Chemical consequences of perovskite fractionation from an ultramafic liquid with application to the dynamics of a basal magma ocean

Leah Ziegler, Hongluo Zhang, Colin Jackson, Matt Jackson, Dave Stegman

Team BMO:

What is a basal magma ocean?

Labrosse et al., 2007

- a dense melt layer that forms at the bottom of the mantle – likely linked to a whole mantle magma ocean

Why a basal magma ocean?-poorly understood lower mantle structures:

-missing geochemical reservoirs:Garnero and McNamara, 2008

Labrosse et al., 2007

How would a basal magma ocean form?

Figures: Stixrude et al., 2009

-Density crossover between liquid and solid at mid-mantle depths is possible – depends on Kd-Fe

-Curvature of mantle liquidus can lead to crystallization initiating at mid-mantle depths

w/ Few/o Fe

Density of liquid and solid with crystallization - f(Xliq, Xsol, XU+Th+K, time)

Concentration of REE and heat producers (U+Th) associated with BMO products

- Both Require:- major element partition coefficients

(Kd)- trace element partition coefficients

Key parameters for evaluating BMO:

Step 1: Compile Kd database for Mg-Pv, Ca-Pv, & Fe-MgO

-Database compiled at CIDER 2012

-Details of database:

Total number of studies: 19

Experiments including -- Mg-Pv: 31 -- Ca-Pv: 10 – Fe-MgO:12

Pressure Range: 23 – 86 GPa

Temperature Range: 2300 – 3500°C

Database will be published with initial reporting of findings

Effect of Al content on the solubility of Fe in MgPv & Kd-TE

Importance of charge coupled substitutions

right: Frost and Langenhorst, 2002left: Liebske et al., 2005

Step 2: Parameterize Kds for Mg-Pv

ln(Kd )c1P c2Xmelt c3X solid c4

Models for Mg-Pv major elements:

Observed

Pred

icted

KdAl

KdCa

KdFe

KdMg

KdSi

Equal weight, multiple linear regression

Step 2: Parameterize Kds for Mg-Pv

ln(Kd )c1Casolid c2Alsolid c3Models for Mg-Pv trace elements:

Observed

Pred

icted

KdU

KdTh

KdSm

KdNd

Similar TE substitutions as pyroxene

Team BMO met in October at BU:- Refined Kd parameterizations- Developed stoichiometric Mg-Pv crystallization model- Quantified uncertainties in crystallization model predictions

using monte carlo approach - Accounts for covariation in model parameters

- Thank you Don Forsyth

Step 3: Apply Kds for Mg-Pv BMO crystallization

Model Results

Al

Si

Fe

Mg

Ca

U

Th

Sm

Nd

Liquid evolution with crystallization (batch):

Majors: cation mol fraction, Trace: PPM*Bulk: McDonough & Sun BSE, 40 GPa

Xmol%

Al-free

Al-bearing

Model Results Kd

Fe with crystallization (BSE bulk, 40 GPa):

ln(KdFe )2.6Siliq 11.5Alsolid 2.4

Results from 78.5 GPa: Andrault et al., 2012

Al + Si causing offset between Nomura and Andrault ?

KdAl

Model Results Removal of heat producing elements and REE from liquid (batch):

High Al in MgPv makes it significant reservoir for TEsStill working to incorporate full uncertainties into TE Kds

-Al partitions equally between MgPV and liquid

-REE+U+Th have similar substitutions into pyx and MgPv- Al AND Ca are important

-Fe is a moderately incompatible element– Fe is strongly controlled by Al, but other effects

appear significant (Siliq?)

Preliminary Findings

Post-AGU Team BMO Meet up:

- Develop new crystallization model- Weaver and Langmuir (1990)

- Working with Stephane Escrig to implement his code

- stoichiometry- self-consistent saturation of phases- Predict past Fe-MgO or Ca-Pv saturation

- Couple chemical model to dynamical model

Current Directions

unknowns : liqsi, liqAl ,liqFe, liqMg ,liqCa

solidi (bulki liqiF)(1 F)

1)KMgPv LiqAl Si liqAl

liqSisolidAlsolidSi

2)solidSi solidAl 0.5

3)KMgPv LiqFe Mg liqFe

liqMgsolidFesolidMg

4)KMgPv LiqCa Mg liqCa

liqMg solidCasolidMg

5)solidFe solidMg solidCa 0.5

System of Equations:

Crystallization Model

-Maintains stoichiometry on Si and Mg site

Effect of Al content on Kd-Fe at high P & T

Al-free

Al-bearing

Andrault et al., 2012

Kds

Kdi

Al

Si

Fe

Mg

Ca

U

Th

Sm

Nd

Solid Evolution

Xmol%

Al

Si

Fe

Mg

Ca

U

Th

Sm

Nd