Origin and composition of LLSVPs in the lowermost mantle
Reidar Trønnes, Natural History Museum, Univ. of Oslo
First-order Earth structure: PREM
Equatorial
section:
?
?
Trønnes (2010)
Dziewonski &
Anderson (1981)
LLSVP: Large low shear-velocity province: thermochemical pile (?)
- Material, origin, age
Seismic tomography models
Large vS-amplitudes at the top
and bottom of the mantle
Harvard model, Masters and Laske, website
S-wave models, lowermost mantle (D”-zone)
Two large anti-podal, slow provinces - LLSVP
Africa – Pacific (near equator - 180º apart)
Dziewonski et al.
(2010, EPSL)
The main, degree-2 velocity anomaly
was recognized about 35 years ago !
e.g. Dziewonsky et al. (1977, JGR)
Dziewonski & Anderson (1984, Am Sci)
Earth’s rotation axis related to mantle mass distribution and geoid
Steinberger and Torsvik (2010, GGG)
Actual rotation axis
Combined contributions: LLSVPs +
shallow slab mass contributions
Calculated rotation axis from
LLSVP-contributions, only
Comparison of seismic tomography (LLSVPs)
and slab-sinking model at 2800 km depth
Dziewonski et al. (2010, EPSL)
Spherical harmonics modeling
Power spectra
Cumulative power spectra
Tomographic
models
Slab model
Degree 2
Degree 2
Tentative conclusions
Lithgow-Bertelloni & Richards
(1998, Rev. Geophys.)
1. The observed degree-2 pattern is only partly reproduced
by calculated slab-accumulation
2. The LM-structure may thus be old ( > 300-500 Ma)
Plume generation along the margins of LLSVPs:
evidence from relocated LIPs
SC
Large igneous provinces (LIPs)
- age span: 16-297 Ma
- irregular distribution
Paleogeographic relocation
→ LIPs cluster near LLSVP-margins
- long-term stability
- dense and hot
SC
Africa
Pacific
–3%
slow
Burke & Torsvik, 2004, EPSL
Torsvik et al., 2006, GIJ
Burke et al. 2007, EPSL
Torsvik et al. 2008, EPSL
Additional kimberlite and LIP data
Torsvik et al. (2010, Nature)
LLSVP-stability may exceed 540 Ma
Seismological image of
Plume Generation Zones (PGZ)
S-wave model
S-wave model
NE part of Pacific LLSVP
Samoa quakes, recorded in N-America
Lay et al.
(2006, Science)
Double crossing of the
pv-ppv-transition
Large lateral variation
 horizontal flow (and PGZ)
Mantle flow model
Bin 1-3
D”-discontinuities
Lay and Helmberger (1983, GJRAS): S-wave triplication: S, Scd and ScS (in certain areas, at least)
The Scd and Scd2 may be
ascribed to double-crossing of
the post-perovskite boundary
- large thermal gradient in the D”
- large dp/dT-slope of phase bound.
pv-ppv transition
- wide phase loop: pv and ppv coexist
through the entire D"-zone
and
-the Al2O3-component stabilizes pv
and widens the phase loop
Catalli et al. (2009, Nature)
Possible rheological explanation for sharp D”-discontinuities
(Amman et al. 2010)
Strong alignment and dislocation creep
in ppv at a ”critical” phase proportion
(40-50% ?)
With this model:
- D”-discontinuities: rheological changes
(steadily increasing ppv-fraction with depth)
- High T may facilitate diffusion creep
below the lower discontinuity.
- Additionally: the lower discontinuity could
also be caused by back-reaction to pv.
But probably no ppv inside LLSVPs
- hot and rich in basaltic material (?)
Locally steep thermochemical pile margins
Requirements:
- moderate density contrast (2-5 %)
- pile material: higher bulk modulus than ambient mantle
Garnero & McNamara
(2008, Science)
High thermal conductivity and low thermal expansivity in the
lowermost mantle may help to stabilize the thermochemical piles
Possible LLSVP-material
Basalt-rich
- separated from subducted lithosphere
- age: 3-0 Ga
Perdotitic (or komatiitic) with elevated Fe/Mg-ratio
- cumulates, deep-level partial melts
- age: mainly Hadean
Mantle mineralogy
Irifune & Tsuchiya (2007, Treatise on Geophys.)
Shim et al. (2011, this meeting)
Density relations
peridotite - basalt
Irifune & Tsuchia, 2007,
Treatise on Geophys.
K0 (GPa)
Mg-pv 230-260 (perid. is stiffer ?)
Ca-pv
236
softest: ferroper. 158-152 (FeO-MgO)
stiffest: silica 314-325 (stish. - aPbO2)
K’: poorly constrained
Basalt: pv, Ca-pv, SiO2, Al-phase
high r, possibly higher K0
Peridotite: pv, fp, Ca-pv
low r, possibly lower K0
Density contrast: 1-3%
- sufficient ?
Peridotite (or komatiite) with elevated Fe/Mg-ratios
Origins:
Magma ocean cumulates from late-stage, residual melts
- crystallization near CMB
- crystallization in TZ or UM, followed by density-driven sinking to CMB
Deep-level Hadean melting in hot plumes at >300 km depth,
followed by downward or upward migration to 410 km depth,
crystallization, cooling and sinking to CMB
(possible plume initiation by density overturn of cumulate sequences)
Compared to ambient” peridotite:
- Similar mineralogy (in D” mainly pv/ppv and fp)
- Higher density (possibly higher than basalt)
- Higher bulk modulus
→ LLSVP-requirements may easily be fulfilled
Dense cumulates from crystallization of lower mantle magma ocean
Depend on relative slopes of peridotite liquidus and melt isentropes
z
If melt adiabat intersects
the curved liquidus here,
the magma ocean will strart
crystallizing in the middle
Scenario with two magma ocean
Labrosse et al (2007, Nature)
Stixrude et al. (2009, Earth Planet. Sci. Lett.)
Stage 1
Core
Stage 2
Cumulates with lower Fe/Mg
Cumulates with higher Fe/Mg
Core
Stage 3
Inner magma ocean:
melt density > crystal density (pv, fp)
(Fe/Mg)melt > (Fe/Mg)crystals
Fe-rich cumulates
starting point for
thermochemical piles
Core
Sinking of solidified melts from 410 km depth
melts formed in hot plumes at 300–900 km
Suggested by:
Lee et al. (2010, Nature)
→ Intermediate age span: Hadean-Archean (between scenarios 1 and 2)
Based on:
Zhang & Herzberg (1994, JGR)
Tønnes & Frost (2002, EPSL)
Ito et al. (2004, PEPI)
Lee et al. (2010, Nature)
Melt accumulation zone
Solidified, thermally
equilibrated melt
sink to the CMB
Unresolved issue: pseudo-invariant melt compositions at 20-30 GPa
- liquidus phase variation can guide
- systematic experimentation on a range of model compositions
Further experiments
with D.J. Frost,
BGI-Bayreuth
Other important tasks
Better data on phase transitions and EoS in basaltic material
Na-Al-phase (15-20 %): Ca-ferrite to Ca-titanite structure (??)
Silica phase (10-15 %): CaCl2- to aPbO2-structure
- p-T-condition of transition, including Clapeyron slope
- compositional relations (silica-phases may contain up to 12% Al2O3)
Possible silica analogue compositions : TiO2, ZrO2, CaCl2 and aPbO2 (at var. T)
For all minerals, better data on:
- thermal conductivity, incl. radiative conductivity
- Fe-spin transitions (in the minerals pv, ppv and fp)
- thermal expansivity (and EoS in general)
- mechanical propertis, diffusivity, deformation style (viscocity)
Geochemistry
The relations of ULVZs and LLSVPs with possible long-lived, enriched (fertile)
mantle reservoirs
Large thermal boundary layer at CMB
Mantle-core mixing is prevented by contrasts in
density (5500 - 9900 kg/m3) and viscosity
Large T-increase → viscosity decrease in the D”
CMB
From: Steinberger and Calderwood (2006, GJI)
Thermochemical piles (LLSVPs)
3 possible origins – 3 different age scenarios
Independant evidence for long-term stability:
e.g. several studies by Torsvik et al., Dziewonsky et al. (2010, EPSL)
Mechanism 1: Segregation and accumulation of basaltic parts of subducted slabs
→ slow growth over most of Earth history
The plume generation zone:
density-driven separation
basalt – peridotite
Trønnes (2010, Mineral. Petrol.)
In spite of viscocity decrease in D”:
The rheology of the mantle imposes the convective
and thermal regime of the core
Terrestrial planets with liquid cores:
”Mantle is the master - core is the slave”
(Dave Stevenson, Caltech)
Crystal structures
Pv: HIGH entropy
Post-pv: LOW entropy
pv-ppv transition has large,
positive dp/dT-slope
MgSiO3 (Murakami et al. 2004)
Phase boundary:
not well constrained
by DAC-experiments
Analogue system:
CaIrO3
DFT-model of pv-ppv in CaIrO3
Stølen & Trønnes (2007, PEPI)
dp/dT = 19 MPa/K
DK: negative (reaction: pv→ppv)
DG: positive
(Dr: negative)
Similar DFT-results for MgSiO3
(e.g. Wookey et al. 2005, Nature)
vs2 = G/r
vf2 = K/r (bulk sound speed)
vp2 = (K + 4/3*G)/r
Consistent with the
anti-correlated vS and vF
Compressibility of
pv and ppv, CaIrO3
Experiments: Boffa-Ballaran et al. 2007, Am Min.
DFT: Stølen & Trønnes 2007, PEPI
So the D"-discontinuities disappear !?
But then:
Amman et al. (2010, Nature)
DFT-computation of diffusion rates: pv, fp and ppv
Step 1: Testing of agreement between existing experimental data
and computations for pv and fp.
Result: good agreement
Step 2: Computation of diffusion rates for ppv
Result: strongly anisotropic diffusion in ppv
with fast diffusion along a-axis
low diffusion creep viscosity along a-axis
Most of the lower mantle
- no seismic anisotropy, small grain size and low stress (Solomatov et al. 2002)
- high viscosity
diffusion creep is likely
D”:
- cold areas: strong seismic anisotropy (high VSH)
- low viscosity
extensive deformation and LPO is likely
deformation-related dislocation creep is likely
CMB
Steinberger and Calderwood (2006, GJI)
Ammann et al. (2010, Nature), Hunt et al. (2009, Nature Geoscience):
For dislocation creep: ppv may be 4 orders of magnitude weaker than pv
Rheology changes dramatically at critical phase fraction of 30-50% ppv
New model for D"-discontinuities
With this model:
- D”-discontinuities: rheological changes
(steadily increasing ppv-fraction with depth)
- High T may facilitate diffusion creep
below the lower discontinuity.
- Additionally: the lower discontinuity could
also be caused by back-reaction to pv.
Structure and dynamics of D”
- Basalt is denser and stiffer (higher K0) than peridotite (consistent with LLSVPs and PGZs)
- Deformation / LPO of ppv at critical phase fraction may eplain the seismic D"-disc. in low-T areas
Important unresolved issue:
Seismic observation of discontinuities inside the hot LLSVPs (thermo-chemical piles, basaltic?)
cannot bed due to the pv-ppv-transition (relative stabilization of pv by the FeAlO3-component precludes this)
T
Could other phase transitions in basalt-rich material be responsible ?
- Possible candidate: CaCl2- to aPbO2-structure of SiO2
- Do we know the Clapeyron slope of this transition ?
- Ohta et al. (2008, EPSL) indicate positive dp/dT, but this is not well established
- Positive slope could explain a double-crossing scenario
p
CMB
core
Earth dynamics –
a modified working hypothesis
Modified from Trønnes (2009/10, Mineral. Petrol.)
?
?
The plume generation zone:
density-driven separation of basalt and peridotite
Schematic
equatorial
section
CaIrO3-based analogues (for MgSiO3)
- space group match: pv: Pbnm, ppv: Cmcm
- phase transition at 1-3 GPa and 1400-1600 ºC
- both phases are quenchable to ambient conditions, enabling single-crystal XRD:
single-crystal structure refinement, DAC-compressibility, thermal expansivity
- bulk and shear moduli changes for the pv-ppv-transition correspond to MgSiO3-based comp.
- deformation mechanisms and slip systems may be similar to D” and MgSiO3-based comp.
E.g. Walte et al. (2009), Hunt et al. (2009), Amman et al. (2010)
But also contradictory indications for slip mechanisms, e.g. Miyagi et al. (2010)
Substitutions in CaIrO3-based systems
(possible studies of phase relations, mineral physics and deformation)
Divalent A-site subsitutions for Ca: Sr, Ba
- corresponding to Mg-Fe-substitutions
Trivalent A- and B-site substitutions for Ca and Ir: In, Sc, Y
- corresponding to end members: Al2O3, FeAlO3, MgAlO2.5
Large T-gradients in D”
large positive dp/dT-slope of pv to ppv transition
 re-stabilization of pv near CMB
the "double-crossing" scenario of Hernlund et al (2005, Nature)
Thermal gradients
S-wave speed
Another characteristic feature:
anti-correlated vS and vF