magma transport and metasomatism in the mantle: a critical ... · magma transport and metasomatism...

American Mineralogist, Volume 78, pages 1117-1134, 1993

Magma transport and metasomatism in the mantle: A criticalreview of current geochemical models

J. E. NrnlsoN, fI. G. WrlsHrnnU.S. Geological Survey, 345 Middlefield Road, M.S. 975, Menlo Park, California 94025, U.S.A.

Ansrnrcr

Conflicting geochemical models of metasomatic interactions between mantle peridotiteand melt all assume that mantle reactions reflect chromato$aphic processes. Examinationof field, petrological, and compositional data suggests that the hypothesis of chromato-graphic fractionation based on the supposition oflarge-scale percolative processes (Navonand Stolper, 1987) needs review and revision. In the hypothesis, melts develop enrichmentfronts of incompatible elements as the melt percolates through a porous mantle columnof refractory peridotite composition and imprint the fractionation patterns on peridotiteelsewhere.

Current models that use Navon and Stolper's (1987) chromatographic fractionationconcept are applied to rocks of the Lherz and Horoman massifs. The calculations producepoor or limited results for the sequence of compositional variations in time and space,and the assumptions do not accord with field relations and estimates of mantle conditionsfrom experiments. In the Lherz model, modest LREE enrichments require melt percola-tion for as long as 25000 yr, an unrealistic life span for mantle dike conduits that supplymetasomatizing melts. Continuous porous flow of melts also requires host peridotite tem-peratures at or above the liquidus.

Models of regional pervasive porous flow conflict with structural and seismic evidencethat fractures control fluid transportation in the upper mantle. Effects of porous-mediumflow have been inferred in studies of mantle peridotite samples on scales of tens of metersat most, but are well documented only on scales of centimeters or decimeters. In all thesehypotheses, porous flow is fundamentally controlled by proximity to magma-filled frac-tures.

Well-constrained rock and mineral data from xenoliths indicate that many elements thatbehave incompatibly in equilibrium crystallization processes are absorbed immediatelywhen melts emerge from conduits into depleted peridotite. After reacting to equilibriumwith the peridotite, melt that percolates away from the conduit is largely depleted ofincompatible elements. Continued addition of melts extends the zone of equilibrium far-ther from the conduit. Such a process resembles ion-exchange chromatography for H.Opurification, rather than the model of chromatographic species separation proposed byNavon and Stolper (1987).

INrnooucrroN

Debate about the scale of mantle metasomatism hasbeen ongoing for two decades. Early speculative hypoth-eses identified cryptic mantle metasomatism as a region-al-scale process in contrast to demonstrably local modalmetasomatism (Harte, 1983). Formation of harzburgiteon a regional scale was postulated to result from disso-lution of clinopyroxene due to pervasive melt infiltrationof lherzolite (Kelemen et al., 1992), and regional forma-tion of biotite pyroxenite was ascribed to metasomaticconversion of peridotite (Lloyd et al., 1987; Dawson andSmith, 1992). These hypotheses are essentially untest-able, but recent studies of massifs and xenoliths promiseto give us a better framework for understanding both thescales and processes of mantle metasomatism.

We have long studied composite mantle xenoliths and

also have made field observations and petrologic studiesof the Lherz and other massifs. Our purpose in this paperis to point out problems inherent in current geochemicalmodels that assume large-scale infiltration (percolation)processes. We question some interpretations of the pro-cesses that formed the observed geochemical patterns andthe inferred scales and consequences of those processes.By doing so, we hope to focus debate on whether themodels support or prove a priori assumptions of broadscale porous-medium flow in the upper mantle, com-pared with magma transfer through fractures, accompa-nied by infiltration of wall rocks. We again emphasize thesimilarity of rock types, contact relations, and reactionsbetween massifs and xenolith suites (Wilshire and Pike,l 97 5).

Extraction and infiltration of magma are two sides of

0003-004v93 /|rr2-r I l7$02.00 trtT

l l l S

the same coin, and either or both can involve flow througha porous medium. Melts form when a portion of the man-tle exceeds the solidus temperature by progressive heatingor depressurization, and the melts may percolate somedistance before finding or creating a fracture and openinga conduit (e.g., Nicolas, 1986; Rabinowicz et al.. 1987:Wilshire and Kirby, 1989). Theoretical considerationssuggest that the asthenosphere is a region ofmelting andmelt percolation; however, abundant field evidence in al-pine massifs and ophiolites and data from mantle xeno-liths indicate that large volumes of magma are transmit-ted through the upper mantle in fractures. Metasomaticcompositional gradients adjacent to magma conduitsclearly show that melts react with and infiltrate wall rock(e.g., Wilshire and Shervais,19751' Irving, 1980; Wilshireet al., 1988; Nicolas, 1989, 1990; O'Reil ly et al., l99l;Nielson et al., 1993). These processes may affect relative-ly large volumes of the upper mantle by repetition.

Models of mantle processes that assume large-scalepercolation use an analogy drawn by Navon and Stolper(1987) between mantle processes and chromatographicfractionation in laboratories to explain observed com-positional variations in mantle rocks (e.g., Bodinier et al.,1990; Vasseur et al., l99l; Takazawa et al., 1992). Inthose models, elements (moslly REE) are thought to be-come separated in a melt as a result of partitioning re-actions while the melt percolates distances between I mand I km through a column of mantle matrix. The meltsimpose their fractionated patterns on volumes of deplet-ed mantle somewhere along the percolation path.

Mantle models are difficult to constrain because, in mostcases, the original compositions of melt and matrix areuncertain. If the natural samples have been subjected torelatively few identifiable episodes of whatever processesoperate in the mantle, simple experimental or theoreticalmodels may be applicable and subject to constraints.However, samples that have undergone numerous events,involving similar or different processes, are likely to pro-duce equivocal or misleading evidence for testing models.We can attest that even a relatively simple, well-docu-mented single-stage metasomatic event cannot be fullyconstrained (Nielson et al., 1993). Therefore, the contextand characteristics ofthe samples studied are critical el-ements of models of mantle processes and will limit theapplication of simple model assumptions.

Mr T AND FLUID TRANSPoRT IN THE MANTLE

Current models of mantle metasomatism differ fun-damentally in basic assumptions about melt transportmechanisms that operate in the directly observable partsof the mantle, as well as about the processes of melt-mantle interaction related to the contrasting modes oftransport. Compositional models that use the Navon andStolper (1987) hypothesis of chromatographic fraction-ation assume porous-medium flow through substantialdistances. Models based on fracture-controlled move-ment of melts, with limited porous-medium flow, alsouse concepts of chromatographic interaction between melt

NIELSON AND WILSHIRE: METASOMATISM IN THE MANTLE

and peridotite but with different boundary conditions.Both hypotheses are consistent with the possibility of po-rous-medium flow over large distances in the astheno-sphere, where the peridotite is above solidus temperature.

Porous-medium flow

Substantial experimental and theoretical studies havefocused on the factors that influence the degree of per-meability in partly melted peridotite. Such factors includemelt fraction and dihedral (wetting) angle (e.g., WaffandBulau, 1982; Watson, 1982; Toramaru and Fujii, 1986;Walker et al., 1988) and the possible presence of freeC-O-H fluids in the mantle (e.g., Eggler and Holloway,1977; Wyllie,1978; Watson and Brenan, 1987; Eggler,1987; Watson et al., 1990). Low dihedral angles are shownby mafic and ultramafic melt compositions in peridotiteat upper mantle T and P, which favor interconnection ofintergranular melts and indicate the most likely condi-tions for porous-medium flow. Driving forces that maydirect the flow of intergranular melts include melt buoy-ancy and deformation in upwelling mantle and rising dia-pirs (Stolper et al., l98l; Ribe, 1985;Nicolas, 1986, 1989,1990; Rabinowicz et al., 1987), compaction (McKenzie,1984), local pressure gradients (Sleep, 1988; Takahashi,1992), and reduction of interfacial energy (Watson, 1982;Riley and Kohlstedt, 1990). Except under restricted con-ditions (ambient temperature near the peridotite solidus),HrO and CO, have high dihedral angles in dunite, a factthat suggests the mantle may be effectively impermeableto fluids that are rich in these components (Watson et al.,r 990).

Studies of natural peridotites have led to conflictingconclusions about the transmission of mafic melts throughthe mantle. Waff and Holdren (1981) examined duniteand lherzolite xenoliths from several localities and con-cluded that refractory mantle may be effectively nonpo-rous to mafic magma. In contrast, experimental studiesusing pure olivine samples (Waff and Faul, 1992) wereinterpreted as indicating substantial grain-edge wetting bymafic melt, where the melt fraction is > lolo in a rock withpreferred olivine grain orientation. Thus, in mantle regionswhere peridotite is anisotropic, substantial permeabilityto mafic melts may be possible.

Experiments by Toramaru and Fujii (1986) on lher-zolite indicated that dihedral angles were significantlysmaller for melt in contact with olivine-olivine bound-aries than for olivine-pyroxene and pyroxene-p).roxeneboundaries. They concluded that the modal proportionof olivine and grain s:'z;e are the most important deter-minants of mafic melt connectivity behavior; higher ol-ivine contents and smaller grain sizes favor connectivity.Evidence supporting this conclusion was reported byBussod and Christie (1991).

Experiments by Fujii et al. (1986) on synthetic peri-dotite powders showed that melt connectivity in dry ma-trix is lowered by the presence of orthopyroxene. How-ever, connectivity in orthopyroxene-bearing rocks mayincrease in the presence ofgreater melt fractions or higher

HrO contents of the melt. Experiments by Daines andKohlstedt (1993) showed no difference in melt migrationbetween olivine samples with 20 and 00/o orthopyroxene.Field observations indicating selective formation of du-nite at the exp€nse ofpyroxene-rich bands (Quick, 1981)also do not support a significant effect of orthopyroxenecontent.

Melt transport in fractures

Numerous field-oriented studies of alpine massifs,ophiolites, and xenoliths demonstrate that a commonmode of melt transport in the upper mantle is in conduitsformed by hydraulic fracturing (also called hydrofractur-ing and extensional fracturing). Nicolas ( 1986, I 989, I 990)modeled the earliest segregation of melts from field evi-dence. He suggested that melts in a mantle diapir migrateby porous flow over distances of a few centimeters andconverge into small fluid-assisted gash fractures. Suchfracturing occurs even in rocks with large melt fractions(Shaw, 1980); the process is generally accompanied byplastic deformation.

The process of hydraulic fracturing in the mantle re-quires melt pressures in excess of the peridotite yield stress,which is estimated by Nicolas (1986) to be <50 MPa.The yield stress is exceeded when the column of con-nected (intergranular) melt attains a height on the orderof l0 km (hydrostatic pressure gradient in the melt of 5MPa/km). At this stage melt extraction is dominated byhydraulic fracturing, even in peridotite that contains alarge melt fraction (Nicolas, 1990). Various assessmentssuggest that, when compared with porous flow driven bya dynamic pressure gradient (matrix deformation, cf.Sleep, 1988), hydraulic fracturing and transport throughfractures is more effective for transporting melts than po-rous flow where the vertical dimensions of the systemexceed a few meters or tens of meters (Phipps-Morgan,1987;Nicolas, 1990).

Gravity-driven porous flow is more difficult to assessin comparison with fracture transport of fluids becausemechanisms that could induce large-scale permeabilityare poorly understood (Navon and Stolper, 1987; Nico-las, 1990). According to the model of Nicolas, conditionsfavoring porous flow on scales larger than - 100 m occurat depths below the column height needed for hydraulicfracturing and also in the few hundred meters below theMoho, where diapiric flow diverges horizontally and meltfractions are high. Because fracturing is a necessary con-dition for entrainment of xenoliths, this model impliesthat only the uppermost of possible porous flow zoneswould be sampled by xenolith-bearing magmas. How-ever, xenoliths representing the probable uppermostmantle commonly show intensive fracturing on the scaleof centimeters and dike emplacement on scales of centi-meters and larger (Wilshire and Kirby, 1989), rather thanthe porous flow that Nicolas's model permits and whichis observed in some ophiolite peridotites (Nicolas, 1989).Direct evidence of porous flow phenomena in xenolithsfrom the uppermost mantle is related to infiltration of

l 1 1 9

melts from dikes into their wall rocks (e.g., O'Reilly etal., l99l: Wilshire et al., l99l), and indirect evidence isprovided by Smith et al. (1991, 1993).

Active volcanic areas also provide evidence for hy-draulic fracturing in the mantle. From detailed geologicand seismic records for Kilauea, Ryan (1988) inferredthat a system ofnumerous concentrically arranged dikesextends beneath the volcano to a 38-km depth. In accordwith earlier models of Shaw (1980), the lateral extent ofthe hydraulic fracturing zone is oval-shaped, with the longaxis between l0 and 15 km and a mean short axis ofabout 6 km. During periods of low to moderate magmadischarge, only the core of the conduit system is used,and outlying dikes may individually stagnate. In the zoneof fracturing, bursts of seismicity about 2-3 km in di-ameter and 3-4 km high identifu hydraulically discon-nected dike systems that are likely to stagnate (Ryan,1988). These tightly columnar sets of seismic sigtals aredisplaced from the main conduit region (Ryan, 1988, hisFig. 2l).

Consistent with seismic evidence elsewhere, the con-ditions and observed phenomena beneath Hawaii favorhydraulic fracturing in the mantle to depths of about 60km (cf. Nicolas, 1986). Bursts of long-lived harmonictremors characteize a belt about 20 km wide, which ex-tends a lateral distance of 80 km at depths between 40and 60 km beneath Loihi, Kilauea, and Mauna Loa vol-canos. The seismic signature of this belt indicates releasesof large melt volumes through hydraulic fracture systems(M. P. Ryan, 1993 personal communication).

Morrr,s oF FLUrD-MANTLE rNTERAcrroNs

Chromatographic mantle hypothesis of Navon andStolper

Navon and Stolper (1987) examined the possible con-sequences of reactions between enriched melts percolat-ing long distances through relatively depleted mantle. Theyproposed that elements would fractionate in the perco-lating melts by chromatographic processes analogous tolaboratory methods in which similar species are separat-ed. In both laboratory and mantle processes, separationsoccur because of reactions between matrix in the columnand the fluid moving through it (Fig. la; cf. Samuelson,1953; Helferrich,19621, Ritchie, 1964). Navon and Stol-per (p. 286) state: ". . . as [the fluid] flows upward, itinteracts with the matrix and trends toward equilibriumwith the initial matrix [composition]. After a time thecolumn becomes 'dirty' or 'used up' and the fluid intro-duced at the base passes through the column with nointeraction." As fluid interacts with matrix, "each ele-ment behaves differently . . . and hence fractionation canoccur. . . . Properties of interest are assumed to changeover a length scale that is large compared with the matrixgrain size."

Both in laboratory processes and by the scheme of Na-von and Stolper, the properties of interest governingchromatographic fractionation are the equilibrium par-


A B D A B D

tition coefficient (Ku, matrix to melt) and the equilibriummass fraction of an element in the liquid, Xr, which re-lates bulk properties of melt and matrix and includes thefactor l/Ko (Navon and Stolper, 1987, Eq. 2, p. 287).Compatible elements react with the matrix before incom-

(3) More aliquots react (4) Final state

(3) More aliquots react (4) Final state

6 *

A B D

II2O NIELSON AND WILSHIRE: METASOMATISM IN THE MANTLE

il. Frontal Oreakthrough) method (for separation of species), no common ions

C. Ion exchange column (for purification of liquids), with or without common ions

(3) Final state

patible elements and thus move through the column at alower rate (Fig. l; D is the most compatible species, A isthe most incompatible). In the laboratory, a column ma-trix is selected for compatibility and incompatibility re-lations toward the species to be separated. With respect

(1) Initial (2) First aliquot reacts

matrlx

D \ B

A B D

malnx

D \B+D

C;matrixL I

Filr - lFUI

6A B D

0 1 2 3 rtime

,,F-q'[ffill,..HX] gJ

b. Frontal method, with common ions

(1) Initial (2) First aliquot reacts

C; matrix

A B D

A---B-T^ - -A B D

(1) Liquids continuously enter matrix

L+Ju

(2) More aliquots

6 -1

U

0

1

Q

U

0

6 *matrix

B+D. liquid ,

100Hr.oF L I

A B D

Cimatrixroo l-t.ol- / cr I

A B D

matrrx

D B+D

Fig. l. Sketch illustrating chromatographic processes. Com-patibility of species (l) is in the order: A < B < D. Liquids shownentering column at base because natural percolative processesmay not be gravity driven. (a) Frontal or breakthrough labora-tory method, no common ions. (l) Volumes of solution (com-position L) pass through ideal matrix (composition C"), whichcontains no species to be separated (C, : 0). Quantitative sepa-ration ofA and B requires that A is highly incompatible relativeto matrix and to B. (2) Reaction fronts shown as concentrationof each species compared with original matrix (C,/C.), plottedagainst percolation distance (A). (3) After several aliquots ofmix-ture enter matrix, reaction front of compatible species (D) gainson B. This happens quickly, and so quantitative separation ofBand D in the emerging liquid (C.) is rarely achieved (Helferrich,1962). (4) Matrix compositions ((,) in final state: zone D, nolonger has the potential to exchange or sorb D; it expands ifadditional liquid enters the column. Depending on actual valuesof exchange potentials (Ko), some quantity of B and A will besorbed by or held in matrix at end of column (dr, 6r) but maynot become concentrated in matrix unless the fractionated liquidis trapped in the column. (b) Frontal method with common ions(direct analogy to mantle process). (l) Solid and liquid both sho*'nwith unfractionated A, B, D content; liquid (L) is enriched 100 xover matrix (C,). (2) One aliquot of liquid enters column, matrixpreferentially exchanges or sorbs D, with the result that speciesform steep reaction fronts in matrix in l, above. Liquids in-volved in a mantle process are not observed (? on faucet); L,(r,)should be enriched in species A,L(t,) in B, and L(/r) in D. Com-(-

LI21

positions of matrix (C) vary because species D accumulates toequilibrium with the liquid composition L behind the reactionfront (zone d,). Depending on Ko values, species A and B accu-mulate in zones d. and 6r, respectively, where they are concen-trated in liquid. (3) More aliquots of liquid L enter column andreaction front of species D moves through matrix. A is still en-riched over other species in liquid l,(t'), but B and D also in-crease because the capacity of matrix to exchange these speciesis reduced by earlier reactions. Last liquids, L(/r), pass throughmatrix unchanged. (a) The d, represents pattern of A, B, and Dwhen matrix reaches equilibrium with liquid, L, at reaction con-ditions. Equilibrium concentrations of D moved through col-umn in same direction as liquid, but matrix first reached equi-librium with A and B in zones d, and 6r, respectively, then movedback through column to 6,. If enough liquid reacts, all matrix inthe column will achieve composition of 6,, and no longer reactwith liquid L. (c) Ion-exchange for liquid purification (H'O soft-eners); process evolves the same whether or not matrix containscommon ions. (1) As aliquots of liquid L enter matrix, unwantedspecies (8, D) are removed by exchange with matrix and con-centrate behind the reaction front. Liquids L(r") contain all ormost of desired species (A). (2) More aliquots of liquid continueto react; reaction front moves through matrix, and liquid ex-tracted has same composition as in Fig. 3a. (3) Reaction frontexpands to full length ofcolumn, which can no longer exchangeB and D, and liquid L passes through unchanged. Species Anever becomes significantly concentrated in matrix.

NIEISON AND WILSHIRE: METASOMATISM IN THE MANTLE

to a mantle process, Navon and Stolper (1987) definedcompatible elements as having high K. relative to peri-dotite and low X, values; for a specific melt compositionand volume ratio reacting in the mantle column, a lowvalue of X, is due to the high Kd.

The mantle chromatographic column discussed by Na-von and Stolper (1987) is peridotite depleted in light rare-earth elements (REE) relative to chondrites (LREE"" of0.008-0.03), which resembles relatively refractory com-positions of harzburgite or lherzolite that are widely be-lieved to represent samples of mantle lithosphere. Liq-uids (for example, silicate melts produced in theasthenosphere or volatile fluids from melting of a sub-ducted slab) might be enriched in LREE relative to chon-drites, like the composition postulated by Navon andStolper (1987). Theoretical values of Ku and X, used byNavon and Stolper predict that relatively compatiblemiddle or heavy REE (HREE) will be retained in thematrix to a greater degree than incompatible LREE, andthey stated, "After small quantities of melt pass throughthe column, the fronts of the more incompatible, lightREE reach the top" of the mantle column ahead of themore compatible REE and "the matrix in this region de-velops U-shaped REEr""r patterns. . . ." Thus, relative toother REE, the LREE should increase in the infiltratingmelt.

Laboratory processes

The most common laboratory methods used to sepa-rate a suite ofions or other species introduce the species

to the column in a small volume of reactive solvent. Theinitial reactions separate species based on Kd, as discussedabove (sorption step, cf. Samuelson, 1953). Large vol-umes of a different liquid (elutriant), commonly an acidof a normality different from the original solution andthat contains none ofthe solute to be separated, is usedto flush the column. Species are removed in an effiuentthat emerges at the end of the column. Species of low Kuhave a low potential to exchange with the matrix andmove quickly through the matrix, whereas solutes of pro-gressively higher Ku move at progressively slower ratesthrough the column. Each may be collected separately insequential aliquots ofeffiuent, given large enough differ-ences of Ko values, large volumes of elutriant, and a longenough column (cf. Helferrich,1962; Ritchie, 1964).

The frontal or breakout method (Samuelson, 1953;Helferrich, 1962;Fig. la), in which the same solvent in-troduces and carries the species, more closely resemblesthe mantle chromatographic process hypothesized by Na-von and Stolper (1987). For small volumes of solvent,the compatible species are extracted in the matrix, andthe most highly incompatible species is effectively sepa-rated into the first aliquots of efluent at the end of thecolumn (shown as the top of the column in Fig. la). Thecompatible species accumulate quickly in the matrix iflarger volumes of solution enter the column. As noted byNavon and Stolper (1987; extract quoted above), the re-acting matrix reaches equilibrium with the solution andloses the potential to exchange compatible elements, whichtherefore move farther through the column. If enough

ll22 NIELSON AND WILSHIRE: METASOMATISM IN THE MANTLE

il. Diagram of compositions and relations in Lherz boulder

la Ce Nd SmEu Tb Yb Lu

Fig. 2. Diagram and data for the Lherz boulder, basis oftheBodinier et al. (1990) metasomatic models. (a) Diagram of theLherz talus block adapted from Figs. I and 8 ofBodinier et al.(1990, p. 599, 610). Distance from the dike contact plotted onabscissa corresponds to sample numbers. Except for the Yb con-tent of sample I 2, values of CelYb". are recalculated from tablesusing chondrite normalization factors of Anders and Ebihara(1982). (A Yb"" value for sample 12 was estimated from Fig. 6ofBodinier et al., because the tabulated value produces CelYb""10x greater than shown by Bodinier et al., in their Fig. 8.) O)Assumed REE- pattern of melt that reacted with harzburgitebetween 0 and 21 cm from the dike contact. This is the theo-retical composition of melt in equilibrium with amphiboles of

Amphibole-free,granular harzburgite

b. Model of melt composition

laCe NdSmEu Tb YbLu

REE pattems in cryptic zone

amphibole-bearing dikes from Lherz and FreychinEde. (c) RE[,pattems of Lherz subsamples, showing dike composition andrange of chondrite-normalized patterns for samples in the am-phibole-bearing zone. The U-shaped pattem of amphibole-bear-ing sample 2l is not ascribed to chromatographic fractionationby percolation but is very close to that ofsample 25 (not shown)in the amphibole-free zone (Bodinier et al., 1990, their Fig. l).(d) REE", patterns of subsamples in amphibole-free harzburgite.Dotted horizontal lines show correspondence between compo-sitional ranges of samples in amphibole-bearing and amphibole-free zones. Compared with amphibole-bearing harzburgite, thesesample patterns have slightly enriched LREE-, slightly depletedHREE-, and marked depletion of some middle REE.

Amphibolepyroxenite- -

dike

Amphibole.bearing,sheared harzburgite

1 6

1 2 l .

q)

u l-/

c. REE patterns of dike and peridotitein frrodallv metasomatiZed zone

c,

E r m)..<U

. q l

'to

d.'tr

U

(d(t)

Centimeters (from dike)

solution is introduced into the column, all the matrixloses the potential to exchange compatible elements, andthe liquid then passes through unchanged.

The goal of laboratory processes is quantitative sepa-ration of species. Therefore, a real laboratory exchangecolumn ideally contains no ions in common with the so-lution. Moreover, compared with the compatible species,only very small amounts of incompatible species carriedin the liquid are likely to remain in the matrix. By con-trast, in the mantle process of elemental separation pos-tulated by Navon and Stolper (1987), matrix and meta-somatizing melt or fluid contain common ions (Fig. lb);also, the patterns created by preferential fractionation ofLREE into liquid during percolation must be recorded inmatrix. According to Navon and Stolper (1987) theevolving composition of percolating melts appears in ma-trix at the top of the column (implicitly a fixed distancefrom the source); transient incompatible-element frac-tionation is recorded early in the process. Fractionatedmelt compositions are progressively replaced by ones thattend toward and ultimately achieve equilibrium with theoriginal melt composition (Fig. la, lb; Navon and Stol-per, 1987, their Fig. 4). The actual compositions pro-duced in liquid and recorded by the matrix may be nearlyany shape; the critical controls are the original composi-tions and relative volumes of matrix and melt. the actualpartition coefficients at the conditions of reaction, andthe actual process ofexchange.

Important questions that are unanswered by Navon andStolper (1987) include: What region of the mantle con-stitutes the top of a reacting column? Why does the mi-grating melt react in this region if the matrix is the sameas that within the column? Why does the matrix patternso accurately reproduce that in the melt if Kl values re-main constant? What is the likely proportion of melt rel-ative to matrix that can percolate far enough to producethe transient fractionation patterns?

Lherz massif. France

Structural, petrologic, and geochemical feafures. Mod-els of mantle reactions were formulated by Bodinier et al.(1990) from systematic sampling of a harzburgite talusboulder. The boulder (Fig. 2a) contains an amphibole py-roxenite dike (vein, cf. Bodinier et al., 1990), and twobranching "amphibole veinlets" occur within -25 cm ofthe dike contact. Harzburgite within 25 cm of the dikecontains amphibole, and the remainder lacks amphibole.The amphibole pyroxenite dike is garnet-bearing, and pe-ridotite across the entire sampling traverse contains ir-regularly distributed secondary spinel and minor apatite(Woodland et al., 1992).

Mineral compositions are variable in the zone of am-phibole-bearing harzburgite; interstitial amphibole andminor phlogopite are relatively poorer in Ti, A1, and Fe,compared with the same minerals in the dike and vein-lets. The compositions of clinopyroxene, orthopyroxene,olivine, and spinel close to amphibole veinlets show re-

1123

action gradients of major and minor elements (Bodinieret al., 1990, their Fig. 3) such as are observed at contactsbetween peridotite and mafic dikes in composite xeno-liths (Irving, 1980; Wilshire et al., 1980; Kempton et al.,1984; Wilshire et al., 1988).

REE"" patterns of whole rock samples within - 15 cmof the dike have relatively unfractionated shapes thatclosely resemble that of the dike, but at slightly lowerLREE/HREE." (Fig. 2c); however, even assuming that theYb content of sample l2 is 0. 104 ppm (not 0.014 as tab-ulated) the CelYb"", Sm/Yb",, and Tb/Yb." ratios forsamples at 3 and 12 cm are 1.5-2 times those of othersamples in the amphibole-bearing zone (Fig. 2a). Begin-ning in the amphibole-bearing harzburgite at -16 cmfrom the dike contact and extending througlr the amphi-bole-free harzburgite, whole rock samples have U-shapedREE patterns, and samples 18, 21, 22 all have valuescomparable with those of samples 3 and 12 (Fig.2a,2c).Samples >22 cm from the dike contact, in the amphi-bole-free zone, have REQ, patterns with negative slopesand progressively higher CelYb"n values (Fig. 2a,2d).Therelative LREE-enrichments reach a maximum of 16.4times chondrite at 49 cm and then decrease to less thaneight times chondrite in the sample at 65 cm (Fig. 2a).The progression of LREE enrichment in the Lherz boul-der shows the sort of effect postulated by Navon andStolper (1987), and thus provides a test ofthat particu-lar hypothesis of chromatographic fractionation in themantle.

Lherz models. Bodinier et al. (1990) explained thecompositional variations of the Lherz sample as relatedto a single metasomatic event. They proposed that a meltinfiltration process deposited amphibole in harzburgitebetween the dike contact and samples aI -21 cm withoutmarked enrichment fronts of incompatible-elements inthe peridotite. They ascribed REE"" patterns in amphi-bole-free harzburgite to chromatographic fractionation ofmelt having the same composition as the liquid that pro-

duced the modal metasomatism. The fractionated meltwas buffered by passage through the amphibole harzburg-ite. Bodinier et al. (1990) questioned whether the metaso-matizing agents of the two processes are from the sameor different sources but concluded that they representevolution of infiltrating melt derived from the same source(the dike). Vasseur et al. (1991) also indicated that thetwo pro@sses represent "a variation from bulk REE en-richment close to the vein, to selective LREE enrichmentfuither away from it," thus ascribing both effects to thesame metasomatic source.

Modal rnetasomatism: Diffusion controlled model. Thediffi.rsion-controlled model of modal metasomatism of theLherz boulder proposed by Bodinier et al. (1990) requiresinfiltration of parent magma from the amphibole pyrox-

enite dike into adjacent wall rock (Fig. 3). The compo-sition of infiltrating melt (Fig. 2b) was constructed byBodinier et al. (1987) as a model of magmas (likened to

alkali basalt) considered to be parental to the amphibole


tt24

[-75 ouerpyroxenite

[_-l Amphibole py'oxeniteAmphiboleveins and

segreganons

Fig. 3. Diagrams illustrating ditrering assumptions of meta-somatic models. (a) Lherz model of Bodinier et al. (l 990, adapt-ed from their Fig. 9). Thick open arrows indicate supposed di-rection of dike htrusion, and thick black arrows show movemenrof melt diffusing into wall rock peridotite from dikes. Thin blackarrows show percolation direction ofalkali basalt melt inferredto have produced cryptic metasomatism. Rectangle indrcates areaof the sampled boulder. (b) Dish Hill model of Nielson et al.(1993; see Fig. 5). Relations ofperidotite wall rocks, older py-roxenite dikes, amphibole pyroxenite dikes, and related amphi-bole veins, based on field relations observed at Lherz (Fig. a)and on relations in xenoliths. Open arrow pointing at box indi-cates one of many possible positions of the Dish Hill xenolithin the original context.

pyroxenite dikes at Lherz and Freychindde. The calculat-ed REE contents of such parent magmas are based on aliquid in equilibnum with amphibole in the dike. Am-phibole veinlets are considered to provide evidence ofmelt circulation through the wall rock, and the crystalli-zation of amphibole is thought to result from "interactionof the alkali basalt and peridotite" (Bodinier et al., I 990),consuming mainly clinopyroxene and spinel from wallrock. Steep compositional gradients of major elementsadjacent to contacts with the dike are interpreted as re-sulting from diffusion between magma in the dike con-duit and the infiltrating melt.


b Our field and petrographic observations at Lherz (Wil-shire et al., 1988; Nielson et al., in preparation) do notsupport this reaction sequence of magmas with perido-tite. The relations between amphibole pyroxenite dikesand amphibole veinlets strongly indicate that amphibolewas the last major phase to crystallize from amphibolepyroxenite dikes and that many amphibole veinlets orig-inated as late segregations that emerged from the dikes.Intrusions that formed amphibole pyroxenite dikes firstreacted with peridotite to form thin borders of amphi-bole-free pyroxenite, contrary to the inference ofBodinieret al. (1990; Fig. 3a). Within the reaction borders, frac-tional crystallization appears to have concentrated theamphibole component into late crystallizing liquids, whereit forms segregations (Figs. 3b, 4a,4b) and, in some dikes,interstilial poikilitic grains.

Some exposures at Lherz show that an amphibole-richresidual liquid infiltrated peridotite wall rock for a fewcentimeters, creating zones of amphibole peridotite suchas those described by Bodinier et al. (1990) (Fig. 4a), butmany other outcrops or talus boulders show dikes withinternal amphibole segregations or dikes with irregularamphibole-rich zones that meander from axis to margin.Locally, residual liquid was injected into fractures withinor external to the dike in which the liquid evolved (Fig.4a, 4b). The residual melts injected into wall rock alsoreacted with and infiltrated the peridotite (lVilshire, 1984,1987; Wilshire er al., 1988; Mukasa et al., 1991; Nielsonet al., 1993, in preparation). Similar relations are ob-served in xenoliths (Fig. 4c).

The unmetasomatized composition of the Lherz boul-der is not known (the model composition is highly LREE-depleted) (Bodinier et al., 1990, their Fig. l2b), but whole-rock REE patterns of amphibole peridotite in the zone ofdifusion metasomatism reflect those of the dike (Bodi-nier et al., 1990, their Fig. 6). The REE compositionscould be explained by the mixing of depleted peridotiteand melt of the dike composition; therefore, we see nocompelling reason for construction of a parent magmacomposition to represent the infiltrating liquid. We con-tend that actual compositions of the amphibole pyroxe-nite dike in the Lherz boulder or residual liquids fromcrystallization of the dike, represented by amphibole seg-regations that intruded peridotite as smaller veinlets (Fig.3b; cf. Nielson et al., 1993), are plausible sources andcompositions of the metasomatic melts.

Data presented by Bodinier et al. (1990) further sug-gested that the Lherz boulder contains more than onesource of metasomatizing melt; for example, mineralcompositions show gradients in the vicinity of the vein-lets (Bodinier et al., 1990, rheir Fig. 3). According to themodel calculations (Bodinier et al., 1990, their Fig. I l),a sample I cm from the amphibole pyroxenite dike shouldhave the highest overall REQ" and LREE

", and values

should decline systematically with progressively greaterdistance from the contact. To the contrary, the sample at12 cm from the pyroxenite dike contact (Fig. 2a,2c) hasthe highest REE"", LREE.", and CelYb"" values reported

l - ' l l Lherzol i te

F.i=I Harzburgite

Fill 4mph16s1spe'ridotite

Fig. 4. Photographs ofrelations between dike and wall-rockperidotite. (a) Amphibole pyroxenite dikes in peridotite, Lherzmassif. Some of the amphibole pyroxenite dikes have thin am-phibole-free pyroxenite reaction rims and irregularly distributedamphibole-rich segregations (dark bands). (b) Amphibole pyrox-enite dikes, Lherz massif. Left side of photo, from top to bottom,amphibole-rich band (dark) originates in amphibole pyroxenite,

tt25

moves to margin, then bifurcates, with one branch emergingfrom pyroxenite and cutting across peridotite. Wider dike toright contains discontinuous parallel hornblendites at left contactand within the amphibole pyroxenite. Left of hand lens, widerdike crosscuts steeply, dipping (to right) chromium diopsidewebsterite band; below hand lens, contacts ofwebsterite parallelfoliation, which is superimposed on the amphibole pyroxenite.These relations indicate that the amphibole grains in peridotiteare not formed by reaction between dike parent magma andperidotite. (c) Xenolith from San Carlos, Arizona, showing thinamphibole-rich veinlet emerging from parent amphibole pyrox-enite dike (nearly horizontal, lower part of photo) and cuttingperidotite. (d) Photograph of anastomozing homblendites, show-ing opposing directions of branching. Lherz massif.


, - 4

tri:'"

t126

from the zone of amphibole-bearing harzburgite (CelYb""1.6; the value for sample I is 0.90; Bodinier et al., 1990).Samples 3 and 18 also do not fit the model; sample 3 hasthe highest LREE of any sample in the amphibole-bear-ing zone, and sample 18 has a value of CelYb", identicalto that of sample 3. We conclude, therefore, that multiplesources of metasomatic fluids are present in the boulder,and their presence probably determined the width of theamphibole-bearing zone.

Cryptic metasomatism: Percolation controlled model.The cryptic metasomatic model devised by Bodinier etal. (1990) to explain REE patterns in the amphibole-freeperidotite zone of the Lherz boulder "involves the evo-lution of a magma buffered by an amphibole peridotitemineralogy, probably as a result of interaction betweenthe infiltrated alkali basalt and the peridotite matrix inthe modally metasomatized wall rocks." LREE "escapethe buffering effect, because of their strongly incompatiblecharacter" and thus become enriched in the melt duringpercolation. Contrary to Navon and Stolper (1987), themelts do not begin to fractionate until they percolate > Im from the source (Bodinier et al., 1990; their Fig. 12b).This distance seems to establish an upper dimension forthe width of reaction zones at dike contacts, which areascribed to diffusion-controlled reactions (see above).Taken together, these statements imply that the meltswere buffered by the contact zones.

Various aspects of this model are inconsistent with ex-perimental data and other considerations. The uppertemperature limit of amphibole stability (pargasite andkaersutite - 1050-l I l0'C; Millhollen et al., 1974; Mer-rill and Wyllie, 1975) at the emplacement pressure of l0-l5 kbar estimated by Bodinier et al. (1990) is too low topermit crystallization of amphibole at the near-solidusperidotite temperatures needed to propagate infiltrationby this scheme. Therefore, the initial formation of anamphibole peridotite modal metasomatic zone needed tobutrer additional infiltrating melts from the dike appearsunlikely.

Other problems with the model are exemplified in Fig-ure l2 of Bodinier et al. (1990), which shows the modelanomalies plotted against the length of the percolationpath for a range of values dependent on melt fraction,elapsed time, grain size ratios, and diffirsive coefficients.The chromatographic calculation produces composition-al changes that resemble results of Navon and Stolper(1987, their Fig. 4), but the sequence of reactions is re-versed (Bodinier et al., 1990, their Figs. l2b and l3). Thecalculations of Navon and Stolper (1987) showed pro-gtessive compositional changes over time in a matrix thatis located a fixed distance from the source of melt (top ofthe column), and this matrix region progresses towardequilibrium with the original melt. In contrast, Bodinieret al. (1990) showed progressive changes in the degree ofREE fractionation, which are recorded in peridotite vol-umes at progressively more distant locations from thesource of melt. High CelYb." values occur because themelts retain LREE and are depleted, first in HREE and


then in middle REE, at progressively larger distances alongthe percolation path. After 50 m of percolation the HREEcontents merge into those of the assumed premetaso-matic harzburgite (the actual composition is unknown).When the effect of melt volumes is accounted for, thedegree of middle REE depletion increases for greater per-colation distances. CelYb." values comparable with thoseof sample 49 are produced after percolation of about100 m.

The time factors of this model are very large. For thepercolating melt, Bodinier et al. (1990) defined a param-eter /" as "the time it takes for the melt to pass throughoutthe distance of percolation." Assuming reasonable ratesof percolation (u), they state that no CelYb." anomaly iscalculated for distances (z) that require <100 yr ofper-colation (z : t"u) (p. 619). Between 5000 and 25000 yrare required to develop the small degrees offractionation(CelYb"" between 8 and 16) that occur over distances of< I km, according to the calculation. Refinements to themodel by Vasseur et al. (1991) show evolution of thepatterns extending over 50-2000 yr, much smaller per-colation distances (3 m), and lower degrees of HREE de-pletion.

The expectation that melt can be supplied through nar-row mantle conduits for periods longer than a few weeksis not supported either by considerations ofduration andperiodicity of mantle melt extraction events (Nicolas,I 989), or ofheat loss (Spera, I 987). Moreover, at the hightemperatures required to propagate the infiltration, meltin conduits as wide as I m might disperse into wall rocksin 100 yr (Watson, 1982). Since the Lherz massif containsa variety of older pyroxenite dikes (Figs. 3b,4a), it seemsevident that ambient temperatures were well below theperidotite solidus when the amphibole pyroxenites wereemplaced.

To prevent small veinlets and infiltrating melt fromrapidly freezing, the younger dikes would have to heatadjacent peridotite over the distances of measured anom-alies and postulatod scales of infiltration, to temperaturesat or above the liquidus of the supposed parental alkalibasalt (about 1275-1300 "C; e.9., Green and Ringwood,1967;Takahashi, 1980). Liquidus temperatures of alkalibasalt are slightly above the dry peridotite solidus andwell above the wet peridotite solidus (Wyllie, 1979). Ev-idence of partial melting in the peridotites and especiallyin older pyroxenites crosscut by amphibole pyroxeniteshould be found if the hypothesis is correct. At lowertemperatures, interstitial crystallization of liquidus andnear-liquidus phases, or reaction products of the parentmagma, ought to be evident in the amphibole-free harz-burgite, but only green spinel and apatite, products ofresidual liquids, have been identified so far (Woodlandet al.. 1992).

In Figure 3, the model relations assumed by Bodinieret al. (1990) are compared with observed field relationsamong amphibole pyroxenite, derivative hornblendites,older dikes, and peridotite atLherz, or reconstructed fromxenoliths. The model suggests that the direction of veinlet

branching indicates the upward direction of infiltrationfor intrusions and melts in the mantle, if the infiltrationwas driven by melt buoyancy, as seems implied. Figure3b schematically illustrates that dike branching may showopposing directions in a single outcrop, an observationdocumented in Figure 4d; therefore, the direction ofbranching is not a criterion for the absolute direction ofmagma movement in the plane of a tabular intrusion.

Xenolith from Dish Hill. California

Structural, petrologic, and geochemical features. Mod-els of a mantle process resembling those in ion exchangecolumns for purifying liquids, such as HrO softeners (Fig.lc; Fletcher and Hofmann, 1974), were developed usingdata from a well-characterized amphibole-bearing spinellherzolite xenolith from Dish Hill, California. The lher-zolite xenolith is 17 cm long and has a thin (l cm orthinner) hornblendite selvage (dike remnant) at one end(Wilshire et al., 1980, 1988; Nielson and Noller, 1987Nielson et al., 1993; McGuire et al., l99l). Progressivechanges in modal abundance and major-minor elementcomposition of the interstitial amphibole grains and mapsof fluid and solid inclusion abundances (Noller, 1986)show that interstitial amphibole grains and the inclusionsare secondary and related to the selvage.

Steep compositional gradients of magmaphile major,minor, and trace elements, including LREE and somehigh field-strength elements (HFSE) such as Ta, formedin the rock a few centimeters from the contact betweenthe hornblendite selvage and lherzolite. Within the mostaltered zone, HREE and some other compatible elementsalso are enriched, but to a lesser extent than elementsthat are commonly considered incompatible. Isotopiccompositions also vary with respect to the contact (Fig.5), such that the narrow (1.5-2 cm wide) contact zone ofperidotite is close to isotopic equilibrium with the selvage(Nielson et al., 1993). Enrichment fronts of the variouselements (Nielson etal., 1993, their Fig. l0) are slightlyseparated in peridotite within 4 cm of the dike selvage.These compositional variations are mostly due to reac-tion between clinopyroxene grains of the host rock andmetasomatic melt from the dike, and wall rock relationsaccord more with the model of Bodinier et al. (1990) thanthe chromatographic fractionation scheme of Navon andStolper (1987).

Dish Hill xenolith model. The reaction gradients in theDish Hill xenolith can be modeled fairly well by simplefinite-plate calculations (Fig. 6; Helferrich, 1962; Farmerand DePaolo, 1987), using end-member compositions inthe sample. Amphibole in the selvage was used for themodel liquid and refractory peridotite farthest from theselvage was presumed close to the unreacted compositionof peridotite. The calculations employed by Nielson et al.(1993) are a simplification of Navon and Stolper's ap-proach and agree with the concept that mantle reactionsmay be modeled by chromatographic processes. In con-trast to tho hypothesis of Navon and Stolper (1987), mag-maphile elements in the Dish Hill sample including Rb,

t t 2 7

Distance (cm)

Fig. 5. A model of metasomatism for Dish Hill, CA xenolithBa-2- I (Nielson et al., I 993) based on reaction fronts ofNd- andSr-isotopic ratios in lherzolite near contact with amphibole sel-vage (at 0). Data from five subsamples, from next to the contactto a distance of about 17 cm (rock compositions are shovrn bydashes and open circle symbols, clinopyroxenes by dotted lines,solid circles). Calculation ofrock compositional variations (afterFarmer and DePaolo, 1987) used mineral compositions and best-fit partition coefficients and assumed that the composition ofinfiltrating melt was that of dike. Concentration profiles calcu-lated with a pure infiltration model (solid lines) neglected theeffect ofdiffirsion but are remarkably close to the position, shape,and compositional values of actual reaction fronts.

Ba, LREE, HrO, and CO' (normally considered incom-patible relative to peridotite) are enriched in peridotiteclose to the metasomatic source, as are compatible con-stituents of the melt. A small level of LREE enrichmentis also seen in the peridotite subsample between 2 and 3cm from the dike, and a slight U-shape is developed inthat pattern. Therefore, although impoverished in LREE(as hypothesizedby Bodinier et al., 1990) liquid contin-ued to deposit LREE in the depleted peridotite as it movedfarther from the contact zone. The decreased ratio ofLREE/HREE"" in this reacted liquid is reflected in thedecreased ratios recorded in the matrix at larger distancesfrom the dike (Nielson et al., 1993, their Fig. 4). There-fore, the liquid became impoverished in incompatible el-ements with distance of percolation away from the dikecontact zone, contrary to the assumptions ofall percola-tive models (Navon and Stolper, 1987; Bodinier et al.,1990: and Harte et al.. 1993).

NIELSONAND WIISHIRE: METASOMATISM IN THE MANTLE

I 128

a" Sampling hav€rse

c Model compositions

:;sffitrFig. 6. Rock types and REE

" variations in clinopyroxenes

of the Horoman massif, compared with percolation models(combined Figs. 3 and 4 of Takazawa et al., 1992). Figs. rear-ranged for direct comparison of rock types, traverse distance,and sample REE"" variations. (a) Rock type of sample, and dis-tance on traverse, anchored by olivine gabbro (composition notgiven). Numbers are those samples given by Takazawa et al.(1992). (b) REE"" compositions related to traverse position ofsamples. All compositional plots are at the same scale; labels ofordinate scale are removed for ease of reading-all are ratios ofsample to chondrite. (c) Models of REE"" patterns produced byinteraction of peridotite varieties and melts from 15 and,25o/omelting, at 100-m percolation distance; models for I km wereomitted because they are much less successful in matching pat-terns of the samples. The model plots have been changed to thesame scale as all the sample plots.

REE patterns of the dike and metasomatized peridotitein the Dish Hill xenolith are similar to those of the Lherzboulder. The 1.5 cm of wall rock closest to the contact inthe Dish Hill sample can be succossfully modeled by sim-ple mixing of end-member dike with unreacted peridotite(Budahn, 199 I ; Nielson et al., 1993). Convergence of Nd-isotopic values in rocks and clinopyroxenes (Fig. 5) showsthat exchange of isotopic species between melt and cli-nopyroxenes can account for the metasomatic whole-rock


b. C-ompositionsofcpx stparahs

composition, without crystallization interstitial amphi-boles. In contrast, crystallization of new minerals (am-phibole and solid inclusions) must account for higherSr-isotopic compositions in the reacted lherzolite (cf.Nielson et al., 1993). A model of the reaction based onIsocon calculations ofGrant (1986) suggest that equilib-rium between the melt and depleted wall-rock peridotitemay have required interaction between equivalent vol-umes of LREE-enriched liquid and rock in the 1.5-cm-thick contact zone (Nielson et al., 1993). The data suggestshort time factors for these interactions.

Horoman massif, Japan

Structural, petrologic, and geochemical features. TheHoroman massif crops out in an area of about 800 km,.The rocks are well exposed along drainages, but substan-tial differences between maps of the complex (Niida, 1974;Obata and Nagahara, 1987; Takahashi, l99l; Takazawaet al., 1992) suggest that intervening areas are poorly ex-posed. The massif consists mostly of alternating bands ofplagioclase lherzolite, lherzolite, and harzburgite that varyfrom a few meters to several hundred meters in thickness.Within the peridotite are gabbro dikes and phlogopite-rich veins (Takahashi, l99l1,Takazawa et al., 1992). Allperidotite variants (including dunite) have porphyroclas-tic and locally mylonitic textures. Zones of plagioclase-free lherzolite separate bands ofplagioclase lherzolite andharzburgite, and parallel layers of dunite as much as 15m thick occur locally in the harzburgite (Takahashi, 1992;Takazawa et al., 1992).

Magmaphile minor and trace elements (bulk perido-tite-melt coemcients >l: Cr, Mn, Fe, Co, Ni, andZn)show little variation between lherzolite varieties andharzburgite (Frey et al., 1991). Larger variations are shownby Ti, Sc, V, and Ga; plagioclase lherzolite has the highestand harzburgites have the lowest contents of these ele-ments. Most whole rocks and clinopyroxenes show con-gruent REE patterns that are depleted in LREE- and haverelatively flat patterns for all middle and HREE. Excep-tions are two harzburgite samples with relatively flatwhole-rock patterns and clinopyroxenes enriched inLREE", (Frey et al., 1991). Takazawa et al. (1992) rc-ported variations of REE in clinopyroxene from a sam-pling traverse across adjacent layers ofthe main perido-tite series, located by reference to a gabbro band (Fig. 5).The data show a variety of REE", patterns (Fig. 5b).

The bulk compositions of Horoman samples from a100-m traverse (Fig. 6) all have values of La." and Ce."that are lower than chondrites. Three of the lherzolitesamples (two plagioclase lherzolites) have chondrite-nor-malized values of l-2 for the REE from Nd through Tb(Frey et al., l99l), and two harzburgite samples haveLa/Sm"" values > L0, although LREE

" contents are 0.I (Frey

et al., l99l). REE"" patterns of clinopyroxene grains inthese rocks are variously LREE".-depleted, slightly en-riched (LalSmc" as much as 5), or have relatively hrghREE contents (Nd"" as much as 13, Sm"" of 16) and rel-

BZ:IQ

Lhflolib

atively unfractionated REE"" patterns. Some of the LREE-enriched patterns are U-shaped, so that LREE/HREE""values do not reflect the level of LREE and middle REEenrichment. Frey et al. (1991) concluded that the pres-ence of LREE enrichments is not consistent with meltextraction by which the main rock types formed andtherefore postulated a later enrichment event.

Iloroman model. The lherzolite and harzburgite bandsare interpreted as restites from partial melting of plagio-clase lherzolite protolith by Takahashi (199 1, 1992), F reyet al. (1991), and Takazawaetal. (1992). Takahashi (1992)also interpreted the dunite zones interlayered with harz-burgite, lherzolite, and plagioclase lherzolite to be a cu-mulate of melts related to the formation of restites. Araiand Takahashi (1989) interpreted phlogopite-rich veinsand interstitial phlogopite as having crystallized from flu-ids released by evolving alkali basaltic magmas of un-specified origin. Takahashi (1992) further presented evi-dence that the gabbros formed as intrusions and that theymetasomatized host dunite and harzburyite, locally de-positing titanium pargasite, kaersutite, titanite, and il-menite.

Because the REE patterns change abruptly in the rockssampled, Takazawaetal. (1992) modeled the varied REEpatterns of Horoman peridotite samples as being causedby chromatographic fractionation in a melt during per-colation. The melt may or may not be related to a gabbroat one end of the transect. Following Navon and Stolper(1987) the model calculation assumes that changing frac-tionation patterns are recorded in matrix a fixed distance(100 m and I km) from the source (Takazawaetal., 1992).In modeling the progression of REE patterns in the peri-dotites, the authors postulated that harzburgite and lher-zolites had different premetasomatic REE contents, dueto the different degrees of partial melting (25 and l5o/o,respectively) that is thought to have created the rock types(Fig. 6c; only the 100-m model is shown). Residual cli-nopyroxene in harzburgite and lherzolite would have dif-ferent metasomatic compositions after reaction with thesame melt composition.

Taking these factors into account, the authors recog-nize that their calculations do not produce a simultaneousfit among pattern shape, level of enrichment, or perco-lation distance or time, if the percolation direction is par-allel to the traverse (Takazawa et al., 1992). Close to thegabbro, patterns in harzburgite show a progressive en-richment in LREE with distance from the contact (sam-ples at 2.4-14 m; Fig. 6), and these compositional changescorrespond to those modeled for a sample 100 m fromthe source at progressively greater percolation times(sample at 2.4 m resembles the model of 25o/o melt resi-due metasomalized at l,; sample at 14 m resembles thesame residue metasomatized slightly later; Fig. 6c). Sim-ilarly, the lherzolite sample at 57 m fits the model of aresidue from 150/o melting, metasomatized at a time clos-er to tz, and the pattern of a clinopyroxene core at 23 mrepresents the pattern for a stage between that time andl, (Fig. 6). However, the rim of that clinopyroxene has a

tt29

deep U-shaped REE pattern with higher HREE than thecore, which resembles the model for 1,.

According to the Navon and Stolper (1987) hypothesis,these fractionation patterns are transitory; patterns rep-resenting metasomatism by the earliest melts reachingthe top of a mantle column should have been changed byinteraction with melt aliquots that came later. In addi-tion, the plagioclase lherzolite sample at 85 m shouldrepresent matrix that has reached equilibrium with theoriginal melt composition; however, evidence of such anequilibrated peridotite column is not represented on thesampling traverse. Thus, a diferent trajectory of perco-lation is indicated, but the data do not support assess-ment of this possibility. The scheme of Bodinier et al.(1990), with some of the refinements of Vasseur et al.(1991), could explain variations between 99 and 57 m,but only ifa nearby source could be identified.

The lack of congnrence between data and the Horomanmodel, especially the HREE zonation in clinopyroxenegrains, is of concern, and Takazawaetal. (1992) admittedthat they need a more complex model. Also, more recentdata for the gabbro layer in the 100-m section sampledindicate that it is not the source of the metasomatic fluids(F. A. Frey, personal communication). Thus, more dataare needed to establish better constraints on the compo-sition of the metasomatic agents, the directions and ve-locity of flow, and permeability.

DrscussroNMantle samples and process models

The models reviewed here share the assumption thatreactions between mantle melts and peridotite are chro-matographic processes, but they conflict in the choice ofchromatographic processes that may emulate the reac-tions. Some models of elemental fractionation due to melt-peridotite interaction are based on the undocumented hy-pothesis that large melt volumes are transported throughthe upper mantle by pervasive, regional-scale percolation(cf. Navon and Stolper, 1987). These models contain ad-ditional assumptions that conflict with each other andwith evidence from constrained studies ofxenoliths andmassifs.

The samples chosen for modeling are critical to estab-lishing appropriate constraints. Models based on xeno-liths cannot be tested unless composite samples (with ev-idence of their context) are used. This limitation shouldnot be a problem for massifs, because sample contextsare observable. The ability to observe context makes thestudy of massifs particularly valuable, but it also impliesmajor logistical commitments for adequate testing of hy-potheses and models.

Data from composite samples can provide importantobservations relevant to the debate about mantle pro-cesses. Composite xenoliths contain all the rock types andcontact relations that are observed in massifs. Xenolithsare relatively fresh, having been excavated from the man-tle lithosphere in a matter of hours or days. In contrast,


I 130

peridotite massifs of the Pyrenees and Ivrea Zone havecomplex crustal emplacement and metamorphic histories(e.g., Quick and Sinigoi, 1992). Electron probe traversesacross samples from contact zones between peridotite andpyroxenite or amphibole dikes of massifs show mineralcompositional gradients like those in composite xenoliths(Hofmann et al., 1992; Nielson et al., in preparation), butmany of these data are highly scattered due to partialreequilibration.

Model similarities and differences

The models of reactions in a composite xenolith fromDish Hill are based on observable compositions of un-metasomatized peridotite and metasomatic melt, whichcan be used to test the assumptions of percolative modelsand the hypothesis of Navon and Stolper (1987). TheDish Hill sample shows conclusively that chromato-graphic reactions can produce steep compositional gra-dients and abrupt reversals in REE." patterns (cf. Taka-zawa et al., 1992) over a distance of 17 cm. However,contrary to Navon and Stolper (1987), the most elevatedLREE- contents and fractionated patterns of LREE." vs.HREE

" occur within 2 cm of the dike contact. In thiszone, the originally depleted peridotite is in isotopic equi-librium with the dike composition and has a pattern ofREE." that is similar to, but slightly lower than, that ofthe dike.

The Dish Hill data also raise questions about the ap-plication to mantle processes of the frontal method ofchromatographic separation, which is the basis of Navonand Stolper's (1987) hypothesis. In the xenolith, LREEare deposited in the peridotite along with HREE, ratherthan being preferentially carried in the melt. Beyond thezone of greatest peridotite-melt reaction, the REQ" pat-tern is slightly U-shaped. The cause of this fractionationis unclear, but it can be modeled as selective partitioningof middle REE by interstitial amphibole (Budahn, l99l).If the observation that REE are not preferentially re-tained in percolating liquid can be generally applied tomantle processes, the LREE probably do not fractionateduring percolation.

Metasomatism of harzburgite is observed in the dikecontact zone of the Lherz boulder, in accord with rela-tions in the Dish Hill xenolith. Similar to Nielson et al.(1993), Bodinier et al. (1990) postulated that peridotiteclosest to the source of melt reflects equilibrium with theoriginal melt composition. However, the Dish Hill modelis consistent with a metasomatic melt composition thatis very similar to the intrusive rock present in the sample.The REE patterns of amphibole peridotite in the Lherzboulder suggest that the metasomatic melt had a com-position like that of the amphibole pyroxenite dike, in-stead of the theoretical parent magma (cf. Bodinier et al.,r990).

The Lherz models assume two metasomatic processes:the contact zone is metasomatized by a diffusion-con-trolled process, and, no matter what melt volume is dis-charged into the wall rock, the zone in equilibrium with

NIEI-SON AND WILSHIRE: METASOMATISM IN THE MANTLE

melt apparently remains about I m wide. Beyond I m,the melt compositions are controlled by percolation; thefractionating melt is said to be buffered by the mineralsin an amphibole peridotite, presumably in this contactzone. However, the melt percolates long distances beyondthe diffusion-controlled zone (Bodinier et al., 1990, theirFig. 9), and HREE are progressively removed from themelt as the distance of percolation increases. Therefore,this part of the model does not accord with classical def-initions of buffer systems, that compositions in a systemof interacting liquid and solid remain relatively un-changed for components that are common to, and equil-ibrated with, both parts of the system (Krauskopfi 1967).

Preferred chromatographic model

On the basis ofthe Dish Hill data, the chromatographicprocess that most resembles observed and documentedmantle reactions is similar to those of HrO-purificationcolumns (Fig. lc; Helferrich, 1962; Hofmann, 1972;Fletcher and Hofmann, 1974). In a mantle region, rela-tively depleted peridotite reacts immediately with incom-patible element-enriched melt, and a zone nearest thecontact reaches equilibrium with a proportional volumeof the melt. The liquid that moves beyond this zone hasreduced potential to react further with the matrix (cf. Fig.lc, left side). If additional volumes of the original meltcomposition enter the matrix, they first encounter peri-dotite in equilibrium with the melt composition and passthrough unchanged; in this equilibrated zone the melt isbuffered, sensu stricto.

Upon reaching the reaction boundary and encounteringdepleted peridotite, the buffered melt reacts as before,and the zone of equilibrated matrix moves farther awayfrom the source (Fig. lc, center); the reacted liquid has areduced potential to exchange further, as before. In themantle, this process is limited by the supply of melt be-cause the abundance of depleted peridotite must be verymuch greater than volumes of enriched melts (Ni€lson etal., 1993). However, large melt volumes would produceincreasingly thick zones of equilibrated peridotite, bothby percolation and dike branching (Fig. 3b).

To preserve transitory patterns offractionated REE inmatrix, the hypothesis of Navon and Stolper (1987) re-quires passage of relatively small volumes of melt, be-cause large volumes ultimately bring a matrix into equi-librium with the melt composition (Fig. la, lb). Incontrast, REE fractionation patterns calculated by Bodi-nier et al. (1990) are not transient for any melt volume.According to their model, matrix in equilibrium with meltis found only in the diffilsion-controlled zone, within Im ofthe dike. Longer times of percolation (/.) correspondto larger percolation distances and greater degrees offrac-tionation (Bodinier et al., 1990, their Fig. l3).

Tests of percolation scale in mantle metasomatism

Models based on regional, pervasive porous-mediumflow in the mantle can be qualitatively evaluated in termsofthe likelihood that the necessary conditions exist, but

NIELSON AND WILSHIRE: METASOMATISM IN THE MANTLE I l 3 l

Fig. 7. Photograph showing the common faceted character ofperidotite xenoliths in basalts. The facets are shown to be preen-trainment joints (and shear fractures) by the common presence of intrusions along them that have compositions ditrerent fromthose of the host basalts (Wilshire and Kirby, 1989). From left to right, top: Victoria, Australia; Queensland, Australia; MassifCentral, France; bottom: Cima, California; Hualalai, Hawaii; Jilin Province, China.

essentially they cannot be tested. In contrast, evidence ofhydraulic fracturing is observed in xenolith samplesworldwide, and the common observation of dikes rep-resenting multiple episodes of magmatism in mantlemassifs points to the same conditions in their sourceregions. Particularly the xenolith samples, which do notshare the crustal deformational history of massifs, showthe evidence of intrusion in all their discernible magmatichistory and hydraulic fracturing on centimeter scales intheir most recent, preentrainment history Gig. 7; Wil-shire and Kirby, 1989). Where conditions in the sourceareas of mantle samples are favorable for hydraulic frac-turing, porous flow probably could not successfully com-pete as a mechanism for fluid transport over substantialdistances.

In our opinion, the evidence for metasomatism due toporous-medium flow on scales of tens of meters or kilo-meters has not yet been documented in massifs. Never-theless, the trace element behavior in massifs showingIarge-scale (tens of meters) symmetrical lithologic varia-tions remain to be explained. Current studies of the Lherzand Horoman massifs will add greatly to our knowledge,especially ifhypotheses are adequately tested. From their

recent study of the Lherz boulder, Woodland et al. (1992)concluded that throughout, "melt infiltrated non-perva-sively, migrating along select grain boundaries, and in-teracting with the host to variable degrees." We note alsothat the refinements to the model of Bodinier et al. (1990)by Vasseur et al. (1991) deal with considerably reducedscales (<3 m) and times (<2000 yr), which shifts theboundary conditions of melt percolation closer to thoseobserved in xenolith reactions and field relations. Wetherefore believe that the collection of systematic datafrom both massifs and xenoliths, focusing on all perido-tite varieties and testing the assumptions about likely meltcompositions, will ultimatoly bring convergence to chro-matographic mantle models.

Geochemical studies of Horoman are at an early stageand thus amenable to tests of model assumptions, partic-ularly because the field relations have been closely ex-amined (Niida, 1974; Obata and Nagahara, 1987;Taka-hashi, 1991). In addition to sampling scale, the lack of fitbetween data from Horoman samples and the model couldbe tested by sampling for small-scale metasomatism dueto localized melt conduits. For example, metasomaticagents might have been distributed through a complex of

tt32

now-healed fractures that allowed local infiltration. Allcurrent hypotheses for the origin of the main lithologicvariants alLanzo in the Ivrea Zone (Boudier and Nicolas,1977; Boudier, 1978) and the Trinity massif in California(Quick, l98l), as well as at Horoman, propose an earlydepletion of the protolith by various degrees of partialmelting. These hypotheses all require later healing of con-duits that must have removed the partial melts; therefore,our working hypothesis is plausible. A more complexmodel may be needed for Horoman, as suggested by Tak-azawa eI al. (1992), but it seems more appropriate to firsttest model assumptions where the data conflict, by ana-lyzing flner scale samples. Test of percolation directioncan be made by sampling on traverses that parallel theindividual bands.

CoNcr,usroNs

Mantle rocks provide limited and commonly ambigu-ous evidence for models of metasomatic processes, andthe boundary conditions of those models conflict. How-ever, evidence abounds in massifs and xenolith suites formantle metasomatic processes by short-scale percolationin wall rocks close to single or multiple sources of melts.Most well-studied occurrences display multiple episodesof magmatism. In contrast, melt percolation on scalesgreater than several meters is still speculative.

Well-documented reactions between mantle peridotiteand basaltic melts with enrichment fronts of incompati-ble elements, both in xenoliths (Nielson et al., 1993) andmassifs (Bodinier et al., 1990; Woodland et al., 1992;Hofmann, et al., 1992), show that dike contact zones aremost likely to reach equilibrium with the melt composi-tions, and that mantle metasomatism imprints the com-positions of melts in conduits upon relatively depletedperidotite in these zones.

Studies of samples or sample suites that are carefullyselected for well-defined contact relations and evidenceof single-stage reactions are more likely to produce suc-cessful models of mantle processes than are large-scalestudies of bodies with highly complex histories. We sug-gest that congnrence among the models will come fromsuch well-constrained studies. Refinements of one per-colative model already trend toward smaller time framesand areal effects.

AcxNowr-nncMENTS

We are indebted to Darby Dyar, University of Oregon, Anne McGuire,University of Houston, Adolphe Nicolas, University of Montpellier, Jo-seph Arth, U.S. Geological Survey, and an anonymous reviewer for ex-tremely helpful criticisms and corrections of our misconceptions. Addi-tional clarifications were provided by conespondence with F.A. Frey,Massachusetts Institute ofTechnology, and J.L. Bodinier, University ofMontpellier and discussions with Douglas Smith, University of Texas.

RlrnnnNcns crrnnAnders, E., and Ebihara, M. (1982) Solar-system abundances ofthe ele-

ments. Geochimica et Cosmochimica Acta. 46.2363-2380.Arai, S, and Takahashi, N. (1989) Formation and compositional varia-

tion of phlogopites in the Horoman peridotite complex, Hokkaido,

NIEISON AND WILSHIRE: METASOMATISM IN THE MANTLE

northern Japan: Implications for origin and fractionation of metaso-matic fluids in the upper mantle. Contributions to Mineralogy andPetrology, l0 l , 165-175.

Bodinier, J.L., Fabrids, J., Lorand, J P., Dostal, J., and Dupuy, C. (1987)Geochemistry ofamphibole pyroxenite veins from the Lherz and Frey-chinOde ultrarnafic bodies (Aridge, French Pyrenees). Bulletin de Mi-n6ralogie, I 10, 345-358.

Bodinier, J.L., Vasseur, G., Vemidres, J., Dupuy, C., and Fabribs, J. (1990)Mechanisms of mantle metasomatism: Geochemical evidence from theLherz orogenic peridotite. Journal of Petrology, 31, 597-628.

Boudier, F. (1978) Structure and petrology ofthe Lanzo peridotite massif(Piedmont Alps). Geological Society of America Bulletin, 52, 39-56-

Boudier, F., and Nicolas, A. (1977) Structural controls on partial meltingin the Lanzo peridotites. In H.J.B. Dick, Ed., Magma genesis, p. 63-78. Oregon Department of Geology and Mineral Industries, Bulletin96, Portland, Oregon.

Budahn, J.R. (1991) An empirical equation to model rare earth elementmineral-mineral partitioning: An application to mantle metasomatism.U S. Geological Survey Open-File Report, OF 9l-0596.

Bussod, G.Y., and Christie, J.M. (1991) Textural development and melttopology in spinel lherzolite experimentally deformed at hypersolidusconditions. Journal ofPetrology, Special Lherzolites Issue, l7-39.

Daines, M.J., and Kohlstedt, D.L. (1993) A laboratory study of meltmigration. Philosophical Transactions of the Royal Society of London,Series A. 342.43-52.

Dawson, J.B, and Smith, J.V. (1992) Olivine-mica pyroxenite xenolithsfrom northern Tanzania: Metasomatic products of upper-mantle peri-dotite Journal ofYolcanology and Geothermal Research, 50, l3l-142.

Eggler, D.H. (1987) Solubility of major and trace elements in mantlemetasomatic fluids: Experimental constraints. In M A. Menzies andC.J. Hawkesworth, Eds., Mantle metasomatism, p 2l-42. Academic,New York

Eggler, D.H., and Holloway, J.R. (1977) Partial melting of peridotite inthe presence of H,O and COr: Principles and review. In H.J.B. Dick,Ed., Magma genesis, p. l5-36. State of Oregon Department of Geologyand Mineral Resources Bulletin 96, Portland, Oregon

Farmer, G.L., and DePaolo, D.J. (1987) Nd and Sr isotope study of hy-drothermally altered granite at San Manuel, Arizona: Implications forelement migration paths during the formation ofporphyry copper oredeposits. Economic Geology, 82, ll42-ll5l.

Fletcher, R.C., and Hofmann, A.W. (1974) Simple models of diffusionand combined difihsion-infiltration metasomatism- In A.W. Hofmann,B.J. Giletti, H.S. Yoder, Jr., and R.A. Yund, Eds., Geochemical trans-port and kinetics. Camegie Institution ofWashington Publication, 634,243-2s9

Frey, F.A, Shimizu, N., kinbach, A., Obata, M., and Takazawa, E. (1991)Compositional variations within the lower layered zone of the Horo-man peridotite, Hokkaido, Japan: Constraints on models for melt-solidsegregation Journal of Petrology, Special Lherz olites Issue, 2l l-227 .

Fujii, N., Osamura, K., and Takahashi, E. (1986) Effect ofwater satura-tion on the distribution ofpartial melt in the olivine-pyroxene-plagio-clase system. Journal of Geophysical Research, 91, 9253-9259.

Grant, J.A. (1986) The isocon diagram: A simple solution to Gresens'equation for metasomatic alteration Economic Geology, 81, 1976-t982.

Green, D.H., and Ringwood, A.E. (1967) The genesis of basaltic magmas.Contributions to Mineralogy and Petrology, 15, 103-190.

Harte, B. (1983) Mantle peridotites and processes: The kimberlite sampleIn C J. Hawkesworth and M.J. Norry, Eds., Continental basalts andmantle xenoliths,p.46-91 Shiva, Cheshire, U K.

Harte, B., Hunter, R H , and Kinny, P.D (1993) Melt geometry, move-ment and crystallization in relation to mantle dykes, veins and meta-somatism. Philosophical Transactions of the Royal Society of London,Series A. f42. l-21.

Helferrich, F. (1962) Ion exchange, 624 p New York, McGraw-Hill.Hofmann, A.W. (1972) Chromatographic theory of infiltration metaso-

rnatism and its application to feldspars. American Joumal of Science,272 ,69 -90 .

Hofmann, A.W., Lu, M., Obermiller, W., Rivalenti, G., and Mazzuc-chelli, M. (1992) Contrasting styles ofcrustal contamination in Ivreamagma chambers and peridotites (abs.). In J.E. Quick and Silvano

Sinigoi, Eds., Ivrea-Verbano zone workshop, 1992, p. lO. U S. Geo-logical Survey Circular 1089.

Irving, A.J. (1980) Petrology and geochemistry ofcomposite ultramaficxenoliths in alkalic basalts and implications for magmatic processeswithin the mantle American Journal of Science, 28O-A, 389-426

Kelemen, P B , Dick, H.J.B., and Quick, J.E. (1992) Formation of harz-burgite by pervasive melt/rock reaction in the upper mantle. Nature,3 58. 63s-64 l .

Kempton, P.D, Menzies, M.A., and Dungan M.A. (1984) Petrography,petrology and geochemistry of xenoliths and megacrysts from the Ge-ronimo volcanic field, southeastem Arizona. In J. Kornprobst, Ed.,Kimberlites. II. The mantle and crust-mantle relationships, p 7l-83.Elsevier, New York

Krauskopf, K.B. (1967) Introduction to geochemistry, 721 p. McGrawHill, New York.

Lloyd, F.E, Nixon, P.H., Hornung, G., and Condliffe, E. (1987) RegionalK-metasomatism in the mantle beneath the west branch of the EastAfrican Rifl: Alkali clinopyroxenite xenoliths in higlly potassic mag-mas. In P H. Nixon, Ed., Mantle xenoliths, p. 641-659. Wiley, NewYork.

McGuire, A.V., Dyar, M.D., and Nielson, J.E. (1991) Metasomatic oxi-dation of upper mantle peridotite. Contributions to Mineralogy andPetrology, lO9, 252-264

McKenzie, D.P. (1984) The generation and compaction of partially mol-ten rock. Journal of Petrology, 25,7 l3-765

Merrill, R.B., and Wyllie, P.J. (1975) I(aersutite and kaersutite eclogitefrom Kakanui, New Zealand: Water-excess and water-deficient meltingto 30 kilobars. Geological Society ofAmerica Bulletin, 86, 555-570.

Millhollen, G L., Irving, A.J., and Wyllie, P.J. (1974) Melting interval ofperidotite with 5.7 p€rcent water to 30 kiloban. Journal ofGeology,82, 57 5-587 .

Mukasa, S.B., Wilshire, H.G., Nielson, J.E., and Shervais, J.W. (1991)Intrinsic Nd, Pb and Sr isotopic heterogeneities exhibited by the Lherzalpine peridotite massif, French Pyrenees. Journal ofPetrology, SpecialLherzolites Issue, I l7-134.

Navon, O., and Slolper, E (1987) Geochemical consequences of meltpercolation: The upper mantle as a chromatographic column. Journalof Geology, 95, 285-307 .

Nicolas, A. (1986) A melt extraction model based on structural studies inmantle peridotites. Journal of Petrology, 27 , 999-1022.

-(1989) Structures of ophiolites and dynamics of oceanic litho-sphere Kluwer Academic, Boston, Massachusetts.

-(1990) Melt extraction from mantle peridotites: Hydrofracturingand porous flow, with consequences for oceanic ridge activity. In M.P.Ryan, Ed., Magma transport and storage, p 159-173 Wiley, NewYork.

Nielson, J.E., and Noller, J.S. ( I 987) Processes of mantle metasomatism.In E.M. Morris and J.D. Pasteris, Eds , Mantle metasomatism andalkaline magmatism. Geological Society of America Special Paper, 215,6 t - 76 .

Nielson, J E., Budahn, J.R., Unruh, D., and Wilshire, H G. (1993) Ac-tualistic models of mantle metasomatism. Geochimica et Cosmochim-ica Acta. 57. 105-121.

Niida, K. (1974) Structure of the Horoman ultramafic massif of the Hi-daka metamorphic belt in Hokkaido, Japan. Geological Society ofJa-pan, 80, 3 l -44

Noller, J.S (1986) Solid and fluid inclusions in mantle xenoliths: Ananalyticaf dilemma? Geology, 14, 437 -44O.

Obata, M., and Nagahara, N. (1987) Layering of alpine-type peridotiteand the segregation of partial melt in the upper mantle. Joumal ofGeophysical Research, 92, 3467 -347 4

O'Reilly, S Y., Griffin, W.L., and Ryan, C.G. (1991) Residence of traceelements in metasomatized spinel lherzolite xenoliths: A proton-mi-croprobe study. Contributions 1o Mineralogy and Petrology, 109, 98-I l 3

Phipps-Morgan, J. (1987) Melt segregation beneath mid-ocean spreadingcenters. Geophysical Research Letters, 14, 1238-1241.

Quick, J E (1981) The origin and significance of large, tabular, dunitebodies in the Trinity Peridotite, northern California. Contributions toMineralogy and Petrology, 7 8, 413-422

l 133

Quick, J.E., and Sinigoi, S, Eds. (1992) Ivrea-Verbano zone workshop,1992. U.S. Geological Survey Circular 1089.

Rabinowicz, M., Ceuleneer, G., and Nicolas, A. (1987) Melt segegation

and flow in mantle diapirs below spreading centers: Evidence from the

Oman ophiolite. Journal of Geophysical Research , 92, 347 5-3486.Ribe, N.M. (1985) The generation and composition of partial melts in

the earth's mantle Earth and Planetary Science ktters, 73,361-376.Riley, G.N., Jr., and Kohlstedt, D.L. (1990) An exp€rimental study of

melt migration in an olivine melt system. In M.P. Ryan, Ed., Magma

transport and storage, p- 77-86. Wiley, New York.Ritchie, A.S. (1964) Chromatography in geology, 185 p. Elsevier, New

York.Ryan, M.P. (1988) The mechanics and three-dimensional intemal struc-

ture of active magmatic systems: Kilauea volcano, Hawaii. Journal of

Geophysical Research, 93, 4213-4248.Samuelson, O. (1953) Ion exchanges in analytical chemistry,29l p. Wiley,

New York.Shaw, H R. (1980) The fracture mechanisms of magma transport from

the mantle to the surface. In R B. Hargraves, Ed., Physics of magimaticprocesses, p. 201-264.Pinceton University Press, Princeton, New Jer-sey.

Sleep, N.H. (1988) Tapping ofmelt by veins and dikes. Journal ofGeo-physical Research, 93, 10255-1027 2.

Smith, D., Grifrn, W L., Ryan, C.G., and Sie, S.H' (1991) Trace-elementzonation in garnets from The Thumb: Heating and melt infiltration

below the Colorado Plateau. Contributions to Mineralogy and Petrol-

ogy, 107,60-79.Smith, D., Grifrn, W.L., and Ryan, C.G. (1993) Compositional evolution

of high-temperature sheared lherzolite PHN 161 1. Geochimica et Cos-

mochimica Acta. 57. 605-613.Spera, F.J. (1987) Dynamics of translithospheric migration of metaso-

matic fluid and alkaline magma. In M.A. Menzies and C.J. Hawkes-

worth, Eds., Mantle metasomatism, p. l-20. Academic, London.

Stolper, E., Walker, D., Hager, 8.H., and Hays, J.F. (1981) Melt segre-gation from partially molten source regions: The importance of melt

density and source region size. Journal of Geophysical Research, 86,

6261-627 | .Takahashi, E. (1980) Melting relations of an alkali-olivine basalt to 30

kbar, and their bearing on the origin of alkali basalt magmas' CarnegieInstitution of Washington Year Book, 79, 27 l-276.

Takahashi, N. (1991) Origin of three peridotite suites from Horomanperidotite complex, Hokkaido, Japan: Melting, melt segxegation and

solidification processes in the upper mantle. Journal of Mineralogy,

Petrology, and Economic Geology, 86, 199-215'- (1992) Evidence for melt segregation towards fractures in the Hor-

oman mantle peridotite complex. Nature, 359, 52-55.Takazawa, E., Frey, F.A , Shimizu, N., Obata, M., and Bodinier, J.L'

(1992) Geochemical evidence for melt migration and reaction in the

upper mantle. Nature, 359, 55-58.Toramaru, A., and Fujii, N. (1986) Connectivity of melt phase in a par-

tially molten peridotite. Journal of Geophysical Research,91,9239-9252.

Vasseur, G., Verndrs, J., and Bodinier, J.L (1991) Modelling oftraceelement transfer between mantle melt and heterogranular peridotite

matrix. Journal ofPetrology, Special Lherzolites Issue, 41-54.Waff, H S., and Bulau, J.R. (1982) Experimental studies ofnear-equilib-

rium textures in partially-molten silicates at high pressure. Advances

in Eanh and Planetary Sciences, 12,229-236.Waff, H.S , and Faul, U.H. (1992) Etrects of crystalline anisotropy on fluid

distributjon in ultramafic partial melts. Journal of Geophysical Re-

search.97. 9003-9014.Waff, H.S., and Holdren, G.R., Jr. (1981) The nature of grain boundaries

in dunite and lherzolite xenoliths: Implications for magma transport in

refractory upper mantle material. Journal ofGeophysical Research, 86,

3677-3683.Walker, M E.L, Jurewicz, S.R., and Watson, E B (1988) Adcumulus du-

nite growth in a laboratory thermal gradient. Contributions to Miner-

alogy and Petrology, 99,306-319.Watson, E.B. (1982) Melt infiltration and magma evolution Geology, l0'

236-240.Watson. E.B.. and Brenan, J.M. (1987) Fluids in the lithosphere. I. Ex-


rl34

penmentally-determined wetting characteristics of CO,-HrO fluids andtheir implications for fluid transport, host-rock physical properties, andfluid inclusion formation Earth and Planetary Science Letters,85,497-) l ) -

Watson, E.B , Brenan, J.M., and Baker, D.R. (l990) Distribution of fluidsin the continental mantl€. In M.A. Menzies, Ed., Continental mantle,p. I I l-125. Clarendon Press, Oxford, U.K.

Wilshire, H.G (1984) Mantle metasomatism: The REE story Geology,| 2, 395-398.

-(1987) A model of mantle metasomatism. Geological Society ofAmerica Special Paper, 215,47-60.

Wilshire, H.G., and Kirby, S.H. (1989) Dikes, joints and faults in theupper mantle. Tectonophysics, 16 l, 23-31.

Wilshire, H.G., and Pike, J.E.N. (1975) Upper-mantle diapirism: Evi-dence from analogous features in alpine peridotite and ultramafic in-clusions in basalt. Geology, 3,467-470.

Wilshire, H.G., and Shervais, J.W. (1975) Al-augite and Cr-diopside ul-tramafic xenoliths in basaltic rocks from western United States. physicsand Chemistry ofthe Earth, 9,257-272

Wilshire, H.G., Pike, J.E.N., Meyer, C.E., and Schwarzman, E.C. (1980)Amphibole-rich veins in lherzolite xenoliths, Dish Hill and DeadmanLake, California. American Journal ofScience. 280-4,. 576-593.


Wilshire, H.G., Meyer, CE., Nakata, J.K., Calk, LC., Shervais, J.W.,Nielson, J.E., and Schwarzman, E.C (1988) Mafic and ultramafic xeno-liths from volcanic rocks ofthe western United States U S. GeologicalSuruey Professional Paper, 1143,179 p.

Wilshire, H.G., McGuire, A V., Noller, J.S., and Turrin, B.D. (1991) Pe-trology of lower crustal and upper mantle xenoliths from the Cimavolcanic field, California. Joumal of Petrology, 32, 169-200.

Woodland, A., Bussod, G, Kornprobst, J., Bodinier, J.L., Macpherson,E , and Menzies, M. (1992) The effect of mafic dike emplacement onsurrounding peridotite: Evidence from spinel compositions and esti-mated redox states. Geological Society of America Abstracts with Pro-grams, 24, 85.

Wyllie, P.J (1978) Mantle fluid compositions buffered in peridotite-COyHrO by carbonate, amphibole, and phlogopite. Journal of Geology, 86,687-7 13.

-(1979) Petrogenesis and the physics of the earth In H.S. Yoder,Jr., Ed., The evolution of the igneous rocks, p. 483-520. PrincetonUniversity Press, Princeton, New Jersey.

M,c.Nuscmm REcETvED Mencg 23, 1993Mer.ruscnrrr AccEprED Jurv 15, 1993

magma transport and metasomatism in the mantle: a critical ... · magma transport and metasomatism...

Documents