Crystal Chemistry of Fe3+ and H+ in Mantle Kaersutite: Implications For

American Mineralogist, Volume 78, pages 968-979, 1993

Crystal chemistry of Fe3+and H+ in mantle kaersutite: Implications for mantle metasomatism

M. D.qnnv Dv.q.n Department of Geology and Astronomy, West ChesterUniversity, West Chester,Pennsylvania, 19383, U.S.A. STBpHBNJ. Ml.crwn'r-r Department of Geosciences,The PennsylvaniaState University, University Park, Pennsylvania 16802, U.S.A. AuNn V. McGurnn Department of Geosciences,University of Houston, Houston, Texas 77204, U.S.A. Launa R. Cnoss, J. Dlvro RonenrsoN Department of Chemistry and Center for Applied Energy Research,University of Kentucky, kxington, Kentucky 40506, U.S.A.

Ansrru.cr Chemical and crystal chemical analyseshave been performed on a suite of subcontinental, mantle-derived hornblende (kaersutite)samples. Mdssbauer techniques have been utilized to investigateFe valenceand site occupancies,U extraction techniqueshave been used to determine bulk H contents, proton-induced y-ray emission (PIGE) analysis was employed to measure F, and electron microprobe techniques coupled with the above measurementshave been utilized to determine major-element contents of hornblende. Similar analyseswere performed on a suite of metamorphic amphibole samples from Coscaet al. (1991).Comparison with their wet chemicalresults on Fe3+/Fe2+permitted determination of C : 1.22, the correction for differential recoil-free fraction effects,which was used to correct the mantle sample M0ssbauerdata. The results of the analysesfor the kaersutite samplesshow a nearly l:l inverse relationship between the Fe3+and H+ contents. Although the range of Fe3+/H+ in the less oxidized kaersutite samples may be explained by partial H loss during entrainment and ascent,the nearly total dehydrogenation of the Fe3*-rich megacrystswould require time scalessignificantly longer than what is expectedfor transport. Thus, it seemslikely that these oxykaersutite samplesgrew in a more oxidized metasomatic fluid, where incorporation of H was not required for charge compensation. As megacrystsfrom the same location show wide variation in Fe3* and H+, it appearslikely that significant variations in the oxidation stateof the mantle metasomatic fluid occurred over relatively small temporal or spatial scales.

INrnonucrroN while others lack hydrous phases.Improved understand- volatile in metasomatism and A 1964paper by Oxburgh was among the first to sug- ing of the role of species phasesis needed. gestthat the Earth's upper mantle must contain td;;- the crystal chemistry of the hydrous studies bv Dvar et al' (1989' eral amphibolein ordei to provide sufficienrK - ^l:ttlt^Sineralogical wood' virgo' and alkalis to serve as a source region for ururi ";e;il;r 1992a' 199.2b)among others including on the presenceof Fe3* Becausethe presenceof amphibole implies "or"u"lt.".,ig"ifl"*t coworkers have focusedattention " in tvpical mantle phases(McGuire et al', 1989;Wood volatile component in the mantle, this rugg"rtToi- *i !o1! relaredworkbyvarne(1970)ledronumero"*;li;r;" ung virgo, 1989;Dyar et al., 1989;Canil et al.' 1990) in metasomatizedperidotites (McGuire et al'' 1991; the influenceof volatiles on the melting uetra.,^ororpe- 1na 1992a). The studies on peridotite indicate ridotite (e.g., Bailey, lg7}, lg72) and on tfre efects ana Dyar et al., metasomatismare associated characteristicsofmetasomatism(e.g.,Menzies,1983; Dt"k that contrastingstyles of with sets of Fe3*/Fe,",ratios in component et al., 1984).Field observations(Wilshire and Sh;;;i;, distinctive phases; of an interrelationship 1975;Harteetal., l975;StoschandSeck, 1980;R"J;; they are also suggestive Fer+ and H* in individual minerals' et al., 1984; Wilshire, 1987; Nielson et al., f Sglj h""" between led to a variety of models for mantle metasomatic pro- The goal of our presentresearch is to examine the crys- cesses.There is current debate over the mechanismsihat tal chemistry of mantle-derived amphibole by studying produce the different styles of metasomatism. For ex- its major-element contents, Fe-site occupanciesand va- ample, it is not understood why some styles of metaso- lences,and H* content and partitioning behavior. Data matism are characterizedby the presenceof amphibole, on the isotopic behavior of H are presentedin brief in 0003-004x/93l0910-0968$02.00 968 DYAR ET AL.: Fe3+AND H+ IN MANTLE KAERSUTITE 969

250

200 1.75

o. 1.50 $ r.zs B t.oo

0.50

1.5 3.0 4.5 6 0 7.5 9.0 10.5 12.0 13.5 15.0 Wt% Fe2O3 Fig. 1. FerO. and HrO contentsof oxykaersutitesand mantle amphiboles. Symbols representdata from Barnes (1930), solid circles;Aoki (1963),solid squares;Best (1970),solid triangles; Frg.2. Schematicdrawing ofthe octahedralsites adjacent to Wilshire and Trask (1971),solid diamonds;Aoki (1971),open theH* atomsin thehornblende structure. O sitesare indicated circles;Boettcher and O'Neil (1980), open squares;and Graham by largeoutlined circles and identified by theirnumbers; cation et al. (1984),open triangles.The lack ofa closelyaligned linear sitesare small, solid circles. The H atomsits close to the03 site trend among these variables arises from three main factors: (1) with an O-H distanceof 0.934A lPecharet al., 1989),and so group part polyhedra HrO was, in most cases,determined by weight loss with no dis- theOH takes in thecoordination of both crimination of possibleCO, contribution; (2) FerO, was, in most Ml andM3. cases,detennined by calorimetry, which could not detect contributions from Ti and Mn oxidation; and (3) since the true (1984) examined the major-elementand volatile com- substitution here involves oxidation ofFer+ to Fe3+due to loss ponents of amphibole in xenoliths from the Grand Can- of H* ions, the plot of oxide data obscuresany evidence of a yon and found a strong positive correlation between Ti crystal chemical trend. contents and H+ deficiency.An intriguing study (Brynd- zia et al., 1990) investigated the relationship among fo,, Dyar et al. (1992b) and will be pursued, togerher wirh O OH- content, DD, and Fe3+/Fe,.,in samplesfrom Shep- isotope data, in a later work (Dyar et al., in preparation). pard and Epstein (1970) and Kuroda et al. (1978) and We present here the results of a detailed study of 20 kaer- found only small ranges of dD and Fe3+/Fe,o,;however, sutite samples, in order to assessthe relationships among no primary data were given. Recent work by Deloule and H+, Fe3+, and other major elements within the amphi- coworkers(e.g., Deloule et al., 1990, l99l) has concen- bole structure. trated solely on the isotopic variation displayed by hornblende. Unfortunately, none of these more recent studies B,c,cxcnouNo included complete (published) Fe3* data on all samples, Beginningwith Barnes(1930) and then Winchell (1945), and so other workers (e.g.,Boettcher and O'Neil, 1980) Aoki (1963, 1970; l97l), Best(1970, 1974),and Ernsr could do little more than note a generalinverse correla- and Wai (1970),geochemists have used atomic absorp- tion betweenFerOr/FeO and HrO content(Fig. l). tion and wet chemical analysesto study the dehydration Many crystallographic studies of the kaersutite struc- of amphibole. Incomplete characteization of minor con- ture using X-ray diffraction have been published, includ- stituents such as Cr3* and Tia+ and suspicionsthat Ti3+ ing Papike and Clark (1968),Papike et al. (1969),Kita- might contribute to errors in Fe3+determinations ham- mura and Tokonami (1971), and Hawthorne and Grundy pered these workers in evaluating the detailed crystal (1973).Powder (Jirdk et al., 1986)and single-crystal(Pe- chemistry of their samples. Furthermore, HrO analyses char et al., 1989) neutron diffraction studies of this min- were generally determined by weight loss, a technique eral have further improved our understandingofthe cat- that yielded resultsgreatly biasedby the contents ofother ion distributions and H positions in kaersutite (Table l, volatile phases, especially COr, which are now known Fig. 2). from fluid inclusion studies to be abundant in minerals Additional relevant X-ray data complemented by ex- derived from the upper mantle (seee.g., Roedder, 1965; perimental work has been performed recently by Popp et Bergman,198 1; Andersenet al., 1984). al. (1990)at TexasA&M University (Phillipset a1.,1988, More recent geochemicaland petrologic studies of am- 1989;Clowe et al., 1988).Results showed that dehydro- phibole have focused more closely on major-element genation accompanying heat treatment causespreferen- analysescoupled with studies ofH isotope behavior. Ku- tial ordering oftrivalent cations at the cis-Ml and trans- roda et al. (1978)examined 6D ofhornblendefrom intru- M3 sites. Some trivalent cations in M2 may relocate to sions in alkali basalts;Boettcher and O'Neil (1980) fol- Ml or M3 as oxidation occurs.Fe3+-H+ oxysubstitution lowed with a thorough examination of DD,HrO content, and Al substitution were identified as the primary mech- and D'8Oin amphibole;and Graham et al. (1984)derived anisms accompanying oxidation and dehydrogenation. activation energiesfor H isotope exchange.Matson et al. Popp et al. (1990) surveyedthe literature on Fe3+-rich 970 DYAR ET AL.: FC3*AND H* IN MANTLE KAERSUTITE

Tmle 1. Cationdistributions in kaersutitebased on structurerefinements

Site Hawthorneand Grundy(1973) Pechar et al. (1989) Phillipset al. (1989)

A Na, + Kyand Nao543, and Ko4s y' 0.265Na+ 0.205K 0.43Na T1 0.53Si + 0.47A1 0.57Si + 0.43A1 0.35Ar+ 0.65Si r2 0.94Si + 0.06A1 0.90si + 0.10A1 0.02Ar+ 0.98si M1 0.343Fe3*+ 0.657M9 0-26Ti+ 0.03ca + 0.32FeF*+ 0.39M9 0.41sFe+ 0.585M9 M2 0.056Fe3*+ 0.502M9 + 0.185At+ O.257Ti 0.17A1+ 0.18Fe3*+ 0.65M9 0.214Fe+ 0.321M9 M3 0.235Fe3*+ 0.765M9 0.93M9 + 0.01Fe'++ 0.06Fe3* 0.588Fe+ 0.412M9 M4 0.009Mn+ 0.958Ca+ 0.010Fe3*+ 0.025Mo 0.99Ca+ 0.01Mn 0.91SCa+ 0.099Ca H 0.51H 1.0H ' The variabfesx and y are related by the equation x + 1.8y : 0.802. calcic and subcalcicamphibole and found that the rela- implies low distortion for Fe2* in silicates,high A implies tion betweenthe Fe3+-oxyamphibolecomponent and the high distortion for Fe3+.Therefore we would predict the OH- content of amphibole could not be quantified, al- highest A for Fe3+in the M2 site (assuming no Fe3+in though the lack of large numbers of high-quality amphi- M4) and gradually decreasingA for Fe3+in the order M3 bole analysesgreatly limited their conclusions. to Ml, which can be used in site assignmentsfor the Numerous Mdssbauer studies of amphibole samples observeddoublets. exist in the literature, as summarized by Hawthorne One final comment regarding the interpretation of (1983).As he noted,the interpretationofspectra ofcalcic Mossbauer spectraof amphibole is important within the amphibole minerals, in particular, is "not straightfor- context of the literature. It is well known that the area of ward." Most of the controversy revolves around proper a M6ssbauerdoublet (pair of peaks)is a function of peak interpretation ofthe doublets resolved in the spectrarel- width t, sample saturation G(x), and the Mdssbauer re- ative to their assignmentto the known crystallographic coil-free fraction / Bancroft (1973) used the following sites in the structure. In principle, four doublets corre- formulation for area ratios: spondingto Fe2* in eachof the metal sitesof the structure Ar",, might be observedin a spectrum of Fe'?*-richamphibole. :_ t,'ttt1n^N(Fe3*) Additional doublets correspondingto Fe3+in those sites Ar,,: , are also possible,although Fe3* in the M4 site would be where N is the amount of each species found by wet unlikely except in Ca or Na-deficient samples.Goodman chemical analyses.It follows that C may be expressedas and Wilson (1976), in fact, resolved five doublets in a Ie.,- *.fre,* G(x)o.,- single hornblende spectrum, which they assignedto Fe2* c : in Ml-M4 and to Fel;f . Other workers have observed *.- * 1",-r C1x1..*' smaller numbers of doublets(Burns et al., 1970; Burns Bancroft and Brown (1975) determined that saturation and Greaves,l97l; Bancroftand Brown, 1975; Goldman corrections are unnecessaryfor samplesmixed with sug- and Rossman, 1977) on tremolite-ferroactinolite solu- ar. The Ip.:. irlld fp.:. were also determined to be equal tions. The question of Fe2+ occupancy in M4 remains by them. However, when the Miissbauer data on horn- controversial(e.g., Aldridge et al., 1982);it is apparent blende reponed by Bancroft and Brown (1975) were com- that Fe2+will occupy the M4 site only in the rare cases pared with wet chemical Fe3+/Fe2+ratios determined by where Ca2+ + Na+ * K* is insufficient to fill the site. Dodgeet al. (1968),a C value of l.l3 + 0.02 wasderived Occupancy of Fe3+ in M4 seems, therefore, extremely (on the basis of sevensamples). A later study reported C unlikely. One additional conclusion from these previous valuesof 1.36, 1.25, and 1.24 for a hastingiteand two studies is germane to the discussionat hand: the distor- kaersutite samples,respectively (Whipple, I 972). tions of the M sitesin calcic amphibole, which are known More recently, four samples studied by Schwartz and to decreasein the order M4, most distorted, to M2 to Irving (1978) yielded Fe3+/Fe2+values that were higher M3 to Ml, leastdistorted, can be related to the quadruple when determined by chemical means than by Mdssbauer splittings observedfor Fe2* in the order M4, lowest quad- spectroscopy(resulting in a C value of < l). They attrib- ruple splitting (A), to M2 to M3 to Ml (Hawthorne, 1981). uted the difference to the presenceof Ti3" in their sam- No Mdssbauer studies of Fe3*-rich kaersutite samples ples. The only other comparison of chemical and Mdss- could be found in the literature, with the exception of bauer Fe3*/Fe2+ratios was performed by Bahgat and Schwartzand Irving (1978) and McGuire et al. (1989); Fayek (1982), who derived a C value of l.l5 basedon both of these focused more on the amount of Fe3+than one sample.One of the goalsof this work was to constrain on its distribution. However, on the basis of the increas- this relationship and determine a realistic value for C. ing distortion and decreasingA relationship observedby Hawthorne(1978, 1981),it would be possibleto predict pR-EpARATroN, relative quadrupole splittings of Fe3+doublets if they could S.lvrpr,r sELEcrroN, AND ANALvSES be resolved (i.e., in the spectrum of a Fer*-rich amphi- Hornblende megacrystswere selectedfor this study from bole). As noted by Burns et al. (1985),although high A the extensivecollections of Howard Wilshire and Robert DYAR ET AL.: FC]+AND H* IN MANTLE KAERSUTITE 971

Coleman of the U.S. Geological Survey. The Coleman TABLE2, Hornblendelocalities represented collection has since been curated by the National Muse- DL-s Deadman Lake, Califomia, Hill2237 um of Natural History, Smithsonian Institution. Source DL-7 Deadman Lake, Califomia, Hill2237 localities for all samplesstudied are listed in Table 2; all DL-g Deadman Lake, Califomia, Hil2284 AK-M1 Harrat al Kishb, Saudi Arabia, Coleman locality H271 specimenscame from basalt hosts erupted in continental AK-M2 Harrat al Kishb, Saudi Arabia, Coleman locality H271 rift environments. Small piecesfor thin sectionswere re- AK-M3 Harrat al Kishb, Saudi Arabia, Coleman locality H271 moved from eachmegacryst. Additional pieceswere bro- AK-M4 Harrat al Kishb, Saudi Arabia, Coleman localiryH271 AK-Ms Harrat al Kishb, Saudi Arabia, Coleman locality H274 ken off and ground by hand under acetone (to prevent Ba-s Dish Hill. California oxidation) into fragments < I mm to facilitate hand-pick- Tm Easy Chair Crater, Lunar Crater, Nevada ing of grains. Only pristine hornblende grains were Fr-11 Massif Central, France,just east of Alleyras basalt flow used Fr-12 Massif Central, France,Montgros, trom oxidizedvent agglom- for this study; any grains containing visible alteration, erate inclusions,or impurities wererejected. Approximately 680 84-BR Dreiser weiher, Germany, Briick-Raders cone, cognate SABM Bullenmenimaar, Victoria, Australia mg of separateswere prepared: 150 mg for Mdssbauer 86-LEH' Lake Eacham maar. Queensland,Australia analyses,100 mg for PIGE, 30 mg for O isotope study, H366A Harrat Hutaymah, Saudi Arabia and 400 mg for H+ extraction. In addition, 300-mg ali- H366-92 Harrat Hutaymah, Saudi Arabia S.C. San Carlos,Arizona, USNM 111312.0023 quots of metamorphic amphibole samples (Coscaet al., K.N.Z. Kakanui,New Zealand,USNM 109647/50 1991) were obtained from Michael Coscafor the com- Eifel Eifel Volcanic Field, USNM 114856.0006 parative study of C. Note: although the sample numbers listed here may resemblenumbers Electron microprobe analyseswere performed on the given in previousstudies, such as Boettcher and O'Neil (1980),they bear 733 electron microprobe at SouthernMethodist no relation to previouslystudied samples except for locality. Jeol Uni- ' Sample 86-LEHconsists of amphibolegrains, rather than a megacryst, versity, using a Krisel automation package.Routine an- separated from a spinel peridotite xenolith found at this locality. alytical conditions were used: l5-keV accelerationvolt- age, 20-nA beam current measured on a Faraday cup, 30-s count times, and the Bence-Albeecorrection routine Samplesare then fused in an induction furnaceto liberate processes supplied by Krisel. At least five points were analyzed on structural HrO. Distillation involving transfer gases each megacryst, and only unzoned samples were used. of evolved through a seriesof traps employing liq- Compositionsfor AK-Ml, AK-M2, and H366A are de- uid N, and a slush of methanol and dry ice are used to scribed in McGuire et al. (1989); the analysis used for separateHrO molecules effectively from other condens- gases. passed H366-92 was made using XRFtechniques at Washington able and noncondensable HrO vapor is then State University. Analytical errors are +0.5-2o/o for ma- over hot U (>750 "C) to liberate free H+. A Hg-piston jor elementsand + l0-200/o for minor elements. Toepler pump is used to collect vapor in a volumetrically yield Mdssbauer analyses were performed in the Mineral calibratedreservoir for measurement.For this study, SpectroscopyLaboratory at the University of Oregon. A every fifth sample was analyzedtwice; distilled HrO was source of 50-20 mCi 57Coin Pd was used on an Austin used for calibration. All data are corrected for a sample (roughly Science Associates constant acceleration spectrometer. blank of 2.5 tor' 0.5-2o/oof the total wto/oHrO, volume HrO vapor). Resultswere calibrated againstan a-Fe foil of 6 pm thick- depending on sample size and of =0.1 nessand 990/opurity. Spectrawere fitted using a version Errors are estimated at wto/oHrO based on repli- of the program Stone modified to run on IBM and com- cations. patible personal computers. The program usesa nonlin- F concentrations were determined by external-beam proton-induced (PIGE) ear regressionprocedure with a facility for constraining 7-ray emission analysis at the Van Accelerator Lab- any set of parametersor linear combination of parame- University of Kentucky de Graaf proton reF ters. Lorentzian line shapeswere usedfor resolving peaks, oratory. The inelastic scattering reactions on proton as there was no statisticaljustification for the addition of at bombarding eneryiesof 3.5 and 2.5 MeV were a Gaussian component to the peak shapes.Fitting pro- utilized. Details of the method are included in Wong et (1992). ppm with ceduresin generalfollowed those describedin Dyar et al. al. Minimum detection limits of 5 stan- ppm (1989) and McGuire et al. (1989). A statisticalbest fit dard deviationsof20-140 were obtained. was obtained for each model for each spectrum using the Rnsur,rs y2 and Misfit parameters; practical application of these parameters is discussedelsewhere (Dyar, 1984). Errors Miissbauer data are estimated at +3o/ofor doublet areasand +0.02 mm/s Interpretation of the Mdssbauerdata measuredfor this for peak width, isomer shift, and quadrupole splitting. study proves extremely difficult becauseofthe wide range H contents were determined by means of a procedure of the Fe3+ contents of the samples (from 100 to 27o/o for collecting and measuring all the structural H using a Fe3+). Both Fe3+and Fe2+can occupy (in theory) any of volumetric measurement of HrO vapor extracted from the four M sites in the structure, although Fe3+ in M4 silicates(see Bigeleisen et al., 1952; and Holdaway et al., would be unlikely in this composition range.This would 1986, for details of the technique).In general,clean min- give rise to a total of eight possibledoublets (realistically, eral separatesare degassedunder vacuum for at least 8 h seven) in each spectrum. In practice, it is impossible to at 50-85 oCto evaporateadsorbed atmospheric moisture. resolve those doublets in a statistically meaningful fash- 972 DYAR ET AL.: Fe3* AND H+ IN MANTLE KAERSUTITE

TABLE3. Mdssbauer results for six peak fits

Fe3* Sample os. r.s. o.s. DL-s 0.38 0.78 0 54 38 1.13 2.56 0.33 29 1.12 2.05 0.41 32 0.06 0.01 621 DL-7 0.42 0.69 0 49 46 1.12 2.56 0.32 25 1.08 2:t3 0.41 29 0.08 0.01 735 DL-g 0.39 1.40 0.s1 43 0.40 0.85 0.43 56 1.06 2.38 0.30 1 0.03 0.01 561 AK-M1 0.38 o.74 0 49 32 1.13 2.52 0.33 30 1.12 2.01 0.40 38 0.04 0.01 591 AK-M2 0.38 0.75 0.49 29 1.13 2.51 0.34 33 1.13 1.99 0.38 38 0.00 0.00 507 AK-M3 0.38 0.73 0.50 32 1.13 2.51 0.35 33 1.12 2.00 0.39 35 0.12 0.01 812 AK-M4 0 38 o.71 0.s4 33 1.13 2.48 0.3s 35 1.12 1.96 0.37 32 0.13 0.01 812 AK-Ms 0.38 0.74 0 54 41 1.14 2.54 0.30 23 1.12 2.02 0.40 36 0.19 0.02 819 Ba-S 0.38 0.81 0 55 40 1.'t2 2.57 0.37 30 1.'t2 2.04 0.41 30 0.01 0.01 531 Tm 0.41 0.94 0 57 80 1.03 2.70 0.27 4 1.12 2.06 0.38 16 -0.01 -0.01 531 Fr-l1 0.41 0.90 0.57 76 1.07 2.59 0.34 12 1.10 2.05 0.34 12 0.01 0.00 532 Fr-12 0.38 1.56 0.44 23 0.40 1.16 0.46 53 0.40 0 70 0.34 24 0.00 0.00 515 84-BR 0.38 0.84 0.56 49 1.12 2.62 0.34 29 1.10 2.13 0.40 22 0.13 0.01 933 86-BM O.37 0.76 0.55 27 1.13 2.s6 0.33 33 1.13 2.O4 0.43 40 0.03 0.01 566 86-LEH 0.40 0.75 0.53 31 1.13 256 0.35 32 1.13 1.99 o.42 37 0.09 0.01 ott H366A 0.39 0.79 0.45 28 1.14 2.59 0.30 32 1.15 2.08 0.45 44 0.06 0.01 585 H366-92 0.39 1.31 0.59 50 0.40 0.81 0.42 43 1.09 2.40 o.27 4 1.1 1 1.92 023 J 0.00 0.00 524 San Carlos 0.43 0.63 0.61 37 1.'t2 2.40 0.40 24 1.08 1.90 052 39 o.02 0.01 527 Kakanui O.37 0.72 0.57 33 1.13 2.60 0.33 31 1.12 2.09 042 36 0.03 0.01 550 Eifel 0.37 0.73 0.55 46 1.12 2.56 0.31 24 1.11 2.O5 0.40 30 0.02 0.01 525 Average 0.39 0.87 0.51 41.3 1.06 2.41 0.31 24.4 1 11 2.05 0.39 28.9 s.D. 0.o2 0.24 0.06 14.2 0.25 0.57 0.08 10.7 0.02 0.10 0 06 12.1 ssA-s 0.35 0.79 0.35 33 1.13 2.70 o.28 33 1.16 2.18 0.43 34 0 01 0.01 525 HL862C 0.36 0.77 0.37 38 1.13 2.69 0.29 36 1 17 2.13 0.44 26 0.03 0.01 537 MR-865A 0.36 0.76 0.34 25 1.13 2.68 0.30 41 1.15 2.06 046 34 0.01 0.01 521 MrN864 0.36 0.77 0.37 38 1.13 2.69 o.29 36 1 17 2.13 o.44 26 0 03 0.01 536 ssA-10 0.35 0.79 0.38 28 1.14 2.67 o.27 32 1 14 2.15 0.42 40 o.02 0.01 522 FA86-1 0.3s o.78 0.37 21 1.13 2.59 0.31 40 1.13 2.O7 0.41 39 0.02 0.01 528 H18611 0.34 0.81 0.36 36 1.13 2.70 0.26 29 1 17 2.20 0.40 3s 0 01 0.01 520 sAS-13 0.37 0.76 0.41 38 1.13 2.62 0.30 34 1.13 2.O9 0.40 28 0.04 0.02 546 Nofe. l.S. : isomer shift, in millimetersper second. Valuesgiven are t0 02 mm/s (Dyar, 1984) O.S. : quadrupolesplitting, in millimetersper second. Valuesgiven are +0.02 mm/s (Dyar,1984). W : peak width,in millimetersper second.Values given are +0.02 mm/s(Dyar, 1984). A : peak area,in percent of total area. Values given are +27o of the total area (Dyar, 1984).Other abbreviations:o7"y;s : percentof Misfit (Ruby, 1973); o/.Un: percent of uncertaintyof fit (Ruby, 1973); x2: chi squared; s.D. : standard deviation. Mdssbauer parametersare referencedto Fe metal foil calibration.

ion becauseof their heavily overlapping peak positions. even these fits are questionable becausethe peak sepa- As noted above, previous workers have resolveddoublets ration is lessthan that required (t/zpeak width minimum) assignedto as many as four Fe2+sites and one Fe3+site. for well-resolved peaks(Bancroft, 1973).The assignment Such assignmentsare difficult to determine in our mantle of such doublets is complicated further by the fact that kaersutite suite becausenone of these hornblende sam- the multiplicity of types of environments (as discrimi- ples has a stoichiometric amount of H+. Some of the nated by the Miissbauer efect) no longer correspondsto cations in the Ml and M3 sites will not have their ex- the known site multiplicity of MI:M2:M3:M4 equal to pectedgeometries, with bonding to four O atoms and two 2'.2:l:2. OH molecules; instead, they may be bonded to five O In light of theseuncertainties in interpretation of mul- atoms and only one OH molecule or to six O atoms and tiple doublet fits, the strategy used in the present study no OH molecules.Because the M6ssbauermeasurements was to fit six peaks initially (a three doublet fit) to each are sensitive primarily to coordination environment and spectrum.We were able to resolvethree doublets in every chargedensity, the Mtissbauerspectrum of a H+-deficient spectnrm,as shown in Table 3 and Figure 3. This method hornblende might theoretically yield doublets corre- resulted in low values of 12, percent Misfit, and percent sponding to Fe atoms in up to eight types of polyhedra: uncertainty for each spectrum and yielded a relatively M2,M4, Ml, and M3 with stoichiometricH+, Ml and consistent set of peak parameters,as shown by the small M3 with one H*, and Ml and M3 with no H+. Either standard deviations in Table 3. The values of percent Fe3+or Fe2+may occupy any of those sites, giving rise Fe3+that resulted are thus a valid representation of the (theoretically)to as many as 16 doublets in the spectrum total percentofFe3* contributions to the spectra.It should of even a slightly H+-deficient hornblende. In practice, also be noted that the number of doublets fitted to model we were able to resolve no more than seven doublets in Fe3+does not affect the final percent ofFe3+ determined. any spectnrmbecause ofthe heavy overlap betweenpeaks; For thesespectra, even six doublet fits yielded exactly the DYAR ET AL.: Fe3+AND H" IN MANTLE KAERSUTITE 9',13

100 20 TaeLE4. Resultsof comparativestudy of amphiboles 100 00 U.U. S.M.U. U.U U.O. o/oFe3+ ohFe3+ C 99 80 Sample Wt% HrO Wt% H,O

DJA-C 1.68 1.63 29 33 1.210 99 60 't.92 HL862C 1.88 35 38 1.135 99 40 2.26 MR-865A 1.78 1.68 22 25 1.178 o 99 20 F 1.62 MrN864 180 1.73 33 38 1.264 99 00 SSA-10 1.75 ''.95142 23 28 1 314 G FA86-1 1.23 18 21 1 236 F 98 80 HL8611 1.64 31 36 1.264 1.143 98 60 SAS-13 1.41 1.38 35 38 Average 1.218+ 0.059 98 40 Nofe; analysesperformed at Universityof Utah (U.U.),Southern Meth- 98 20 odist University(S M.U.),and Universityof Oregon(U.O.) * Insufficientsample was availablefor analysis s8 00

Fig. 3. Mtissbauerspectrum of sampleBa-5, fitted with a been biasedby lack of good H analysesfor their samples. r6rFe2+ threedoublet model. Two doubletscorresponding to and However, our value of 1.22 agreeswell with the two kaer- I6lFe3+ onedoublet corresponding to areshown. A threedoublet sutite C values determined by Whipple (1972): 1.25 and modelwas used to fit all spectrain this study;results are given 1.24. Therefore, we have applied our correction factor of in Table3 and usedin the formularecalculations in Table4. 1.22 to all the Mdssbauerresults presentedin this paper, and corrected values for percent of Fe3+have been used in Table 5 and in related figures. samepercent ofFe3* total peak area as the three doublet fits; i.e., the percent Fe3* determined by M6ssbauer is model-independent for these samples. fI+ contents However, the three doublet model does not allow ac- H analyseswere obtained for sevenofthe eight samples curate interpretation of the location of the Fe atoms in provided from the Coscaet al. (1991)study of metamor- the structure becausethe area contained in each doublet phic samples;results are shown in Table 4. The largest may represent contributions from Fe in more than one difference between the two laboratories is the 0.33 wo/o site (becauseof the similarity between an Ml or M3 site HrO differencefor sample SSA-10. The remaining anal- in an anhydrous amphibole and a normal M2 site, as ysesagree well within the commonly stated error bar of discussedabove). This conclusion is supported by the +0. I wto/oHrO. The results suggestthat there is no sub- peak widths given in Table 3, which are much larger than stantive diference between the yields obtained using the the line widths typically observedfor Fe in single sites in apparatusat Utah (asused by Coscaet al., l99l) and the silicates:0.27-0.28 mm/s for Fe2* and 0.28-0.35 mm/s line usedat S.M.U. for Fe3*(Bancroft, 1973). Only in the caseof sampleFr- 12 were we able to resolve three unique doublets corre- F contents sponding to three sites (for Fe3+ We interpret the dou- ). PIGE analysesof the kaersutite samples showed only blets,which have 6 : 0.38, 0.40, and 0.40 mm/s and A minor variations in F contents,ranging from a low of 167 : 1.56,1.16, and 0.70mm/s, to representFe3+ occupan- ppm for the San Carlos sample to a high of 3054 ppm cy in M2, M3, and Ml sites,respectively, as explained for the Eifel specimen. In contrast with conclusions earlier. For other samples, four, five, and six doublet reachedby Kyser (1992)on a suiteof kaersutitesamples, models were also fitted to each spectrum as far as possi- we find that halogen variation alone cannot explain the ble. However, the heavy peak overlap that results from range of observed H* contents in our sample suite. such fits renders them mathematically possible but crystal-chemically unbelievable. In order to evaluate the true Fe3+contents of our man- Chemical formulae tle kaersutite samples,it was necessaryto establish a vi- Chemical compositions for all samplesand for the av- able value for C the correction factor. Eight samplesfrom erageof all samplesstudied were determined from probe, a metamorphicsuite analyzed by Coscaet al. (1991)were Mtissbauer, PIGE, and H-extraction data and are given fitted with six peak fits, as shown in the bottom portion in Table 5. The Fe and FerO, contents were calculated of Table 3. Direct comparison of the results is shown in from the probe data using the corrected Mtissbauermea- Table 4. An averagecorrection factor, or C value, of 1.22 surementsfrom the six peak fits given in Table 3 and the + 0.06 was obtained. Previous work by Bancroft and C value of 1.22. Data were normalized to formula units Brown (1975)and Bahgatand Fayek(1982), who derived with 24 O atoms and adiusted to account for the contri- valuesof l.l3 + 0.02,and 1.15,respectively, may have butions of F and Cl. 974 DYAR ET AL.: FEsf AND H] IN MANTLE KAERSUTITE

Tlel-e 5. Hornblendecompositions

DL.5 DL.7 DL-g AK-M1 AK-M2 AK-M3 AK-M4 AK-Ms Ba-s tm Fr-11 Ft-12 sio, 38.91 39.85 39.01 39.73 39.93 4028 40.01 39.20 40.41 39.52 40.05 39.82 3929 Al,o3 14.64 14.39 14.26 14.85 14.83 1482 15.20 14.11 14.60 14.03 14.48 14.53 13.85 Tio, 4.84 4.51 4.52 5.21 5.31 5.32 4.98 5.14 4.52 5.87 4.49 4.30 3.27 FeO 7.92 8.17 0.14 8.20 8.27 8.67 8.83 7.95 6.89 2.84 3.58 0.00 5.88 Fe.O. 4.43 6.36 13.27 3.53 3.07 3.73 3.97 5.04 4.37 10.40 10.34 13.95 5.15 MgO 12.16 11.26 1231 12.11 12.84 12.22 1201 11 .68 13.07 11.71 11.92 12.12 13.77 MnO 0.17 o.20 015 0.08 0.09 0.13 0.12 0.11 0.14 0.17 0.12 014 0.10 CaO 11.02 10.82 1071 9.98 9.87 9.79 9.39 9 37 10.36 11.10 10.73 10.82 12.10 NarO 2.61 2.74 2.61 2.86 2.91 3.30 3.37 3.10 2.66 2.81 2.73 277 1.96 K.O 1.66 1.50 1.64 1 14 1.10 1.24 0.92 1.12 1.55 1.29 1.35 1.38 2.18 Cr,O" 0.03 0.00 0.02 0.03 0-o2 0.04 0.00 0.00 0.01 0.00 o.02 0.02 0.02 F 0.14 o.17 0.20 0.15 0.17 0.16 0.12 0.13 0.16 0.24 0.16 0.11 0.22 cr 0.01 0.02 0.02 0.00 0.00 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 H,O 1.05 0.90 0.16 1.01 1.03 0.96 0.92 0.84 1.03 0.07 0.48 0.17 1.32 Total 99.59 100.89 99.02 98.88 99.44 100.66 99.84 97.80 99.77 100.05 r00.45 100.13 99.11 Cations to 24 O atoms minus F and Cl atoms Si 5.81 5.89 5.85 5.92 5.90 5.92 5.93 5 94 5.95 5 88 5.92 5.96 5.84 AI 2.58 2.51 2.52 2.61 2.58 2.57 2.65 2.52 2.54 2.46 2.52 2.57 2.43 Ti 0.54 0.50 0.51 0.58 0.59 0.59 0.50 't.02 0.55 0.59 0.50 0.66 0.48 0.37 Fe,* 0.99 103 0.21 1.02 1.06 1.09 1.02 0.86 0.46 0.54 0.00 0.75 Fe3* 0.49 0 69 1.31 0.40 0.34 0.41 0.44 0.57 0 48 1.06 1 06 1.37 0.56 Mg 2.71 2.48 2.75 2.69 2.83 2 68 2.65 2.64 2.87 2.60 2.63 2.71 3.05 Mn 002 0.03 0.02 0.01 0.01 0.02 0.02 0.01 0.02 0.02 0.02 0.02 0.01 1.76 1.71 1.72 1.59 1.56 1.54 1.49 1.52 1.64 1.77 1.70 1.74 1.93 Na 0.76 0.79 0.76 0.83 0.83 0.94 0.97 0.91 0.76 0.81 0.78 0.80 0.57 K 0.32 0.28 0.31 0.22 0.21 0.23 017 0.22 o.29 0.24 0.25 0.26 0.41 0.00 0.00 0 00 0.00 0.00 0 00 0.00 0.06 0.00 0.00 0.00 0.00 0.00 F 0.07 0.08 0.09 0.07 0.08 0 07 0.06 0.06 0.08 0.11 0.08 0.05 0.10 cl 0.00 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 H 1.05 0.89 0.16 1.00 1.02 0.94 0.91 0.85 1.01 0.06 0.47 0.17 1.31

eral agreement the H+ calculatedin this way and DrscussroN between the measured H+, the calculated values are consistently Measured vs. calculated Fe3+and H+ too high. In addition to the problems associatedwith as- Calculations of amphibole Fe3+contents basedon var- sumptions of perfect stoichiometry and charge balance ious charge balance arguments abound in the literature discussed by Hawthorne (1983), such a recalculation (e.9.,Stout, 1972; Paplke et al., 1974). Schumacher( I 99 I ) schemeis also strongly biasedby its dependenceon mea- discussedthe effectsof such empirical corrections on am- sured Si content, an element that is particularly difficult phibole geothermometry, and Hawthorne (1983) thor- to measureaccurately by microprobe. Note also that this oughly covered the many reasonswhy there is "(unfor- recalculation depends on accurate assessmentof Fe3* tunately) no significant correlation" between Fe3+values contents. thus calculatedand those that have been measured.This conclusion is also true for our data, as shown in Figure Mechanisms for Fe3* substitution 4a, where Fe3+is calculated solely on the basis ofcharge As discussedabove, Popp et al. (1990)have identified balance. three likely mechanismsfor the accommodation of Fe3+ Clowe et al. (1988) describeda method for calculating in the amphibole structure,which can be expressedas the the oxyamphibole content of amphibole for which both following substitutionvectors: (l) Fe'z+H+Fe1l,oxysub- electron microprobe and Fe3*/Fe,o,data are available (in stitution (or deprotonation) with H+ providing charge their case,the Fe3+/Fe,".values were determined by cal- balancefor the Fe2+, (2) t6lAl3+Fe3+,substitution of Al3* orimetry). The processinvolves correction of an analysis for Fe3+,(3) tAr(K,Na)Fe2+-tAlE_rFe1],replacement of an based on 23 O atoms by recalculating the probe-deter- Fe3* and a vacant A site by a Fe2* and either a K or Na mined total elementalwto/o Fe to elementalFe3+ and Fe2+, in the A site. Each of these relationships is plotted in followed by conversionto weight percentoxide and moles Figure 5. It is immediately obvious from the nearly l:l of Fe3+and Fe2+,and renormalization to 23 O atoms + inverse correlation betweenH* and Fe3+that the bulk of YzFe3+pfu. They note that this method makes the as- the Fe3+variation can be explained by the relationship sumption that Fe3+is present as an oxyamphibole com- representingthe oxysubstitution mechanism l. A linear ponent, with one Fe3* for every H* lost. regressionfit to the plot of H* vs. Fe3* yields a slope of Their method was tested on the samplesin the present -0.80 + 0.10, with a correlationcoefficient of 0.88. The data set using the Fe3+/Fe,o,values determined by Miiss- deviations from a l:l inverse relationship (slope of - l) bauer and correctedfor recoil-free fraction; the resultsare are probably indicative of other less important substitu- shown in Figure 4b. Although this figure showssome gen- tions and of the limitations of the electron microprobe DYAR ET AL.: Fe3* AND H+ IN MANTLE KAERSUTITE 975

Tlp,te S.-Continued 100

86-BM 86-LEH H366A H366-92 S.C. K.N.Z. 90 41.91 41.52 39.76 39.61 40.87 3932 38.00 14.46 15.05 14.04 14.77 14.76 1354 13.69 80 3.23 2.50 4.34 5.19 3.29 4 89 +.o I 6.11 7.51 9.97 o.72 7.50 7.85 8.09 70 2.06 3.07 3.37 12.34 4.01 3.53 6.30 !t o 15.16 14.70 11.29 12.41 12.90 12.31 1143 (u 011 0.10 0.15 0j2 0.17 0.14 031 60 10.74 10.17 't0.24 10.43 10.94 o 9.96 10.27 (d 3.18 2.88 2.90 2.82 2.97 2.56 2.18 o 50 1.21 0.96 1.52 1.44 1.05 2.06 2.37 + o.77 012 0.00 0.00 0.00 0.00 0.00 o 40 0.08 0.07 0.15 0.16 0.o2 0.18 0.31 l!s 0.00 0.00 0.00 0.00 0.01 0.02 0.07 30 1.r3 1.23 1.18 0.13 0.69 1.14 1.18 100.15 9988 98.91 100.14 99.18 97.50 98.57 20 Cationsto 24 O atomsminus F andCl atoms 6.10 6 07 5 98 5.85 6.10 5.95 574 10 2.48 2 59 2.49 2.57 2.60 2.42 2.44 0.35 0.27 0.49 0.58 0.37 0.56 0.50 0 0.74 0.92 1.25 0.09 0.94 0.99 1.04 10 20 30 40 50 60 70 80 90 100 o.23 0.34 0.39 1.37 0.45 0.40 0.70 3.29 3.20 2.53 2.73 2.87 278 z-51 o/oFe3+(measured/corrected for c) 0.01 0.01 0.o2 0.02 0.02 0.02 0.04 1.67 1.59 1.65 1.75 1.62 1.66 1.68 0.90 0.82 0.85 0.81 0 86 0.75 0.64 2.1 0.22 0 18 0 29 0.27 0.20 0.40 0.46 0 09 0.01 0.00 0.00 0.00 0.00 0.00 0.04 0.03 0.07 0.07 0.01 0.09 0.15 1.9 0.00 0.00 0.00 0.00 0.00 0.01 0.02 T 1.10 1.20 1.18 0.13 0.69 115 1.19 1.7 I f 1.5 do(D and wet chemical techniquesin performing such sensitive at =(, 1.3 measurements. ()(u Examination of Figure 5 yields two further observa- + 1.1 tions relating to H+ contents.At one extreme,where Fe3* x reachesits theoretical minimum of zero, H+ contents do not equal the stoichiometric 2 H+ pfu; instead, the intercept is 1.32 + 0.18. This observationunderscores the 0.7 idea that bulk composition, particularly substitution of tetravalent ions such as Tio* on trivalent sites and tri- 0.5 L- valent ions on divalent sites,probably controlsthe amount 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 of H+ that can be contained in the amphibole structure H+Measured, pfu at mantle pressuresand temperatures.At the other ex- Fig.4. Calculatedvs. measuredFe3" (a) and H+ (b) contents treme, the intercept at the hypothetical zero H+ is very of amphiboles.Calculations follow the proceduredescribed in close to three data points representing almost H+-free Cloweet al. (1988).Measured and correctedFe3* data points : amphibole. Thus, true oxykaersutite probably exists and areMrissbauer data corrected for C 1.22. is certainly stable at surfacepressures and temperatures.

Irrpr-rcnrroNs FoR MANTLE METASoMATTSM sampleswere dehydrogenatedduring a metasomatic event As indicated above, the H+/Fe3+ data plotted in Figure in the upper mantle subsequentto amphibole growth but 5 display a near l: I inverse relationship between the H+ prior to entrainment and transport; or (4) the amphibole and Fe3+contents in the kaersutite samples.The data fall samplescrystallized in regions of the upper mantle that into two clusters:the first, Fe3+contents ofabout 0.4 pfu were chemically distinct (in terms of O and/or H activi- and H* contentsof about 1.1 pfu, generallyshow <250lo ty). Each ofthese interpretations will be discussedbelow dehydrogenation;the second,with much higher Fe3* (near in light of petrographic observations of the megacrysts 1.3 pfu) and lower H+ (near 0. I pfu), show about 900/o and their host rocks, as well as experimental constraints hydrogenation.There are four possibleinterpretations of on dehydrogenationof the hornblende. these data: (l) the amphibole sampleswere partially de- hydrogenated during ascent from the upper mantle; (2) Kinetics of dehydrogenation dehydrogenation occurred after the amphibole samples The correlation between H* and Fe3* in these horn- were erupted onto the Earth's surface;(3) the amphibole blende samplesmay be compared with observationsfrom 976 DYAR ET AL.: Fe3+AND H* IN MANTLE KAERSUTITE

1.5 essary for partial or total dehydrogenation of the kaer- 1.4 sutite crystals.Althoueh they determinedtheir diffirsivities t.3 from isotope exchangeexperiments, as opposed to de- 1.2 hydrogenationexperiments, the generalagreement of their 1.1 data with the results from the experimentsof Skogbyand Rossman(1989) suggests that the kineticsof the two pro- 1.0 cessesare similar. We modeled the dehydrogenationpro- r 0.9 o- cessassuming the reaction given above for olivine and o 0.8 using a simple calculation of the diffusivity for H neces- 3 0.7 sary to produce partial dehydrogenation of hornblende C' o 0.6 crystals of a particular grain size. In this calculation, we assumedthat diffusion is essentiallyone-dimensional and obeys a relationship determined for difusion in a finite slab with an infinite sink for H at the grain edges(see e.g.,Schmalzried, l98l;p. 85). It is assumedin this calculation that dehydrogenationis driven by the partition- 0.1 F I ing of H/HrO from the crystalline hornblende into the f T 0-0 L surrounding melt phase, and that the rate of defect re- 0.0 o.4 1.0 1.2 1.4 equilibration on the major ion sublatticesin the crystals Fe3+pfu is slow. Although no data exist for reequilibration of point Fig.5. Fe3*substitution mechanisms for amphiboles,as sug- defectsin amphibole,Mackwell et al. (1988)have shown gestedby Poppet al. (1990). that, for olivine crystalsat 1200 'C, defectreequilibration on the octahedral cation sublattice takes on the order of l0 d for crystalswith a grain size of about l0 mm. experiments investigating the hydrogenation of olivine crystalsat temperaturesfrom 800 to 1000 oC.From their Dehydrogenationof kaersutite? results,Mackwell and Kohlstedt (1990)concluded that H The kaersutite samplesinvestigated in this study were was incorporated as protons in the crystal structure of generally crystals on the order of l0 mm, although this olivine through a reaction involving the consumption of figure should probably be taken as an underestimate,as polarons (the excesscharge on the Fe3+): most of the crystals showed evidence of surface breakdown during transport. The partial breakdown of the *[H,]*" + Felf + HJr + Fe3l hornblende also provides some constraints on the tem- resulting in a predicted l:l inverse correlation between peratures (probably 1100-1250 'C) and time scales H* and Fe3+.At higher temperatures(1300'C), Bai and (probably betweenseveral hours and a month) of entrain- Kohlstedt (1992\ have shown that defect relaxation on ment and ascent(e.g., Spera, 1984). In Figure 6, the hor- the major element sublatticespermits a decoupling of the izontal extent ofthe boxes in the upper left indicates this H+ and Fe3*. A similar process might be expected to temperature range, while the vertical extent of each box occur in reversein hornblende, with dehydrogenationre- indicates the time span from 3 h (bottom) to 30 d (top) sulting in a parallel increasein Fe3+at moderate temper- for l0 mm of hornblendeto reach 10, 25, or 900/odehy- atures (Dyar eL al., 1992b). drogenation. Although the cluster of H+-Fe3+ data with Graham et al. (1984)have measuredH diffusivitiesin lesserdegrees (<25o/o) ofdehydrogenation (upper left, Fig. hornblende experimentally from isotope exchangemea- 5) might be explained by partial dehydrogenationduring surementsat temperaturesin the range 350-850 "C. We ascent,the data that show almost total dehydrogenation have replottedthe data ofGraham et al. (1984) for par- (lower right, Fig. 5) would require difftrsivities that are gasitic hornblende 322 and ferroan pargasitichornblende far in excessofthose predicted by the results ofGraham 6099 in Figure 6. We seeno reasonwhy there should be et al. (1984).This latter observationmay be interpreted significant differencesin difftrsivities for hornblende sam- as indicating that the Graham et al. (1984)data are not ples of such similar compositions and thereforehave gen- valid at the higher temperaturesof the hornblende source erated linear regressionfits to their data for both horn- region. However, it seems unlikely that the diffirsivity blende samples assuming plate and cylinder geometries. would changeby 3-4 orders of magnitude over the range Also plotted on the figure is a data point calculated from from 850 "C in the experimentsto 1000-1200'C in the the infrared observationsof Skogbyand Rossman (1989) Earth, particularly given the similar chemical composi- for the samplethat they dehydratedat room pressureand tions of the two suites of hornblende specimens. 700 "C in air. This result is in good agreementwith those The right axis for Figure 6 shows the amount of time ofGraham et al. (1984). necessaryfor 900/odehydrogenation ofa hornblende with As the major element compositions of the hornblende a l0 mm grain size.If we extrapolatethe data of Graham samples investigated in this study are quite similar to et al. (1984)to 1200'C, the time requiredfor significant those of Graham et al. (1984), we extrapolated their re- loss of H (900/0)from the kaersutite samples is on the sults to higher temperaturesto determine the times nec- order of 3-30 yr. Thus, despite our earlier speculation DYAR ET AL.: FC3*AND H* IN MANTLE KAERSUTITE 977

(Dyar et al.,1992b), it seemsunlikely that the entire rang- T (.C) es of H* and Fe3* observedin this hornblende suite can -8 1200 900 600 300 be explainedby dehydrogenationduring transport, as only limited H loss is possible by this mechanism. The secondpossible explanation for the observedrange Hornblende (GHS) in H+/Fe3+would be that the kaersutitesamples main- -10 Plates tained a sufficiently high temperature over a protracted 90% c 322 period subsequentto emplacementon the Earth's surface, ^ 6099 during which time dehydrogenation could occur. Al- 25% Cylinders though this possibility may be plausible for megacrysts _19 x BZZ entrained in a thick sequenceof lava deposits or em- (h + 6099 placed adjacent to a long-lived lava lake, it is unable to LO% account for the low H* observed in samples that show F A 4a evidenceof eruption as ejecta. -74 The third possibility, that of a long-term interaction +) with a mantle fluid after crystallization but prior to en- blt trainment is also unsupported by petrographic analysis. o0 For a silicate melt to produce nearly total dehydrogena- 8- -16 \o tion of kaersutite would require an interaction over a period of years that should leave clear petrographic evidence and zoning in the major element compositions. X\ Such evidence has been looked for but is not observed. -rtt Hornblende (SR) The final explanation requires distinct chemical difer- X\ encesin the metasomatic fluids from which the various a Plates t( hornblende samples crystallized in the subcontinental 10 mantle, in accord with the recent observations of Ionov _96 and Wood (1992), and Amundsen and Neumann (1992). 0.6 0.8 1.0 1.2 1.4 1.6 1.8 Thus, for the hornblende samples in the lower right in ro3/r (/K) Figure 5, the high Fe3* contents might reflect relatively Fig. 6. A plot ofH diffusivity vs. temperature for pargasitic more oxidizing conditions in the metasomatic fluid, with hornblende using the data ofGraham et al. (1984) (GHS) and a the low H+ contents occurring becauseH incorporation value inferred from the observations of Skogby and Rossman would not be required for chargebalance during growth: (1989)(SR). The boxesin the upper left ofthe figureshow the the hornblende would grow as oxykaersutite.The suite of diffusivities necessaryto produce the percentageindicated ofH hornblende samples in the upper left of Figure 5 could loss for l0-mm hornblendes over the temperature range from 'C then be explained as normal kaersutite specimens,some I100 to 1250 on a time scaleof 3 h to 30 d. of which partially dehydrogenatedduring entrainment and ascent.Although both types of kaersutite may be found in the samegeographic locality (e.g.,the Saudi and Dead- probably indicative of chemical heterogeneity of the man Lake samples),the entraining basaltic melt traversed mantle source regions. large regions of the upper mantle during its ascentto the surface and may have sampled hornblende megacrysts CoNcr-usroNs from a number of quite distinct chemical reservoirs.Such l. Mdssbauerspectra measured on a suite of 20 mantle a model requireseither that the metasomaticfluids varied kaersutite samples reveal a range of Fe3+contents from chemically over time within the region sampled by the 1000/0Fe3+ to only 2lo/oFe3+. Spectra can be interpreted basalt, or that local heterogeneitiesin the chemical sig- with simple three doublet models. More complicated fits nature of the metasomaticfluids occur over a spatial scale are mathematically possible but lack physical relevance of a few tens of kilometers or less. becauseheavy overlap between peaks renders the resul- The experimental data clearly indicate that significant tant fits nonunique. dehydration of the hornblende is unlikely on the time 2. Comparison with wet chemical results of Fe3+/Fe2* scalesof mobilization and transport of hornblende mega- determinations on a suite of metamorphic amphibole crysts with crystal sizes near l0 mm. Thus, we concur samples permits calculation of an improved value of C with Graham et al. (1984) that "amphibole crystalsin : 1.22 for the recoil-free fraction correction factor in rapidly quenched lavas will, in the absenceof extensive kaersutite. oxidation of alteration, preserve their high-temperature 3. H+ contents as determined by U extraction tech- (magmatic) H isotope composition, and they should niques range from 0.07 to 1.32 wto/oHrO. Careful deter- therefore preserve useful information on the H isotope mination of F by PIGE spectroscopyand Cl by electron compositions and origins of 'magmatic' waters." Conse- microprobe showsthat theseelements are presentin only quently, we would argue that, at least for the kaersutite minor amounts; thus, H+ variation cannot be explained samplesinvestigated in this study, high Fe3* contents are by halogen contents. 978 DYAR ET AL.: Fe3* AND H+ IN MANTLE KAERSUTITE

4. H+ contentsof the kaersutitesamples vary in a nearly metasomaticfluid? Geological Societyof America Abstracts with Pro- l: I inverse relationship with Fe3* contents.Although the grams,13,408 Best, M G. (1970) range of Fe3+/H+ in the less oxidized kaersutite mega- Kaersutite-peridotite inclusions and kindred megacrysts in basanitic lavas, Grand Canyon, Arizona. Contributions to crysts may be explained by partial H loss during entrain- Mineralogy and Petrology, 27, 25-44. ment and ascent, the near total dehydrogenation of the - (1974) Mantle-derived amphibole within inclusions in alkalic-ba- Fe3*-rich megacrystswould require time scales signifi- saltic lavas. Journal of GeophysicalResearch, 79 (14), 2107-2113. cantly longer than is expectedfor transport. Thus, it seems Bigeleisen,J., Perlman,M.L., and Prosser,H.C. (1952)Conversion of hydrogenic materials to hydrogen for isotopic analysis. Analytical likely that these oxykaersutite samples grew in a more Chemistry,24, 1356-1357. oxidized metasomaticfluid, whereincorporation of H was Boettcher, A.L., and O'Neil, J.R. (1980) Stable isotope, chemical, and not required for charge compensation. As megacrystsfrom petrographicstudies of high pressureamphiboles and micas: Evidence the same location show wide variation in Fe3+and H+, for metasomatism in the mantle source regions of alkali basalts and kimberlites. American it appearslikely that significant variations in the oxida- Journal of Science.280-A. 594-62 -. Bryndzia, L.T., Davis, A.M., and Wood, B.J. (1990) The isotopic com- tion state of the mantle metasomatic fluid occurred over position (6D) ofwater in amphibolesand the redox stateofamphibole- relatively small temporal or spatial scales. bearingupper mantle spinel peridotites GeologicalSociety of America Abstracts with Progams, 22 (7), A254. Burns, R G., and Greaves,C.J. (1971) Conelations of infrared and M6ss- AcxNowr,nocMENTS bauer site populationsofactinolites. American Mineralogist, 56, 2010- 2033. Supported by National Science Foundation grants EAR-90-17167 Bums, R.G., Greaves, C.J , Law, A.D., Tew, M J., and Prentice, F.J. (A.v.M.), EAR-93-05187 (S.J.M ), and EAR-93-04304(M D.D.). We thank (1970) Assessmentof the reliability of the Mdssbauer and infrared Howard Wilshire for the loan of samples from his personal collection, methods in site-population studies of amphiboles. Geological Society Michael Coscafor use of samplesfrom Coscaet al. ( I 99 l), and the staff of America Abstractswith Programs,2,509-511. at the National Museum of Natural History for assistancein obtaining Burns,R.G., Bums,Y.M Dyar, M.D Ryan,V.L., and Solberg,T. (1985) samplesfrom the Coleman collection We also acknowledgeKipp Bajaj, , , Iron coordination symmetries in silicates: Correlations from Mdss- Tim Day, Mike Harrell, Ma{orie Taylor, and Heather Wilcox for assis- bauer parametersofFer* and Fer+. GeologicalSociety ofAmerica Ab- tance with sample preparation, and M.T. Colucci, R. Gregory, and C.B. stractswith Programs, 17 (7), 535. Douthitt for logistical support of the isotope data. We are grateful for Canil, D., Virgo, D., and Scarfe,C.M. (1990) Oxidation state of mantle thoughtful reviewsby Roger Burns, Michael Cosca,Etienne Deloule, Col- xenoliths from British Columbia, Canada.Contributions to Mineralogy in Graham, and ClaudeHerzberg. Mdssbauer and H* analyseswere fund- and Petrology, 104, 453-462. ed by the Mineral SpectroscopyLaboratory at the University ofOregon. Clowe,C.A, Popp,R.K., and Fritz, S.J.(1988) Experimental investiga- tion of the effect of oxygen fugacity on ferric-ferrous ratios and unit RprsnrNcns crrED cell parametersof four natural clinoamphiboles.American Mineralo- gist,73, 487-499. Aldridge, L.P., Tse, J.S.,and Bancroft, G M. (1982) The identification of Cosca,M.A., Essene,E.J., and Bowman,J.R. (1991)Complete chemical Fe'?+in the M4 site of calcic amphiboles: Discussion.American Min- analyses of metamorphic homblendes: Implications for normaliza- eralogist,67 , 335-339. tions, calculatedHrO activities, and thermobarometry. Contributions Amundsen, H.E.F., and Neumann, E.-R. (1992) Redox control during to Mineralogy and Petrology, 108,472-484. manfle/melt interaction.Geochimica et CosmochimicaActa. 56.2405- Deloule, E., Albardde, F., and Sheppard,S.M F (1990) D/11 anatysisof 24r6 amphibolesfrom mantle lherzolite xenoliths by ion probe. Internation- Andersen,T., O'Reilly, S.Y.,and Griffin, W.L. (1984)The trappedfluid al Conferenceon Geochronology,Cosmochronology, and Isotope Ge- phasein upper mantle xenoliths from Victoria, Australia: Implications ology,7, 26. from mantle metasomatism.Contributions to Mineralogy and Petrol- -(1991) Hydrogen isotope heterogeneitiesin the mantle from ion ogy,88, 72-85. probe analysisofamphiboles from ultramafic rocks. Earth and Plane- Aoki, K.-L (1963) The kaersutitesand oxykaersutitesfrom alkalic rocks tary Sciencelrtters, 105,543-553 ofJapan and surrounding areas.Journal ofPetrology, 4,198-210. Dick, H.J.B., Fisher,R.L., and Bryan, W.B. (1984)Mineralogic variability - (l 970) Petrologyof kaersutite-bearingultramafic and mafic inclu- ofthe uppermost mantle along mid-ocean ridges.Earth and Planetary sionsin Iki Island, Japan.Contributions to Mineralogy and Petrology, Sciencektters. 69. 88-106. 25,270-283. Dodge,F.C.W., Papike,J.J., and Mays, R.E. (1968)Hornblendes from -(1971) Petrology of mafic inclusions from Itinome-gata, Japan. granite rocks ofthe Central Sierra Nevada Batholith, Califomia. Jour- Contributions to Mineralogy and Petrology, 30, 3 l4-33 l. nal of Petrology,9, 378-410 Bahgat,A.A , and Fayek, M.K. (1982) Absolure evaluation offerrous and Dyar, M.D. (1984) Precision and interlaboratory reproducibility of mea- ferric concentrationin Ca amphibole. PhysicaStatus Solidi A, 7 l, 575- surementsof the Mdssbauereffect in minerals. American Mineralogist, 581. 69, tt27-tt44. Bai, Q, and Kohlstedt, D.L. (1992) Substantial hydrogen solubility in Dyar, M.D., McGuire, A.V., and Ziegler,R.D. (1989)Redox equilibria olivine and implications from water storagein the mantle. Nature, 357, and crystal chemistry of coexistingminerals from spinel lherzolite man- 672-674. tle xenoliths. American Mineralogist, 74, 969-980. Bailey,D.K. (1970)Volatile flux, heat focusing,and generarionof magma. Dyar, M.D., McGuire, A.V., and Harrell, M.D. (1992a) Redox behavior GeologicalJoumal SpecialIssue, 2, 177-186. in contrastingstyles of metasomatismin the upper mantle. Geochimica -(1972) Uplift rifting and magmatism in continental plates. Leeds et Cosmochimica Acta. 56. 2579-2586. Journal of Earth Sciences,8. 225-239. Dyar, M.D., McGuire, A.V., and Mackwell, S.J.(1992b) Fe.* /H+ and D/ Bancroft, G.M. (1973) Mtissbauer spectroscopy:An introduction for in- 11in kaersutites:Misleading indicators of mantle sourcefugacities. Ge- organic chemistsand geochemists,p 34 McGraw-Hill, New york. ology,20, 565-568 Bancroft, G.M., and Brown, J.R. (1975) A M6ssbauerstudy of coexisting Ernst, W.G., and Wai, C.M. (1970) Mdssbauer,infrared, X-ray, and op- hornblendes and biotites: Quantitative Fe3+/Fe,+ ratios. American tical study ofcation ordering and dehydrogenationin natural and heat- Mineralogist,60, 265-272. treated sodic amphiboles.American Mineralogist, 55, 1226-1 258 Barnes, V.E. (1930) Changesin hornblende at about 800qC.American Goldman, D.S., and Rossman,G.R. (1977) The identification of Fer* in Mineralogist, 15, 393-417. the M(4) site of calcic amphiboles. American Mineralogist, 62, 205- Bergman,S.C. (1981) Fluid inclusionsin xenoliths: SamDlesof the mantle 2t6. DYAR ET AL.: Fe3+AND H+ IN MANTLE KAERSUTITE 979

Goodman, B.A., and Wilson, M.J. (1976) A Mtissbauer study of the Pechar, F., Fuess,H., and Joswig, W. (1989) Refinement of the crystal weatheringof hornblende.Clay Minerals, 2, 153-163. structure of kaersutite (Vlditbra, Bohemia) from neutron diffraction. Graham,C.M., Harmon, R.M., and Sheppard,S M.F. (1984)Experimen- NeuesJahrbuch {iir Mineralogie Monatshefte, 89 (3), 137-143. tal hydrogen isotope studies:Hydrogen isotope exchangebetween am- Phillips, M.W., Popp, R.K., and Clowe, C.A. (1988) Structural adjust- phibole and water American Mineralogist, 69, 128-138. ments accompanyingoxidation-dehydrogenations in amphiboles. Harte, B., Cox, K.G., And Gurney, J.J (1975)Petrography and geological American Mineralogist, 73, 500-506. history of upper mantle xenoliths ftom Matsoku Kimberlite pipe. Phys- Phillips, M.W., Draheim, J.A., Popp, R.K., Clowe, C.A., and Pinkerton, ics and Chemistryofthe Earth,9,477-506 A.A. (1989) Effects of oxidation-dehydrogenation in tschermakitic Hawthorne, F.C. (1978) The crystal chemistry of the amphiboles VI The hornblende.American Mineralogist, 74, 764-7 73. stereochemistryofthe octahedralstrip. CanadianMineralogist, 16, 37- Popp, R.K., Phillips, M.W., and Harrell, J.A. (1990) Accommodation of Fe3* in natural Fer*-rich calcic and subcalcic amphiboles: Evidence --11f eSf 1 Crystal chemistry of the amphiboles.In Mineralogical So- from published chemical analyses.American Mineralogist, 75, 163- ciety of America Reviewsin Mineralogy,9A, l-102. I 69. -(1983) The crystalchemistry of the amphiboles.Canadian Min- Roden, M.F., Frey, F.A., and Francis, D.M. (1984) An example of con- eralogist,21, 173-480. sequent mantle metasomatism in peridotite inclusions from Nunivak Hawthorne, F.C., and Grundy, H D. (1973) The crystal chemistry of the Island, Alaska. Journal of Petrology, 25, 546-577. amphiboles. II. Refinement of the crystal structure of oxy-kaersutite. Roedder, E. (1965) Liquid CO, inclusionsin olivine-bearingnodules and Mineralogical Magazine, 39, 390-400 phenocrystsfrom basalts.American Mineralogist, 50, 1746-1782. Holdaway,M.J, Dutrow, B.L., Borthwick,J., Shore,P., Harmon, R.S., Ruby, S.L. (1973) Why Misfit when you already have 1'??In I.J. Gruver- and Hinton, R.W. (1986) H content of staurolite as determined by H man and C.W. Seidel, Eds., Miissbauer effect methodology, p. 263- extraction line and ion microprobe. American Mineralogist, 71, I I 35- 276 Plenum, New York. I 141 Schmalzried,H. (1981)Solid statereactions, 254 p. VerlagChemie, Wein- Ionov, D.A., and Wood, B.J. (1992) The oxidation stateofsubcontinental heim, Germany. mantle: Oxygen therrnobarometry ofmanfle xenoliths from central Asia. Schumacher,J.C. (1991) Empirical ferric iron corrections:Necessity, as- Contributions to Mineralogy and Petrology, I I I , I 79- 193 sumptions,and effectson selectedgeotherrnobarometers. Mineralogical lirirk,2., Pechar,F., and Vratislov, S. (1986) Distribution ofcations and Mag;azine,55,3-18. the proton location in kaersutite CrystallographicResearch and Tech- Schwartz,K.B., and lrving, A.J. (1978) Cation valencedeterminations in nology,21,l5l7-1519. kaersutiteand biotite megacrystsfrom alkalic basalts:Evidence for Tia* Kitamura, M., and Tokonami, M. (1971) The crystal structure of kaer- in the upper mantle. Eos, 59, 1214-1215. sutite.Scientific Repons ofthe Tohoku University,Series 3,11,125- Sheppard,S.M.F., and Epstein,S. (1970) D/ H and '80/'60 ratios of min- 141. erals of possible mantle or lower crustal origin. Earth and Planetary Kuroda, Y., Hariya, Y., Suzuoki,T., and Matsuo, S. (1978) Hydrogen Sciencel€tters. 9. 232-239. isotope study on biotite and homblende from Finnish granitic rocks. Skogby,H., and Rossman,G.R. (1989) OH in pyroxene:An experimen- GeochemicalJournal, 12, 259-263. tal study of incorporation mechanismsand stability. American Min- Kyser, T.K. (1992) Hydrogen isotope systematicsof mafic magmas and eralogist,74, 1059-1069. mantle minerals (abs.).Eos, 73, 348 Spera, F.J. (1984) Carbon dioxide in petrogenesis.III. Role ofvolatiles Mackwell, S J., and Kohlstedt, D.L. (1990) Diffusion of hydrogen in ol- in the ascent of alkaline magmas with special referenceto xenolith- ivine: Implications for water in the mantle. Journal of Geophysical bearing mafic lavas. Contributions to Mineralogy and Petrology, 88, Research.95. 5079-5088. 2t7-232. Mackwell,S.J., Dimos, D., and Kohlstedt,D.L. (1988)Transient creep Stosch, H.-G., and Seck, H.A. (1980) Geochemistry and mineralogy of of olivine: Point defect relaxation times. Philosophical Magazine, 57, two spinel peridotite suitesfrom Dreiser Weiher, West Germany. Geo- 7'79-789 chimica et Cosmochimica Acta, 44, 457-47O. Matson, D.W., Muenow, D.W., and Garcia, M.O. (1984) Volatiles in Stout, J.H. (1972) Phasepetrology and mineral chemistry ofcoexisting amphibolesfrom xenoliths, Vulcan's Throne, Grand Canyon, Arizona. amphibolesfrom Telemark, Norway. Journal ofPetrology, 13,99-145. Geochimica et CosmochimicaActa. 48. 1629-1636. Varne, R. (1970) Hornblende lherzolite and the upper mantle. Contri- McGuire,A.V , Dyar, M.D., and Ward, K.W. (1989)Neglected Fe.*/Fe,* butions to Mineralogy and Petrology,27, 45-51. ratios: A study ofFer* content of megacrystsfrom alkali basalts Ge- Whipple, E.R. (1972) Quantitative Mtissbauer specua and chemistry of ology, 17, 687-690 iron, 205 p. Ph.D. thesis,Massachusetts Institute of Technology,Cam- McGuire,A.V., Dyar, M.D., and Nielson,J E (1991)Metasomatic oxi- bridge, Massachusetts. dation of upper mantle peridotite. Contributions to Mineralogy and Wilshire, H.G. (1987) A model of mantle metasomatism.Geological So- Petrology,107, 252-264. ciety ofAmerica SpecialPaper, 215,47-60. Menzies, M.A. (1983) Mantle ultramafic xenoliths in alkaline magmas: Wilshire, H.G., And Shervais,J.W. (1975) Al-augite and Cr-diopside ul- Evidence for mantle heterogeneitymodified by magmatic activity In tramafic xenoliths in basaltic rocks from westem United States:Struc- C.J. Hawkesworth,and M.J Norry, Eds.,Continental basaltsand man- tural textural relationships.Physics and Chemistry ofthe Earth, 9, 257- tle xenoliths,p. 92-110. Shiva,Nantwich, U.K. 272. Nielson, J.E., Budahn, J.R., Unruh, D.M., and Wilshire, H.G. (1993) Wilshire, H.G., and Trask, N.J. (1971) Structural and textural relation- Actualistic models of mantle metasomatismdocumented in a compos- ships of amphibole and phlogopite in peridotite inclusions, Dish Hill, ite xenolith from Dish Hill, California. Geochimica et Cosmochimica California. American Mineralogist, 56, 240-253. Acta,57,105-121. Winchell, A.N. (1945) Variation in composition and propertiesof calcif- Oxburgh, E.R. (1964) Petrologic evidencefor the presenceof amphibole erous amphiboles.American Mineralogist, 30, 27 -5O. in the upper mantle and its petrogeneticand geophysicalimplications. Wong, A.S., Robertson,J.D., and Francis, H.A. (1992) Determination of GeologicalMagazine, l0l, 1-19 fluorine in coal and coal fly ashby proton-inducedgamrna-ray emission Papike, J.J, and Clark, J.R. (1968) The crystal structure and cation dis- analysis.Joumal ofTrace and Microprobe Techniques, 10, 169-181. tribution of glaucophane.American Mineralogist, 53, 11 56- I I 73. Wood, B.J., and Virgo, D. (1989) Upper mantle oxidation state: Ferric Papike, J.J., Ross, M., and Clarke, J.R. (1969) Crystal-chemicalcharac- iron contentsof lherzolite spinelsby "Fe Mtissbauerspectroscopy and terization of clinoamphibolesbased on five new structure refinements. resultant oxygen fugacities. Geochimica et Cosmochimica Acta, 53, MineralogicalSociety ofAmerica SpecialPaper,2,1 156-1173. 1277-1291. Papike, J J., Cameron, K.L., and Baldwin, K. (1974) Amphiboles and pyroxenes:Characterization of OTHER quadrilateral componentsand estimatesof ferric iron from microprobe data. Geological Society of Maxr;scnrpr RECETvEDOcrosen 21, 1992 AmericaAbstracts with Programs,6, 1053-1054. MlNuscrrsr AccEPTEDMeY 13, 1993