the density structure of subcontinental lithosphere through time...

The density structure of subcontinental lithosphere throughtime

Yvette H. Poudjom Djomani a;*, Suzanne Y. O'Reilly a, W.L. Gri¤n a;b,P. Morgan a

a GEMOC ARC National Key Centre, Department of Earth and Planetary Sciences, Macquarie University, Sydney,N.S.W. 2109, Australia

b CSIRO Exploration and Mining, P.O. Box 136, North Ryde, N.S.W. 1670, Australia

Received 5 October 2000; accepted 23 November 2000

Abstract

This study uses information on composition, thermal state and petrological thickness to calculate the densities ofdifferent types of subcontinental lithospheric mantle (SCLM). Data from mantle-derived peridotite xenoliths andgarnet^xenocryst suites document a secular evolution in the composition of SCLM: the mean composition of newlyformed SCLM has become progressively less depleted, in terms of Al, Ca, mg# and Fe/Al, from Archean, throughProterozoic to Phanerozoic time. Thermobarometric analyses of xenolith and xenocryst suites worldwide show that themean lithospheric palaeogeotherms rise from low values (corresponding to surface heat flows of 35^40 mW/m2) beneathArchean terranes, to higher values (s 50 mW/m2) beneath regions with Phanerozoic crust. The typical thickness of thelithosphere (defined as a chemical boundary layer), ranges from about 250 to 180 km, 180^150 km and 140^60 km forArchean, Proterozoic and Phanerozoic terranes respectively. The depth of this lithosphere^asthenosphere boundarycorresponds to a temperature of 1250^1300³C. Using the estimated compositions, average mineral compositions andexperimental data on the densities of mineral end-members (tables 1 and 2), we calculate mean densities at 20³C forPrimitive Mantle (3.39 Mg m33) and for SCLM of Archean (3.31 þ .016 Mg m33), Proterozoic (3.35 þ 0.02 Mg m33)and Phanerozoic (3.36 þ 0.02 Mg m33) age. Curves of density and cumulative density versus depth, which take intoaccount variations in geotherm with tectonothermal age, have been constructed for each age type of lithospheric sectionto assess the buoyancy of these columns relative to the asthenosphere, modelled as a Primitive Mantle composition. Thedensity curves show that Archean SCLM is significantly buoyant relative to the asthenosphere at depths greater thanabout 60 km. Proterozoic sections deeper than about 100 km thick also are significantly buoyant. The buoyancy ofArchean and Proterozoic SCLM sections, combined with their refractory composition, leads to high viscosities andexplains the longevity and stability of old SCLM. Replacement of Archean lithosphere, as beneath the present-dayeastern Sino^Korean craton, probably involves mechanical dispersal by rifting, accompanied by the rise of hot, fertileasthenospheric material. Fertile Phanerozoic lithosphere is buoyant when the geotherm is sufficiently high, as in manyCenozoic volcanic provinces. However, as the geothermal gradient relaxes toward a stable conductive profile,Phanerozoic SCLM sections thinner than about 100 km become denser than the asthenosphere, and hencegravitationally unstable. This could help to induce delamination of the SCLM and upwelling of asthenosphericmaterial, beginning a new cycle. The tectonic consequences of such lithosphere replacement would include uplift and

0012-821X / 01 / $ ^ see front matter ß 2001 Elsevier Science B.V. All rights reserved.PII: S 0 0 1 2 - 8 2 1 X ( 0 0 ) 0 0 3 6 2 - 9

* Corresponding author. Tel. : +61-2-9850-9673; Fax: +61-2-9850-8943; E-mail: [email protected]

EPSL 5713 4-1-01

Earth and Planetary Science Letters 184 (2001) 605^621

www.elsevier.com/locate/epsl

magmatism, and basin formation during subsequent thermal relaxation. ß 2001 Elsevier Science B.V. All rightsreserved.

Keywords: lithosphere; composition; mantle; density; geothermal gradient; xenoliths

1. Introduction

The subcontinental lithospheric mantle (SCLM)carries a geochemical, thermal and chronologicalrecord of large-scale tectonic events that haveshaped the Earth's crust. The SCLM is part ofthe continental plate, and moves with the plateover the weak asthenosphere. The idea that `old'(cratonic) lithosphere is relatively thick, depletedand cold has been long accepted by petrologists([1] and references therein). Recognition that`young' lithosphere is relatively thin, fertile andhot is more recent; Jordan [2] observed that thethermal boundary layer (TBL) beneath late-Pro-terozoic and Phanerozoic mobile belts is moresimilar to oceanic mantle than to cratonic mantlein its geophysical characteristics.

Development of the 4-D Lithosphere Mappingmethodology [1] has allowed the construction ofrealistic geological sections of the SCLM in awide variety of tectonic settings and at di¡erenttimes. Xenoliths and garnet and chromite xeno-crysts from mantle-derived volcanic rocks (e.g.basalts, lamproites, kimberlites) provide samplesof the lithospheric mantle at the time of eruption.Where su¤cient xenoliths and/or xenocrysts ofappropriate composition are available, determina-tion of the palaeogeotherm, the depth to thecrust^mantle boundary, the detailed distributionof rock types with depth within the SCLM, andthe depth to the petrological lithosphere^astheno-sphere boundary (LAB) within the tectosphere ispossible [1]. Mantle sections constructed in thisway provide a basis for the calculation of physicalproperties (such as density) of speci¢c lithospheredomains.

2. Palaeogeotherms and lithosphere thickness

2.1. Construction of palaeogeotherms

The distribution of temperature with depth atthe time of a given volcanic eruption (the palaeo-geotherm) is one of the key parameters that mustbe determined to derive the density of a columnof SCLM. Empirical palaeogeotherms are derivedby geothermobarometric calculations based onmineral chemistry, using methods with experimen-tal or theoretical calibration relevant to the min-eral assemblage, bulk composition and equilibra-tion conditions of particular samples (e.g. [3,4]).Where xenoliths with mineral assemblages appro-priate for calculation of both pressure and tem-perature of equilibration are available, severalgeothermobarometers commonly give concordantresults for a xenolith suite, even though resultsfrom the application of di¡erent techniques to asingle sample may show a high variance (e.g.[5,6]).

The limited geographic and temporal distribu-tion of xenolith suites is a major problem in thisapproach, and in many suites (especially thosefrom basalts) only a small number of xenolithsin a suite may have appropriate mineral assem-blages. The situation has been improved by thedevelopment of single-element thermometers (Niin garnet, Zn in chromite) based on partitioningbetween these phases with mantle olivine, and amethodology for derivation of geotherm parame-ters from suites of garnet and spinel xenocrysts[6]. Where suites of both xenoliths and xenocrystsare available, the two approaches give concordantresults ([6,7]).

Surface heat-£ow values and xenolith-derivedmantle geotherms are commonly linked throughassumptions such as those made in the downwardextrapolation of temperature from surface heat£ow by Pollack and Chapman [8]. However, theseassumptions include poorly constrained models of

EPSL 5713 4-1-01

Y.H. Poudjom Djomani et al. / Earth and Planetary Science Letters 184 (2001) 605^621606

the distributions of thermal conductivity and ra-diogenic heat production with depth which arelaterally variable. In general, surface heat £ow isa function of both the geotherm at depth andupper crustal heat production, and the mantlegeotherm and surface heat £ow may be onlyweakly linked. In the Pollack and Chapman mod-el [8], the mantle geotherm is directly linked to thesurface heat £ow by assuming that 60% of thesurface heat £ow is derived from below the uppercrust. We use this model here to illustrate a roughglobal average surface heat £ow equivalent to themantle geotherms that we model, not an attemptto model any particular local geotherm. In mostcratonic areas, and in Phanerozoic regions with-out active Neogene volcanism, palaeogeothermsderived from both xenoliths and xenocrysts tendto parallel theoretically constructed model con-ductive geotherms such as those of Pollack andChapman [8]. These geotherms are strongly mod-el-dependent, involving assumptions that includethe distribution of heat production and thermalconductivity with depth and the variation of ther-

mal conductivity with temperature. The same sur-face heat £ow can result from a range of heat-production distributions, and similar lithosphericmantle geotherms may give rise to di¡erent sur-face heat £ows, depending on the structure andcomposition of the crust (e.g. [9,10]). However,the model geotherms of Pollack and Chapman[8], which are parameterised in terms of the sur-face heat £ow, provide a convenient referenceframe for comparison of the xenolith and xeno-cryst data between di¡erent areas. We have con-structed, or collected from the literature, xenolith/xenocryst-based palaeogeotherms for more than300 localities worldwide, and classi¢ed these bytheir tectonothermal age (i.e. the age of the lastmajor thermal event in the crust [1]). These em-pirically determined palaeogeotherms are typicallylow beneath Archean cratonic areas, higher be-neath Proterozoic regions, and still higher beneathareas of Phanerozoic tectonic activity, corre-sponding broadly to global variations in surfaceheat £ow ([9] ; Fig. 1). Furthermore, the observeddi¡erences in lithospheric mantle temperature are

Fig. 1. (A) The range of P^T conditions commonly derived from xenolith and xenocryst suites in volcanic rocks that penetratecrust of di¡erent tectonothermal age (crustal age domains modi¢ed from Janse [15]: Archean, v 2.5 Ga; Proterozoic, 2.5^1.0Ga; Phanerozoic, 6 1 Ga). Also shown is the range of depths to the LAB typical of each group, as de¢ned by the maximumdepth of Y-depleted garnets beneath Archean and Proterozoic cratons (Ryan et al. [6]), and the maximum depth of sampling inPhanerozoic regions. Reference geotherms are conductive models of Pollack and Chapman [8], labelled with corresponding sur-face heat £ow in mW/m2, and the southeastern Australia advective geotherm (SEA) derived from xenoliths in basalts [12].(B) Variation of surface heat £ow measurements with tectonothermal age of the crust (after Morgan [9]). The crosses at the centreof the boxes represent the mean heat £ow for each age group; note the general correlation with the range of geotherms in (A).

EPSL 5713 4-1-01

Y.H. Poudjom Djomani et al. / Earth and Planetary Science Letters 184 (2001) 605^621 607

essentially those expected from a constant mantleheat £ow, but variable crustal heat production([9,11]). In areas of basaltic volcanism, such aseastern Australia and eastern China, xenolithsand xenocrysts in basalts generally record high,strongly convex-upward geotherms, consistentwith advective heat transport by magmas andunderplating of basaltic magmas in the upperpart of the lithospheric mantle ([12]).

2.2. Estimation of lithosphere thickness

Once the palaeogeotherm has been estimatedfrom xenolith/xenocryst data or heat-£ow model-ling, temperature estimates for individual samplescan be projected to this palaeogeotherm to deter-mine their depth of origin. This approach hasproved especially fruitful for use with xenocrystsuites, because statistically meaningful numbersof samples can be gathered for each section,something which often is di¤cult to achievewith xenoliths alone. By determining the NickelTemperature (TNi) of individual garnet grains in aheavy-mineral concentrate, and the Zinc Temper-ature (TZn) of individual chromites, the geochem-

ical information contained in each grain can beplaced in stratigraphic context, to map the verticaldistribution of rock types and geochemical pro-cess signatures ([6,13]).

The resulting lithological/geochemical columns(Fig. 2) provide 1-D maps, equivalent to drill holelogs, through individual lithospheric sections.Gri¤n and Ryan [13] and Ryan et al. [6] haveshown that beneath many volcanic provincesthere is a sharply de¢ned maximum depth atwhich garnets depleted in yttrium (a lithospheresignature) are common (Fig. 3). They have arguedthat this depth represents the base of the petro-logically de¢ned lithosphere, and thus corre-sponds to the LAB. In most localities worldwide,this petrologically de¢ned LAB corresponds totemperatures of 1250^1300³C [6]. The depth ofthe LAB mapped in this way varies broadlywith tectonic setting, being deepest (250^150 km)beneath Archean terrains (e.g. [7,14]) and shallow-est beneath Phanerozoic terrains (Fig. 1), aswould be expected from the observed variationin the palaeogeotherms beneath terrains of di¡er-ent age (Fig. 1a). Where high-quality deep seismicdata are available, this LAB commonly corre-

Fig. 2. Lithologic sections through the lithospheric mantle of several Archean cratons, constructed using the compositions andcalculated temperatures of garnet xenocrysts ([47,7,14]). The grey shading represents the asthenosphere. Note that for the Kaap-vaal craton, the two time slices are shown representing mantle xenoliths sampled by kimberlites intruded before and after 90 Ma.

EPSL 5713 4-1-01


sponds to the L (Lehman) discontinuity in craton-ic areas (e.g. [1]). Various geophysical and geo-chemical de¢nitions of the nature of the LABhave been discussed in ([1] and references therein).These di¡erent estimates converge at similardepths and may indicate the interdependence ofthermal state, rheology and phase relationships indetermining the location of this boundary in dif-ferent sections.

3. Secular variation of lithosphere composition

Isotopic studies, especially of the Re^Os sys-tem, show that the lithospheric mantle beneathArchean cratons ( = 2.5 Ga, as de¢ned by theage of the last major tectonothermal event to af-fect the crust; [15]) is largely Archean in age,while that beneath Proterozoic crust is largelyProterozoic in age ([16^19]). These studies providea priori evidence that the formation of the crustand the SCLM are linked processes, and suggestthat a detailed analysis of the tectonothermal his-tory of a crustal volume provides a basis for esti-mating the formation age of the underlyingSCLM.

Boyd ([20,21]) recognised a fundamental dis-tinction between Archean cratonic SCLM, repre-sented by xenoliths of peridotitic mantle from

African and Siberian kimberlites, and Phanero-zoic circumcratonic mantle, represented by xeno-liths of peridotitic mantle in intraplate basalts andby orogenic lherzolite massifs. Compared to Pha-nerozoic mantle, the peridotite xenoliths from on-craton kimberlites are not only more depleted onaverage, but have higher Si/Mg (higher opx/oli-vine). This distinctive feature is common to bothlherzolites and harzburgites, and to rocks of bothgarnet and spinel facies [22].

Analysis of a large database of xenolith compo-sitions ([23,24]) also demonstrates that Archeanxenoliths have lower Ca/Al, Fe/Al and Cr/Al ra-tios at equivalent mg# than xenoliths from Pro-terozoic and Phanerozoic regions. ArcheanSCLM is distinctive in another important respect:subcalcic (clinopyroxene-free) harzburgites arewell represented in Archean xenolith and xeno-cryst suites, but essentially absent in youngerones ([23,24]). These data clearly indicate thatthe SCLM beneath Archean cratons and youngerterranes is compositionally di¡erent. A larger da-taset, covering a larger geographic range, is pro-vided by geochemical analyses of garnet xeno-crysts (disaggregated from peridotitic mantlexenoliths) in volcanic rocks and derived alluvialconcentrations. Analysis of s 13 000 Cr^pyropegarnet xenocrysts from volcanic rocks worldwideshows a clear correlation of average garnet com-position with the tectonothermal age of the crustpenetrated by the volcanic rocks [14].

3.1. Calculating the composition of lithosphericmantle

If the composition of SCLM varies with tecto-nothermal age, the interpretation of geophysicaldata and understanding of tectonic processes re-quires that we consider di¡erences in the meancomposition of the di¡erent types of SCLM be-neath the major tectonic divisions noted above.The great bulk of the existing analytical data forArchean xenoliths comes from a small area of theKaapvaal craton in South Africa, and from a sin-gle kimberlite pipe in Siberia (Udachnaya); thereare very few xenolith data for Proterozoic ter-ranes. However, abundant data from garnet con-centrates are available for many of the world's

Fig. 3. Y^TNi (Ni temperature; [6]) plot for garnets fromkimberlites of Shandong Province, eastern China, showingthe LAB de¢ned by the deepest occurrence of Y-depletedgarnets. After Gri¤n et al. [47].

EPSL 5713 4-1-01


cratonic areas and these expand our picture ofSCLM variability.

In xenolith suites, the Cr2O3 content of garnetcorrelates well with the Al2O3 content of the hostrock ([23,24]). These suites also show good corre-lations between the content of Al2O3 and those ofother major and minor elements. Such correla-tions make it feasible to calculate the bulk com-position of a mantle section, given the medianCr2O3 content of garnet xenocrysts from that sec-tion; the technique gives good agreement with theaverage compositions of xenolith suites, wheresuitably large data sets are available (Table 1;see other examples in Gri¤n et al. [24]).

The mean composition of SCLM beneath ter-ranes of Archean, Proterozoic and Phanerozoictectonothermal age, calculated in this way, showsa secular evolution in all measures of depletion,such as Al, Ca, mg#, and Fe/Al (Table 2). Theexisting database is not su¤cient to establishwhether this secular evolution is continuous ordiscontinuous; for the purposes of this work weadopt the broad divisions recognised by Gri¤n etal. [25,26], who divide the datasets into Archean(tectonothermal age = 2.5 Ga), Proterozoic (2.5^1.0 Ga) and Phanerozoic ( = 1.0 Ga).

Unmodi¢ed Proterozoic SCLM is moderatelydepleted, and intermediate in composition be-tween Archean and Phanerozoic SCLM. Cenozoic

SCLM, exempli¢ed by the Zabargad peridotitesof the Red Sea and by garnet peridotite xenolithsfrom young extensional areas of China, Siberiaand Australia, is only mildly depleted relative toPrimitive Upper Mantle (PUM: [3,23^25]).SCLM beneath some Phanerozoic terrains, espe-cially in Europe, is more depleted and may repre-sent reworked Proterozoic SCLM (Table 2: pre-ferred values). The correlation of SCLMcomposition with crustal age is strong evidencethat crustal volumes and their underlying litho-spheric mantle formed quasi-simultaneously, andthat syngenetic crust and mantle can, and in mostcases do, remain linked for periods of aeons.

4. Calculating the density of SCLM

For each of the mean SCLM compositions de-¢ned above (Table 2), we have estimated meanmineral compositions, based on broad correla-tions between mineral and rock compositions inxenoliths ([23]). These mineral compositions havethen been used to calculate the modal composi-tions of the average Archean, Proterozoic andPhanerozoic SCLM (Table 2). We have approxi-mated the composition of the asthenosphere bythe Primitive Mantle composition of McDonoughand Sun [26], which is close to the pyrolite com-

Table 1Comparison of mean mantle compositions calculated from garnets, with average compositions of xenolith suites (after Gri¤n etal. [24])

Kaapvaal Kaapvaal Kaapvaal Kaapvaal Vitim Vitim6 90 MA 6 90 MAGarnet lherzolite Lherzolite

xenolithsGarnet harzburgite Harzburgite

xenolithsGarnet lherzolite Lherzolite

xenoliths(Calculated fromGarnets)

(Median) (Calculated fromGarnets)

(Median) (Calculated fromGarnets)

(Median)

Wt%SiO2 46.0 46.6 45.7 45.9 44.5 44.5TiO2 0.07 0.06 0.04 0.05 0.15 0.16Al2O3 1.7 1.4 0.9 1.2 3.7 4.0Cr2O3 0.40 0.35 0.26 0.27 0.40 0.37FeO 6.8 6.6 6.3 6.4 8.0 8.0MnO 0.12 0.11 0.11 0.09 0.13 0.10MgO 43.5 43.5 45.8 45.2 39.3 39.3CaO 1.0 1.0 0.5 0.5 3.3 3.2Na2O 0.12 0.10 0.06 0.09 0.26 0.32NiO 0.27 0.28 0.30 0.27 0.25 0.25

EPSL 5713 4-1-01


Tab

le2

Cal

cula

ted

mea

nco

mpo

siti

ons

for

Arc

hean

,P

rote

rozo

ican

dP

hane

rozo

icSC

LM

(aft

erG

ri¤

net

al.

[24]

),m

odes

,m

iner

alco

mpo

siti

ons,

and

sele

cted

calc

ulat

edra

-ti

osan

dco

nsta

ntva

lues

Arc

hean

Pro

tero

zoic

Pro

tero

zoic

Pha

nero

zoic

Pha

nero

zoic

Pri

mit

ive

Man

tle

Gnt

-cal

cSC

LM

Gnt

-cal

cSC

LM

Xen

olit

hs,

mas

sifs

Gnt

-cal

cSC

LM

spin

elpe

rido

tite

(pre

ferr

ed)

[26]

Wt%

SiO

245

.744

.744

.644

.544

.445

.0T

iO2

0.04

0.09

0.07

0.14

0.09

0.2

Al 2

O3

0.99

2.1

1.9

3.5

2.6

4.5

Cr 2

O3

0.28

0.42

0.40

0.40

0.40

0.38

FeO

6.4

7.9

7.9

8.0

8.2

8.1

MnO

0.11

0.13

0.12

0.13

0.13

0.14

MgO

45.5

42.4

42.6

39.8

41.1

37.8

CaO

0.59

1.9

1.7

3.1

2.5

3.6

Na 2

O0.

070.

150.

120.

240.

180.

36N

iO0.

300.

290.

260.

260.

270.

25R

atio

sm

g#92

.790

.690

.689

.989

.989

.3M

g/Si

1.49

1.42

1.42

1.33

1.38

1.25

Ca/

Al

0.55

0.80

0.80

0.82

0.85

0.73

Cr/

Cr+

Al

0.43

0.30

0.30

0.17

0.18

0.05

Fe/

Al

4.66

2.64

2.64

1.66

2.23

1.30

Mod

esol

/opx

/cpx

/gnt

69/2

5/2/

470

/15/

7/8

70/1

7/6/

760

/17/

11/1

266

/17/

9/8

57/1

3/12

/18

Min

eral

com

posi

tion

sO

livin

eF

o 93

Fo 9

1F

o 91

Fo 9

0F

o 90

Fo 9

0

Ort

hopy

roxe

neE

n 93

En 9

1E

n 91

En 9

0E

n 90

En 9

0

Clin

opyr

oxen

eaD

i 70

Hd 1

0D

i 70

Hd 1

0D

i 70

Hd 1

0D

i 70

Hd 1

0D

i 70

Hd 1

0D

i 70

Hd 1

0

Jd10

Cc 2

En 8

Jd10

Cc 2

En 8

Jd10

Cc 2

En 8

Jd10

Cc 2

En 8

Jd10

Cc 2

En 8

Jd10

Cc 2

En 8

Gar

netb

Py 7

0A

lm15

Py 7

0A

lm25

Py 7

0A

lm25

Py 7

0A

lm25

Py 7

0A

lm25

Uv 1

5U

v 5U

v 5U

v 5U

v 5D

ensi

ty(M

g/m

3)

3.31

3.34

3.34

3.37

3.36

3.39

Con

stan

tsa 0U

103

40.

2716

50.

2701

4^

0.26

970.

2776

8^

a 1U

103

81.

0497

11.

0594

5^

1.01

920.

9545

1^

a 23

0.15

031

30.

1243

^3

0.12

823

0.12

404

^k

129

130

^13

012

813

4aD

i=di

opsi

de,

Hd

=he

denb

ergi

te,

Jd=

jade

ite,

Cc

=co

smoc

hlor

e,E

n=

enst

atit

e.bP

y=

pyro

pe,

Alm

=al

man

dine

,U

v=

uvar

ovit

e.

EPSL 5713 4-1-01


position of Ringwood [27]. The assumption thatthe major-element composition of the modern as-thenosphere is only slightly depleted relative toPUM is supported by the common occurrenceof Phanerozoic xenoliths with near-PUM compo-sitions, as noted above.

The calculated modes (Table 2) illustrate theincrease in modal garnet+clinopyroxene and cli-nopyroxene/garnet from Archean to Proterozoicto Phanerozoic SCLM, and the relatively smalldegree of depletion of Phanerozoic SCLM relativeto Primitive Mantle compositions. The mean den-sity of each SCLM composition at surface tem-perature (T) and pressure (P) has then been cal-culated by combining the modes and mineralcompositions with the end-member mineral den-sities of Smyth and McCormick [28].

The major controls on density are mg# and theproportion of olivine to clinopyroxene+garnet,and these are the parameters that probably arebest constrained in these calculations. A changeof 1% (s 1c) in the mg# of olivine results in achange in density of 0.008 Mg m33. In Archeanrocks, a variation of 50% (1c ; [23]) in the abun-dance of (cpx+gnt) also produces a variation indensity of þ 0.008 Mg m33. These uncertaintieswill be correlated, because a more iron-rich oli-vine is likely to be associated with a higher fertil-ity, as expressed in higher (cpx+gnt). The total 1cuncertainty in density therefore is on the order ofþ 0.016 Mg m33. In the average Proterozoic orPhanerozoic compositions, where the mean(cpx+gnt) is higher, a þ 50% variation in (cpx+gnt) has a larger e¡ect, producing a 1c uncer-tainty of 0.02 Mg m33.

Our calculations show a signi¢cant increase inmean standard temperature and pressure (STP)density from Archean (3.31 þ 0.016 Mg m33) toProterozoic (3.34 þ 0.02 Mg m33) to Phanerozoic(3.36 þ 0.02 Mg m33) SCLM. The most depletedProterozoic SCLM overlaps only the most fertileArchean SCLM at the 1c level, while the di¡er-ence between Proterozoic and Phanerozoic SCLMis much less. The average Archean SCLM is 2.5%less dense than Primitive Mantle (3.39 Mg m33 ;Table 2) at the same temperature and pressure;for the less-depleted Phanerozoic mantle the dif-ference is about 1%. If Cenozoic lithosphere is

assumed to be similar to the more fertile Zabar-gad peridotites, or the Phanerozoic garnet perido-tites (Garnet SCLM of Table 2), the di¡erencedrops to about 0.6%.

Jordan [2] used a suite of kimberlite-bornexenoliths to derive a linear relation between nor-mative (at STP) density and mg#, given byb= 5.093^0.019 mg#. This equation reproducesour derived densities to within 0.01 Mg m33 ; weare encouraged by the agreement.

Boyd and McAllister [29] used modal estimatesand cell-volume data on separated minerals tocalculate room-T densities for a strongly depletedArchean garnet lherzolite xenolith (samplePHN1569; b= 3.30 þ 0.02 Mg m33) and a high-T-sheared peridotite xenolith (sample PHN1611; density = 3.39 þ 0.02 Mg m33). The densityestimate for PHN1569 agrees well with our calcu-lated average for Archean SCLM (Table 1). Thedensity of the sheared xenolith PHN1611 is higherthan that of our average Archean lherzolite. How-ever, the garnet and cpx content of PHN1611 hasbeen greatly increased (and its bulk mg# de-creased) by high-T melt-related metasomatismshortly prior to its entrainment in the host kim-berlite [30]. This `refertilisation' has produced abulk composition similar to some `pyrolite' com-positions, and the density of PHN1611 is identicalwithin error of our estimated density for PrimitiveMantle, and that of Jordan [31]. Jordan's [31]higher estimated density (3.35 Mg m33) for `aver-age continental garnet lherzolite' included manyhigh-T-sheared lherzolites like PHN1611. How-ever, this class of xenolith appears to be signi¢-cant only in the deepest parts of the SCLM, andis excluded from the initial analysis given below.

4.1. Variation of density with depth

Density variations in the Earth depend on thecomposition of the mantle section, and are alsoa¡ected by temperature variations, which in turna¡ect the elasticity of the minerals ; the calculateddensity at a given depth is a function of the tem-perature, the bulk thermal expansion coe¤cientand the bulk compressibility of rock, as deter-mined by the relative proportions of mineralend-members. In this section, we calculate the

EPSL 5713 4-1-01


variation in density with depth for each averagemantle type, assuming a constant compositionwith depth and using the range of geothermscharacteristic of Archean, Proterozoic and Pha-nerozoic SCLM (Fig. 1).

As noted above, the palaeogeotherms (exceptfor the southeastern Australian `SEA' geotherm(Fig. 1a)) derived from xenolith and xenocrystdata tend to parallel the model conductive geo-therms of Pollack and Chapman [8]. In this sec-tion of the paper, we therefore have used Pollackand Chapman's conductive model geotherms forcalculating T-depth relations because these geo-therms are easily parameterised.

For each lithosphere type, we assumed a two-layer model with an average crustal thickness of35 km and a total lithosphere thickness of up to280 km. For the crust, we assume a constant ther-mal conductivity of 2.5 W m31 K31. FollowingPollack and Chapman [8], we assumed that about40% of the mean £ux arises from surface radio-genic sources, and the rest comes from greaterdepths. We used this assumption to de¢ne an em-pirical equation relating the reduced heat £ow andthe mean heat £ow at depth, which then de¢nesthe distribution of heat production with depth. Inour models, we assumed a uniform characteristicdepth of heat source distribution, equal to 8 kmas de¢ned by Pollack and Chapman [8]. In thecase of the mantle, we assumed that there wasno heat production, and we used Schatz and Sim-mons' method [32] to calculate the thermal con-ductivity of olivine as a function of the temper-ature and depth. We have parameterised the SEAadvective geotherm ([12]) using a spline function.The resulting parameterised geotherms were usedto calculate the pressure and temperature-depen-dent density variations for each mantle type.

To calculate the temperature-dependent densityvariations, we used the relation:

bT b 203b 20TK 1

where b20 is the density of the section estimated ata room temperature of 20³C (Table 2), bT is thedensity of the section at a given temperature, andK, is the bulk thermal expansion coe¤cient calcu-lated from the relevant mineral end-members. A

polynomial expression of the thermal expansioncoe¤cient is given by [33] :

K a0 a1T a2T32 2

where a0, a1 and a2 are constants which are esti-mated from the mineralogical composition ofeach mantle type.

The density variations with depth/pressure arealso a function of the bulk compressibility of therocks. A change in pressure will result in a frac-tional change in the volume of a given mass ofmaterial, and this is given by the thermal dilata-tion or the compressibility L of the material,which can be related to the density by the rela-tion:

Nb bT LP 3

where bT is the density at a given temperature Tand pressure P (GPa).

Table 2 lists the mean composition, the con-stants of the thermal expansion coe¤cients (a0,a1, a2) and the bulk moduli (k = 1/L), calculatedfrom the mineral end-members (Table 2), of theaverage Archean, Proterozoic, Phanerozoic andPrimitive Mantle compositions used in our den-sity calculations.

As temperature changes with depth in theEarth, the competing e¡ects of the compressibilityand thermal expansion of the solid phases havesigni¢cant e¡ects on the density variations withdepth [34]. Therefore, we used Eq. 1 to calculatethe temperature-dependent density variation withdepth, and Eq. 3 to compute the compressibilitye¡ect on the density variations with depth. Thesetwo e¡ects are then combined to express the tem-perature and pressure-dependent density variationwith depth for Archean, Proterozoic and Phaner-ozoic mantle sections. In the Phanerozoic section,we have shown the e¡ect on density of the spinelperidotite^garnet peridotite transition at ca 55km. In the more depleted Proterozoic and Arche-an sections, the e¡ect of this phase transition isvery small (and a¡ects a narrower depth intervalof the section as the transition depth decreaseswith decreasing temperature) and has beenignored; the e¡ect is to give a maximum density

EPSL 5713 4-1-01


for these sections. The density of the astheno-sphere at the LAB has been calculated using aPrimitive Mantle composition (Table 2) (densityat 20³C and surface pressure = 3.288 Mg m33).We represent the LAB by an adiabat with a po-tential temperature of 1300³C, independent of thedi¡erent thicknesses of the lithospheric columnfor SCLM of di¡erent types. The density of theasthenospheric (Primitive Mantle) composition asa function of depth is shown in Fig. 4 as a solidheavy line. The density of the asthenosphere at220 km depth, derived from two standard Earthmodels based on seismic data (PEM and PREM;B.L.N. Kennett, pers. commun., 1998) showsgood agreement with our calculated values (Fig.4).

The results show that the density of typical Ar-chean lithospheric mantle, within a typical rangeof cratonic geotherms (35^40 mW m32 ; Fig. 1a,increases with depth at a slower rate than that ofthe asthenosphere (Fig. 4), becoming less densethan asthenosphere at about 80^100 km. TypicalProterozoic mantle is denser than the astheno-sphere at depths shallower than 145^175 km (de-

pending on the geotherm), but buoyant relative toasthenosphere at greater depths (Fig. 4). In areasof active volcanism, where advective heat trans-port by magmas raises the geotherm (e.g., theSEA geotherm; [12]), Phanerozoic mantle actuallydecreases in density with depth due to the over-riding e¡ects of thermal expansion. Where thegeotherm approximates a steady-state conductivemodel, and within the range typically observed inolder, tectonically inactive Phanerozoic terrains,the density of Phanerozoic mantle increases withdepth in the shallow spinel lherzolite stability¢eld, rises sharply at the spinel^garnet lherzolitetransition (ca 55^65 km) and then rises again withdepth in the garnet lherzolite stability ¢eld (belowca 55 km). The density increase due to the phasechange is about 3%. At depths of 160^185 km inareas with a typical Phanerozoic conductive geo-therm, the density of Phanerozoic SCLM ap-proaches that of the asthenosphere; in thinnersections the lithosphere is denser than the subja-cent asthenosphere.

To evaluate the stability of di¡erent lithospher-ic sections relative to the asthenosphere, it is use-

Fig. 4. Plots of density versus depth for lithospheric mantle sections of di¡erent age, calculated using the preferred compositionsshown in Table 2 and the range of geotherms and LAB depth shown in Fig. 1. SEA, southeastern Australia xenolith-based geo-therm, representing advective heat transport in active volcanic areas. This geotherm will decay toward the high conductive geo-therms typical of Phanerozoic areas with a time constant of W10 Ma ([8]). PEM and PREM are the values for the density ofthe asthenosphere at 220 km depth, based on seismic data and derived from these Earth models (B.L.N. Kennett, pers. commun.,1998). The density of the asthenosphere as a function of depth has been calculated using a Primitive Mantle composition and anadiabat of 0.5³C/km. The shaded areas for each mantle type and age correspond to the sections of the lithosphere that are unsta-ble at any depth relative to the underlying asthenosphere. The rapid increase in density around 55 km on the Phanerozoic curvescorresponds to a change in composition from spinel lherzolite (for depths 6 55 km) to garnet lherzolite (for depths s 55 km).The 1300³C point marking the asthenosphere temperature is shown with grey circles for each geotherm. The horizontal stripedarea tracks the locus of thermal relaxation from the high advective geotherm.

EPSL 5713 4-1-01


ful to calculate the cumulative density of eachcolumn as a function of its thickness, and to com-pare this density with that of a column of asthe-nospheric mantle whose temperature follows the1300³C adiabat (Fig. 5). These curves demon-strate that sections of Archean lithospheric mantlethicker than ca 60 km are buoyant relative to theunderlying asthenosphere. Typical Archean litho-sphere is 150^250 km thick (Fig. 1) and thereforethese sections are signi¢cantly buoyant; a 200-kmsection is 2.5% less dense than the asthenosphereat the LAB. A Proterozoic SCLM section thickerthan ca 125 km is buoyant relative to the astheno-sphere, but less so than Archean lithosphere; typ-ical sections 160 km thick are ca 1% less densethan the asthenosphere at the LAB. PhanerozoicSCLM sections with advective geotherms decreasein density with depth, and are strongly buoyantrelative to the asthenosphere. However, older sec-tions that have cooled to steady-state conductivegeotherms are buoyant relative to the astheno-sphere only if they are more than 110^120 kmthick.

Fig. 6 summarises lithospheric density pro¢lesfor each SCLM age type and the upper astheno-sphere. The geotherm used for each SCLM sec-tion is within the typical range shown in Figs. 4and 5: the asthenosphere thermal state assumes a

potential temperature of 1300³C and an adiabatof 0.5³C/km. This ¢gure emphasises the densitycontrasts between lithospheric sections of di¡erentages and the relationship between the density ofeach lithospheric section and that of the under-lying asthenosphere.

5. Discussion

5.1. Comparison with other models

Jordan [2] showed that the elevation of the con-tinents, and particularly of the cratonic areas re-quires a chemical boundary layer beneath the con-tinents that is buoyant relative to oceanic mantle.He also gave a formulation for the condition formarginal stability in a layer with such a lateralvariation in density, which gives a minimum valuefor the compositional density contrast across thelayer. One pro¢le that satis¢es this condition forArchean lithosphere is the isopycnic curve, inwhich the density of the lithospheric column ateach depth is the same beneath oceans and con-tinents [2]. The negative buoyancy produced bythe lower temperature gradient of continentallithosphere is balanced by a positive buoyancydue to a more depleted composition. This formu-

Fig. 5. Plots of cumulative density versus lithosphere thickness for lithospheric mantle sections of di¡erent age, calculated usingthe preferred compositions shown in Table 2 and the range of geotherms and LAB depth shown in Fig. 1. The shaded areas foreach mantle type and age correspond to the integrated sections of the lithosphere that are unstable relative to the underlying as-thenosphere. The thick line indicating the density of the asthenosphere as a function of depth assumes a Primitive Mantle compo-sition (Table 2) and an adiabat of 0.5³C/km for the LAB. The striped line on the Archean plot shows the density of a hotter as-thenosphere (1500³C) calculated with an adiabat of 0.5³C/km. Same abbreviations as in Fig. 4.

EPSL 5713 4-1-01


lation requires that the normative density (atSTP) of the continental section increases withdepth. The isopycnic curve, corrected for the ef-fects of T and P, would necessarily approximatethe density of the asthenosphere as shown on Fig.5. This represents a minimum buoyancy; a con-tinent with a lithospheric density pro¢le corre-sponding to the isopycnic curve would have littleor no freeboard relative to the oceans.

Jordan [2] used xenolith data to argue that thenormative density of the Archean lithosphere in-creases with depth as required by the isopycnichypothesis. A more detailed analysis based onthe study of garnet xenocrysts in volcanic rocks

from cratonic areas worldwide ([35]) con¢rms ageneral decrease in mg# with depth, implyingthat the density of the lower parts of Archeanand Proterozoic sections can approach that ofPhanerozoic lithosphere. However, this e¡ect ismost pronounced in the deeper parts of the litho-sphere (s 180 km depth) and represents a rapidincrease in density over a short vertical distance;a similar distribution of density with temperatureis seen in the data used by Jordan ([2], Fig. 3).

The cumulative density curves shown in Fig. 5,which assume a constant normative density, prob-ably give the maximum buoyancy contrast to beexpected for each type of lithospheric section. TheArchean lithosphere approximated in this wayshows signi¢cant buoyancy for sections with typ-ical Archean lithosphere thicknesses of 180^220 km, while Proterozoic sections with typicalthicknesses of 150^180 km are only moderatelybuoyant, and Phanerozoic sections (100^130 kmthick) are neutrally buoyant. The low to moderatebuoyancy of Proterozoic and Phanerozoic litho-sphere, relative to the asthenospheric mantle, sat-is¢es the minimum requirements for marginalstability in the plate. However, in contrast tothe isopycnic situation, the density^depth curvesderived here show lithospheric densities greaterthan that of the asthenosphere at shallow depths,and less than that of the asthenosphere at greaterdepths. The very low normative densities at shal-low depths implied by the isopycnic relationwould require Archean-type densities at Mohodepths beneath all sections, and there is no evi-dence in the xenolith record for this except insome Archean sections. The overall buoyancy ofArchean sections (Fig. 5) probably is a maximumestimate, if the rapid decrease in mg# at depths of180^220 km observed by Gaul et al. [35] is a gen-eral feature of Archean lithosphere, but even withthis caveat, it appears that Archean lithosphere issigni¢cantly positively buoyant with respect to theunderlying asthenosphere. Is such buoyancy com-patible with observation?

Jordan [36] argued that cratons are neutrallybuoyant relative to oceanic mantle because theyare not observable in the long-wavelength gravity¢eld. Richards and Hager [37] and Forte et al.[38] found that in fact there is a weak association

Fig. 6. Density pro¢les for each lithospheric mantle age typeand for the upper asthenosphere: Primitive Mantle composi-tion from Table 2 is used for the asthenosphere compositionwith an adiabat of 0.5³C/km. The representative average (la-belled) geotherm for each lithosphere section is within thetypical geotherm range shown in Figs. 3 and 5. The starsshow where the geotherms for each lithosphere section crossthe LAB. The average LABs taken from Figs. 4 and 5 are200, 150 and 100 km for the Archean, Proterozoic and Pha-nerozoic lithosphere, respectively.

EPSL 5713 4-1-01


between geoid perturbations and the continentalshields, while Shapiro et al. [39] argued that thisdi¡erence is not signi¢cant at the 2c level. How-ever, the wavelengths used (of necessity) in thesestudies may be too long to recognise the signi¢-cant buoyancy of Archean domains. The param-eterisations have grouped Archean and Protero-zoic shield areas, which will tend to combine largeareas of neutrally to slightly buoyant Proterozoicshields with smaller areas underlain by morebuoyant Archean keels, leading to a small totalsignal at such long wavelengths. On a smaller (re-gional) scale, analysis of upward-continued grav-ity data indicates a signi¢cant density de¢cit be-neath Archean cratons, but not beneath thesurrounding Proterozoic mobile belts (PoudjomDjomani et al., in prep.).

5.2. Can continental lithosphere `delaminate'?

The cumulative density curves in Fig. 5 providesome constraints on tectonic and geochemicalmodels that invoke the gravity-driven detachment(`delamination') of SCLM and its recycling intothe deep mantle. Our results show that a typicalsection of Archean lithospheric mantle is signi¢-cantly buoyant relative to the asthenosphere.Lowering the geotherm below the 35 mW/m2 con-ductive model could increase the density of thecolumn, but even if the geotherm were greatlydepressed, such a section of Archean SCLM isunlikely to become negatively buoyant if it isthicker than 100 km. Tectonic stacking, often in-voked as a mechanism for delamination, will sim-ply increase the buoyancy of the Archean SCLMsection relative to the asthenosphere, as the den-sity of the asthenosphere at the depressed LABincreases faster than that of the thickened litho-sphere, assuming constant composition. This typeof lithospheric mantle therefore cannot be delami-nated through gravitational forces alone.

The compositional buoyancy of ArcheanSCLM also acts to stabilise the TBL against con-vective disruption by reducing the stress, whichleads in turn to a reduction in viscosity and anincrease in strength. Such compositional buoy-ancy appears to be required by £ow models thatinvoke activation energies less than those required

for the £ow of olivine [40]. Finally, the refractorynature and low heat production of the Archeanlithosphere reduces the probability of its beingweakened by partial melting, even when subjectedto relatively large temperature perturbations. Thecombined e¡ects of their highly depleted compo-sitions on buoyancy, strength and resistance tomelting provide a simple explanation for thethickness and apparent long-term stability of Ar-chean lithospheric roots.

Proterozoic SCLM, while denser than ArcheanSCLM at the same temperature because of its lessdepleted composition, also typically has some-what higher geotherms (Fig. 1). Any Proterozoicsection more than about 150^180 km thick ismoderately buoyant (Fig. 5), consistent with theobserved preservation of lithosphere with Prote-rozoic Re^Os ages beneath Proterozoic cratons(e.g. [16^19]) and beneath some areas where Pha-nerozoic tectonic (but not magmatic) activity hasreworked Proterozoic crust, such as the Caledo-nides of western Norway ([41]) and parts of west-ern Europe ([23]). Because the density contrastbetween asthenosphere and Proterozoic SCLMat the LAB is less than for Archean SCLM, it ispossible that a decrease in the geotherm belowthose modelled here could destabilise a sectionas thick as 150 km.

Phanerozoic terranes have a wide range of geo-therms (Fig. 1). With the high advective geo-therms, typical of areas of active or recent basalticvolcanism (e.g. [12,42]), Phanerozoic SCLM is ex-tremely buoyant. Some Phanerozoic SCLM sec-tions may sustain relatively high conductive geo-therms because they have high internal heatproduction. However, with lower geotherms ob-served in older Phanerozoic areas (45^50 mW/m2

conductive models ; Figs. 1, 4 and 5), a section ofPhanerozoic SCLM thinner than about 110 km isdenser than the asthenosphere at the LAB, andhence is gravitationally unstable. This is a mini-mum estimate: if the denser Cenozoic lithosphericmantle found as garnet peridotites in some re-gions (e.g. eastern China; [3]) is considered, sec-tions up to about 150 km thick are unstable evenon the higher conductive geotherms characteristicof many Phanerozoic terranes.

Lithosphere delamination, accompanied by the

EPSL 5713 4-1-01


upwelling of the asthenosphere to shallow depths,is widely invoked to explain tectonic uplift andcrustal magmatism and metamorphism. The datapresented here suggest that this mechanism is un-likely to have been relevant during the growth ofArchean cratons, unless the lithospheric roots ofthese cratons formed after the major magmatic/metamorphic episodes. However, available Re^Os data suggest that such roots are at least asold as the overlying crust ([16] and referencestherein).

On the other hand, the density^depth system-atics of Phanerozoic SCLM suggest that such de-lamination is highly likely as a result of coolingfollowing crustal formation.

5.3. `Lithosphere erosion' ^ mechanisms and e¡ects

The data presented here suggest that tectonic ormagmatic events that lead to the replacement ofold SCLM by younger material will cause changesin the density and geothermal pro¢le of the litho-spheric column, with major e¡ects at the surface.These e¡ects will include regional uplift due to thelower density of material on an elevated geotherm[43], followed by subsidence as the elevated geo-therm relaxes [44] and the overall density of thesection increases due to the replacement of olderbuoyant mantle with younger, denser mantle.

Several studies have documented situations inwhich thick, depleted (Archean) lithospheric man-tle has been wholly or partially removed and re-placed by thinner, hotter and more fertile SCLM.Eggler et al. [45] showed that beneath the Wyom-ing craton, an original Archean lithosphere about200 km thick, with a low conductive geotherm,has been largely replaced by fertile CenozoicSCLM with a thickness of about 120 km. In theKaapvaal Craton, Brown et al. [46] used ¢ssiontrack dating and xenolithic material from kimber-lites of di¡erent ages to show that the removal ofca 40 km of old, depleted lithospheric mantle andits replacement by hotter and chemically re-charged (metasomatised) lithosphere around 90Ma ago, is correlated with signi¢cant uplift anderosion of the craton. In the eastern Sino^Koreancraton, China, the removal of about 100 km ofArchean lithosphere during the late Mesozoic was

accompanied by uplift, basin formation and wide-spread magmatism [47].

If, as argued above, Archean SCLM cannot be`delaminated' and is too refractory to be meltedsigni¢cantly, how does such lithosphere replace-ment occur? A detailed discussion of possible sce-narios is outside the scope of this paper. However,at least one mechanism is suggested by geologicaland geophysical studies of the Sino^Korean cra-ton. Detailed seismic tomography of part of theeastern Sino^Korean craton, an area of very highheat £ow, shows a lithospheric mantle made up ofvertically and laterally extensive blocks of high-velocity (probably Archean) mantle embedded ina matrix of lower-velocity (presumably Cenozoic)mantle [48]. Yuan [48] suggests that this heteroge-neity re£ects the disruption of the Archean litho-sphere due to rifting processes that allowed up-welling of young fertile mantle along breaks in theArchean root. This mechanism has resulted inmechanical dispersal, heating and `dilution' ofthe Archean root, rather than its removal. Wesuggest that the prolongation of this processwould involve heating and metasomatism of therelict Archean blocks by asthenospheric material,leading to a progressive increase in normativedensity and decrease in their buoyancy and vis-cosity, to the point where they would be di¤cultto distinguish from young lithosphere, except pos-sibly by isotopic techniques.

5.4. Implications for lithosphere generationprocesses

The material of Archean SCLM is so buoyantrelative to asthenosphere that even relatively smallvolumes would rise and accumulate, especially ifthey were generated in high-temperature events[49] so that their density would be lower thanmodelled in Figs. 4 and 5. Fig. 5 shows the e¡ectof a hotter asthenosphere in the Archean, mod-elled as an adiabat with a potential temperatureof 1500³C. Even in this situation, Archean litho-sphere thicker than ca 75 km would be signi¢-cantly buoyant. This buoyancy o¡ers a mecha-nism for generating Archean protocontinentsfrom the depleted residues of small-scale events.Indeed, it would be di¤cult not to accumulate

EPSL 5713 4-1-01


such material, and the current lack of Re^Os de-pletion ages greater than ca 3.5 Ga for mantlematerials may imply that these lithosphere-form-ing processes did not begin earlier.

The relatively low density of a hotter astheno-sphere could have led to the preferential preserva-tion of only the least dense lithosphere sections,thus exaggerating the apparent contrast in norma-tive density between Archean and younger litho-sphere. However, if this density-sorting mecha-nism was the primary cause, we would expect to¢nd Archean-type material in younger lithospheresections, and this has not been identi¢ed to date([24]). Thus even if only the most buoyant parts ofthe Archean lithosphere have survived, the di¡er-ences between Archean and younger SCLM arguefor a secular change in the processes that havegenerated SCLM through time. Discussions ofthe nature of these processes are given elsewhere[25,26,49].

Proterozoic lithospheric sections are typically150^180 km thick, but are only gravitationallystable if they are 130 km thick. These thick sec-tions would have to have accumulated either rap-idly, or at high enough temperatures to remainbuoyant until they have thickened to the pointwhere thermal relaxation would not make themunstable. This accumulation would be assistedby secular cooling of the Earth, as the cooler as-thenosphere became more dense. The processesthat produced these relatively thick stable rootsappear to be largely con¢ned Archean and Prote-rozoic time, though present-day analogues mayexist under regions such as the Sierra Nevada ofCalifornia.

A section of Phanerozoic SCLM cooling from ahigh advective geotherm (such as the SEA geo-therm (Figs. 4 and 5) toward a typical high con-ductive geotherm undergoes a dramatic change indensity (s 2% for a section 100 km deep). Whilethis section is very buoyant on the high advectivegeotherm it is gravitationally unstable and subjectto delamination (Rayleigh^Taylor instability ; [50]and references therein) on lower geotherms. Theremoval of such a section, and its replacement byupwelling asthenospheric material, would producea new advective geotherm, starting the cycle overagain. This instability may be a common feature

of Phanerozoic orogenic belts, and could explainthe apparent scarcity of oceanic or island^arc de-pleted mantle components in xenolith suites fromyoung mobile belts such as eastern China andeastern Australia, as documented by Gri¤n etal. [23,24].

6. Conclusions

Di¡erences in SCLM composition and geo-therms between Archean, Proterozoic and Pha-nerozoic regions result in marked, age-related dif-ferences in mean lithospheric density and in thedensity^depth relationship within the SCLM.

Archean SCLM sections (depth to base oflithosphere 180^240 km) are signi¢cantly buoyantrelative to asthenosphere under any reasonablegeological scenario, and this buoyancy reducesboth stress and viscosity. Therefore ArcheanSCLM cannot be delaminated by gravitationalprocesses alone, and will tend to be preserved.Tectonic processes such as rifting may allow itsdisruption and replacement by upwelling, morefertile asthenospheric (or plume) material.

Proterozoic SCLM is less depleted and there-fore denser (at the same T) than ArcheanSCLM. Thin Proterozoic SCLM (less than about120 km thick) is denser than asthenosphere andwould be gravitationally unstable. However, Pro-terozoic lithospheric sections are typically 150^180 km thick; such SCLM columns are moder-ately buoyant and, like Archean SCLM, are un-likely to be delaminated.

Phanerozoic lithospheric sections are com-monly less than about 100 km thick. These sec-tions are stable under the elevated geotherms as-sociated with volcanic activity, but gravitationallyunstable once they have cooled to typical steady-state conductive geotherms. This instability sug-gests that lithospheric delamination may be acommon feature of the history of Phanerozoicmobile belts. Replacement of such a section byupwelling asthenospheric material will start an-other cycle of cooling, instability and delamina-tion. This process may explain the common oc-currence of highly fertile SCLM beneathPhanerozoic mobile belts, where we might instead

EPSL 5713 4-1-01


expect to ¢nd depleted oceanic and arc-relatedperidotites.

Acknowledgements

The concepts discussed here owe much to dis-cussions with many colleagues, including JoeBoyd, Rod Brown, Geo¡ Davies, Buddy Doyle,Kevin Kivi, Lev Natapov, Norm Pearson, ChrisRyan, Yuan Xuecheng and Ming Zhang. Wethank Brian Kennett for providing his indepen-dent estimates of asthenosphere density at 220km and A. Lenardic for helpful comments onan earlier version of this manuscript. Commentsby I. Artemieva and an anonymous reviewer haveclari¢ed issues in the manuscript. Funding for thiswork has come from several sources, includingARC Large Grants (S.Y.O'R. and W.L.G.), Mac-quarie University Industry Collaborative Grants,MURG (Y.H.P.D., S.Y.O'R. and W.L.G.), Aus-AID grants and the mineral exploration industry.This is contribution no 228 from the ARC Na-tional Key Centre for Geochemical Evolution andMetallogeny of Continents (www.es.mq.edu.au/gemoc/).[AC]

References

[1] S.Y. O'Reilly, W.L. Gri¤n, 4-D lithospheric mapping: areview of the methodology with examples, Tectonophysics262 (1996) 3^18.

[2] T.H. Jordan, Structure and formation of the continentaltectosphere, J. Petrol. Special lithosphere issue (1988) 11^37.

[3] X. Xu, S.Y. O'Reilly, W.L. Gri¤n, X. Zhou, X. Huang,The nature of the Cenozoic lithosphere at Nushan, east-ern China, in: M. Flower, S.L. Chun, C.H. Lo, T.Y. Lee(Eds.), Mantle Dynamics and Plate Interactions in EastAsia, Geodynamics Series 27, Am. Geophys. Union,Washington, DC, 1998, pp. 167^196.

[4] D. Smith, Temperatures and pressures of mineral equili-bration in peridotite xenoliths: review, discussion and im-plications, in: Y. Fei, C.M. Bertka, B.O. Mysen (Eds.),Mantle Petrology: Field Observations and High-pressureExperimentation: A Tribute to Francis R. (Joe) Boyd,Geochemical Society Special Publication #6, The Geo-chemical Society, Houston, TX, 1999, pp. 171^188.

[5] A.A. Finnerty, F.R. Boyd, Thermobarometry for garnetperidotite xenoliths: a basis for mantle stratigraphy, in:

P.H. Nixon (Ed.), Mantle Xenoliths, John Wiley andSons, New York, 1987, pp. 381^402.

[6] R.G. Ryan, W.L. Gri¤n, N.J. Pearson, Garnet Geo-therms: a technique for derivation of P^T data fromCr^pyrope garnets, J. Geophys. Res. 101 (1996) 5611^5625.

[7] W.L. Gri¤n, B.J. Doyle, C.G. Ryan, N.J. Pearson, S.Y.O'Reilly, R.M. Davies, K. Kivi, E. Van Achterbergh,L.M. Natapov, Layered mantle lithosphere in the Lacde Gras area, Slave Craton: composition, structure andorigin, J. Petrol. 40 (1999) 705^727.

[8] H.N. Pollack, D.S. Chapman, On the regional variationof heat £ow, geotherms and lithosphere thickness, Tecto-nophysics 38 (1977) 279^296.

[9] P. Morgan, The thermal structure and thermal evolutionof the continental lithosphere, Phys. Chem. Earth 15(1984) 107^193.

[10] R.L. Rudnick, W.F. McDonough, R.J. O'Connell, Ther-mal structure, thickness and composition of continentallithosphere, Chem. Geol. 145 (1998) 395^411.

[11] C. Jaupart, J.C. Mareschal, The thermal structure andthickness of continental roots, Lithos 48 (1999) 93^114.

[12] S.Y. O'Reilly, W.L. Gri¤n, A xenolith-derived geothermfor southeastern Australia and its geophysical implica-tions, Tectonophysics 111 (1985) 41^63.

[13] W.L. Gri¤n, C.G. Ryan, Trace elements in indicator min-erals: Area selection and target evolution in diamond ex-ploration, J. Geochem. Explor. 53 (1995) 311^337.

[14] W.L. Gri¤n, F.V. Kaminsky, C.G. Ryan, S.Y. O'Reilly,L.M. Natapov, I.P. Ilupin, The Siberian Lithosphere Tra-verse Mantle terranes and the assembly of the Siberiancraton, Tectonophysics 310 (1999) 1^35.

[15] A.J.A. Janse, Is Cli¡ord's rule still valid? A¤rmative ex-amples from around the world, in: H.O.A. Meyer, O.Leonardos (Eds.), Diamonds: Characterization, Genesisand Exploration, CPRM Spec. Publi. 1A/93, Dept. Na-tional da Prod. Mineral., Brazilia, 1994, pp. 215^235.

[16] R. Carlson, D.G. Pearson, F.R. Boyd, S.B. Shirey, G.Irvine, A.H. Menzies, J.J. Gurney, Re^Os systematics oflithospheric peridotites: Implications for lithosphere for-mation and preservation, vol. 1, Proc. 7th Int. Kimb.Conf., 1999, pp. 99^108.

[17] J.S. McBride, D.D. Lambert, A. Greig, I.A. Nicholls,Multistage evolution of Australian subcontinental mantleRe^Os isotopic constraints from Victorian mantle xeno-liths, Geology 24 (1996) 631^634.

[18] M.R. Handler, V.C. Bennett, T.M. Esat, The persistenceof o¡-craton lithospheric mantle: Os isotopic systematicsof variably metasomatised southeast Australian xenolith,Earth Planet. Sci. Lett. 151 (1997) 61^75.

[19] B.G. Hoal, K.E.O. Hoal, F.R. Boyd, D.G. Pearson, Ageconstraints on crustal and mantle lithosphere beneath theGibeon kimberlite ¢eld, Namibia, S. Afr. Tydskr. Geol.98 (1995) 112^118.

[20] F.R. Boyd, Composition and distinction between oceanicand cratonic lithosphere, Earth Planet. Sci. Lett. 96 (1989)15^26.

EPSL 5713 4-1-01


[21] F.R. Boyd, Origin of peridotite xenoliths: major elementconsiderations, in: G. Ranalli et al. (Eds.), High Pressureand High Temperature Research on Lithosphere andMantle Materials, University of Siena, 1997.

[22] F.R. Boyd, N.P. Pokhilenko, D.G. Pearson, S.A. Mertz-man, N.V. Sobolev, L.W. Finger, Composition of theSiberian cratonic mantle: evidence from Udachnaya peri-dotite xenoliths, Contrib. Mineral. Petrol. 128 (1997) 228^246.

[23] W.L. Gri¤n, S.Y. O'Reilly, C.G. Ryan, O. Gaul, D. Ion-ov, Secular variation in the composition of subcontinentallithospheric mantle, in: J. Braun et al. (Eds.), Structureand Evolution of the Australian Continent, GeodynamicsSeries, Am. Geophys. Union, vol. 26, 1998, pp. 1^25.

[24] W.L. Gri¤n, S.Y. O'Reilly, C.G. Ryan, The compositionand origin of subcontinental lithospheric mantle, in: Y.Fei, C.M. Bertka, B.O. Mysen (Eds.), Mantle Petrology:Field Observations and High-pressure Experimentation:A Tribute to Francis R. (Joe) Boyd, Geochemical SocietySpecial Publication #6, The Geochemical Society, Hous-ton, TX, 1999, pp. 13^45.

[25] D.A. Ionov, I.V. Ashchepkov, H.-G. Stosch, G. Witt-Eickschen, H.A. Seck, Garnet peridotite xenoliths fromthe Vitim volcanic ¢eld, Baikal region: the nature of thegarnet^spinel transition zone in the continental mantle,J. Petrol. 34 (1993) 1141^1175.

[26] W.F. McDonough, S. Sun, The composition of the Earth,Chem. Geol. 120 (1995) 223^253.

[27] A.E. Ringwood, The chemical composition and origin ofthe Earth, in: P.M. Hurley (Ed.), Advances in Earth Sci-ence, MIT Press, Cambridge, MA, 1996, pp. 287^356.

[28] J.R. Smyth, T.C. McCormick, Crystallographic data forminerals, in: T.J. Ahrens (Ed.), Mineral Physics and Crys-tallography: A Handbook of Physical Constants, Am.Geophys. Union, Washington, DC, 1995, pp. 1^17.

[29] F.R. Boyd, R.H. McAllister, Densities of fertile and ster-ile garnet peridotites, Geophys. Res. Lett. 3 (1976) 509^512.

[30] D. Smith, W.L. Gri¤n, C.G. Ryan, Compositional evo-lution of high-temperature sheared lherzolite PHN1611,Geochim. Cosmochim. Acta 57 (1993) 605^613.

[31] T.H. Jordan, Mineralogies, densities and seismic velocitiesof garnet lherzolites and their geophysical implications,in: F.R. Boyd, H.O.A. Meyer (Eds.), The Mantle Sample:Inclusions in Kimberlites and Other Volcanics, Am. Geo-phys. Union, Washington, DC, 1979, pp. 1^14.

[32] J.F. Schatz, G. Simmons, Thermal conductivity of Earthmaterials at high temperatures, J. Geophys. Res. 77 (1972)6966^6983.

[33] Y. Fei, Thermal expansion, in: T.J. Ahrens (Ed.), MineralPhysics and Crystallography: A Handbook of PhysicalConstants, Am. Geophys. Union, Washington, DC,1995, pp. 29^44.

[34] F. Cella, A. Rapolla, Density changes in upwelling man-tle, Phys. Earth Planet. Int. 103 (1997) 63^84.

[35] O.F. Gaul, W.L. Gri¤n, S.Y. O'Reilly, N.J. Pearson,Mapping olivine composition in the lithospheric mantle,Earth Planet. Sci. Lett. 182 (2000) 223^235.

[36] T.H. Jordan, The continental tectosphere, Rev. Geophys.Space Phys. 13 (1975) 1^12.

[37] M.A. Richards, B.H. Hager, The Earth's geoid and thelarge-scale structure of mantle convection, in: S.K. Run-corn (Ed.), The Physics of Planets, Wiley, New York,1988, pp. 247^272.

[38] A.M. Forte, A.M. Dziewonski, R.J. O'Connell, Conti-nent^Ocean chemical heterogeneity in the mantle basedon seismic tomography, Science 268 (1995) 386^388.

[39] S.S. Shapiro, B.H. Hager, T.H. Jordan, The continentaltectosphere and the Earth's long-wavelength gravity ¢eld,Lithos 48 (1999) 135^152.

[40] S.S. Shapiro, B.H. Hager, T.H. Jordan, Stability and dy-namics of the continental tectosphere, Lithos 48 (1999)115^133.

[41] B. Jamtveit, D.A. Carswell, E.W. Mearns, Chronology ofthe high-pressure metamorphism of Norwegian garnetperidotites/pyroxenites, J. Metamorph. Geol. 9 (1991)125^139.

[42] A.P. Jones, J.V. Smith, J.B. Dawson, E.C. Hansen, Meta-morphism, partial melting and K metasomatism of gar-net^scapolite^kyanite granulite xenoliths from Lashaine,Tanzanian J. Geol. 91 (1982) 143^165.

[43] P. Morgan, Constraints on rift thermal processes fromheat £ow and uplift, Tectonophysics 94 (1983) 277^298.

[44] D.P. McKenzie, Some remarks on the development ofsedimentary basins, Earth Planet. Sci. Lett. 40 (1978)25^32.

[45] D.H. Eggler, J.K. Meen, R. Welt, F.O. Dudas, K.P. Fur-long, M.E. McCallum, R.W. Carlson, Tectonomagmatismof the Wyoming Province, Colo. Sch. Mines Q. 83 (1988)25^40.

[46] R.W. Brown, K. Gallagher, W.L. Gri¤n, C.G. Ryan,M.C.J. De Wit, D.X. Belton, R. Harman, Kimberlites,accelerated erosion and evolution of the lithospheric man-tle beneath the Kaapvaal craton during the mid-Creta-ceous, Ext. Abst. 7th Int. Kimb. Conf., 1999, pp. 105^107.

[47] W.L. Gri¤n, A. Zhang, S.Y. O'Reilly, C.G. Ryan, Pha-nerozoic evolution of the lithosphere beneath the Sino^Korean Craton, in: M. Flower et al. (Eds.), Mantle Dy-namics and Plate Interactions in East Asia, GeodynamicsSeries, vol. 27, Am. Geophys. Union, Washington, DC,1998, pp. 107^126.

[48] X. Yuan, Velocity structure of the Qinling lithosphere andmushroom cloud model, Sci. China Ser. D 39 (1996) 235^244.

[49] C. Herzberg, Formation of cratonic mantle as plume res-idues and cumulates, in: Y.Fei, C.M. Bertka and B.O.Mysen (Eds.), Mantle Petrology: Field Observations andHigh-pressure Experimentation: A Tribute to Francis R.(Joe) Boyd, Geochemical Society Special Publication #6,The Geochemical Society, Houston, TX, 1999, pp. 241^258.

[50] G.A. Houseman, P. Molnar, Gravitational (Rayleigh^Taylor) instability of a layer with non-linear viscosityand convective thinning of continental lithosphere, Geo-phys. J. Int. 128 (1997) 125^150.

EPSL 5713 4-1-01


the density structure of subcontinental lithosphere through time...

Documents