Contributions to Contrib Mineral Petrol (1983) 84:1-5 Mineralogyand Petrology Springer-Verlag 1983

Density Constraints on the Formation of the Continental Moho and Crust

C.T. Herzberg 1, W.S. Fyfe 2, and M.J. Carr 1 1 Department of Geological Sciences, Rutgers University, New Brunswick, New Jersey, USA 08903 z Department of Geology, University of Western Ontario, London, Ontario, Canada N6A 5B7

Abstract. The densities of mantle magmas such as MORB- intermediate composition favoured by Tarney and Wind- like tholeiites, picrites, and komatiites at 10 kilobars are ley (1979) to more basic compositions than the basaltic- greater than densities for diorites, quartz diorites, granodio- andesite preferred by Taylor (1979). Uncertainties in pro- rites, and granites which dominate the continental crust. viding a more exact composition stem from obvious sam- Because of these density relations primary magmas from pling problems and from the model-dependent nature of the mantle will tend to underplate the base of the continen- mass balance calculations which proportion a granodiorite tal crust. Magmas ranging in composition from tholeiites upper layer to various assumed bulk crust compositions. which are more evolved than MORB to andesite can have Exposures of obducted portions of the upper and lower densities which are less than rocks of the continental crust continental crust are now thought to exist in cross-section at 10 kilobars, particularly if they have high water contents. at a number of sites on the Earth's surface; these have The continental crust can thus be a density filter through been reviewed in detail by Fountain and Salisbury (1981). which only evolved magmas containing H20 may pass. This An important outcome of the raw geological data is that explains why primary magmas from the mantle such as it illustrates how overly simple is the above two-layer model the picrites are so rare. Both the over-accretion (i.e., Moho for continental structure. Superimposed on prominent penetration) and the under-accretion (i.e., Moho underplat- metamorphic facies changes is an extensive vertical and lat- ing) of magmas can readily explain complexities in the litho- eral compositional variability in addition to complexities logical characteristics of the continental Moho and lower associated with deformation. Although mafic granulites are crust. Underplating of the continental crust by dense mag- important constituents of the lower crust in some areas, mas may perturb the geotherm to values which are charac- in others the prominent rock type is intermediate to acid teristic of those in granulite to greenschist facies metamor- granulite. phic sequences in orogenic belts. An Archean continental New high resolution seismic data are now showing that crust floating on top of a magma flood or ocean of tholeiite the original Moho concept for the crust-mantle boundary to komatiite could have undergone a major cleansing pro- beneath continents is also much too simple. Obviously, a cess; dense blocks of peridotite, greenstone, and high den- petrologically complex lower continental crust and upper sity sediments such as iron formation could have been re- mantle must be separated by an equally complex "inter- turned to the mantle, granites sweated to high crustal levels, face". To quote Oliver (1982) "Many recent observations and a high grade felsic basement residue established. indicate that the current view of the Moho, the crust-mantle boundary, needs revision. In particular, the widely held con- cept of the Moho as a sharp laterally uniform boundary appears misleading in its simplicity, and may even be a Introduction major barrier to a better understanding of the nature and evolution of continents". Recent COCORP (Brown et al. Despite our inability to directly observe the inaccessible 1981) and BIRPS (Matthews 1982) seismic data for deep regions of the lower continental crust, there is good evi- dence that it differs compositionally from crust near the crustal regions clearly show the complexity of deep crustal surface. This has arisen from heat flow constraints which structure, Data like Fig. 3 from Brown et al. (1981) is rather require the upper crust to be enriched in the LIL elements typical and shows for a region of Texas that there is no such as U, Th, and K relative to deeper levels (Heier 1973). single well defined continental Moho. Similar data led A number of simple two-layer geometries have thus Missner (1973) to conclude that the Moho in some areas emerged, granodiorite being an average composition of the is a transition zone characterized by velocities greater than uppermost part (Taylor 1979) and granulite facies rocks mean deep crustal values and yet less than mantle P-wave of uncertain composition below. The formation of granu- velocities; the interfingering of ultramafic intrusions and lites by dehydration and partial melting processes is likely to granulites is an interpretation of this transition zone which be important for transporting the LIL elements from the is receiving support from field and petrological studies of lower to the upper crust (e.g., Fyfe 1973; Heier 1973), and the peridotites and associated rocks in the Invrea Zone (e.g., could be compatible with a lower crust ranging from an Fountain and Salisbury 1981 ; Rivalenti et al. 1981). Brown et al. (1981) show that structures may change radically over Offprint requests to: W.S. Fyfe very short lateral distances and could be strongly controlled by local rheology, density or composition. The new seismic

data also show the existence of rather low angle thrust faults 2.5 in continental crust, thrusts which may go far below con- ventional Moho depths (Smythe et al. 1982; Matthews 1982). Certainly there are places where a good 30 km Moho 2.7 appears to be present, but one wonders if this is not often simply a reflection of the resolution of seismic techniques (e.g., see Bolt 1982; p 74-78). It is clear that any hypothesis which attempts to explain 2.9 the origin of the structure of the continental crust and Moho must provide a mechanism for understanding the lithological diversity involved. This mechanism must ad- 3.1 dress the control of mass transfer during igneous and meta- morphic processes throughout geological time. As a step in this direction, we wish to point out some rather simple density relationships amongst rocks and magmas which D~D place constraints on the distribution of mass in the conti- nental crust and upper mantle. I I A number of recent observations tend to suggest that 10 20 PRESSURE KILOBARS all models have underestimated the importance of a basaltic component in both the lower crust in general and in the Fig. 1. Densities of rocks and magmas. Rock densities are averages listed in Clark (1966) calculated for lower crustal temperatures of formation of present-day subduction zone magmatism in 600-I ,000~ C using i/V(dV/dT)= 3 x I 0- 5o C- 1 and -1/V(dV/dP) particular. Basaltic and kimberlitic pipes commonly contain =1.1 x l0 -3 kbar -1. Magmas are Mt. Hood andesite given in pyroxene and garnet granulite nodules which are basaltic Bottinga and Weill (1970), average FAMOUS MORB glass in E1- in composition and which originated at pressures corre- thon (1979), and P9-118 komatiite in Arndt et al. (1977) at 1,100, sponding to those of the lower crust (e.g., Rogers 1977; 1,250, and 1,600~ C respectively. Densities of magmas at 1 atmo- Arculus and Smith 1978; Padovani and Carter 1977; sphere are from Nelson and Carmichael (1979). The range of McCulloch et al. 1982). Subduction zone magmas of Cen- magma densities at high pressures arise from uncertainties in the tral America are typically basaltic rather than andesitic, isothermal bulk modulus K ~ and its pressure derivative K' in an and include primitive MORB-like tholeiites (Carr 1983). equation of state (Stolper et al. 1981; Herzberg 1983; Kushiro 1982); the values chosen are K~ 150 to 200 kbar at K'= 7. Den- Similar primitive magmas have been modelled as parental sities of the high pressure mineralogies increase continuously and to the high alumina basalts in the Aleutians (Conrad and discontinuously due to solid solution effects and the stabilization Kay 1982). Indeed, basaltic inclusions in crystals of olivine of jadeite, garnet, kyanite, and coesite at pressures higher than found in andesites, dacites as well as basalts have lead An- 10-15 kbar (Stern et al. 1975; Stern and W~cllie 1981; Herzberg derson (1982) to the general conclusion that basaltic liquid 1978) is parental to the formation of new continental crust at convergent plate boundaries. In Central America many of the magma chambers of rocks which make up the crust and upper mantle, in which fed the volcanoes appear to be located at the base addition to those of magmas which range from komatiite of the crust (Carr 1983), in indicating that the Moho is to andesite. The most important observation is that primi- a boundary layer under which magmas from the mantle tive MORB-like tholeiitic magmas are more dense than are temporarily ponded. A magma ponding or underplating diorite, quartz diorite, granodiorite, and granite at 10 kilo- process has previously been considered in an attempt to bars; they are also denser than most common sedimentary explain the formation of continental crust during the Ar- rocks. A maximum H20 content of such magmas which chean (Fyfe 1973, 1974, 1978), and is now being encorpor- can be parental to high alumina basalts and andesites is ated into models of the formation of continental flood ba- likely to be less than 1.6 wt.% (D.M. Harris, pers. comm.), salts (Cox 1980) and all magmatic processes at convergent thus reducing their density by about 0.06 gm cm- 3 at these plate margins (Hildreth 1981). The idea is that basaltic pressures (Bottinga and Weill 1970). A water content of magma will from a stable fluid layer beneath 'granitic' ma- 3 % contained in some basaltic glasses considered by Ander- terial of lower density as long as partial melting or plastic son (1982) would reduce the magma density even more, deformation of this buoyant layer can temporarily seal frac- but it would still be less than granodiorite at 10 kilobars. tures and prevent the formation of a hydraulic head via Although the dissolution of H20 can clearly reduce the extrusion. In the model of Cox (1980) the magma is a picrite densities of magma at the base of the crust by a significant which underplates and fractionates until the residual liquid amount, this can be partly offset by the effect of suspended becomes basaltic; the Moho becomes the petrological inter- crystals in a vigorously convecting parcel of magma. For face between olivine cumulates added to the mantle and example, a dry tholeiitic liquid laden with 30% olivine crys- gabbros added to the lower crust. tals in suspension can have a bulk density of about 2.9 gm cm -3 at 10 kilobars, far greater than diorite and Magma Densities quartz diorite. For relatively dry tholeiitic and picritic magmas, the With the densities of magmas to high pressures now better continental Moho can be an effective boundary layer across understood (Stolper et al. 1981; Herzberg 1983; Kushiro which only heat could be transferred to the crust above. 1982) it is possible to examine the underplating model with Indeed, these density relations explain why primary mag- new constraints. Figure 1 shows the densities of a range mas from the mantle (i.e., picrites and komatiites) are so rare; quite simply, their densities require that they remain Zimbabwe have been interpreted to be juvenile material in the mantle or pond below a low density boundary layer. produced from an irreversibly differentiating Earth (Moor- A breakdown of this boundary layer may only occur if bath and Taylor 1981). However, the possibility that there the major element chemistry of these magmas evolve and has been mixing of continental crust and mantle on a large if they gain volatiles such as H20. The continental crust scale will render this interpretation ambiguous (Fyfe 1981). could thus become a kind of density filter through which Regardless of whether the mass of the continental crust evolved tholeiites and magmas of intermediate composition has grown or has been conserved with time, the isotopic could pass, particularly if they gained H20. Magmas which data indicate that large volumes of calcalkaline magmas underplate the crust or penetrate the Moho and reside at could have originated from the mantle during certain epi- various crustal levels could have their trace element and sodes. For a detailed discussion of the relationship between isotopic signatures extensively modified by contamination isotopic "age peaks" and "quantum" jumps in convection (e.g., Carter etal. 1978; Thompson etal. 1982; Carlson styles in the mantle, the reader is referred to Herzberg and et al. 1981). Forsythe (1983). Of immediate concern here, however, is An objection which is periodically raised to a basaltic the fate of large quantities of these low density magmas. lower continental crust is that it would crystallize to eclogite Figure 1 shows that such magmas of intermediate composi- with densities and P-wave velocities which are higher than tion are less dense than granodiorite at 10 kilobars. These the known geophysical properties of the crust at 10 kilobars could have easily disrupted any pre-existing ancient Moho (e.g., Heier 1973). This objection is based on the use of during their ascent to high crustal levels. This corresponds phase diagrams of the gabbro to eclogite transformation to the "over-accretion" model of Wells (1980) rather than which were constructed from high temperature experiments the "under-accretion" or under-plating model which we extrapolated linearly to the temperatures of interest at the have discussed above. Intuitively, the consequences of each base of the crust (Green and Ringwood 1972; Kennedy model on the spatial distribution of igneous rocks, regional and Ito 1972). Because these extrapolations require eclogite metamorphism, and partial melting in the crust would be to be the stable basaltic mineralogy at all points along a profoundly different. Whereas the underplating model normal geotherm, hypothetical reaction kinetics associated would tend to distribute the heat of crystallization of basal- with this transition has been invoked to explain why we tic magmas over a larger lateral extent and produce a more rarely see eclogite in crustal rocks (Kennedy and Ito 1972). sharply-defined Moho (i.e., but see below), the overaccre- However, it has been pointed out that these phase diagrams tion model could result in larger lateral modifications to based on linear extrapolations of high temperature experi- pre-existing lithologies and a severely disrupted Moho. mental data may be in error (Herzberg 1978). Indeed, recent Since the geological record points to a nonsteady state cool- experimental and thermochemical work has shown that gar- ing Earth with periods of enhanced magmatic activity and net granulite rather than eclogite is the stable assemblage metamorphism followed by relatively quiescent periods for many anhydrous metabasic compositions at 9 kilobars (e.g., Herzberg and Forsythe 1983), we should expect that and 700-900~ C (Newton and Perkins 1982). Additionally, there have been various ways in which mass has been trans- any H20 introduced could transform both garnet granulite ferred from the mantle to the crust. The production of a and eclogite into amphibolite assemblages (e.g., Wyllie wide range of lower crustal lithologies and Moho geome- 1977). Finally, large volumes of magmas ponded below the tries by these different accretion mechanisms is totally con- crust would cool very slowly and differentiate much more sistent with the geological and seismic observations outlined efficiently than similar magma batches at higher crustal lev- above. els. Strong flotation of plagioclase and sinking of olivine With these alternative accretionary mechanisms in and pyroxene cumulates could produce a layered ultra- mind, we think that the obducted portions of the upper mafic-anorthosite complex, not eclogite. As pointed out and lower crustal columns reviewed by Fountain and Sal- above, both seismic and field observations support the con- isbury (1981) point to the overall importance of a magma cept of the Moho as a transition zone of interlayered intru- underplating model. Common to the cross-sections at the sions and granulites at the base of the crust in some areas. Invrea Zone, Fraser and Musgrave Range, Pikwitonei Belt, It is worth considering, however, some consequences and the Kasila Series is a change in metamorphic grade of an eclogite mineralogy in the lower portions of a very from the granulite facies at the base to greenschist facies thick Andean-type continental crust where pressures in the at the top. To quote Fountain and Salisbury (1981) "the 10 20 kilobar range may be reached. Layers or pods of most prominent layering in the crust is not compositional eclogite with densities higher than that of mantle peridotite but metamorphic". Although detailed geothermometry and below would certainly be unstable; decoupling of eclogite geobarometry remains to be done for these areas, most from the lower crust and subsequent foundering into the granulite facies rocks from all terrains studied in detail re- mantle wedge above the subduction zone could occur. The cord 700-1,000 ~ C and 5 10 kilobars (Bohlen and Boettcher fate and identify of the lower crust may then be very com- 1981; Johnson and Essene 1982; Newton and Perkins 1982; plex and depend on the total thickness and tectonic history Bohlen et al. 1983). If these equally apply to the granulite of the continental crust being ponded. A basaltic compo- terrains cited above, the metamorphic geotherm would nent at the base of a very thick crust could begin to autodes- range from about 20-55 ~ C km- 1, and could be representa- truct on formation, leaving behind a stable lower crust with tive of the actual geothermal gradient at their time of for- a buoyant mineralogy appropriate to compositions which mation. Although high metamorphic geotherms have been are less basic. interpreted as the product of rapid uplift and erosion for The nature of the lithological diversity bounded on each the Dalradian (e.g., England and Richardson 1977), we side of a Moho can be even more complex when the origin think that they are also the consequence of a vast "contact of large volumes of Precambrian amphibolite fades ortho- aureole" effect due to magma underplating. Indeed, both gneisses are considered. Those from West Greenland and wocesses could have competed to boost the thermal state of the crust particularly if significant crustal updoming oc- the nature of primary magmas and crustal evolution in the curs above sites of rising mantle convection cells or diapirs. Aleutian Arc. J Volcanol Geotherm Res (in press) A continental crust floating on an underplate magma Cox KG (1980) A model for flood basalt volcanism. J Petrol flood would be subject to partial melting or degassing, and 21 : 62%650 Elthon D (1979) High magnesia liquids as the parental magma become mechanically weak. Complex styles of deformation for ocean floor basalts. Nature 278:514-518 near the Moho with rapid lateral changes (Brown et al. England PC, Richardson SW (1977) The influence of erosion upon 1981) would hardly be surprising. Further, underplating the mineral facies of rocks from different metamorphic environ- provides an elegant mechanism for "cleaning" continental ments. J Geol Soc London 134: 201-213 crust by a density filtration process. For example, if granite- Fountain DM, Salisbury MH (1981) Exposed cross-sections greenstone terrains of the Archean were underplated, high through the continental crust: implications for crustal structure, density lithologies such as peridotites, greenstone and iron petrology, and evolution. Earth Planet Sci Lett 56: 263-277 formations could have been returned to the mantle, low Fyfe WS (1973) The granulite facies, partial melting and the Ar- melting granites sweated out to higher crustal levels, and chean crust. Phil Trans R Soe London A273:457-461 a high grade felsic basement residue established with the Fyfe WS (1974) Archean tectonics. Nature 249:338 Fyfe WS (1978) The evolution of the Earth's crust: modern plate constraint that P basement < P underplate magma. Sinking tectonics to ancient hot spot tectonics? Chemical Geology of high density H20-CO 2- S rich materials could flood 23: 89-114 the lower crust with such volatiles. We suggest that this Fyfe WS (1981) How do we recognize plate tectonics in very old could have been particularly important in the Archean char- rocks? In: Kroner A (ed) Precambrian Plate Tectonics. Elsevier acterized by the flotation of continental crust on a magma 549 560 flood or ocean of dense tholeiite to komatiite. Green DH, Ringwood AE (1972) A comparison of recent experi- mental data on the gabbro-garnet granulite-eclogite transition. Acknowledgements. This research was partly supported by NASA J Geol 80 : 277-288 Under Grant NAGW-177. This paper was motivated by the orga- Heier KS (1973) Geochemistry of granulite facies rocks and prob- nizers and participants of the Workshop on the Lower Continental lems of their origin. Phil Trans R Soc London A273:429-442 Crust held in Princeton, 1982. Herzberg CT (1978) Pyroxene geothermometry and geobarometry: Experimental and thermodynamic evaluation of some subsoli- dus phase relations in the system CaO-MgO-AI203-SiO 2. References Geochim Cosmochim Acta 42:945-957 Herzberg CT (1983) Solidus and liquidus temperatures and minera- Anderson AT, Jr (1982) Parental magmas in subduction zones: logies for anhydrous garnet-lherzolite to 15 GPa. Phys Earth implications for continental evolution. J Geophys Res Planet Inter (in press) 87: 7047-7060 Herzberg CT, Forsythe RD (1983) Destabilization of a 650 kilome- Arculus RJ, Smith D (1979) Eclogite, pyroxenite and amphibolite ter chemical boundary layer and its bearing on the evolution inclusions in the Sullivan Buttes latite, Chino Valley, Yavapai of the continental crust. Phys Earth Planet Inter (in press) County, Arizona. In: Boyd FE, Meyer HOA (eds) The mantle Hildreth W (1981) Gradients in silicic magma chambers: implica- sample: Inclusions in kimberlites and other volcanics. Am Geo- tions for lithospheric magmatism. J Geophys Res 86:10153- phys Union 309 317 10192 Arndt NT, Naldrett AJ, Pyke DR (1977) Komatiitic and iron-rich Johnson CA, Essene EJ (1982) The formation of garnet in olivine- tholeiitic lavas of Munro Township, Northeastern Ontario. J bearing metagabbros from the Adirondacks. Contrib Mineral Petrol 18:319-369 Petrol 81:240-251 Bohlen SR, Boettcher AL (1981) Experimental investigations and Kennedy GC, Ito I (1972) Comments on: "A comparison of recent geological applications of orthopyroxene geobarometry. Am experimental data on the gabbro-garnet granulite-eclogite tran- Mineral 66: 951-964 sition". J Geol 80:28%292 Bohlen SR, Wall VJ, Boettcher AL (1983) Experimental investiga- Kushiro I (1982) Density of tholeiite and alkali basalt magmas tion and application of garnet granulite equilibria. Contrib at high pressures. Carnegie Inst Wash Yearbook 81:305-309 Mineral Petrol in press Matthews DH (1982) BIRPS: deep seismic reflection profiling ar- Bolt BA (1982) Inside the Earth. WH Freeman and Co, San Fran- ound the British Isles. Nature 298:709-710 cisco, 191 pp McCulloch MT, Arculus RJ, Chappel BW, Ferguson J (1982) Bottinga Y, Weill DF (1970) Densities of liquid silicate systems Isotopic and geochemical studies of nodules in kimberlite have calculated from partial molar volumes of oxide components. implications for the lower continental crust. Nature 300:166- Am J Sci 268:169-182 169 Brown LD, Oliver JE, Kaufman S, Brewer JA, Cook FA, Schilt Meissner R (1973) The "Moho" as the transition zone. Geophys FS, Albaugh DS, Long GH (1981) Deep crustal strucutre: Im- Surveys 1 : 195 216 plications for continental evolution. In: O'Connell RJ, Fyfe Moorbath S, Taylor PN (1981) Isotopic evidence for continental WS (eds) Evolution of the Earth. Geodynamics series vol. 5, growth in the Precambrian, In: Kroner A (ed) Precambrian Am Geophys Union 38-52 Plate Tectonics, Elsevier 491-526 Carlson RW, Lugmair GW, MacDougall JD (1981) Columbia Nelson SA, Carmichael ISE (1979) Partial molar volumes of oxide River volcanism: the question of mantle heterogeity or crustal components in silicate liquids. Contrib Mineral Petrol contamination. Geochim Cosmochim Acta 45:2483-2499 71 : 117-124 Carr MJ (1983) Crustal influence on magmatic diversity along the Newton RC, Perkin D (1982) Thermodynamic calibration of geo- Central American volcanic front. J Volcanol Geotherm Res, barometers based on the assemblages garnet-plagioclase-ortho- submitted pyroxene (clinopyroxene)-quartz. Am Mineral 67:203-222 Carter SR, Evensen NM, Hamilton PJ, O'Nions RK (1978) Nd- Oliver J (1982) Changes at the crust-mantle boundary. Nature and Sr-isotopic evidence for crustal contamination of continen- 299 : 398-399 tal volcanics. Science 202 : 743-747 Padovani ER, Carter JL (1977) Aspects of the deep crustal evolu- Clark SP, Jr (1966) Handbook of Physical Constants. Geol Soc tion beneath south central New Mexico. In: Heacock JG) The Amer Mere 97 Earth's crust: its nature and physical properties. Am Geophys Conrad WK, Kay RW (1982) Ultramafic and mafic inclusions Monogr Ser 20:19-55 from Adak Island: Crystallization history, and implications for Rivalenti G, Garuti G, Rossi A, Siena F, Sinigoi S (1981) Existence of different peridotite types and of a layered igneous compiex Tarney J, Windley BF (1979) Continental growth, island arc accre- in the Invrea Zone of the Western Alps. J Petrol 22:127-153 tion and the nature of the lower crust - a reply to SR Taylor Rogers NW (1977) Granulite xenoliths from Lesotho kimberlites and SM McLennan. J Geol Soc London 136:501-504 and the lower continental crust. Nature 270:681-684 Taylor SR (1979) Chemical composition and evolution o the conti- Smythe DK, Dobinson A, McQuillan R, Brewer JA, Matthews nental crust: the rare earth element evidence. In: McElhinney DH, Blundell DJ, Kelk B (1982) Deep structure of the Scottish MW (ed) The Earth: Is origin, structure, and evolution. Aca- Caledonides revealed by the MOIST reflection profile. Nature demic 353 : 376 299 : 338-340 Thompson RN, Dicken AP, Gibson IL, Morrison MA (1982) Ele- Stern CR, Huang W-L, Wyllie PJ (1975) Basalt-andesite-rhyolite- mental fingerprints of isotopic contamination of Hebridean Pa- H20: crystallization intervals with excess H20 and H20-under- laeocene mantle-derived magmas by Archean sial. Contrib Min- saturated liquidus surfaces to 35 kilobars, with implications eral Petrol 79 : 159-168 for magma genesis. Earth Planet Sci Lett 28:189-196 Wells PRA (1980) Thermal models for the magmatic accretion Stern CR, Wyllie PJ (1981) Phase relationships of I-type granite and subsequent metamorphism of continental crust. Earth with HzO to 35 kilobars: The Kinkey Lakes biotite-granite Planet Sci Lett 46:253-265 from the Sierra Nevada batholith. J Geophys Res Wyllie PJ (1977) Crustal anatexis: an experimental review. Tectono- 86:10412-10422 phys 43:41-71 Stolper E, Walker D, Hagar BH, Hays JF (1981) Melt segregation from partially molten source regions: the importance of melt density and source region size. J Geophys Res 86:6261-6271 Accepted May 3, 1983