tters 243 (2006) 701–710www.elsevier.com/locate/epsl

Earth and Planetary Science Le

Diffusion-driven extreme lithium isotopic fractionation in countryrocks of the Tin Mountain pegmatite

Fang-Zhen Teng ⁎, William F. McDonough, Roberta L. Rudnick, Richard J. Walker

Geochemistry Laboratory, Department of Geology, University of Maryland, College Park, MD 20742, U.S.A.

Received 26 October 2005; received in revised form 15 January 2006; accepted 16 January 2006Available online 28 February 2006

Editor: R.W. Carlson

Abstract

Lithium concentrations and isotopic compositions in the country rocks (amphibolites and schists) of the Tin Mountainpegmatite show systematic changes with distance to the contact. Both Li and δ7Li decrease dramatically along a ∼10m traversefrom the pegmatite into amphibolite, with Li concentration decreasing from 471 to 68ppm and δ7Li decreasing from +7.6 to −19.9.Rubidium and Cs also decrease from the pegmatite contact into the country rock, but only within the first 2m of the contact, afterwhich their concentrations remain constant. Neither mixing between pegmatite fluids and amphibolite, nor Li isotope fractionationby Rayleigh distillation during fluid infiltration is a likely explanation of these observations, due to the extremely light isotopiccomposition required for the amphibolite end-member in the mixing model (δ7Li=−20) and the similarly extreme isotopicfractionation required in a Rayleigh distillation model. Rather, these variations are likely due to isotopic fractionationaccompanying Li diffusion from the Li-rich pegmatite (Li=450 to 730ppm) into amphibolites (Li=20ppm). The fact that otheralkali element concentrations vary only within 2m of the contact reflects the orders of magnitude faster diffusion of Li relative toheavier elements.

Quartz mica schists in contact with the pegmatite also show large variations in both Li and δ7Li as a function of distance fromcontact (∼1wt.% to ∼70ppm and +10.8 to −18.6, respectively), but over a longer distance of N30m. Lithium concentrations of theschist decrease from ∼1wt.% adjacent to the contact to ∼70ppm 300m from the contact; the latter is a typical concentration inmetapelites. The nature of the δ7Li variations in the schists is different than in the amphibolites. Schists within the first 2m of thecontact have nearly identical δ7Li of +10, which mimics that of the estimated bulk pegmatite (+8 to +11). At a distance of 30m theδ7Li reaches the lowest value in the schists of −18.6 (similar to the lowest amphibolite measured). At a distance of 300m the δ7Liclimbs back to +2.5, which is within the range of δ7Li of other schists in the region and metapelites worldwide. The behavior of Liin the schists can also be modeled by Li diffusion, with the effective diffusion coefficient in the schist being ∼10 times greater thanthat in the amphibolite. The effective diffusion coefficients of Li in the amphibolite and schist are N2 orders of magnitude greaterthat those in minerals, which implicates the importance of fluid-assisted grain-boundary diffusion over solid-state diffusion intransporting Li through these rocks.© 2006 Elsevier B.V. All rights reserved.

Keywords: lithium diffusion; isotope fractionation; fluid infiltration; Tin Mountain pegmatite

⁎ Corresponding author. Tel.: +1 301 405 2707; fax: +1 301 4053597.

E-mail address: tfz@geol.umd.edu (F.-Z. Teng).

0012-821X/$ - see front matter © 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.epsl.2006.01.036

1. Introduction

Lithium is a fluid-mobile, moderately incompatibleelement having two stable isotopes with ∼17% relative

702 F.-Z. Teng et al. / Earth and Planetary Science Letters 243 (2006) 701–710

mass difference. This large mass difference gives rise tosignificant isotopic fractionations in different geologicalenvironments, especially those involving fluid–rockinteractions [see review of [1]]. These characteristicsresult in distinct Li concentration and isotopic compo-sition in the hydrosphere, continental crust and mantle.In general, the hydrosphere is isotopically heavy (δ7Liof seawater=+32), the mantle is intermediate at +4,whereas the weathered, upper continental crust isisotopically light (∼0) [1,2].

Equilibrium Li isotope fractionation, like other stableisotope systems [3], is due to differences in bondenergies between different phases for the two isotopes(e.g., mineral and fluid), with heavy Li (7Li) preferringthe higher-energy bond. Bond energy is, in turn, relatedto coordination number, with lower coordinationnumber sites having higher bond energy [4]. Thus Lisubstituting for Mg in octahedral mineral sites (e.g.,pyroxenes) will be isotopically lighter than Li found influids, which is believed to be in tetrahedral coordina-tion [4,5]. Recent studies using both laboratory experi-ments and natural rocks have demonstrated theexistence of another type of Li isotopic fractionation,produced by the different diffusivities of 6Li and 7Li insilicate melts and solids. Diffusion couple experimentshave reported ∼40‰ Li isotopic fractionation betweenjuxtaposed molten rhyolite and basalt, with an initial Liconcentration ratio of ∼15 [6]. Large Li isotopicvariation (N10‰) was also observed along a ∼2m Liconcentration gradient in peridotite adjacent to discor-dant dunites (former melt channels), and explained by Liisotopic fractionation during a combined process ofdiffusion and melt extraction [7]. Compared withequilibrium isotopic fractionation produced by bond-energy differences, diffusion-driven (kinetic) isotopicfractionation is controlled by the concentration gradientand the relative mobility of isotopes. Thus, it can occurat high temperatures where equilibrium fractionationdiminishes [8].

In this paper we report Li concentrations and isotopiccompositions of country rocks adjacent to the Li-rich TinMountain pegmatite, Black Hills, South Dakota. Em-placement and crystallization of this pegmatite produceda large Li concentration gradient between the pegmatiteand country rocks and led to Li diffusion into the countryrocks. Our results demonstrate that up to 30‰fractionation of Li isotopes was produced in this manner.

2. Geological background and samples

Samples studied here are from the country rocks ofthe ca. 1.7Ga Tin Mountain pegmatite, Black Hills,

South Dakota, which crops out ∼12km to the southwestof the main body of the Harney Peak Granite [9]. TheLi-rich, Tin Mountain pegmatite discordantly intrudedboth schists and a tabular amphibolite unit, which arethe focus of this paper. The pegmatite consists of fivemajor structural/mineralogical zones, with the wall zoneforming a shell that encloses inner zones. Fracturefillings are compositionally similar to the core andtraceable from the core into the surrounding zones [10].The estimated formation pressure is ∼0.3GPa [9] withcrystallization temperature varying from ∼600 down to340°C [10–12]. Lithium isotopic compositions ofminerals and whole rocks from the different zones ofthe Tin Mountain pegmatite, as well as the Harney PeakGranite, associated simple pegmatites and representativecountry rocks are reported elsewhere [13]. All samplesmeasured here have been characterized in previousinvestigations [14,15] and previously published chem-ical data are provided in the electronic supplement (seeAppendix).

2.1. Amphibolites

The amphibolite is similar in composition totholeiitic basalt. It overlies much of the wall zone ofthe pegmatite, and reached amphibolite facies prior tothe intrusion of the pegmatite [9]. The Li isotopicvariation within the amphibolite was determined foreight samples collected along a 10m vertical traverseabove the pegmatite body, where the outcrop ends at thecliff top. An additional sample, collected from adifferent amphibolite body 300m away from thepegmatite, was also analyzed.

The amphibolite is a medium-grained rock composedpredominantly of hornblende and plagioclase with minoramounts of quartz and biotite. Significant amounts ofcalcite (2–7%) are present in two samples (samples 12-2and 12-3). The amphibolite is layered, with segregationsof oriented hornblende alternating with layers ofcrystalloblastic plagioclase [14]. Most major and traceelements have relatively constant concentration alongthe traverse (Fig. 1), which suggests that modalvariations are insignificant. Some fluid-mobile elements,however, show large variations in amphibolites near thecontact (Fig. 1). Relative to the mean of the othersamples, the sample taken from the contact (sample 12-8) is significantly enriched in K2O, Rb, Cs, Sb, δ

18Oand depleted in Na2O [14,15]. By comparison, sample12-7, taken 0.5m from the contact, shows considerablyless enrichment/depletion of these elements.

Feldspars show little compositional variation alongthe traverse, and range from An30–40, except for the

Fig. 1. Concentration profiles of SiO2, Sc, Cs, Rb, Sb, Na2O in theamphibolite and Na2O in the amphibole vs. distance from the contactwith the Tin Mountain pegmatite. Data from Walker [14] and Laul etal. [15] and are provided in the electronic supplement.

Table 1Lithium isotopic composition and concentration of samples from theTin Mountain pegmatite

Sample ID δ7Li a Li b Distance

(ppm) (m)

Amphibolites14-1 +0.9 20 30012-1 −19.5 72 9.512-1 replicate c −19.912-2 −14.2 68 812-3 −13.7 98 6.512-4 −8.3 140 512-4 replicate −7.512-5 −0.9 209 3.512-6 +6.5 260 212-7 +7.5 438 0.512-8 +7.6 471 0.0312-8 replicate +7.3

Quartz mica schists40-1A −3.1 62 N8000WC-4 +1.6 79 N800026-2 +2.3 150 N800023-2 +2.5 68 30023-4 −18.0 124 3023-4 replicate −18.69-4 +10.0 960 1.59-1 +8.9 1607 0.39-3 +9.0 1179 0.113-1 +10.8 ∼10,000 0

Wall zone whole rock of Tin Mountain pegmatite10-3 +11.1 4539-2 +7.5 50443-1 +11.1 735

a Analytical uncertainty is b±1‰ (2σ), based on both pure Lisolutions and natural rocks (see text for details).b Li measured by comparison of signal intensities with 50 or 100ppb

L-SVEC.c Replicate: repeat column chemistry from the same sample solution.

703F.-Z. Teng et al. / Earth and Planetary Science Letters 243 (2006) 701–710

sample at the contact, which is more anorthite-rich(∼An50). Amphibole compositions vary on the scale ofa thin section, but, except for Na, no systematic changeis observed along the traverse [14]. FeO contents rangefrom 12.3 to 19.9wt.% and MgO contents range from7.6 to 12.4wt.%. Amphibole at the contact containsmore Na than amphiboles elsewhere along the traverse,a trend that is different from that observed from wholerock samples [14].

2.2. Quartz mica schists

Schists are the dominant country rocks in this area.Their protolith was shale that was metamorphosed to thesillimanite zone during regional metamorphism prior tothe intrusion of Tin Mountain pegmatite [9]. Comparedwith the amphibolites, fresh outcrops of the schists arelimited and it thus proved difficult to collect systemat-ically along a traverse away from the pegmatite.Consequently, a variety of schists were sampled in andaround the pegmatite in order to determine the extent ofa metasomatic halo on a lateral scale. These samplesinclude an inlier of schist enclosed by the wall and thirdintermediate zones of the pegmatite (13-1), three sam-ples collected near the lower contact of the pegmatite (9-3, 9-1 and 9-4), and two samples taken from scatteredoutcrops to the northwest and southwest ends of thepegmatite at distances of 30 (23-4) and 300m (23-2). Inorder to characterize any compositional variation withinthe Black Hills terrane, three more schist samples weretaken in this region far away (e.g., N8km) from knowngranite and pegmatite outcrops.

The schists are composed of quartz, biotite, plagio-clase and minor muscovite. Both modal mineralogy andmajor element composition vary considerably. The va-riations are probably due to the original sedimentarylayering of the protolith [14]. Trace element concentra-

tions do not generally correlate with modal biotite+muscovite abundances. Relative to regional schists, theinlier sample (13-1) is extremely enriched in Rb, Cs, andZn and depleted in δ18O. Lesser amounts of alkalienrichment are found in two samples taken within 0.5mof the pegmatite (i.e., 9-1 and 9-3) and no discernableenrichments are seen in samples taken from over a meterfrom the contact (9-4, 23-4 and 23-2) [14].

3. Analytical methods

Sample powders were the same as those used inprevious studies [14–17]. Procedures for sampledissolution and column chemistry are described inRudnick et al. [18] and the method of instrumentalanalysis is found in Teng et al. [2].

704 F.-Z. Teng et al. / Earth and Planetary Science Letters 243 (2006) 701–710

The external precision of Li isotopic analyses, basedon 2σ of repeat runs of pure Li standard solutions androck solutions, is b±1.0‰. For example, pure Listandard solutions (IRMM-016 and UMD-1) alwaysyield values falling within previously established ranges(−0.1‰±0.2‰ and +54.7‰±1‰) [2]; in-house rockstandard AO-12, a shale from the Amadeus Basin (partof the PAAS group of Nance and Taylor, [19]), givesδ7Li=+3.5±0.6 (2σ, n=36 runs with 4 replicate samplepreparations); and BCR-1 gives δ7Li=+2.0±0.7 (2σ,n=10 runs). The uncertainty of Li concentrationmeasurement, determined by the comparison of signalintensities with that measured for the 100 or 50ppb L-SVEC standard and then adjusting for sample weight, isb±10%. The accuracy of this method has beenestablished in Teng et al. [2] to be b±5% as based onisotope dilution methods.

4. Results

Lithium concentration and isotopic composition forboth schists and amphibolites are reported in Table 1 andplotted in Fig. 2 as a function of distance from thecontact. The Li concentration and isotopic composition

Fig. 2. Plots of Li and δ7Li verses distance form the contact for a)amphibolite samples taken along a vertical profile and b) schistsamples taken at different distances to the contact, but not along asingle profile. Stars in a) represent Li (filled) and δ7Li (open) of theregional amphibolite taken from another unit, 300m from the contact.Data from Table 1.

of the Tin Mountain pegmatite and related igneous rockshave been reported in Teng et al. [13]. Data for threewall zone whole rock samples that represent the averagecomposition of the Tin Mountain pegmatite arepresented for comparison in Table 1.

4.1. Amphibolite

Both Li and δ7Li decrease with distance from thecontact (Fig. 2a). The amphibolite near the pegmatitecontact is isotopically heavy (+7.6), similar to wall zonesamples of the pegmatite (+7.5 to +11.1), and it has thehighest Li concentration (471ppm). The sample at theend of the traverse, 10m away from the contact, has thelowest Li concentration (70ppm) and the lightestisotopic composition (−19.9). The regional amphibolite,taken 300m from the Tin Mountain pegmatite, has a Liconcentration (20ppm) and isotopic composition (+0.9)comparable to typical upper crustal lithologies [2].

4.2. Quartz mica schist

Lithium concentrations and isotopic compositions ofschists are also highly variable (Fig. 2b). The schistenclosed by the pegmatite (13-1) is extremely rich in Li(∼1wt.%) with pegmatite-like δ7Li (+10.8). Threesamples within 1.5m of the pegmatite also showpegmatite-like Li isotopic compositions and high Liconcentrations, with the furthest one (9-4, 1.5m from thecontact) showing no discernable enrichments in otherfluid mobile elements. A sample collected 30m from thepegmatite has a very light Li isotopic composition(−18.6) and a Li concentration of 124ppm. Samplescollected much further from the contact show typicalshale-like Li concentrations (∼70ppm) and isotopiccompositions (−3 to +2.5) [2].

5. Discussion

The Tin Mountain pegmatite is extremely enriched inisotopically heavy Li [13] relative to typical amphibo-lites and schists [20–22]. Thus, the compositionalvariations observed within the country rocks as afunction of distance from the pegmatite must haveresulted from mass transfer of elements enriched in thepegmatite into the country rocks. This transfer may haveoccurred by fluid infiltration and/or solid-state diffusion.

Crystallization of the Tin Mountain pegmatiteexsolved a large amount of fluids [10,14]. Since Li isfluid-mobile [23], infiltration of these fluids into thecountry rocks would have carried Li, which could thenequilibrate with the country rocks. This process can be

Fig. 3. Plots of Li and δ7Li for amphibolites and two end-membermixing model. Mixing equation used: δ7Li= (δ7Li)1× f1+ (δ

7Li)0×(1−f1), f1 = fraction of Li in end member 1. End-members are shown assolid symbols and represent the measured values of amphibolites ateither end of the transect. Star represents composition of the regionalamphibolite taken from another unit, 300m from the contact (Table 1).

705F.-Z. Teng et al. / Earth and Planetary Science Letters 243 (2006) 701–710

considered as simple mixing between pegmatite fluidsand country rocks if no isotopic fractionation occurred,or modeled as Rayleigh Distillation if Li isotopicfractionation occurred during fluid infiltration. Alterna-tively, the large compositional contrast between peg-matite and country rocks could have led to diffusion ofLi from the pegmatite into country rocks [24]. Suchdiffusion can produce large Li isotopic fractionations, asdemonstrated by previous studies [6,7]. Below wemodel each process and evaluate their potentials forexplaining the Tin Mountain data. While all processescould have modified the Li concentrations and isotopiccompositions of country rocks, we show that diffusion isthe most likely process to have produced the trends weobserve.

5.1. Modeling the Li profile in amphibolite countryrocks

Amphibolites show a ten-fold decrease in Li concen-tration and a ∼30‰ drop in Li isotopic compositionalong the 10m traverse. Variations in both Li and δ7Livalues do not correlate with either the modes ofamphibole or plagioclase or the amphibole or plagioclasecompositions. They therefore likely reflect interactionsbetween the amphibolites and the Tin Mountainpegmatite.

5.1.1. Mixing modelIf no Li isotopic fractionation during fluid-rock

interactions occurred, then fluid infiltration can bemodeled as two-end-member mixing between isotopi-cally heavy Li from the pegmatite and lighter Li in theoriginal amphibolite. Using amphibolites at either end ofthe transect as end-members, the mixing model fits thedata well (Fig. 3).

The high Li concentration and δ7Li of the amphiboliteat the contact is similar to the Tin Mountain pegmatite,and suggests that Li in this sample is dominated by acomponent coming from the pegmatite. The other am-phibolite end-member, however, is richer in Li (70ppm)and isotopically lighter (−19.9) than regional amphib-olite (sample 14-1) or amphibolites from other regions[20,21]. The regional amphibolite was likely not affectedby the intrusion of this pegmatite or other plutons, andcan be considered as representative of unalteredamphibolites in this area. The large difference in Liand δ7Li between the amphibolite end-member used inthese mixing calculations and typical amphibolitessuggests that the end-member has been substantiallyaffected by Li infiltration from the pegmatite, and thusthis simple mixing model cannot explain the data.

5.1.2. Rayleigh distillation modelPrevious studies have shown that Li isotopes can be

strongly fractionated during low-T fluid-rock interac-tions, with isotopically heavy Li partitioning into fluidsrelative to rocks [18,25,26]. Although fluid infiltrationof amphibolite country rocks occurred at relatively hightemperature (TN340°C) [10–12], Li isotopes may stillbe fractionated (the isotopic exchange between amphi-bole, plagioclase and fluids has not yet been quantifiedas a function of temperature). Therefore, we explorefluid infiltration accompanied by isotopic fractionationusing a Rayleigh fractionation law, which is consideredas an extreme end-member process that fractionatesisotopes in the most effective way.

A series of Rayleigh distillation curves can becalculated to fit most of the data (Fig. 4), assumingthat the Li isotopic fractionation factor between fluidsand minerals (α) is ∼1.013±0.002, i.e., there is ∼13‰difference between fluids and amphibolites. Since, inreality, fluid infiltration may not perfectly obey a Ray-leigh distillation process, which is the most effectiveprocess at fractionating isotopes, these estimated αvalues are considered to be minimum values. Althoughexperimentally measured fractionation factors betweenhydrothermal fluids and amphibolites are not currentlyavailable, data for α determined for seafloor alterationcan be used for comparison. At ∼2 °C, the Li isotopicfractionation factor between seawater and clays pro-duced by alteration of basalts (α) is inferred to be 1.019[25]; while at 350 °C, α between hydrothermal fluidsand basalts decreases to 1.003–1.007 [27,28]. Given aminimum crystallization temperature for this system of

Fig. 4. Plots of Li and δ7Li for amphibolites and Rayleigh distillationmodel. Rayleigh distillation equation: δ7Liamph = (δ

7Li0 +1000)f (α−1)−1000; α: Li isotopic fractionation factor defined as 7Li /6Lifluid /

7Li /6Liamph; f: the fraction of Li remaining in the rock,calculated from Liamph /Li0; 0: amphibolite at the contact. The grayfield represents the area where data would plot for a Rayleighdistillation model that uses more appropriate empirical α values (1.003to 1.007). The greater α values (on lines) compared with thoseempirically observed in similar conditions rule out this process as amechanism explaining the data.

706 F.-Z. Teng et al. / Earth and Planetary Science Letters 243 (2006) 701–710

∼340 °C [10–12], the fractionation factor required to fitthe data (1.013) is significantly greater than thoseobserved at comparable temperatures in the oceanicenvironment. Therefore, this process is unlikely toexplain the data.

5.1.3. Diffusion modelThe above discussions suggest that fluid infiltration,

with or without isotopic fractionation, is unlikely toproduce the Li profile observed in the amphibolitecountry rocks. Instead, we explore the likelihood thatdiffusion of Li played an important role in producing theextreme isotopic fractionation observed. The impor-tance of diffusion in producing the Li variations isconsistent with the differences in scales observed inchemical modification of the country rocks for Li versusother elements. All other elements that change system-atically within the amphibolite transect show concen-tration variations only within 2m of the contact (Fig. 1).The much larger scale of Li variation (≥10m) mayreflect greater Li diffusivity relative to the otherelements examined. Although Li diffusivity in amphi-bole is unknown, studies of other silicate minerals andmelts demonstrate that Li diffusion is orders ofmagnitude faster than other elements [6,29–31]. Forexample, Li diffuses more than four orders of magnitudefaster than other alkali elements in plagioclase feldspar[30] and 100 times faster than most trace elements in aliquid basalt [6].

To model the data, the pegmatite is considered as aninfinite Li reservoir relative to the country rocks;

therefore, the diffusion model used to fit the data is aone-dimensional diffusion model with the boundaryconditions of a constant Li concentration at the contactand uniform initial Li concentration in the countryrocks. Using these boundary conditions, the solution tothe Fick's second law of diffusion is [32]:

½ðCx−C1Þ=ðC0−C1Þ� ¼ erfc½x=2ðDtÞ1=2�

where here, x = distance from the contact; Cx = elementconcentration at distance x from the contact;C0 = elementconcentration at the contact at x=0; C1 = elementconcentration in unaffected country rocks; D = diffusioncoefficient; t = duration of the diffusion process; erfc =complementary error function.

6Li and 7Li are treated as two different elements andeach diffusion curve is modeled separately. The 7Li / 6Liratio or δ7Li is calculated by combining concentrationsat different distances. By using the regional amphiboliteto represent the unaltered amphibolite country rock, andthe pegmatite as representative of the concentration atthe contact, the following parameters are derived:C1=20ppm and δ7Li=+0.9. This, with C0=471ppmand δ7Li=+7.6, allows C1,6 (or C1,7) and C0,6 (or C0,7)to be calculated. Since neither 7Li and 6Li diffusioncoefficients in amphibolites, nor the duration ofdiffusion is known, the value of √Dt (characteristicdiffusive transport distance) is calculated to best fit theLi concentration profile (Fig. 5a). The best value of√Dt=3.1 (m) was determined by using the MicrosoftExcel® solver function. The effective diffusion coeffi-cient derived here corresponds to the transport of Lithrough a mixed phase material (amphibolite+ fluid, seeSection 5.3 for more details). Moreover, this diffusioncoefficient is the average D of both 7Li and 6Li, butsince N90% of Li is 7Li, D is considered to represent D7

and the Li concentration profile can be considered to bethat of 7Li.

In order to model the δ7Li profile, the diffusionprofile for 6Li is needed. If the relationship between D7

and D6 can be found, then the6Li diffusion curve can be

modeled, followed by the δ7Li curve. Here theassumption of Richter et al. [8] is used: the ratio of theeffective diffusion coefficients of the isotopes 1 (light)and 2 (heavy) of element i are related to their mass bythe following equation: Di,1 /Di,2= (mi,2 /mi,1)

β, where βis an empirical parameter determined from experimentaldata. Given the same conditions (e.g., same concentra-tion gradient, temperature etc.) but different diffusionmedia, the larger the β, the stronger the isotopicfractionation. For ideal gases, β is equal to 0.5, whilein condensed matter and fluids it is b0.5. For example, a

Fig. 5. Diffusion calculations (curves) compared to data (symbols) forLi and δ7Li of amphibolites vs. distance from the contact, a) diffusionmodel for the amphibolite profile and b) shows how varying the initialLi isotopic composition of the unaltered amphibolite has little effect onoutcome. Star in b) represents the Li (filled) and δ7Li (open) of theregional amphibolite, which is used as the initial Li and δ7Li values inthe diffusion model in a); see text for details.

Fig. 6. Models of diffusion-induced lithium isotope fractionation forschists as a function of distance from the contact. Due to the limitedsample numbers, the model is non-unique. See text for further details.

707F.-Z. Teng et al. / Earth and Planetary Science Letters 243 (2006) 701–710

β value of ∼0.215 was determined experimentally for Lidiffusion in silicate melts [6] and 0.071 [33] and 0.015[34] were reported for Li diffusion in water.

The effective diffusion coefficients depend ontemperature, pressure and matrix (composition, miner-alogy, fluid, etc.) [35]. The estimated crystallizationtemperature of the Tin Mountain pegmatite varied from∼600 to 340°C [10–12] at an intrusion depth of∼10km[9]. At a normal geothermal gradient of∼35°C/km [36],the country rocks at this depth should have had atemperature of∼350°C, similar to that at the final stagesof pegmatite crystallization. Therefore, the temperaturegradient produced by the intrusion in the country rocksshould be small. Moreover, both the major elementcomposition and the mineralogy of the amphibolitesalong the traverse change little, so the effective diffusioncoefficients of both 6Li and 7Li in the amphibolite areassumed to have remained constant, i.e., D6 /D7= (m7 /m6)

β = constant, which means β is constant. The β valueyielding the best fit to the data is 0.12 (Fig. 5a) and thecorresponding ratio of D6 /D7=1.018. This β value is

smaller than that measured in silicate melt [6] butgreater than those measured in water [33,34]. Thissuggests that the difference in the effective diffusioncoefficient between 7Li and 6Li is larger in fluid-infiltrated amphibolite than in pure water, which mayresult from different bonding environments for Li inpure water compared to supercritical fluids [13] andinteractions of fluids with amphibolites. Furthermore, itis important to note that the initial Li isotopiccomposition of unaltered amphibolite is not importantfor this modeling (Fig. 5b). The calculated curves still fitthe data well even if the δ7Li values of the amphiboliteare varied from −4 to +6.

In summary, the Li profile observed in the amphib-olite country rocks cannot be explained by mixingbetween pegmatite fluids and amphibolite due to theextremely light δ7Li required for the amphibolite end-member (≤−20). Nor can Li isotope fractionation byRayleigh distillation during fluid infiltration explain thetrend due to the extreme fractionation factors required(≥1.013). Instead, these variations can be explained byisotopic fractionation accompanying Li diffusion fromthe pegmatite into amphibolites, assuming reasonablediffusion coefficients, β values and δ7Li value for theunaltered amphibolite.

5.2. Modeling the Li profile in schist country rocks

The Li concentrations and isotopic compositions ofschists taken from the inlier within the pegmatite andfrom near the contact with the pegmatite are dominatedby Li from the Tin Mountain pegmatite. Samples takenfar from the pegmatite (N300m) have Li similar toschists and shales worldwide [2,22,37]. These distalschists may, therefore, represent samples that have notbeen affected by interactions with any pluton. Thesample taken 30m from the pegmatite, however, has

Fig. 7. Plot of D vs. t, assuming √Dt=3.1 and 10 in amphibolites andschists, respectively;D values for Li in water from Li and Gregory [39]and for Li in plagioclase and pyroxene from Giletti and Shanahan [30]and Coogan et al. [29], respectively.

708 F.-Z. Teng et al. / Earth and Planetary Science Letters 243 (2006) 701–710

extremely light Li isotopic composition and was likelyinfluenced by the pegmatite.

This isotopically light schist, like the isotopicallylight amphibolite, may also result from Li isotopicfractionation through diffusion-related processes. Ap-plying the approach outlined above to the data for theschist provides a best-fit √Dt=10 (m) and β=0.15,although the lack of a well-constrained diffusion profilethrough the schist means that these values are not unique(Fig. 6). Nevertheless, the schist's relatively large √Dtvalue compared with that in amphibolite is consistentwith earlier findings that demonstrated that Li canmigrate to a greater distance in schist country rocks (theEtta pegmatite near the Harney Peak Granite) thanamphibolite country rocks of the Tin Mountainpegmatite [15].

5.3. Implications for the nature of diffusion

Data from the country rocks of the Tin Mountainpegmatite demonstrate that Li isotopes can be stronglyfractionated by diffusion at moderate temperature (downto 340°C) and over large length scales (N30m). Thisobservation agrees well with the theoretical model,implying that diffusion-induced isotopic fractionationmay be important and Li isotopic heterogeneity mayexist at both large and small scales (e.g., zoned mineralsor minerals that interacted with fluids). Moreover,results of this modeling can be used to constrain theeffective diffusion coefficients in the country rocks andto constrain the type of diffusion: grain-boundarydiffusion vs. volume diffusion.

Based on the best fits to the Li concentration profiles,the value for √Dt differs between the amphibolite andschist with Dschist/Damphibolite=∼10. Although neitherthe effective diffusion coefficient (D) nor the duration ofthe diffusion process (t) is known, the minimum D canbe calculated by assuming that the maximum time ofdiffusion is equal to the age of the Tin Mountainpegmatite (1.7Ga). As a result, the minimum Damphibolite

for Li is calculated to be 2×10−12 cm2/s and theDschist isaccordingly ∼10 times higher (Fig. 7). These estimatedminimum D values are N2 orders of magnitude higherthan those measured in silicate minerals (e.g., feldsparand pyroxene) [29,30], clearly ruling out the possibilityof volume diffusion as the mechanism by which Lidiffused through the country rocks, and suggesting thatdiffusion occurred in an interconnected fluid phase thatinteracted with mineral grains.

As discussed in Watson and Wark [38], the effectivediffusion coefficient is related to the diffusion coeffi-cient in the fluid, effective porosity and interactions

between fluids and minerals. Whereas Li diffusivity inpure water has been measured [39], it is unknown insupercritical fluids (e.g., fluids exsolved during crys-tallization of the Tin Mountain pegmatite). It is likelythat the different bonding environment for Li in purewater (Li coordinated with H2O) is different from thatin supercritical fluid (Li coordinated with Cl) [13].Furthermore, the extent and nature of interactionsbetween fluids and country rocks (e.g., fluid/rockratios, Li partition coefficients between supercriticalfluids and schist/amphibolite minerals) are not wellcharacterized. In the simple case, if we assume that thediffusion coefficients in fluids are the same betweenamphibolites and schists and the partition coefficientsfor Li between the fluid and rock are not significantlydifferent, the effective diffusion coefficient is mainlycontrolled by the effective interconnected porosity. Inthis case, the effective interconnected porosity inschists is thus ∼10 times larger than in amphibolite,reflecting a greater permeability in the schist comparedto amphibolite.

6. Conclusions

The country rocks of the Tin Mountain pegmatiteexhibit ∼30‰ Li isotopic change over a ∼10m traversein the amphibolites and N30m in the schist. Theseobservations are unlikely to be due to either mixingbetween pegmatite and country rocks or Li isotopefractionation by Rayleigh distillation during fluidinfiltration. Instead, they reflect extreme Li isotopicfractionation accompanying Li diffusion from the Li-rich

709F.-Z. Teng et al. / Earth and Planetary Science Letters 243 (2006) 701–710

pegmatite into country rocks. These extreme fractiona-tions are due to the large differences in diffusioncoefficients between 6Li and 7Li, and in Li concentra-tions between the pegmatite and country rocks.

Lithium diffusion in the schists is inferred to be ∼10times faster than in the amphibolites. However, in bothtypes of rocks, Li diffuses much faster than other alkalielements, as concentrations of these elements vary onlywithin 2m of the contact. The effective diffusioncoefficient of Li in both rocks are N2 orders of magnitudegreater than in minerals, which indicates Li diffusion incountry rocks was dominated by fluid-assisted grain-boundary diffusion over solid-state diffusion.

Acknowledgements

We thank George Helz for very helpful discussionsand Boz Wing for discussions and guidance with themodeling presented here. The constructive commentsfrom Lui-Heung Chan, Craig Lundstrom and efficientediting from Richard Carlson are greatly appreciated.We appreciate the efforts of Dr. Richard Ash in PlasmaLab. This work was supported by the N.S.F (EAR0208012).

Appendix A. Supplementary data

Supplementary data associated with this article canbe found, in the online version, at doi:10.1016/j.epsl.2006.01.036.

References

[1] P.B. Tomascak, Developments in the understanding and applica-tion of lithium isotopes in the Earth and planetary sciences, in: C.Johnson, B. Beard, F. Albarede (Eds.), Geochemistry of Non-Traditional Stable Isotopes, vol. 55, Mineralogical Society ofAmerica, Washington, DC, 2004, pp. 153–195.

[2] F.-Z. Teng, W.F. McDonough, R.L. Rudnick, C. Dalpe, P.B.Tomascak, B.W. Chappell, S. Gao, Lithium isotopic compositionand concentration of the upper continental crust, Geochim.Cosmochim. Acta 68 (2004) 4181–4192.

[3] T. Chacko, D.R. Cole, J. Horita, Equilibrium oxygen,hydrogen and carbon isotope fractionation factors applicableto geological systems, in: J.W. Valley, D.R. Cole (Eds.),Stable Isotope Geochemistry, Reviews in Mineralogy andGeochemistry, vol. 43, Mineralogical Society of America,Washington, DC, 2001, pp. 1–82.

[4] M. Wenger, T. Armbruster, Crystal-chemistry of lithium—oxygen coordination and bonding, Eur. J. Mineral. 3 (1991)387–399.

[5] K. Yamaji, Y. Makita, H. Watanabe, A. Sonoda, H. Kanoh, T.Hirotsu, K. Ooi, Theoretical estimation of lithium isotopicreduced partition function ratio for lithium ions in aqueoussolution, J. Phys. Chem., A 105 (2001) 602–613.

[6] F.M. Richter, A.M. Davis, D.J. DePaolo, E.B. Watson, Isotopefractionation by chemical diffusion between molten basalt andrhyolite, Geochim. Cosmochim. Acta 67 (2003) 3905–3923.

[7] C.C. Lundstrom, M. Chaussidon, A.T. Hsui, P. Kelemen, M.Zimmerman, Observations of Li isotopic variations in thetrinity ophiolite: evidence for isotopic fractionation by diffusionduring mantle melting, Geochim. Cosmochim. Acta 69 (2005)735–751.

[8] F.M. Richter, Y. Liang, A.M. Davis, Isotope fractionation bydiffusion in molten oxides, Geochim. Cosmochim. Acta 63(1999) 2853–2861.

[9] J.A. Redden, J.J. Norton, R.J. McLaughlin, Geology of theHarney Peak granite, Black Hills, South Dakota, in: F.J. Rich(Ed.), Geology of the Black Hills, South Dakota and Wyoming,Amer. Geol. Inst., 1985, pp. 225–240.

[10] R.J. Walker, G.N. Hanson, J.J. Papike, J.R. O'Neil, J.C. Laul,Internal evolution of the Tin Mountain pegmatite, Black Hills,South-Dakota, Am. Miner. 71 (1986) 440–459.

[11] M.L.C. Sirbescu, P.I. Nabelek, Crystallization conditions andevolution of magmatic fluids in the Harney Peak Granite andassociated pegmatites, Black Hills, South Dakota—Evidencefrom fluid inclusions, Geochim. Cosmochim. Acta 67 (2003)2443–2465.

[12] M.L.C. Sirbescu, P.I. Nabelek, Crustal melts below 400 °C,Geology 31 (2003) 685–688.

[13] F.-Z. Teng, W.F. McDonough, R.L. Rudnick, R.J. Walker.Lithium isotopic systematics of granites and pegmatites from theBlack Hills, South Dakota, Am. Miner. (In review).

[14] R.J. Walker. The origin of the Tin Mountain pegmatite, BlackHills, South Dakota. PhD. dissertation, Univ. New York at StonyBrook, 1984.

[15] J.C. Laul, R.J. Walker, C.K. Shearer, J.J. Papike, S.B. Simon,Chemical migration by contact metamorphism between pegma-tite and country rocks: natural analogs for radionuclidemigration, Material Research Society Symposium Proceedings,vol. 26, 1984, pp. 951–958.

[16] R.J. Walker, G.N. Hanson, J.J. Papike, Trace-element constraintson pegmatite genesis—Tin Mountain pegmatite, Black Hills,South-Dakota, Contrib. Mineral. Petrol. 101 (1989) 290–300.

[17] R.J. Walker, G.N. Hanson, J.J. Papike, J.R. O'Neil, Nd, O and Srisotopic constraints on the origin of Precambrian rocks, southern-Black-Hills, South-Dakota, Geochim. Cosmochim. Acta 50(1986) 2833–2846.

[18] R.L. Rudnick, P.B. Tomascak, H.B. Njo, L.R. Gardner, Extremelithium isotopic fractionation during continental weatheringrevealed in saprolites from South Carolina, Chem. Geol. 212(2004) 45–57.

[19] W.B. Nance, S.R. Taylor, Rare-earth element patterns and crustalevolution: 1. Australian Post-Archean sedimentary-rocks, Geo-chim. Cosmochim. Acta 40 (1976) 1539–1551.

[20] F.-Z. Teng, W.F. McDonough, R.L. Rudnick, S. Gao, Lithiumisotopic composition of the deep continental crust, EOS 85 (47)(2004) (Abstr. V51C-0597).

[21] D.J. Bottomley, L.H. Chan, A. Katz, A. Starinsky, I.D. Clark,Lithium isotope geochemistry and origin of Canadian Shieldbrines, Ground Water 41 (2003) 847–856.

[22] F.-Z. Teng, W.F. McDonough, R.L. Rudnick, B.A. Wing, Lack oflithium isotopic fractionation during progressive metamorphicdehydration in metapelite: a case study from the Onawa contactaureole, Maine. Chem. Geol. (in preparation).

[23] J.M. Brenan, F.J. Ryerson, H.F. Shaw, The role of aqueous fluidsin the slab-to-mantle transfer of boron, beryllium, and lithium

710 F.-Z. Teng et al. / Earth and Planetary Science Letters 243 (2006) 701–710

during subduction: experiments and models, Geochim. Cosmo-chim. Acta 62 (1998) 3337–3347.

[24] W. Jost, Diffusion in Solids, Liquids, and Gases, Academic Press,New York, 1960. 558 pp.

[25] L.H. Chan, J.M. Edmond, G. Thompson, K. Gillis, Lithiumisotopic composition of submarine basalts: implications for thelithium cycle in the oceans, Earth Planet. Sci. Lett. 108 (1992)151–160.

[26] Y. Huh, L.-H. Chan, L. Zhang, J.M. Edmond, Lithium and itsisotopes in major world rivers: implications for weathering andthe oceanic budget, Geochim. Cosmochim. Acta 62 (1998)2039–2051.

[27] L.H. Chan, J.M. Gieskes, C.F. You, J.M. Edmond, Lithiumisotope geochemistry of sediments and hydrothermal fluids of theGuaymas Basin, Gulf of California, Geochim. Cosmochim. Acta58 (1994) 4443–4454.

[28] L.H. Chan, J.M. Edmond, G. Thompson, A lithium isotope studyof hot-springs and metabasalts from Mid-Ocean ridge hydrother-mal systems, J. Geophys. Res.-Solid Earth 98 (1993) 9653–9659.

[29] L.A. Coogan, S.A. Kasemann, S. Chakraborty, Rates ofhydrothermal cooling of new oceanic upper crust derived fromlithium-geospeedometry, Earth Planet. Sci. Lett. 240 (2005)415–424.

[30] B.J. Giletti, T.M. Shanahan, Alkali diffusion in plagioclasefeldspar, Chem. Geol. 139 (1997) 3–20.

[31] C.C. Lundstrom, An experimental investigation of the diffusiveinfiltration of alkalis into partially molten peridotite: implications

for mantle melting processes, Geochem. Geophys. Geosyst. 4(2003) 8614, doi:10.1029/2001GC000224.

[32] J. Crank, The Mathematics of Diffusion, Clarendon Press,Oxford, 1975. 414 pp.

[33] S.J. Fritz, Measuring the ratio of aqueous diffusion coefficientsbetween 6Li+Cl− and 7Li+Cl− by osmometry, Geochim. Cosmo-chim. Acta 56 (1992) 3781–3789.

[34] F.M. Richter, R.A. Mendybaev, J.N. Christensen, I.D. Hutcheon,R.W. Williams, N.C. Sturchio, J.A.D. Beloso, Kinetic isotopicfractionation during diffusion of ionic species in water, Geochim.Cosmochim. Acta 70 (2) (2006) 277–289.

[35] D.R. Cole, S. Chakraborty, Rates and mechanisms of isotopicexchange, in: J.W. Valley, D.R. Cole (Eds.), Stable IsotopeGeochemistry, Reviews in Mineralogy and Geochemistry, vol.43, The Mineralogical Society of America, Washington, DC,2001, pp. 83–224.

[36] G.C. Brown, A.E. Mussett, The Inaccessible Earth: An IntegratedView of its Structure and Composition, Chapman and Hall,London, 1993. 276 pp.

[37] T. Moriguti, E. Nakamura, Across-arc variation of Li isotopes inlavas and implications for crust/mantle recycling at subductionzones, Earth Planet. Sci. Lett. 163 (1998) 167–174.

[38] E.B. Watson, D.A. Wark, Diffusion of dissolved SiO2 in H2O at1GPa, with implications for mass transport in the crust and uppermantle, Contrib. Mineral. Petrol. 130 (1997) 66–80.

[39] Y.-H. Li, S. Gregory, Diffusion of ions in sea water and in deep-sea sediments, Geochim. Cosmochim. Acta 38 (1974) 703–714.