile in A is about 200 times wider (d) \( D_A/D_B \approx 14 \); the profile in A is about 3.7 times wider
Answer: (b)
\( \ln(D_A/D_B) = \frac{E_a}{R}\left(\frac{1}{1073.15} - \frac{1}{1273.15}\right) = 36084\times 1.464\times 10^{-4} = 5.28 \); \( D_A/D_B = e^{5.28} \approx 197 \). Width scales as \( \sqrt{D} \): \( \sqrt{197} \approx 14 \). Option (c) forgets the square root; option (a) takes the ratio of absolute temperatures.
Quick Revision Table
| Concept | Key point |
|---|---|
| Nernst \( D \) | \( C_{\text{min}}/C_{\text{melt}} \); constant by Henry’s law |
| Olivine–melt \( K_D \) (Fe–Mg) | about 0.30 |
| Bulk \( D \) | \( \sum x_jD_j \) |
| Batch melting | \( C_L/C_0 = 1/[D + F(1 - D)] \); limits \( 1/F \), \( 1/D \) |
| Rayleigh crystallisation | \( C_L/C_0 = F^{D-1} \) |
| Fractional melting | instantaneous \( (1/D)(1 - F)^{1/D-1} \) |
| Source ratios | ratios of two highly incompatible elements are preserved |
| Critical nucleus | \( r^* = -2\gamma/\Delta G_v \); \( \Delta G^* = 16\pi\gamma^3/3\Delta G_v^2 \) |
| Heterogeneous factor | \( f(\theta) = (2 + \cos\theta)(1 - \cos\theta)^2/4 \) |
| Texture | small \( \Delta T \): coarse; large \( \Delta T \): fine or glassy; two-stage: porphyritic |
| Fick | \( J = -D\,\partial C/\partial x \); \( \partial C/\partial t = D\,\partial^2C/\partial x^2 \) |
| Length scale | \( x \sim \sqrt{Dt} \) |
| Arrhenius | \( D = D_0e^{-E_a/RT} \) |
| Closure \( T \) (Dodson) | rises with grain size and cooling rate; \( A \) = 55, 27, 8.7 |
Key Takeaways
What’s Next
From high-temperature processes we move to the surface, where water, oxygen and electrons control element mobility: Redox, Eh–pH, Waters and Aerosols. For a refresher on how batch and Rayleigh equations reappear with isotopes, see Stable Isotopes and Geochemical Cycles.