0361-0128/01/3493/203-22 $6.00 203

100th Anniversary Special Paper:Secular Changes in Global Tectonic Processes and Their Influence on the

Temporal Distribution of Gold-Bearing Mineral Deposits

DAVID I. GROVES,

Centre for Global Metallogeny, School of Earth and Geographical Sciences, University of Western Australia, Crawley, Western Australia 6009, Australia

KENT C. CONDIE,Department of Earth and Environmental Science, New Mexico Institute of Mining and Technology, Soccorro, New Mexico 87801

RICHARD J. GOLDFARB,U.S. Geological Survey, Box 25046, Mail Stop 964, Denver Federal Center, Denver, Colorado 80225-0046

JONATHAN M.A. HRONSKY,WMC Resources Ltd., 191 Great Eastern Highway, Belmont, Western Australia 6104, Australia

AND RICHARD M. VIELREICHERCentre for Global Metallogeny, School of Earth and Geographical Sciences, University of Western Australia,

Crawley, Western Australia 6009, Australia

AbstractMineral deposit types commonly have a distinctive temporal distribution with peaks at specific periods of

Earth history. Deposits of less redox-sensitive metals, such as gold, show long-term temporal patterns thatrelate to first-order changes in an evolving Earth, as a result of progressively declining heat production andattendant changes in global tectonic processes. Despite abundant evidence for plate tectonics in the early Pre-cambrian, it is evident that plume events were more abundant in a hotter Earth.

Episodic growth of juvenile continental crust appears to have been related to short-lived (

Introduction

THE ADVENT of the theory of plate tectonics generated con-siderable interest in the relationship between mineral de-posits and their global tectonic setting (e.g., Sawkins, 1972,1990; Sillitoe, 1972; Mitchell and Garson, 1981; Brimhall,1987; Hutchinson, 1993; Titley, 1993; Kesler, 1997; Barley etal., 1998; Kerrich et al., 2000; Blundell, 2002). This has led tothe realization that different mineral deposit types are relatedto specific convergent, divergent, or anorogenic settings andthat the recognition of these settings is useful in exploration.At the same time, there has been a growing recognition thatthe various mineral deposit types have a heterogeneous tem-poral distribution (Fig. 1), with characteristic peaks in theirabundance at specific times in Earth history (e.g., Meyer,1981, 1988; Barley and Groves, 1992; Goldfarb et al., 2001a).

The uneven temporal distribution can be explained bychanges in the processes that combine to produce the mineraldeposits and/or the preservation potential of their deposi-tional environments. In turn, temporal changes in the process

of mineral deposit formation can be ascribed to (1) the evolu-tion of atmosphere-hydrosphere-biosphere systems (e.g.,Holland, 1984), (2) the widely accepted secular decrease inglobal heat flow (e.g., as indicated by the virtual restriction ofkomatiite-associated nickel-copper-sulfide deposits to theearly Precambrian; Lesher, 1989), and (3) long-term changesin tectonic processes (e.g., Windley, 1984; Meyer, 1988;Veizer et al., 1989; Barley and Groves, 1992). As discussedbelow, secular decrease in global heat flow directly affects ir-reversible changes in tectonic processes, and both affect thelong-term preservation potential of terranes of different ageand, hence, represent important first-order controls on thetemporal distribution of mineral deposit types. Whether ornot the temporal distribution of near-surface mineral de-posits, particularly sediment-hosted deposits, for which metaltransport and deposition are highly affected by redox condi-tions, is controlled by an evolving atmosphere-hydrosphere-biosphere system is not discussed here. There is considerablecurrent controversy on whether there was a sudden appear-ance of an oxygenated world (e.g., Lasaga and Ohmoto, 2002)

204 GROVES ET AL.

0361-0128/98/000/000-00 $6.00 204

Phanerozoic. Late Archean formation of buoyant subcontinental lithospheric mantle was particularly impor-tant in the anomalous preservation of some earlier formed deposit types located inboard of craton margins andin providing critical conditions for the formation of others. Development of negatively buoyant subcontinentallithospheric mantle can explain the lack of preservation of some deposit types that formed in the later Pro-terozoic. A single fundamental concept of coupled episodic crustal growth and preservation in the Archean andPaleoproterozoic, evolving to decoupled episodes of growth and preservation from the Mesoproterozoic on-ward, can thus explain the temporal distribution of a number of gold-bearing mineral deposit types.

Cyprus-type

Abitibi-type

Kuroko

Orogenic gold

Uranium inweathered profile

Kiruna-type

Olympic Dam-type

Ilmenite-anorthosite

Lead-zinc inclastic sediments

Lead-zinc in carbo-nates: MississippiValley-type

Copper in clasticsediments

SED

EXIO

CG

Paleoplacerand placer gold

Porphyry Cu

Porphyry Mo

Porp

hyry

VH

MS

Ore type Ore type3Ga 3Ga2Ga 2Ga1Ga 1GaP PA) B)

FIG. 1. Distribution through time of the number of preserved specific mineral deposits ascribed to (A) orogenic-conver-gent margin settings and (B) anorogenic or continental-basin settings. Peaks in abundance of anorogenic and continental-basin metal deposits appear to correspond to (1) possible breakup, or incipient breakup, of a Paleoproterozoic superconti-nent, (2) formation of Rodinia at about 1 Ga, and (3) formation and breakup of Gondwana and Pangea. Adapted from Barleyand Groves (1992). Peaks for deposit types in (A) are better defined and discussed in the text. Note that the temporal pat-tern for orogenic gold deposits, in particular (see also Fig. 8), have evolved as better dating techniques have become em-ployed. IOCG = iron-oxide copper-gold deposits, SEDEX = sedimentary-exhalative deposits, VHMS = volcanic-hosted mas-sive sulfide deposits.

or more gradual evolution to a present-day atmosphere (e.g.,Farquhar et al., 2000) and discussion of these issues is beyondthe scope of this paper.

This paper examines current evidence for the evolution ofthe continental crust, the supercontinent cycle, the nature ofmantle plumes and their influence on the supercontinentcycle, and the transition from Archean- to modern-style platetectonics in order to develop a model that explains secularvariations in metallogeny in terms of changes in tectonicprocesses and their preservational consequences. In particu-lar, these consequences are examined with respect to theglobal metallogeny of gold deposits. The paper deals withglobal secular patterns of mineral deposits in terms ofchanges in tectonic processes but does not deal in any detailwith the environmental consequences (e.g., Titley, 1993;Kesler, 1997) of those tectonic processes. The best possibletemporal patterns for >90 percent of the resources of eachdeposit type are presented within the constraints on compat-ibility of resource data and uncertainties of precise timing forsome deposits. The paper attempts to integrate data and con-cepts across several disciplines and focuses only on the mostrecent overviews. The interested reader is referred to thesereviews for exhaustive lists of references within the differentdisciplines. The review by Kerrich et al. (2000) is a particu-larly complete source of references that relate to the geody-namics of world-class gold deposits.

Juvenile Continental Crust and the Supercontinent CycleIn order to understand the temporal evolution of mineral

deposits that are preserved within continental crust, it is im-portant to understand the temporal evolution of that crustand its relationship to the configuration of continental massesthrough timethe supercontinent cycle. The evolution of theEarth is intrinsically linked to a gradual decay of heat pro-duction, lowering of mantle temperatures and viscosity, andthickening of the subcontinental lithosphere (e.g., Pollack,1986). Some authors consider this to have been more or lessa continuous evolution (e.g., Pollack, 1997), but the episodicdistribution of radiogenic isotopic ages, first recognized byGastil (1960), has led to models of more episodic growth ofjuvenile continental crust (Taylor and McLennan, 1985; Steinand Hoffmann, 1994; Condie, 1998, 2000). The distributionof U-Pb zircon ages, coupled with Nd and Hf isotope data,suggest that significant crustal growth commenced at or be-fore 3.0 Ga with two major peaks in juvenile crustal produc-tion rate, one at ca. 2.7 Ga and another at ca. 1.9 Ga (Fig. 2).The global dominance of greenstone belts and granitoids thatformed in the interval from ca. 2.75 to 2.60 Ga, in particular,cannot be a sampling artifact, as such terranes are knownfrom all continents (e.g., Yilgarn craton of Australia, southernSuperior and Slave provinces of Canada, Zimbabwe craton,Tanzania craton, and Sao Francisco craton of Brazil), and ro-bust geochronologic techniques have been employed increas-ingly in these terranes in the past decade. In addition, al-though it is evident that the precise timing of peaks in crustalgrowth is progressively evolving as more data become avail-able (compare Condie, 1998, with Condie, 2001a), with aprobable uncertainty of 50 Ma, the relative patterns are ro-bust. Thus, in addition to the two major peaks in crustal pro-duction rate, smaller peaks may be present at ca. 2.8, 2.5, 2.1,

1.7, 0.48, 0.28, and 0.1 Ga (Condie, 2001b). As discussedbelow, the pattern of these peaks and the peaks themselvescorrespond broadly with those of orogenic gold provinces andhence are an important control on global metallogeny. It isstill not clear if, and how, these peaks in the rate of crustalproduction relate to the supercontinent cycle, as defined, forexample, by Hoffman (1988) and Murphy and Nance (1992),and to mantle overturn events (e.g., Condie, 1998). They may,or may not, correlate with the accumulation or breakup phaseof supercontinents. In the last 1 b.y., the formation andbreakup of three such major supercontinents (Rodinia,Gondwana, and Pangea), and a possible short-lived supercon-tinent (Pannotia) in the latest Proterozoic, have been recog-nized (Unrug, 1997; Condie, 2002a). Geologic data supportthe existence of at least two earlier supercontinents, one atthe end of the Archean and one in the Paleoproterozoic(Hoffman, 1989; Rogers, 1996; Aspler and Chiarenzelli, 1998;Pesonen et al., 2003).

Ages from Archean cratons suggest that the first supercon-tinent (or supercontinents; Aspler and Chiarenzelli, 1998)formed during frequent collisions and suturing of older con-tinental blocks and juvenile oceanic terranes (principally arcsand oceanic plateaus) between 2750 and 2650 Ma (Fig. 2).The Late Archean peak in the rate of juvenile crust produc-tion is also centered at 2700 50 Ma (Fig. 2), further sug-gesting a correlation between supercontinent formation andjuvenile continental crust production. Major gold provincesdeveloped in the southern Superior, Slave, Yilgarn, Zim-babwe, Tanzania, and Sao Francisco cratons at about this time(as summarized by Goldfarb et al., 2001a).

The final breakup of the Late Archean supercontinent(s)occurred in the Paleoproterozoic between 2200 and 2100 Ma.Subsequently, a new supercontinent formed between 1900and 1800 Ma, with most collisions in Laurentia, Baltica, andSiberia occurring ca. 1850 Ma (Condie, 2002b; Pesonen et al.,2003; Fig. 2). The giant Homestake gold deposit formed ataround this time (Dahl et al., 1999).

100th ANNIVERSARY SPECIAL PAPER: TECTONIC PROCESSES & Au DISTRIBUTION 205

0361-0128/98/000/000-00 $6.00 205

14

2.7 Ga12

10

8

6

4

2Volu

me

perc

ent

grow

th

3.8 3.4 3.0 2.6 2.2 1.8 1.4 1.0 0.6 0.2Age (Ga)

1.9 Ga

FIG. 2. Frequency distribution of juvenile continental crust based on atotal volume of continental crust of 7.177 109 km3. Juvenile crust ages areU/Pb zircon ages used in conjunction with Nd isotope data and lithologic as-sociations. Modified after Condie (1998, 2000).

In the Phanerozoic, crustal growth peaks are indicated atca. 480, 280, and 100 Ma (Condie, 2001b). The first two peaksbroadly correlate with the rapid growth of Pangea between480 and 250 Ma and the third with the growth of a new su-percontinent beginning at ca. 150 Ma. Importantly, thesepeaks broadly coincide with major gold events in the Tasmanorogen, Central Asia, and circum-Pacific (Goldfarb et al.,2001a, b). Increased rates of production of juvenile crust ap-pear to correlate with formation and not breakup phases ofthe supercontinents over the last 1,200 m.y. (Condie, 2004).

Mantle-Plume Events and the Supercontinent CycleThe following section discusses the relationship between

crustal growth and mantle-plume events and how the latterimpact on the supercontinent cycle, which, in turn, impactson the temporal distribution of mineral deposits.

Major peaks in the rate of production of juvenile crust areinterpreted to be caused by mantle overturn events involvingnumerous mantle plumes in a short time interval (Condie,1998, 2000). Mantle plume events are short-lived episodes(100 m.y.) during which many plumes impact on or affectthe base of the lithosphere, and they appear to have been im-portant throughout Earth history (Condie, 1998; Isley andAbbott, 1999). During such events, plume activity may beconcentrated in one or more mantle upwellings, as during themid-Cretaceous (ca. 100 Ma) superplume event focused inthe Pacific basin (Larson, 1991).

Mantle plume events can be identified by the recognitionof igneous rocks associated with mantle plumes, commonlycalled plume proxies (Ernst and Buchan, 2002; Isley and Ab-bott, 2002). Using a combination of plume proxies, includingflood basalts, komatiites, and high MgO lavas, giant dikeswarms, and layered intrusions, Abbott and Isley (2002) rec-ognized 36 mantle-plume events in the last 3.8 b.y. (Fig. 3). Aweighted time series analysis shows that most mantle-plumeevents, regardless of age, lasted about 10 m.y., with majorPrecambrian events at ca. 2.75, 2.45, 1.8, 1.75, and 1.65 Ga,and several events in the Phanerozoic. Peaks in the abun-dance of plume proxies at ca. 2.75, 1.8, 0.25, and 0.12 Ga areclose to calculated peaks in the production of juvenile crust,although not all such events show such correlations (Condie,2004).

There may be two types of mantle-plume events; one asso-ciated with supercontinent formation and one with supercon-tinent breakup (Condie, 2004). It is commonly believed that

plumes associated with mantle upwellings are responsible forfragmenting supercontinents (Condie, 2001b). Computermodels suggest that it takes 200 to 400 m.y. for shielding of alarge supercontinent to cause a mantle upwelling beneath it(Lowman and Jarvis, 1996). This is followed by developmentof numerous mantle plumes within the upwelling and, finally,by supercontinent breakup over a period of up to approxi-mately 200 m.y. (Fig. 4). If, in a mantle-plume event (Condie,2004), the supercontinent is too small to provide sufficientmantle shielding to produce an upwelling, the small super-continent may move laterally with the downwelling (e.g., Tru-bitsyn et al., 2003) or may not completely fragment, as ap-pears to be the case with the 1.9-Ga supercontinent (Condie,2002c). Thus, there may be long-term survival of the roots ofrelatively small cratonic blocks such as the Yilgarn and Pilbaracratons of Western Australia (Trubitsyn et al., 2003). Al-though plume heads can have diameters of up to 2,500 km insuch a shielding mantle-plume event, the volume of juvenilemafic crust associated with a given plume (e.g., estimated bythe volume of Phanerozoic flood basalts and mafic under-plates) is probably relatively small, as is the volume of oceanicplateaus accreted to the continents, at least since the end ofthe Archean (Condie, 2001b).

There must be, therefore, something unique about themantle-plume events associated with peaks in the productionof juvenile crust at about 2.7 and 1.9 Ga. These events, calledcatastrophic mantle-plume events (Condie, 2004), must betriggered by some process other than plate shielding. Theyare also short lived (less than 100 m.y.) in contrast to shield-ing events (>200 m.y.) and must be more intense and perhapsmore widespread. The cause of the catastrophic mantleplumes is unclear. Peltier et al. (1997) and Condie (1998) sug-gested that the breakup of Precambrian supercontinents trig-gered slab avalanches at the 660-km discontinuity in the man-tle, resulting in catastrophic plume production in a (D)layer just above the core. However, this model does not ap-pear to work for the 2.7-Ga event, since there is no evidencefor significant earlier supercontinent fragmentation.

The most voluminous banded iron formations (BIF) weredeposited in intracratonic, passive-margin or platform basinsduring stands of high sea level in the Late Archean and Pale-oproterozoic (Simonson and Hassler, 1997). The iron and sil-ica enrichments in these rocks appear to have been derivedchiefly from hydrothermal vents on the deep sea floor (Kleinand Beukes, 1992; Isley, 1995; Barley et al., 1999). Two majorpeaks in BIF deposition at 2.7 and 1.9 Ga correlate well withand can be explained by mantle-plume events at these times(Klein and Beukes, 1992; Isley and Abbott, 1999). The en-hanced submarine volcanism and hydrothermal venting asso-ciated both with ocean-ridge and oceanic plateau volcanismduring a mantle-plume event may be the source of iron andsilica. Also, the elevated sea level at 1.9 Ga caused by a plumeevent could provide the extensive shallow marine basins alongstable continental platforms necessary to preserve BIF (cf. Ti-tley, 1993).

Crustal growth associated with catastrophic mantle-plumeevents is chiefly by addition of magmatic arcs directly withincontinental margins or by accretion of oceanic arcs to theedge of the continents (Rudnick, 1995; Condie, 2001b), al-though at least in the last 2.5 b.y., the latter played a relatively

206 GROVES ET AL.

0361-0128/98/000/000-00 $6.00 206

1.5

0.60.9

0.3

1.2

Hei

ght

4.0Age (Ga)

3.5 3.0 2.5 2.0 1.5 1.0 0.5

FIG. 3. Distribution of mantle plume events deduced from time seriesanalysis of plume proxies (i.e., igneous rocks associated with mantle plumes)from Abbott and Isley (2002). Peak height depends on the number of plumeproxies and the errors of the age, the latter set at 5 m.y.

minor role. That supercontinent formation occurred nearlysimultaneously with the 1.9 Ga plume event may not be coin-cidental. Perhaps breakup of the Late Archean superconti-nent at 2.2 to 2.1 Ga served as a trigger for the 1.9 Ga plumeevent, and, in this sense, a catastrophic event provided posi-tive feedback for crustal growth that began during a shieldingplume event. A growing supercontinent also may contributeto the preservation of juvenile crust by trapping it in colli-sional and accretionary orogens.

Broadly, the mineral deposits shown in Figure 1 have tem-poral distributions related, at least in part, to the superconti-nent cycle (e.g., Barley et al., 1998). As shown in Figure 1A,both Abitibi-type volcanic-hosted massive sulfide (VHMS)and orogenic gold deposits correlate with a possible cata-strophic mantle-plume event at 2.7 Ga, and ore types in Fig-ure 1B may correlate with shielding-type plume events.

The Transition from Archean to Modern-Style Plate Tectonics

If the Earth has steadily cooled with time, as is widely ac-cepted (e.g., Pollack, 1986), there is good reason to suspectthat tectonic processes were different in the Archean (e.g.,Fyfe, 1978). For example, a hotter mantle in the Archeanwould have produced more melt at ocean ridges and hencethicker oceanic crust (e.g., Bickle, 1990). Model calculationsindicate that the Archean oceanic crust should have beenabout 20 km thick, and perhaps even thicker in the EarlyArchean, compared to the present thickness of about 7 km(Sleep and Windley, 1982). It is also possible that there weremore numerous, but smaller, plates in the Archean (e.g., Pol-lack, 1997; deWit, 1998) and that there was a coupled reduc-tion in the number of plates and increase in the size of plateswith time, perhaps leading to a progression to larger super-continents with time.

Modern oceanic lithosphere reaches neutral buoyancy inabout 20 m.y. (Davies, 1992), after which it becomes nega-tively buoyant. The observed mean age of oceanic crustwhen it begins to subduct is 60 (Parsons, 1982) to 80 m.y.(Sprague and Pollack, 1980). In the presence of a hottermantle, it would take oceanic plates longer to become neu-trally buoyant, and thus, neutral buoyancy would not bereached until about 80 m.y. in the Late Archean (Fig. 5). In

addition, given a hotter mantle, it is probable that plateswould move faster and, thus in the Late Archean, spreadingcrust may have subducted sooner, possibly in about 35 m.y.and perhaps as rapidly as 20 m.y. (Bickle, 1986). Whereas thetime to reach neutral buoyancy in modern oceans is less thanthe time it takes for spreading crust to reach an active trench,oceanic crust in the Archean may have encountered activesubduction zones while it was still buoyant. The crossover inthese two ages depends on which values are accepted forsubduction rates but is likely to be between 2.5 and 2.0 Ga(Fig. 5), the time that Archean buoyant plate tectonics(deWit, 1998) may have started to evolve into modern platetectonics.

The Archean and probably Paleoproterozoic plate tectonicdynamics were likely controlled by various types of plumeevents, the largest of which were responsible for voluminouscontinental growth and accretion of buoyant oceanic lithos-phere. The voluminous komatiites and related basalts in the

100th ANNIVERSARY SPECIAL PAPER: TECTONIC PROCESSES & Au DISTRIBUTION 207

0361-0128/98/000/000-00 $6.00 207

3 2 1 0Age (Ga)

? R

R G P

P

N

Breakup

JuvenileCrust

Formation

Superplume EventsSuper continents

FIG. 4. Formation and breakup of supercontinents during the past 3.0 b.y, based on available paleogeographicreconstructions. Also shown are times of maximum production rate of juvenile continental crust and proposed catastrophicmantle-plume events. Data from Condie (1998, 2001b, 2002a-c). G = Gondwana, N = new supercontinent, P = Pangea, R =Rodinia.

4.010

20

30

40

50

60

70

80

3.5 3.0 2.5 2.0 1.5 1.0 0.5Age (Ga)

Tim

eto

reac

hneu

tral

buoy

ancy

or

for s

ubd

uctio

nto

begi

n(m

.yr.)

SubductionNeutral Buoyancy

FIG. 5. Effect of mantle cooling on the time needed for a plate to reachneutral buoyancy and the time needed for subduction to commence. Modi-fied after Davies (1992).

Archean and, to a lesser extent, Paleoproterozoic granitoid-greenstone terranes attest to the widespread influence ofplume tectonics (e.g., Campbell et al., 1989; Pirajno, 2000).There have been arguments against Archean plate tectonicsto produce the magmatic and deformational features ofgreenstone belts (e.g., Hamilton, 1998). However, thenonkomatiitic volcanic and volcano-sedimentary successionsin the greenstone belts are remarkably similar in their com-positional range and geochemistry to volcanic successions inmodern convergent settings, such as arcs, backarcs, and in-terarc rifts, with differences readily explained by contrastsbetween shallow subduction of hot, old, Archean oceaniclithosphere versus steeper subduction of colder, youngeroceanic lithosphere (e.g., Barley et al., 1998). This implicatesthe existence of Archean-Paleoproterozoic plume-influ-enced, modified plate tectonics. This is strongly supportedby the recognition of fossil subduction zones in the Abitibibelt of Canada during the LITHOPROBE project (e.g., Lud-den and Hynes, 2000). The similarity in structures andchronology of deformation events in the Precambrian green-stone belts and modern convergent margins and the abun-dance in both of orogenic gold deposits (Groves et al., 1998)and VHMS deposits (e.g., Franklin et al., 1981) is discussedfurther below.

Evolution of the Subcontinental Lithospheric MantleThe higher heat flow and greater plume activity in the early

Earth should also be reflected in the mantle lithospherebelow the continental crust. There is evidence for this in theunusual, broadly equidimensional (~1,0001,500-km diam)shapes of the Archean and Paleoproterozoic cratons and theanomalously high abundance of granitoids that make thesecratons readily identifiable in maps or remote sensing imagesat the regional scale. Although individual greenstone belts areelongate in a pattern similar to that of modern orogenic belts,the largely granitic cratons in which they are encased areequidimensional. Each craton (i.e., granitoid-greenstone ter-rane) has its own volcanic history and specific substrate, andits own distinctive metallogenic associations. For example,only the Yilgarn craton contains both giant gold and nickel de-posits such as those of the Eastern Goldfields province, andonly the Abitibi belt of the Superior craton contains bothworld-class gold and VHMS deposits, each hosted in ca. 2.72to 2.65 Ga greenstone belts. The different metallogeny ofthese cratons may partly reflect the smaller size of Archeanplates and certainly relates to the nature of the crust (andlithosphere) in the various provinces at the time of mineral-ization. For example, the VHMS deposits of the Abitibi beltwere mostly formed on primitive crust (Ayer et al., 2002),whereas the komatiite-hosted nickel deposits in the Yilgarncraton formed on rifted continental crust (Krapez et al.,2000), as also shown by the predominance of granitoids andtonalites in the Abitibi belt compared with monzogranites inthe Yilgarn craton (Champion and Sheraton, 1997). DeWitand Thiart (2003) further demonstrate that individual cratonsat a global scale have their own distinctive metallogenic asso-ciations regardless of the timing of mineralization (pre-, syn-,or postcratonization). The postcratonization example suggestsnot only that each craton is unique but also that it has uniquesubcontinental lithospheric mantle reflected by the nature of

the mineral deposits (e.g., PGE, diamonds, Fe oxide Cu-Au)that form after craton development. It is thus important froma metallogenic viewpoint to examine the evolution of the sub-continental lithospheric mantle (e.g., Richter, 1985; Jordan,1988) and its potential effect on the temporal distribution ofmineral deposits.

The density of the subcontinental lithospheric mantle be-neath stabilized continental crust of varying ages has been de-termined using data on composition, thermal state, and petro-logical thickness (e.g., OReilly and Griffin, 1996; PoudjomDjomani et al., 2001). Data from mantle-derived peridotitexenoliths and garnet xenocrysts provide a pattern of secularevolution of such lithosphere, with progressively less deple-tion in Al and Ca and lower Mg no. and Fe/Al from theArchean to the Phanerozoic. Thermobarometric data fromthe xenolith and xenocryst suites show that paleogeothermswere lower in Archean than in Phanerozoic subcontinentallithospheric mantle and that the typical thickness of thelithosphere, defined as a chemical boundary layer, was greaterin the Archean (250180 km) than in the Proterozoic(180150 km) and Phanerozoic (14060 km). From thesedata, and data on mineral end members that constitute thesubcontinental lithospheric mantle, Poudjom Djomani et al.(2001) calculate mean densities (at 20C) for Archean sub-continental lithospheric mantle of 3.31 0.016 Mg/m3 com-pared to Proterozoic and Phanerozoic equivalents of 3.35 0.02 and 3.36 0.02 Mg/m3, respectively. Thus, not only arethere significant variations in the compositions of subconti-nental lithospheric mantle with time (Fig. 6), but these trans-late, in an analogous way to oceanic lithosphere discussedabove, into temporal variations in density, and hence buoy-ancy, of the subcontinental lithospheric mantle (Fig. 7). Thissecular evolution of the subcontinental lithospheric mantleimplies broadly synchronous formation of crust and its lithos-pheric root and their linkage through their subsequent history(Griffin et al., 2003; Sleep, 2003). This elegantly explains theindividual metallogenic signatures of Precambrian cratons(deWit and Thiart, 2003), as each craton may have a lithos-pheric root of different depth extent and composition, de-pendent on degree of mantle melt extraction, mantle incom-patible-element metasomatism, and degree of modificationalong craton margins. Griffin et al. (2003) also argue that theunique Archean subcontinental lithospheric mantle roots rep-resent residues and/or cumulates from deep high-degreemelting related to the unique abundance of major mantle-plume events of varying type, as discussed above. The exactprocesses by which subcontinental lithospheric mantleformed and was coupled to the crust are beyond the scope ofthis paper, but various models are discussed by Arndt et al.(2002), and a schematic model of lithosphere accretion in-volving plume-arc interaction is presented in Kerrich et al.(2000). Whatever the process, massive melting events in themantle, and subsequently in the lower and middle crust toproduce many of the voluminous granitoids that typify theoverlying terranes (e.g., Champion and Sheraton, 1997),probably were the crustal-scale cause of the broadly equidi-mensional surface expressions of early Precambrian cratons.

There are major implications from these studies in terms oftectonics and metallogeny. As summarized by PoudjomDjomani et al. (2001), Archean subcontinental lithospheric

208 GROVES ET AL.

0361-0128/98/000/000-00 $6.00 208

mantle would have been buoyant relative to asthenosphere inany reasonable geologic scenario. As this buoyancy reducesstress and viscosity, Archean subcontinental lithospheric man-tle could not have been delaminated solely by gravitationalprocesses and would have been preserved unless disrupted byrifting and replacement by more fertile asthenosphere. Thus,mineral deposits formed late in Archean craton evolution, orin anorogenic settings after Archean cratonization, shouldhave enhanced preservation potential.

The calculations by Poujom Djomani et al. (2001) indicatethat Proterozoic subcontinental lithospheric mantle wasdenser than the Archean equivalent, but that typical 150- to

180-km-thick subcontinental lithospheric mantle was likelymoderately buoyant nonetheless, and thus also unlikely tohave been delaminated. As in Archean cratons, Paleoprotero-zoic mineral deposits that formed late in the cratonizationprocess should still have had relatively high preservation po-tential.

In contrast, the commonly 90 Ma

Siberia NE Siberia E ChinaSlave cratonCanada

DepletedHarzburgite

DepletedLherzoliteFertileLherzoliteLherzolite (LRE Emetasomatized)Peridotite (Melt-metasomatized)

LAB

LABLAB

Projec

tedD

epth

(km)

Cumulative % of each rock type0 0000

100 100100

50 5050

150 150150

200 200200

50 50505050 100 100100

LAB

FIG. 6. Depth and depth variation in composition of Archean, Proterozoic, and Phanerozoic lithospheric sections, in-cluding changes in the composition of subcontinental lithospheric mantle with time. Profiles for each age of lithosphere showestimated proportions of components (see legend) at various depths in the lithosphere. LAB = lithosphere-asthenosphereboundary. From Griffin et al. (2003).

Dep

th(km

)

50

100

150

200

2503.26 3.30 3.34 3.38 3.42

Upper MantleCrust

Archean(35mW/m )2

Proterozoic(40mW/m )2

LAB

LAB

A)

FIG. 7. Density profiles and average heat flow for (A) Archean and Proterozoic and (B) Phanerozoic subcontinental lithos-pheric mantle and for the asthenosphere. LAB = lithosphere-asthenosphere boundary. Adapted from Poudjom Djomani etal. (2001). The stars show where the isotherms for each lithosphere cross the LAB.

LAB

Dep

th(km

)

50

100

150

200

2503.26 3.30 3.34 3.38 3.42

Upper MantleCrust

SoutheasternAustralia

Phanerozoic(50mW/m )2

Cumulative density (g/cm )3

B)

older Phanerozoic and Neoproterozoic deposits are leastlikely to have been preserved.

The temporal evolution of subcontinental lithosphericmantle, which directly reflects the secular change fromstrongly plume-influenced Archean tectonics to Phanerozoic-style plate tectonics, must be considered a major factor influ-encing secular variations in both mineral deposit types andthe present abundance of specific deposit styles.

Tectonic Evolution and Secular Change in Metallogeny: Useful Markers

To determine links between secular tectonic and metallo-genic evolution, it is imperative to consider mineral depositstyles that are at least only mildly affected by other evolution-ary changes. Thus, it is important to avoid deposits of stronglyredox-sensitive metals (e.g., sediment-hosted deposits), thetemporal evolution of which may be linked to any temporalchange in the atmosphere-hydrosphere-biosphere system aswell as tectonic environment.

Epigenetic gold deposits are potentially useful for testingsecular changes in tectonic processes because most such de-posits formed below the influence of surficial processes and,hence, were unlikely to have been influenced by any secularvariation in atmosphere-hydrosphere-biosphere systems. Por-phyry Cu-Au deposits, some of which are significant gold de-posits in their own right (e.g., Kesler et al., 2002) and ep-ithermal Au-Ag Cutype deposits have a strong tectoniccontrol in convergent margin settings (e.g., Sillitoe, 1997),particularly in tectonic settings with anomalous, high K mag-matism (e.g., Barley et al., 2002). However, they form at shal-low crustal levels (

Late Archean and Paleoproterozoic peaks broadly reflect themajor periods of episodic growth of Precambrian continentalcrust, centered at ca. 2.7 and 1.9 Ga (Fig. 8B), althoughpeaks in the formation of orogenic gold deposits clearly flankthe latter peak in crustal growth. Similarly, the Phanerozoicorogenic gold deposits broadly mimic the more continuous,shorter wavelength distribution of Phanerozoic Cordilleran-type orogenic events and associated periods of crustalgrowth, although accurate correlations are limited by uncer-tainties, particularly in the ages of some gold provinces. De-spite these uncertainties, the secular evolution of orogenicgold deposits clearly reflects the proposed evolutionary trendfrom strongly episodic, plume-influenced plate tectonics inthe Archean to more cyclic modern-style plate tectonics.Archean and, to a lesser extent, Paleoproterozoic orogenicgold provinces related to the catastrophic mantle-plumeevents of Condie (2004) have equivalent gold production(Fig. 8A), despite their age, deep erosion, and common deepregolith cover. This, combined with evidence for relativelythin lithosphere at the time of gold deposit formation and

prior to cratonization for some of the larger Archean goldprovinces (e.g., Abitibi belt and Kalgoorlie terrane; Groves etal., 2003) and Paleoproterozoic gold provinces (e.g., Ashantibelt; Pigois et al., 2003), suggests that crustal-scale thermalinput may have played a critical role in the formation of giantorogenic gold provinces in the youngest and most juvenilecrust preserved.

Despite the strong Mesoproterozoic to Neoproterozoicrecord of crustal growth from 1.8 to 1.2 Ga (Fig. 8B), in partcorrelating with the formation of Rodinia at ca 1.3 to 1.0 Gathrough a number of continental collisions (e.g., Dalziel,1991; Hoffman, 1991), there is a lack of gold deposits be-tween about 1.7 Ga and 600 Ma (Fig. 8A). The Olympiadadeposit of southwestern Siberia (ca. 820 Ma; Safonov, 1997),as shown in Figure 8A, may be an important exceptionalthough it could be as young as 600 Ma (Konstantinov et al.,1999). This suggests that an additional factor, other than justthe growth of juvenile crust, was important in determiningthe temporal distribution of orogenic gold deposits preservedin the geologic record. Clues to the nature of this factor lie in

100th ANNIVERSARY SPECIAL PAPER: TECTONIC PROCESSES & Au DISTRIBUTION 211

0361-0128/98/000/000-00 $6.00 211

Gol

dR

esou

rce

(Moz

)

10

30

50100200

Bar

berto

n

Supe

rior

Hom

esta

ke

SWSi

beria

Cent

ral A

sia-

Tien

Shan

Vict

oria

Bai

kal

Ara

bian

-N

ubi

anSh

ield

Rus

sia-

East

Chin

a

Mot

her L

ode

Wes

t Afri

ca

Kol

ar

Urals

Yilg

arn

Quad

rilate

roFe

rrife

ro

Am

azon

ian

2.7 Ga

50%

Volu

me

per c

ent

grow

th

25%

1.9 Ga

Age (Ga)3

2

4

6

8

10

12

14

2 0.51

A)

B)

FIG. 8. Timing of orogenic gold deposits vs. periods of crustal growth. A. Distribution of major orogenic gold provinceswith time. Adapted from Goldfarb et al. (2001a). See Groves et al. (2003) for updated version. B. Temporal evolution of con-tinental crustal growth. Note, y-axis shows relative crustal growth. Adapted from Condie (2000). Variable bar width in (A)due to varying uncertainties in age of mineralization, and in (B) to better illustrate major periods of crustal growth. The an-notations 50 and 25 percent refer to approximate percentages of recorded crustal growth in the Late Archean and mid tolate Paleoproterozoic, respectively.

the contrasting regional-scale shapes of the major goldprovinces of different age. Archean and Paleoproterozoic oro-genic gold deposits are distributed along elongatesupracrustal belts, but these are hosted within and commonlyin the central part of (cf. fig. 5.2 of Solomon and Groves,1994) roughly equidimensional cratons as noted above. Incontrast, Phanerozoic deposits are distributed in elongatebelts along the margins of these cratons or along the marginsof older, marginal Phanerozoic belts (fig. 5 of Goldfarb et al.,2001a). The time of transition between these styles of oro-genic belts is unclear, but reconstructions of Rodinia (Fig. 9)suggest that modern-style orogenic belts were in existenceprior to 1.0 Ga. These observations, combined with evidencefor timing of changes in buoyancy of the oceanic lithosphere(Fig. 5) and subcontinental lithospheric mantle (Figs. 67),suggest that an important time in the transition from plume-influenced to modern-style plate tectonics was somewherenear the end of the Archean to early Paleoproterozoic.

The unusual lack of Mesoproterozoic to Neoproterozoicorogenic gold deposits (Fig. 8A) can be explained largely interms of the preservation potential of the hosting terranes.Deposits embedded in Archean and Paleoproterozoic cra-tons would be underlain by relatively buoyant subcontinentlithospheric mantle, which would be difficult to delaminate.Hence, there would be a very high chance of preservation ex-cept adjacent to craton margins or in relatively small cratonicblocks, as probably existed in the Middle Archean, where

later orogeny could cause uplift and erosion. In contrast,modern-style, highly elongate, metasedimentary andmetavolcanic rock-dominated accretionary belts along themargins of the cratons, with their negatively buoyant sub-continent lithospheric mantle, would be much more prone touplift and erosion. Thus, it is likely that Mesoproterozoic toNeoproterozoic orogenic gold deposits did form duringmajor continental crust-forming events in the period fromca. 1.7 Ga to 600 Ma, but most were removed by long-livederosion of the narrow continental margin orogens down totheir high metamorphic grade root zones; such zones arebelow the crustal depths that typically would contain oro-genic deposits (Groves et al., 1998). The reappearance ofabundant orogenic gold deposits at ca. 600 Ma suggests thatthis is an approximate threshold for preservation, or lack ofcomplete destruction, of deposits in modern-style orogenicbelts. The abundance of gold placers spatially associated withmany Phanerozoic orogenic gold provinces attests to theirprogressive erosion such that, within 500 to 600 m.y. of ini-tial unroofing, an orogenic gold province and associated plac-ers may be totally lost from the geologic record under condi-tions of Cordilleran-style plate tectonics. Similarly, the lackof any economically significant orogenic gold provinces inrocks younger than ca. 50 Ma suggests that 50 m.y. may bethe minimum period to uplift and expose a productiveprovince (Goldfarb et al., 2001a). Uplift rates were lowerthan those for the arc-related porphyry and epithermal

212 GROVES ET AL.

0361-0128/98/000/000-00 $6.00 212

Rodinia

AMAZONIA

WestAfrica

La Plata

Kalahari

MadagascarTibet

Mawson

S Australia

N Australia

S China N China

SIBERIA

Berentsia

BALTICA

LAURENTIA

INDIA

Malaysia

1.3 - 0.9 Ga OrogensJuvenile Crust

FIG. 9. Schematic representation of Rodinia, a supercontinent formed between 1.3 and 0.9 Ga and fragmented between750 and 600 Ma. The Kalahari, La Plata (Rio de La Plata) and West Africa cratons were probably never part of Rodinia. Re-construction modified after Tohver et al. (2002) and Pisarevsky et al. (2003). Note the location of the linear, elongate, Meso-proterozoic cratonic belts (orogens; shaded medium gray) around relatively equidimensional Archean-Paleoproterozoic cra-tonic blocks.

provinces of the present-day circum-Pacific, perhaps be-cause the orogenic gold deposits formed largely during trans-pressional, rather than compressional, tectonics (e.g., Gold-farb et al., 2001a).

Thus, it appears that orogenic gold deposits are inherentproducts of crust-forming events throughout Earth history.The processes responsible for the formation of this gold de-posit type were broadly similar through time, reflecting mod-erate- to high-temperature tectonic events superimposed forthe first time on growing, volatile-rich (i.e., H2O, CO2, H2S)juvenile continental crust. However, the change from aplume-influenced plate tectonic style in the latest Archean toearly Paleoproterozoic to a modern tectonic style affected thepreservation potential of terranes of different age. Hence, thetemporal distribution of the deposits (Fig. 8A) reflects a com-bination of processes of formation and preservation ratherthan a fundamental change due to secular changes in tecton-ics. In the Archean and Paleoproterozoic, crustal growth andpreservation processes worked together to produce the abun-dant deposits in those time periods, whereas preservation wasnot favored during the critical tectonic transition in theMesoproterozoic to Neoproterozoic; that is, crustal growthand associated metallogenesis and preservation were coupledin the Archean and Paleoproterozoic but largely decoupledthereafter.

Examples of Tectonically Induced Selective PreservationOther mineral deposit types that are little influenced by

near-surface redox conditions are examined below. Mostshow temporal patterns that relate to preservation processes(e.g., VHMS, paleoplacer gold), whereas the distribution ofothers can be related to the development of Archean or Pale-oproterozoic subcontinental lithospheric mantle (e.g., IOCG,intrusion-related gold).

Volcanic-hosted massive sulfide deposits

VHMS deposits, many of which are gold rich (Huston,2000), although commonly subdivided into subtypes based ongeographic settings (e.g., Meyer, 1981; Fig. 1A) or metal ra-tios (e.g., Large, 1992), are a coherent class of depositsformed at or below the sea floor by circulating hot seawater(e.g., Barrie and Hannington, 1999). Their distribution

through time has been examined by Meyer (1981), Titley(1993), and Barrie and Hannington (1999), among others.Production data for many deposits, particularly those in theformer Soviet Union and other Asian and Eastern Blockcountries, are incomplete or contradictory, as, importantly,are accurate ages for some deposits. Modern VHMS systemsare present at sea-floor spreading ridges, such as the East Pa-cific Rise, and also in backarc basins, such as the Lau basin.Where the former eventually reach the continental margin,they may be incorporated into the accreting margin in tec-tonic environments similar to those in which younger oro-genic gold deposits were generated (e.g., Goldfarb, 1997).Solomon and Quesada (2003) point out that some depositssuch as those of the Iberian pyrite belt may form in terranesafter their accretion to a margin. Where VHMS deposits andorogenic gold deposits occur in the same terrane, the formerare always older (e.g., Spooner and Barrie, 1993), possiblyleading to overprinting of VHMS deposits by orogenic goldmineralization, as summarized by Groves et al. (2003), al-though this appears to be rare (e.g., Hannington et al., 1999).

The VHMS deposits (Fig. 10) show a preservation historythat is broadly similar to that of orogenic gold deposits, withepisodic development in the early Precambrian and a morecyclic pattern in the Paleozoic. The oldest, but noneconomicVHMS deposits in the east Pilbara and Barberton terranes, atca. 3.5 to 3.25 Ga (e.g., Vearncombe et al., 1995), correspondbroadly with the oldest orogenic gold events globally (e.g.,Zegers et al., 2002) and with the earliest record of formationof significant continental crust (Fig. 2). VHMS deposits arenot uniformly distributed in greenstone belts globally, withmost belts being devoid of major deposits. The economicallysignificant ca. 2.7 Ga VHMS deposits are mainly from theAbitibi belt. The ca. 1.85 Ga VHMS deposits are in the FlinFlon district, Manitoba, Canada (Syme and Bailes, 1993) andWisconsin. These occurrences broadly coincide with hypoth-esized mantle-plume events at 2.7 and 1.9 Ga, respectively(Fig. 8B). Both regions are composed of terranes displayingevidence for thin, dominantly oceanic lithosphere at the timeof VHMS mineralization (e.g., Stern et al., 1999; Ayer et al.,2002; Hart et al., 2004). Similarly, VHMS deposits appear tohave formed and been incorporated into continental crust al-most continuously since the latest Neoproterozoic, although

100th ANNIVERSARY SPECIAL PAPER: TECTONIC PROCESSES & Au DISTRIBUTION 213

0361-0128/98/000/000-00 $6.00 213

13 2 00.5

1000

3000

2000

Mt(

ore)

Abi

tibi(

Kid

dCr

eek)

Man

itoba

(Flin

Flon

)

Kaz

akhs

tan

()

Mai

kain

Ura

ls(G

ai)Ib

eria

(Rio

Tin

to)

NW

Brit

ishCo

lum

bia

(Win

dyCr

aggy

)Tu

rkey

(Murg

al)

Japa

n(H

okuro

ko)

W Tasmania(Mt Lyell)

Age (Ga)FIG. 10. Temporal distribution of volcanic-hosted massive sulfide deposits. Data derived from Barrie and Hannington

(1999)examples of specific deposits given in parentheses. Due to some uncertainties in age, deposits are grouped in placesinto broad time groups rather than absolute time periods. Note expanded time scale from 0 to 0.5 Ga.

there are distinct peaks in the early to middle Paleozoic andin the Mesozoic (e.g., Titley, 1993), as there are for orogenicgold deposits. Similarly, although VHMS deposits are recog-nized between ca. 1.7 Ga (e.g., Jerome; McCandless et al.,1999) and 600 Ma (e.g., Barrie and Hannington, 1999), theytoo are markedly under-represented in this age range. Again,it appears most likely that the VHMS deposits were largelylost from the geologic record because the linear belts inwhich they formed occur above less buoyant subcontinentallithospheric mantle and were uplifted and deeply eroded totheir roots. The youngest exposed VHMS deposits and/orprospects are of Tertiary age, for example Ufuo in PNG (Cor-lett and Akiro, 1999), deposits in Prince William Sound ofsouthern Alaska (Goldfarb, 1997), and deposits in Ecuador(Chiaradia and Fontbot, 2001), corresponding broadly to theage of the youngest exposed, orogenic gold deposits.

The most logical conclusion is that, although VHMS de-posits formed throughout Earth history, their temporal distri-bution, as for orogenic gold deposits, largely represents a pat-tern of preservation due to changes in the lithosphere causedby changing global tectonic processes.

Placer and paleoplacer gold

Most giant placer gold deposits were deposited in foreland(commonly retroarc) basins in Mesozoic-Cenozoic conver-gent margins of the circum-Pacific (e.g., New Zealand, Cali-fornia, Alaska) through the erosion of Paleozoic to Mesozoicorogenic gold deposits (e.g., Henley and Adams, 1979; Ed-wards and Atkinson, 1986; Goldfarb et al., 1998). SimilarPaleozoic margins were the source for additional large placerfields in Victoria (Hughes et al., 2004) and the EasternCordillera of South America (e.g., Haeberlin et al., 2003), al-though much of the central Asian Paleo-Tethyan margin was

preserved by subsequent Himalayan continent-continent col-lision. In addition, some significant gold placers formed inpermafrost regions (e.g., Eastern Russia, Siberia) or in areasof deep tropical weathering of Precambrian orogenic gold de-posits, as for example in the Ashanti belt of Ghana and theTapajos region of the Amazon (e.g., Santos et al., 2001). Mostplacer deposits were mined from Recent river systems andbeaches, although some Tertiary paleoplacers were preservedby overlying volcanic and volcaniclastic rocks (e.g., Ballaratand Bendigo, Victoria, Australia). The fact that most world-class placers are associated with source lodes older than ca.100 Ma (i.e., Fairbanks, Nome, Eastern Russia) suggests that~50 m.y. is an approximate threshold between unroofing of agold province in a Cordilleran-style orogen and loss of a sig-nificant percentage of gold to the secondary environment.

Paleoplacer gold deposits are rare in the geologic recordbefore the Tertiary (Fig. 11), yet the giant Late Archean Wit-watersrand deposits represent the largest gold province onEarth. Their origin has been controversial, with both modi-fied placer (e.g., Minter et al., 1993) and hydrothermal (e.g.,Phillips and Myers, 1989; Barnicoat et al., 1997) models beingproposed. However, a variety of recent evidence, particularlyfrom Re-Os dating of both gold and associated roundedpyrites, which yields ca. 3.0 Ga ages, which are presedimen-tation of the host conglomerates (e.g., Kirk et al., 2001, 2002),suggests that the Witwatersrand gold ores are modified pale-oplacers, as summarized by Frimmel et al. (2005).

Significantly, the Witwatersrand gold was deposited in theCentral Rand Group in a retroarc foreland basin setting(Kositcin and Krapez, 2004) that is similar to many moderndepositional settings of placer gold. Extreme environmentalconditions, specific to the early Earth, including a potentiallymore acidic and chemically aggressive atmosphere (Holland,

214 GROVES ET AL.

0361-0128/98/000/000-00 $6.00 214

Age (Ga)2.0 02.5 1.5 1.03.0 0.5

t(Au

)

100

200

300

400

500

100,000

10,000Witwatersrand

Tarkwa

Jacobina and Roraima

Modern Placers

2

15

3000

Moz

(Au)

300

FIG. 11. Temporal distribution of placer and paleoplacer gold deposits. Main sources of data are Boyle (1979), Bache(1987), Goldfarb et al. (1998), Milesi et al. (2002), and Frimmel et al. (2005). The Tertiary-Recent placer total production isa minimum as a significant proportion of placer gold production by small-scale miners may have been unrecorded in officialproduction records. The total production is given for Tertiary to Recent placers as a single bar as there are considerable un-certainties in the age of some placers.

1984), the lack of vegetation and other organisms, and hencethe predominance of braided stream environments and po-tential for effective wind erosion and sorting (e.g., Minter,1999), can potentially explain the giant size of the Witwater-srand deposits. Some present-day placers are derived fromadjacent orogenic gold deposits (e.g., Hughes et al., 2004).However, many modern placers are not derived from unusu-ally large primary orogenic deposits (i.e., Nome, Alaska;Klondike, Yukon) but owe their origin to tectonic, erosional,and sedimentary processes that combine to produce ex-tremely effective sorting and detrital gold concentration. Thesource of the Witwatersrand gold is unclear. Frimmel et al.(2005) suggest that the Witwatersrand gold, with anomalouslyhigh Re and Os concentrations and 187Os/188Os of 0.108,equivalent to the Os isotope composition of the 3.0 Ga man-tle (e.g., Kirk et al., 2002), was likely derived from dispersedmagmatic phases in mantle-derived mafic and/or ultramaficrocks of now eroded granitoid-greenstone terranes in the hin-terland to the Central Rand basin. A source region with a bulkconcentration between about 0.5 and 7 ppb Au, in agreementwith a value of 4 ppb Au for Barberton-type granitoid-green-stone crust (Robb and Meyer, 1990), is all that is required toform the Witwatersrand deposits (Loen, 1992), provided thatconcentration mechanisms were very effective (Frimmel etal., 2005). High background gold values in the early Archeankomatiites may have correlated with mantle plumes derivedfrom the mantle boundary with the gold-rich core (Kerrich,1999). Alternatively, the gold could have been derived fromgold deposits equivalent in age to, or slightly younger than,the orogenic Barberton gold deposits, in the proposed hinter-land of the Witwatersrand basin, for example in the Amalia-Kraaipan, Murchison, Pietersburg and Giyani belts, but theage of this gold is poorly defined (e.g., Barton, 1984; Beattieand Barton, 1992). A source of some gold from epithermal orepizonal orogenic gold deposits is also reasonable becausesome of the gold has low fineness and high mercury contents(Feather and Koen, 1975). Interestingly, orogenic deposits inthe Murchison greenstone belt of the potential hinterlandalso have gold with high mercury contents (cf. Poujol et al.,1996). As stated above, the precise source or sources of thegold are currently unclear, but mass-balance arguments(Frimmel et al., 2005) clearly show that there was sufficientgold in the hinterland of the Witwatersrand basin to producethe recorded gold concentrations.

The question remains as to why these giant Witwatersranddeposits are preserved. On a larger scale, the Witwatersrandores almost certainly owe their preservation to their locationin old subcontinental lithospheric mantle (e.g., Shirey et al.,2003), where buoyancy protected them from subsequent de-struction. Whereas the host Kaapvaal craton is not unique interms of the antiquity of its subcontinental lithospheric man-tle (e.g., Richardson et al., 2004), the age of its earliest wellpreserved basins (ca. 3.02.8 Ga) and the remarkable preser-vation of its sedimentary and volcanic basins from ca. 3.5 Gato the Mesoproterozoic attests to its long-term stability andhigh preservation potential. Importantly, the Re-Os age of thedetrital gold and pyrite in the Witwatersrand paleoplacers isca. 3.0 Ga, 300 m.y. earlier than the first significant crust-forming event at 2.7 Ga (Figs. 2, 8B), so that early subconti-nental lithospheric mantle could have been in place at this

time, as also indicated by the occurrence of detrital diamondsin the Witwatersrand rocks. The detrital gold was presumablyderived from the upthrusted margins of relatively small con-tinental blocks to the north and west, which would have beensusceptible to destruction despite potentially buoyant sub-continental lithospheric mantle. The Central Rand Group ofthe Witwatersrand basin was probably deposited in the fore-land of a collision between the amalgamated Witwatersrandblock to the south (Schmitz et al., 2004) and the Pietersburgblock to the north and was not related to the major interac-tion between the larger Zimbabwe and Kaapvaal cratons dur-ing Limpopo orogeny, more than 100 m.y. after Witwater-srand Supergroup sedimentation (Frimmel et al., 2005, andreferences therein). The involvement of relatively small Mid-dle Archean crustal blocks, rather than large cratons, can ex-plain the widespread uplift and erosion of the now erodedterranes that are the suggested source of Witwatersrand gold.

There are also clearly regional-scale processes that assistedpreservation. The most important was the outpouring of theKlipriversberg Group basaltic lavas over the Witwatersrand sed-imentary basin, but there was also the possible formation of aresistant veneer of impact melt over at least part of the basin asa result of the Vredefort meteorite impact in the center of theWitwatersrand basin, as summarized by Frimmel et al. (2005).

The other significant gold paleoplacer provinces are Paleo-proterozoic in age (Fig. 11), with Tarkwa in Ghana at ca. 2.1Ga (Pigois et al., 2003) being the largest, and Jacobina (ca. 2.0 0.1 Ga) and Roraima (ca. 1.96 Ga) representing smallerprovinces (Frimmel et al., 2005). The gold production fromthese probable foreland basin settings is two orders of magni-tude lower than for the Witwatersrand (Fig. 11). These de-posits are 100 to 200 m.y. older than the 1.9-Ga peak in crustproduction and presumably formed from unroofing and ex-posure of early orogenic gold source provinces (see above). InGhana, at least, some orogenic gold deposits overprint preex-isting Tarkwa-type paleoplacers (Pigois et al., 2003). As sug-gested for the Witwatersrand, it is, thus, most likely that thepaleoplacer deposits owe their preservation to the buoyancyof the underlying Paleoproterozoic subcontinental lithos-pheric mantle, developed during the protracted orogenicevents in which the primary gold deposits formed.

It is suggested that placers and paleoplacers, like orogenicgold and VHMS deposits, display a temporal pattern largelydictated by preservation. Presumably, they formed through-out Earth history whenever orogenic gold provinces were up-lifted and eroded, particularly after 1.7 Ga, but depositsformed before the Tertiary only survived where regional-scaleprocesses assisted their preservation above buoyant subconti-nental lithospheric mantle.

Examples of Tectonically Induced Selective Formation and Preservation

Iron-oxide copper-gold (IOCG) deposits

The IOCG deposit type (e.g., Hitzman et al., 1992) has beenexpanded to include many different styles of iron-oxiderichmineralization that formed in a variety of tectonic settings (cf.Hitzman, 2000), but only those deposits with significant cop-per- and iron-bearing sulfides and gold resources are consid-ered here. As summarized by Groves and Vielreicher (2001),

100th ANNIVERSARY SPECIAL PAPER: TECTONIC PROCESSES & Au DISTRIBUTION 215

0361-0128/98/000/000-00 $6.00 215

their origin is equivocal. However, there is a clear connectionto alkaline magmatism, particularly if the giant Palabora Fe-P-REE-Cu-Au-PGE deposit of South Africa is included.Neodymium isotope data for the giant Olympic Dam deposit,South Australia, combined with gravity and magnetic data,also implicate a mafic alkaline body at depth (e.g., Campbellet al., 1998) that could be the feeder to lamprophyre dikes inthe deposit. Dating of the giant Salobo (Requia et al., 2003)and Igarap Bahia (Tallarico et al., 2005) deposits at Carajs,Brazil, also suggests a temporal association with A-typegranitic magmatism, as does the presence of smaller, youngerIOCG (Bi Sn W) and Au-PGE deposits associated withtwo subsequent events of alkaline magmatism in the same re-gion (Groves et al., 2004). Importantly, all significant Pre-cambrian IOCG deposits are sited within about 100 km of themargins of Archean cratons (e.g., Palabora, Carajs, OlympicDam) or close to the boundary of Archean and Paleoprotero-zoic lithosphere as interpreted from remote sensing geophys-ical data (e.g., Cloncurry, Queensland, Australia).

The temporal distribution of economically significant Pre-cambrian IOCG deposits (Fig. 12) shows major peaks in thelatest Archean (ca. 2.57 Ga; e.g., Carajs deposits), Paleopro-terozoic (ca. 2.05 Ga; e.g., Palabora), and Mesoproterozoic(ca. 1.59 Ga; e.g., Olympic Dam) that are significantly offsetfrom the main periods of crustal growth at ca. 2.7 and 1.9 Ga(Fig. 2). The oldest Carajs deposits are sited in a region ofLate Archean postcratonization platformal sequences of sim-ilar age to those in the Pilbara and Kaapvaal cratons, ratherthan greenstone sequences (e.g., Grainger et al, 2002). TheIOCG deposits formed in the Amazon craton, however, at ap-proximately the same time as late-orogenic gold deposits inthe eastern Dharwar block of the Indian Shield (Hamiltonand Hodgson, 1986; Balakrishnan et al., 1999), emphasizingthe diachroneity of Archean cratonization despite the majorpeak in crustal growth at about 2.7 Ga. Presumably, large

IOCG deposits could also have formed in the Pilbara andKaapvaal cratons in the Late Archean if alkaline magmatismhad occurred at that time. The Olympic Dam deposit formedat ca. 1.59 Ga, very soon after amalgamation of Archean andPaleoproterozoic blocks to form proto-Australia at ca. 1.75 to1.70 Ga (e.g., Betts et al., 2002), in association with outpour-ing of the Gawler lavas and associated intrusions of alkalineaffinity. In each case, some form of anorogenic tectonism andalkaline magmatism is implied.

The association of the giant Precambrian IOCG depositswith alkaline magmatism and their location near craton mar-gins suggests that the magmas were derived by small degreesof post-tectonic partial melting of subcontinental lithos-pheric mantle previously metasomatized (K, U, Th, Au,LREE) during and after the events related to cratonization.This could explain the offset between the ca. >2.6 Ga ages ofcrustal growth and/or cratonization and the younger periodsof alkaline magmatism and associated IOCG mineralization.Thus, there is a major tectonic control on the formation oflarge Precambrian IOCG deposits close to the margins ofpreexisting cratons during periods of magmatism of alkalineaffinity. However, the temporal distribution of IOCG de-posits (Fig. 12) also reflects preservation, with giant andworld-class Precambrian examples of the deposit type selec-tively formed and preserved in buoyant subcontinentallithospheric mantle.

There are Neoproterozoic deposits of IOCG affinity atKhetri in India (probably ca. 850 Ma in age; Knight et al.,2002) and in the Lufilian intracratonic rift basin of Zambia, al-though large examples of the latter have low copper grades(

shows gross similarities to the Precambrian deposits, althoughthe sulfide assemblage is more sulfur rich, zinc and silver con-centrations are anomalous, and the REE concentrations, de-spite similar LREE enrichment, are more erratic. Candelariaclearly formed in a different tectonic setting, being related totranspression and basin inversion in a long-lived, arc-parallelstrike-slip fault system linked to subduction in the coastalbatholith of Chile, outboard from the South American Shield(e.g., Marschik and Fontbot, 2001). The ore-forming fluid isconsidered to have been magmatic, and spatially related sub-alkaline to alkaline granitoids have Sr, Nd, and Pb isotope ra-tios that indicate derivation of the parent magmas from a sub-duction-modified mantle source (Marschik et al., 2003a, b),as in Precambrian analogues. The shape of the orebody ismantolike and grossly conformable to layering, rather thanpipelike, as for the Precambrian examples (Marschik andFontbot, 2001). This suggests emplacement of the ore intomore permeable horizons rather than brecciation under highfluid pressures and may indicate a different genetic process.

Intrusion-related gold deposits

During the past decade, there has been an emphasis on thediversity in gold deposit types within metamorphic belts (e.g.,Robert and Poulsen, 1997), with increasing interest in a groupof deposits termed intrusion-related gold deposits (e.g., Silli-toe and Thompson, 1998). The original definition of this de-posit type included a large number of different styles of de-posits with different metal associations (e.g., Sillitoe, 1991),but only the intrusion-related gold systems in the sense ofLang and Baker (2001) are considered here. According toThompson et al. (1999) and Lang and Baker (2001), they arecharacterized by (1) spatial association with relatively reducedgranitoids, (2) carbonic hydrothermal fluids, (3) gold as thedominant economic element, but with anomalous Bi, W, Mo,Te, and/or Sb, (4) a low sulfide content, with no iron oxides,(5) spatially restricted and weak hydrothermal alteration inmesozonal examples, (6) a tectonic setting well inboard ofrecognized convergent plate boundaries, with complex gran-ite petrogenesis, and (7) an association with tungsten and tinprovinces. There may also be regional metal zoning, withmore silver and/or base metal-rich deposits distal to the in-trusion-related gold systems (e.g., Lang et al., 2000). As sum-marized by Groves et al. (2003), there is general acceptancethat the Fort Knox deposit in Alaska, various small and as yetuneconomic deposits in the eastern part of the Tintina goldprovince of Yukon, Canada, and disseminated gold prospectsat Timbarra, NSW, Australia, are intrusion-related gold de-posits. A notable difference compared to orogenic-type golddeposits is the order of magnitude lower gold grades. Allother large deposits that are sometimes classified as intrusion-related gold deposits by different authors are of uncertain ori-gin and have many characteristics more akin to orogenic golddeposits (Groves et al., 2003). Hence, no temporal pattern ispresented here.

An important difference between deposit types is the tec-tonic setting of the undoubted intrusion-related gold depositsthat occur well inboard from the convergent margins andorogenic gold deposits that form in the margins. The intru-sion-related gold deposits are located in deformed and meta-morphosed shelf sequences adjacent to cratonic margins,

rather than in the turbiditic or mafic volcanic rock sequencesin commonly forearc settings that characterize orogenic golddeposits. This also explains their spatial association withbroadly contemporaneous tin and tungsten deposits, whichalso form adjacent to older continental crust well inboardfrom subduction zones (e.g., Solomon and Groves, 1994). Thesuites of felsic intrusions associated with intrusion-relatedgold systems are unusual. Lang and Baker (2001) mention co-eval intrusions of alkalic, metaluminous calc-alkalic, and per-aluminous compositions. Mair et al. (2003) show that, in theTombstone belt of the Yukon, the intrusion-related gold de-positassociated intrusions have mixed mantle and crustal sig-natures. They show that the complex array of intrusions islikely due to mantle-derived mafic alkaline magmas imping-ing at the base of the crust, consequent melting of the crustto produce felsic magmas, and contamination of these mag-mas by the basement and shelf sequences they intrude. Thesefactors may combine to produce fertile magmas with suitableredox switchovers for the concentration of gold and its subse-quent release in relatively reducing magmatic-hydrothermalfluids.

The conjunction of anomalous granitic magmas and un-usual inboard tectonic setting for their generation is likely tohave been rare in geologic history. Such settings would havebeen extremely rare, or absent, in the early Precambrianwhen the earliest cratons formed, as also shown by the lack oftin and tungsten deposits related to fractionated granites insimilar settings. The formation of the deposits in crust aboveyounger subcontinental lithospheric mantle near the marginsof cratons also would have limited their preservation. Thustheir rarity and the restriction of undoubted examples to thelate Phanerozoic (Timbarra, Permo-Triassic; Tombstone de-posits, Late Cretaceous) are expected.

SynthesisBased on the synthesis of observations presented above, it

appears that Archean, and probably Paleoproterozoic, tec-tonic processes were dominated by relatively buoyant oceaniclithosphere and plume events, with the largest, short-lived,catastrophic events being responsible for voluminous conti-nental growth. Plume-influenced to -dominated plate tecton-ics operated rather than a modern-style of plate tectonics.The progressive decline of heat flow and decrease in plumeactivity also affected the nature of the subcontinental lithos-pheric mantle, which records a distinctive temporal evolution(e.g., Poudjom Djomani et al., 2001; Griffin et al., 2003)linked to the synchronous formation of crust and its lithos-pheric root. Archean subcontinental lithospheric mantle, rep-resenting residues and/or cumulates from deep high-degreemelting related to the early abundance of mantle-plumeevents, had a relatively low density and hence was buoyant.The more or less similarly sized, broadly equidimensionalearly Precambrian cratons were probably a result of theseprocesses and consequent melting of the lower and middlecrust. Subsequent Proterozoic lithosphere was more denseand had positive to neutral buoyancy, whereas Phanerozoiclithosphere was even more dense and negatively buoyant(Poudjom Djomani et al., 2001). These changes in buoyancyproduced global tectonic patterns in which post-Mesopro-terozoic orogenic belts surrounded early Precambrian nuclei.

100th ANNIVERSARY SPECIAL PAPER: TECTONIC PROCESSES & Au DISTRIBUTION 217

0361-0128/98/000/000-00 $6.00 217

Gold-bearing deposits that form at shallow crustal depths inconvergent margins (e.g., porphyry-skarn-epithermal Cu-Mo-Au-Ag systems) are highly susceptible to tectonic uplift anderosion and are only sporadically and selectively preservedfrom times older than the Mesozoic by accretion or collisionof their host arcs with continental blocks. In contrast, oro-genic gold deposits formed over a much larger range ofcrustal depths in deforming convergent margins undergoingdeformation for a period of at least 3.4 b.y. (Goldfarb et al.,2001a, b). They have a temporal distribution that broadly mir-rors that of juvenile crustal growth, particularly the changefrom episodic to more cyclic growth with time, although theyare rare between 1.7 Ga and 600 Ma. The exceptional en-dowment of Archean and Paleoproterozoic provinces, despitetheir age, suggests that plume-influenced plate tectonics pro-duced large gold deposits due to associated high thermal flux.They were incorporated into crust above buoyant subconti-nental lithospheric mantle and preserved, particularly in thecenters of cratons. In the Mesoproterozoic, plume activity de-clined, tectonics akin to modern plate tectonics evolved, andthe subcontinental lithospheric mantle became negativelybuoyant. It appears that uplift and erosion in these newlyformed orogenic belts along the margins of early Precambriancratons destroyed the majority of any orogenic gold depositsthat formed between 1.7 Ga and 600 Ma. From 600 untilabout 50 Ma, the temporal distribution of the deposits grosslyrepresents that of the cycle of orogenesis and crustal growth.Thus, formation of the orogenic gold deposits was broadlycontrolled by the timing and intensity of crust-formingevents, but their temporal distribution was strongly affectedby the buoyancy of the subcontinental lithospheric mantlebelow their host terranes. VHMS deposits, which have a sim-ilar, almost 3.5-b.y. history of formation, have a very similartemporal distribution, particularly their contrasting Precam-brian and Phanerozoic patterns, suggesting that preservationwas a dominant factor in their present distribution. Highthermal flux related to plume-influenced plate tectonics pro-duced highly endowed Archean and Paleoproterozoic VHMSprovinces, as for orogenic gold.

Gold placers have probably formed by erosion of orogenicgold deposits since the Middle Archean, but paleoplacers andplacers display a highly anomalous temporal distribution withpeaks in the early Precambrian and in the Tertiary to Recent.The giant, extraordinarily gold rich palaeoplacers of the Wit-watersrand probably owe their formation to effective fluvialsorting under extreme climatic conditions and their preserva-tion to generation of early buoyant subcontinental lithos-pheric mantle under the Kaapvaal craton. It is also possiblethat greenstone source rocks in the hinterland to the hostforeland basin were enriched in gold due to mantle-plume ac-tivity in the early Earth. Giant Precambrian IOCG depositsalso appear to have required the preexistence of buoyantArchean (and/or Paleoproterozoic) subcontinental lithos-pheric mantle for their formation and subsequent preserva-tion. A concentration of deposits in the Late Archean and latePaleoproterozoic to early Mesoproterozoic appears to haveformed near craton margins during alkaline magmatism de-rived from previously metasomatized mantle lithosphere.

The intrusion-related gold deposits are rare because theyrequire the conjunction of near-craton setting and shelf

sedimentary sequences to provide the correct mantle andcrustal ingredients for generation of the causitive granites.Such settings were unlikely in the early Precambrian, andyounger intrusion-related gold deposits that formed near themargins of cratons would have had only limited chances ofpreservation.

A summary diagram (Fig. 13) shows the relative locationsof deposit types discussed in the text in terms of their tec-tonic setting at the time of formation and where they mighthave been preserved in the context of underlying subconti-nental lithospheric mantle. It is concluded that the majorfactor affecting the temporal distribution of many gold-bearing deposits, in terms of evolution of tectonic processes,is the nature of subcontinental lithospheric mantle formedat different times in Earth history. Archean (and/or Paleo-proterozoic)-style lithosphere favored the formation ofIOCG deposits and hence their temporal distribution. Theswitchover from plume-influenced buoyant plate tectonicsto modern-style plate tectonics, with the shift from buoyantto negatively buoyant subcontinental lithospheric mantle,strongly influenced the patterns of preservation of other de-posits, for example, orogenic gold and VHMS deposits, ex-amples of which started to form before the widespread oc-currence of Archean cratons. The intrusion-related golddeposits may represent a special case in that they requirenear-craton settings, but form outside the cratons in nega-tively buoyant lithosphere, and hence are rare and largelyrestricted to the Phanerozoic.

The temporal distribution of most gold-bearing depositsdiscussed here reflects the first-order evolution from mantle-plumeinfluenced plate tectonics to a modern style of platetectonics in a cooling Earth. Coupled crustal growth andpreservation in the Archean and Paleoproterozoic evolved todecoupled episodes of growth and preservation from theMesoproterozoic onward as a result of irreversible changes tothe subcontinental lithospheric mantle with time.

Future Research and Exploration SignificanceIn the past few years, the global mining and exploration in-

dustry has changed dramatically, with amalgamation intolarger global companies and consequent loss of exploration-focused medium-sized companies. The large companies havebecome more risk averse at the very time when near-surfacetargets in relatively well exposed terranes are nearing exhaus-tion, particularly in mature mineral provinces. Future signifi-cant discoveries will have to be made at depth in covered ter-ranes, perhaps in remote locations, posing a dauntingchallenge to the industry. Clearly, conceptual targeting will berequired and, in turn, will necessitate even greater integrationof theoretical and empirical geoscience into mineral explo-ration (e.g., McCuaig and Hronsky, 2000). In a risk-averse en-vironment, relatively small, but potentially highly productive,segments of the globe will have to be selected for intensiveconceptually based exploration, rather than continuation ofthe presently less focused exploration effort, often highly re-liant on near-surface geochemical anomalies, particularly inthe case of gold exploration.

Over the past few decades, economic geology research hasbecome very deposit centric and forensic, with much of theresearch seeking to understand ore genesis through the use of

218 GROVES ET AL.

0361-0128/98/000/000-00 $6.00 218

sophisticated mineralogical, geochemical, isotopic, and fluidinclusion techniques. This has resulted in a much betterknowledge of ore fluids and their sources, metal transport anddeposition, and deposit-forming processes. However, thesestudies mostly help understand the mechanism of ore depositformation (the how) but not the specific location of the

deposit (the where). Thus, very few of the researched para-meters can be used directly to select specific terranes withinspecific segments of the Earth that can be intensivelyexplored for world-class to giant mineral deposits. This is con-firmed by research studies which demonstrate that giant hy-drothermal deposits form essentially from the same fluids and

100th ANNIVERSARY SPECIAL PAPER: TECTONIC PROCESSES & Au DISTRIBUTION 219

0361-0128/98/000/000-00 $6.00 219

- 20- 10

102030

100

150

200

km PHANEROZOICLITHOSPHERE

NEOPROTEROZOICLITHOSPHERE

CONTINENTAL CRUST

Young negativelybuoyant SCLM- moderate uplift

Old negativelybuoyant SCLM- major uplift Moderately buoyant

SCLM - minor upliftBuoyant SCLM- minor uplift

VHMS OGD IRGD IOCGMinorOGD OGD IOCG VHMS OGD

Witwatersrand - typePaleoplacer

***

*

****

x x

xCONTINENT

Accretionary wedge

Granitoids

Asthenosphere

Continental crust

Oceanic crust

Older cratonSubcrustallithosphere

Extensional fault

Deformed shelf sequence

Compressionalfault/thrust

OCEANIC ARC

EpithermalAu-Ag

VHMSCu-Au Orogenic

Au

EpithermalAu

Paleoplacer Au(Witwatersrand)

Intrusion-related AuPorphyry Cu-Au(skarns)

Porphyry Cu-Au-Mo (skarns )

Carlin-style Au

BACK ARCACCRETEDTERRANES

CONTINENTALARC

FORELAND BASIN

BACK-ARC

CRATONMARGI N

ARCHEANLITHOSPHERE

PALEOPROTEROZOICLITHOSPHERE

A)

B)

Shelf sequences

Conglomerates

Sedimentary rocks

Volcanic rocks

Metamorphic belts

Fold belts

FIG. 13. Schematic lithosphere-scale sections showing (A) the formational environments of gold-bearing deposit types dis-cussed in the text (modified from Groves et al., 1998, with palaeoplacer, intrusion-related, and Carlin-style gold depositsadded), and (B) the environments of preservation of the same deposit types. Note that only the spatial positions of the envi-ronments are shown, whereas it is evident that the deposit types formed and/or were preserved in those environments at dif-ferent times in the evolutionary history of the hosting terranes. Both sections are of necessity generalized and simplified toinclude all environments and deposit types. IOCG = iron-oxide copper-gold deposits, IRGD = intrusion-related golddeposits, OGD = orogenic gold deposits, VHMS = volcanic-hosted massive sulfide deposits.

by similar processes as smaller deposits (see, e.g., papers inWhiting et al., 1993, and Cooke and Pongratz, 2002). It ap-pears much more likely that it is the conjunction of province-scale characteristics of a terrane, rather than deposit-scale pa-rameters, that dictate whether a giant deposit will be presentor not, as summarized, for example, by Groves et al. (2003)for gold deposits in metamorphic belts and noted by Richards(2003) for porphyry Cu deposits.

In order to understand these province-scale controls onworld-class to giant deposits, and utilize this in predictivemineral discovery under cover, it will be necessary to under-stand the four-dimensional evolution of potentially prospec-tive terranes globally. To achieve this will not only requiregovernment or multiclient state of the art remote sensing andairborne geophysical databases, such have become routinelyavailable in many mature exploration terranes globally (e.g.,Australia, Canada, southern Africa, southwestern UnitedStates), but integrated research at the global to provincescale. For example, there needs to be more research on thespecific tectonic settings of mineral deposits. This includesknowledge of the evolution of the geometries of plates, sub-duction slabs and transform faults in convergent margin set-tings, and their structural and petrogenetic signals in the rockrecord of ancient terranes. The role that mantle plumes playin modifying tectonic processes and regimes, and in accentu-ating global metallogenic epochs, also needs to be better un-derstood, particularly in the early Precambrian when theymay have been the dominant control. There is also a need tounderstand the nature and thickness of subcontinental lithos-pheric mantle beneath prospective terranes at different timesduring their evolution, for example by using a combination ofseismic tomology with petrology, mineralogy and age con-straints on mantle xenoliths and inclusions in diamonds (e.g.,Shirey et al., 2003). This will not only provide an indication ofthe prospectivity of the terranes at different periods in Earthhistory but also provide strong indications of the likelihood oftheir preservation.

For most mineral deposit types (e.g., Meyer, 1988; Barleyand Groves, 1992; this study), there are particular times inEarth history when world-class to giant deposits formed andwere preserved at a global scale. Thus, not only do the tec-tonic settings and lithospheric structures of potentiallyprospective terranes need to be understood, but the ages oftheir component units and metamorphic, deformational, andintrusive events need to be established by robust geochronol-ogy. In concert with this, there needs to be more emphasis onaccurate dating of mineral deposits, using robust geochrono-logical methods on clearly ore related minerals, to better de-fine the critical age peaks of major mineral deposit types. Fi-nally, in order to trace prospective, but now dispersed,Paleozoic to Precambrian terranes across the globe, there willbe need for more research on the paleogeographic relation-ships between terranes, particularly the timing and geometri-cal arrangement of supercontinent assembly, using appropri-ate geochronology and paleomagnetic methods, as times ofassembly and breakup appear particularly significant in globalmetallogeny. It will also be important to link these data to thepatterns of global mantle convection.

In summary, in order to meet changing exploration re-quirements, there is a need to change the emphasis of the

scale of economic geology research with a shift to integratecurrent deposit-centric and forensic research into more mul-tidisciplinary studies that emphasize global tectonics andmetallogeny. This should bring attendant advances in concep-tual targeting that will lead to world-class to giant discoveriesto satisfy the resource demands of the next generation.

AcknowledgmentsThis paper was inspired by the pioneering academic con-

cepts of Chuck Meyer and the global exploration vision ofRoy Woodall. We are grateful to colleagues at the Centrefor Global Metallogeny, University of Western Australia,particularly Noreen Vielreicher, the U.S. Geological Sur-vey, and Centre for Ore Deposit and Exploration Studies(CODES), University of Tasmania, particularly MikeSolomon, for useful discussions on this topic. The paperwas improved by the useful reviews of Dallas Abbott, PhilBrown, Rob Kerrich, Steve Kesler, and Henry Pollack, andincisive editorial comments by Steve Kesler and Mark Han-nington. This paper is a contribution to the Centre forGlobal Metallogeny and Tectonics Special Research Centrepublication 304.August 30, October 25, 2004

REFERENCESAbbott, D.H., and Isley, A.E., 2002, The intensity, occurrence and duration

of superplume events and eras over geological time: Journal of Geodynam-ics, v. 34, p. 265307.

Arndt, N.T., Lewin, E., and Albarde, F., 2002, Strange partners: Formationand survival of continental crust and lithospheric mantle: Geological Soci-ety of London Special Publication, v. 199, p. 91103.

Aspler, L.B., and Chiarenzelli, J.R., 1998, Two Neoarchean supercontinents?Evidence from the Paleoproterozoic: Sedimentary Geology, v. 120, p.75104.

Ayer, J., Amelin, Y., Corfu, F., Kamo, J., Ketchum, J., Kwok, K., and Trowell,N., 2002, Evolution of the southern Abitibi greenstone belt based on U-Pbgeochronology: Autochtonous volcanic construction followed by plutonism,regional deformation and sedimentation: Precambrian Research, v. 115, p.6395.

Bache, J.J., 1987, World gold deposits: A geological classification: New York,Elsevier, 178 p.

Balakrishnan, S., Rajamani, V., and Hanson, G.N., 1999, U-Pb ages for zirconand titanite from the Ramagiri area, southern India: Evidence for accre-tionary origin of the eastern Dharwar craton during the late Archean: Jour-nal of Geology, v. 107, p. 6986.

Barley, M.E., and Groves, D.I., 1992, Supercontinent cycles and the distrib-ution of metal deposits through time: Geology, v. 20, p. 291294.

Barley, M.E., Krapez, B., Groves, D.I., and Kerrich, R., 1998, The Late Ar-chaean bonanza: Metallogenic and environmental consequences of the in-teraction between mantle plumes, lithospheric tectonics and global cyclic-ity: Precambrian Research, v. 91, p. 6590.

Barley, M.E., Pickard, A.L., Hagemann, S.G., and Folkert, S.L., 1999, Hy-drothermal origin for the 2 billion year old Mount Tom Price giant iron oredeposit, Hamersley province, Western Australia: Mineralium Deposita, v.34, p. 784789.

Barley, M.E., Rak, P., and Wyman, D., 2002, Tectonic controls on magmatic-hydrothermal gold mineralization in the magmatic arcs of SE Asia: Geo-logical Society Special Publication, v. 204, p. 3947.

Barnicoat, A.C., Henderson, I.H.C., Knipe, R.J., Yardley, B.W.D., Napier,R.W., Fox, N.P.C., Kenyon, A.K., Muntingh, D.J., Strydom, D., Winkler,K.S., Lawrence, S.R., and Cornford, C., 1997, Hydrothermal gold mineral-ization in the Witwatersrand basin: Nature, v. 386, p. 820824.

Barrie, C.T., and Hannington, M.D., 1999, Classification of volcanic-associ-ated massive sulfide deposits based on host-rock composition: Reviews inEconomic Geology, v. 8, p. 111.

Barton, J.M., Jr., 1984, Timing of ore emplacement and deformation,Murchison and Sutherland greenstone belts, Kaapvaal craton: GeologicalSociety of Zimbabwe Special Publication, v. 1, p. 629644.

220 GROVES ET AL.

0361-0128/98/000/000-00 $6.00 220

Beattie, M., and Barton, J.M., Jr., 1992, Alteration and occurrence of goldmineralization in the Roodepoort goldfield, Pietersburg granite-greenstoneterrane: South African Journal of Geology, v. 95, p. 131140.

Betts, P.G., Giles, D., Lister, G.S., and Frick, L.R., 2002, Evolution of theAustralian lithosphere: Australian Journal of Earth Sciences, v. 49, p.661695.

Bickle, M.J., 1986, Implications of melting for stabilisation of lithosphere andheat loss in the Archaean: Earth and Planetary Science Letters, v. 80, p.314324.

1990, Mantle evolution, in Hall, R.P., and Hughes, D.J., eds., Early Pre-cambrian basic magmatism: Glasgow, Blackie and Sons, p. 111135.

Blundell, D.J., 2002, The timing and location of major ore deposits in anevolving orogen: The geodynamic context: Geological Society Special Pub-lication, v. 204, p. 112.

Boyle, R.W., 1979, The chemistry of gold and its deposits: Geological Surveyof Canada, Bulletin 280, 584 p.

Brimhall, G.H, Jr., 1987, Metallogenesis: Reviews in Geophysics, v. 25, p.10791088.

Campbell, I.H., Griffith, R.W., and Hill, R.I., 1989, Melting in an Archaeanmantle plume: Heads its basalt tails its komatiites: Nature, v. 339, p. 697699.

Campbell, I.H., Compston, D.M., Richards, I.P., Johnson, J.P., and Kent,A.R., 1998, Reviews of the application of isotopic studies to the genesis ofCu-Au mineralization at Olympic Dam and Au mineralization at Porgera,Tennant Creek district and Yilgarn craton: Australian Journal of Earth Sci-ences, v. 45, p. 201218.

Champion, D.C., and Sheraton, J.W., 1997, Geochemistry and Nd isotopesystematics of Archaean granites of the Eastern goldfields, Yilgarn craton,A