• Composition of the source• Partial melting process• Fractional crystallization• Crustal assimilation (continental or oceanic)

Factors controlling magma composition

Effect of tectonic setting on chemistry ofmantle-derived melts

Lherzolite Basalt

1) Ridge volcanism (MORB - "Mid Ocean Ridge Basalts")- creation of oceanic lithosphere

2) Intraplate volcanism- Large Igneous Provinces (LIP)

Oceanic Flood Basalts (OFB)Continental Flood Basalts (CFB)

- Ocean Island Basalts (OIB)

3) Arc volcanism (oceanic and continental)- destruction of oceanic lithosphere- creation of continental crust?

Tectonic contexts of mantle partial melting

Volcanism beneath mid-ocean ridges

Sample types available:• MORB (Mid Ocean RidgeBasalts)• Gabbros• Abyssal peridotites• Ophiolites

Gale et al. (2014)

60,000 km of mid-ocean ridges

MORB• Most abundant volcanic rocks on Earth• Very large sample set available relative to deeper rock types of ridges• But only 10 - 15% of oceanic crust

Generation of oceanic crust(rapid spreading ridge)

MORB

adiabat

As the mantle rises beneath the ridge, it follows an adiabatic geotherm, in whichthe decrease in temperature is linked uniquely to the decrease in pressure(thermodynamic effect). At around 50 to 100 km, this geotherm intersects theperidotite solidus, and partial melting begins.

Why does themantle meltbeneath theridges?

Nelson course notesTulane Univ. website

First melts

Peridotite

DiopsideCaMgSi2O6

ForsteriteMg2SiO4

EnstatiteMgSiO3

Simplifiedperidotitemelting

Peridotite melting producesbasalts

Major element chemistry of MORBmay depend on:• Source compostion• H2O content• T and P of melting• Extent of melting• Fractional crystallization• Assimilation of pre-existing crustand sediments

Given the many factors that control MORB chemistry,how can we make sense of their compositions?

Klein and Langmuir (1987)

• Between ridges, large differences in Na2O at each MgO value.

To understand the variations between ridges, must remove effects of fractionalcrystallization.

Major elements in MORB vary with ridge depth

• AAD (Antarctic discordance) deep

• Tamayo Fracture Zone - mediumdepth

• Kolbeinsey Ridge - shallow

Concentrations averaged over100 km ridge segments to filterout small scale variations.

MgO wt%

Na 2

O w

t %deep

mediumdepth

shallow

1098765Increasingfractionation

• On each ridge, NaO2 varies with MgO - consistent with fractional crystallization

Olivine is the liquidus phase. Plagioclase and clinopyroxene appearwith decreasing temperature

Langmuir et al. (1992)

Typical low pressure fractionation sequence of MORB

FractionalCrystallization

Olivine

Ol+Pl

Ol+Pl+Cpx

Mg (cation %)

T (o

C)

Ol = olivinePl = plagioclaseCpx = clinopyroxene

Tamayo Region, East Pacific Rise

Effect of fractionation on liquidcomposition:

Example of Mg, Na and Al

Primitive basalt~ 5.8% Mg (9.7% MgO)~ 1.8% Na~ 8.4% Al

Effect of olivine fractionation on liquid:

Mg Na

Olivine(Fo 90 = Mg-rich olivine)~ 30 wt% Mg~ 0 wt% Na~ 0 wt% Al

Al

Effect of plagioclase fractionation on liquid:

Mg Na

Plagioclase(An 90 = Ca-rich plag)~ 0 wt% Mg~ 1 wt% Na~ 19 wt% Al

Al

• increase in TiO2 and Na2O:incompatible in all major phases

• decrease in Al2O3 and increase inFeO: plagioclase crystallization

• kink, then decrease in CaO: onsetof cpx crystallization

W+L : program of Weaver et Langmuir (1990)N : program of Nielsen (1985)

Langmuir et al. (1992)

Fractional crystallizationEffect of low pressurefractionation on MORBmajor element composition

Clipperton fraction zone (EPR)

Langmuir et al. (1992)

MgO = 8%

Comparison at 8 wt% MgO(Na 8.0, Fe8.0, Ti8.0, Ca8.0…)

Use 8% MgO because:1) only moderate degrees of fractionation2) most basalt series include 8% MgO samples

To see effects of differences in partial melting linked to ridgedepth must first correct for fractional crystallization

Gale et al. (2104)

Alternative fractionation correction

Back-correction until magma in equilibriumwith mantle olivine (Fo90)(Na90, Fe90, Ti90, Ca90...)

• Advantage: Can directly compare results withperidotite melting experiments

• Disadvantage: Must extrapolate beyond datarange

Results of two methods are strongly correlatedand lead to similar interpretations

After correction for fractional crystallization, aglobal correlation exists between ridge depthand major element composition of MORB(Klein et Langmuir, 1987).

Each point represents the average of basaltvalues from ~100 km of ridge.

Black squares : MORB from "normal" ridgesWhite squares : MORB influenced by hot spotsCrosses : MORB next to hot spotsSmall diamonds : Back-arc basin basalts

Langmuir et al. (1992)

Correlations confirmedby recent results

Results fromcompilation of Gale etal (2014) using threetimes as much data.

Outlined fields: normal ridges (solid), back-arcs (dashed); Langmuir et al. (1992)

plumes

back-arcs

Crustal thickness andchemical composition alsocorrelate.

In general, where the ridge isshallow, the crust is thick.

Implication : In each ridge segment, the MORB compositionis linked to the quantity of magma produced.

Quantity of magma depends on depth at which melting starts. Thisdepth depends on pressure at which peridotite solidus is crossed.

Where the potential temperature is high, mantle material rising alongan adiabat crosses the peridotite solidus at greater depth, and alarger quantity of magma is produced.

Potential temperature = thetemperature that an adiabaticgeotherm would have at thesurface of the Earth

10%

Higher potential temperature partial melting starts at greater depthTwo effects :

1) Greater quantity of magma2) Higher average pressure of partial melting

Polybaric partial melting in an upwelling mantle column

Cold Mantle Hot Mantle

ResidualMantleColumn

ResidualMantleColumn

RidgeAxis

Melting Regime

Solidus

Solidus

40% removed

20% removed

10% removed

0% removed

20%

30% 30% removed

RidgeAxis

Langmuir et al. (1992)

1) Effect of magma quantity on MORB composition

Low potentialtemperature

Small quantityof magma

Thin oceaniccrust

High Na8.0 concentration

Deep mid-ocean ridge

Concentration of incompatible element,(e.g. Na) decreases as extent of partialmelting increases (by dilution) Na8.0correlated with ridge depth

Models of decrease in magma Na2O contentwith increasing extent of partial melting (F).Langmuir et al. (1992)

2) Effect of melting pressureon magma compositions

MgO and especially FeO increase,SiO2 decreases with melting pressure.

Experimental magma compositions from partialmelting of "pyrolite" at various pressures.

5

10

5

5

5

10

10

10

15

15

15

15

numbers: P (kb)

Pyrolitemelts

pyrolite = particular fertile peridotitecomposition taken by many as primitivemantle composition

Figure from Langmuir et al. (1992) usingdata of Jacques and Green (1980).

Low potentialtemperature

Small quantityof magma

Thin oceaniccrust

High Na8.0 concentration

Deep mid-ocean ridge

Low Fe8.0 concentration

Negative correlation between Na8.0 and Fe8.0

Shallowaverage meltingpressure

Polybaric melting, with melt onsetdepending on potential temperature canexplain negative correlation betweenFe8.0 and Na8.0.

However, some differences existbetween regions and tectonic contexts:probably reflect source variations.

Black squares : MORB from "normal" ridges (N-MORB)White squares : MORB influenced by hot spotsCrosses : MORB adjacent to hot spotsDiamonds : back-arc basin basalts

⇐ Crustal thickness

Langmuir et al. (1992)

Variations according to tectonic context

Na8.0

Among all MORB, a rough correlation exists between Ti8.0 and Na8.0 asexpected from model. But MORB from certain tectonic contexts (back-arcbasins, adjacent to hot spots) plot off trend.

Mostly reflects variation of source composition.

Black squares : MORB from "normal" ridges (N-MORB)White squares : MORB influenced by hot spotsCrosses : MORB adjacent to hot spotsDiamonds : back-arc basin basaltsTi

8.0

Na8.0

For Ti8.0, variations among N-MORB from different ocean basins. Reflectsubtle differences between the MORB sources beneath each basin.

Variations between ocean basins:N-MORB only

Ti8.

0

Na8.0 Langmuir et al. (1992)

MORB: Message from major elements

• Melt composition strongly controlled by fractionation

• Removing fractionation effects reveals correlation betweencrustal thickness and melt chemistry

• Likely explanation: polybaric melting during mantle upwelling.Temperature controls both average pressure and extent of melting

• Removing melting effects reveals subtle source differencesbetween ocean basins and tectonic contexts

What can we learn from trace elements inMORB?

MORB

Continental crust

CH

ON

DR

ITE

NO

RM

ALIZ

EDC

ON

CEN

TRAT

ION

Moderately incompatibleelements have highestnormalized abundancesin MORB.

Hofmann (1988)

Magmas produced bypartial melting ofresidual mantle

The hump in moderately incompatible elements results from partialmelting of mantle material that has already been depleted in highlyincompatible elements by crustal extraction.

How are MORB trace element patterns produced?

Hofmann (1988)

D = concentration in residueconcentration in liquid

F = fraction of melting

PRIM

ITIV

E M

ANTL

EN

OR

MAL

IZED

CO

NC

ENTR

ATIO

N

0

2

4

6

8

10

12

Rb Pb U Th Ba K La Ce Nb Pr Sr Nd Hf Zr Sm Eu Gd Tb Dy Ho Er Y Tm Lu Cu Sc Co Ni

Pri

mitiv

e M

antle

norm

aliz

ed

Bulkoceancrust

MORB

Lowercrust

Increasing compatibility

MORB vs. bulk and lower crustal concentrations

Only 10 to 15% of the ocean crust is composed of MORB.Most of crust is composed of gabbro.

Crustal concentrationsgiven by White (2013)

Factors controlling trace element fractionationduring partial melting

• Liquid remains in contact withresidue until end of melting

• Moderate depletion inincompatible elements

• High porosity

1) The melting process

Liquid"Batchmelting"

• Liquid removed as soon as it isproduced

• Very strong depletion inincompatible elements

• Vanishingly small porosity

Liquid"FractionalMelting"

Reality undoubtedly between the two cases

Elemental abundances of partial melt determined by mineralogicalcomposition of rock and partition coefficients of each phase.

concentration in phase iconcentration in liquid

kdi =

Global partition coefficient, D = Σ kdi xi = Csolid/Cliquid(xi = proportion of phase i )

Phase proportions can change due to loss of most fusible phases (e.g.,clinopyroxene) or to pressure changes (garnet to spinel as Al-bearingphase) D changes during melting.

Partition coefficients

Derivation of batchmelting equation

From mass balance: Co = FCL + (1-F)CS

By definition: D = CS/CL

Substituting for CS: Co = FCL + (1-F)DCL

Rearranging terms: Co = (F + D -DF) CL

Co

CL 1D + F(1 -D)

=F = fraction of liquidCo = original concentration in solidCL = concentration in liquidCS = concentration in residue

Liquid"Batchmelting"

Non-modal melting: phases do not enter the melt in the same proportionsthat they are present in the original solid (more realistic than modal melting).

P = bulk partition coeffient of phases melting to form liquid

P = Σ kdi yi (yi = proportion of phase i entering melt)

At any point during melting: D = (Do - PF)/(1-F)(Do = original value of D)

Modal melting Non-Modal melting

As developed by Shaw (1970):

MIT OpenCourseWare, basedon equations of Shaw (1970)

Equations for fractional melting

Modal melting Non-Modal melting

MIT OpenCourseWare, basedon equations of Shaw (1970)

Concentrations inliquids and residuesproduced by batchand fractional melting

CL : batch melting liquidCl : fractional melting instanteous meltCL : fractional melting aggregated melt

CS : batch melting residueCs : fractional melting residue

Example: Variation of Laconcentration during non-modalmelting of garnet pyroxenite (La ishighly incompatible).

Factors controlling trace element fractionationduring partial melting

2) Source mineralogyLherzolite mineralogy: olivine (55 - 85%), orthopyroxene (10 - 35%),clinopyroxene (5 - 20%), phase rich in Al2O3 (0 - 10%)

As T and P vary with depth, the mineralogy of the Al2O3 rich phase changes

• Down to ~24 km, plagioclase is stable

• Between ~24 and ~70 km, spinel is stable

• Below ~70 km, garnet is stable

These mineralogical changes produce large changes in bulk partition coefficientsduring partial melting.

Most incompatibleelements are inclinopyroxene

Many trace elements,particularly the HREE, arehosted mostly in garnet

Effect of source mineralogy on trace elementcompositions of melts: Example of REE

kdclinopyroxene

garnet

For garnet, kd increases dramatically with mass of REE. Melting ingarnet facies can therefore strongly fractionate REE.

kd's of McKenzie and O'Nions (1991)

compatible

incompatible

LREE HREE

• At > 70 km, garnet is stable inperidotite. Strong fractionation of REE, inliquids and residues.

• At ~ 70 to 24 km, spinel replaces garnetas Al2O3 bearing phase. REE partitioningcontrolled mostly by clinopyroxene, sovery little fractionation of HREE.

garnetfacies liquid

garnet faciesreisdue

spinel facies reisdue

spinel facies liquid

CL /

Co

or C

S/C

o Effect of pressure, and thusmineralogy, on REE in melt

Assumptions: 8% melting, partition coefficients of McKenzie & O'nions (1981), non-modal batch melting model. Phaseproportions garnet facies: 55%ol, 20%opx, 15%cpx, 10% gt; melting proportions: 50%cpx, 50%gt; phase proportions spinelfacies: 55%ol, 22%opx, 18% cpx, 5% sp; melting proportions: 5% opx, 95% cpx.

Strong fractionation of HREEimplies deep melting

White (2009)

1

10

100

1000

La Ce Pr NdSmEuGdTb DyHo Er TmYb Lu

Cal

cula

ted

liqui

d/ch

ondr

ites

Depleted mantle values of Salters andStracke (2004) used for mantle source

Calculated melts fromdepleted mantle

MORB generation mainly occursin spinel facies (< 70 km)

melting in spinelfacies

melting ingarnet facies

Factors controlling trace element fractionationduring partial melting

3) Extent of meltingC

L /Co

CL /C

o

Spinel Facies Garnet Facies1% 1%

10% 10%

At large extents of melting typical of MORB magmatism, very littlefractionation between incompatible trace elements because nearlyall are found in melt phase.

Trace elementratios in MORB:Useful source

tracers

• Partial melting creates only limited trace elementvariation in MORB because of high extents of melting

• Fractional crystallization has almost no effect onincompatible trace element ratios

Incompatible trace element ratios in MORBcan be used to trace source variations.

E-MORB: "Enriched"(La/Sm)N > 1.8 Requires an enriched source

N-MORB: "Normal"(La/Sm)N < 1

T- MORB: "Transitional"(between the others)

Schilling et al. (1983)

Ulrich et al. (2012)

Trace element and isotopic variationsalong mid-Atlantic Ridge

Implication : MORB source is not completely uniform ata ~100 km length scale

MORB are isotopically quite homogeneous comparedto OIB and crustal magmas

Radiogenicisotopes

Ideal source tracersbecause notfractionated bymelting process

White (2009)

Still, some systematic variationsare observed between oceanbasins (Sr, Nd, Pb…)

Indian Ocean MORBEvidence of an enrichedcomponent: recycledpelagic sediments?

Continentalcrust

White (2009)

MORB - Points to remember

• After effects of fractionation removed, MORB major elementcompositions correlate with crustal thickness. Probably explained bypolybaric melting, with onset of melting dependant on temperature.

• Relative enrichment in moderately incompatible elementsobserved, resulting from remelting of a mantle depleted by crustalextraction, ie, MORB source is depleted in incompatible elements.

• Trace element abundances suggest MORB produced by highdegrees of melting in spinel facies.

• MORB source relatively homogeneous compared to other Earthreservoirs. However major and trace elements and isotopes allindicate subtle variations at large and small length scales.