q 2004 Geological Society of America. For permission to copy, contact Copyright Permissions, GSA, or editing@geosociety.org.Geology; September 2004; v. 32; no. 9; p. 765–768; doi: 10.1130/G20590.1; 2 figures. 765

Inhibited eclogite formation: The key to the rapid growth of strongand buoyant Archean continental crust

M.G. BjørnerudGeology Department, Lawrence University, Appleton, Wisconsin 54912, USA

H. AustrheimPhysics of Geological Processes Project and Department of Earth Sciences, Postbox 1047, University of Oslo,

Oslo N-0316, Norway

ABSTRACTSubduction is the principal mechanism by which the hydrosphere and interior of

Earth interact. Today, subduction involves the dehydration of ocean crust at depths of60–120 km depending on the age of the slab. Release of the water leads to generation ofarc magmas (future continental crust), and the slab is then transformed into denser eclo-gite that helps to pull more of the slab into the trench. However, it is unlikely that thefirst continental crust formed this way. Growing geochemical evidence indicates that largevolumes of continental crust were produced over a short period of time in the Archean,when the planet was probably too hot for modern plate tectonics to operate. A significantincrease in the kinetics of eclogite-forming reactions may have been the key to the tran-sition from Archean to modern tectonics. Under the higher geothermal gradients of theArchean, tectonically buried ocean crust would have been severely dehydrated beforereaching eclogite facies pressures. Because rapid eclogitization is dependent on water asa medium for advective ion transport, the very shallow dehydration in the Archean mayhave inhibited the formation of eclogite facies minerals. The importance of water in eclo-gite metamorphism is illustrated by a complex of partly eclogitized mafic granulites inHolsnøy, western Norway, in which reaction progress was limited by the availability ofwater. When water is scarce or absent, metastable granulite facies mineral assemblagescan persist at eclogite facies depths owing to the extremely slow reaction kinetics whendiffusion is the only chemical transport mechanism. Such dehydrated but uneclogitizedmafic crust would have been very strong and too buoyant to sink into the mantle, and itmay have formed the substrate for the first continental lithosphere.

Keywords: Archean tectonics, continental crust formation, eclogite, slab dehydration, kineticsof metamorphic reactions.

ARCHEAN TECTONICSTwo of the most fundamental unresolved questions about the solid

phase Earth are (1) when subduction-dominated plate tectonics beganand (2) how—and how early in the planet’s history—the first signifi-cant volumes of continental crust formed. These questions are closelyrelated. New continental crust forms principally in arc settings where,at depths of 60–120 km, water is released from hydrous minerals insubducting ocean slabs and acts as a flux for the overlying mantlewedge, producing low-temperature melts. The slab is eventually con-verted to dense eclogite, which then helps to pull more ocean crustdown into the mantle for dehydration and transformation. Eclogite for-mation is arguably the essential component in Earth’s self-sustainingtectonic system, setting the pace for subduction and preventing thecrust from becoming too thick (Anderson, 1989). However, this mech-anism for incremental production of continental crust does not explainthe growing evidence for a burst of crustal formation in Middle Ar-chean time, or possibly earlier (Bowring and Housh, 1995; Wilde etal., 2001; Hamilton, 2003), nor is it consistent with the higher radio-genic heat production and higher geothermal gradients in ArcheanEarth, which would probably have precluded modern-style subduction(Davies, 1992; Vlaar et al., 1994; Hamilton, 2003; Martin and Moyen,2002). That is, most of Earth’s original continental crust appears tohave formed long before the onset of true plate tectonics (Rudnick,1995). The key to both the early formation of continental crust and the

change to subduction-dominated tectonics may have been an acceler-ation in the rates of eclogite-forming reactions, related to the coolingof Earth and associated changes in the extent and depth of slab de-hydration near the Archean-Proterozoic boundary.

In modern subduction zones, the dehydration of downgoing oceancrust is dependent on the manner in which a subducting slab ‘‘drags’’its surface isotherms with it as it descends. The depths at which theslab begins to dehydrate and later recrystallize to eclogite are artifactsof Earth’s modern thermal regime. In Archean time, when Earth’s geo-thermal gradient was higher, the physics of the process would havebeen quite different. First, mantle convection would have been morevigorous, and slabs would have been younger, hotter, and thicker thanthey are today upon arrival at a subduction zone (or the Archean equiv-alent) (Davies, 1992). Second, a subducted or tectonically buried slabwould have encountered significantly higher temperatures at any givendepth than is true today. Geophysical models and the geochemistry ofArchean peridotites and komatiites suggest that the upper mantle was200–400 8C hotter than present-day asthenosphere (Hamilton, 2003;Breuer and Spohn, 1995; Herzberg, 1993). Under these higher Archeantemperatures, ocean crust being buried via protosubduction or crustalthickening (Fig. 1) would have encountered granulite facies conditions(700–900 8C at 0.5 to ;1.3 GPa) before being subject to eclogite faciespressures (.1.3 GPa). Even under the comparatively cold conditionsin modern subduction zones, the slab-dehydration process is highly

766 GEOLOGY, September 2004

Figure 1. A: Schematic drawing showing how shallow dehydration ofthickened mafic crust under higher geothermal gradients of the Ar-chean may have built strong, buoyant lower crust and also led toformation of tonalite-trondhjemite-granodiorite (TTG) magmas that aretypical of Archean terranes. P, T—pressure, temperature. B: Contrast-ing P-T-time paths (wide arrows) for modern subducted slab vs. Ar-chean tectonically buried crust. Ellipses show approximate ranges ofmetamorphic facies fields (P-P—prehnite-pumpellyite facies). Thinblack arrow shows partial P-T history of Holsnøy rocks, which un-derwent granulite facies conditions ca. 945 Ma and eclogite conditionsca. 420 Ma.

efficient. Studies using cerium as an analogue for hydrogen (and thuswater content) in ocean basalts have shown that 97% of the water isextracted from a subducting slab before it reaches the middle of theupper mantle (Dixon et al., 2002). In the hotter world of the Archean,the dehydration process would presumably have been even more effi-cient, and would have occurred at a shallower depth.

ECLOGITIZATION AS A HYDRATION PHENOMENONToday, eclogite formation is related to progressive dehydration of

a subducting slab, although in detail the reaction mechanisms are morecomplex. First, the reaction rates are slowed by the coldness of theslab, and large pressure oversteps are required before significant con-version occurs (Kirby et al., 1996). Second, although the mineral phas-es that are diagnostic of the eclogite facies are not hydrous, water iscritical in facilitating the essential eclogite reaction albite → jadeite 1quartz (Hacker, 1996). Consequently, eclogite reactions tend to occurat depths where still-hydrous gabbros release water and rehydrate al-ready dry gabbro on the outer part of the slab (Kirby et al., 1996). Inthe Archean, slabs would have been hotter and diffusion rates faster,but thorough dehydration would have happened before a slab everreached eclogite conditions, and this circumstance would likely haveinhibited the formation of eclogite phases.

A remarkable complex of partly eclogitized granulites in Holsnøy,western Norway, provides a possible analogue for a phenomenon that

may have been common in Archean time. Although these rocks residedfor perhaps several millions of years at eclogite facies conditions, eclo-gite mineral assemblages occur only in patches and zones, mainlyalong fractures and in reaction fronts emanating from them (Fig. 2A).This observation suggests that the dry granulite persisted metastablyunder eclogite pressures until fracturing allowed the infiltration of wa-ter that then facilitated the transformation to eclogite through advectiveion transport (Austrheim and Griffin, 1985; Boundy et al., 1997; Jamt-veit et al., 2000; Bjørnerud et al., 2002).

The Holsnøy eclogites are within the Bergen arcs, a stack of ar-cuate thrust sheets of Proterozoic gneisses and lower Paleozoic ophiol-ite and island-arc complexes that were emplaced onto the margin ofthe Baltic craton during the formation of the Scandinavian Caledoni-des. The eclogites occur near the western edge of the largest exposedallochthonous sheet, the Lindas nappe. The island of Holsnøy is un-derlain by Middle(?) Proterozoic anorthositic, gabbroic, and locally pe-ridotitic rocks—perhaps originally a layered mafic intrusive complex—that were subject to temperatures of 800–900 8C and pressures of ;1.0GPa ca. 945 Ma (Austrheim and Griffin, 1985; Boundy et al., 1997).This granulite facies event left the rocks almost completely dehydrated,as indicated by nearly anhydrous mineral assemblages: garnet, Al-richdiopside, and plagioclase (An45–60), together with minor amounts oforthopyroxene, hornblende, and scapolite (Austrheim and Griffin,1985). Loss on ignition (LOI) values for the granulite rocks are ,0.2wt%, and the rocks may have been even drier immediately followingthe metamorphic peak.

The rocks on Holsnøy cooled by an unknown amount between945 Ma and ca. 420 Ma, when tectonic burial subjected them to pres-sures of .1.5 GPa (.50 km) and temperatures of 650–700 8C, asrecorded by the eclogite facies assemblages omphacite (Jd50), garnet,kyanite, and clinozoisite (Austrheim, 1987). The conversion of therocks from granulite to eclogite was virtually isochemical except forthe addition of an aqueous fluid in a concentration of one volumetricpart in 107–108 (Jamtveit et al., 1990). However, the granulite mineralassemblage persisted in ;50% of the rock volume, and the transfor-mation of the rock to eclogite was localized and heterogeneous, limitedalmost entirely by the availability of water (Engvik et al., 2000;Bjørnerud et al., 2002).

The anhydrous Holsnøy granulites were apparently very strongprior to their conversion to eclogite. They contain pseudotachylytesheets, some traceable for hundreds of meters, in which eclogite faciesminerals occur as microlites (Austrheim and Boundy, 1994). This ev-idence for frictional melting indicates that the granulites, while residingmetastably at eclogite conditions in the lower crust, stored largeamounts of strain energy until a series of massive earthquakes fracturedthe complex. These seismic events appear to have set in motion thelong-delayed eclogitization process by allowing water to enter therocks via macroscopic fractures. Reaction rates that had previouslybeen limited by diffusion were suddenly increased by orders of mag-nitude in the presence of a medium for advective ion transport (Rubie,1986; Baxter and DePaolo, 2000). Once the conversion to eclogitebegan, it was an initially self-perpetuating process, because the asso-ciated volume reduction (;10%) gave rise to a network of microfrac-tures that allowed water to make further inroads into the rock mass(Jamtveit et al., 2000). Ultimately, however, the process was self-limiting, because the hydrated eclogite was mechanically much weakerthan the dry, rigid granulite (Fig. 2B). Once the volume of granulitefell below a critical value (;70%), ductile eclogite limited the bulkrock strength, so that fracturing was no longer possible and fluids couldnot readily reach unconverted rock (Bjørnerud et al., 2002). An inter-esting implication of this change in rheology is that once a lower-crustal rock mass is thoroughly dehydrated, it may be impossible torehydrate it completely. What remained when no more water could

GEOLOGY, September 2004 767

Figure 2. A: Band of eclogitized rock (dark, upper left to lower right) on either side of meter-scale fracture in mafic granulite, Holsnøy,western Norway. Compositional foliation—layers of garnet and pyroxene alternating with plagioclase-rich domains—developed duringgranulite metamorphism and can be traced across eclogitized zone. Shorter transecting fractures may be caused by volume reductionassociated with formation of dense eclogite phases. Pencil is 15 cm long. B: Eclogite shear zone showing strong rheological contrastbetween ductile eclogite (right) and dry, brittle foliated granulite (left). Reduction in rock strength upon conversion to eclogite may limitextent of further eclogitization, because once strong granulite bodies are separated from each other, they no longer form stress-supporting framework capable of large-scale fracturing, and no new conduits can carry water necessary for eclogite reactions. Lenscap is 5 cm in diameter.

enter the rocks was a partly eclogitized granulite mass too buoyant tobe subducted.

WATER-STARVED ECLOGITE REACTIONS LED TOARCHEAN CONTINENTS

The Holsnøy rocks may have relevance for Archean Earth. Evenif subduction in the modern sense was not possible in Archean time,there had to be some process complementary to the production of newocean crust at upwelling convective sites (primitive spreading centers).Whether by shallow subduction or collision and thickening, someocean crust would have become tectonically buried and, under the highgeothermal gradients of the Archean, would have been dehydrated atdepths shallower than those at which slab dehydration occurs today.Geochemical studies of the tonalite-trondhjemite-granodiorite (TTG)suites that are typical of Archean terranes (Condie, 1981; Wyllie et al.,1997) show that the source melts for TTG rocks older than 3.0 Ga didnot interact with the mantle (Smithies, 2000; Martin and Moyen, 2002).These analyses suggest that prior to Late Archean time, convergencebetween slabs of mafic crust occurred by stacking or low-angle under-thrusting (Fig. 1A), not subduction sensu stricto, in which a wedge of

mantle material overlies the downgoing plate. Collisions between oce-anic plateaus could have produced areas of particularly thick crust thatmay have formed the substrates for the first continents (Stein and Gold-stein, 1996).

One implication of this style of convergence is that the pressure-temperature-time path of a tectonically buried rock on its way to theeclogite facies field would have been very different in Archean timethan it is today (Fig. 1B). While modern oceanic slabs typically passthrough the low-temperature, high-pressure blueschist field before un-dergoing eclogite facies conditions, Archean ocean crust probably tra-versed the amphibolite and granulite fields en route to the eclogitefacies. The rarity of blueschists and eclogites older than 2 Ga is con-sistent with this hypothesis, although it could also be an artifact ofpreservation (Moller et al., 1995). The Holsnøy rocks demonstrate thatthe history of a rock mass can have an important effect on the waythe rock responds to subsequent metamorphic conditions. If maficrocks arrive in the eclogite field with very little water, they may beunable to equilibrate to the new pressure conditions and will retainlower-density, metastable mineral assemblages. The Holsnøy complexsuggests that such thoroughly dehydrated rocks may never transform

768 GEOLOGY, September 2004

entirely to eclogite even if external water later becomes available, be-cause the conversion to eclogite progressively reduces rock strength,which eventually suppresses the formation of large-scale fractures thatact as pathways for water to reach the interior of the rock mass. Thispoint should be taken into consideration in models that invoke density-driven delamination as a mechanism for crustal recycling in Archeantime (Rudnick, 1995; Zegers and van Keken, 2001).

An uncertainty in the proposed model is the exact means by whichunderthrust rocks would have been dehydrated during tectonic burialor protosubduction in Archean time. In modern subduction zones, theprocess seems to be very efficient (Dixon et al., 2002), and under thehigher geothermal gradient of the Archean, prograde devolatilizationreactions would likely have proceeded even more quickly. However,the geometry of metamorphic dehydration of thickened Archean crustwould have been different from that for a slab entering a relatively drymantle environment, and metamorphism alone may not have been ableto drive all the water out of the rocks. It is possible that extraction ofsmall amounts of melt that absorbed free water from the lower maficcrust may have assisted in the dehydration of these rocks. This is con-sistent with the geochemical signatures of the oldest TTG suites, whichappear to reflect melting of hydrous basalt at the base of thickenedmafic crust (Wyllie et al., 1997; Smithies, 2000; Martin and Moyen,2002).

We suggest that prograde metamorphic dehydration reactions to-gether with partial melting of the lowermost crust in Archean colli-sional zones would have resulted in a relatively buoyant raft of drymafic granulite below the tonalitic upper crust. Even if this lower crustunderwent eclogite facies pressures, it would have been resistant toeclogitization in the absence of free water, because reaction kineticswould have been limited by the rates of solid-state diffusion. Thesedry, buoyant masses would also have been exceptionally strong andmay have constituted the first true lithosphere. Only when upper-mantletemperatures had moderated by 100–200 8C (probably around the timeof the Archean-Proterozoic transition) could downgoing oceanic crustremain cold enough to retain its water to eclogite-forming depths. Oncethese conditions were met, modern tectonics—with efficient eclogiteproduction and generation of arc magmas from water-fluxed mantle—could finally begin.

SUMMARY AND CONCLUSIONSAlthough modern eclogite formation is generally considered a de-

hydration process, rocks that have been strongly dehydrated prior toundergoing eclogite pressures do not convert readily to eclogite, asshown by the partly eclogitized mafic granulites of the Bergen arcs.The higher geothermal gradient of the Archean could have caused tec-tonically buried ocean crust to be thoroughly dehydrated at shallowerdepths than occurs today. The residual, dehydrated crust would havebeen very strong and too buoyant to sink into the mantle. The emer-gence of modern plate tectonics around the beginning of the Protero-zoic Eon may therefore reflect a change in eclogite reaction mecha-nisms and rates, from slow, diffusion-dominated processes in dryoceanic crust to much faster, fluid-assisted recrystallization of still-hydrated gabbro. In a sense, plate tectonics required a deeper connec-tion between Earth’s hydrosphere and interior than was possible on thehot, young Earth.

ACKNOWLEDGMENTSWe thank two anonymous reviewers for comments that significantly im-

proved the manuscript.

REFERENCES CITEDAnderson, D., 1989, Where on Earth is the crust?: Physics Today, v. 42, no. 3,

p. 38–44.Austrheim, H., 1987, Eclogitisation of lower crustal granulites by fluid migra-

tion through shear zones: Earth and Planetary Science Letters, v. 81,p. 221–232.

Austrheim, H., and Boundy, T., 1994, Pseudotachylytes generated during seis-mic faulting and eclogitization of the deep crust: Science, v. 265,p. 82–83.

Austrheim, H., and Griffin, W., 1985, Shear deformation and eclogite formationwithin granulite facies anorthosites of the Bergen arcs, western Norway:Chemical Geology, v. 50, p. 267–281.

Baxter, E., and DePaolo, D., 2000, Field measurements of slow metamorphicreaction rates at temperatures of 500 to 6008C: Science, v. 288,p. 1411–1414.

Bjørnerud, M., Austrheim, H., and Lund, M., 2002, Processes leading to den-sification (eclogitization) of tectonically buried crust: Journal of Geo-physical Research, v. 107, no. B10, 2252, doi: 10.1029/2001JB000527.

Boundy, T., Mezger, K., and Essene, E., 1997, Temporal and tectonic evolutionof the granulite-eclogite association from the Bergen arcs, western Nor-way: Lithos, v. 39, p. 159–178.

Bowring, S., and Housh, T., 1995, The Earth’s early evolution: Science, v. 269,p. 1535–1540.

Breuer, D., and Spohn, T., 1995, Possible flush instability in mantle convectionat the Archaean-Proterozoic boundary: Nature, v. 378, p. 808–610.

Condie, K., 1981, Archaean greenstone belts: Amsterdam, Elsevier, 434 p.Davies, G., 1992, On the emergence of plate tectonics: Geology, v. 20,

p. 963–966.Dixon, J., Leist, L., Langmuir, C., and Schilling, J.-G., 2002, Recycled dehy-

drated lithosphere observed in plume-influenced mid-ocean-ridge basalt:Nature, v. 420, p. 385–389.

Engvik, A., Austrheim, H., and Andersen, T.B., 2000, Structural, mineralogicaland petrophysical effects on deep crustal rocks of fluid-limited poly-metamorphism, Western Gneiss Region, Norway: Geological Society[London] Journal, v. 157, p. 121–134.

Hacker, B., 1996, Eclogite formation and the rheology, buoyancy, seismicity,and H2O content of oceanic crust, in Bebout, G., et al., eds., Subductiontop to bottom: American Geophysical Union Geophysical Monograph 96,p. 337–346.

Hamilton, W., 2003, An alternative Earth: GSA Today, v. 13, no. 11, p. 4–12.Herzberg, C., 1993, Lithosphere peridotites of the Kaapvaal craton: Earth and

Planetary Science Letters, v. 120, p. 13–29.Jamtveit, B., Bucher-Nurminen, K., and Austrheim, H., 1990, Fluid controlled

eclogitization of granulites in deep crustal shear zones, Bergen arcs, west-ern Norway: Contributions to Mineralogy and Petrology, v. 104,p. 184–193.

Jamtveit, B., Austrheim, H., and Malthe-Sørensen, A., 2000, Accelerated hy-dration of the Earth’s deep crust induced by stress perturbations: Nature,v. 408, p. 75–78.

Kirby, S., Engdahl, E.R., and Denlinger, R., 1996, Intermediate-depth earth-quakes and arc volcanism as physical expressions of crustal and uppermostmantle metamorphism in subducting slabs, in Bebout, G., et al., eds., Sub-duction top to bottom: American Geophysical Union Geophysical Mono-graph 96, p. 195–214.

Martin, H., and Moyen, J.-F., 2002, Secular changes in tonalite-trondhjemite-granodiorite composition as markers of the progressive cooling of Earth:Geology, v. 30, p. 319–322.

Moller, A., Appel, P., Mezger, K., and Schenk, V., 1995, Evidence of a 2 Gasubduction zone: Eclogites in the Usagaran belt of Tanzania: Geology,v. 23, p. 1067–1070.

Rubie, D., 1986, The catalysis of mineral reactions by water and restrictions onthe presence of aqueous fluid during metamorphism: Mineralogical Mag-azine, v. 50, p. 399–415.

Rudnick, R., 1995, Making continental crust: Nature, v. 378, p. 571–578.Smithies, R., 2000, The Archean tonalite-trondhjemite-granodiorite (TTG) se-

ries is not an analogue of Cenozoic adakite: Earth and Planetary ScienceLetters, v. 182, p. 115–125.

Stein, M., and Goldstein, S., 1996, From plume head to continental lithospherein the Arabian-Nubian Shield: Nature, v. 382, p. 773–778.

Vlaar, N., van Keken, P., and van den Berg, A., 1994, Cooling of the Earth inthe Archean: Earth and Planetary Science Letters, v. 121, p. 1–18.

Wilde, S., Valley, J., Peck, W., and Graham, C., 2001, Evidence from detritalzircons for the existence of continental crust and oceans on Earth 4.4 Gyrago: Nature, v. 409, p. 175–178.

Wyllie, P.J., Wolf, M., and van der Laan, S., 1997, Conditions for formation oftonalities and trondhjemites: Magmatic sources and products, in de Wit,M., and Shwal, L., eds., Greenstone belts: Oxford, UK, Clarendon Press,p. 256–266.

Zegers, T., and van Keken, P., 2001, Middle Archean continent formation bycrustal delamination: Geology, v. 29, p. 1083–1086.

Manuscript received 20 February 2004Revised manuscript received 26 May 2004Manuscript accepted 9 June 2004

Printed in USA