Magma Genesis, Plate Tectonics, and Chemical Differentiation of the Earth

Home , Igneous petrology, Magmatism, Mineralogy, Petrology

REVIEWS OF GEOPHYSICS, VOL. 26, NO. 3, PAGES 370-404, AUGUST 1988

PETER J. WYLLIE

Division of Geolo•7icaland Planetary Sciences,California Institute of Technolo•Ty,Pasadena

Magma genesis,migration, and eruption have played prominent roles in the chemical differentiation of the Earth. Plate tectonics has provided the framework of tectonic environments for different suites of igneousrocks and the dynamic mechanismsfor moving massesof rock into melting regions.Petrology is rooted in geophysics.Petrological and geophysicalprocesses are calibrated by the phase equilibria of the materials. The geochemistry of basalts and mantle xenoliths demonstrates that the mantle is heterogeneous.The geochemical reservoirs are related to mantle convection, with interpretation of a mantle layered or stratified or peppered with blobs. Seismic tomography is beginning to reveal the density distribution of the mantle in three dimensions,and together with fluid mechanical models and interpretation of the geoid, closer limits are being placed on mantle convection. Petrological cross sectionscon- structed for various tectonic environments by transferring phase boundaries for source rocks onto assumedthermal structuresprovide physical frameworks for consideration of magmatic and metasomatic events,with examplesbeing given for basalts,andesites, and granites at ocean-continentconvergent plate boundaries, basalts and nephelinitesfrom a thermal plume beneath Hawaii, kimberlites in cratons, nephelinites from continental rifts, and anorogenic granites. The fluid dynamics of rock-melt-vapor systemsexerts strong control on igneous processesand chemical differentiation. Unravelling the processesduring subduction remains one of the major problems for understandingmantle heterogeneities and the evolution of continents.

INTRODUCTION the magmatic processesoccurring nearer the source of magma The Union Lecture on which this review is based was pre- generation.There remains a gap betweenvolcanoes, their plu- sented on behalf of the International Association for Vol- tonic roots, and the magma sourceswhich has to be filled by canology and the Chemistry of the Earth's Interior (IAVCEI). indirect studies including geophysics,geochemistry, and fluid It was an over-illustrated, somewhat superficial ramble dynamics. Lavas reach the surface through volcanoes,and through the whole Earth emphasizing volcanic activity, the lavas can provide information about the source rocks from more dramatic aspectsof the Earth's chemical differentiation. which they were derived if the materialsand processescan be I have been charged by President Lal of the International suitably calibrated in the laboratory. The calibration involves Union of Geodesy and Geophysics (IUGG) to prepare the the geochemistryof major, minor, and trace element distri- review more or less along the lines of the lecture, without butions (including isotopes)between minerals and melts and writing the book and compilingthe enormouslist of references the conditions for melting of the various source materials required to do formal justice to all these topics. The list of under various conditions.Some lavas also bring to the surface referencescited is admirably supplementedby the comprehen- samplesof the host rocks from which they were derived by sive bibliographies provided in the U.S. National Report to partial melting or of the rocks through which they rose.The the IUGG for 1983-1986, published in Reviews of Geophyics lavas releasegases to the atmosphere and hydrosphere which (volume 25, numbers 2, 3, 5, and 6, 1987). also provide clues about the nature of the source material at Volcanoes have played a significant role in the development depth. of geology since Abraham Werner at the beginning of the last The theme common to the different parts of this review century concluded that all rocks were formed by precipitation relates to the conditions for melting, transfer, storage, and from an ocean, with volcanoes being simply local phenomena eruption of melts. The approach is as follows. Plate tectonics caused by burning coal seams. The challenge by James provides a framework of tectonic environments and internal Hutton, who concluded that magmas were formed by the in- processes,which can be calibrated by laboratory experiments ternal heat of the Earth, began the study of volcanoes as the at high pressures.Within this framework I next outline the overflow edifices for plutonic magma reaching the surface. development of ideas on mantle convection and the contri- Most people who have stood and listened and sniffed at the butions from geochemistrythat identify the separatechemical rim of Aetna or any other active volcano can easily be per- reservoirsyielding basalts.This approach is complementedby suaded that the sounds and smells emanate from the bowels of seismic tomography, which is beginning to provide three- the Earth and that volcanism is the outward and visible sign dimensional pictures of motions in the mantle. The subduction of deep internal processes,of mantle convection, and of the process and the role of volatile components lead to further chemical differentiation of the Earth. differentiation of the Earth and to the formation of continents Volcanoes are beautiful structures,but ephemeral in geo- associatedwith a variety of igneous rocks. Constraints for the logical time, and the progressiveexposure of the deep-seated conditions of melting are provided by thermal structures and plutons feeding volcanoes provides a more detailed history of the calibrations from experimental petrology. After reviewing some critical phase relationships of magmas and source rocks Copyright 1988 by the AmericanGeophysical Union. and recapitulating models of mantle convection,I then consid- Paper number 8R0242. er magmatic processesin selectedplate tectonic environments: 8755-1209/88/008R-0242505.00 basalts at divergent plate boundaries,andesites and batholiths

370 WYLLIE'MAGMA GENESIS, PLATE TECTONICS, DIFFERENTIATION 371

Mid-Atlantic magma generation and volcanoes, and the framework has changedthe way we considermagmatic processes.Among the .Ridge more significant change are the following: (1) it provides a framework of tectonic environments within which different suites of igneous rocks can be located, a framework long sought by petrologists;(2) it provides a dynamic mechanism for moving masses of rock up and down within the Earth, Ocean changing pressureand temperature, causing partial melting, Tre and therefore initiating magmatism; (3) it provides a broader view of igneous petrology in terms of the chemical differ- Pa c, if i c ••••'"'••'•'•'•'•••••••• •'"' ..a_•! • i! 6 !_•_h-e •il...... '••"•••••••'•:•••i''••,,\ entiation of the Earth instead of the preoccupation with the diversity of igneousrocks; (4) it has changed the subject from descriptive petrography and philosophical petrogenesisto a k-Lithosphere Lithosphere--;'study of processesand products firmly rooted in geophysics; (5) it involves processesthat can be calibrated in part by Fig. 1. Outline of plate tectonics,vintage late 1960s. laboratory experiments at high pressuresand temperature; and (6) it provides a mechanism for recycling crustal rocks and volatile componentsback into the mantle, thus generating at subduction zones, ocean island lavas and anorogenic grani- volatile fluxesthat redistributeelements at depth. toides associatedwith mantle plumes, and the subsilicic kimberlites that puncture continental plates. Unravelling the pro- Tectonic Environmentsand Petrographic cessesduring subductionremains one of the major problems Associations for understanding both the origin and evolution of the conti- Figure 1 shows the main tectonic environments:(1) diver- nents and the heterogeneitiesin mantle reservoirs. gent plate boundaries, (2) convergent plate boundaries, (3) oceanicplates, and (4) continental plates. These may be subdi- MAGMA GENESIS AND PLATE TECTONICS vided. Divergent plate boundaries include (1A) oceanic ridges The main variables to be considered in magma genesisare and (lB) continental rift systems.Convergent plate boundaries (1) the compositionsof materials,including both the sourcesof include (2A) ocean-ocean,(2B) ocean-continent, and (2C) mammasand the magmaticproducts (the mineralogyis depen- continent-continentboundaries. Oceanic plates may be (3A) dent on the other variables), (2) the pressure or equivalent stable or (3B) flowing acrossa hot spot. Continental plates depth, and (3) the temperature.These variables can be con- include, in addition to rift zones,(4A) stable platform or cra- trolled in laboratory experiments, providing a static flame- tions,(4B) continentsmoving acrossa hot spot, and (4C) pas- work of phase boundariesfor the Earth materials. The Earth sive margins. is a dynamic system,and changesthrough time exert a strong There are many different kinds of igneousrocks and petro- influenceon processes.MaMmas are generated(1)when bodies graphic associations,but in this review we will consideronly a of rock are transported acrossthe depth-temperature limits of selection,the basalts occurring in environments 1 and 3B; the melting curves by physical convection, (2) through temper- andesites,rhyolites, and granitoid batholiths characteristic of ature increase arising from tectonic conditions, or (3) by de- environments 2B, 2C, and 4B; the kimberlites of environment pression of melting boundaries to lower temperatures by 4A; nepheline-normative basic lavas of environments lB and influx of volatile components or by change in oxidation state 3B; and carbonatites of environment lB. of volatile components with subsequentlowering of the solidus. Physical processessuch as mantle convection transport Chemical Differentiation of the Earth massesof material across phase boundaries and cause chemi- The first chapter of Bowen's [1928] book was entitled "The cal transformations. The flow of fluids through rocks trans- Problem of the Diversity of Igneous Rocks." This was con- ports heat and material and changesbulk composition, caus- sidered to be the primary task of igneous petrology. Bowen ing chemical differentiations.Temperature changesas a func- acceptedas a fundamental thesis the parental nature of basalt tion of processoccurring, and thereforeas a function of time. and the derivation of other igneousrock types from basalt by Time available for reactions also influences the extent to fractional crystallization. He found "the most likely source of which equilibrium is attained in reacting systemsof rock- basaltic magma in the selective fusion of a portion of the magma-vapor. peridotite layer." These statementsare acceptabletoday, but Figure 1 is a simpleversion of the theory of plate tectonics they are not sufficient. Sources other than mantle peridotite as it was presentedin the late 1960's. There are large, rigid can yield magmas, and processesother than fractional crys- lithosphereplates migrating acrossthe surface of the Earth, tallization of basalt are influential in the derivation of some separated along boundaries where adjacent plates are in igneousrocks. Petrologiststoday are more concernedwith the motion relative to each other. Uprise of mantle is associated complementary relationship between magmas and source with midoceanicridges, and sinking of lithosphereand mantle rocks, viewing igneous petrogenesisas one approach to the is associated with convergent plate boundaries. Convection interpretation of the chemical differentiation of the Earth. Dif- within the mantle is implied. We can add to this thermal ferentiation of magma occurs after it is separated from its plumes bringing deep material upward to the plates. Deeper source,and the chemicaland mineralogicalchanges following down, we have the interactions that must occur between core this separation constructscreens between observedmagmatic and mantle providing other heat and material exchanges,and product and inferred source rock. Geochemists,petrologists, probably initiating vertical processeswithin the mantle. Plate and experimental petrologistswork to unravel the paths in tectonicshas provided an excellent framework for studying order to see through those screens. 372 WYLLIE' MAGMAGENESIS, PLATE TECTONICS, DIFFERENTIATION

The Earth is differentiated into core, mantle, crust, hydrosphere, and atmosphere, with the biosphere concentrated in 0.4%(•1CRUST 0'.1 si ....•I_F_eI._.A. LI?.'I;?Rest the boundary layers between crust and the fluid envelopes. MANTLE0 Si ' Mg FelAI The major differentiationwas accomplishedthrough magmat- -: 80 68.• % --Co ic processes.Figure 2 comparesestimated average compo- • 60 -No-- Rest sitions for the core, mantle, and crust. The iron-nickel core o 40 containslight elements,with sulfur or oxygenbeing the main •-e20 3•.5% Fe S Ni candidates.More than 90% by weight of the mantle is repre- F: o "COREI I I I I I , I , I sentedby the componentsMgO-FeO-SiO 2. The crust is en- 20 40 60 80 I00 Wt % elements riched in Si, A1, Ca, Na, and K, with Mg being left behind in the solid mantle when partial melting occurs.The mantle and Fig. 2. Averagechemical compositions of core,mantle, and crust. crust are both heterogeneous.Heterogeneities are produced in See Wyllie [1981] for data sources.Oxygen is now a candidatefor the mantle peridotite by partial melting and extraction of melt, light elementin the core. passage,or eraplacementof melt through or in mantle regions removed from the melt source, by the action of metasoma- ingthat the lithosphere plates that transport d{ifting conti- tizing fluids, and by the recyclingof crustal materials through nents may be 400 km thick beneath the continental shields. convergent plate boundaries. Magmatic processesduring the Other seismologistsconclude that systematiccontinent-ocean growth of the Earth may have have accompaniedby the for- differencesdo not appear to extend below 200-kin depth. The mation of a magma ocean causing differentiation of the low-velocity layer beneath the East Pacific Rise, however, does average composition of Figure 2 into two distinct rock types, extendas deep as 400 km. The ratio Vp/Vscannot be explained peridotite and piclogite(see Figure 4c). by changesin chemistry or temperature alone, and at least in Magmatic processesare also responsiblefor the diversity of the upper 200 km partial melting is believed to be involved; igneous rock types within the crust. The extreme chemical the differencein properties extending to 400-km depth may be differentiation occurring within the thin veneer of sediments causedby material rising to generatethe mid-ocean ridge ba- that blankets the solid surface was accomplished by weather- salts. ing and transportation and redeposition of material by inter- The velocity gradients between 400 and 650 km are higher action between the fluid envelopes and the solid surface. than expectedfor a homogeneousself-compressed region. The Water and other fluid components also played an important 650-kin discontinuity is a good reflector of seismic energy, role in the extreme geochemical concentrations associated requiring that its width be lessthan 4 km. Another significant with the formation of ore deposits, either with igneous rocks observation is that no earthquake loci occur deeper than 700 or within sedimentary bodies. Igneous petrology thus involves kin. The changesin physical propertieslocated by seisinology the chemical differentiation of the Earth, and the mantle is must be explained in terms of either chemical composition or thus an essentialpart of igneous petrology. There is growing phase changes, solid-solid transformations, or melting reac- evidence that physical and chemical interactions between tions. solids and fluids (melts and vapors) are important in deep We are on the verge of dramatic increasein our understand- crust and mantle, as well as in the more obvious convection ing of the Earth's interior, based on improved seismictech- cells associatedwith plutons, and within the oceanic crust. niques,improved computers,and the deploymentof new digi- tal seismicarrays. Three-dimensional mapping of the interior GeophysicalRoots of Volcanism is progressing rapidly with techniques similar to medical to- Petrology is rooted in geophysics.The mantle is the source mography being used for imaging with seismic body waves of all good things. It is the dynamics of the mantle that moves and surfacewaves. This representsa dramatic breakthrough in lithosphere plates causing plate interactions associated with solid Earth physics and chemistry, as discussedin more detail specific petrographic associations, initiating magma gener- below. ation within the mantle or within subducted oceanic crust or within the continental crust through upward transfer of heat Calibration at High Pressures and material. It is the thermal evolution of the Earth and the and Temperatures consequent movement of material that controls petrological The classicalapproach to petrology is to study rocks at the processes. surface,analyze them in detail, and develop hypothesesabout Seismologyhas been the major source of information about what happeneddown below. A basic tool is the geological the Earth's interior, and fruitful developments have been hammer, and the petrologist attacks mountains of rock with a achieved by cross fertilization between seismology, experi- small pieceof steel.This approachis now supplementedby the mental petrology and phase equilibria, and mineral physics. experimental approach, calibrating at high pressuresin the Increasingly precise seismic data are producing more infor- laboratory the reactions expectedto occur at depth. In the mation about the structure of the mantle. High-resolution ve- experimentalapproach the petrologisttakes a small sampleof locity profiles have been determined for various tectonic prov- mineral or rock and crushesit together with a furnace in a inces. The lithosphere overlies a low-velocity zone commonly massivesteel press. For even higher pressures,a more delicate correlated rather loosely with the asthenosphere. There are approachis adopted usingeven smaller samplessqueezed be- discontinuities at depths of about 200, 400, and 650 km. There tween two minute diamond anvils, heated by laser. If these is some evidence suggestingthat between ocean basins and approaches fail to achieve the required conditions, the more continental shields, thermal or compositional differences or drastic approach of shock-waveexperiments is adopted and both persist to depths of at least 400 km. Travel times of S the sample is blasted, yielding its secretsonly during the last waves from deep-focus earthquakes to oceanic islands are hot nanosecondsof its existence[e.g., Williams et al., 1987]. about 5 s later than the times to continental stations,suggest- With the calibrationsof laboratory experimentson suitable WYLLIE: MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION 373

the origin of a particular rock, thus dealing with specific,more localized petrological problems. A recent report from the Na- tional Academy of Sciences[Committee on Physicsand Chemis- try of Earth Materials, 1987] noted that the base of phase 200 600 equilibrium data remains inadequate. Many of the phase equilibrium data in use to 30-40 kbar are based on results obtained 10-20 years ago, when experimental difficulties were being discovered and sorted out and when microbeam instru- / !iil I ments for analysis of products were inferior in comparison with those of today. Critical data on element distribution iilt 1 among phases are still fragmentary. On page 75 the report •-•oo• ß states: "The great tragedy of the present situation is that there is a general perception that such information is already well / /:.:'iiiiiiiii! doo known." In fact, the lack of quantitative information "severely cripples all attempts to model accurately the chemical evolution of magmas during ascent." A glance ahead to Figure 33a shows the range of results reported for the solidus of perido- 0 1000 2000 3000 tite. Temperoture,øC Figure 3 expands the data available for peridotite from 60 Fig. 3. Experimentally determined phase diagram fo• peridotitc to 200 kbar, and recent work with the diamond anvil cell has ET•k•h•shi, 1986]. Abbreviations: LHZ, ]hc•zolitc; P], plagioclasc; provided information about melting in the lower mantle at Sp, spinel; Ga, garnet; O1, oilyinc; Cpx, clinopy•oxcnc;• and 7 pressures greater than 200 kbar [Heinz and Jeanloz, 1987]. high-pressurepolymo•phs of olivine. The stippled band is the melting These represent significant advances for high-pressure calibra- interval. tion of mantle processes,but the authors of these new results will surely not be offended by the suggestion that in due materialsat high pressuresand temperatures,we hope to pro- course, these pioneering results will also need revision as the duce a static picture of the structureof the mantle, with com- experimental problems are identified and corrected. position and mineralogy matching the measured physical properties. The interesting geophysicaland petrological pro- SOURCE MATERIALS, MANTLE RESERVOIRS, cesses,however, involve a dynamic mantle, and knowledge of AND CONVECTION the structure is incomplete without information on mantle The bulk compositionsof Earth, mantle, and crust provide convection. Inferences about convection are related to the dis- important limits on petrological processes.More important tinct chemical reservoirsrequired by interpretation of the geo- are the compositions and mineralogies of the specific rocks chemistryof basalts,and the general problem is further com- constituting the mantle and crust. There are three major poundedby the invocationof mantle metasomatismto adjust sourcesfor igneous rocks, which constitute separate magmatic mantle geochemistryto fit the observationsin lavas. There is a hearths: (1) the mantle, (2) subducted oceanic crust, and (3) large variety of igneous rocks representinga variety of pro- continental crust. Each hearth is heterogeneous containing cessesoccurring within the upper mantle and crust. These too several rock reservoirs, but each has its distinctive geochemi- are calibrated by experimental studies on suitable materials cal characteristics.In fact, more than one source is commonly under appropriate conditions. involved in magmatic processes,and the mantle also contains The phase relationshipsof mantle peridotite are reviewed components recycled from sources(2) and (3). A great deal of below, but Figure 3 is presented here as an example of the attention has been directed toward distinguishing what pro- experimental calibration of the mantle or of that part of it portions of a magma are derived from each of these sources composedof peridotite, specificallylherzolite. There are sub- and reservoirs, as part of the problem of working out the solidus phase changesinvolving increasesin density with in- transfer of chemical elements from one reservoir to another, creasingpressure, and at pressurescorresponding to a depth and the history of crustal growth [Alldgre, 1987]. of about 650 km, where there is a seismic discontinuity, it appears that most mantle components find a comfortable home in the mineral structure of perovskite. The solidus curve Mantle Sources for the beginning of melting is the most important phase Cosmochemistryis an important approach to mantle com- boundary in connection with chemical differentiation of the position giving information about the composition of the Earth. The composition of the melt produced within the melt- whole mantle, a somewhat fictitious average, primordial ing interval above the solidus varies as a function of pressure mantle [Wasserburg, 1987]. The approach is model depen- and temperature.A basic knowledge of the phase equilibria of dent, requiring assumptions about the origin of the solar mantle and crustal rocks is required in order to trace the system.There are currently two significantlydifferent interpre- generation of melts and their chemical differentiation as they tations of mantle mineralogy, involving the distribution of ma- move from high- to lower-pressureregimes. terial within the mantle, with several variants. One of the main features of this review is to present a Until the late 1970s, most models for the mantle were based samplingof the phase diagramsused to calibrate geophysical on an average composition corresponding to peridotite, with and petrologicalprocesses. Many of thesediagrams are quite chondritic trace elements[Ringwood, 1975]. The main exercise old and are generalizedto cover general processessuch as the was to fit the major and minor elements into a suitable se- partial melting of mantle or crust. Many of the more recent quence of phase transitions to satisfy the seismic data on phasediagrams have been determinedto place constraintson properties and discontinuities, although there were some 374 WYLLIE: MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION challengesto the concept of a fairly uniform composition from lating with the geophysical efforts to map convection in the core to crust, with proposals that the lower mantle has higher mantle. Si/Mg ratio than the upper mantle. Interpretation of the rock Simple petrogenetic schemes related to the derivation of compositions in terms of mantle compositions involves either basalt by partial melting of peridotite fail to explain the abun- mixing models involving basalt and residual peridotite to re- dances of trace elements associated with the major elements. constitute pyrolite or attribution of the composition of the The geochemistryof trace elementsin many basalts appears to "most primitive, undepleted peridotite" to the composition of be decoupled from that of major elements. Basalts are said to the whole mantle. Results in terms of major elements (Figure be depleted, undepleted, or enriched in trace element con- 2) are broadly consistentwith the cosmochemicalapproach. centration patterns, trace element ratios, or isotopic ratio sig- A new approach by Anderson [1983] is independent of the natures. No single definition of these terms can encompassall assumptions made in other petrological and cosmochemical of the usages.For example, a low Nd isotope signature may approaches. He estimated the composition of the primitive accompany an enriched rare Earth element (REE) pattern. mantle by mass balance calculations that do not make as- These descriptive geochemicalterms are also assignedto the sumptions about basalt/peridotite ratios (pyrolite model) or source regions of the basalts with a view to mapping mantle the large ion lithophile (LIL) components of these compo- reservoirs. The term "depleted" is also used for major and nents. Nor does the approach assume that Rb/Sr, K/U, and minor elementsor their ratios. A peridotite is said to be deple- other ratios must be the same in crust as in mantle. Anderson ted in basaltic components by the extraction of basalt, giving treated the primitive upper mantle as a four-component the peridotite a higher Mg/Fe raio. Peridotites from the system, involving crust, peridotite, LIL-depleted basalt mid- mantle give us direct measurementsof the chemistryof poten- ocean ridge basalt (MORB), and a LIL-enriched component. tial source rocks. Peridotite xenoliths in alkali basalts and In order not to prejudge the major element chemistry of the kimberlites are described as depleted (or "barren" or "infer- mantle, two ultramafic components with different oli- tile," or "refractory") and "fertile," depending on their con- vine/orthopyroxene are needed. Basalt is decomposedinto a centrations of basaltic components.They are also describedas depleted component, MORB, and an enriched component depleted or enriched according to their trace element and iso- with composition KIMB, an average kimberlite. The compo- topic characteristics.A peridotite may be depleted in basaltic sitions of many basalts can be approximated by mixing the components yet enriched in some incompatible trace elements two componentsMORB and KIMB. which should have been carried away by the basalt fraction. The equations solved by Anderson defined the weight frac- The lack of correlation between isotopic composition and tions of each of the five components that yields chondritic incompatible elements in many volcanic rocks is now com- ratios of the 18 refractory oxyphile elements. Element con- monly explained by metasomatic alteration of parts of the centrations normalized to carbonaceous chondrite abun- upper mantle. The mantle is said to have been metasomatized dances show good agreement with the cosmochemicalabun- to the extent necessaryto explain the geochemistryof the lava dances. The results show extensively differentiated Earth and or nodule. Metasomatic fluids rich in LIL elements are as- efficient upward concentration of the incompatible trace ele- sumed to have percolated upward through the mantle, enrich- ments into the crust. Anderson [1983] therefore concluded ing the peridotite by altering the clinopyroxene composition that the whole mantle has contributed to upper mantle and or by crystallizing new hydrous minerals and other minerals. crustal elemental abundances and that the lower mantle is The nature of the fluids is rarely defined, but recently the depleted in the incompatible elements, including K, U, and distinction between silicate melt and hydrous solution has Th. If this depletion occurred during the early differentiation been more consistentlymade. The effects of metasomatismin of the Earth, it need not imply present-day communication the mantle can be examined in samplesfrom peridotite massifs between upper and lower mantle. and peridotite xenoliths. These include refractory rocks which The observed or derived seismic and elastic properties for nevertheless contain concentrations of compatible elements the Earth should be matched by the aggregate properties of and light REE/heavy REE ratios greater than chondritic. The the mineral assemblageat each depth. According to Anderson evidence for depletion of elements assignedto partial melting, and Bass [1986], an olivine-rich mantle can be ruled out be- followed by enrichment in trace elements in a subsequent tween 450- and 650-km depth, whereas an eclogite mantle is metasomatic event, is well displayed. permitted by the data. Both pyrolite and eclogitemineral as- Recent discoveries of significant variations in geochemistry semblagessatisfy the seismic properties between 200- and and petrology of basalts on local scalessupport the concept of 400-km depth. The lower mantle has seismicproperties consis- a mantle heterogeneous on a small scale. These topics were tent with (Mg, Fe)SiO3 in the perovskitestructure. reviewed by Alldgre [1987], Carlson [1987], McCallurn [1987], and Marsh [1987]. The geochemistry of basalts from Geochemical Reservoirs an individual oceanic island may be distinguishablefrom that Basaltsand other magmasare derived from the mantle, and of basalts from other islands. Recent detailed bathymetric variations in trace elementsand in particular the variations in maps of the ocean ridge system have revealed tectonic and isotopic compositionsof Nd, Sr, Pb, and rare gasesindicate constructional features on the scale of a few meters, including that there are distinct sources or reservoirs in the mantle that a series of segmented ridge sections, each with a topographic have remainedgeochemically distinct and thereforephysically high probably representinga volcanic center, with overlapping separatefrom each other. It has been demonstratedby geo- ridge axes, propagating ridge tips, and transform faults. There chemicalstudies that pervasive,long-lived chemicalheteroge- are significantgeochemical and petrologicaldifferences among neitieshave persistedin the mantle on time scalesof a billion basalts associatedwith these local tectonic parts of the oceanic years,and the exerciseof mappingthe heterogeneitiesin space ridges, as well as along the ridges. The evidence suggeststhat and time remains a major objective of geochemistry,corre- mantle heterogeneitiesmay persist on the scale of tens of ki- WYLLIE: MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION 375

CFB CFB MORB

A. 2-LAYER CONVECTION B. 1-LAYER CONVECTION

C.F...B...... •cGMI:• Enriched CGRM•::: CF.B:..:.:•i:.:.l••.:...,.....:...MORB •_ De pie te d

:.::::iiiiiiiii•!!!i:!:::::: ...... ii;/""•..:.:.:.:.:.:.;,.,.....,...... i):.,:...... ,..,., , ,,, . . , , , , , , _ ., , , C. PICLOGITE LAYER D. HARZBURGITE MEGALITHS Fig. 4. Four contrastingmodels for mantle convectionproposed mainly on the basis of geochemistry.See text for sources. lometers beneath the ridges, and there is some evidence of expanded to 3, 4, and 5 components as the number of isotopic metasomaticallyenriched regions under the ridges. systems used simultaneously expanded from Sr and Nd to A_lld•lre[1987] has presenteda comprehensivetreatment of include Pb, Hf, He, and Xe." The mantle is chemically hetero- the Earth as a systemdivided into severalreservoirs, consider- geneous,different sourcecomponents have been distinguished, ing the Earth's behavior to be described as a set of distri- and the exerciseof relating these reservoirsto physical layers, butions of chemicalelements constantly modified by a set of strata, or blobs within the mantle has led to a proliferation of operators which characterizethe principal geologicalfunc- convection models. The blank interior of Figure 1 has been tions. All•gre has identified five distinct reservoirs on the basis replaced by a variety of possibilities,some of which are sum- of multi-isotope studies:(1) depleted and outgassedmantle, marized in Figure 4. the source of MORB's; (2) metasomatized continental litho- Figure 4a representsthe two-layer mantle proposedby De- sphere,the source of continental flood basalts; (3) a mixed Paolo and Wasserbur•t[1976, 1979] and developed by Jac- reservoir,the sourceof oceanicisland basalts, containing por- obsenand Wasserburg [1981], who presented a cartoon of the tions of 1 and 2 above and subducted oceanic slab with crust mechanism for crustal growth according to their model I. and lithosphere; (4) undepleted, nonoutgassedpart of the Figure 4a shows the continental crust and a thick layer of mantle; and (5) the segmented,enriched continental crust into depleted mantle overlying undepleted primitive mantle, from which elements incompatible in the mantle have been con- which the two overlying reservoirshad been derived by mag- centrated. matic processesand differentiation. The major mantle reservoir, the source of mid-oceanic ridge basalts, is depleted in Mantle Convection incompatible elements. According to the model of Jacobsen Figure 1 shows a simple interpretation of plate tectonics, and Wasserbur•t [1981], the depleted reservoir has grown emphasizingthe tectonicenvironments demarcated by moving downward through time at the expenseof primitive, undeple- plates with plate boundariesas the loci of major geological ted mantle. The boundary between the two reservoirsat pre- activity. The conceptsof seafloor spreadingand subductionof sent approximates the depth of the 650-km-deep seismic dis- oceanic plates involve uprise of hot mantle material and sink- continuity, as well as the depth at which the mantle minerals ing of cold plates, which leads smoothly into the formulation are transformed to perovskite (Figure 3). Ocean island basalts of models for mantle convection. The formulations were not (OIB) and continental flood basalts (CFB) are enriched in uniformly acclaimed.Powerful argumentswere producedby incompatible trace elementscompared with MORB. Jacobsen authoritative scientistsduring the 1960s objecting to the idea and Wasserbur•t [1981] proposed that these basalts were of convectionin the viscous,solid mantle. Sullivan [1974, p. derivedfrom blobs of undepleted,primitive mantle rising from 84] referredto a rejectionslip from a scientificjournal citing a the deeper layer. Island arc basalts (IAB) are more compli- refereewho said, "This is another in a number of papers by cated because of the subduction of oceanic crust, and more Runcorn on convection in the mantle and I hope it will be the than one source is involved. The question arises as to what last." The objections faded as geochemical studies of the extent the enriched characteristics of the IAB are due to sub- mantle and crust provided firm evidence for the existence of ducted sources, to mantle source reservoirs, or to contami- distinct reservoirs of magma sourcematerial that had retained nation of the lavas by lithosphereor crust. their separateidentities for a billion years or more, which was Figure 4b represents an opposing interpretation without explained in terms of different convecting regions within the separate convecting layers but with mantlewide convection. mantle. Carlson [1987], in his review of the geochemicalevo- Hofrnann and White [1982] proposed that subducted slab ma- lution of the crust and mantle for the period 1983-1986, noted terial sank to the mantle-core boundary where it was stored that "two component compositional models for the mantle until a thermal plume lifted it to higher levels where chemical 376 WYLLIE: MAGMA GENESIS,PLATE TECTONICS, DIFFERENTIATION heterogeneities were transferred to the magmas generated. zolite of the lower lithosphere is stripped off during subduc- Several studies have reported seismic anomalies extending tion and resorbed into the upper mantle, which is thus regen- into the lower mantle deeper than the seismic zones con- erated as the depleted source reservoir for MORB (heavy stip- sidered to delineate the subducted lithosphere [e.g., Creager pling). The depleted, cool, subducted lithosphere forms a and Jordan, 1984]. If the subducted slabs penetrate deep into second reservoir of megaliths composed of harzburgite en- the mantle as indicated in Figure 4b, two-layer convection is closing eclogite from oceanic crust, density-balanced at about prohibited, and mantlewide motions are required. Carlson 650-km depth between peridotite above and perovskitite [1987] summarized arguments from geochemistry and geo- below. The megaliths float like icebergs at this level, gradually physics against the two-layer mantle, as well as problems flowing to produce a continuous layer. The megaliths reach facing mantlewide convection. The physical interpretation of thermal equilibrium with surrounding mantle on a time scale Figure 4b involves increasing viscosity as a function of depth of 1-2 billion years, and at this stage the harzburgite rises [Davies, 1984; Hager, 1984]. The effect of the viscosity in- buoyantly forming mantle plumes. The depleted harzburgite crease is to reduce the rates of flow and mixing in the deeper has been rendered newly fertile by liquids generated by partial mantle, so that despite the mantlewide convection the upper melting of the entrained eclogite. The plumes provide the and lower mantle tend to remain separate from each other sources for OIB an CFB, and incorporation of the enriched through long periods. harzburgite into the lithosphere yields the alkaline basaltic In Figure 4b the mantle is composed of peridotite, with suite by small degreesof partial melting. Eclogite in the mega- phase change to perovskite at the 650-km level, as indicated liths may become converted to perovskitite, and sink down by the dashed lines. The flows pass through the phase transi- through the lower mantle. tion region. Distributed through the mantle are blobs and In a mantle with layers convecting separately, it has been stringersor layers of other material, carried into the mantle at suggestedthat motionsare related through the boundary layer convergentplate boundaries. These heterogeneitiesvary with (Figures 4a, 4c, and 4d). The cool subductedlithosphere reach- respect to their contents of incompatible and radiogenic ele- ing the 650-km level, for example, could initiate downwelling ments. Basalts are generatedby partial melting of the convect- in the layer below, and the upwelling associatedwith mid- ing mantle, or of isolated heterogeneities,or of a mixture of oceanic ridges might be initiated by plumes rising from the host mantle with some enclosed heterogeneity. The more core-mantle boundary in the layer below. This kind of cou- radiogenicmasses enriched in subductedcrustal material may pling implies opposingflows above and below the boundary become heated after a billion years or so and deliver mantle layer in suchlocations, with the developmentof shear. plumes [Hofmann and White, 1982]. These plumes are as- All•gre's [1987] interpretation contains elements of the sumed to be the sources of OIB and CFB. Partial melting to models in Figures 4a and 4d, but his schemewas determined produce MORB is much more extensive, and the contri- on the basis of multi-isotope studies. The Earth's evolution is butions of trace element and isotopic signatures from hetero- treated as a transfer matrix made of transfer functions that geneousenclosures is correspondinglydiluted. describe the transfer of mass among reservoirs. The upper Figure 4c illustratesAnderson's [1982] more complexmodel mantle, the source of MORBs, is depleted, 99% outgassed, for the recycling of oceanic crust, based on the geochemistry and well mixed by convection,and it may extend down to the of basalts, kimberlites, and peridotites and consistent with 670-km seismic discontinuity, a convenient, known horizon, seismic data. He suggestedthat extensive partial melting of coinciding with a large density change associatedwith the primitive mantle during accretionof the Earth was followed phasetransition to perovskite(Figure 3). The limit of the out- by formation of a cumulatelayer (unshaded)of olivine eclogite gassedmantle does not coincide exactly with the limit of the (piclogite), with an overlying peridotite layer (stippled) en- depleted mantle. The lower mantle has been outgassed to riched by residual kimberlite fluids from below. The depleted about 17% and is thus more primitive [All•gre et al., 1987]; it peridotite from which the basaltic melt was removed consti- is probably convecting.In theserespects, All•gre's model cor- tutes the lower mantle, precisely the opposite of the model in respondsto the "standard model" of Figure 4a, but there are Figure 4a. The upper mantle of peridotite composition con- additional complexities. The source of continental flood ba- tains a density-balancedlayer of piclogite perched between salts is identified as metasomatized continental lithosphere, 220 and 670 km. The low-velocity zone is attributed to incipi- which was probably formed early at the same time as the ent melting. The piclogite layer is a depleted reservoir, the continents and has remained below the continents as the source of MORB, and the upper peridotite layer is the reser- plates have migrated. The enormous dispersion of chemical voir enriched by kimberlite fluids giving it the geochemical data for the ocean island basalts involves a mixture of MORB requirements of the source of OIB and CFB. Convection source and at least four different end-members. All•gre places within the eclogite layer deforms the upper boundary, raising this mixed sourcein a mesosphereboundary layer between the it above its solidus temperature. Partial melting produces a upper and lower mantle, resembling Figure 4d in geometry. buoyant diapir with adiabatic ascent leading to extensive The layer, fed by depleted material from the subductedocean- melting and eruption at mid-oceanridges. The eclogitelayer is ic crust and by delaminated continental lithosphere, is continuously replenishedby subduction of the oceanic crust warmed by conductionfrom the lower mantle and then mixed derived directly from the layer some millions of years earlier. back into the upper mantle by convection, producing a Ringwood [1982] proposed a more elaborate two-layer "marble cake" mantle with the mixed material deformed by mantle illustrated in Figure 4d, with an additional reservoir elongation. Large blobs from boundary layer instabilities rise occurring as megalithsthat flowed to form a layer separating through the convecting upper mantle and yield the ocean the two convectinglayers (solid shading)and with eclogite(no island basalts. The largest of these blobs may entrain lower shading) playing a special role. The extraction of MORB at mantle sources, as suggestedby the rare gas analyses from mid-ocean ridges generatesa depleted lithosphereof harzbur- Loihi seamount [All•gre et al., 1987]. The formation of conti- gite and lherzolite.Ringwood suggested that the depletedlher- nental crust follows scavenging of subducted oceanic litho- WYLLIE: MAGMA GENESIS,PLATE TECTONICS, DIFFERENTIATION 377 sphere for incompatible elements by dehydration and partial need to take into account the flow that the density variations melting. Each new crustal segment is a mixture between ma- induce in the mantle. Fluid mechanical models were used to terial newly extracted from the mantle and recycled old conti- compute the dynamic topography caused by mantle convec- nental material. Continental collisions destroy continental tion at the Earth's surface and the core-mantle boundary. The massesby doubling the crust and exposing it to erosion. The dynamically maintained relief at these two levels represent behavior of volatile components is important in processesof large mass anomalies which contribute to the geoid. These crust formation. anomalies overwhelm the effects of the internal density contrast driving the flow and thus determine the sign of the geoid. Mantle Tomography and the Geoid Hager and Clayton [1987] considered in addition the effects of Recentdevelopments in tomographyare beginningto yield subducted slabs and viscosity variations. From many calcula- a three-dimensionalpicture of density variations attributable tions, they concluded that the best fit was obtained with a to temperaturedifferences and related directly to mantle flow nonlayered mantle with an increase in effective viscosity from and convectionpatterns. These results,correlated with fluid the asthenosphere to the lower mantle by a factor of 300. A mechanicalmodels and geoid calculations,promise rapid ad- layered mantle like the mantles in Figures 4a, 4b, and 4d does vancementin our understandingof mantle structure and con- not satisfy the data. A mantle with such a change in viscosity vectionand in interpretationof mantle geochemistry.The geo- with depth, however, could lead to a compositionally stratified chemistry,when correlated with present mantle convection mantle. Rates of flow would be much higher in the upper patterns,may be appliedmore readily to ancientmantle dy- mantle than in the lower mantle, and elements or geochemical namics. reservoirs would have a much longer residence time in the Surfacewaves from large earthquakescircle the Earth many lower mantle than in the upper mantle. Figure 4b represents times, and different wavelengthsof the surfacewaves sample the kind of mantle favored by the tomography and geoid differentdepths. Average velocities along many paths are con- studies, and Hager and Clayton [1987] refer to a lower mantle verted to three-dimensionalimages of the seismic velocity with the structure of "ribbon candy" rather than that of the structure,in specificdepth intervalsdown to 900 km. Present more familiar convection experiments.All•gre [1987] similarly data are sparseand resolvingpower is limited, but character- refers to a "marble cake" mantle. Another prediction from the isticlengths are relevantfor plate tectonics.Global mapshave new investigations of the geoid is that there is about 5-km been preparedshowing lateral heterogeneities,anisotropy re- relief at the core-mantle boundary, which has implications for lated to horizontal flow directions, and anisotropy related to the Earth's magnetic field, magnetic reversals,and the orienta- vertical flow directions [Woodhouse and Dziewonski, 1984; An- tion of the Earth's spin axis. derson, 1987; Tanimoto, 1987]. The seismic velocities to Density inhomogeneities in the mantle grow and subside, depthsof more than 100 km remain closelycorrelated with depending on the location of continents and subduction zones. major tectonicfeatures of the surface.Correlation with conti- The resulting geoid highs change slowly. Anderson[1984] has nental shields indicates that continents have roots of 200-300 speculatedthat continental stability leads to the development km that have drifted with the continents. Slow material be- of thermal and geoid anomalies, followed by a shift of the neath oceanicridges is distinguishedto 400 km. Fast material Earth's axis of rotation relative to the mantle. The geoid beneath island arcs representingcold subductedlithosphere would evolve stably and slowly, while the lithosphere would beginsto loseits closecorrespondence with surfacefeatures at break up and drift away from the thermal anomalies that the depthsgreater than 250 km. There is a dramatic changeat continents had caused. The continents would drift toward the about 450 km comparedwith the pattern of shallowerdepths. colder parts of the mantle, toward the geoid lows, and they Two methods have been used to invert P wave travel time would make the mantle cold as they overrode the oceanic residual data sets to obtain the velocity structure of the lower lithosphere. These speculations have implications for mantle mantle. Dziewonski [1984] has expressed velocity pertur- and igneouspetrology. bations in terms of smooth functions. R. W. Clayton and R. P. Comer (in the work of Hager and Clayton [1987]) adopteda Oceanic Crust, Continental Crust, and tomographicmethod with iterativeback projectionsof travel Volatile Components time residualsalong ray paths.The advantagesand disadvan- The crust, the atmosphere, and the oceans are reservoirs tagesof eachinversion method were reviewedby Hager et al. unmixed from the mantle and core. The reincorporation of [1985] and Hager and Clayton [1987]. There is encouraging oceanic crust into mantle reservoirs has been outlined in con- agreementbetween the two methodsin many cases.Hager and nection with Figure 4. During the processof subduction, how- Clayton[1987] reviewedthe velocityvariations displayed in a ever, the oceanic crust composed of basalt/gabbro/eclogite, at series of mantle cross sections and maps for specific depths, least partly hydrated to amphibolite and enclosing bodies of correlatingthem with surfacetectonic features as far as possi- serpentinite and with some sediment attached, is a separate ble. One of the most striking characteristicsin the context of magmatic hearth, a distinct chemical reservoir. It is subjected mantle convection is the radial variation in velocity structure. to dehydration, decarbonation, and partial melting reactions, Coherent structurescommonly have shallow dips. with the volcanic island arcs or continental margin volcanic Hager [1984] and Hager and Clayton [1987] have worked arcs being the most obvious products. with the long-wavelengthnonhydrostatic geoid which pro- The deep continental crust, composed largely of gabbro, vides a fundamental observational constraint on mantle con- tonalite, and granite with some sediments (together with their vection,and they demonstratedthat most of the variations are metamorphosed equivalents) with average composition close ultimately causedby density contrastsin the lower mantle. to that of andesite, is another magmatic hearth yielding more The tomographicresults were usedto infer a densityfield and silicic partial melts with distinctive geochemical signatures. to calculate geoid anomalies.The calculatedgeoid was op- Armstrong [1968] has concluded on the basis of isotope stud- positein sign from the measuredgeoid, which indicatedthe ies that continental crust is regularly recycled through the 378 WYLLIE:MAGMA GENESIS, PLATE TECTONICS, DIFFERENTIATION

A. REGIONAL METAMORPHISM B. CONTINENTAL COLLISION

He CO2 H20

100krn He CwH--O

C. SUBDUCTION D. METEORIC WATER AND RIFT

Sea

Oce.c CO, Lithosphere "'•• •-

Fig. 5. Four different tectonicenvironments involving migrations of large volumesof volatile components.

mantle at convergentplate boundaries.Schreyer et al. [1987] from the Earth's mantle during its early history, and Figure 5b have describedpyrope-coesite rocks in the Western Alps indi- shows primordial volatile components rising from the deep cating that continental sedimentary rocks were subducted to mantle. The major componentsare probably C-H-O, but their depths of about 100 km at 700ø-800øC.These exampleswere molecular form is controversial, depending on conclusions elevated to the surfaceagain, but an alternative fate is disper- about the oxygen fugacity in the mantle as discussedin a later sal through the convectingmantle. section. All•gre et al. [1987] concluded from rare gas system- It has been established that some CFB has been contami- atics that about half of the mantle is 99% outgassedand that nated by its passage through continental crust, and the geo- a very early intense stage of degassingwas followed by a slow chemical effects are similar to those that have been attributed processwhich continuesto the present. to metasomatism of a mantle source rock. There remains un- Continental collisions or subduction of oceanic crust at certainty about interpreting the trace element and isotope convergentplate boundariesare associatedwith the releaseof variations observed in basaltic lavas in terms of the relative abundant fluids. Figure 5b representsthe dissociation reac- contributions of mantle metasomatism, crustal contamination, tions associatedwith continental overlap like that occurring or recyclingof crustal material through the mantle. beneath the Himalaya at present. The overlying continental The distribution and effectsof volatile componentsare par- crust is flushedby large quantities of H•.O and CO•. released ticularly significant in magma genesisassociated with subdue- from the sedimentary cover of the subductedcontinental crust. ted oceanic crust and the continental crust, but they are also This is related to the processesdescribed by Schreyer et al. assignedsignificance in mantle metasomatism.The migration [1987], who emphasized that magma generation in conti- of volatile components as vapors or solutions transfers ele- nent/continent collision zones is influenced or even governed ments from one source reservoir to another, changing their by the fluid and melt production within slabs of continental geochemicalsignatures. H20 is the most abundant volatile crust that may be subductedto considerabledepths within the componentin crustalprocesses, but CO 2 is also plentiful,and mantle. other components such as halogens and sulfur, although pres- Figure 5c illustrates the general location of vapors released ent in much smaller proportions, may be very influential. from subducted oceanic crust. Many of the volatile compo- Figure 5 summarizes some of the ways in which volatile com- nents make their way back to shallower levels along the ponents are made available for participation in magmatic pro- boundary zone between subducted slab and overlying mantle cesses. or crust, and this effect is surely dominant at shallower levels, Figure 5a represents the occurrence of dissociation reac- but it is generally assumedthat at deeper levels there is signifi- tions releasingmainly H•.O and CO2 from sedimentaryrocks cant migration of volatile components into the overlying as they are buried and as isotherms rise during regional meta- mantle. This assumption may require modification if the morphism. Similar reactions occur in other environments recent research by Watson and Brenan [1987] on the wetting when physical conditions change so as to bring rocks across characteristicsof CO•.-H20 fluids leads to a firm conclusion phaseboundaries. The occurrenceof helium in basaltsis taken that such fluids in the upper mantle are incapable of migration as evidence that not all volatile components were expunged except by hydrofracture. WYLLIE.' MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION 379

H20 CO2 boundaries with a sequence of processesinitiated in the subducted slab (by dehydration or partial melting), and at ocean- hydrosphere carbonatiteI continent plate boundaries these magmas are joined by rhyo- oceanic basalt lites escaping from granitoid batholiths (GR). Kimberlites rise from volatile-enriched mantle near the asthenosphere- metasomatism lithosphere boundary beneath cratons. Various nephelinitic 002 magmas are associated with rift zones in continental plat- subduction•C?c 0 concentratedforms, and carbonatites may be formed by their differ- mantle )C' mantle entiation. We aim to determine the petrological and geophysi- trace H20 C02 ! G-H-O various ' cal processeslinking the observed magmatic products and the reservoirs inferred mantle and crustal reservoirs, stripping off the effects of mixing or contamination, and differentiation of magmas. Fig. 6. Schematic diagram illustrating magmatic sources and selected products at divergent plate boundaries, convergent plate The interpretation of geochemical data with respect to boundaries, and continental plates. mantle convection continues to be discussed,as outlined by Carlson [1987]. The situation remains about the same as summarized by Davies [1984, p. 6017], except that there are now Ocean water penetrates deeply into oceanic crust near the even more data to be interpreted: midocean ridges, causing important chemical changesin the compositionsof both crust and water. The emerginghot water With the current proliferation of isotopic systemsand data being forms spectacularchimneys with their black and white smok- applied to the problem of the structure and evolution of the ers, depositing concentrations of minerals, including sulfide mantle, any attempt to synthesizea model consistentwith a large variety of geophysical and geochemical data risks being rapidly ore deposits. Wickham and Taylor [1987] have concludedon outdated. However, it seemsnot to be widely appreciated that as the basis of detailed petrological and oxygen isotope study of some models of the recent past have been elaborated to accom- a sequenceof metamorphic rocks and granites in the Pyrenees modate new data, some of their basic assumptions have been that the granitic magmas were derived by partial melting in- progressively undermined by the same data. Furthermore, many duced in the metamorphic rocks by the flux of meteoric water of the geophysical and geochemical constraints are in apparent conflict, and there has been, as a result, a proliferation of mu- penetratingas deep as 12 km. They suggested,as illustratedin tually exclusive models [e.g., Wasserburg and DePaolo, 1979; Figure 5d, that crustal rifting was responsibleboth for the Alldgre et al., 1980; Davies, 1981; Anderson,1982b; Hofrnann and fracturesthat permitted deep penetration of water from a shal- White, 1982]. The natural tendency to ignore inconvenient con- low sea and for the uprise of isotherms associatedwith the straints is reinforced by the breadth of material relevant to this subject, while a preferencefor one model can easily blind one to metamorphism and partial melting. alternative interpretations of data.

Review Davies [1984] went on to discussassumptions and overinter- We have progressedin plate tectonicsfrom the concept of a pretations that have been made in both geochemical and geo- dynamic Earth activated by unspecified internal convective physical approaches to the problems represented in Figure 4 motions, as indicated in Figure 1, to a situation where we have and presented a structure for a heterogeneous mantle with several varieties of convection to choose from, as indicated by increasing viscosity as a function of depth, corresponding to Figure 4. These varieties are based on geochemical data, pet- Figure 4b. A_lldgre's[1987] global synthesisof the geochemical rology, and seismology.Figure 6 outlines some of the sources data transcends most previous models, and his conclusion is and magmatic products that are products of the mantle con- that the lower mantle has remained physically separate from vection. These are related to the tectonic environments of the upper mantle, except for intermittent entrainment of lower plate tectonics (Figures 1 and 4) and to the fluxes of volatile mantle in blobs rising from the mesosphere boundary layer. componentsin some environments (Figure 5). Whole mantle convection is precluded by the rare gas data Starting at the left-hand side of the figure at divergent plate FA_lldgreet al., 1987]. This conclusion contradicts that of boundaries and in oceanic plates, basalts (MORB and OIB) Hager and Clayton [1987], who find whole mantle convection with small amounts of dissolvedH20 and CO 2 are certainly (Figure 4b) to be more in accord with the geophysical evi- derived from the mantle. The submarine basalts react with dence. seawater with the formation of carbonates and hydrated min- The models for mantle structure and dynamics must satisfy erals which are carried into the mantle in subduction zones. not only the data of geochemistry and seismology JAiltigre, The oceanic crust when subducted may also carry with it 1987; Hager and Clayton, 1987], but also the constraints im- pelagic sediments,including limestones (in recent geological posed by thermal structures and phase boundaries, particu- times). During subduction it appears that most of the water in larly solidus curves. These two topics are considered in the subducted oceanic crust is released to participate in conver- following sections and used to test the models in Figure 4. At gent boundary processes[Wyllie, 1979]. A_lldgreet al. [1987] present, they provide constraints but not definitions. These concluded from rare gas systematics that subduction vol- topics become more definitive for the shallower processesas- canism is a barrier preventing volatile components from sociated with subduction and the lithosphere. deeper return into the mantle. Some of the carbonate may escapethese processesand be carried deeper into the mantle THERMAL STRUCTURES for long-term storage of carbon [Huang et al., 1980]. Carbon- ate in contact with peridotite will be converted to dolomite Magmas are generated because from place to place and then to magnesite,or it will be reduced to graphite or dia- from time to time, the Earth's internal temperature exceeds mond dependingon redox conditions(see Figures 20 and 21). that for the beginning of melting of some source rock at depth. Basalts (IAB) and andesitesare erupted at convergentplate The thermal history of the Earth is thus a fundamental factor. 380 WYLLIE.'MAGMA GENESIS,PLATE TECTONICS, DIFFERENTIATION

The present temperature distribution depends upon the Temperature øC Earth's temperature when formed, heat added during its early o lOOO 2000 3000 growth and history, the amount of heat generated as a func- I I tion of depth and time, and the rate of outward flow of heat. Processesfor the outward flow of heat include (1) conduction, Layer I (2) radiation, (3) convection, and (4) migration of fluids through the lithosphere. Geotherms have been calculated as- craton signingdifferent significanceto theseprocesses. 2OO It is now evident that the mantle convection associated with Layer 2 plate tectonics is the major processof heat transfer from the Earth's interior to the surface. Despite this understanding,a history of revisions of what scientistshave believed to be the thermal state of the Earth's interior should generatea certain 400 skepticism about current estimates. Lord Kelvin's authoritative estimate of the Earth's youthful age because of his cal- e- d culated cooling time from a molten state did not include the unknown effect of radioactivity. The concept of cold accretion of planets was reversedfollowing the study of lunar rocks and development of the idea of a magma ocean. It is now believed 6OO that the Earth became hot during accretion and that magma genesiswas probably an integral part of the growth history, rather than a process that began only when the Earth had heated sufficiently. Even during the past decade estimates of Layer 3 geotherms for the mantle have varied considerably, with the uncertaintiesincreasing as a function of depth. b c

Geotherrns A variety of geotherms calculated with assignment of differ- Fig. 7. Geotherms for different tectonic environments' see text for ent values to conduction, radiation, and convection of heat locations and sources. was compared by Wyllie [1971, Figure 3-12]. Figure 7 com- pares a more recent selection. Ringwood's[1975] calculated that calculated from heat loss [Boyd and Gurney, 1986]. The geotherm (a-a) is based on conduction and radiative heat geotherm for many kimberlites is inflected to higher temper- transfer. Jeanloz and Richter [1979] presented an improved atures at a depth of about 175 km (f-g), somewhatdeeper than geothermb-b for a conductinglithosphere (layer 1 about 120 the graphite to diamond transition. The inflection occurs at km thick) overlying a convecting mantle. Richter and Mc- shallower levels (f-h) in mobile belts. As reported by Boyd and Kenzie [1981] added the third layer 3, convecting indepen- Gurney [1986], the inflection has been interpreted in terms of dently of layer 2, based on the two-layer convection model of upriseof mantle diapirs associatedwith the generationof kim- Figure 4a. They determined from laboratory and numerical berlites or in terms of local magma chambers. Nickel and experiments that there is a marked increase in temperature Green [1985] refined empirical garnet-orthopyroxene ge- between the two discrete convecting layers as shown by geo- obarometry and presenteda distinctivepattern for South Afri- therm c-c. There is a conductive geotherm in layer 1, the can xenoliths where the high-temperature xenoliths give near- lithosphere lid; there are two steep adiabatic gradients in the isobaric estimates,corresponding to a depth of 150-160 km, at two convecting layers; there is a double thermal boundary 900 ø-1400 øC. layer between the convecting layers 2 and 3, resulting in an Recent measurementsof the melting curve for iron at high average temperature increase of about 500øC near 700-km pressuresusing shock wave methods and diamond anvil appa- depth if the lower mantle is isolated from the upper layer. If ratus have improved the constraint on the geotherm at the the deep layer 3 is assumedto have been depleted of its radio- core-mantle boundary, 2900 km down, and at the Earth's active elements through magmatic processesearly in Earth center [Williams et al., 1987]. Temperature estimates at the history, as proposed by Anderson [1982] in Figure 4c, the core-mantle boundary are in the range 3000ø-3500øC, with a geotherm would not exhibit a step, but would remain near distinct thermal boundary layer. b-b. We need to know the Earth's thermal structure in order to understand magma genesis,but we also need to know the Isotherms Earth's magmatic history in order to determine the present The perturbationscaused by uprisingmantle material, or by geotherm. In regions of active upward mantle movement, as sinkinglithosphere in subductionzones, cannot be represented associated with mid-oceanic ridges and mantle plumes, the completelyby geotherms.We need to see the thermal struc- lithosphereis thinned, and the adiabatic part of the geotherm ture as depictedby isotherms.The thicknessof the lithosphere corresponding to layer 2 in Figure 7 is extended upward, as is an important variable becauseof the change in rheology shown for geotherms d-d and e-e, respectively.The estimated occurring through the asthenosphere-lithosphereboundary positions of the changesin slope of the geotherms are based layer, which is assumedto be near the 1200øCisotherm in the on independent evidence or assumptions about lithosphere following diagrams. This layer is depicted as a line near thicknessor melting temperatures (seebelow). 200-km depth beneatha continentalcraton in Figures 8 and From the geothermometry and geobarometry of mantle 10 and near 90-km depth beneathan oceanin Figure 9. nodules we have a geotherm for cratons (f-f) consistentwith Figure 8 illustratesthe regular distributionof isothermsas- WYLLIE.' MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION 381

_600o__---- ..-_600 ø-- ,800 I00 - _.... 800ø'"-"-" _ IOO I000o• , •ooO

200 200 ..1300ø ---i/<_•__•,.• .... . -x•øøø' ...... - I I 300 300 - I I i 1,4ooo,, Fig. 8. The craton with isothermssuperimposed according to the Fig. 10. The lithosphere sectioncorresponding to Figure 8, with uninflectedgeotherm (f-f) in Figure 7. isotherms perturbed by a plume rising from the deep mantle and diverging within the asthenosphere beneath the lithosphere- asthenosphereboundary layer. sociated with a thick craton and a somewhat thinner lithosphere. If a thermal plume rises below the lithosphere, the The mantle wedge may be relatively cool, chilled by the sub- asthenosphere-lithosphereboundary rises to shallower levels, ducted slab, or it may be heated by induced convection causingthinning of the lithosphere, as indicated by the 1200øC [Toks6z and Hsui, 1978]. Schematic isotherms for these two point on the inflected geotherm (f-g) of Figure 7. In a different mantle conditions are illustrated in Figure 12. Combinations environment the effect of a plume on the isothermsis depicted of thesefour conditionsprovide four limiting cases.An impor- in Figure 9, which representsan oceanic plate moving from tant fifth condition must be included. Marsh, [1979] and right to left. The asymmetry of the plume and of lithosphere Maaloe and Petersen [1981] have appealed to extreme in- thinning is causedby movement of the plate. The geotherm in duced convection cells in the mantle wedge, raising the tem- the center of the plume correspondsto e-e in Figure 7. The perature of the subducted oceanic crust at 100-km depth to lithosphere is heated above the plume, and associatedwith the 1250øC or 1450øC. rise in isotherms is thinning of the lithosphere. A mantle plume rising beneath a static craton changes the isotherms as EXPERIMENTAL CALIBRATIONS: SOURCE ROCKS AND MAGMAS shown in Figures 8 and 10. There is evidence that beneath Figure 13 outlines some of the ways that phase diagrams for major continental rifts the lithosphere is thinned. The contin- minerals and rocks may help to decipher the processesoc- ued heat flux permits uprise of the asthenosphere[Gliko et al., curring between the partial melting of the inferred source 1985]. This is representedby the isothermsin Figure 11. rocks and the observed magmatic products. Wyllie et al. There have been so many different calculated thermal struc- [1981] discussedthe use of experimental petrology in forward tures published for subduction zones that we may conclude and inverse approachesto the problem of source-magmarela- either that the thermal structures are not yet known in detail tions and reviewed in detail the inverse approach for basalts. or that there is considerable variation. The subducted oceanic Interpretation of geochemicalanalysis of magmatic products crust may be relatively warm or significantly cooled by en- in terms of the chemistry of source rocks is an inverse ap- dothermic dehydration reactions [Anderson et al., 1978, 1980]. proach. Forward and inverse approaches supplement each other as illustrated in Figure 13. This is followed by a seriesof 200 100 seo-• 0 crust-x 100 200km phasediagrams providing limits and calibrations for processes :...... ,...... •.•... ß ...... L...... ,I ---- Moho-• involving the major materials of source reservoirs in mantle 4000 - _ I_ithos_phere 600o_ and crust.

lOO o L 3•..• ..... 6øøø--'"-t •., 200 'øø1=/...... ?=•I'i_J...... 8ooø---_ • -[ ;IIr iI,k _- 'øøøø'.-" / ._x:E .... ,,F'½.-II II ......

30O • 200

z60 lOO 6 ' lOO Dist(]nce,km Fig. 9. Schematicisotherms for mantle plume based on plumes of Courtney and White [1986], the requirement for melting at M in 1400 o Figure 21, and with asymmetry causedby motion of lithosphereplate I I above the plume. The change in rheology associated with the asthenosphere-lithosphereboundary layer is arbitrarily represented Fig. 11. Schematic isotherms in craton as in Figure 10, with ad- by the heavy line for the 1200øCisotherm. ditional thinning of the lithosphere beneath a rift zone. 382 WYLLIE: MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION

WARMMANTLE

Fig. 12. Distribution of isotherms in mantle wedge (between the two stippled crusts) for (a) a relatively cool mantle wedgeand (b) a mantle wedgewarmed by inducedconvection.

Phase Diagrams and Their Interpretation hydrous minerals.The soliduscurve at low pressuresis the The main variables for representing the phase relationships sameas that for Figure 14f At higher pressures,however, the are pressure, temperature, and parameters for composition. soliduscorresponds to the reaction where the hydrous min- The simplest, most convenient geometrical framework for eral breaks down, providing H20 for solution in experimental data, considering the variables that can H20-undersaturatedliquid. Figure 14e representsa rock be measured and controlled in the laboratory, is systemwith hydrous minerals and subsolidusvapor but with P-T-X(SiO2)-X(H20) [Wyllie, 1977, 1979]. For an individual insufficientH20 to saturatethe liquid. The solidusis the same rock, X(SiO2) is constant,and the frameworkis reducedto the as that in Figure 14f, but the melting interval includesa large three-dimensional model P-T-X(H20 ) illustrated in Figure area of crystalsplus H20-undersaturatedliquid. Similar dia- 14a. This is too complicated for convenient viewing, and re- grams can be constructedfor silicate-CO2systems including sults are commonly presentedin sectionsthrough the model. carbonate minerals. Figure 14b is an isobaric T-X(H20 ) section.Note the three In order to trace geologicalprocesses for a given rock, we different solidus temperatures, one for excessvapor, one for a should follow polybaric-polythermalpaths through the phase rock with all H20 stored in a hydrous mineral, and one for volumesin Figure 14a, which is graphically difficult. However, the completely anhydrousrock without any hydrous minerals. the types of processescan be illustrated in isobaric or iso- Note also the large area for H20-undersaturated liquid with thermal diagrams. crystals. The phase equilibrium data for rocks and magmasprovide Isoplethal sectionsthrough the model in Figure 14a provide a set of constraints that can be used in the following ways as four types of P-T diagrams correspondingto different ranges outlined in Figure 13:(1) to define the mineralogy of possible of H20 contents.Each of theseis correlatedin Figure 14 to source rocks as a function of depth; (2) to define the compo- the isobaric section,Figure 14b. Figures 14c and 14frepresent sitionsof liquids producedfrom possiblesource materials; (3) the familiar extremes for the melting interval of a rock either to definepolybaric polythermal paths of crystallizationor dif- completelyanhydrous or with excessH20. Figure 14d repre- ferentiation;(4) to determine,in conjunctionwith geochemical sentsa rock without pore fluid, but with some H20 stored in data, which composition in an igneous suite representsthe

INVERSE FORWARD Hydrous H20-under F. C. DRY D.m•neral E. saturated ExcessH•O OBSERVEDROCK SUITESIGNEOUS F D,fferentmtion I • P P PIx,• •q L+I •_ xls+L Parentalmoõmo ß ,I, ' I Temperature•/xls+L+V.• I t • T• Moslliqu,dprimitive A. t I An/primary , I I•uid? , ...... MeltIna;,•s:b•L I PhaserelationsP.T. volatiles I ? Dry Liquidusminerals Phase relationsof Multiplesaturahon Matchwith source • poss•blesource rocks rockm•neralogy P.T. valorlies I 0 wt •1INFERRED ROCKSSOURCE • Fig. 13. The inverseand forward approachesfor testingthe relations between observedigneous rocks and inferred source rocks as Fig. 14. Phaserelationships for a singlerock representedwithin a depth, using phaseequilibrium results from experimentalpetrology P-T-X(H20) modeland illustrated by isobaricand isoplethal sections (simplifiedfrom Wyllie et al. [1981, Figure 3.1.4]). through the model. WYLLIE.' MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION 383

I GARNET•ERI'DOTI'T•- ',•lii'' ' ' schemeis wrong becauseit violatesthe constraintsof experi- I 1611 mental petrology. 30I-My'sen ½Kush,ro 1977 lii'i ,, - LHERZOLITE/ I00 / (•o•o.• • z71,'!;!"',• Lherzolite 01+Opx Alkoh+ Cpx +•.'• • 20,- /./fi Ohvme•.../ • There have been many experimentalstudies on various as- • PYROLITE/ / / •.• •oso.,,•F",•' " pectsof the phaserelationships of mantle peridotite,lherzolite, JJequeseGreen 1980// /// o..o,, • l(10kbend 15kb)// /// and harzburgite,with most pressureslimited to about 40 kbar. IUF I I /11 The persistenceand ingenuity of Japanesescientists with the developmentof massivepresses and relatively large volume 0-/ .... •/,,,,,,,•/,/,,'•',', • , , , .... •'•)"'• Z_• •' • 0 ooo o'o5 500 I000 1,500 2000 high-pressureexperiments have been rewardedrecently with A Temper•lureøC B. Takahashi's[1986] most extensivephase diagram for lherzolite as illustratedin Figure 3. The diagram showsphase re- Fig. 15. Phase relations for peridotites.(a) Limited data for percentageof liquid within melting interval above solidus,after Mysen lationshipsdetermined for a lherzolite up to pressuresof 200 and Kushiro [1977] and Jaquesand Green[1980]. (b) Compositionsof kbar, with distinctivecusps associated with subsolidusphase liquids generated within melting interval, after Jaques and Green changes,where the aluminousminerals change from plagio- [1980]. clase at low pressure,to spinel, and to garnet at pressures corresponding to depths of about 75 km. Some indication of primary or most primitive liquid; and (5) to determine the experimentalproblems is given by the variety of resultspub- conditions under which specific liquid compositionscould lished for soliduscurves. Figure 33a showsthe soliduscurves coexistwith residual minerals in a source rock. These ap- previously published for different peridotites [Wyllie, 1984, proaches, in combination with the thermal structure within Figure 2]. A linear extrapolationalong c-d providesvery high the Earth, provide a means to test geophysicaland pet- solidustemperatures at depth. The calculatedSimon-type fit rological models in different tectonic environments. along c-e [Griggs, 1972] is a better approximation to the The mantle, subducted oceanic crust, and continental crust recent high-pressuremeasurement (Figure 3). The calculated containseveral different source rock typesthat may contribute solidusis paired with an estimatedliquidus curve in Figure to magmas. In a forward approach, we should know for each 33b [Wyllie, 1981]. rock: (1) the subsolidusphase relations at least to the depths The resultsof two attempts to measurethe percentageof of magma generation, (2) the solidus curve, which defines the liquid producedabove the solidusof peridotiteare compared conditionsfor beginningof melting,(3) the percentageof melt- in Figure 15a. Mysen and Kushiro [1977] reported near- ing as a function of temperatureabove the solidus,and (4) the isobaricinvariant melting with productionof muchliquid just changingcomposition of the liquid within the melting interval above the solidusat 20 kbar and 35 kbar in a garnet perido- as a function of pressure. For some environments, we should rite, with distinct change in slope of the melting curve wher- know (5) the effect of vclatile componentssuch as H20 and ever one or more minerals disappeared.According to Jaques CO2 on the previous items and (6) the effect of oxygen fu- and Green [1980], their results for pyrolite demonstrated that gacity in changing vapor phase composition and solidus. In the steplike melting characteristicof simple syntheticsystems order to translate these chemical parameters into magmatic was smoothed out by solid solution effectsinvolving Fe-Mg, processes,we need to know in addition (7) the physical A1-Si, Na-Ca, etc., Despite the different solidus curves for the characteristics of the partially molten rocks and their vari- two rock compositions,the melting resultsappear to converge ation with pressure,temperature, percentage of melting, and for about 50% melting. The reason for the difference between liquid-mineral compositions.This information is required to thesetwo setsof resultshas not been establisheddespite their establish the range of conditions through which a partly significancewith respectto the partial melting of the mantle. molten source may rise as a unit, the conditions through The fact that the solidus curves for the two materials com- which a liquid may percolate upward through the crystalline pared in Figure 15a are different may be due to differencesin mush, maintaining equilibrium with its host, and the condition composition, but differencesin experimental techniques and where the liquid may escape from its host and rise as an sample containers could contribute. When Takahashi and Ku- independentbody of magma in a crack or diapir. Dense melts shiro [1983] redetermined the solidus for a lherzolite mea- formed at depth may sink [Rigden et al., 1984]. suredearly by Kushiroet al. [1968], usingbetter experimental A primary magma is a liquid whose composition has not techniques,the solidus temperature was lowered by about changedsince it was generatedby partial melting in the source 70ø-100øC and the solidus curve had the cusps as given in region. Figure 13 outlines the proceduresin the inverse ap- Figure 3, instead of the original linear solidus like those proach of deducing source rock chemistry and mineralogy shown in Figure 15. The presenceof detectablecusps on the from the phase relationshipsof primary magmas. Wyllie et aL peridotite solidus,corresponding to the subsolidusphase tran- [1981] gave detailed accounts of the procedures, of a search sitions,was first recordedby Presnall et al. [1979] in a syn- for possibleprimary magmas, and of a case study for oceanic thetic system. More detailed study would probably reveal tholeiites. They concluded that primary magmas were rare, similar cuspsfor the soliduscurves in Figure 15. most magmas having experiencedlow-pressure crystal-liquid Figure 15b shows the subsolidusphase relations and esti- fractionation. One critical test is to match multiple-mineral mated melting interval for garnet lherzolite, with the P-T saturation on the liquidus of a primitive magma with the re- rangesfor liquids of various compositionswithin the melting sidual minerals of a multimineral source rock at the same interval. Liquid compositions correspond to those calculated pressure-temperatureconditions. Although an internally con- by Jaquesand Green [1980], who concludedthat "most, per- sistent solution may be constructed for source-magma rela- haps all, previously publishedpartial melt compositionsob- tions, this is not necessarilyunique. The approach may be tained by direct analysis of quenched glasses"are in error, more useful in demonstrating that a proposed petrogenetic becauseof iron lossfrom sampleas documentedby Nehru and 384 WYLLIE' MAGMAGENESIS, PLATE TECTONICS, DIFFERENTIATION

4O Wyllie [1975] and growth of metastable crystals during O•,+•px+Cx . - quench [Green, 1976]. Takahashi and Kushiro [1983] reached conclusionsconsistent with Figure 15b based on microprobe 30 ...... • measurementsof glass using a technique designedto over- come the Fe loss and quench problems. emcv . 0S+0px+Gpx. '"---••'•-x • ? /%"• _

Lherzolite-H20 and Lherzolite-CO2 --• +Go+V. ''• //// The phaserelationships for the systemlherzolite-CO2-H20, involving the minerals amphibole, phlogopite, dolomite, and - • 0J+0px+Cpx// Peridotire magnesite,provide the framework for upper mantle petrology •.+• x+• +P•+V !/ Iconreining and for evaluationof the phaserelations with reducedoxygen 0 u,up •px // ,•oC0,2 0 fugacity. There remains uncertainty about the position of the 600 800 I000 1200 1400 1600 solidus, which also varies as a function of composition and TemperotureøC oxygen fugacity. Fig. 17. An interpretation of phase relationships for The solidus curve for lherzolite is compared in Figure 16 peridotire-CO2, based on various sources [Wyllie, 1978b]. The sub- with some of the important phase boundariesintroduced by solidusexchange reaction with dolomite replacedby magnesiteis now the addition of H20. The solidus curve is lowered consider- known to be at lower pressures(see Wyllie et al. [1983] for a review). ably by the presence of an aqueous vapor phase, and the Abbreviations:O1, olivine; Opx, orthopyroxene;Cpx, clinopyroxene, Ga, garnet; Sp, spinel; P1, plagioclase; Cd, dolomite solid solution; hydrous minerals amphibole and phlogopite coexist with the Cm, magnesite solid solution; V, vapor. Dotted lines are estimated liquid; see Wyllie [1978a] and Wyllie et al. [1981] for sources geotherms.See Figure 7. and for different experimentallyreported curvesfor amphibole and the solidus.Note that if thereis only enoughH20 present to form amphibole or phlogopite, with none left over for a cate compositionsfrom the CO2-richextreme, as indicatedby separate vapor phase, the solidus curves for phlogopite- the estimatedliquid compositionsmarked in the diagram.The peridotite or for amphibole-peridotite would be at somewhat principleis clear, but there remainssome uncertaintyabout higher temperaturesthan the phaseboundaries for phlogopite the pressureof Q, as shownby the variety of estimatesgiven and amphibole, respectively,in Figure 16 (compare Figure in Figure 19c. Wyllie and Rutter [1986] have new experi- 14d). Hydrous minerals talc and serpentine are produced in mental data placing Q at 22 kbar. peridotite at moderate temperatures,and at high pressures densehydrous magnesiansilicates (DHMS) are formed. Lherzolite-H 20-C 0 2 Figure 17 showsan interpretation of the phaserelationships Some aspects of the phase relationships in this system for peridotitewith a smallamount of CO 2. This remainsvalid remain controversial [Wyllie, 1987b, c; Egglet, 1987; Taylor for up to about 5% CO 2, which would be sufficientto com- and Green, 1987], but we can adopt some diagrams and con- plete the carbonation reaction represented by the line TQ sider applications without becoming embroiled in the contro- where dolomite (Cd) is being generated at the expense of clin- versies.In the presenceof H20-CO 2 vapors, the solidus be- opyroxene; as it is, all CO 2 is used up in reaction TQ, and no comesa divariant surfaceconnecting the H20 and CO 2 solidi free vapor remains at higher pressures. in Figures 16 and 17. The reactions for amphibole and dolo- The solidusfrom P to Q is depressedonly slightly below the mite also become divariant, and all three phase surfacesover- peridotite solidusPS becausethe solubility of CO 2 is so low, lap in a broad region. The details of the solidus in this region in contrast with that for H20 (compare solidi in Figures 16 vary according to the ratio and amounts of H20 and CO 2 and 17). The formation of dolomite in the peridotite (reaction present. TQ) changes this situation, and along QR the partial melting If there is excess vapor, more than enough to make the of dolomite-lherzolite generates a liquid rich in carbonate, maximum amounts of dolomite and amphibole within their with about 40% CO 2 in solution; this is not a siliciateliquid, stability ranges, then the solidus surface is divariant, as shown but a carbonatitic liquid. With increasing temperature above in Figure 18. Figure 18 reproduces the liquid compositions the solidus, the liquid composition migrates back toward sili- and solidus for volatile-free peridotite from Figure 15b and shows in addition the solidus curves for peridotite-H20 , peridotite-CO 2, and the divariant solidus surface connecting these curves. Details of the phase relations and the geometry PERIDOTITE . •. 75 of isoplethal sectionsare given by Wyllie [1978c, 1979, 1980]. 2OO There have been several experimental studies and many con- 5O troversial reviews concerned with the composition of near- solidus liquids in this system. Figure 18 is an attempted com- 100 promise that illustrates the major trends, with chemical variation shown by the lines representing changes in normative composition of the near-solidusliquids [Wyllie et al., 1981, pp. e ,/I 553-556]. The amount and composition of liquid developedat 500 I000 1500 the solidus depends on the amount of (CO 2 + H20), on TemperotureøC CO2/H20, and on the presenceor absenceof buffering minerals amphibole,dolomite, and phlogopite.The effect of H20 Fig. 16. Selectedreactions showing the effect of H20 on melting and hydration of peridotite. DHMS. dense hydrous magnesiansili- on liquid compositions is to cause enrichment in SiO2, cates (see Wyllie [1978a] for sourcesand referenceto other experi- whereasthe effect of CO 2 is to causedepletion in SiO2. With mentally determined curvesfor the solidus and for amphibole). progressive melting at mantle pressures, the liquids follow WYLLIE: MAGMA GENESIS, PLATE TECTONICS, DIFFERENTIATION 385

500 1500 0 I I P\ ,<) CRUST f02 30 \ "'''-'OQ -...... ~/.: WM c '~\ -IW ~ ~ , NNO ~ ~ " ~ 20 = Ci "WM ~ Cl. 50~ '\ \ 10 E WM -'£ ·-IW ..c LLithosphere 0' , , , , , , , , , I •. , I ' , , 0. 200 ... . WM Q) o 500 1000 1500 o FMQ Temperature °C Fig. 18. Projection of the solidus surface for peridotite-COz-HzO, with tentative boundaries showing the changes in composition of the near-solidus, vapor-saturated liquids [Wyllie, 1979]. Solidus with CO"~ P-Q-COz; with HzO, P-HzO; and for dolomite-peridotite with 300 buffered vapor (Vb), Q-N. Peridotite- C- H-° paths through wide temperature intervals toward the basaltic with X = and picritic liquids near the solidus of volatile-free peridotite. C02 /(C02 + H20) =0.8 For a particular rock with defined volatile component con 4001 II II 1') tents, we need isopleths through the complete phase diagram Fig. 20. Solidus curves for peridotite-HzO (dashed lines) and for [Wyllie, 1979J. If the lherzolite contains less than 5% COz peridotite-COz-H 2 0 with defined ratio of COz/HzO (PMQR) extrap and less than 0.4% HzO, then the solidus curve involves dolo olated to high pressures and compared with the cratonic geotherm, mite, amphibole, or both in buffered melting reactions. For a with and without inflection (Figure 7). The positions of selected phase boundaries relevant to mantle processes and the origin of kimberlites wide range of conditions, the solidus curve remains univariant and nephelinites are given. Possible oxygen fugacities at various levels in terms of pressure and temperature. The only two experi are indicated by the buffers listed [Haggerty, 1988]. The peridotite mental studies published on amphibole-dolomite-peridotite in solidus is from Figure 3. For DHMS dense hydrated magnesian sili dicate that the amphibole stability volume overlaps the sol cates (DHMS), see Ringwood [1975, p. 295]. Abbreviations: Hb, am phibole; Ph, phlogopite; Do, dolomite; idus with dolomite. There is a· difference of several kilobars Mc, magnesite; V, vapor. between the results of Brey et al. [1983J and those from the detailed investigation of Olafsson and Eggler [1983J, as shown in Figure 19. peridotite-C-H-O for a system with volatile components in the An estimated (but constrained) and extrapolated solidus ratio COz/(COz + HzO) = 0.8, if oxidized. With variation in curve for amphibole-dolomite-peridotite is compared with this ratio the solidus temperature changes at pressures shal curves for peridotite dry and with HzO in Figure 20. This is lower than point Q at about 75 km but remains unchanged at modified from a previous version [Wyllie, 1980J incorporating higher pressures. Experimental data extend no deeper than new data [Olafsson and Eggler, 1983; Wyllie, 1987a]. Despite about 180 km. differences in experimental results in Figure 19, the existence Important levels in the mantle are those where volatile com of the solidus ledge near Q, where the solidus changes slope ponents would be released by activation of a dissociation reac and becomes subhorizontal, appears to be established (level 4 tion. One such reaction is represented in Figure 20 by the line in Figure 20). In the following discussion the extrapolated DHMS giving the boundary for the reaction of olivine to form solidus for peridotite-HzO, which is believed to remain close dense hydrous magnesian silicates in the presence of HzO to that for peridotite-COz-HzO at pressures greater than (Figure 16). The estimated position of the reaction for the about 35 kbar, is adopted as the solidus at depth for conversion of forsterite to brucite and enstatite in

30 I- -pe-rid .. I -l 30 -0 ~ Hb-Do-H1/r H2 lperidotite J ~ ~ "'u v", .. II 20 Do-Hb- \withV ~ 20 ~ Q) ct I ,. HtI-oerl!j I I lOr /Hb-:~rid / l 10 / / I / /El8 0& 1200 -800 1000 1200 800 1000 1200 1400 1600 Temperature °C

Fig. 19. (a, b). Experimentally determined solidus curves for hornblende-dolomite-peridotite. (e) Locations for the

invariant point 16 on the solidus for peridotite-C02 deduced or estimated from experimental data in other systems. Points G and OE are deduced from Figures 19a and 19b, respectively [Wyllie, 1987b, c]. 386 WYLLIE'. MAGMA GENESIS,PLATE TECTONICS, DIFFERENTIATION

MgO-SiO2-H20 (Figure 21) is closeto the DHMS curve at RO' /' '/ 300 km [Ellis and Wyllie, 1979]. GABB_ •.•/Solidusfor/ - 75 200 •, ._o'/qua rtz/coes fie/ EE cj eclogite 50 Lherzolite-C-H-O H20dry - The redox state of the deeper mantle appears to be such '-' 100 that the componentsC-H-O exist as carbon and H 2 or H20 -' ecløgdel - Solidus- CL with CH 4, rather than as CO2and H20[Deines, 1980;Ryab- •- /.•-••-..•'J Hb-gobbroJ chikov et al., 1981; Green et al., 1987], although this remains controversial.According to D. H. Eggler (personalcommuni- 500 I000 1500 cation, 1987), "My sourcesnow indicate that it is by no means TemperatureøC certain that asthenosphericoxygen fugacity is low enough Fig. 22. Selectedreactions showing the effect of H20 on melting (well below magnetite-wustite)that volatile componentsin and hydration of basalt (and eclogite).After Wyllie [1978a]. C-H-O would existas H20 and CH4 rather than as H20 and CO2." Figure 20 summarizesHaggerty's [1986, 1988] picture of the possiblevariations of oxygen fugacity in terms of stan- Gabbro-Eclogite dard buffers down to 250 km. Basalt, the most abundant magma derived from the mantle, With low oxygen fugacitiesat high pressures,C-H-O exists crystallizesas basalt or gabbro at crustal pressuresand as as H20 with CH• or graphite/diamond.Under these con- eclogiteat mantle pressures.Part of this phase transition is ditions,the peridotite-C-H-O solidusmay be raisedin temper- shown in Figure 22. There has been extensivedebate about ature compared with the oxidized solidusas indicated by the whether or not the Moho could be explained by the gabbro- arrows in Figure 21. Accordingto someinvestigators, the sol- eclogitephase transition [Wyllie, 1971,chapter 5], but present idus may remain close to that for peridotite-H20, with car- evidencefavors a chemicalchange. There have been extensive bonate ions being generatedin the melt when CH• or graph- studies of the liquidus minerals of various basaltic compo- ite/diamond dissolvesrEggler and Baker, 1982; Ryabchikovet sitions in efforts to relate them through the inverse approach al., 1981; Woermann and Rosenhauer, 1985]. According to D. (Figure 13) to peridotitesource rocks at variousdepths. The H. Green (personalcommunication, 1987), however,recent ex- resultsand approacheswere reviewedin detail by Wyllie et al. perimentswith pyrolite-CH•-H20 at various H20 activities [1981], with particular attention to MORB's. It is expected, with reduced oxygen fugacitiesdo move the solidus to higher for example,that a basaltliquid deriveddirectly from lherzol- temperatures. ite would be in equilibrium with the lherzolite minerals oli- The dashed lines in Figure 20 show geothermsfrom mantle vine, orthopyroxene,clinopyroxene, and a fourth aluminous xenoliths in kimberlites (Figure 7, f-f, f-g), corresponding to mineral, or at least olivine plus orthopyroxeneif partial melt- isotherms in Figures 8 and 10. In Figure 21 the solidus is ing had beenextensive. The resultsfor the basaltcomposition compared with two geothermsassociated with a suboceanic givenin Figure 23 show olivine on the dry liquidusonly to mantle plume (Figure 9). The temperaturesare higher and the about 20-km depth and without orthopyroxeneon the liqui- lithosphereis thinner than in Figure 20. Note that the reac- dus. One conclusionis that this particular basalt composition tions of olivine with water to form DHMS or brucite probably wasnot a primarymagma and had experienceddifferentiation intersect the geotherm somewherebetween 300 and 400 km, en route to the surface. According to Figure 15b, basalt suggestingthat perhapsno H20 but only reducedgases CH• magmacan be derivedfrom peridotiteonly at depthsof less and H• can exist at deeperlevels. These species are not stored than 50-60 km; melts formed deeper than this are richer in in high-pressureminerals as far as we know. The depths2 and magnesium,corresponding to picritesor komatiites.Limits of 3 are variable, dependingon conditions. these kinds are fundamental in checkingthe validity in detail of the processesrepresented broadly in diagramssuch as those in Figure 4.

TemperatureøC Gabbro/Basalt-H2 0 2) 500, lO,OO 1,.%00 2000 The main source rock in subducted oceanic crust is basalt -'--,Go%,,.['HiRZOLiTE or gabbro, metasomatized to uncertainextent by penetration loo BASALT- H20 with reduced _LHERZOLITEvolotiles C-H-O •}-- o ""•11,•• I '/ I < •' I /I C02/H20:0.8 I coes,*e/ 701 ø 2 ?l En o 200 30 3 O(J p 'øø

300 • t l / I• • • JO• Hb+Cp'x' ,'/ I , , 40qkm -•lkm I 400 o 600 I000 1400 Fig. 21. Soliduscurves for lherzolitewith and without volatile TemperatureOc componentsand deep dehydration reactions(see Figure 20), comparedwith geothermslocated at centerof mantleplume axis and 400 Fig. 23. Phasereactions for basalt-H20,modified after Sternet al. km away (seeFigure 9). [1975] by Wyllie [1979]. WYLLIE' MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION 387

L • I • I1 I I ' ¬Si02 1400 50 k b,; 7.5% H2 0 / tl \ ////////•.•Noturol r oc ks ili]cii...•.. Liquid / 30kb 1200 moleYo /Im • • '•';:?;:•;;':LiquidpothsThisstudy /½• •, • = Sternand iooo DRY iii'"'"'"""'•••"'"'"'"' '""'" Cryst.Is .fb••• • Wyllie1978 800

_iiA Coesite- kydnite- - • eclogite+ V 600

o io 20 30 40 Wt % H•,O CaO• •MgO Fig. 24. (a) Isobaric sectionfor gabbro/eclogite-H20at 30 kbar, showingpath through phasefields for rock or melt with 5% H20. (b) The calc-alkalinerock sequenceB-A-D-R comparedwith equilibrium paths for liquids generated experimentallyin quartz eclogiteat 30 kbar with 7.5% H20 , using natural basalt (bB [Stern and Wyllie, 1978]) and syntheticbasalt in systemCaO-MgO-A1203-SiO2-H20 (mM). "This study"is from Sekineet al. [1981].

of seawater and metamorphosed during subduction through Tonalite/Andesite-H2 0 greenschist, amphibolite, and eclogite facies. Amphibolite is Andesite is the characteristic lava of convergent plate also an important source rock in continental crust. Addition boundaries,and tonalite (with similar compositionin terms of of water to gabbro introduces hydrous minerals and lowers major elements)is a major component of many batholiths in the melting temperatures. Figure 22 shows the stability range the sameenvironments. Tonalitic gneissis probably an impor- of the hydrated subsolidus assemblage in the greenschist tant component of the source rocks in the lower continental facies, with dehydration of chlorite at about 600øC marking crust. Figure 25 is directly analogousto Figure 23, showing the transition to amphibolite facies. Amphibole stability ex- the liquidussurface for the magmacontoured in termsof H20 tends to temperatures above the solidus curve in the presence content and the effect of H•O on the liquidus mineralogy, as of H20, which is lower by several hundred degreesthan the well as the effect of excessH20 on the soliduscurve and the dry solidus for the rock. The greenschistfacies assemblage subsolidusmineralogy. The liquidus surfaceprovides the key extendsto high pressures,whereas the amphibole is limited to evidencein the inverse approach concerningpossible source pressures of less than 30 kbar because it breaks down to the rocks for an andesite of this composition, if it is consideredto dense eclogite phase assemblage that is responsible for the be a primary magma. An isobaric section at 15 kbar is shown changein slope of the soliduswith H20 at about 15 kbar. If in Figure 26 rHuang and Wy!!ie, 1986]. This is analogousto the rock contains only enough H20 to form amphibole with Figure 24a for basalt, with the addition of details of the sub- none left over for free vapor, then the solidus is associated liquidus mineralogy. Note that the vapor-absent tonalite with the breakdown of vapor-absent hornblende-gabbro or beginsto melt at a higher temperaturethan the vapor-present hornblende-eclogite, releasing H •O for solution in rock, following the example of Figures 14b and 14d. Huang H20-undersaturated liquid (Figure 14d). The phase bound- and Wy!lie [1986] reported the compositionsof glassesand aries in Figure 22 are those used in the forward approach of mineralsin the sequenceof runs with 5% H•O. calibrating what happens to basalt under various conditions. Rutter and Wy!lie [1988a] studied the progressivemelting For the reverse approach of considering possible source for of vapor-absent tonalite at 10 kbar, representingpartial melt- this basaltic composition with dissolvedH•O, we need infor- ing of tonalite gneissin the lower crust in the absenceof pore mation about the liquidus. fluids. They measuredthe percentageof H20-undersaturated Figure 23 outlines phase relationshipsfor basalt-H•O ex- melt produced as a function of temperature as well as the tending from the solidus to the liquidus surface which is con- compositionsof glassesand minerals.The resultsin Figure 27 toured in terms of H•O content. The solidus is the same as show that melting begins at about 825øC with the breakdown long as there is a free vapor phase,but the liquidus curve for of biotite to produceH20-undersaturated liquid; 20% liquid any specificH20 content is given by the appropriate contour. This is illustrated in Figure 24a, an outline of the isobaric sectionfor the system(Figure 13) at 30 kbar, above the stabili- ANDESITE-H20 ty range of amphibole. Notice the liquidus for o 20 io 5 2% H20-undersaturated basalt. A gabbro with a few percent oli- Coes•te / / / a:•30 •- ecloglteI O/ / I00 vine is converted to coesite-eclogiteat 30 kbar (Figure 23). I - •...-d •/,/ '•//./ /cx I The results of two attempts to determine the compositions of -- / q,' liquids produced during progressivefusion of quartz-eclogite • 20 Crystals in the presenceof a few percent of H•O are compared in Figure 24b. Stern and Wyllie [1978] plotted liquid compo- o_ I0 +pc •-• sitions ranging through intermediate SiO• contents, with o, +L+V 7 01 I • I TM I Ca/(Mg + Fe) higher than the average composition trend for 600 I000 1400 calc-alkaline dacite-andesite-basalt. Results of Sekine et al. TemperatureøC [1981] for synthetic basalt composition in the system Fig. 25. Phase relationshipsfor andesite-H20 modified after Stern CaO-MgO-A1203-SiO2-H20 reinforcedthe conclusion. et al. [1975]. 388 WYLLIE' MAGMAGENESIS, PLATE TECTONICS, DIFFERENTIATION

1400

-- ' TONALfiE ' 15 Kb 1200 - ß ', --

SOLIDUS 0 L ' L + V

lOOO_ '-.

\ • '-. ,' Hb+Cpx+ Ky+ L+ V Hb----•• 8i.•. / Hb+ Cpx+ Ga+ Ky+ L+ V II 800•P..• '%, ' •' tJim ' PI+++ei(••.06,7 ,' "'• ' / _,,'/ Hb+ Cpx +(Ga) + ei +Ky +L + Vrl (õa)+L /•..... ß L_J 600 .. - SOLIDUS .-,...-Hb + Cpx+ PI+ Oz+ (Ga)+ Bi Hb+ Cpx+ PI+ Oz+ (Ga)+ Bi+ V

0fROCK 10 20 30 40 50 WATER WT. %

Fig. 26. Isobaricsection for tonalite-H20 at 15 kbar [Huangand Wyllie, 1986]. is producedduring the biotite reaction between825øC and evidencefor depth of crystallization.The two empiricalcurves 900øC.By 1000øCthe horblendehad also melted,producing publishedare reproducedin Figure 28 [Hammarstromand 35% liquid containinga calculated2.3 weightpercent H20. Zen, 1986; Hollister et al., 1987]. These are applicable only The compositionof hornblendein granitic intrusionsis a under the specifiedconditions, because of the large number of function of the magma compositionand the conditionsof variables that can affect the hornblende composition. The ex- crystallization.An empiricalgeobarometer has beenproposed perimental results for hornblende composition in partly recentlyon the basisof total A1 contentof calcichornblendes melted tonalites determined by Rutter and Wyllie [1988b] in tonalitesand granodioritesfor which there was independent provide encouragingsupport for the empiricalgeobarometer.

Granite/Rhyolite-H 20 I I I I I I I I I Figure 29 providesthe basisfor consideringthe inverseap- 1200 proach for the origin of rhyolite and granite. The

i i50 _ TONALITE 101 H20-undersaturatedliquidus surfaceshows the range of - depth-temperatureconditions for the formationof a liquid of i lOO _ Pressure= 10 _

i i i i i i i i i i ! !

-- 1050 11

lO iOO0 950*•- -- 7• PI+ Cpx +Mt +l Hb-ot• 9 T 950 _ • . _ Cpx-in_ • 8 _.+lkbar-•i ,///// y p'+ Hb +O_px +G_ø +M_t + S_Ph +.L Qz-ou, ..• 7 ',/," 900 PI + Qz + Hb + Opx + Go+ Mt + Sph + L Bi-out' Hollisteret al. •Hammarstromand Zen 1986 •'[• PI+ Qz + Bi+ Hb +Opx +Go +Mt + Sph +L Or-out m 5 85O opx-in-L], - • Or+PI+Qz+Bi+ Hb+ Go+ Mt+ L Sph-inJ Q. 4 Solidus 800, Or + PI+ Qz+ Bi + Hb + Go+ Mt

750 12 I0 20 30 40 50 60 70 80 90 AIT in hornblende % Liquid Fig. 28. Empiricalcurves of total A1 in hornblendeversus pres- Fig. 27. Dehydration melting of tonalite at 10 kbar. Phase sure of crystallization,extrapolated to 10 kbar. The crossesare for boundariesand percentageof liquid through the melting interval microprobeanalyses of hornblendesin partlymelted tonalite experi- [Rutter and Wyllie, 1988a]. mental productsat 10 kbar. WYLLIE:MAGMA GENESIS, PLATE TECTONICS, DIFFERENTIATION 389

RHYOLITE- HzO 15t , • ' •B•?•w•I H•O-undersaturated Coes,t• /' / • ..... f o_. , ]so J ,odeire / /20 IO b r ½J!ii;:::;•iiiiiillliquid compositions J /x. /Ct/•// I00 • • /,// / / ,/ / a'OJ-H20-soturated •iiiiiliij• •I•.:•:;.•:•:•:•:•::: ::::::::::::::::::::::: "======O•'u::::::::::::::: • • ///// // • ••:•:•:•:•:•:•:•:: ••"••/••]...... '•• I 5 •:•:•:•:•:•:•:•:•:•:•:::;:•:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::...... •:•:•:•:•-..::.. 20 renlte CONTNENTAL 01 I •• I/ • I 600 I000 1400 [-2% H•O 7 I TemperatureOC 0• / • I , , "• .... -'-• • I 0 400 600 800 I000 1200 Fig. 29. Phase relationshipsfor rhyolite-H20 modified by Wyllie TemperatureøC [1979] after Stern et al. [1975]. Fig. 31. Compositionsof liquids generatedwithin the melting interval of rocks of continentalcrust in the presenceof H•,O [after Wyllie, rhyolite compositionwith various amounts of dissolvedH20. 1977]. If this particular rhyolite (or others with similar compositions) was a primary magma, the residual minerals left behind in the feldspars and quartz have been replaced by eclogitic minerals partly melted source rock should include the minerals on the plus kyanite (in proportions quite different from those in liquidus. At crustal pressuresthese include quartz and plagio- gabbro, of course).Biotite and amphibole are no longer pres- clase. At mantle pressuresthe liquidus mineral is quartz, and ent at 30 kbar, but the dense mineral sanidine hydrate has quartz crystallizes alone through a large temperature interval become stable, and muscovite remains in more aluminous below the liquidus [Stern and Wyllie, 1981]. There is no granites [Huang and Wyllie, 1981]. The composition of the source rock in the mantle that could leave residual quartz. subsolidus vapor phase has not been determined, but this Rhyolite, therefore, cannot be a primary magma from mantle would be of interest [Schreyer et al., 1987]. peridotite or eclogite. Rhyolite is not generated from mantle source rocks, and Gneiss-H20 granite is a crustal rock. Crustal rocks, however, may be car- The dominant rocks of the deep continental crust consist of ried to considerable depths during the geologically brief the calc-alkaline plutonic rocks and their metamorphic equiv- periods when continental thickness is greatly increased by alents, together with metamorphosed shales and greywackes. continental collisions,as indicated in Figure 5b. Schreyeret al. Wyllie [1977] presenteda detailed review of experimentsre- [1987] have presented evidence for the recrystallization of lated to crustal anatexis for conditions with a pore fluid and granitic rocks beneath the convergent Alps at depths of about for vapor-absentmelting of rocks with hydrous minerals mus- 100 km. The isobaric sectionfor granite-H20 at 30 kbar illus- covite, biotite, or amphibole. trated in Figure 30 is, therefore, relevant for interpretation of Figure 31 shows the approximate P-T ranges for the gener- such processes.Note that the subsolidus mineralogy at 30 ation of liquids of various compositions from continental kbar is completely different from that of normal granite. The gneisses,with a generousallowance of 2% H20 in pore fluids. The temperatureinterval for the persistenceof H20-saturated granitic liquids is narrow. At lower pressures, H20- undersaturated granite (rhyolite) liquid coexists with residual minerals through a wide range of temperatures. Very high J4001 \•\\GF•,b,N/TE' temperatures are required to reach liquid compositions corre- Solidus•,• \ :50 Kb sponding to tonalites (andesites).Increasing pressureproduces ___l/\ Liquid \, / liquids with lower SiO 2, and rhyolite liquid cannot be produced at the base of the normal continental crust. For thick-

Temperature,øC 0(• 500 1000 1500 ooor v_ou, Migmatite •i?:: Granodiorite and t•::--•.'"..".•...... "ilil Bi Hb f CRUST granite•.....:'i...,..'•i [:iii.Jii!:'!i!:.:-,• •"' Gneiss H[:':-..':::....::.•i [.. •-::iKl!::i::i::iiii: Moho

• Ja out 50 MANILE

= • ••. •... "•o •00- o -. o - •V-out •'• - • i -• Ct+Jd+ Ky+OrH+Ga +V -- 2 + 4000 , I0I , 20I , 30I , 40 TOTAL Wt. % HpO IN THE SYSTEM Fig. ]2. The major sitesof magmageDeratioD (stippled) in conti- Fig. 30. Isobaric sectionfor granite-H20 at 30 kbar [Stern and DeDtaJcrust, maDtie, add subductedoceaDic crust (at JOO-km depth). Wyllie, 1981]. Soliduscurves are transferredfrom previousfigures, as cited in text. 390 WYLLIE: MAGMA GENESIS, PLATE TECTONICS, DIFFERENTIATION

A. LAYERED MANTLE C. ECLOGITE LAYER 0 1000 2000·C 0 g~ofh Loyer 1 ~'0

200

.c Loyer 2 0. 400 0'"

600

Perovskite Loyer 3 +(Mg,FelO 800 Richter a McKenzie 1981 Ringwood 1975 1982 Anderson 1982 Fig. 33. Proposed mantle structure and geotherms compared with experimentally determined and extrapolated melt ing conditions for peridotite and eclogite. (a) Experimental peridotite solidus curves extrapolated long either ce or cd. (b) Estimated peridotite melting interval, after Wyllie [1981]. The curve ab is the solidus QR with buffered vapor, from Figure 20. (e) Eclogite melting interval extrapolated from 100 km, from Wyllie [1984, Figure 1]. See text for comparison with Figure 4.

ened crust, the near-solidus liquid approaches syenite compo linearly along c-d. The calculated Simon-type fit along c-e sition [Huang and Wyllie, 1981]. The vapor-absent melting of [Griggs, 1972] is probably a better representation of a perido tonalite gneiss at 10 kbar was illustrated and described above tite solidus, and it matches more closely the recent experi (Figures 26, 27, and 28). mental determination (Figure 34). The peridotite melting in terval in Figures 33b and 33c was estimated by Wyllie [1981]. Depth-Temperature Framework The geotherms in Figure 33 are from various sources, and The solidus curves for the various materials reviewed above these merit comparison with Figure 7. provide a depth-temperature framework useful for consider Jacobsen and Wasserburg's [1981] mechanism for crustal ation of magmatic sources and magma interactions in various growth according to their model I was illustrated in Figure 4a. tectonic environments. Figure 32 provides specific depth This model is represented in 'Figure 33a by the layered mantle temperature limits for some of the magmatic processes sum considered by Richter and McKenzie [1981]. They assumed marized in Figure 6. It shows the continental crust overlying that the separate reservoirs situated below the lithosphere the mantle, with a layer of subducted oceanic crust 100 km (layer 1) consisted of superimposed layers 2 and 3, convecting below the surface. The sites of possible magma generation in without significant mass transfer across the boundary between these various magmatic hearths are identified by shading. them. Jacobsen and Wasserburg [1981] proposed that blobs of High-temperature conditions for the generation of basalts and un depleted mantle rise from layer 3 and intersect the solidus

picrites above AB are from Figure 15b. The influence of H 2 0 at shallower levels. According to the recent determination of or C-H-O species on magma generation in mantle peridotite is the peridotite solidus (Figure 34), material rising from layer 3 given by CD and EQD from Figures 20 and 21. With the with the geotherm of Figures 33a and 33b would begin to melt

availability of H 20 in subducted oceanic basalt, melt can be at depths considerably below the lithosphere. generated in the interval FG (Figure 24). The continental crust begins to melt above H in the presence of aqueous pore fluids (Figure 31). In the absence of pore fluids, melting begins near K when biotite or amphibolite breaks down to form

H 2 0-undersaturated liquid (Figure 27). Figure 32 shows the potential for interaction of rising melts of different compo sitions and temperatures, considering simply mantle magmas rising into the crust or the more complex situation of magmas from subducted oceanic crust rising into mantle and conti nental crust. ~ <:; .= ~ c.. Q) CONVECTION, GEOTHERMS, AND SOLIDUS CURVES Cl ~ '"Cl.) The geotherms and solidus curves in Figure 33 provide tests ct for the different mantle convection models summarized in Figure 4. The solidus curves plotted for peridotite are older versions that should be compared with the curve in Figure 3 by Takahashi [1986], which is reproduced in Figure 34 along with a geotherm (Figure 7) corresponding to the mantle model in Figure 4a. Fig. 34. Phase diagram for lherzolite from Figure 3 [Takahashi, Figure 33a reproduces the different experimentally deter 1986], compared with proposed upper mantle structure and geotherm mined solidus curves for peridotites. These are extrapolated from Figure 33a. WYLLIE' MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION 391

Figure33b illustratesRingwood's [1982] modificationof the physical properties. The densities of rocks and melts are im- pyrolite-basaltmodel (see Ringwood [1975] for a summary),as portantvariables in controllingconvection and the migration sketchedin Figure4d. The lithosphereis composedof deple- of magmas.Viscous forces which act to impedemelt segre- ted harzburgiteand lherzoliteoverlying pyrolite which resorbs gationdepend critically upon the rheologicalresponse of the the depletedmaterial strippedoff during subduction,thus crystal-liquid systemto local conditions. maintaining a continuous source for MORBs. The seismic A magmais definedas a melt-crystalassemblage. There are low-velocityzone is attributedto incipientmelting at temper- threeways for magmaor meltto migratein themantle: (1) the aturesabove the solidus,a-b, for peridotite-H20-CO2 (Figure partiallymolten assemblage may becomedynamically unsta- 18). Ringwood[1975] usedthe calculatedgeotherm shown in ble and migrateupward as a diapir throughits host; (2) the Figure33b, with diapiricuprise of mantlemasses tracing adia- interstitialmelt may percolateupward through the crystalline batic pathssuch as that risingfrom point fi The secondmantle host along grain boundaries;or (3) the melt fraction may reservoircomposed of megalithsof depleted,subducted, cool segregatefrom the crystallinehost in either example,and rise lithosphereenclosing eclogite from oceaniccrust is density through a propagatingcrack. The escapingmagma could balancedat about650-km depth. Attainment of thermalequi- transportwith it someof the hostcrystals. librium after 1-2 billion yearsis followedby upriseof the The questionof how much melt needsto be producedin a refertilizedharzburgite. The modelneeds testing and reviewin partlymolten rock beforethe meltcan segregate(the critical termsof new soliduscurves and geotherms(Figures 34 and 7). melt fraction) is an important topic. Another is how much Anderson's[1982] model of Figure 4c is representedin residualmelt is retainedwithin the pore spacesand intergra- Figure 33c. The low-velocityzone is attributedto incipient nular boundariesof a rock that has beeneffectively drained of melting.The density-balancedeclogite layer perchedbetween magma. This has implicationsfor interpretation of the seismic 220 km and 670 km is the depletedreservoir, and the upper low-velocityzone and the chemistryof the mantle,especially peridotite layer is a reservoirenriched by kimberlite fluids. with respect to LIL elements and to mantle metasomatism. The schematic,extrapolated melting interval for peridotiteof Answers to these questions will be different for mantle Figure 33b is replacedbetween 220 km and 670 km by an magmasand for the more viscouscrustal magmas. estimateof the meltinginterval for eclogite.The extrapolation Maaloe and $cheie[1982] consideredthe processof partial preservesthe narrowmelting interval known to occurthrough melting of lherzolitein a rising mantle plume.They defineda 30 kbar and the proximity of the solidusfor quartz-freeeclo- partial meltingboundary, the level wheremelting begins, and gite with that for peridotite[Wyllie, 1984, Figure 1]. Ander- a permeabilityboundary at a higher level where the partly son's[1982] mantle in Figure 33c has an additional convect- molten source became permeable. They demonstrated that ing layer, compared with the model of Richter and McKenzie once permeable,the mantle would undergo compaction and [1981] in Figure 33a. This introducesanother step into the the melt would be squeezedout, accumulatingin the form of geotherm.The solid line in Figure 33c is a conservativeesti- horizontallayers. The magmamay be tappedfrom a dispersed mate for the geotherm with this mantle model, but the dashed sourceconsisting of a multitude of layersand interconnecting line is a more realistic estimate of what the additional convect- veins.With divergentflow as in the topmostpart of the plume, ing layer would do to the geothermaccording to F. M. Rich- the layers come apart and a large magma chamber may be ter (personalcommunication, 1983). formed [Maaloe, 1981]. Ribe and $mooke[1987] also con- The soliduscurves provide constraintson the dimensionsof cludedthat layersof segregatedmelt are likely to form magma layersin Figures33a, 33c, and 34, becauseof the largesteplike chambersabove plumes at the baseof the lithosphereand that increasesin temperatureassociated with eachboundary. With about 60-80% of the melt generatedwithin the plume is ex- the resultsused by Richterand McKenzie [1981] in Figure tracted from an area comparableto that of the plume itself. 33a, the temperatureat 700 km is not high enoughto cause Nayon and $tolper [1987] modeled aspects of the chemical partial melting of layer 3. They pointed out, however,that if interaction between rising melts and the mantle in terms of the bottom of layer 2 were shallowerthan about 500 km, then ion exchangeprocesses with particular attention to trace ele- widespreadmelting is suggestedby their calculation,and the ments and including specifictreatment of melt in a rising two-layer arrangement would be untenable.Therefore it is not plume. clearhow the two-layerconvecting mantle (Figure 4a) could In a review of magmatic processes,Marsh [1987] con- have developedinto what is presentlya stablecondition, if cluded, "No other quadrennium has contributed so much to layer 2 is the depletedmantle reservoirwhich has grown advancing our understandingof magmatic processes."He downwardthrough time at the expenseof undepletedsource noted that we are in the midst of an extremelyfertile burst of layer3, asin modelI of Jacobsenand Wasserburg[1981]. The activity with petrologistsactively seeking evidence for and problem is exacerbatedif mantle temperatureswere even againstvarious physicalprocesses and applying the essentials higherin early Earth history. of the mechanicsof magmas with quantitative insights.He Considerationof soliduscurves and geothermsis essential reviewed developmentsin understanding the role of com- for testing the validity of mantle models, and as data from paction in deforminga solid to closethe poresin a rock and experimentalpetrology and geophysicsimprove, they may fa- thus force out interstitial liquid [McKenzie, 1984] and the cilitate discriminationamong interpretationsbased on geo- transport of liquids in solitons, solitary waves of increased chemistry. They do provide close constraints on subduction percentagesof liquid which travel as pulsesdeforming the and lithosphereprocesses. rock matrix [Scott and Stevenson,1986]. Theseprocesses may be important in the mantle. In the crust, however, where silicic MAGMA GENERATION AND TRANSPORT melt viscositiesare much larger, the conditions over which The combination of isotherms and solidus curves shows compactionis effectiveare morelimited. Marsh's [1987] possible sites of magma generation, but the mechanism of reviewalso covers additional physical topics crucial to under- meltingand the processof magmamigration are controlledby standingdiapirism, chemical plumbing systems,processes of 392 WYLLIE: MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION

convection and mineral sorting in magma chambers, and the Distance from trench km withdrawal and eruption of magmas. 0 50 i00 150 The variation of density of melts as a function of pressure, Oceanic Continental temperature, and composition is a fundamental parameter in all magmatic processes.The densitiesof molten silicatesup to 230 kbar have been measuredfor the first time recently using 50- CRUSTJ •,. \ , agma Moho- shock wave techniques [Rigden et al., 1984]. Results were obtained for a synthetic model basaltic composition and were i SG':••EIHornblende peridotireSP• compared with density-depth relations for the Earth derived H3 from recent seismologicalmodels. Some implications of the results are as follows: (1) melts become denser than host serpentineS61-S62 SE mantle at pressuresof 60-100 kbar (200-300 km in the Earth), 100StobilityHornblende limits in of[ HI -H2 (2) there is a maximum pressurefrom which basaltic (or pic- peridotireH3-H 4 ritic) melt can rise within planetary interiors, (3) the slope of Soliduswi•h H20 for silicate melting curves should become very shallow at high 150 6obbro/eclogHeSEi-SE2 pressures(see Figure 3), and (4) enriched reservoirsformed by sinking liquids may exist at depth within the Earth. Fig. 35. Cross section of a highly idealized 45 ø subduction zone, showing (for a particular set of isotherms) the petrological structure of These results have important implications for early evolu- subducted oceanic crust, the overlying mantle wedge, and the conti- tion of the Earth and the origin of komatiites. Komatiite melt nental crust, with particular referenceto the regions of potential melt- formed at depth would be too denseto rise. It would sink, and ing if H20 is present. the olivine with which it could coexistwould float to generate a refractory cap. The concept of a magma ocean formed in MORB's was reviewed in detail by Wyllie et al. [1981]. The this way and implications for the thermal and geochemical cycle involving piclogite, basalt, and subducted eclogite was evolution and stratification of the mantle have been discussed developed by Anderson [1982]. It is now evident that many by Nisbet and Walker [1982], Anderson [1982], and Ohtani complex processes including magma mixing and differ- [1985]. entiation have occurred in shallow level magma chambers be- There is another kind of control exerted by density on neath the ridges. A recent review by Fallcon and Green [1987] magma migration. The melt in equilibrium with mantle perid- presents new experimental data supporting the idea that otite changes composition as it rises from picritic (more dense) MORBs are derivative from primary picritic magmas gener- to basaltic (lessdense). Picritic melt, if emplaced into a basaltic ated at depths greater than 30 km. magma chamber in the crust, is denser than the basalt melt and may form a separate layer beneath it. This has interesting CONVERGENT PLATE BOUNDARIES petrological consequencesthat have been explored recently. Picritic melt is also denser than continental crust, and low- There are many variables in petrology and geophysicsat density crustal rocks could act as a density filter preventing convergent plate boundaries, but the emergent plutonic and the eruption of picritic lavas. volcanic rocks have common, distinctive chemical character- With crystallization of a magnesian melt at any depth, the istics and organized structural geology [Gill, 1981]. These subtraction of crystals from the melt causes a progressive characteristicsindicate crustal differentiation modified by dis- change in its density. Stolper and Walker [1980] calculated the solved H20 [Grove and Kinzler, 1986], but the evidencefor change in density of such liquids versus fractionation. The multiple source materials is clear [Carlson, 1987]. The sources densities of residual liquids pass through a minimum when include subducted oceanic crust, overlying mantle, and conti- pyroxene and plagioclasejoin the crystallization sequenceand nental crust if present. Subducted pelagic sedimentsmay con- then rise to a maximum corresponding to the stage where tribute to the magmas. Bodies of serpentinite in the oceanic ilmenite begins to crystallize. They also demonstrated that crust are a source of aqueous solutions. The subduction of oceanic basalt compositions cluster around the composition volatile components distinguishes this tectonic environment correspondingto the density minimum for differentiating liq- from others, and the influence of these components after ther- uids. Magnesian melts trapped in the crust by their density mal dissociation produces magmas of distinctive composition. may evolve by crystallization to melts with densities lower Ocean-Continent Boundary than the overlying crust, which then permits their eruption. The "window of eruptibility" is restricted by the subsequent Applying experimentally determined phase boundaries for increase in melt density, and it could be this filter effect which the three source regions in the presence of aqueous fluids favors the eruption of the observed basalts compared with (Figures 16, 22, 31, and 32) to specific thermal structures denser,more or lessfractionated liquids. (Figure 12) provides a framework locating sites of dehydration and potential melting. Figure 35 shows the boundaries for an idealized 45 ø subduction zone, for a relatively warm crust and DIVERGENT PLATE BOUNDARIES cool mantle. The points where wet solidus curves cross the The major chemical differentiation of the Earth is accom- amphibole-out curves are located in Figures 16 (P) and 22 (Q); plished at mid-ocean ridges by the separation of MORBs from these points occur at fixed pressures,but they may move hori- mantle lherzolite or from piclogite or from some mixture of zontally in Figure 35 as the thermal structure varies. Magmas peridotite and eclogite (Figure 4). The lithosphere is thin, and are developed in the shaded areas only if aqueous vapors pass mantle uprise yields an oceanic geotherm approximating d-d through them. The locations of the phase boundaries and thus in Figure 7, crossing the solidus of lherzolite (Figures 2, 33, the petrological structure are sensitiveto changesin thermal and 34). The application of phase diagrams to the origin of structure. Four examples are shown in Figure 36, correspond- WYLLIE'MAGMA GENESIS, PLATE TECTONICS, DIFFERENTIATION 393

A v C 0 0••..

I0•on,'%••••?l _ _ --]...... _S•i •,_ _ _

] Amphibølite••••• • ,, •'k•• •mpn,0ole-out ,00 ,00 I •. _t •7•,• %.. 150 1501 • Eclogite• •-•

200 200 ----- 0• nt '---'--' - ,H•,O • H20 B ,c• .... Continental• ..... 1---ICo inentol Crust 50 Greenschist• • "• ""'"•"'•"••'•'••••'•'•••'••••• H20Mantle A phibolite E 100 • mphi 00

150 -•o • I• •,u•ou• ,,ueou, ' , I 7'u"ø. PoteMialsolidus • 001 withH20 ....'•'• Met0somotic withH20 • ["•?•./,;•0mphibole • generationMoore0 '• pyroxenitePhlogopite- Fi•. 36. Locationsof dehydration, mctasomatism, andmcltin• in subductionzones from results of experimental pctrolo•y,according tofour thermal structures obtained bycombinations ofwarm or cool (by cndothcrmic dehydration) subductedocean crust, with cool or warm (by induced convection) mantle. (a) Cool mantle-warm crust. (b) Cool mantle- coolcrust. (c) Warm mantlc•ool crust. (d) Warm mantlewarm crust [Wfilie and Sekine, 1982]. ing to the four combinationsof warm and cool mantle and largeshaded areas of mantleand continental crust and gener- crustdiscussed in connection with Figure 12. These figures are ate magma in the continentalcrust (C). For a cool mantle- designedto illustratethe kinds of processesand products that warm crust (Figure 36a), most solutions enter the mantle canoccur, rather than what actually happens in anyparticular alongD-E, but somegenerate magma from subductedcrust at subduction zone. M. For a coolmantle-cool crust (Figure 36b), the metasoma- Dehydrationoccurs wherever rock masses are transported tic solutionsdo not enterthe meltingregions of eithersubdue- acrossthe dehydration boundaries in Figure 36. The dehy- ted crust or mantle peridotite.For a warm mantle-cool crust drationfront for serpentinite(D-G) approximatesthe greens- (Figure36c), induced convection brings the regionfor poten- chistfacies boundary, and the dehydration front for amphibo- tial meltingof the mantleabove the dehydrationfronts, and le separatesamphibolite from eclogite in the subductedocean- metasomaticfluids cause partial meltingof peridotite.Note ic crust.Partial melting occurs if H20 passesinto theregions how dehydrationof metasomaticmantle amphibolecould on the high-temperature sides of the dashed solidus bound- localizeenhanced melting at N, perhapscontrolling the posi- ariesfor rock-H20(shaded areas in Figure35). tion of the volcanic front. Aqueoussolutions rising from the dehydrationfront D-G Figure 36d represents an intermediate thermal structure may migrateback toward the trench along the dislocation with relativelywarm mantleand relativelywarm subducted abovethe slab,but directupward migration of solutionsinto crust. Magmas are generatedsimultaneously in all three theoverlying mantle wedge is commonlyassumed (Figure 5c). sourcematerials at M, N, C, and H' [Andersonet al., 1980]. Thesesolutions may precipitate metasomatic amphibole in the Nichollsand Ringwood [1973] proposed that hydroussiliceous 394 WYLLIE' MAGMA GENESIS,PLATE TECTONICS, DIFFERENTIATION

magma from M would react with overlying mantle, producing occurs subsolidusin tonalite and granite at pressurescorre- hybrid olivine pyroxenite. Wyllie and Sekine [1982] concluded sponding to a depth of about 100 km (Figures 26 and 29), fromtheir experiments [Sekine and Wyllie, 1982, 1983] that a confirming the conclusionsof Schreyer et al. [1987] based on seriesof discreteor overlapping bodies of hybrid phlogopite- other phase studies. They estimated temperatures of 700ø- pyroxenite would be transported downward with the slab to 800øC for the metamorphism, which would place the rocks R, where vapor-absentmelting definesanother possiblesite of closeto or above their soliduscurves in the presenceof aque- magma generation. Solutions releasedduring hybridization at ous pore fluid accordingto Figures 26, 29, and 30. The possi- H cause metasomatism, followed by partial melting in the bility that the subductedmetasediments released trachytic shallower mantle at N. Another possibilityis that the hydrous melts or KMg-rich fluids into overlying mantle was evaluated magma from M could enter the region N for hydrous melting by the authors, who noted that the trace element signatureof of peridotite. lamproites is best explained by contamination of normal Release of vapors from deep dissociation fronts in subduc- mantle with material from continental crust prior to magma ted oceanic crust leads to migration of fluids through large formation. volumes of upper mantle. The geometry of the arrangement of metasomatic volumes, and volumes where the fluids dissolve Magma Contamination and Mixing in magmas, is strongly dependent on the geometry of subduc- The dehydration and partial melting of subducted oceanic tion, the thermal history of the plate being subducted,and the and continental crust introduces material for contaminating thermal structure of the subduction zone. Note in Figure 36d mantle-derived magmas, and basalts from the mantle have the complicated sequenceof events that could be experienced similar potential for contaminating magmas derived in the by water stored initially in serpentine in the oceanic crust. continental crust. Magma mixing was not a popular process Figure 14 shows the types of phase fields traversed by the until about a decade ago. In their influential textbook, Carmi- water. It starts in the "vapor-absent rock," then alternates chael et al. [1974, p. 67] wrote: "Mingling of magmas is no between fields for "rock + V" and "xls + L" (with very short longer consideredto be a principal factor in evolution of such intervals of "xls + L + V" between).Following dehydration at continuous magma series," although they did not rule out about 100-km depth (Figure 36d), the aqueous vapor would completely the possible role of magma mixing in volcanic rise through eclogite, cross the solidus, and dissolve in inter- rocks of the orogenic, calc-alkaline series.The frequencywith stitial melt near M. The magma would solidify by hybridiza- which magma mixing and contamination are now called upon tion, releasing H20 at N, forming vapor that would rise demonstrates that they are once again respectable processesin through peridotite until it crossed the mantle solidus near N, the petrogenesisof many igneous rocks. Marsh [1987] noted generatinga secondbatch of magma. This would rise into the that although magma mixing continues to be a popular pro- base of the continental crust at H', perhaps escapingin erup- cess,its nomenclature is in a shambles and the processis not tion B but more likely becoming involved in additional reac- clearly defined. To some the process means homogenization, tions between H' and the surface. to others it means simply mingling, and yet others mean The fifth arrangement advocated by Marsh [1979] and merely contamination. Metasomatism may be superimposed Maaloe and Petersen [1981] with convection in the mantle for some mantle magmas. Sparks and Marshall [1986] pre- wedge raising the temperature of the subducted oceanic crust sented a detailed evaluation of thermal and mechanical con- at 100-km depth to 1250ø-1450øC is hot enough to generate straints on mixing between mafic and silicic magmas. primary basaltic andesite directly from the subducted crust. The transfer of melts and vapors from one region of the Johnston[1986] studied the liquidus phase relationshipsfor a mantle to another producing a nonequilibrium situation was near-primary high-alumina basalt and concluded using an in- illustrated in Figures 32 and 36 (see also Figures 42, 43, 44, verse approach (Figure 13) that this compositioncould be and 46). In a subduction zone environment, there is an es- derived as a primary melt from subductedMORB. Brophy and pecially high potential for interactions among melts, rocks, Marsh [1986] considered the partial melting of subducted and vapors in both mantle and crust, as illustrated in Figure quartz eclogite (with H20 playing a minor role) and con- 36. There are five situations involving magmas and rocks of cluded that gravitational instability of the entire sourceregion particular interest: (1) the intrusion of hydrous, siliceous would occur before melt extraction was possible.Uprise of the magma from the subducted slab into cooler mantle peridotite mush with increasing melting to 50% liquid would continue (H, Figure 36d); (2) the intrusion of the same melt into partly through 15-20 km, at which stage the melt could escapefrom melted, hydrous peridotite at somewhat shallower levels (N, the crystalline host and rise diapirically. The liquid at this Figure 36d); (3) the intrusion of basic magmas into the deep stage would have a pattern of REE equal to that observedfor continentalcrust; (4) the intrusionof basicmagmas into partly islandarc magmas.They presentedother arguments involving melted continental crust or into siliceousmagma chambers(H the mechanicsof ascentand melt extraction along with phase and H', Figures 36c and 36d); and (5) the intrusion of grani- equilibria to support their conclusion that partial melting of toid magmas into shallow crust. There are few experimental quartzeclogite best satisifes the geochemicalconstrain ts of data providing information about the complex processesoc- island arc basalts. curring in these regions of hybridism, contamination, or mixing. Some examples and their relationship to subduction Continent-ContinentBoundary are illustrated in Figure 37. Schreyer et al. [1987] reviewed convincing evidence that Hybridism between hydrous siliceous magma and mantle metasedimentsfrom the Alpine/Mediterranean area had been peridotite at 100-km depth, a process proposed by Nicholls subductedto depths of at least 100 km. Pyrope-coesiterocks and Ringwood [1973], has been explored experimentally by with individual pyrope crystals up to 20 cm in diameter in- Sekineand Wyllie [1982, 1983] using either mixtures of gran- clude varieties with kyanite and jadeite, as well as phengite ite and peridotite with H20 or reaction couplesof hydrous and talc. The assemblage pyrope-coesite-jadeite-kyanite graniticmelt adjacent crystallineperidotite (Figure 37, WYLLIE' MAGMA GENESIS,PLATE TECTONICS, DIFFERENTIATION 395

0 magma by potassium is likely. He used the results to illustrate quantitatively the principles of selectivecontamination during Amphibolite G2kb basalt/crust interaction. Watson and Jurewicz [1984] de- •,Granite scribed additional experimentsaimed specificallyat the chemi- melt cal diffusivity of potassium and the partitioning of sodium. Basalt melt These experiments were conducted with dry materials, and 10-15kb Mantle Watson [1981] concluded that diffusion rates in crustal r• G magmas are extremely sensitiveto H20 content. Johnstonand Peridotite Wyllie [1986b, 1988] reported results of a first set of experi- 100 -- ments at 10 kbar designed to explore the properties of Mantle coexisting basaltic and rhyolitic magmas and to determine how these properties vary as a function of pressure, temper- Fig. 37. Locations of four reaction experiments designed to ex- ature, H20 content, and other variables (Figure 37, number 3). plore interactions between magmas and rocks from subducted slab to They found that even with H20-undersaturated conditions at shallow crust. 920øC, Na20 and K20 were rapidly equalized between the two melts, CaO, MgO, and FeO migrated somewhat less rap- number 1). Figure 38 is a sketch of a polished surface through idly, and the gradient in SiO 2 and A120 3 persistedfor a longer a typical experimental reaction charge. In the mineral reaction period. zones produced by hydridization of H20-undersaturated M. R. Carroll and P. J. Wyllie (manuscript in preparation) granite liquid and the overlying peridotite, orthopyroxene and also conducted experimentswith H20-undersaturated granite jadeitic pyroxene are ubiquitous products with compositions glass adjacent peridotite at 15 kbar designed to study iso- different from those in peridotite. thermal contamination and element diffusivities. They dis- Figure 39 shows Sekine and Wyllie's [1982] framework of covered instead that the granite melt experienced remarkable liquidus field boundaries at 30 kbar associated with peridotite convective motion, as shown in Figure 40. Unmelted crystals and the calc-alkalinerock seriesin the presenceof H20. John- of feldspar and quartz at the far end of the granite melt ston and Wyllie [1986a] studied the phase fields intersectedby became entrained in the flow defined by trails of graphite, mixtures on the join tonalite-peridotite-H20 at 30 kbar, and which formed by reduction of the small amount of carbonate confirmed the position of the boundary between the liquidus in the granite. Despite the reduction of viscosity by dissolved fields for garnet and orthopyroxene in Figure 39, on the MgO H20 and the relatively high temperature of 920øC at 10 kbar, side of the calc-alkaline series BAR. Similar experiments con- convection was not expected in such a small sample. The phe- ducted by Carroll and Wyllie [1987] at 15 kbar, correspond- nomenon is reproducible and is currently being explored in ing to the depth of the Moho in orogenic belts where the crust detail. Similar contamination experiments between is thickened (Figure 37, number 2), show that the garnet- H20-saturated granite and crystalline amphibolite at 800øC orthopyroxene boundary (with H20 ) becomes almost coin- and 2 kbar (Figure 37, number 4) produced results as ex- cident with the calc-alkaline seriesin this projection. pected: diffusivities in granite melt are very slow under these Watson [1982] studied experimentally the chemical pro- conditions [Rutter and Wyllie, 1986]. Figure 41 is a sketch of cessesresulting from contact of molten basalt with felsic min- a typical experimental sample. Bubbles in the glass are caused erals, concluding that selective contamination of the basaltic by the small amount of CO 2 in the granite. Note the bubble- free zone against the contact. This was slightly wider than the zone of element diffusion.

i i 0.25mm A • Peridotire Co0 IlOOFm::•:"'•• HydrousReactionglass Zone Fig. 38. Sketch of polished sectionof quenchedrun products of Fig. 39. Pseudoternary liquidus field boundaries at 30 kbar with crystalline peridotite above H20-undersaturated granite melt. The excessH20, separatingareas for the primary crystallizationof min- run was at 30 kbar, 1000øC for 45 hours. Minerals in the reaction erals, based on experiments with calc-alkaline rock series, BG, and zone are orthopyroxene,jadeitic pyroxene, and quartz [Sekine and with mixturesbetween peridotite and granite,PG [Stern and Wyllie, Wyllie, 1983]. 1978; Sekine and Wyllie, 1982]. 396 WYLLIE:MAGMA GENESIS, PLATE TECTONICS, DIFFERENTIATION

.....' :•::-•. • ,. ß ,:.z•.•..:::....::::.-:•:.... ;•:::•::...... -.::-:: .: ....::`:``::.•:.:...:.:...:.;.:•::.:::::..:;•.::.:.:.:.•:z.•:•.::•:•::::•:.•:•t•:•::•:•:•;•?•:•;•:•;::•:*•::;:::•;•:•:::•:• -•.:..::..•-.;. ß'• ...... : ....: , •

•:;'•:.;•;-•dv•f.:.'.,':"'" '"' .-:.'•7'•;;.:.;•;'•,,½•:;:,t;•'•;'. . ,,.--.,..•½• " ...... • ==. • ' ,• '.'

:. .;. ,.' ..

•:?.?? ...'

...... ,...:'•...... '"•'"•'•-• ...... ß..... '" ......

Fig.40. Thinsection (transmitted light, nicols partly crossed) ofreaction experiment between largely molten hydrous graniteand solid peridotitc (left side) at 15 kbar and 1050øC. Vertical field of view is 0.2 mm. Position number 2 in Figure 37.Convection isshown by trails of graphite (from trace of reduced carbonate) emanating from cluster of unmelted quartz plusfeldspar crystals atleft of charge. From M. R.Carroll and P. J. Wyllie (manuscript inpreparation).

The proposalthat hydroussiliceous magmas rising from foundthat the positionof this reactionline movedvery close subductedoceanic crust react with overlyingmantle peridotite to the calc-alkaline series at 15 kbar. Thus if melts from the [N•chollsand Ringwood,1973] can be evaluatedin termsof subductedslab experiencedcontinued reaction with the host Figure39. The liquidsgenerated in subductedcrust follow mantle as they rose through the mantle, their compositions pathson the CaO sideof thecalc-alkaline series (j-c). $ohnston might be directedtoward the calc-alkalinecompositions as andWyllie [1986a] demonstrated that the liquidschanged to theyapproached the mantle-crustboundary. For a relatively the MgO sideof the volcanicline if theyreacted with perido- warm mantle capableof yieldingbasalts, however, the slab- tite at 30-kbar pressure(line i-k). Carroll and Wyllie [1987-[ derived melts would have to traverse a region of partially meltedperidotite (Figure 36d).

OCEAN PLATE The hot spotsgenerating the enrichedOIB's in Figure 4 are Gold Cal•sule commonlyassumed to involvethermal plumes, although other processesinvolving propagating fractures and shear melting have been proposed.The concept of plumes may include Molten sourcematerial rising from the core-mantleboundary (Figure IoOdo o o o granite 4b), from a primitive layer below 650 km (Figure 4a), from a o o o megalithor mesosphereboundary layer at the 650-km level o o o o o (Figure 4d), from a shallow enrichedlayer below the litho- o o o o sphere(Figure 4d), or from large-volumeisolated blobs within With bubbles o o o o the mantle (Figure 4b).

o o o o o Bubble free Hawaiian Islands The Hawaiian Islands are the most intensively studied Crystalline OIB's, but accordingto Frey and Roden[1987, p. 434-1,"There hornblende is no consensuson a general petrogeneticmodel for Hawaiian -gabbro volcanism." This leaves considerable latitude in setting up a thermaland petrologicalframework, an exerciseI attempted elsewhere[Wyllie, 1988a] and will summarizehere. The physicalframework is developedby superposingisotherms ' 2mm and phaseboundaries and consideringthe flow lines trans- portingmaterial. Some of thestarting points are controversial, but similar frameworks can be constructed with alternative Fig. 41. Sketchof samplerun in goldcapsule to studyreaction assumptions.The key phaserelationships were summarized in betweensolid amphibolite (below) and hydrous granite liquid (above). Figure 21, and the temperaturesare givenin Figure 9. The Positionnumber 4 in Figure 35. Bubblesare from traceof CO• in thermal structure is based on theoretical models constructed granite;note the bubble-free zone adjacent to theamphibolite [Rutter and W ylhe, 1986]. by Courtheyand White [1986], with asymmetryestimated, WYLLIE' MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION 397

200 100 0 100 200 km o boundariesare static,fixed in positionas long as the iso- thermsdo not change.Material migrates through the frame- <:Lithosphere' !'I ii', ',! • ' work,following flow lines represented by the arrowsin Fig-

IOO ures42 and 43. The positionsof the phaseboundaries in this structurebecomes modified by energychanges associated with Solid..• \,/',11• •m meltingand the extraction of melts,but those plotted provide ,'11 i 'u - Troceofmelt ?!'" i,\U i •/ ' U//':• v,-- - a guide as to what happenswhere, within the crosssection. 200 w,h l'11 I - If thereare volatile components in the deep mantle, H20 is ...... /'/:".'..'•:...... :•'•;.;..• (.;-I'l-U releasedfrom the hydratesas the plumematerial crosses the dehydrationboundary in Figure 42 rWoerrnannand Ro- •00 senhauer,1985], joining the carbonspecies rDeines, 1980]. Greenet al. [1987] consideruprise of CH,• and H 2. The URising • plume II •rH•iSte mantle plume with volatilesrises until it reachesthe solidus oI ' boundary, where the volatilesbecome dissolved in a trace of melt. In the centralpart of the plume, for the thermal struc- Distance,km Fig. 42. Pctrological structure beneath Hawaii based on the i$o- turein Figure9, thehydrated peridotite melts directly without thermsin Figure9 and the phaseboundaries in Figures20 and 21. an intervalfor a separatevapor phase. Significant melting of 'Materialfollowing flow linescrosses the phaseboundaries, causing the plume occursin the shadedarea M, which is situated thedistribution of vaporand melt in risingplume as shown.Major abovethe volatile-freesolidus (see M in Figure21). In the plumemelting occurs at M, withmagma chambers forming at various outerenvelope of the plume,the entrainedvolatile-charged levels. Figure 43 shows more details in the region of the meltbypasses the regionM, enteringthe lithospherenear the asthenosphere-lithosphereboundary layer (heavy line). 1200øCisotherm. At thislevel the changein rheologywould significantlyretard the upwardmovement of the smallper- centage of interstitial melt. and with the constraint that Hawaiian lavas were derived from a sourcerock containinggarnet (still controversial)re- Figure43 showsin moredetail what may happenin the quiringthat meltingoccurred at pressuresgreater than the lowerlithosphere. The compositions of the magmas reaching soliduscusp for garnet(Figure 3), that is in the regionM in the lithosphereare labelledaccording to the experimentally Figure 21. With this constraint,temperatures in the centerof determinedcompositions of meltsunder the specifiedcon- theplume must exceed 1500øC. If thelithosphere plate moves ditions(Figures 15b, 17, and 18).The picriticmagmas enter fasterthan the plume,the plumemay be bent overto oneside thelithosphere and remain there in magmachambers [Maaloe or shearedoff [Presnalland Helsley, 1982]. and Scheie,1982; Ribe and Srnooke,1987] until suitabletec- The relationshipof geothermsto the phaseboundaries for tonic conditionspermit their uprisethrough cracks to shal- lherzolite-C-H-Ois shownin Figure21. Mappingthe phase lower magma chambers,where they differentiateto the boundariesfrom Figure 21 in terms of the pressure-tholeiites and alkali basaltserupted at the surface.The concentration of magmas in chambers at the base of the litho- temperaturepoints of Figure 9 yieldsthe petrologicalstructure in Figure 42, with an expanded view of the spheremay cause partial melting of the overlyinglithosphere [Chen and Frey, 1985]. asthenosphere-lithosphererelations in Figure43. Thesephase Figure 21 indicatesthat the solidustemperature for peridotite-C-H-Ois increasedwith reducedoxygen fugacity, and D. H. Green(personal communication, 1987) now has 0 200, 100, ,• • • 10,0 20,Okm experimentalconfirmation that this is so. If the oxygenfu- gacityin the mantlewere known, then followingthe experi- .. • • Moho fholm•fe'•I ilI I'I Lithosphere mentaldata the deepsolidus boundary below the major melt- olEoS,, p,cr,e, !, I!I I . ing regionM in Figures43 and 44 could be remappedat 50-.. nephelm•te'' '• •1 I,I, •--•q'"•o I' p•cr,eor shallowerdepths, making a smallerenvelope of mantlewith a vopor '1 • • I/f•..•p • •.... N :••;::• i:" • ,-"•...... , '-•.• :•;•nephelinite trace of volatile-chargedmelt. The fate of the volatile-chargedmagma differs according to its positionwithin the plume.The melt in the centralregion is incorporatedinto the muchlarger volume of picriticmelt. The ' • elt•;•••• •:: • melt on either sideof the regionM reachesthe lithosphere and then followsdifferent routes according to its location on the upstream(U) or downstream(D) sideof M. Depending upon the relativerates of mantle flow and melt percolationin 200 100 0 100 200 the regionU (Figure 43), magmaon the upstreamside could Disl•nce,km percolateslowly through the lithospherewith decreasingtem- Fig. 43. Petrologicalstructure of lithosphereand upperplume perature to the solidus,or percolate upward and downstream correspondingto Figure 42, with expandedvertical scale. Picritic from U alongthe asthenosphere-lithosphereboundary layer, magmasenter lithosphere from meltingregion M. Volatile-chargedand become incorporated into the hotteralkali picrite magma magmasenter the lithosphereon eitherside of M, with subsequentabove the upstreammargin of the meltingregion M. Melt percolationretarded. As they approachthe solidus,vapor is evolved enteringthe lithosphereon the downstreamside, D, would be (e.g.,at N) enhancingthe prospectof eruptiondirectly through the lithosphere;paths followedby thesemagmas to the solidusdiffer carriedaway from the plume along lithosphereflow lines as accordingto theirposition on the upstream(U) or downstream(D) shown in Figure 43, remaining above the solidus of the side of the plume. peridotite-meltsystem as temperaturedecreased along DN. 398 WYLLIE.' MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION

The distance downstream that it would reach the solidus de- ß..-- Kim berlite • CRUST pends on the relative rates of lithosphere movement and of migration of small bodies of magma through the lithosphere. NMoho When magma approaches the solidus N, vapor is released, LITHOSPHERE lOO - Failed - and this may enhance crack propagation and permit eruption Kimberlite, Metasomatism• ':'" of nephelinitesor basanitesdirectly from the asthenosphere- lithosphere boundary region to the surface. Fundamental __. questions to be addressedinclude the relative rates of move- .:.:.'!i!'2•:' •.",:." ':•.,-•Ji: Solidu s ment of solid, vapor, and melt within the plume and in the Traceof melt) lithosphere and the chemical consequencesof differential flow W•H b-•]-u rates. 300 -- ...... Vapor .:,;:...: Solidus . ß .'.. -- ß.. C_H_O----.-- .•(; /.:.: .. Macd•natic Sequence Sparsehydrated minerals MagmaI PS-ffl If the Hawaiian Islands were generated over a thermal DHMS Brucite Vapor :.:• plume, the sequenceof magmatic events should correspond to Fig. 44. Craton section corresponding to Figure 8, with phase those illustrated in Figure 43. A specific surface area receives boundaries transferred from Figure 20. Migration of deep volatile magmas in successionfollowing the processesdepicted in components, generation of melts in asthenosphere, and their en- Figure 43 from right to left. The first event could be eruption trapment in lower lithosphere. Escape of volatile components as of nephelinites (or basanites),followed by, or almost contem- magmas approach the solidus leads to metasomatism, crack propagation, and intrusion of kimberlites. Various stages of this process are poraneous with, the eruption of alkali basalts. Then comes illustrated covering a long time interval. major eruption of tholeiites with no time break. The width of M and the time interval for the eruption of tholeiites are extensive in comparison with the initial alkalic magma phase. CONTINENTAL PLATES The third phase is the eruption of alkali basalts generated at the downstream edge of the melting region M; if some of the Many varieties of magmas and petrographic associations alkali picrite becomes trapped in the lithosphere and is trans- are represented on continental plates. In this section we con- ported in the direction downstream before escaping to shal- sider three groups, two of which are associated with volatile lower levels, this stage could be somewhat extended in time components from the mantle in cratonic or rift environments and intermittent. The final stage occurs much later, when the (kimberlites, nephelinites and carbonatites) and the third as- small pockets of nephelinitic magma are erupted from N. By sociated with a hot spot bringing basalt and heat into the the time this occurs, the original volcanic pile, migrating with crust and generating anorogenic granitoid rocks. the plate above DN, has become deeply eroded. Thereafter, Cratons and Kirnberlites magmatic activity ceases until the lithosphere crosses some other thermal disturbance in the mantle capable of raising the Kimberlites are interesting igneous rocks not only because geotherm to a level where it crossesthe solidus. This sequence they bring diamonds to the surface, but also because they of events appears to satisfy the broad geological and pet- bring rock specimensfrom as deep as 150-200 km within the rological history of the Hawaiian Islands [Decker et al., 1987]. lower lithosphere. All kinds of geochemical and geophysical There have been many proposals to explain the geochemis- information can be gleaned from these specimens including try of the different lavas [Wright and Helz, 1987; Carlson, fossil geotherms for the time of eruption of the kimberlite 1987; McCallum, 1987], but there is no consensus[Frey and (Figure 7). Approximately 25 million tons of rock were hauled Roden, 1987]. There are at least three distinct mantle reser- out of the famous Kimberley mine, and this included about 3 voirs involved in the melting process. If the central plume tons of diamond. The ratio carbon/kimberlite is very small, material is assumed to rise from an undepleted lower mantle but a 3-ton pile of diamonds would look rather impressive.In layer, or some other deep enriched mantle sourcesuch as a addition to diamond, many kimberlites contain carbonate, mesosphere or core-mantle boundary layer (Figure 4), the much of which precipitated at shallow levels from CO2 carried outer envelope of the plume may be derived from the upper from greater depths. There is no doubt that carbon was pres- mantle layer (depleted),and the lithosphereis further depleted ent in the depth region from which kimberlite was erupted, and differentiated by removal of MORB constituents. The and experimental petrologistsso far know of no way to gener- framework outlined in Figures 42 and 43 indicates that differ- ate kimberlite magma from mantle peridotire without involv- ent deep source materials could be involved in the formation ing CO 2 [Ecdcdler,1978; Wyllie, 1978c]. of the tholeiites and of the nephelinites, with a prospect that The question of oxygen fugacity in the mantle and its effect more than one deep sourcemay provide material for the alka- on speciationin C-H-O vapors and the consequencesfor melt- lic magmas near the margins of the melting zone. The litho- ing temperatures and magmatic products remain contro- sphere reservoir may contaminate nephelinitesand alkali ba- versial, and it has specialsignificance for the origin of kimber- salts on the upstream side of the plume, U, but this is less lites. A relatively oxidized mantle associated with kimberlite likely on the downstream side, where the thickening litho- genesisis preferred by Eggler and Baker [1982]. Wyllie [1980] sphereis composedof plume material. The tholeiites from the presented an interpretation of kimberlites involving the uprise deep mantle reservoir may be contaminated with the litho- from below the asthenosphere of reduced gases in the system sphere to the extent that the hot picrite magmas can cause C-H-O which dissolved in melts in the asthenosphare,became local melting in the lithosphere [Chen and Frey, 1985]. The oxidized during magmatic processesinvolving diapiric uprise scheme in Figures 42 and 43 is presented as a flexible frame- into the lithosphere, and were released at shallower levels as work that can be modified with new data. H20 q-GO 2. Several alternative models were suggested,one WYLLIE.'MAGMA GENESIS,PLATE TECTONICS, DIFFERENTIATION 399

K,mbeJ-Ilte degreesof melting between these two levels if volatile compo- CRUST nents are present. For simplicity, the two dehydration bound- - MOHO aries of Figures 20 and 21 were merged into a single dehydration boundary. Intermittent local disturbances could re-

IO0 _ lease vapors from the deep dissociation front [Woermann and Fa•led - Rosenbauer, 1985]. The volatile components would rise Metasomatlsrn..,. . Klrnberllte, through mantle to the solidus at level 2, where they would .... ß. • • :,: : .. dissolve in melt. The trace of interstitial melt so developed

200 - ... •.."'.'. would rise through the asthenosphere to the asthenosphere- lithosphere boundary at level 1. The rate of percolation of magmas entering the depleted lithosphere would be reduced. Small magma chambers could form and remain sealed within \qz• / Solidus the more rigid lithosphere, maintained at temperatures above 300 _ • 1I•1 I the solidus for lherzolite-C-H-O. The magmas have little tend- ',•JI mMagma :.'•Va por ency to crystallize or to evolve vapors until they approach level 3, 10-15 km above the asthenosphere-lithospherebound- Fig. 45. Upper mantle section correspondingto Figure 10, with ary. Contact of the lherzolite-derived magma with harzburgite, phaseboundaries transferred from Figure 20. The mantle plume (with volatile components)melts, with concentrationof magma above the however, should cause reaction and the precipitation of min- plume, and entrainment of melts in flowing lherzolite of the astheno- erals through magma contamination. This slow process may sphere.Magma in the boundary layer of the lithosphereapproaches cause the more or less isothermal growth of large minerals the solidus,level 3, with consequencesas illustrated in Figure 44. resembling the discrete nodules in some kimberlites. Those magmas managing to insinuate their way near to level 3, the of which involved the failure of geotherm and solidus to over- solidus,will evolve H20-rich vapors. CO2-rich vapors cannot lap at depth [Wyllie, 1980, p. 6905]: "A third possibility is exist in this part of the mantle [Egglet, 1978; Wyllie, 1978c]. that the componentsC-H-O may occur in the mantle perido- The vapors may enhance crack propagation, permitting rapid tite without the occurrenceof melting if reduced oxygen fu- uprise of the kimberlite magma. Many intrusions from this gacity raisesthe solidusfor the systemperidotite-C-H-O, es- level will solidify through thermal death before rising far peciallyif the subcontinentalgeotherm is lower than that plot- [Spera, 1984], but others will enter the crust as kimberlite ted. Partial melting would then occur only where temper- intrusions [Artyushkov and Sobolev,1984]. atures were raised or oxygenfugacity was increased."Green et Figure 44 illustrates a series of events as they might have al. [1987], using the data of Taylor and Green [1987], have occurred at successivetimes. It does not represent a set of developedthis idea in elegantdetail, referringto the processas simultaneousevents. The depleted, refractory base of the litho- redox melting. The partial melting of peridotite in the pres- sphere, 150-200 km deep, has probably been invaded intermit- enceof risingCH,• + H 2 is delayeduntil a more oxidizedlevel tently by small bodies and dikes of kimberlite or magmas of in the lithosphere is reached where CO 2 forms and fluxes related composition through billions of years. Most of these melting; with the reducedgases the solidusis at temperatures aborted and gave off vapors. The combination of failed kim- above the geotherm, but with oxidation it is lowered to a betlite intrusions and their vapors has contributed to the gen- temperaturebelow the geotherm.It is obviousfrom Figure 20 eration of a heterogeneous layer in this depth interval, that with the cool cratonic geotherma small upward displace- characterized by many local volumes metasomatized in di- ment of the solidus (Figure 21) would be sufficientto permit verse ways at various times. The aqueous vapors may perco- the passageof reducedvolatile componentsthrough the depth late upward through many kilometers to shallower levels,per- interval (2-3) without partial melting,but note that the solidus haps generating the sequenceof metasomatizedgarnet perido- displacementrequired to accomplish this with the inflected tites described in detail by Erlank et al. [1987]. Schneider and geothermis considerablywider. Eggler [1986] suggestedon the basis of their experimental I have modified the 1980 diapiric model for one in which data that aqueous vapors and solutions rising through the the mantle melt percolates into the base of the lithosphere lithosphere would leach the mantle with minor precipitation where it stalls [Wyllie, 1987a]. Once the relationshipsamong until they reached a level of about 70 km where a "region of oxygenfugacity, water activity, experimentallydetermined sol- precipitation" is associatedwith the formation of amphibole idus curves, and actual redox conditions in the mantle have and consequentchange in vapor compositiontoward CO 2. been established,the appropriate solidus boundaries can be If a thermal plume rises below the cratonic lithosphere used with isotherms to produce petrological structures.In the (Figure 10), the petrological structure of Figure 44 becomes presentstate of uncertainty,however, let us examine the dia- modified in the way illustrated in Figure 45. The depth of gramsthat can be constructedfor the limiting situation,where intersection of the geotherm with a dissociation reaction is the solidus for peridotite-C-H-O is assumed to remain near displaced to greater depths, releasing vapor [Woerrnann and that for peridotite-CO2-H20 (Figure 20). Subsequentchange Rosenbauer, 1985]. Interstitial melt is generated at level 2 in the temperature of either solidus or geotherm will involve where the lherzolite is transported across the solidus curve. movement of the phase boundaries in the petrological struc- With continued convection, the geotherm rises to a higher, tures. inflected position as depicted in Figure 20, and the depth 2 at Figure 44 representsa stable craton with uninflectedgeo- which melting begins is increased considerably. The intersec- therm, correspondingto Figure 8. Phase boundariesat the tion points of geotherm with solidus and dissociation bound- specifieddepths are drawn at intersectionsof the uninflected ary could overlap, causing the hydrated solid to melt directly geothermwith the phaseboundaries in Figure 20. Intersection without the opportunity for a vapor phase to exist through a of the geothermwith the solidusat levels2 and 3 impliessmall discrete depth interval (see Figure 45). As the plume diverges 400 WYLLIE' MAGMA GENESIS,PLATE TECTONICS, DIFFERENTIATION

RIFT1•11 i below the lithosphere,the melt becomesconcentrated in layers o or chambersin the boundary layer above the plume, as shown Nephel•n•te K •m berl ite in Figure 45. Ribe and $mooke [1987] concluded from their Carbonat•te CRUST - ...:...... MOHO - analysis of a dynamic model for melt extraction from a mantle • .:• •'Metasomat•sm plume that layers of segregatedmelt are likely to form at the base of the lithosphere above a plume. Lateral divergenceof the asthenospheretransports some of the entrained plume ..... i.l.[ Kmber te -•vleTasomaTisml,lLk ...'. , m m m ••- melt, and this penetratesthe lithosphere,forming small dikes .... [•' 2• : '• : .. or magma chambers. Kimberlites may rise either from the early magma accumulating above the plume or from the later- 200 al magma chambers in the lithosphere base. If the plate is moving across the plume, modifications are required as sug- ...... - gested by Figure 9. x, • • / fi/ Solidus 300 Rift Valleys, Nephelinites, and Carbonatites There is evidence that beneath major rifts the lithosphere is .. •)•'1• 1Magma•Vapor thinned. The continued heat flux from a rising plume and the Fig. 46. An upper mantle section correspondingto conditions in concentration of hotter magma at the asthenosphere- Figures 10 and 45, followed by lithospheric thinning in Figure 11. Thinning of the lithosphere above the plume permits upward migra- lithosphere boundary (see Figure 45) will promote further tion of the magma to level 4, where magma chambers develop, and thinning of the lithosphere. Gliko et al. [1985] discussedthe vapor is evolved with strong metasomatic effects. This level is the rates of thinning. Figure 46 depicts the petrological structure source of parent magmas for nephelinitic volcanism, and the associ- in a craton with rifting crust. ated carbonatites. The magma near the boundary layer at levels 1-3 in Figure 45 will rise with the layer as the lithosphere thins, either per- colating through the newly deformable rock matrix or rising Anorogenic Granites as a series of diapirs with the amount of liquid increasing in Granitoid batholiths are characteristic rocks of subduction amount as the diapirs extend further above the solidus for zones and compressiveplate boundaries,but there are many peridotite-C-H-O. This magma approaches the solidus ledge occurrencesof granite intrusions within stable plates. The gen- at level 4 in the depth range 90-70 km (Figure 20). Magma eral depth-temperature relationships for these anorogenic chambers may be formed as the magma solidifies (Figure 46), granites are shown in Figures 31 and 32. Pore fluids in the and vapors will be evolved causing metasomatism in the deep continental crust are likely to be trivial in amount except overlying mantle and facilitating intermittent crack propaga- in special environments such as in Figures 5b and 5c. Some tion, which permits the escape of magmas through the litho- anorogenic granites may be produced by differentiation of sphere. The metasomatic vapors will be aqueous if released at basalt, but partial melting of crustal rocks with hydrous min- depth greater than Q and CO2-rich if releasedat depths shal- erals such as muscovite, biotite, and amphibole under vapor- lower than Q (Figure 20; Wyllie [1980, Figure 8]), which is absent conditions is probably more likely for the generation of consistent with Haqqerty's [1988] account of metasomes at granitic magmasin stablecontinental crust (K in Figure 32). these depths. Magmas passing from level 3 to level 4 will also Figure 27 showsthe results of Rutter and Wyllie [1988a] on have left metasomatic traces beneath former rift zones. tonalite gneissat 10 kbar, representingpartial melting of the A variety of alkalic magmas may be generated at level 4, deep crust. The biotite-melting reaction produced 20% depending sensitivity upon conditions. There are large H20-undersaturated melt by 900øC, and the hornblende- changesin the compositions of melts and vapors in this region melting reaction produced 35% granodioritic melt with 2.3% for small changes in pressure and temperature I-Wyllie, 1978c; H20 at 1000øC.The partly molten rock can be mobilized only Wendlandt and Eqqler, 1980; Wendlandt, 1984]. Magmas when the critical melt fraction is reached. According to recent rising from this level may include the parents of olivine neph- estimates, the critical melt fraction for granitic melts is 30- elinites, melilite-bearing lavas, and other igneous associations 50%. According to the results in Figure 27, temperatures be- differentiating at shallower depths to carbonatites. The ig- tween 950øC and 1000øC are required for segregation of a neous sequenceassociated with the thinning of the lithosphere granodioritic melt from a tonalite gneisssource at the base of between the conditions in Figures 45 and 46 was discussedby the crust through dehydration melting. Wendlandt and Morgan [1982]. For the formation of carbon- There is a problem of heat supply. The geotherm for stable atites in this environment, see Le Bas [1977] and Wyllie continents (e.g., b or f in Figure 7) is too low for the conti- [1987a, 1988b]. nental crust to be partly melted to provide granitic magmas under normal circumstances.In a detailed review on the origin Continental Flood Basalts of silicic magma chambers and associatedigneous rocks, Hil- Not .all magmas are affected by the solidus ledge at level 4 dreth [1981] presented strong evidence to support his thesis in Figures 20 and 21. For those magmas that have already that granitoid magmas were the end product of a complex escaped from equilibrium with a peridotite host, the solidus series of events initiated by flux of basalt into the crust and ledge has no significance.For those diapirs or melts of higher followed by a long history of partial melting of the crust and temperature rising from greater depths, adiabatic paths would slow migration of magmas, with many opportunities for con- miss the solidus at M in Figure 20, and these might reach the tamination of the magmas and for mingling of the various solidus for volatile-free peridotite, at shallower levels. The materials. There are now many examples of the intrusion of magmatic product is basalt, generated from continental litho- basic magma into a silicic magma chamber. Figure 32 indi- sphere as required by Alldqre [1987]. cates that temperatures of basalts from mantle are high WYLLIE: MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION 401

enough to melt stable continental crust. The topics of con- and addition of magma on the one hand and the passageof tamination and mixing of basaltic and granitic magmas were vapors or solutions on the other, is quantitative, involving discussedbriefly in connection with Figure 37. The intrusion mass of material added to the rock. The final stage of meta- of basaltinto continentalcrust appearsto be a prerequisitefor somatism by magmas is solidification of the magma with the anorogenic granites. resultant introduction of new material into the rock. Schneider and Eggler [1986], following their experimental studies on the METASOMATISM solubility of peridotite componentsin H20 and H20-CO 2 The study of mantle xenoliths has revealed two kinds of fluids at high pressures,calculated that high fluid/rock ratios metasomatism. Infiltration (or patent or modal) metasomatism would be required to effect signifciant changes in the chemis- has introduced elements leading to the formation of new min- try of rocks with such fluids passingthrough them. They con- erals such as phlogopite, ilmenite, rutile, sulfides, richterite, cluded that the fluids were not effective metasomatic agents barium titanite, and carbonate, as well as changesin chemistry and that most mantle metasomatism was accomplished by of preexisting minerals. The metasomatizing fluids cause en- magmas. It is also established,however, that metasomatism of richmentin Fe, K, Ti, H20, S, Nb, Zr, Hf, Sr, Cu, Rb, Ba, Na, individual mantle sampleshas been accomplishedboth by the CO 2, and light rare earth elements,among others.Micas and intrusions of magmas and by the infiltration of solutions ema- amphiboles contain F, C1, and O as well as OH. Diffusion (or nating from the magmas during solidification of the melt cryptic) metasomatisminvolves the formation of no new min- [Wilshire, 1984]. In time we should be able to erect criteria to erals, but the addition of elements such as light REE, Ba, and make the distinction even when the source rock is not in hand. Sr to clinopyroxene.These observationsneed to be correlated CONCLUDING REMARKS with the inferences from inverse interpretations of the geochemistry of basalts. Two recent books are informative The petrological cross sectionsshown in Figures 35, 42, 43, [Nixon, 1987; Menzies and Hawkesworth, 1987]. 44, 45, and 46, derived by transferring phase boundaries onto The appeal to metasomatism to explain the geochemical assumedthermal structures,provide physical frameworks with idiosyncraciesof basalt geochemistry,supported and illumi- specificboundaries for consideration of magmatic and meta- nated by the observationsof metasomatismin hand specimens somatic events. The boundary positions will change in re- of mantle xenoliths, is certainly reinforcedby the proposalsfor sponseto the dynamics of mantle flow and the chemical differ- widespreadmigration of vapors and volatile-chargedmelts il- entiation caused by separation of melts and vapors from lustrated in several of the preceding figures. We assume that rocks. The fluid dynamics of rock-melt-vapor systems is of most of these vapors and melts can percolate through rocks paramount importance in deciding what will actually happen under some conditions, despite the warning signal from the to the phasesin the phase fields depicted in these petrological recent research of Watson and Brenan [1987] on the wetting cross sections [Marsh, 1987]. The physical frameworks may characteristicsof CO2-H20 fluids suggestingthat these fluids be adjusted to satisfy the geochemical constraints as the geo- may be incapable of movement except by hydrofracture. The chemical models are themselvesadjusted [Carlson, 1987] and chemical consequencesinferred and observedfor mantle meta- in order to satisfy the geophysical constraints. The frame- somatism were outlined above. There are very few experi- works appear to be robust enough to accommodate modifi- mental data to supplementthe geochemicalalgebra. cations, and it is precisely through such modifications arising Much effort has been devoted to determination of the com- from the interplay of geochemistry,geophysics, fluid dynamics, positions of liquids coexisting with mantle rocks at various thermodynamics, and experimental petrology that under- pressuresand with various volatile components,but much less standing of the petrological consequencesof volatile fluxes is known about the solubilities of the solid components in from the mantle may be advanced. vapor phases of various compositions under mantle con- It is mantle convection that controls the motion of the ditions. Wyllie and Sekine [1982] reviewed briefly data appli- plates,the geologyof the continents,and the eruption of vol- cable to conditions above subducted oceanic crust, and Schnei- canoes. As we take an overview of the North American conti- der and E•7•Tler[1984, 1986] reviewed existing data and pre- nent, we can seeourselves sitting in Vancouver surroundedby sented new data on the compositions of solutes coexisting mountains and volcanic activity. The divergent plate bound- with peridotite at high pressures.The solubility of components ary of the Juan de Fuca ridge off Vancouver and its more is sensitiveto the mineralogy of the host rocks as well as to southern extension rifting Baja California from the main con- depth and temperature. Metasomatic vapors rising from great- tinent are the sites of basaltic eruptions and submarine hot er to lower depths commonly follow paths of decreasing tem- springs. Subduction of the Pacific Plate beneath North perature, but in subduction zones they may rise from lower- America was responsiblefor Mount Garibaldi just up the temperature subducted oceanic slab into higher temperature road from this conference hall, for the Cascade volcanoes to mantle wedge. Metasomatic vapors either scavengeperidotite the south, and for the Sierra Nevada batholith. Mount St. for additional components or change its composition by ex- Helens,after its dramatic eruption of 1980, still puffs out water change or precipitation of components. Schneiderand E•l•ller and carbon dioxide laced with sulfur dioxide,the by-products [1986] demonstrated that addition of even small amounts of of digestingthe Pacific Ocean crust. CO 2 to a hydrous solution causes a significant decrease in My favoriteslide, the volcano Mount Fuji seen from Tokyo solubilities of peridotite components. Bay with the sun setting behind it, summarizes the available It is easy to appeal to metasomatism by fluids. It is impor- energy sources. First, the sun represents our direct external tant to attempt to distinguish between metasomatism by sourceof energy.Second, the lava representsan uprise of the vapors and by liquids. We can expect differencesin the nature Earth's internal energy,and it heats groundwaterproviding of the chemical changes caused by the passage of a vapor geothermalenergy. The magma brings to the surfaceincreased phase compared with the intrusion of a liquid, or magmatic concentrations of radioactive elements such as Th and U, phase. One distinction between these two processesof passage which may be redistributedby crustalwaters, becoming con- 402 WYLLIE: MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION

centrated into uranium/thorium ore deposits. The volcanic Ryabchikov, Pyroxene-carbonaterelations in the upper mantle, gasescontributed to the atmosphereand oceanswhen exposed Earth Planet. Sci. Lett., 62, 63-74, 1983. Brophy, J. G., and B. D. Marsh, On the origin of high-alumina arc to sunshine participate through the marvels of photosynthesis basalt and the mechanics of melt extraction, J. Petrol., 27, 763-789, in chemical reactions involving the formation of organic ma- 1986. terials from CO 2 and H20. These molecules may become Carlson, R. W., Geochemical evolution of the crust and mantle, Rev. trapped into fossil fuels, coal and petroleum, and geological Geophys.,25, 1011-1020, 1987. Carmichael, I. S. E., F. J. Turner, and J. Verhoogen, Igneous Pet- history dictates whether or not these materials are saved as rology, McGraw-Hill, New York, 1974. fossil sunshine. Carroll, M. R., and P. J. Wyllie, Phase equilibria and liquid lines of Magmas and volcanoes,the magmatic overflow valves, have descentin the systemtonalite-peridotite-H20 at 15 kb (abstract), clearly played a major role in Earth evolution and in the Eos Trans. AGU, 68, 451, 1987. distribution of energy. The Earth's fluid envelopes interact Chen, C.-Y., and F. A. Frey, Trace element and isotopic geochemistry of lavas from Haleakala volcano, East Maui, Hawaii: Implications with the magmatic rocks most vigorously at divergent plate for the origin of Hawaiian basalts,or. Geophys. Res., 90, 8743-8768, boundaries where the ocean transforms fresh basalt and in 1985. subduction zones where most of the trapped ocean is expelled Committee on Physicsand Chemistry of Earth Materials, Earth Ma- either as solutions or in magmas. The major problem of iden- terials Research, National Academy Press, Washington, D.C., 1987. tifying the geochemical reservoirs in the Earth and relating Courtney, R. C., and R. S. White, Anomalous heat flow and geoid these to mantle convection patterns JAiltigre, 1987] cannot be acrossthe Cape Verde Rise: Evidence for dynamic support from a solved without more information on the details and mass bal- thermal plume in the mantle, Geophys.or. R. Astron. Soc.,87, 815- ance of interactions between the fluid envelopes and crust and 867, 1986. mantle in these two plate tectonic environments. Creager, K. C., and T. H. Jordan, Slab penetration into the lower mantle, J. Geophys.Res., 89, 3031-3049, 1984. Davies, G. F., Earth's neodymium budget and structure and evolution Acknowledgments. This research was supported by the Earth Sci- of the mantle, Nature, 290, 208-213, 1981. ences section of the U.S. National Science Foundation, grants Davies, G. F., Geophysical and isotopic constraints on mantle con- EAR84-16583 an EAR85-06857. For preprints and comments,I thank vection: An interim synthesis, or. Geophys. Res., 89, 6017-6040, C. J. All•gre, D. L. Anderson, R. W. Clayton, B. H. Hager, and T. 1984. Tanimoto. California Institute of Technology Division of Geological Decker, R. W., T. L. Wright, and P. H. Stauffer, Volcanismin Hawaii and Planetary Sciencescontribution 4565. 2 vols., U.S. Geol. Surv. Prof. Pap., !350, 1987. The Editor thanks J. Longhi for his assistancein evaluating this Deines, P., The carbon isotopic composition of diamonds: Relation- paper. ship to diamond shape, color, occurrenceand vapor compositions, Geochim. Cosmochim. Acta, 44, 943-961, 1980. DePaolo, D. J., and G. J. Wasserburg, Inferences about mantle REFERENCES sourcesand mantle structurefrom variationsof x'•3Nd/X'•'•Nd,Geo- All•gre, C. J., Isotope geodynamics, Earth Planet. Sci. Lett., 86, 175- phys.Res. Lett., 3, 743-746, 1976. 203, 1987. DePaolo, D. J., and G. J. Wasserburg, Petrogenetic mixing models All•gre, C. J., O. Brevart, B. Dupre, and J.-F. Minster, Isotopic and and Nd-Sr isotopic patterns, Geochim. Cosmochim.Acta, 43, 615- chemicaleffects produced in a continuouslydifferentiating and con- 627, 1979. vecting Earth mantle, Philos. Trans. R. Soc. London, Ser. ,4, 297, Dziewonski, A.M., Mapping the lower mantle: Determination of 447-477, 1980. lateral heterogeneity in P velocity up to degree and order 6, or. All•gre, C. J., T. Staudacher, and P. Sarda, Rare gas systematics: Geophys.Res., 89, 5929-5952, 1984. Formation of the atmosphere, evolution and structure of the Egglet,D. H., The effectof CO 2 upon partial melting of peridotitein Earth's mantle, Earth Planet. Sci. Lett., 81, 127-150, 1987. the systemNa20-CaO-A1203-MgO-SiO2-CO 2 to 35 kb, with an Anderson, D. L., The chemical composition and evolution of the analysisof melting in a peridotite-H20-CO 2 system,Am. or.Sci., mantle: Advances in Earth and planetary sciences,in High-Pressure 278, 305-343, 1978. Researchin Geophysics,edited by S. Akimoto and M. H. Mangh- Egglet, D. H., Discussionof recent papers on carbonated peridotite, nani, pp. 301-318, D. Reidel, Hingham, Mass., 1982. bearing on mantle metasomatism and magmatism: An alternative, Anderson, D. L., Chemical composition of the mantle, d. Geophys. Earth Planet. Sci. Lett., 82, 398-400, 1987. Res., 88, B41-B52, 1983. Eggler, D. H., and D. R. Baker, Reduced volatiles in the system Anderson, D. L., The Earth as a planet: Paradigms and paradoxes, C-O-H: Implications to mantle melting, fluid formation, and dia- Science, 223, 347-355, 1984. mond genesis,in High PressureResearch in Geophysics,Adv. Earth Anderson, D. L., Global mapping of the upper mantle by surface Sci. 12, edited by S. Akimoto and M. H. Manghnani, pp. 237-250, wave tomography, in The Composition,Structure and Dynamics of D. Reidel, Hingham, Mass., 1982. the Lithosphere-,4sthenosphereSystem Geodyn. Ser., vol. 16, edited Ellis, D., and P. J. Wyllie, Hydration and melting reactions in the by K. Fuchs and C. Froidevaux, pp. 89-97, AGU, Washington, systemMgO-SiO2-H20 at pressuresup to 100 kilobars,Am. Min- D.C., 1987. eral., 64, 41-48, 1979. Anderson, D. L., and J. D. Bass, Transition region of the Earth's Erlank, A. J., F. G. Waters, C. J. Hawkesworth, S. E. Haggerty, H. L. upper mantle, Nature, 320, 321-328, 1986. Allsop, R. S. Rickard, and M. Menzies, Evidence for mantle meta- Anderson, R. N., S. E. De Long, and W. M. Schwartz, Thermal model somatism in peridotite nodules from the Kimberley pipes, South for subduction with dehydration in the downgoing slab, d. Geol., Africa, in Mantle Metasomatism, edited by M. Menzies and C. J. 86, 731-739, 1978. Hawkesworth, pp. 221-311, Academic,San Diego, Calif., 1987. Anderson, R. N., S. E. De Long, and W. M. Schwartz, Dehydration, Falloon, T. J., and D. H. Green, Anhydrous partial melting of MORB asthenosphericconvection and seismicity in subduction zones, or. pyrolite and other peridotite compositions at 10 kbar: Implications Geol., 88, 445-451, 1980. for the origin of primitive MORB glasses,Mineral. Petrol., 37, 181- Armstrong, R. L., A model for Sr and Pb isotope evolution in a 219, 1987. dynamic Earth, Rev. Geophys.,6, 175-199, 1968. Frey, F. A, and M. F. Roden, The mantle source for the Hawaiian Artyushkov, I. V., and S. V. Sobolev, Physicsof the kimberlite mag- Islands: Constraints from the lavas and ultramafic inclusions, in matism, in Kimberlites I: Kimberlites and Related Rocks, edited by Mantle Metasomatism, edited by M. Menzies and C. J. Hawkes- J. Kornprobst, pp. 309-322, Elsevier,New York, 1984. worth, pp. 423-463, Academic,San Diego, Calif., 1987. Bowen, N. L., The Evolution of the Igneous Rocks, Princeton Univer- Gill, J., Orogenic Andesitesand Plate Tectonics, Springer-Verlag, New sity Press, Princeton, N.J., 1928. (Reprinted by Dover, Mineola, York, 1981. N.Y., 1954.) Gliko, A. L., A. F. Grachev, and V. S. Magnitsky, Thermal model for Boyd, F. R., and J. J. Gurney, Diamonds and the African lithosphere, lithosphericthinning and associateduplift in the neotectonicphase Science, 232, 472-477, 1986. of intraplate orogenicactivity and continentalrifts, or.Geodyn. Res., Brey, G_, W. R. Brice, D. J. Ellis, D. H. Green, K. L. Harris, and I.D. 3, 137-138, 1985. WYLLIE.'MAGMA GENESIS,PLATE TECTONICS, DIFFERENTIATION 403

Green, D. H., Experimental testing of "equilibrium" partial melting of Maaloe, S., and T. S. Petersen, Petrogenesisof oceanic andesites,d. peridotite under water-saturated, high-pressure conditions, Can. Geophys.Res., 86, 10,273-10,286, 1981. Mineral., 14, 255-268, 1976. Maaloe, S., and A. Scheie, The permeability controlled accumulation Green, D. H., T. J. Falloon, and W. R. Taylor, Mantle-derived of primary magma, Contrib. Mineral. Petrol., 81, 350-357, 1982. magmas--Roles of variable source peridotite and variable C-H-O Marsh, B. D., Island-arc volcanism, Am. J. $ci., 67, 161-172, 1979. fluid compositions, in Maximatic Processes: Physicochemical Prin- Marsh, B. D., Magmatic processes,Rev. Geophys., 25, 1043-1053, ciples, Geochem. Soc. Spec. Publ. 1, edited by B. O. Mysen, pp. 1987. 139-154, Pergamon, New York, 1987. McCallum, I. S., Petrology of the igneous rocks, Rev. Geophys.,25, Griggs, D. T., The sinking lithosphere and the focal mechanism of 1021-1042, 1987. deep earthquakes, in The Nature of the Solid Earth, edited by E. C. McKenzie, D. P., The generation and compaction of partial melts, J. Robertson, pp. 361-384, McGraw-Hill, New York, 1972. Petrol., 25, 713-765, 1984. Grove, T. L., and R. J. Kinzler, Petrogenesisof andesites,Annu. Rev. Menzies, M., and C. J. Hawkesworth (Ed.), Mantle Metasomatism, Earth Planet. Sci., 14, 417-454, 1986. Academic, San Diego, Calif., 1987. Hager, B. H., Subducted slabs and the geoid: Constraints on mantle Mysen, B. O., and I. Kushiro, Compositional variations of coexisting rheology and flow, J. Geophys.Res., 89, 6003-6016, 1984. phaseswith degree of melting of peridotite in the upper mantle, Am. Hager, B. H., and R. W. Clayton, Constraints on the structure of Mineral., 62, 843-856, 1977. mantle convection using seismic observations, flow models, and the Navon, O., and E. Stolper, Geochemical consequencesof melt perco- geoid, in Mantle Convection,edited by W. R. Peltier, Gordon and lation: The upper mantle as a chromatographic column, J. Geol., Breach, New York, in press, 1987. 95, 285-397, 1987. Hager, B. H., R. W. Clayton, M. A. Richards, R. P. Cramer, and A. Nehru, C. E., and P. J. Wyllie, Compositions of glassesfrom St. Paul's M. Dziewonski, Lower mantle heterogeneity, dynamic topography peridotite partially melted at 20 kilobars, d. Geol., 83, 455-471, and the geoid, Nature, 313, 541-545, 1985. 1975. Haggerty, S. E., Diamond genesisin a multiply constrained model, Nicholls, I. A., and A. E. Ringwood, Effect of water on olivine stabili- Nature, 320, 34-38, 1986. ty in tholeiites and production of silica-saturatedmagmas in the Haggerty, S. E., Source regions for oxides, sulfides and metals in the island arc environment, d. Geol., 81, 285-300, 1973. upper mantle: Clues to the stability of diamonds, and the genesisof Nickel, K. G., and D. H. Green, Empirical geothermobarometry for kimberlites,lamproites and carbonatites,in Proceedin,Isof the 45th garnet peridotites and implications for the nature of the litho- International Kimberlite Conference,Australian Geological Society, sphere,kimberlites and diamonds, Earth Planet. Sci. Lett., 73, 158- in press,1988. 170, 1985. Hammarstrom, J. M., and E. Zen, Aluminum in hornblende: An em- Nisbet, E.G., and D. Walker, Komatiites and the structure of the pirical igneous geobarometer, Am. Mineral., 71, 1297-1313, 1986. Archean mantle, Earth Planet. Sci. Lett., 60, 105-113, 1982. Heinz, D. L., and R. Jeanloz, Measurement of the melting curve of Nixon, P. H. (Ed.), Mantle Xenoliths, John Wiley, New York, 1987. Mg0.9Fe0.•SiO3 at lower mantle conditionsand its geophysicalim- Ohtani, E., The primordial terrestrial magma ocean and its impli- plications, J. Geophys.Res., 92, 11,437-11,444, 1987. cation for stratification of the mantle, Phys. Earth Planet. Inter., 38, Hildreth, W., Gradients in silicic magma chambers: Implications for 70-80, 1985. lithospheric magmatism, J. Geophys.Res., 86, 10,153-10,192, 1981. Olafsson, M., and D. H. Eggler, Phase relations of amphibole- Hofmann, A. W., and W. M. White, Mantle plumes from ancient carbonate, and phlogopite-carbonate peridotite: Petrologic con- oceanic crust, Earth Planet. Sci. Lett., 57, 421-436, 1982. straints on the asthenosphere,Earth Planet. Sci. Lett., 64, 305-315, Hollister, L. S., G. C. Grissom, E. K. Peters, H. H. Stowell, and V. B. 1983. Gisson, Confirmation of the empirical correlation of A1 in horn- Presnall, D.C., and C. E. Helsley, Diapirism of depleted peridotite• blende with pressure of solidification of calc-alkaline plutons, Am. A model for the origin of hot spots, Phys. Earth Planet. Inter., 29, Mineral., 72, 231-239, 1987. 148-160, 1982. Huang, W. L., and P. J. Wyllie, Phase relationships of S-type granite Presnall, D.C., J. R. Dixon, T. H. O'Donell, and S. A. Dixon, Gener- with H20 to 35 kbar: Muscovitegranite from Harney Peak, South ation of mid-ocean ridge tholeiites, J. Petrol., 20, 3-35, 1979. Dakota, J. Geophys.Res., 86, 1015-1029, 1981. Ribe, N.M., and M.D. Smooke, A stagnation point flow model for Huang, W. L., and P. J. Wyllie, Phase relationships of gabbro- melt extraction from a mantle plume, d. Geophys.Res., 92, 6437- tonalite-granite-water at 15 kilobar with applications to differ- 6443, 1987. entiation and anatexis, Am. Mineral., 71, 301-316, 1986. Richter, F. M., and D. P. McKenzie, On some consequencesand Huang, W. L., P. J. Wyllie, and C. E. Nehru, Subsolidusand liquidus possible causes of layered mantle convection, J. Geophys.Res., 86, phase relationshipsin the systemCaO-SiO2-CO 2 to 30 kbar with 6133-6142, 1981. geological applications, Am. Mineral., 65, 285-301, 1980. Rigden, S. M., T. J. Ahrens, and E. M. Stolper, Densities of liquid Jacobsen,S. B., and G. J. Wasserburg, Transport models for crust and silicates at high pressures,Science, 226, 1071-1074, 1984. mantle evolution, Tectonophysics,75, 163-179, 1981. Ringwood, A. E., Compositionand Petrolo•ty of the Earth's Mantle, Jaques, A. L., and D. H. Green, Anhydrous melting of peridotite at McGraw-Hill, New York, 1975. 0-15 kb pressureand the genesisof tholeiitic basalts, Contrib. Min- Ringwood, A. E., Phase transformation and differentiation in subduc- eral. Petrol., 73, 287-310, 1980. ted lithosphere: Implications for mantle dynamics, basalt pet- Jeanloz, R., and F. M. Richter, Convection, composition, and the rogenesis,and crustal evolution, J. Geol., 90, 611-643, 1982. thermal state of the lower mantle, J. Geophys. Res., 84, 5497-5504, Rutter, M. J., and P. J. Wyllie, Experimental evidence for the chemical 1979. effectsof interaction between hydrous silicic melts and hornblende- Johnston, A.D., Anhydrous P-T relations of near-primary high- gabbro at upper crustal pressures(abstract), Eos Trans. AGU, 67, alumina-basalt from the South Sandwich Islands: Implications for 1279, 1986. the origin of island arcs and tonalite-trondhjemite series rocks, Rutter, M. J., and P. J. Wyllie, Melting of vapour-absent tonalite at Contrib. Mineral. Petrol., 92, 368-382, 1986. 10 kbar to simulate dehydration-melting in the deep crust, Nature, Johnston, A.D., and P. J. Wyllie, Interaction of slab-derived melts 331, 159-160, 1988a. with mantle peridotite: Experimental constraints at 30 kbar with Rutter, M. J., and P. J. Wyllie, Experimental calibration of horn- H20 (abstract),Eos Trans.AGU, 67, 405, 1986a. blende as a proposed empirical geobarometer, Eos. Trans. A GU, Johnston, A.D., and P. J. Wyllie, Simulated H20-undersaturated 69, 86-87, 1988b. two-layer magma chambers: Rhyolite above basalt, 10-30 kbar (ab- Ryabchikov, I.D., D. H. Green, W. J. Wall, and G. P. Brey, The stract), in Abstracts With Pro,trams 1986 Cordilleran Section,p. 122, oxidation state of carbon in the reduced-velocity zone, Geochem. Geological Society of America, Boulder, Colo., 1986b. lnt. 1981, 148-158, 1981. Johnston, A.D., and P. J. Wyllie, Interaction of granite and basic Schneider,M. E., and D. H. Eggler, Compositionsof fluids in equilib- magmas: Experimental observationson contamination processesat rium with peridotite: Implications for alkaline magmatism- 10 kbar with H20 , Contrib.Mineral. Petrol.,98, 352-362, 1988. metasomatism, in Kimberlites I: Kimberlites and Related Rocks, Kushiro, I., Y. Syono, and S. Akimoto, Melting of a peridotite nodule edited by J. Kornprobst, pp. 383-394, Elsevier, New York, 1984. at high pressuresand high water pressures,J. Geophys.Res., 73, Schneider,M. E., and D. H. Eggler, Fluids in equilibrium with perido- 6023-6029, 1968. tite minerals: Implications for mantle metasomatism, Geochim. Cos- Maaloe, S., Magma accumulation in the ascending mantle, J. Geol. mochim. Acta, 50, 711-724, 1986. Soc. London, 138, 223-236, 1981. Schreyer, W., H.-J. Massonne, and C. Chopin, Continental crust sub- 404 WYLLIE: MAGMA GENESIS,PLATE TECTONICS,DIFFERENTIATION

ducted to depths near 100 km: Implications for magma and fluid Wendlandt, R. F., and P. Morgan, Lithospheric thinning associated genesisin collision zones, in Magmatic Processes:Physicochernical with rifting in East Africa, Nature, 298, 734-736, 1982. Principles, Geochern.Soc. Spec. Publ. 1, edited by B. O. Mysen, pp. Wickham, S. M., and H. P. Taylor, Jr., Stable isotope constraints on 155-163, Pergamon, New York, 1987. the origin and depth of penetration of hydrothermal fluids associ- Scott, D. R., and D. J. Stevenson, Magma ascent by porous flow, J. ated with Hercynian regional metamorphism and crustal anatexis Geophys.Res., 91, 9283-9296, 1986. in the Pyrenees,Contrib. Mineral. Petrol., 95, 255-268, 1987. Sekine,T., and P. J. Wyllie, The systemgranite-peridotite-H20 at 30 Williams, Q., R. Jeanloz, J. Bass, B. Svendsen, and T. J. Ahrens, The kbar, with applications to hybridization in subduction zone mag- melting curve of iron to 250 gigapascals:A constraint on the tem- matism, Contrib. Mineral. Petrol., 81, 190-202, 1982. perature at Earth's center, Science,236, 181-182, 1987. Sekine, T., and P. J. Wyllie, Experimental simulation of mantle hybri- Wilshire, H. G., Mantle metasomatism: The REE story, Geology, 12, dization in subduction zones, J. Geol., 91, 511-528, 1983. 395-398, 1984. Sekine, T., P. J. Wyllie, and D. R. Baker, Phase relationships at 30 Woermann, E., and M. Rosenhauer,Fluid phasesand the redox state kbar for quartz eclogite composition in CaO-MgO- of the Earth's mantle: Extrapolations based on experimental, A1203-SiO2-H20 with implicationsfor subductionzone magmas, phase-theoreticaland petrological data, Fortschr. Mineral., 63, 263- Am. Mineral., 66, 938-950, 1981. 349, 1985. Sparks, R. S. J., and L. A. Marshall, Thermal and mechanical con- Woodhouse, J. H., and A.M. Dziewonski, Mapping the upper straints on mixing between mafic and silicic magmas, J. Volcanol. mantle: Three-dimensional modeling of Earth structure by inver- Geotherrn. Res., 29, 99-124, 1986. sion of seismicwave forms, J. Geophys.Res., 89, 5953-5986, 1984. Spera, F. J., Carbon dioxide in petrogenesisIII: Role of volatiles in Wright, T. L., and R. T. Helz, Recent advancesin Hawaiian petrology the ascent of alkaline magma with special reference to xenolith- and geochemistry,Volcanism in Hawaii, vol. 1., edited by R. W. bearing mafic lavas, Contrib. Mineral. Petrol., 88, 217-232, 1984. Decker, T. L. Wright, and P. H. Stauffer, U.S. Geol. Surv. Prof. Stern, C. R., and P. J. Wyllie, Phase compositions through crys- Pap., 1350, 625-640, 1987. tallization intervals in basalt-andesite-H20 at 30 kbar with impli- Wyllie, P. J., The DynamicEarth, John Wiley, New York, 1971. cations for subduction zone magmas, Am. Mineral., 63, 641-663, Wyllie, P. J., Crustal anatexis: An experimental review, Tec- 1978. tonophysics,13, 41-71, 1977. Stern, C. R., and P. J. Wyllie, Phase relationships of I-type granite Wyllie, P. J., Water and magma generation at subduction zones, Bull. with H20 to 35 kbar: The Dinkey Lakes biotite granite from the Volcanol., 41-4, 360-377, 1978a. Sierra Nevada batholith, J. Geophys.Res., 86, 10,412-10,422, 1981. Wyllie, P. J., Peridotite-CO2-H20 and the low-velocity zone, Bull. Stern, C. R., W. L. Huang, and P. J. Wyllie, Basalt- Volcanol., 41-4, 670--683, 1978b. andesite-rhyolite-H20:Crystallization intervals with excessH20 Wyllie, P. J., Mantle fluid compositions buffered in and H20-undersaturated liquidus surfacesto 35 kilobars with im- peridotite-CO2-H20 by carbonates,amphibole, and phlogopite,J. plications for magma genesis,Earth Planet. Sci. Lett., 28, 189-196, Geol., 86, 687-813, 1978c. 1975. Wyllie, P. J., Magmas and volatile components, Am. Mineral., 64, Stolper, E., and D. Walker, Melt density and the average composition 469-500, 1979. of basalt, Contrib. Mineral. Petrol., 74, 7-12, 1980. Wyllie, P. J., The origin of kimberlites, J. Geophys.Res., 85, 6902- Sullivan, W., Continents in Motion, McGraw-Hill, New York, 1974. 6910, 1980. Takahashi, E., Melting of dry peridotite KLB-1 up to 14 GPa: Impli- Wyllie, P. J., Plate tectonics and magma genesis,Geol. Rundsch., 70, cations on the origin of peridotitic upper mantle, J. Geophys.Res., 128-153, 1981. 91, 9367-9382, 1986. Wyllie, P. J., Constraints imposed by experimental petrology on pos- Takahashi, E., and I. Kushiro, Melting of dry peridotite at high pres- sible and impossible magma sources and products, Philos. Trans. suresand basalt magma genesis,Am. Mineral., 68, 859-879, 1983. R. Soc. London, Set. A, 310, 439-456, 1984. Tanimoto, T., Three dimensional reconstruction of Pangaea, Geophys. Wyllie, P. J., Transfer of subcratoniccarbon into kimberlites and rare Res. Lett., 14, 235-238, 1987. earth carbonatites, in Magmatic processes:Physicochernical Prin- Taylor, W. R., and D. H. Green, The petrogeneticrole of methane: ciples, Geochern.Soc. Spec. Publ. 1, edited by B. O. Mysen, pp. Effect on liquidus phase relations and the solubility mechanism of 107-119, Pergamon, New York, 1987a. reduced C-H volatiles, in Magmatic Processes: Physicochernical Wyllie, P. J., Discussion of recent papers on carbonated peridotite, Principles, Geochern.Soc. Spec. Publ. 1, edited by B. O. Mysen, pp. bearing on mantle metasomatism and magmatism, Earth Planet. 121-138, Pergamon, New York, 1987. Sci. Lett., 82, 391-397, 1987b. Toks6z, N.M., and A. T. Hsui, Numerical studies of back-arc convec- Wyllie, P. J., Response,Earth Planet. Sci. Lett., 82, 401-402, 1987c. tion and the formation of marginal basins, Tectonophysics,50, 177- Wyllie, P. J., Solidus curves, mantle plumes, and magma generation 196, 1978. beneath Hawaii, J. Geophys.Res., 93, 4171-4181, 1988a. Wasserburg, G. J., Isotopic abundances: Inferences on solar system Wyllie, P. J., Origin of carbonatites--Evidence from phase equilibri- and planetary evolution, Earth Planet. Sci. Lett., 86, 129-173, 1987. um studies, in Carbonatites--Genesis and Evolution, edited by K. Watson, E. B., Diffusion in magmas at depth in the Earth: The effects Bell, Allen and Unwin, London, in press,1988b. of pressureand dissolvedH20, Earth Planet.Sci. Lett., 52, 291-301, Wyllie, P. J., and M. J. Rutter, Experimental data on the solidus for 1981. peridotite-CO2, with applications to alkaline magmatism and Watson, E. B., Basalt contamination by continental crust: Some ex- mantle metasomatism (abstract), Eos Trans. AGU, 67, 390, 1986. periments and models, Contrib. Mineral. Petrol., 80, 73-87, 1982. Wyllie, P. J., and T. Sekine, The formation of mantle phlogopite in Watson, E. B., and J. B. Brenan, Fluids in the lithosphere, 1, Experi- subduction zone hybridization, Contrib. Mineral. Petrol., 79, 375- mentally determinedwetting characteristicsof CO2-H20 fluids and 380, 1982. their implications for fluid transport, host-rock physical properties, Wyllie, P. J., C. J. Donaldson, A. J. Irving, S. E. Kesson, R. B. Merrill, and fluid inclusion formation, Earth Planet. Sci. Lett., 85, 497-515, D.C. Presnall, E. M. Stolper, T. M. Usselman, and D. Walker, 1987. Experimental petrology of basalts and their source rocks, in Ba- Watson, E. B., and S. R. Jurewicz,Behavior of alkalis during diffusive saltic Volcanism on the Terrestrial Planets, Basaltic Volcanism interaction of granitic xenoliths with basaltic magma, J. Geol., 92, Study Project, pp. 492-630, Pergamon, New York, 1981. 121-131, 1984. Wyllie, P. J., W. L. Huang, J. Otto, and A. P. Byrnes, Carbonation of Wendlandt, R. F., An experimental and theoretical analysis of partial peridotites and decarbonation of siliceousdolomites representedin melting in the systemKAISiO,•-CaO-MgO-SiO2-CO 2 and appli- the systemCaO-MgO-SiO2-CO 2 to 30 kbar, Tectonophysics,100, cations to the genesisof potassicmagmas, carbonatites and kimber- 359-388, 1983. lites, in Kirnberlites I: Kirnberlites and Related Rocks, edited by J. Kornprobst, pp. 359-369, Elsevier,New York, 1984. P. J. Wyllie, Division of Geological and Planetary Sciences,Califor- Wendlandt, R. F., and D. H. Eggler, The origins of potassicmagmas, nia Institute of Technology, 170-25, Pasadena,CA 91125. 2, Stability of phlogopite in natural spinel lherzolite and in the systemKA1SiO4-MgO-SiO2-H20-CO 2 at high pressuresand high (Received December 9, 1987; temperatures, Am. J. Sci., 280, 421-458, 1980. acceptedMarch 17, 1988.)