Fe^Ni^Co^O^S Phase Relations inPeridotite^Seawater Interactions

FRIEDER KLEIN AND WOLFGANG BACH*GEOSCIENCE DEPARTMENT, UNIVERSITY OF BREMEN, KLAGENFURTER STRA�E, 28359 BREMEN, GERMANY

RECEIVED SEPTEMBER 8, 2008; ACCEPTED NOVEMBER 28, 2008

Serpentinization of abyssal peridotites is known to produce extremely

reducing conditions as a result of dihydrogen (H2,aq) release upon

oxidation of ferrous iron in primary phases to ferric iron in

secondary minerals by H2O.We have compiled and evaluated ther-

modynamic data for Fe^Ni^Co^O^S phases and computed phase

relations in fO2,g^fS2,g and aH2,aq^aH2S,aq diagrams for tem-

peratures between 150 and 4008C at 50MPa.We use the relations

and compositions of Fe^Ni^Co^O^S phases to trace changes in

oxygen and sulfur fugacities during progressive serpentinization and

steatitization of peridotites from the Mid-Atlantic Ridge in the

158200N Fracture Zone area (Ocean Drilling Program Leg 209).

Petrographic observations suggest a systematic change from awar-

uite^magnetite^pentlandite and heazlewoodite^magnetite^pentlan-

dite assemblages forming in the early stages of serpentinization to

millerite^pyrite^polydymite-dominated assemblages in steatized

rocks. Awaruite is observed in all brucite-bearing partly serpenti-

nized rocks. Apparently, buffering of silica activities to low values

by the presence of brucite facilitates the formation of large amounts

of hydrogen, which leads to the formation of awaruite. Associated

with the prominent desulfurization of pentlandite, sulfide is removed

from the rock during the initial stage of serpentinization. In contrast,

steatitization indicates increased silica activities and that high-

sulfur-fugacity sulfides, such as polydymite and pyrite^vaesite solid

solution, form as the reducing capacity of the peridotite is exhausted

and H2 activities drop. Under these conditions, sulfides will not

desulfurize but precipitate and the sulfur content of the rock increases.

The co-evolution of fO2,g^fS2,g in the system follows an isopotential

of H2S,aq, indicating that H2S in vent fluids is buffered. In contrast,

H2 in vent fluids is not buffered by Fe^Ni^Co^O^S phases, which

merely monitor the evolution of H2 activities in the fluids in

the course of progressive rock alteration.The co-occurrence of pentlan-

dite^awaruite^magnetite indicates H2,aq activities in the

interacting fluids near the stability limit of water. The presence

of a hydrogen gas phase would add to the catalyzing capacity of

awaruite and would facilitate the abiotic formation of organic

compounds.

KEY WORDS: serpentinization; ODP Expedition 209; sulfide;

oxygen fugacity; sulfur fugacity; hydrothermal system; metasomatism;

Mid-Atlantic Ridge

I NTRODUCTIONMantle peridotite is commonly exposed at the seafloor ofultraslow- and slow-spreading mid-ocean ridges by detach-ment faulting that initiated close to the spreading axes (e.g.Cann et al., 1997; Tucholke et al., 1998). Retrograde meta-morphic hydration of these rocks as a result of reactionwith seawater (i.e. serpentinization) is widespread inthese settings. Serpentinization strongly influences therheology of the oceanic lithosphere (Escart|¤n et al., 1997),the geochemical budgets of the oceans (Thompson &Melson; 1970; Snow & Dick, 1995), and microbial processeswithin, at, and above the seafloor (Alt & Shanks, 1998;O’Brien et al., 1998; Kelley et al., 2001). A remarkable fea-ture of serpentinization is the strongly reducing nature ofthe interacting fluids, as indicated by the occurrence ofnative metals or alloys in serpentinites (Nickel, 1959;Chamberlain et al., 1965; Lorand, 1985; Abrajano &Pasteris, 1989). Fluids issuing from ultramafic massifs, forinstance in the Coast Range and Samail ophiolites, exhibitexceptionally high concentrations of dissolved dihydrogen(H2,aq) (Thayer, 1966; Barnes et al., 1967; Neal & Stanger,1983). Serpentinite-hosted seafloor hydrothermal systemsalso show high H2,aq concentrations of 10^15mmol/kg

*Corresponding author. Telephone: 0049-421-218-65400.Fax: 0049-421-218-65429. E-mail: wbach@uni-bremen.de

� The Author 2009. Published by Oxford University Press. Allrights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org

JOURNALOFPETROLOGY VOLUME 50 NUMBER1 PAGES 37^59 2009 doi:10.1093/petrology/egn071

(Charlou et al., 2002; Douville et al., 2002; Kelley et al.,2005; Proskurowski et al., 2006). These high H2,aq concen-trations are due to the oxidation of Feþ2 in the host rock bywater to Feþ3 in magnetite that forms along with serpen-tine and brucite during serpentinization. Seyfried et al.(2007) have presented experimental data suggesting thatincorporation of Feþ3 in serpentine may also generate con-siderable amounts of hydrogen.The combined effects of fluid^rock interaction as a func-

tion of rock composition, water to rock ratio, and tempera-ture can be examined by peridotite^seawater reaction pathmodels (e.g.Wetzel & Shock, 2000; Palandri & Reed, 2004;McCollom & Bach, 2008), but detailed knowledge of themineral^fluid equilibria that govern fluid compositions isrequired. Experimental and theoretical studies (e.g.Seyfried & Ding, 1995; Seyfried et al., 2004, 2007) indicatethat specific mineral^fluid reactions buffer H2,aq and alsoH2S,aq in hydrothermal solutions venting at the seafloor.In basalt systems, for instance, the pyrite^pyrrhotite^mag-netite (PPM) buffer commonly sets H2,aq and H2S,aq inthe interacting fluids (Seyfried & Ding, 1995). The reac-tions that control H2,aq and H2S,aq in peridotite^water systems are less well known. Seyfried et al. (2004)suggested that a bornite^chalcopyrite^magnetite buffercontrols H2,aq and H2S,aq in fluids issuing from submar-ine peridotite-hosted hydrothermal systems. In other stu-dies it has been suggested, however, that the phaserelations actually observed in altered peridotites indicatethe presence of H2,aq concentrations of the order ofhundreds of millimoles in serpentinization fluids (Sleepet al., 2004; Bach et al., 2006; Frost & Beard, 2007). Thesehigh predicted concentrations of hydrogen are corrobo-rated by similarly high H2,aq concentrations in fluidsfrom hydrothermal experiments (Janecky & Seyfried,1986; Berndt et al., 1996; Horita & Berndt, 1999;McCollom & Seewald, 2001; Allen & Seyfried, 2003;Seyfried et al., 2007).The use of mineral^fluid equilibria calculations in

examining peridotite^water interactions and associatedH2,aq and H2S,aq activities in hydrothermal solutions iscurrently limited by the lack of thermodynamic data formany of the phases that are abundant in serpentinite (e.g.pentlandite, awaruite, godlevskite, polydymite, violarite,vaesite; idealized formulae of opaque minerals in serpenti-nized peridotites are given inTable 1). Eckstrand (1975) andFrost (1985) worked out the phase relations in the Fe^Ni^O^S system and concluded that native metals and alloys[e.g. awaruite (Ni3Fe), which is commonly observed in ser-pentinites] indicate extremely low oxygen fugacities (e.g.eight orders of magnitude below PPM at 3008C) alongwith low sulfur fugacities [e.g. 10 orders of magnitudebelow PPM (Frost, 1985)]. Alt & Shanks (1998) recognizedthat awaruite is a common phase forming in the earlystages of serpentinization of abyssal peridotites. Awaruite

and many other phases, however, were not considered inprevious theoretical studies dealing explicitly with seafloorhydrothermal systems because of a scarcity of thermody-namic data. To overcome these difficulties, we added ther-modynamic data for some crucial phases of the Fe^Ni^Co^O^S system to the SUPCRT92 (Johnson et al., 1992) data-base and created a new EQ3/6 (Wolery, 1992) isobaric(50MPa) thermodynamic database for temperatures from0 to 4008C in 258C increments. Values of the dissolutionreaction constant (log K) for phases for which heat capa-city data are unavailable were estimated using the van’tHoff relation. Our approach is to document carefullyFe^Ni^Co^O^S phase relations in altered peridotites fromthe Ocean Drilling Program (ODP) Leg 209, Mid-Atlantic Ridge (MAR) 158N area, and determine phasestabilities in the aH2,aq^aH2S,aq plane. Following Frost(1985), we use these phase relations to deduce the evolutionpaths of peridotite^water interaction. Our goal is to esti-mate H2,aq and H2S,aq concentrations in fluids associatedwith the assemblages observed so as to compare the pre-dicted concentrations with those measured in fieldand experimental studies (Seyfried & Dibble,1980; Janecky&Seyfried,1986;McCollom&Seewald,2001; Charlou et al.,2002; Douville et al., 2002; Allen & Seyfried, 2003;Proskurowski et al., 2006; Seyfried et al., 2007).

GEOLOGICAL SETT INGThe area of the 158200N Fracture Zone (FZ, Fig. 1) at theslow-spreading (53 cm/year, full rate) MAR has been

Table 1: Idealized formulae of opaque minerals in serpenti-

nized peridotites

Mineral Chemical formula

awaruite Ni3Fe

tetrataenite NiFe

pentlandite (FeNi)9S8

godlevskite Ni9S8

heazlewoodite Ni3S2

millerite NiS

polydymite Ni3S4

violarite FeNi2S4

magnetite Fe3O4

pyrite FeS2

pyrrhotite FeS

linnaeite Co3S4

cattierite CoS2

jaipurite CoS

wairauite CoFe

chalcopyrite CuFeS2

JOURNAL OF PETROLOGY VOLUME 50 NUMBER 1 JANUARY 2009

38

explored in detail by numerous surveys (e.g. Rona et al.,1987; Bougault et al., 1988; Cannat et al., 1997; Casey et al.,1998; Escart|¤n & Cannat, 1999; Escartin et al., 2003;Fujiwara et al., 2003). Basaltic volcanism is sparse north ofthe 148N region and the volcanic rocks are mainlyenriched-type mid-ocean ridge basalts (E-MORB) (Dosso& Bougault,1986; Dosso et al., 1993). Two magmatic centersat 14 and 168N coincide with negative mantle Bouguergravity anomalies, whereas the region between these grav-ity bull’s-eyes shows a gravity high consistent with the pre-dominance of mantle rocks at the seafloor (e.g. Cannatet al., 1997; Escart|¤n & Cannat, 1999; Fujiwara et al., 2003).Extensive outcrops of serpentinized peridotites and gab-broic rocks on both rift flanks are indicative of a heteroge-neous lithosphere, a high ratio of tectonic to magmaticextension, crustal thinning and the formation of oceaniccore complexes along long-lived low-angle detachmentfaults (Escart|¤n & Cannat, 1999). ODP Leg 209 drilled 19holes at eight sites north and south of the 158200 FZ intovariably serpentinized peridotites intruded by gabbroicrocks (Kelemen et al., 2004a, 2007). The following briefdescription of drill holes and rock alteration is based onthe work of Kelemen et al. (2004a) and Bach et al. (2004),and focuses on those holes from which samples wereselected for this study.Hole 1268A lies on the western rift valley wall south of

the 158200 FZ and extends 147.6m into completely alteredharzburgite, dunite, strongly altered late magmatic dikesand mylonitic shear zones. The rocks were affected by per-vasive serpentinization and locally a superimposed perva-sive replacement of serpentine by talc (steatitization).

Sulfides are abundant in particular in the lower half ofthe section.Site 1270 is the southernmost of all the drill sites, located

on the eastern flank of the MAR in the vicinity of theLogatchev hydrothermal field (148450N) on the easternmedian valley wall. Peridotites from Hole 1270C are perva-sively serpentinized and steatized adjacent to gabbroicveins. Although alteration of peridotites from Holes 1270Cand 1270D mainly took place under static conditions,some peridotites are closely related to strongly deformed,schlieren-like gabbroic intrusions that have been comple-tely altered to chlorite and tremolite. In addition, Holes1270C and 1270D contain minor carbonate and oxideveins.Site 1271 is located on the inside corner high of the MAR

spreading segment south of the 158200 FZ. Drill core 1271Ais mainly composed of completely serpentinized dunite.Drill core 1271B comprises variably serpentinized duniteand harzburgite. Steatitization is minor in these rocks.Site 1274 is located 31km north of the 158200N FZ on the

western rift valley wall at 3940m water depth and �700mwest of the termination of the detachment fault. Hole1274A penetrates 156m into the basement and recovered35m of core that comprises 77% harzburgite, 20%dunite, and 3% gabbro. Peridotite from this hole repre-sents the least altered rock from Leg 209 with up to 35%of the original minerals preserved. For a comprehensivedescription of all the drill sites and a more detaileddescription of the alteration mineralogy and chemistry werefer the reader to Kelemen et al. (2004a, 2007), Bach et al.(2004, 2006) and Paulick et al. (2006).

Fig. 1. Location of the study area in the vicinity of the 158200N Fracture Zone. Investigated samples are from Sites 1268, 1270, 1271, and 1274;redrawn from Kelemen et al. (2004c).

KLEIN & BACH KLEIN AND BACH PERIDOTITE^SEAWATER INTERACTIONS

39

ANALYT ICAL METHODSMicroscopy and electron microprobeanalysisThin sections were optically investigated in transmittedand reflected light using a Leica DM RXP HC oil immer-sion microscope. Mineral compositions were analyzedwith a ‘JEOL Superprobe JXA 8900 R’ electron microp-robe at the University of Kiel (Germany), equipped withfive wavelength-dispersive spectrometers. Minerals wereanalyzed with an accelerating voltage of 20 kV for a beamcurrent of 20 nA and a fully focused 1 mm beam diameter.Both synthetic and natural mineral standards were used.The raw data were corrected using the ZAF method.Micro-scale element mapping and backscattered electronimages of serpentine meshes and nickeliferous opaquemineral assemblages were used to complement petro-graphic observations. In total, 915 single point analyses ofnickeliferous opaque minerals were obtained. Althoughmany of these analyses had low totals because of the smallgrain size of most of the Fe^Ni^Co^O^S minerals, theirevaluation helped us understand the changing phase rela-tions with progressive serpentinization.

Thermodynamic calculationsThermodynamic calculations were conducted using theSUPCRT92 (Johnson et al., 1992) computer code. Thedatabase of SUPCRT92 consists of standard-state(298�15K and 105 Pa) thermodynamic parameters, Maier^Kelley coefficients, and equation of state parameters forpure minerals, aqueous species and gases for the calcula-tion of equilibrium constants (log K values) for tempera-tures and pressures up to 10008C and 500MPa. Thedatabase used for this study combines all upgrades fromthe slop98.dat and the speq02.dat database (Wolery &Jove-Colon, 2004). Phase diagrams were constructed usingGeochemist’s Workbench� (GWB�) version 7.0.2 (Bethke,2007). A thermodynamic database for GWB� wasassembled for a pressure of 50MPa and temperatures of 0,25, 100, 200, 250, 300, 350, and 4008C. Log K values in thatdatabase were either computed by SUPCRT92 or calcu-lated using a van’t Hoff temperature extrapolation, whileignoring the effect of pressure (see below). Log K valuesfor the dissolution of minerals are given in Table 2.Activity coefficients for H2,aq were calculated followingDrummond (1981) for CO2,aq, whereas the activity coeffi-cient for H2S,aq was assumed to be unity at all tempera-tures (see Helgeson et al., 1970). Actual fugacity^concentration relations for H2S and H2 from Kishima(1989) and Kishima & Sakai (1984) suggest some deviationfrom ideal behavior. However, corrections to the logK values for equilibrium between dissolved and gaseousspecieswere not applied, because (1) fugacity^concentrationrelations are unavailable forT53008C and (2) correctionsare negligible (50�15 log units) between 300 and 4008C.

In the following subsections, we describe the thermody-namic data for the Fe^Ni^Co^O^S phases, which are notpart of the SUPCRT92 mineral set. Uncertainties in thesedata and their propagation in the calculation of phaseboundaries are hard to quantify. Standard state thermody-namic data for minerals, aqueous and gaseous species, aswell as high-temperature heat capacity data for mineralsand interaction parameter for solid solutions are oftenpoorly known. Consequently, these data must be consid-ered preliminary.

PentlanditeBerezovskii et al. (2001) conducted low-temperature heatcapacity measurements for synthetic pentlandite(Fe4�60Ni4�54S8) and reported a standard entropy (S8) of474�9 J/mol per K and H298�15�H0 of 762�80 kJ/mol.Using these data, we calculated a standard enthalpy of for-mation (iH8f) of �847�0 kJ/mol for stoichiometric pen-tlandite (Fe4�5Ni4�5S8), using standard enthalpies offormation for troilite (FeS) and millerite (NiS) fromRobie & Hemingway (1995). Cemic› & Kleppa (1987)reported a iH8f of �837�37�14�59 kJ/mol, which is inagreement with our results. An apparent Gibbs energy offormation (iG8f) of �836�3 kJ/mol was derived using thestandard molar entropies of Fe, Ni and S given by Robie &Hemingway (1995). This number is consistent withiG8f¼�835�2 kJ/mol from Craig & Naldrett (1971).Because high-temperature heat capacity data are lacking,we used the van’t Hoff extrapolation and SUPCRT92data for Fe, Ni and S to compute log K values for dissolu-tion of pentlandite. A molar volume (V8) of 153�3 cm3/molwas calculated for a natural pentlandite from cell constantsgiven by Kouvo et al. (1959).

HeazlewooditeiG8f, iH8f and S8 for heazlewoodite (Ni3S2) were takenfrom Robie & Hemingway (1995). We used high-tempera-ture heat capacity data from St�len et al. (1991) to calculateMaier^Kelley coefficients.V8 (40�655 cm3/mol) for heazle-woodite was calculated using cell constants given by Parise(1980).

AwaruiteHowald (2003) reported iG8f, iH8f and S8 for awaruite(Ni3Fe).We calculated log K values for awaruite by meansof the van’t Hoff extrapolation and SUPCRT92 data for Feand Ni, because high-temperature calorimetric data areunavailable. V8 (26�96 cm3/mol) was calculated from cellconstants given byAnthony et al. (1990).

TetrataeniteHowald (2003) also reported iG8f, iH8f and S8 for tetra-taenite (NiFe). We calculated log K values for tetrataeniteby means of the van’t Hoff extrapolation and SUPCRT92data for Fe and Ni, because high-temperature calorimetric

JOURNAL OF PETROLOGY VOLUME 50 NUMBER 1 JANUARY 2009

40

data are lacking. V8 (13�84 cm3/mol) was calculated fromcell constants given byAlbertsen et al. (1978).

GodlevskiteThermodynamic properties of godlevskite (Ni9S8) wereobtained from heat capacity measurements by St�len et al.(1994) for Ni7S6 and Ni3S2. A iH8f of ^802�92 kJ/mol wascalculated from H298�15�H0 (74920 J/mol) using the stan-dard enthalpy of formation of NiS given by Robie &Hemingway (1995). iG8f was derived from iH8f and

standard molar entropies for Ni and S given by Robie &Hemingway (1995). V8 for a natural godlevskite (148�7cm3/mol) was calculated from cell constants given byFleet (1987).

MilleriteThe standard state thermodynamic properties of millerite(NiS) were taken from Robie & Hemingway (1995). Thetransition of b-millerite at 3798C to �-millerite was

Table 2: Equilibrium constants for dissolution of opaque minerals (P¼ 50MPa)

Reaction Mineral log K

no. 08C 258C 1008C 2008C 2508C 3008C 3508C 4008C

1 awaruite 231�70 196�80 161�91 120�06 104�74 91�77 80�46 72�21

2 tetrataenite 119�43 103�67 83�38 61�90 54�05 47�41 41�63 37�38

3 pentlandite �57�71 �56�73 �56�35 �59�09 �61�67 �65�15 �70�08 �71�77

4 heazlewoodite 30�68 26�61 16�94 7�73 3�93 0�32 �3�23 �5�40

5 godlevskite �87�45 �84�42 �79�62 �78�71 �80�01 �82�44 �86�50 �87�29

6 millerite �9�13 �8�83 �8�42 �8�49 �8�73 �9�09 �9�64 �9�84

7 polydymite �121�00 �113�88 �99�44 �89�24 �86�54 �85�08 �84�96 �83�59

8 violarite �118�14 �110�42 �94�59 �83�35 �80�31 �78�57 �78�21 �76�67

9 vaesite �15�48 �14�63 �13�44 �13�45 �13�90 �14�62 �15�72 �16�65

10 Wairauite 120�48 109�25 84�27 62�91 55�11 48�63 42�79 38�57

11 cobaltpentlandite �82�22 �79�11 �73�66 �71�80 �72�65 �74�64 �78�29 �78�83

12 jaipurite �8�25 �8�00 �7�64 �7�72 �7�94 �8�29 �8�83 �9�03

13 linnaeite �112�41 �105�66 �91�84 �81�96 �79�30 �77�80 �77�56 �75�74

14 cattierite �17�94 �16�81 �14�98 �14�41 �14�64 �15�17 �16�11 �16�91

15 H2S,aq �7�28 �6�86 �6�37 �6�53 �6�82 �7�22 �7�78 �8�49

16 H2O,l �50�66 �46�30 �36�35 �27�58 �24�31 �21�53 �19�09 �16�84

Reaction no.

1 Ni3Fe þ 8Hþ þ 2 O2,aq¼ 3 Ni2þ þ Fe2þ þ 4 H2O

2 NiFe þ 4Hþ þ O2,aq¼ Fe2þ þ Ni2þ þ 2 H2O

3 Fe4�5Ni4�5S8 þ 10Hþ¼ 4�5 Ni2þ þ 4�5 Fe2þ þ 8 HS� þ H2,aq

4 Ni3S2 þ 4Hþ þ 0�5 O2,aq¼ 3 Ni2þ þ 2 HS� þ H2O

5 Ni9S8 þ 10Hþ¼ 9 Ni2þ þ 8 HS� þ H2,aq

6 NiS þ Hþ¼Ni2þ þ HS�

7 Ni3S4 þ 4Hþ¼Ni2þ þ 2 Ni3þ þ 4 HS�

8 FeNi2S4 þ 4Hþ¼ Fe2þ þ 2 Ni3þ þ 4 HS�

9 NiS2 þ H2,aq¼Ni2þ þ 2 HS�

10 CoFe þ 4Hþ þ O2,aq¼ Fe2þ þ Co2þ þ 2 H2O

11 Co9S8 þ 10Hþ¼ 9 Co2þ þ 8 HS� þ H2,aq

12 CoS þ Hþ¼Co2þ þ HS�

13 Co3S4 þ 4Hþ¼Co2þ þ 2 Co3þ þ 4 HS�

14 CoS2 þ H2,aq¼Co2þ þ 2 HS�

15 H2S,aq¼HS� þ Hþ

16 H2O,l¼H2,aqþ 0�5 O2,aq

KLEIN & BACH KLEIN AND BACH PERIDOTITE^SEAWATER INTERACTIONS

41

accounted for in the calculation of the equilibrium con-stants at 4008C.

VaesiteiH8f, S8 and heat capacity data for vaesite (NiS2) weretaken from NIST (Chase, 1998). iG8f (�124�8 kJ/mol) wasderived usingiH8f and the standard molar entropies of Niand S.V8 was taken from Smyth & McCormick (1995).

ViolariteWe used iG8f and S8 of violarite (FeNi2S4) reported byCraig (1971). iH8f was taken from Cemic› & Kleppa(1987). As no heat capacity data are available we calculatedlog Kvalues for dissolution of violarite using the van’t Hoffextrapolation and SUPCRT92 data for Fe, Ni and S. V8was taken from Smyth & McCormick (1995).

PolydymiteiH8f, S8 and heat capacity data for polydymite (Ni3S4)were taken from NIST (Chase, 1998). iG8f(�291�8 kJ/mol) was derived using standard molar entro-pies of Ni and S given by Robie & Hemingway (1995).V8 was taken from Smyth & McCormick (1995).

CobaltpentlanditeiH8f (854�79 kJ/mol) and S8 (463�17 J/mol per K) data forthe Co-endmember (Co9S8) of the pentlandite solid solu-tion series were taken from Rosenqvist (1954). iG8f(836�43 kJ/mol) was computed using iH8f and standardmolar entropies for Co and S given by Robie &Hemingway (1995). High-temperature heat capacity datawere adopted from Kelley (1949). The standard molarvolume (147�1cm3/mol) was calculated using cell para-meters from Rajamani & Prewitt (1975).

WairauiteiG8f and iH8f for wairauite (CoFe) are listed in thescientific group thermodata binary compounds database(Dinsdale, 1991). We calculated dissolution constants bymeans of the van’t Hoff extrapolation and SUPCRT92data for Co and Fe. The molar volume (14�09 cm3/mol)was calculated from cell constants given by Bayliss (1990).

Linnaeite, cattierite, and jaipuriteiH8f, S8 and high-temperature heat capacity data for allthree phases were taken from Mills (1974). The iG8f of allthree minerals was calculated using iH8f and standardmolar entropies for Co and S given by Robie &Hemingway (1995). The molar volumes of linnaeite(Co3S4) and cattierite (CoS2) were taken from Robie &Hemingway (1995), and that of jaipurite (CoS) fromNaumov et al. (1974).

RESULTSPetrographyWe distinguish two types of rock alteration: serpentiniza-tion of peridotite and steatitization of serpentinite. At Site1274, peridotites are partially to fully serpentinized,whereas at Sites 1270, 1271 and 1268, partially to fully ser-pentinized peridotites have undergone additional steatiti-zation to variable degrees (see Bach et al., 2004).Microtextures of the serpentinized peridotites range frompseudomorphic mesh and hourglass textures after olivineto transitional ribbon texture to non-pseudomorphic andinterlocking textures. Typical in partially serpentinizedrocks (65^98% altered) are mesh-textures with fresh oli-vine relicts in the mesh centers and serpentine/brucite-richmesh rims. Magnetite is concentrated away from the oli-vine kernels along former grain boundaries and in areasof complete serpentinization (e.g. Bach et al., 2006).Completely serpentinized rocks have serpentine/brucite/iowaite centers, which often appear dark in transmittedlight. Most samples are extensively veined by paragranularand transgranular serpentine veins. Paragranular veinsform an anastomosing network that is usually foliation par-allel and wraps around porphyroclasts, whereas transgra-nular veins crosscut porphyroclasts (Kelemen et al., 2004a).Late isotropic picrolite veins are also transgranular.Serrate chrysotile veins occur almost exclusively in non-pseudomorphic interlocking textures. In proximity to gab-broic intrusions steatitization is strongest and often invadesadjacent serpentinite by replacing former transgranularserpentine veins. Even in strongly steatized rocks the origi-nal serpentine micro-texture is commonly preserved, indi-cating alteration under static conditions (Bach et al., 2004).Cr-spinel is widespread but is clearly a magmatic relictthat, as mentioned by Eckstrand (1975), serves only as anucleus to secondary overgrowth of magnetite. Incipientalteration of Cr-spinel is indicated by occasional over-growths of grains by a thin (5^10 mm) ferrit-chromite rim.The following petrographic description focuses on

opaque mineral assemblages in variably serpentinized andsteatized peridotites. Because of the small grain size ofmost nickeliferous opaque minerals, their identification byreflected light microscopy was often impossible. To over-come this problem and to document better the changingFe^Ni^Co^O^S phase relations with progressive serpenti-nization we analyzed the chemical compositions of opaquephases by electron microprobe and used the compositionaldata for phase identification.

Primary sulfidesPeridotites from ODP Leg 209 are strongly melt-depletedas indicated by their modal and trace element composition(Paulick et al., 2006; Seyler et al., 2007). Intercumulus sul-fide^oxide ‘blebs’ of clear magmatic origin, as describedby Eckstrand (1975) for the Dumont serpentinite, or sulfide

JOURNAL OF PETROLOGY VOLUME 50 NUMBER 1 JANUARY 2009

42

inclusions hosted by olivine or pyroxene, as described byLorand (1989), were not recognized in samples investigatedhere.Primary sulfides are commonly completely replaced by

secondary sulfides or alloys formed during serpentiniza-tion. Seyler et al. (2007) reported local occurrences of mag-matic sulfides (polyhedral blebs of pentlandite, bornite,and chalcopyrite with concave inward grain boundaries)probably introduced during late melt impregnation of thelithospheric mantle. Chalcopyrite was found only in sam-ples from Site 1268; no Cu-sulfides were present in the lim-ited number of samples investigated from Sites 1270, 1271,and 1274. Sulfide grains residual to melting in ultramaficrocks often consist of pentlandite and minor pyrrhotite.Although pyrrhotite occurs in many serpentinized perido-tites described in the literature (e.g. Shiga, 1987; Abrajano& Pasteris, 1989; Lorand, 1989), it is absent in all the sam-ples we investigated from Leg 209. Miller (2007) reportedthe occurrence of pyrrhotite together with pentlandite insamples from Hole 1268A. This paragenesis is probably ofmagmatic origin and related to gabbroic intrusions fre-quently found in the lower part of this hole. Rare euhedralpentlandite grains with octahedral cleavage (10^30 mm indiameter), which were found in samples from Site 1271(Figs. 2a and e), may also be primary. They usually occurin porphyroclasts of former orthopyroxene (bastite), but nopentlandite inclusions were found in fresh pyroxene.

Secondary opaque phasesThe investigated serpentinites contain50�1vol.% nickeli-ferous opaque minerals. The principal opaque minerals inpartly serpentinized peridotites include, in order ofdecreasing abundance, magnetite, cobaltian pentlandite,pentlandite, and heazlewoodite. Awaruite is common butoccurs only in minor amounts. In general, nickeliferousopaque minerals appear as finely disseminated grains inthe serpentine matrix. Their grain size ranges from51 to50 mm. By far the most abundant mineral assemblages arepentlandite þ awaruite þ magnetite and pentlandite þheazlewoodite þ magnetite (Figs. 2b and f).

Mesh rims

In pseudomorphic serpentine mesh rims, disseminatedopaque phases are generally51 mm in diameter and thustheir mineralogy could not be determined by reflectedlight oil immersion microscopy or conventional quantita-tive electron microprobe analysis. Semi-quantitativemicro-scale element mapping revealed the presence ofmagnetite, pentlandite, heazlewoodite, and minor awar-uite (the presence of awaruite could only be deduced bystrong enrichments in Ni in places with low sulfur). Incompletely serpentinized areas magnetite forms threadsalong former olivine grain boundaries or pre-serpentiniza-tion intra-grain cracks.

Veins

In paragranular serpentine veins, abundant anhedral toweakly subhedral magnetite is aligned along tracks of afew tens of micrometers thickness, which are up to severalhundred micrometers long. Pentlandite, heazlewoodite,awaruite, and godlevskite are preferentially located in thecentral part of the veins together with magnetite. In placespentlandite þ awaruite þ magnetite and cobaltian pent-landite þ awaruite þ magnetite occur in the same vein.In larger transgranular (isotropic picrolite) veins, typi-cally 0�5^1mm thick, magnetite occurs as either patchysingle tracks or as double tracks discontinuously alongboth sides of the vein. In transgranular veins, nickeliferousopaques and their different replacement products areeasily identified, because their grain size is typicallylarger (up to 50 mm in diameter) than in meshes or para-granular veins.In transgranular veins of partly serpentinized perido-

tites pentlandite is often intergrown with awaruite andmantled by magnetite, suggesting that pentlandite andawaruite perhaps grew in equilibrium and were subse-quently mantled by magnetite (Fig. 2d). In other cases,pentlandite is clearly replaced by awaruite and magnetite(Fig. 2b). This is indicated by the elevated Co and Ni con-tents of the magnetite (see below), mantling pentlandite.Remarkably, awaruite and not wairauite replaces cobal-tian pentlandite. Pentlandite is frequently associated withheazlewoodite and/or magnetite (Figs. 2f and g). In someapparently pure heazlewoodite^magnetite assemblages intransgranular veins of almost fully serpentinized perido-tites micro-scale element mapping revealed the presenceof relic cobaltian pentlandite. It occurs as inclusions typi-cally smaller than 1 mm within relatively coarse-grainedheazlewoodite or between heazlewoodite and magnetite,suggesting that heazlewoodite and magnetite grew at theexpense of cobaltian pentlandite. Heazlewoodite co-occur-ring with godlevskite and magnetite may also containsmall cobaltian pentlandite inclusions, which are lackingin godlevskite. This indicates that godlevskite did notdirectly grow at the expense of cobaltian pentlandite.Because godlevskite is exclusively found in fully serpenti-nized rocks, it is more likely that godlevskite replaces hea-zlewoodite (Fig. 2h) in the final stage of serpentinization.In some completely serpentinized peridotites from Hole1268A magnetite in veins is partially replaced by pyrite.

Bastite

Serpentine veins crosscutting bastite (serpentine pseudo-morphic after pyroxene) are devoid of magnetite. In con-trast to pentlandite co-occurring with awaruite and/ormagnetite, the pentlandite occurring as a solitary phasein bastite exhibits a distinct octahedral cleavage (Fig. 2a).Where serpentinization is advanced, pentlandite in veinscrosscutting bastite is rimmed by and/or intergrown with

KLEIN & BACH KLEIN AND BACH PERIDOTITE^SEAWATER INTERACTIONS

43

Fig. 2. Back-scattered electron images of representative sulfide, oxide, and native metal assemblages in variably serpentinized and steatizedperidotite samples from ODP Leg 209. (a) Pentlandite in bastite serpentine with an octahedral cleavage (sample 1271B-7R1-15^22; scale bar5 mm). (b) Grain of cobaltian pentlandite (medium grey) located in a paragranular vein, partly altered to awaruite (light grey) and magnetite(dark grey) (sample 1274A-15R1-106^114; scale bar 10 m m). (c) Euhedral grain of cobaltian pentlandite (medium grey) intergrown with andrimmed by awaruite (light grey); located in a transgranular vein crosscutting bastitic serpentine (sample 1274A-17R1-121^129; scale bar 3 mm).(d) Pentlandite (medium grey) intergrown with awaruite (light grey) and mantled by magnetite (dark grey) located in a transgranularserpentine vein of sample 1274A-10R1-3^10 (scale bar 10 mm). (e) Euhedral grain of cobaltian pentlandite in bastitic serpentine with flame-likeappendages on each corner (sample 1274A-18R1-83^93; scale bar 3 mm). (f) Pentlandite (medium grey) and heazlewoodite (light grey) rimmedby magnetite (dark grey) in sample 1271B-17R1-98^102 (scale bar 10 mm). (g) Heazlewoodite (light grey) and magnetite (medium grey) in atransgranular vein of sample 1274A-20R1-121^126 (scale bar 10 mm). (h) Grain of heazlewoodite (light grey), which is partly replaced bygodlevskite (medium grey) and mantled by magnetite (dark grey) (sample 1271B-7R1-15^22; scale bar 10 mm). (i) Porous polydymitess that hascompletely replaced pentlandite associated with magnetite (sample 1271B-7R1-15^22; scale bar 50 mm). (j) Grain of heazlewoodite (light grey),godlevskite (light to medium grey), pentlandite (medium grey), millerite (medium dark grey) and magnetite (dark grey); heazlewoodite hasreplaced pentlandite during serpentinization. Godlevskite probably replaced heazlewoodite as serpentinization neared completion, whereas theinitiation of transformation of heazlewoodite and godlevskite to millerite is most probably related to steatitization (sample 1271B-10R1-30^35;scale bar 30 mm). (k) Opaque vein crosscutting partly talc altered bastitic serpentine along pseudomorphic cleavage plane. Magnetite(dark grey) is in sharp contact with pyrite (medium grey), which hosts millerite and polydymite-ss (sample 1268A-20R1-8^12; scale bar 100 mm).

JOURNAL OF PETROLOGY VOLUME 50 NUMBER 1 JANUARY 2009

44

awaruite (Fig. 2c), suggesting that awaruite replaced pen-tlandite. Awaruite in bastite is exclusively found in associa-tion with pentlandite.

Steatitization

Serpentine veins host magnetite that is gradually trans-formed to pyrite with increasing degree of steatitization(Fig. 2k). Completely steatized rocks contain pyrite, whichis locally replaced by hematite and/or goethite (see Altet al., 2007). Awaruite, pentlandite, heazlewoodite, andgodlevskite are relics of serpentinization and scarce inpartly steatized rocks (e.g. Fig. 2j). Where present, theyare mantled by magnetite, protecting them from reactionto millerite or other higher sulfur-fugacity phases. Relics ofthe assemblage pentlandite þ awaruite þ magnetite arefound in partly serpentinized peridotites that have under-gone steatitization, whereas relics of the assemblages pen-tlandite þ heazlewoodite þ magnetite and pentlandite þgodlevskite þ magnetite were found in fully serpentinizedperidotites that have undergone steatitization. Withincreasing degree of steatitization, sulfur-poor Ni sulfidesare progressively replaced by sulfur-rich Ni sulfides(Figs. 2i^k). Millerite is the most abundant Ni sulfide inpartly steatized samples, where it replaces heazlewooditeor godlevskite. Where steatitization is advanced, mineralsof the violarite^polydymite solid solution (polydymite-ss)grow at the expense of pentlandite or millerite and magne-tite (Figs. 2i and k). In completely steatized serpentinitesmagnetite is completely replaced by pyrite. Chalcopyriteis absent in samples from Sites 1270, 1271 and 1274. Itoccurs only within a few steatized samples from Hole1268A together with pyrite. Miller (2007) found chalcopyr-ite associated with pyrrhotite or minerals from the mono-sulfide solid solution in samples from the lower half of Hole1268A, but the occurrence of these minerals is clearlyrelated to gabbroic intrusions.

Relations with increasing degree ofserpentinization and steatitizationThe assemblage type changes with increasing extent of ser-pentinization (Table 2). Pentlandite þ awaruite þ magne-tite is exclusively found in partly serpentinized peridotite,whereas pentlandite þ heazlewoodite þ magnetite isusually found in fully serpentinized peridotites.The assem-blages pentlandite þ awaruite (in bastite), magnetite þheazlewoodite, and heazlewoodite þ godlevskite þ mag-netite (in fully serpentinized rocks) are common, but lessabundant (Fig. 2c, g and h). Locally, pentlandite occurs asa solitary phase (Fig. 2a). All examined samples lack cobal-tian minerals other than cobaltian pentlandite (and cat-tierite in pyrite; see below). The sulfur-rich Ni sulfidesdeveloped exclusively in steatized rocks are millerite,vaesite (in pyrite; see below), and minerals of the polydy-mite-ss. Pyrite replacing magnetite veins was foundin some fully serpentinized samples from Hole 1268A.

These pyrite-bearing serpentinites show the first signs ofsteatitization, preferentially of bastite. The assemblagesobserved change with increasing extent of serpentinizationfrom pentlandite þ awaruite þmagnetite to pentlandite þheazlewoodite þ magnetite to heazlewoodite þ godlevs-kite þ magnetite, and continue to change with progressivesteatitization manifested in Hole 1268A to magnetite þpyrite þ millerite (magnetite þ pyrite, if Ni is locally lack-ing) to pyrite þmillerite þ polydymite-ss to pyrite þ poly-dymite-ss (to pyrite þ vaesite as indicated by chemicalanalyses; see below).

Mineral chemistryAwaruite

In partly serpentinized peridotites from Hole 1274A the Niand Fe contents of awaruite vary between 63�0 and73�5mol % and 21�3 and 29�3mol %, respectively(Supplementary Data Table A1at http://petrology.oxford-journals.org/). Co is present only in minor amounts,usually53�5mol %. In one sample from Hole 1274A anawaruite grain has an elevated copper content of 2�8mol%; in all other awaruite grains the copper concentrationwas below the detection limit of the electron microprobe(�300 ppm). Awaruite is scarce in Hole 1268A and has Nicontents ranging from 64�0 to 71�5mol %. The Fe contentvaries between 26�0 and 27�9mol % and, although gener-ally a minor constituent, Co contents can range up to6�6mol %. Awaruite is absent in samples from Hole1271B. Compositional variability of awaruite seems to beindependent of pentlandite composition, as the Co contentis rather uniform and Ni/Fe atomic ratios of awaruite exhi-bit no systematic relation with pentlandite composition.Wairauite (CoFe) was described by Chamberlain et al.(1965) and Abrajano & Pasteris (1989) in serpentinites, butcould not be found in the samples investigated here. Thetransformation of cobaltian pentlandite to awaruite sug-gests that the stability field of awaruite is much largerthan that of wairauite.

Pentlandite

Pentlandite displays a wide compositional range that mayindicate a solid solution between Co9S8 and (FeNi)-pen-tlandite with approximately equal proportion of bothmetals (Supplementary Data Table A2; Fig. 3). The cobaltcontent of pentlandite varies from virtually Co-free(50�3mol %) to Co-rich (442mol %). The atomicmetal/sulfur ratio of most pentlandite grains is close to 9/8(i.e. 1�125). The full range, however, is between 1�06 and1�65. Rather than real variations in pentlandite composi-tion, the elevated ratios are related to finely intergrownawaruite or heazlewoodite in pentlandite. The slightlylower atomic metal/sulfur ratios of the pentlandite fallwithin the metal/sulfur range for natural pentlanditesreported by Harris & Nickel (1972) and synthetic pentlan-dites reported by Kaneda et al. (1986). In Fig. 3 Co-rich

KLEIN & BACH KLEIN AND BACH PERIDOTITE^SEAWATER INTERACTIONS

45

pentlandites trend towards the Ni-rich side. Samples fromHoles 1268A and 1271A contain only Co-rich pentlandite,whereas samples from Site 1270 contain only Co-free pen-tlandites. In Holes 1271B and 1274A Co-rich and Co-freepentlandites are found together in some thin sections oreven in the same serpentine vein.

Heazlewoodite

The atomic metal/sulfur ratio of heazlewoodite is mostlynear stoichiometric and varies between 1�40 and 1�69.Considerable amounts of Fe (55�06mol %) and smallamounts of Co (50�56mol %) were detected(Supplementary Data Table A3) in heazlewoodite fromHole 1274A. However, the presence of minute inclusions ofpentlandite in heazlewoodite, as mentioned above, couldincrease the apparent Fe and Co content, but would alsolower the atomic metal/sulfur ratio systematically.Samples from Hole 1271B, however, have abundant heazle-woodite.Variable amounts of Fe (0�5^7�6mol %) and smallamounts of Co (50�32mol %) were detected in thesegrains. High Fe contents together with low totals suggestthat some magnetite was included in the analyses. In con-trast, where analyses approach 100wt % higher metal/sulfur ratios are positively correlated with a higher Fe con-tent, possibly indicating small-scale intergrowths withawaruite.

Godlevskite

The atomic metal/sulfur ratio of godlevskite rangesbetween 1�08 and 1�25 (Supplementary Data Table A4).Most analyses are consistent with a stoichiometry of Ni9S8

proposed by Fleet (1987). Where godlevskite replacesheazlewoodite, metal/sulfur ratios are slightly elevated,whereas low metal/sulfur ratios correspond to godlevskiteassociated with millerite. Common impurities in godlevs-kite from Leg 209 are Fe and Co, ranging between 0�7and 5�8mol % and 0�0 and 0�8mol %, respectively. Inplaces elevated Fe contents correlate with low totals, sug-gesting that some magnetite was included in the analyses.

Millerite

The atomic metal/sulfur ratio of millerite (SupplementaryData Table A5) in Hole 1268A ranges between 0�97 and1�06.Where millerite replaces godlevskite the metal/sulfurratio is slightly elevated. The Fe content varies from 0�3 to5�0mol % with a bimodal distribution depending on asso-ciated minerals. Millerite with low Fe content (51�2mol %)is always associated with pyrite and polydymite-ss,whereas millerite with high Fe content (41�5mol %) co-occurs with magnetite. Usually millerite associated withpyrite or polydymite-ss contains some Co, but in mostcases this is below 0�3mol %. Remarkably, millerite asso-ciated with relic cobaltian pentlandite in samples fromHole 1271B has elevated Co contents of51�1mol %, indi-cating that millerite grew at the expense of cobaltianpentlandite.

Polydymite^violarite solid solution (polydymite-ss)

In polydymite-ss from Hole 1268A the molar metal/sulfurratio varies between 0�72 and 0�80, whereas polydymite-ssfrom Hole 1271B has higher metal/sulfur ratios as a result ofintergrowths with magnetite (Supplementary DataTable A6). Compositions are intermediate between polydy-mite and violarite, with Fe content ranging from 7�1 to12�3mol %. Polydymite-ss grains associated with milleritehave a mean Fe content of 7�7mol, whereas those thatoccur with pyrite contain on average 11�2mol % Fe. TheCo content is mostly below 0�3mol %, but one grain asso-ciated with millerite contains 3�7mol % Co. Because of theporous texture of polydymite-ss (and intergrowths withmagnetite in samples from Hole 1271B) most electronmicroprobe analyses have low totals (SupplementaryDataTable A6).

Magnetite

Magnetite is rather uniform in composition with smallamounts of NiO (see Supplementary Data Table A7;52�9mol %) and CoO (50�3mol %). Magnetite replacingcobaltian pentlandite is slightly enriched in Co and Nicompared with magnetite that is not associated with cobal-tian pentlandite. Copper is below the detection limit of theelectron microprobe (�300 ppm) in all magnetite analysesof rocks from Hole 1274A. For Hole 1268A magnetite ana-lyses reveal slightly elevated copper and zinc contents(50�05mol %).

Fig. 3. Ternary pentlandite diagram redrawn from Kaneda et al.(1986). Pentlandite forms a continuous Co9S8^Fe4�5Ni4�5S8 solid solu-tion at temperatures above 3008C. Bimodal Co distribution in pen-tlandites from Sites 1271 and 1274 indicates temperatures below2008C, whereas those from Site 1268 indicate temperatures43008C.

JOURNAL OF PETROLOGY VOLUME 50 NUMBER 1 JANUARY 2009

46

Pyrite

The molar metal/sulfur ratio of pyrite varies between 0�48and 0�54 (Supplementary Data Table A8). Nickel contentsin pyrite range from 0�0 to 7�4mol %. The Ni content ofpyrite is low if it is associated with low-Ni minerals (e.g.Co-rich pentlandite or magnetite). In contrast, the Ni con-tent of pyrite is high if it is associated with Ni sulfides suchas millerite or polydymite-ss. The majority of pyrite grainshave high proportions of a vaesite component (4^8mol %).Typical Co contents are50�3mol %, but can be as high as6�1mol % (equivalent to 18�45mol % cattierite) in a fewspots. Copper in pyrite is below the detection limit of300 ppm.

Chalcopyrite

Chalcopyrite in talc altered rocks from Hole 1268A exhi-bits a stoichiometric composition (Supplementary DataTable A9). Small amounts of Co and Ni (50�04mol %)were detected.

Phase diagramsWe constructed diagrams illustrating the Fe^Ni^O^Sphase relations in the log fO2 vs log fS2 and log aH2,aq vslog aH2S,aq plane. Phase boundaries were all calculatedfor a pressure of 50MPa, temperatures between 150 and4008C and aH2O¼1 (Figs 4^6). Minerals that areobviously lacking in natural samples (bunsenite, nativenickel, native iron, and wu« stite) were omitted from the dia-grams to accommodate for metastable equilibrium of kine-tically favored minerals.Seyfried et al. (2004) constructed an activity^activity

diagram, which depicts the phase relations in the NiO^H2S^H2O^HCl system at 4008C and 50MPa. In additionto the stability fields of the minerals considered by Seyfriedet al. (2004), we show stability fields of vaesite, polydymite,godlevskite, and the Fe-bearing phases tetrataenite, awar-uite and pentlandite.Violarite is metastable with respect tomillerite, godlevskite, polydymite and vaesite, and there-fore does not project in the phase diagrams shown.The sta-bility region of millerite as suggested by Seyfried et al.(2004) is significantly reduced in size when pentlanditeand vaesite are considered. Godlevskite is expected to bemore stable over a wide range in H2,aq and H2S,aq activ-ities and would exclude millerite as a stable phase.Millerite is common, however, and millerite and godlevs-kite even co-occur in a single opaque grain. We thereforeshow the phase boundaries of godlevskite as dotted linesto allow for examination of the perhaps metastable phaserelations. The stability field of awaruite is larger than thatof native nickel shown by Seyfried et al. (2004) and shiftsthe appearance of heazlewoodite to higher H2S,aq activ-ities. Remarkably, the stability field of tetrataenite suggeststhat awaruite would form metastably from pentlandite.Because tetrataenite is lacking in the samples we

investigated, we show the phase boundaries of tetrataeniteas grey continuous lines to account for the manifestedcoexistence of pentlandite þ awaruite þ magnetite.Log activity^activity diagrams for the Co^Fe^O^S

system in the H2,aq^H2S,aq plane are shown for tempera-tures between 150 and 4008C at 50MPa (Fig. 6). Because ofthe smaller entropy of cobaltpentlandite compared withthat of pentlandite, the size of the pentlandite stabilityfield is strongly dependent on the Co content. The higherthe Co content of pentlandite the more the field expandstowards lower H2,aq and H2S,aq activities. As awaruiteand not wairauite replaces cobaltian pentlandite, it shouldbe expected that the stability field of awaruite is largerthan that of wairauite.This observation is in apparent con-trast to the actual stable phase relations (Fig. 6). However,the greater abundance of Ni relative to Co in the systemwill promote the stability of nickeliferous phases relativeto cobaltian phases. In addition to cobaltian pentlandite,cattierite coexisting with pyrite was the only other Cophase we detected in our samples. Jaipurite is metastablerelative to cobaltian pentlandite and linnaeite, and doestherefore not project.

DISCUSS IONFe^Ni^Co^S phase relations andtemperature estimates of fluid^rockinteractionThe phase diagrams displayed in Figs. 4^6 indicate consid-erable temperature dependences of the positions of invar-iant points and univariant reaction lines in the H2,aq^H2S,aq activity plane. Hence, before the H2 and H2S con-centrations of the interacting fluids can be estimated con-straints on the prevailing temperatures of fluid^rockinteraction are required. These can be estimated usingphase relationships (Bach et al., 2004) or oxygen isotopecompositions (Alt et al., 2007). Phase relations, specificallythe replacement of olivine by serpentine, brucite and mag-netite in the presence of fresh clinopyroxene from theupper half of Hole 1274A, have been interpreted to indicatelow temperatures of serpentinization (5200^2508C; Bachet al., 2004). Using whole-rock oxygen isotope data, Altet al. (2007) estimated variable serpentinization tempera-tures of peridotites from Leg 209. Consistent with thephase relation estimate, those workers proposed ratherlow alteration temperatures (51508C) based on the high�18O (up to 8�1ø) of samples from Hole 1274A. In con-trast, higher alteration temperatures (250^3508C) are indi-cated by low �18O whole-rock values (2�6^4�4ø) at Site1268.The phase relations in the Fe^Ni^Co^O^S system can

reveal additional information about the formation tem-peratures (Craig, 1971; Kaneda et al., 1986; Alt & Shanks,2003; Kitakaze & Sugaki, 2004). We investigated thin

KLEIN & BACH KLEIN AND BACH PERIDOTITE^SEAWATER INTERACTIONS

47

–8 –7 –6 –5 –4 –3 –2 –1 0 1–8

–7

–6

–5

–4

–3

–2

–1

0

log

a H

2S,a

qlo

g a

H2S

,aq

log

a H

2S,a

q

log a H2,aq log a H2,aq

Heazlewoodite

Mill

erite

Vaesite

Pentlandite

Aw

aru

ite

Pyrite

Pyrrhotite

Hematite Magnetite

Godlevskite

150°C

–8 –7 –6 –5 –4 –3 –2 –1 0 1–8

–7

–6

–5

–4

–3

–2

–1

0

Heazlewoodite

Mill

erite

Vaesite

Pentlandite

Aw

aru

ite

Pyrite

Pyrrhotite

Hematite

Magnetite

Godlevskite

200°C

–7 –6 –5 –4 –3 –2 –1 0 1–7

–6

–5

–4

–3

–2

–1

0

Heazlewoodite

Mill

erite

Vaesite

Pentlandite

Aw

aru

ite

Pyrite

Pyrrhotite

Hematite

Magnetite

Godlevskite

250°C

–6 –5 –4 –3 –2 –1 0 1–6

–5

–4

–3

–2

–1

0

Heazlewoodite

Mill

erite

Polyd

ymite

Vaesite

Pentlandite

Aw

aru

ite

Pyrite

Pyrrhotite

Hematite

Magnetite

Godlevskite

300°C

–5 –4 –3 –2 –1 0 1–5

–4

–3

–2

–1

0

Heazlewoodite

Mill

erite

Poly

dym

ite

Vaesite

Pentlandite

Aw

aru

ite

350°C

Pyrite

Pyrrhotite

Hem

atite

Magnetite

Godlevskite

–5 –4 –3 –2 –1 0 1–5

–4

–3

–2

–1

0

Heazlewoodite

Mill

erite

Polyd

ymite

VaesitePentlandite

Aw

aru

ite

Pyrite Pyrrhotite

Hematite

Magnetite

Godlevskite

400°C

Tetrataenite

TetrataeniteTetrataenite

Tetrataenite

Tetrataenite

Tetrataenite

LHFRHF

Fig. 4. Activity^activity diagrams depicting redox phase equilibria in the Fe^Ni^O^S system from 150 to 4008C at 50MPa. Dashed lines arethe boundaries of the magnetite, hematite, pyrrhotite, and pyrite stability fields (field labels in italics); continuous lines are boundaries of awar-uite, pentlandite, heazlewoodite, millerite, polydymite, and vaesite stability fields. Because the stability fields of godlevskite and tetrataenitecover the stability fields of millerite and polydymite and awaruite (in part), respectively, we show their phase boundaries as dotted and greylines. Phase boundaries represent equal activities of the minerals in adjacent fields. In the 3508C panel, white and grey circles represent the H2

and H2S concentrations of fluids from the Logatchev (LHF) and Rainbow (RHF) hydrothermal fields (Charlou et al., 2002; Douville et al.,2002).

JOURNAL OF PETROLOGY VOLUME 50 NUMBER 1 JANUARY 2009

48

sections from the same samples, for which �18O whole-rockdata are available (Alt et al., 2007; see Table 3). The con-trasting �18O composition between rocks from Holes1268A and 1274A is also reflected in systematic differencesin Fe^Ni^Co^O^S phase relations and mineral composi-tions. Both can be used in deriving rough estimates ofalteration temperature. In particular, the compositions ofpentlandite and polydymite-ss may reveal temperatureinformation. Kaneda et al. (1986) reported that pentlanditeforms a complete solid solution between (Fe, Ni)9�xS8 andCo9�xS8 in the 300^6008C temperature range. At 2008C,there appears to be a solvus that would allow cobaltianpentlandite to coexist with an Fe^Ni endmember pentlan-dite (Fig. 3). Cobaltian and non-cobaltian endmember pen-tlandites do indeed co-occur in veins in some samples fromHole 1274A, indicating low formation temperatures of�2008C or lower (Supplementary DataTable A2). Whole-rock �18O values for these samples (4�8^7�4ø, Alt et al.,2007) corroborate these rather low alteration temperatures.A single sample from Hole 1271B (10-R1-30^35) also fea-tures both pentlandites, which would seem inconsistentwith alteration temperatures43508C deduced by Alt et al.(2007) based on �18O of rocks from Hole 1268A and simila-rities to rocks from Site 1271 in sulfide contents and �34S.

Pentlandite from sample 1274A-15-R1-106^114 and mostpentlandite from Hole 1268A fall just outside the 3008Crange in Fig. 3. Locally, alteration temperatures may haveexceeded 3008C even in Hole 1274A. Unfortunately, no�18O data exist for that sample. All pentlandite in rocksfrom Hole 1268A is cobaltian so that alteration tempera-tures apparently were 43008C, consistent with the low�18O values of those samples (Alt et al., 2007).Polydymite and violarite form a continuous solid solu-

tion at 3008C (Craig,1971). The composition of the polydy-mite-ss grains in rocks from Hole 1268A are consistent witha (Fe,Ni)3S4 phase stable at 4008C, but they are too rich inNi to have formed at 4508C. The polydymite-ss composi-tion of the Hole 1268A samples is consistent with therather high alteration temperatures of 3508C and higherdeduced from oxygen isotope data.

Redox conditions during serpentinizationand steatitizationPrevious studies revealed that sulfides, oxides and alloys inthe Fe^Ni^O^S system are indicative of the redox condi-tions during serpentinization (e.g. Eckstrand, 1975; Frost,1985; Alt & Shanks, 1998). Our petrographic investigationsreveal that serpentinization of abyssal peridotites from

–35 –30 –25 –20–25

–20

–15

–10

–5

0

log

fS

2,g

log fO2,g

Hematite

Magnetite

Pyrite

Pyrrhotite

Awaruite

Millerite

Heazlewoodite

Polydymite

Vaesite

Pentlandite

350°C50 MPa

Pn

AwMt

PnHz

Mt

log aH2S,aq = −3

, a [

mai

n

Mi

Py Pd

, a [

mai

nPy Vs

PnHz

Mt

Mi

Fig. 5. Fugacity^fugacity diagram depicting redox phase equilibria in the Fe^Ni^O^S system at 3508C and 50MPa. Dashed lines are bound-aries of magnetite, hematite, pyrrhotite, and pyrite stability fields (field labels in italics); continuous lines are boundaries of awaruite, pentlan-dite, heazlewoodite, millerite, polydymite, and vaesite stability fields. The H2S isopotential (bold dash line) is for an activity of 1mmol/kg. It iscalculated for the equilibrium S2,g þ H2O,l ¼ O2,g þ H2S,aq using SUPCRT92 and assuming unity activity of water.The chemographs depictthe changing Fe^Ni^O^S phase assemblages observed. A trend from lower left to upper right represents the evolution of oxygen and sulfurfugacity with increasing extent of peridotite^water interaction. It follows the H2S isopotential, suggesting that H2S activities in the interactingfluids may be buffered to values around 1mM.

KLEIN & BACH KLEIN AND BACH PERIDOTITE^SEAWATER INTERACTIONS

49

–8 –7 –6 –5 –4 –3 –2 –1 0 1–8

–7

–6

–5

–4

–3

–2

–1

0

log

a H

2S,a

qlo

g a

H2S

,aq

log

a H

2S,a

q

log a H2,aq log a H2,aq

Hemat

ite

Magnetite

Pyrite

Pyrrhotite

Linn

aeite

Cattierite

Wai

rau

ite

150°C

Cobaltpentlandite

–8 –7 –6 –5 –4 –3 –2 –1 0 1–8

–7

–6

–5

–4

–3

–2

–1

0

Hematite

Magnetite

Pyrite

Pyrrhotite

Linnae

ite

Cattierite

Cobalt

Wai

rau

ite

200°C

Cobaltpentlandite

–7 –6 –5 –4 –3 –2 –1 0 1–7

–6

–5

–4

–3

–2

–1

0

Hematite

Magnetite

Pyrite

Pyrrhotite

Linn

aeite

Cattierite

Cobalt

Wai

rau

ite

250°C

Cobaltpentlandite

–6 –5 –4 –3 –2 –1 0 1–6

–5

–4

–3

–2

–1

0

Hematite

Magnetite

Pyrite

Pyrrhotite

Linn

aeite

Cattierite

Cobalt

Wai

rau

ite

300°C

Cobaltpentlandite

–5 –4 –3 –2 –1 0 1–5

–4

–3

–2

–1

0

Hem

atite

Magnetite

Pyrite

Pyrrhotite

Linn

aeite

Cattierite

Cobalt

Wai

rau

ite

350°C

Cobaltpentlandite

–5 –4 –3 –2 –1 0 1–5

–4

–3

–2

–1

0

Hematite

Magnetite

PyritePyrrhotite

Linn

aeite

Cattierite

Cobalt

Wai

rau

ite

400°C

Cobaltpentlandite

Fig. 6. Activity^activity diagrams depicting redox phase equilibria in the Fe^Co^O^S system from 150 to 4008C at 50MPa. Dashed linesare boundaries of magnetite, hematite, pyrrhotite, and pyrite stability fields; continuous lines are boundaries of cobalt, wairauite,Cobaltpentlandite, linnaeite, and cattierite stability fields. Jaipurite is metastable relative to Cobaltpentlandite and linnaeite, and does thereforenot project.

JOURNAL OF PETROLOGY VOLUME 50 NUMBER 1 JANUARY 2009

50

Table 3: Opaque phase assemblages and �18O isotope data� for the studied samples

Hole: 1274A 1274A 1274A 1274A 1274A 1274A 1274A 1274A 1274A

Core: 10 15 15 16 17 18 20 22 27

Section: 1 1 2 1 1 1 1 1 2

Depth (cm): 3–10 106–114 39–46 44–52 121–129 83–93 121–126 24–32 5–11

Depth (mbsf): 49�33 75�06 75�86 84�14 89�51 94�13 104�11 122�34 147�65

Rock type: Du Hz Hz Hz HZ HZ Du Hz Hz

Lab. code: AP-88 AP-92 AP-93 AP-94 AP-95 AP-96 AP-98 AP-99 AP-103

Pentlandite þþ þ þ

Co-pentlandite þþþ þþþ þþþ þ þþ þþþ þþþ þ þ

Awaruite þþþ þþþ þþþ þ þþ þþ þ þ

Heazlewoodite þþþ þ þþþ þ

Godlevskite þþþ

Millerite

Polydymitess

Magnetite þþþ þþþ þþþ þþþ þþþ þþþ þþþ þþ þ

Pyrite

Chalcopyrite

Serpentine þþþ þþþ þþþ þþþ þþþ þþþ þþþ

Brucite þþþ þ þ þ þ þþ þþþ

Talc þ þ vein

�18O 7�4 6 4�8 5�4 5�7 5�4

Hole: 1268A 1268A 1268A 1268A 1268A 1271A 1271 B 1271 B 1271 B

Core: 2 2 4 13 20 4 7 10 17

Section: 1 2 3 1 1 1 1 1 1

Depth (cm): 10–16 108–115 26–35 46–55 8–12 105–110 15–22 30–35 98–102

Depth (mbsf): 14�10 16�48 28�04 68�74 103�65 29�55 36�35 50�8 85�49

Rock type: Hz Hz Hz Hz Hz Du Du Du Hz

Lab. code: AP-02 AP-03 AP-08 13R1 none AP-55 AP-61 AP-63 AP-67

Pentlandite þþ þ

Co-pentlandite þ þþþ þþ þþ

Awaruite þ

Heazlewoodite þþþ þþþ þþþ

Godlevskite þþ þþ

Millerite þþþ þþþ þ

Polydymite-ss þþ þþ þþ

Magnetite þ þ þ þþ þþ þþþ þþþ þþþ

Pyrite þþþ þþþ þþ þþþ þþþ þþþ

Chalcopyrite þ

Serpentine þþþ þ þþþ þþþ þþþ

Brucite

Talc þþþ þ þþþ þ vein þ þ þ

�18O 5�9 3�7 4�8 4�1

��18O isotope data are from Alt et al. (2007).þ, scarce; þþ, abundant; þþþ, very abundant; Du, dunite; Hz, harzburgite; mbsf, metres below sea floor.

KLEIN & BACH KLEIN AND BACH PERIDOTITE^SEAWATER INTERACTIONS

51

ODP Leg 209 is accompanied by changing Fe^Ni^Co^O^S phase assemblages. In partly serpentinized perido-tites, pentlandite þ awaruite þ magnetite and pentlanditeþ heazlewoodite þ magnetite are the dominant assem-blages. Fe^Ni^O^S minerals in serpentinites from ODPLeg 209 are volumetrically insignificant (50�1 vol.%) andhence incapable of buffering H2,aq. Changes in Fe^Ni^O^S assemblage mineralogy apparently monitor changes inH2,aq activity superimposed by reactions between sea-water-derived fluids and phases in the MgO^FeO^Fe2O3^SiO2^H2O system.A reaction commonly observed in thin section is the

desulfurization of pentlandite to awaruite and magnetitethat is driven by the large quantities of H2,aq releasedduring serpentinization:

Ni4�5Fe4�5S8 þ 4H2,aqþ 4H2O ¼

1�5Ni3Feþ Fe3O4 þ 8H2S,aq:ð1Þ

The reaction indicates extremely low oxygen and sulfurfugacities in the system.Pentlandite þ awaruite þ magnetite equilibria in the

absence of heazlewoodite imply hydrogen concentrationsclose to or at the solubility of dihydrogen in water, in parti-cular between 200 and 3508C (Fig. 4). Awaruite and hea-zlewoodite never co-occur in one assemblage, althoughthey may co-occur in the same thin section. The assem-blage consisting of pentlandite, heazlewoodite, and mag-netite indicates H2,aq activities just below dihydrogensaturation of the fluids. This assemblage suggests that pen-tlandite breakdown is now by the reaction

Ni4�5Fe4�5S8þ 6H2O!

1�5Ni3S2 þ 1�5Fe3O4 þH2,aqþ 5H2S,aqð2Þ

wherease awaruite breaks down by the reaction

3Ni3Feþ 6H2S,aq þ 4H2O!

Fe3O4 þ 3Ni3S2 þ 10H2,aq:ð3Þ

Both reactions indicate increasing oxygen fugacities; thelatter also suggests increasing sulfur fugacities comparedwith conditions during the breakdown of pentlandite toawaruite. Although there is ample evidence for pentlanditebreakdown to heazlewoodite and magnetite in thin section(Fig. 2f), the actual awaruite breakdown reaction is notrecorded in a specific assemblage, so it is uncertain if reac-tion (3) does indeed take place.At 4008C the awaruite stability field expands into the

pyrrhotite field. Here, the assemblage pentlandite þ awar-uite þ magnetite is not stable.

Redox conditions during steatitizationThe opaque phase assemblages we found in steatized ser-pentinites are completely different from those found in

partly to fully serpentinized peridotites. With increasingdegree of steatitization magnetite is replaced by pyriteand sulfur-poor Ni sulfides are progressively replaced bysulfur-rich Ni sulfides.These transitions indicate increasingoxygen and sulfur fugacities (see Eckstrand, 1975; Frost,1985). During steatitization of serpentinized peridotitesmillerite grows at the expense of the sulfur-poor Ni sulfidesheazlewoodite and godlevskite (Fig. 2j).The direct replace-ment of pentlandite or awaruite by millerite was notobserved, although it may take place if fO2 increasesrapidly:

Ni3S2 þH2S,aq! H2,aqþ 3NiS ð4Þ

Ni9S8 þH2S,aq! H2,aqþ 9NiS: ð5Þ

The replacement of magnetite by pyrite is represented bythe reaction

Fe3O4 þ 6H2S,aq!

2H2,aqþ 4H2Oþ 3FeS2:ð6Þ

Reactions (4)^(6) indicate decreasing H2,aq and increasingH2S,aq activities.With progressive steatitization the repla-cement of millerite by polydymite-ss (Fig. 2k), indicates afurther decrease in H2,aq and an increase in H2S,aqactivities:

3NiSþH2S,aq! H2,aqþNi3S4 ð7Þ

6NiSþ Fe3O4 þ 6H2S,aq!

2H2,aqþ 4H2Oþ 3FeNi2S4:ð8Þ

In partly steatized serpentinites magnetite þ millerite þpyrite � polydymite is the dominant assemblage (Fig. 2k).Although this is not an equilibrium assemblage per se, thosephases do represent a small range in H2,aq^H2S,aq activ-ities at 3508C (Fig. 4).The solubility of vaesite in pyrite at temperatures

54008C is below 2mol %, whereas the maximum solubi-lity of vaesite in pyrite is 7�5mol % at temperaturesaround 7008C (Clark & Kullerud, 1963). The high Ni con-centrations in pyrite from Hole 1268A (up to 7�44mol %)in combination with alteration temperatures around 3508Csuggest that the solution of Ni in pyrite is metastable. This,in turn, implies that H2,aq^H2S,aq activities were in thestability region of vaesite. The Ni-rich pyrite in Hole1268A hence is the phase representing the highest sulfurfugacitiesçor lowest aH2,aq / highest aH2S,aq conditions(Figs. 4^6). This type of mineralization appears tightlylinked to the formation of talc in veins and steatitizationof serpentinite.In addition to forming vaesite from polydymite in the

course of increasing sulfur fugacities,

Ni3S4 þ S2,g! 3NiS2 ð9Þ

JOURNAL OF PETROLOGY VOLUME 50 NUMBER 1 JANUARY 2009

52

vaesite-rich pyrite along with millerite may replace poly-dymite-ss:

FexNi3�xS4 ! FexNi1�xS2 þ 2NiS: ð10Þ

Our thermodynamic analyses indicate that this reaction isfavorable with decreasing temperatures (Fig. 4).

Implications for a potential H2S,aq bufferin serpentinite-hosted hydrothermalsystemsWe next examine if the sulfide assemblages observed inveins (i.e. unit water activities) are consistent withH2S,aq concentrations measured in high-temperaturevent fluids from ultramafic-hosted hydrothermal systems.Fluids venting at the Rainbow or Logatchev hydrothermalfields have uniform H2S concentrations of around 1mmol/kg (mM) (Charlou et al., 1998, 2002; Schmidt et al., 2007).We can relate the aqueous H2S concentrations of interact-ing fluids with estimated sulfur and oxygen fugacities fromphase relations, if we explore the equilibrium of thereaction

S2,gþ 2H2O,l ¼ 2H2S,aqþO2,g ð11Þ

(see Frost, 1985). To determine which phases could be instable coexistence with a fluid with 1mM H2S at 3508C,we plotted an H2S,aq isopotential of 1mM in Fig. 5.Remarkably, at 3508C and 50MPa the 1mM H2S,aq iso-potential follows the fS2 / fO2 evolution trend revealed bythe succession of Fe^Ni^O^S phase relations observed. Itcan thus be suggested that H2S,aq in serpentinite-hostedhydrothermal systems is buffered by equilibria betweenthe Fe^Ni^O^S phases. H2S concentrations in the alkalinelower temperature Lost City hydrothermal vent fluids areless well constrained, but exploratory data (Kelley et al.,2001; Dulov et al., 2005) are suggestive of H2S activitiesthat are are in the range of several mmol/kg. Such lowH2S activities are entirely consistent with buffering by pen-tlandite breakdown reactions at estimated reaction zonetemperatures (Allen & Seyfried, 2004) of 2008C (Fig. 4).An alternative explanation was provided by Seyfried

et al. (2004), who hypothesized that the high H2,aq andlow H2S,aq concentrations found in these systems suggestmagnetite þ bornite þ chalcocite þ fluid equilibria at4008C and 50MPa. Although this interpretation is entirelyconsistent with the thermodynamic data, we have not yetobserved the magnetite þ bornite þ chalcocite assemblagein altered peridotite. Perhaps the serpentinites and soap-stones drilled from the area around Logatchev are unlikerocks in the reaction / upflow zones underneath theLogatchev vent field. Our preferred interpretation, how-ever, is that H2S,aq is set by pentlandite-desulfurizationreactions.Metal sulfides are trace components in altered peridotite

(in particular Cu sulfides), so it appears unlikely that they

buffer a major fluid species such as dissolved H2, exceptperhaps in a mineralized upflow zone. In a subsequentpaper we will show that the levels of dissolved H2 in theLogatchev and Rainbow hydrothermal fluids are entirelyconsistent with serpentinization reactions at 4008C. Ourpreliminary conclusion is that there is no unique H2,aq^H2S,aq buffer in peridotite-hosted systems, but H2S,aqshould be set by phase equilibria to values around 1mMat temperatures around 350^4008C. However, we do notseem to have a sample of a rock that represents a reactionzone of a high-temperature vent fluid. Rocks from Hole1274A reveal low alteration temperatures, whereas rocksfrom Hole 1268A have Fe^Ni^O^S phase relations incon-sistent with the H2,aq^H2S,aq systematics of the ventfluids. The problem of whether or not there is a uniqueH2,aq^H2S,aq buffer and what that buffer might berequires further examination.

Sulfur metasomatismTotal sulfur contents of rocks from ODP Leg 209 rangefrom 0�003 to 2�1wt % (Paulick et al., 2006; Alt et al.,2007). They are variably depleted or enriched comparedwith the depleted upper mantle (�0�012wt %; Salters &Stracke, 2004). As our petrographic observations reveal,main-stage serpentinization results in desulfurization ofprimary sulfide (see Alt & Shanks, 1998). Consequently,sulfur should be lost from the rocks during serpentiniza-tion. Indeed, sulfur concentrations in many serpentinitesamples are below 0�012wt % (Fig. 7). In a plot of SiO2 vs�S data partly serpentinized peridotites with modal bru-cite (those with SiO2 540wt %) have distinctly lower�S compared with completely serpentinized and steatizedrocks. The latter can have sulfur contents of the order of2 wt %.What is the source of that sulfur and why do sul-fides become enriched in the course of steatitization?Hydrogen produced in copious amounts during serpen-

tinization will keep sulfur fugacities low and push sulfurout of primary sulfides into dissolved H2S:

S2,gþ 2H2,aq ¼ 2H2S,aq: ð12Þ

Primary sulfides in fresh peridotite will desulfurize andbecome obliterated during serpentinization:

FeSðin primary sulfideÞ þH2,aq ¼

H2S,aqþ Feðin awaruiteÞ:ð13Þ

When serpentinization nears completion the conditionsbecome less reducing and reaction (12) proceeds to the left,allowing mineralization with high sulfur fugacity assem-blages such as observed in Hole 1268A to develop. One pos-sible explanation for the sulfur enrichment in completelyserpentinized peridotites is a moving serpentinizationfront. Sulfur is leached from the peridotite during activeserpentinization, removed by the serpentinization frontand reprecipitated in rocks where serpentinization is

KLEIN & BACH KLEIN AND BACH PERIDOTITE^SEAWATER INTERACTIONS

53

complete. Perhaps the parts of the system that undergoactive serpentinization are regions where percolatingfluids pick up H2S,aq that they will subsequently dump insulfides in areas where increased sulfur and oxygen fugaci-ties prevail. The H2S,aq front, however, would have to befairly subdued, as even the low sulfur fugacity phaseswould buffer H2S,aq concentration to values of the orderof 1mM (see above). To create the sulfide accumulationsobserved in Hole 1268A, one would need to flux a substan-tial amount of serpentinization fluids through a zone thathas externally controlled high sulfur and oxygen fugacities.Such a zone could potentially be the peripheral parts ofhydrothermal upflow zones, where upwelling reducedfluids mix with entrained seawater. Not only would thephysicochemical changes in those areas of fluid mixing beconducive to sulfide precipitation, but also seawater sulfatewould constitute an external source of sulfur, which couldbe reduced thermogenically by dihydrogen dissolved in thehot upwelling fluids. Indeed, the sulfur isotope composi-tion of hydrothermal sulfide veins from Hole 1268A(�34S¼ 5^11ø; Alt et al., 2007) indicates that reduced sea-water sulfate is a significant source of sulfur besides sulfideleached from the basement. Similarly, Beard & Hopkinson(2000) found marcasite accumulations associated with afossil vent in Hole 1068 (Leg173; Iberian Margin) and con-cluded that the sulfide was precipitated from seawaterinteracting with reducing fluids venting from theserpentinite.It appears that both sulfur- and silica-metasomatism are

somehow related. However, some weakly steatized

serpentinites have exceedingly high sulfur contents (Fig. 7;see Paulick et al., 2006) indicating that the sulfur-metaso-matism preceded the silica-metasomatism. In those sam-ples, steatitization starts in serpentine veins or whereserpentine pseudomorphs pyroxene (bastite). In contrast,serpentine replacing olivine is apparently unaffected bysteatitization. Because bastite and vein serpentine areusually devoid of brucite they can be readily transformedinto talc, whereas brucite intergrown with serpentine inthe mesh-textured groundmass sucks up the silica beforeserpentine can be transformed into talc. Apparently theintroduction of silica to the system leads to increasedoxygen and sulfur fugacities that, in turn, promote sulfideprecipitation. Aqueous silica plays a role in setting oxygenfugacities, as hydrogen production tied to magnetite for-mation will be facilitated at low silica activities; forexample,

3Fe2SiO4 þ 2H2O!

2Fe3O4 þ 3SiO2,aqþH2,aqð14Þ

or

Fe3Si2O5ðOHÞ4 !

Fe3O4 þ SiO2,aqþH2OþH2,aq:ð15Þ

As long as brucite is present, it will keep silica activitieslow; that is, at the brucite^serpentine buffer, which is 3^4orders of magnitude below quartz saturation (e.g. Frost &Beard, 2007). As silica activity goes up, reaction (15) maybe reversed and Fe-rich serpentine forms. That reaction

0.0

0.5

1.0

1.5

2.0

30 35 40 45 50 55 60 65 70

SiO2 wt.%

S w

t.%

1268A1270A1270B1270C1270D1271A1271B1272A1274A

Hole

Fig. 7. Whole-rock concentrations of sulfur vs silica for samples from ODP Leg 209. Partly serpentinized peridotites (540 wt % SiO2) havesulfur concentration that are slightly enriched or markedly depleted relative to depleted mantle peridotites (0�012 wt % S). Sulfur is stronglyenriched in silica-metasomatized (i.e. brucite-free) serpentinites (440wt %) and steatites from Hole 1268A. Data plotted are from the literature(Kelemen et al., 2004b; Paulick et al., 2006; Alt et al., 2007). (See text for details.)

JOURNAL OF PETROLOGY VOLUME 50 NUMBER 1 JANUARY 2009

54

would provide a sink for H2,aq, required to pyritize mag-netite [see reaction (5)].Because talc does discriminate against Fe much more

than serpentine, the sulfide impregnation peaks duringsilica metasomatism immediately after brucite has reactedout, but before replacement of serpentine by talc is com-plete. The source of silica is most probably gabbroic intru-sions (Bach et al., 2004); such intrusions have also beenproposed to explain the sulfur and S isotopes systematics(Alt et al., 2007). Both silica- and sulfur-enrichments inrocks from Hole 1268A can, therefore, be best explainedby the involvement of gabbroic lithologies, which are fre-quently found in Hole 1268A and elsewhere in the 15820’NFracture Zone area (see Kelemen et al., 2007).

Possible existence of a free H2-rich vaporphaseAs demonstrated above, pentlandite þ awaruite þ magne-tite equilibria imply hydrogen concentrations close to or atthe solubility of dihydrogen in water, in particular between2008C and 3508C (Fig. 5). Because temperature estimatesfor alteration of rocks from ODP Leg 209 largely overlapwith this temperature range, a free H2-rich vapor phasemay exist in abyssal serpentinization systems. In continen-tal settings active serpentinization produces H2-rich gasemanations (e.g. Thayer, 1966; Coveney, 1971; Barnes et al.,1978; Yurkova et al., 1982; Coveney et al., 1987; Abrajanoet al., 1988; Sturchio et al., 1989). Although the hydrostaticpressure in the deep sea will increase the solubility ofdihydrogen, serpentinization of abyssal peridotites mayproduce a free H2-rich vapor phase. This has been pro-posed previously, based on experimental work(McCollom & Seewald, 2001, 2006) and theoretical consid-erations (Sleep et al., 2004). Our calculations provide addi-tional support for the idea that H2 concentrations close toor exceeding hydrogen solubilities may develop during ser-pentinization. Figure 8 compares the H2 concentrationscorresponding to awaruite^pentlandite^magnetite^heazle-woodite equilibrium

Ni4�5Fe4�5S8þ 9 13H2Oþ 2 1

2Ni3Fe ¼

9 13H2,aqþ 2 1

3 Fe3O4 þ 4Ni3S2ð16Þ

with that of equilibrium H2,g¼H2,aq. As indicated in thephase diagrams presented above, awaruite-bearing assem-blages represent extremely high H2,aq concentrations closeto saturation, in particular in the 200^3008C temperaturerange.The effect of pressure is also considered in the calcu-lations and illustrated in Fig. 8.Hydrogen concentrations actually measured in fluids

venting from peridotite-hosted hydrothermal systems (e.g.Charlou et al., 2002) are one to two orders of magnitudelower than the maximum values expected. However,Lost City fluid hydrogen concentrations of 15mM

(Proskurowski et al., 2006) are undersaturated at a hydro-gen partial pressure of 7�5MPa (ambient pressure at LostCity) by only a factor of five.Awaruite has been assigned a critical role in methano-

genesis and Fischer^Tropsch-type synthesis of organiccompounds in hydrothermal systems (Horita & Berndt,1999; McCollom & Seewald, 2001). We suggest that thepresence of awaruite in a rock indicates that the interactingfluids may have exsolved a H2 gas phase, which would alsomake a very efficient catalyst for organic synthesis reac-tions (McCollom & Seewald, 2006). Whether or not ahydrogen-rich gas phase forms during serpentinizationdepends on the pressure (Fig. 8). Our calculation resultssuggest that a free hydrogen gas phase could potentiallydevelop at pressures550MPa.H2- and CH4-rich fluid inclusions observed in deep-

seated gabbroic rocks from the ocean crust (e.g. Kelley,1997; Kelley & Fru« h-Green, 2001) may provide evidencefor the development of fluid exsolution. Further examina-tion of this hypothesis will rely on improved estimates ofthe pressure^temperature conditions of peridotite^fluidinteraction.

CONCLUSIONSWe propose that Fe^Ni^Co^O^S phase relations provide avery useful monitor for the evolution of temperature andthe fugacities of sulfur and oxygen during peridotite^sea-water interaction. Our results demonstrate that peridotites

−1.5

−1

−0.5

0

0.5

1

1.5

100 200 300 400 500Temperature °C

Lo

g a

H2,

aq

H2,aq solubility

Aw-Hz-Pn-Mtcontrol

100

50

25

10 MPa

Fig. 8. Comparison of the amount of H2,aq corresponding to H2O^awaruite^heazlewoodite^pentlandite^magnetite equilibria (comapreinvariant point in Fig. 4) and the amounts of H2,aq soluble at pressuresindicated by the numbers in italics. It is assumed that the partial pres-sure of hydrogen in the gas phase corresponds to the total hydrostaticpressure. Solution curves terminate just short of the two-phase bound-ary. It should be noted that a hydrogen gas phase could potentiallydevelop at pressures below 50MPa.

KLEIN & BACH KLEIN AND BACH PERIDOTITE^SEAWATER INTERACTIONS

55

from Hole 1274Awere serpentinized at uniformly reducingconditions and relatively low temperatures of 53008C.Steatitization superimposed on serpentinization resultedin increasing oxygen and sulfur fugacities.Sulfur-metasomatism affecting fully serpentinized peri-

dotites is related to steatitization.The evolution of SiO2, H2

and H2S activities is coupled. Dihydrogen drops when bru-cite is exhausted by reaction with high-aSiO2 fluids prob-ably derived from interactions with gabbroic bodies thatmake up a large fraction of the lithosphere in the 158200NFracture Zone area. As H2,aq activity drops, high-sulfur-fugacity phases such as pyrite and polydymite precipitate.The sequence of events leads to early pervasive sulfide leach-ing followed by late and localized sulfur enrichment.Associated with the prominent desulfurization of pen-

tlandite, sulfide is removed from the rock during the initialstage of serpentinization. In contrast, steatitization indi-cates increased silica activities, and high-sulfur-fugacitysulfides, such as polydymite and pyrite^vaesite solid solu-tion, form as the reducing capacity of the peridotite isexhausted and H2 activities drop. Under these conditions,sulfides will not desulfurize but precipitate, leading to anenrichment of S, as seen in rocks from Site 1268.The co-evolution of fO2,g^fS2,g in the system follows an

isopotential of H2S,aq, indicating that H2S in vent fluids isbuffered to about 1mmol/kg at 350^4008C and to micro-molal quantities at 150^2008C. These predicted concentra-tions are consistent with concentrations observed in activeperidotite-hosted hydrothermal systems.The development of pentlandite þ awaruite þ magne-

tite assemblages implies hydrogen concentrations close toor at solubility of dihydrogen in water, in particularbetween 200 and 3008C.The common occurrence of awar-uite indicates that an H2-rich vapor phase may develop inabyssal serpentinization systems, if the pressures of water^rock interaction are550MPa. The presence of such a gasphase would greatly facilitate the abiotic synthesis oforganic compounds. The phase petrological constraints onredox could be tightened if high-temperature calorimetrydata were available for awaruite, pentlandite, and violarite.

ACKNOWLEDGEMENTSWe thank Michael Hentscher for his help in setting up thethermodynamic database. Many thanks go to Niels Jo« nsand Ron Frost for stimulating discussions. We thankBarbara Mader and Peter Appel for their assistance withthe electron microprobe analyses. Carlos Garrido andHolger Paulick provided sample material and thin sec-tions. Carlos Garrido also is thanked for sharing informa-tion on Hole 1268A opaque phase assemblages. Insightfulreviews by Bernhard Evans, James Beard and DionysisFoustoukos as well as helpful comments by editor RonFrost are much appreciated. This research used samplessupplied by the Ocean Drilling Program (ODP). ODP is

sponsored by the US National Science Foundation (NSF)and participating countries under management of JointOceanographic Institutions (JOI), Inc. This work was sup-ported with funds from the Special Priority Program 1144of the German Science Foundation (BA 1605/1-1 and BA1605/1-2) and by the DFG-Research Center/ExcellenceCluster ‘The Ocean in the Earth System’.

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online.

REFERENCESAbrajano, T. A. & Pasteris, J. D. (1989). Zambales ophiolite,

Philippines, II. Sulfide petrology of the critical zone of the AcojeMassif. Contributions to Mineralogy and Petrology 103, 64^77.

Abrajano,T. A., Sturchio, N. C., Bohlke, J. K., Lyon, G. L., Poreda, R. J.& Stevens, C. M. (1988). Methane^hydrogen gas seeps, ZambalesOphiolite, Philippines: Deep or shallow origin? Chemical Geology 71,211^222.

Albertsen, J. F., Jensen, G. B. & Knudsen, J. M. (1978). Structure oftaenite in two iron meteorites. Nature 273, 453^454.

Allen, D. E. & Seyfried, W. E., Jr (2003). Compositional controls onvent fluids from ultramafic-hosted hydrothermal systems at mid-ocean ridges: An experimental study at 4008C, 500 bars.Geochimica et Cosmochimica Acta 67, 1531^1542.

Allen, D. E. & Seyfried, W. E., Jr (2004). Serpentinization and heatgeneration: constraints from Lost City and Rainbow hydrothermalsystems. Geochimica et Cosmochimica Acta 68, 1347^1354.

Alt, J. C. & Shanks,W. C. (1998). Sulfur in serpentinized oceanic peri-dotites: Serpentinization processes and microbial sulfate reduction.Journal of Geophysical Research 103, 9917^9929.

Alt, J. C. & Shanks,W. C. (2003). Serpentinization of abyssal perido-tites from the MARK area, Mid-Atlantic Ridge: Sulfur geochem-istry and reaction modeling. Geochimica et Cosmochimica Acta 67,641^653.

Alt, J. C., Shanks,W. C., III, Bach,W., Paulick, H., Garrido, C. J. &Beaudoin, G. (2007). Hydrothermal alteration and microbial sulfatereduction in peridotite and gabbro exposed by detachment faultingat the Mid-Atlantic Ridge, 15820’N (ODP Leg 209): A sulfur andoxygen isotope study. Geochemistry, Geophysics, Geosystems 8, Q08002,doi:08010.01029/02007GC001617.

Anthony, J.W., Bideaux, R. A., Bladh, K.W. & Nichols, M. C. (1990).Handbook of Mineralogy. Tucson, AZ: Mineral Data Publishing (bypermission of the Mineralogical Society of America).

Bach,W., Garrido, C. J., Harvey, J., Paulick, H. & Rosner, M., (2004).Variable seawater^peridotite interactionsçFirst insights fromODP Leg 209, MAR 158N. Geochemistry, Geophysics, Geosystems 5,Q09F26, doi: 10.1029/2004GC000744.

Bach, W., Paulick, H., Garrido, C. J., Ildefonse, B., Meurer, W. P. &Humphris, S. E. (2006). Unraveling the sequence of serpentinitza-tion reactions: petrography, mineral chemistry, and petrophyscisof serpentinites from MAR 158N (ODP Leg 209, Site 1274).Geophysical Research Letters 33, L13306^doi:13310.11029/12006GL025681.

Barnes, I., Lamarche, V. C. & Himmelberg, G. (1967). Geochemicalevidence of present-day serpentinization. Science 156, 830^832.

JOURNAL OF PETROLOGY VOLUME 50 NUMBER 1 JANUARY 2009

56

Barnes, I., O’Neil, J. R. & Trescases, J. J. (1978). Present day serpenti-nization in New Caledonia, Oman and Yugoslavia. Geochimica et

Cosmochimica Acta 42, 144^145.Bayliss, P. (1990). Revised unit-cell dimensions, space group, and che-

mical formula of some metallic minerals. Canadian Mineralogist 28,751^755.

Beard, J. S. & Hopkinson, L. (2000). A fossil, serpentinization-relatedhydrothermal vent, Ocean Drilling Program Leg 173, Site 1068(Iberia Abyssal Plain): Some aspects of mineral and fluid chemis-try. Journal of Geophysical Research 105, 16527^16539.

Berezovskii, G. A., Drebushchak, V. A. & Kravchenko, T. A. (2001).Low-temperature heat capacity of pentlandite. American

Mineralogist 86, 1312^1313.Berndt, M. E., Allen, D. E. & Seyfried, W. E. (1996). Reduction of

CO2 during serpentinization of olivine at 3008C and 500 bars.Geology 24, 351^354.

Bethke, C. M. (2007).The Geochemist’sWorkbenchVersion 7.0. Urbana, IL:University of Illinois.

Bougault, H., Dmitriev, L., Schilling, J. G., Sobolev, A., Joron, J. L. &Needham, H. D. (1988). Mantle heterogeneity from trace elements:MAR triple junction near 148N. Earth and Planetary Science Letters

88, 27^36.Cann, J. R., Blackman, D. K., Smith, D. K., McAllister, E.,

Janssen, B., Mello, S., Avgerinos, E., Pascoe, A. R. & Escart|¤n, J.(1997). Corrugated slip surfaces formed at North Atlantic 921 ridge^transform intersections. Nature 385, 329^332.

Cannat, M., Lagabrielle,Y., de Coutures, N., Bougault, H., Casey, J.,Dmitriev, L. & Fouquet, Y. (1997). Ultramafic and gabbroic expo-sures at the Mid-Atlantic Ridge: geological mapping in the 158Nregion.Tectonophysics 279, 193^213.

Casey, J. F., Braun, M. G. & Kelemen, P. K. (1998). Megamullionsalong the mid-Atlantic ridge between 148N and 168N: results ofLeg 1, JAMSTEC/WHOI MODE 98 Survey. EOS Transactions,

American Geophysical Union 79, F920.Cemic› , L. & Kleppa, O. J. (1987). High temperature calorimetry of

sulphide systems; 2, Standard enthalpies of formation of pentlanditeand violarite. Physics and Chemistry of Minerals 14, 52^57.

Chamberlain, J. A., McLeod, C. R., Traill, R. J. & Lachance, G. R.(1965). Native metals in the Muskox intrusion. Canadian Journal ofEarth Sciences 2, 188^215.

Charlou, J.-L., Fouquet,Y., Bougault, H., Donval, J. P., Etoubleau, J.,Jean-Baptiste, P., Dapoigny, A., Appriou, P. & Rona, P. A. (1998).Intense CH4 plumes generated by serpentinization ofultramafic rocks at the intersection of the 15820’N fracture zoneand the Mid-Atlantic Ridge. Geochimica et Cosmochimica Acta 62,2323^2333.

Charlou, J.-L., Donval, J.-P., Fouquet,Y., Jean-Baptiste, P. & Holm, N.(2002). Geochemistry of high H2 and CH4 vent fluids issuing fromultramafic rocks at the Rainbow hydrothermal field (36814’N,MAR). Chemical Geology 191, 345^359.

Chase, M. (1998).ThermochemicalTables: NIST-JANAF, 4th edn. Journalof Physical and Chemical Reference Data, Monograph 9.

Clark, L. A. & Kullerud, G. (1963). The sulfur-rich portion of theFe^Ni^S system. Economic Geology and the Bulletin of the Society of

Economic Geologists 58, 853^885.Coveney, R. M. (1971). Hydrogen and serpentinite; their roles in the

localization of gold ores at the Oriental Mine, Alleghany,California. Economic Geology and the Bulletin of the Society of EconomicGeologists 66, 1265^1266.

Coveney, R. M., Jr, , Goebel, E. D., Zeller, E. J., Dreschhoff, G. A. M.& Angino, E. E. (1987). Serpentinization and the origin of hydro-gen gas in Kansas. AAPGBulletin 71, 39^48.

Craig, J. R. (1971).Violarite stability relations. American Mineralogist 56,1303^1311.

Craig, J. R. & Naldrett, A. J. (1971). Phase relations and PS2^Tvaria-tions in the Fe^Ni^S system. Geological Association of Canada^

Mineralogical Association of Canada, Joint Annual Meeting, May 13^15,

1971, Abstracts of Papers.Dinsdale, A. T. (1991). SGTE data for pure elements. Calphad 15,

317^425.Dosso, L. & Bougault, H. (1986). A hot-spot at 148N on the mid-

Atlantic Ridge: isotopic (Sr, Nd) and trace element data. EOS

Transactions, American Geophysical Union 67, 410.Dosso, L., Bougault, H. & Joron, J.-L. (1993). Geochemical morphol-

ogy of the North Mid-Atlantic Ridge, 10^248N: Trace element-isotope complementarity. Earth and Planetary Science Letters 120,443^462.

Douville, E., Charlou, J. L., Oelkers, E. H., Bianvenu, P., JoveColon, C. F., Donval, J. P., Fouquet, Y., Prieur, D. & Appriou, P.(2002). The Rainbow vent fluids (36814’N, MAR): the influence ofultramafic rocks and phase separation on trace metal contents onMid-Atlantic Ridge hydrothermal fluids. Chemical Geology 184,37^48.

Drummond, S. E., Jr (1981). Boiling and Mixing of Hydrothermal Fluids:

Chemical Effects on Mineral Precipitation. University Park, PA:Pennsylvania State University.

Dulov, L. E., Lein, A. Y., Dubinina, G. A. & Pimenov, N. V. (2005).Microbial processes at the Lost City Vent Field, Mid-AtlanticRidge. Microbiology 74, 97^103.

Eckstrand, O. R. (1975). The Dumont Serpentinite: A model for con-trol of nickeliferous opaque mineral assemblages by alteration reac-tions in ultramafic rocks. Economic Geology 70, 183^201.

Escart|¤n, J. & Cannat, M. (1999). Ultramafic exposures and the grav-ity signature of the lithosphere near the Fifteen-Twenty FractureZone (Mid-Atlantic Ridge, 148^16�58N). Earth and Planetary Science

Letters 171, 411^424.Escart|¤n, J., Hirth, G. & Evans, B. (1997). Effects of serpentinization

on the lithospheric strength and the style of normal faulting atslow-spreading ridges. Earth and Planetary Science Letters 151, 181^189.

Escartin, J., Mevel, C., MacLeod, C. J. & McCraig, A. M. (2003).Constraints on deformation conditions and the origin of oceanicdetachments: The Mid-Atlantic Ridge core complex at 15845’N.Geochemistry, Geophysics, Geosystems 4, doi:10.1029/2002GC000472.

Fleet, M. E. (1987). Structure of Godlevskite, Ni9S8. Acta

Crystallographica, C 43, 2255^2257.Frost, B. R. (1985). On the stability of sulfides, oxides and native metals

in serpentinite. Journal of Petrology 26, 31^63.Frost, B. R. & Beard, J. S. (2007). On silica activity and serpentiniza-

tion. Journal of Petrology 48, 1351^1368.Fujiwara,T., Lin, J., Matsumoto,T., Kelemen, P. B.,Tucholke, B. E. &

Casey, J. F. (2003). Crustal evolution of the Mid-Atlantic Ridgenear the Fifteen-Twenty Fracture Zone in the last 5 Ma.Geochemistry, Geophysics, Geosystems 4, doi:10.1029/2002GC000364.

Harris, D. C. & Nickel, E. H. (1972). Pentlandite compositions andassociations in some mineral deposits. Canadian Mineralogist 11(4),861^878.

Helgeson, H. C., Brown, T. H., Nigrini, A. & Jones, T. A. (1970).Calculation of mass transfer in geochemical processes involvingaqueous solutions. Geochimica et Cosmochimica Acta 34, 569^592.

Horita, J. & Berndt, M. E. (1999). Abiogenic methane formation andisotopic fractionation under hydrothermal conditions. Science 285,1055^1057.

Howald, R. A. (2003). The thermodynamics of tetrataenite andawaruite: A review of the Fe^Ni phase diagram. Metallurgical and

MaterialsTransactions A 34, 1759^1769.

KLEIN & BACH KLEIN AND BACH PERIDOTITE^SEAWATER INTERACTIONS

57

Janecky, D. R. & Seyfried,W. E., Jr (1986). Hydrothermal serpentini-zation of peridotite within the oceanic crust: Experimental investi-gations of mineralogy and major element chemistry. Geochimica et

Cosmochimica Acta 50, 1357^1378.Johnson, J.W., Oelkers, E. H. & Helgeson, H. C. (1992). SUPCRT92:

A software package for calculating the standard molal thermody-namic properties of minerals, gases, aqueous species, and reactionsfrom 1^5000 bars and 0^10008C. Computers and Geosciences 18,899^947.

Kaneda, H.,Takenouchi, S. & Shoji, T. (1986). Stability of pentlanditein the Fe^Ni^Co^S system. Mineralium Deposita 21, 169^180.

Kelemen, P., Kikawa, E. & Miller, D. J. & Shipboard Scientific Party(2004a). ODP Leg 209 drills into mantle peridotite along the Mid-Atlantic Ridge from 148N to 168N. JOIDESJournal 30, 14^20.

Kelemen, P. B., Kikawa, E. & Miller, D. J., & Shipboard ScientificParty (2004b). Explanatory notes. In: Proceedings of the Ocean

Drilling Program; Initial Reports, 209. College Station, TX: OceanDrilling Program, 75 pp.

Kelemen, P. B., Kikawa, E. & Miller, D. J., & Shipboard ScientificParty (2004c). Leg 209 summary. In: Proceedings of the Ocean Drilling

Program; Initial Reports, 209. College Station, TX: Ocean DrillingProgram,139 pp.

Kelemen, P. B., Kikawa, E. & Miller, D. J. & Shipboard ScientificParty (2007). Leg 209 summary: processes in a 20-km-thick con-ductive boundary layer beneath the Mid-Atlantic Ridge, 148^168N.In: In: Kelemen, P. B., Kikawa, E. & Miller, D. J. (eds.) Proceedingsof the Ocean Drilling Program, Scientific Results, 209. College Station,TX: Ocean Drilling Program, pp. 1^33.

Kelley, D. S. (1997). Fluid evolution in slow-spreading environments.In: Karson, J. A., Cannat, M., Miller, D. J. & Elthon, D. (eds.)Proceedings of the Ocean Drilling Program, Scientific Results, 153. CollegeStation: Ocean Drilling Program, pp. 399^415.

Kelley, D. S. & Fru« h-Green, G. L. (2001). Volatile lines of descent insubmarine plutonic environments: insights from stable isotope andfluid inclusion analyses. Geochimica et Cosmochimica Acta 65,3325^3346.

Kelley, D. S., Karson, J. A., Blackman, D. K., Fru« h-Green, G. L.,Butterfield, D. A., Lilley, M. D., Olson, E. J., Schrenk, M. O.,Roe, K. K., Lebon, G. T. & Rivizzigno, P. & AT3-60 ShipboardParty (2001). An off-axis hydrothermal vent field near the Mid-Atlantic Ridge at 308N. Nature 412, 127^128.

Kelley, D. S., Karson, J. A., Fruh-Green, G. L., et al. (2005). A serpen-tinite-hosted ecosystem: The Lost City Hydrothermal Field. Science307, 1428^1434.

Kelley, K. K. (1949). Contributions to the data on theoretical metal-lurgy, X, high-temperatures heat-content, heat capacity, andentropy data for inorganic compounds. US Bureau of Mines Bulletin

476.Kishima, N. (1989). A thermodynamic study on the pyrite^pyrrhotite^

magnetite^water system at 300^5008C with relevance to the fuga-city/concentration quotient of aqueous H2S. Geochimica et

Cosmochimica Acta 53, 2143^2155.Kishima, N. & Sakai, H. (1984). Fugacity^concentration relationship

of dilute hydrogen in water at elevated temperature and pressure.Earth and Planetary Science Letters 67, 79^86.

Kitakaze, A. & Sugaki, A. (2004). The phase relations betweenFe4�5Ni4�5S8 and Co9S8 in the system Fe^Ni^Co^S at temperaturesfrom 4008 to 11008C. Canadian Mineralogist 42, 17^42.

Kouvo, O., Huhma, M. & Vuorelainen, Y. (1959). A natural cobaltanalogue of pentlandite. American Mineralogist 44, 897^900.

Lorand, J. P. (1985). The behaviour of the upper mantle sulfide com-ponent during the incipient alteration of ‘alpine’-type peridotites asillustrated by the Beni Bousera (northern Morocco) and Ronda

(southern Spain) ultramafic bodies. Tschermaks Mineralogische und

Petrographische Mitteilungen 34, 183^209.Lorand, J. P. (1989). Mineralogy and chemistry of Cu^Fe^Ni sulfides

in orogenic-type spinel peridotite bodies from Ariege (northeasternPyrenees, France). Contributions to Mineralogy and Petrology 103,335^345.

McCollom, T. M. & Bach,W. (2008). Thermodynamic constraints onhydrogen generation during serpentinization of ultramafic rocks.Geochimica et Cosmochimica Acta doi:10.1016/j.gca.2008.10.032.

McCollom, T. M. & Seewald, J. S. (2001). A reassessment of thepotential for reduction of dissolved CO2 to hydrocarbons duringserpentinization of olivine. Geochimica et Cosmochimica Acta 65,3769^3778.

McCollom, T. M. & Seewald, J. S. (2006). Carbon isotope composi-tion of organic compounds produced by abiotic synthesis underhydrothermal conditions. Earth and Planetary Science Letters 243,74^84.

Miller, D. J. (2007). Sulfide mineralization at Site 1268, Mid-AtlanticRidge, Ocean Drilling Program Leg 209. In: Kelemen, P. B.,Kikawa, E. & Miller, D. J. (eds.) Proceedings of the Ocean Drilling

Program; Scientific Results, 209. College Station, TX: Ocean DrillingProgram,18 pp.

Mills, K. C. (1974).Thermodynamic Data for Inorganic Sulphides, Selenides

andTellurides. London: Butterworth.Naumov, G. B., Ryzhenko, B. N. & Khodakovsky, I. L. (1974).

Handbook of Thermodynamic Data. (Moscow: Atomizdat, 1971.)Springfield,VA: US National Technical Information Service.

Neal, C. & Stanger, G. (1983). Hydrogen generation frommantle source rocks in Oman. Earth and Planetary Science Letters 66,315^320.

Nickel, E. H. (1959). The occurrence of native nickel^iron in the ser-pentine rock of the Eastern Townships of Quebec Province.Canadian Mineralogist 6, 307^319.

O’Brien, D., Carton, M., Eardly, D. & Patching, J. W. (1998). In situ

filtration and preliminary molecular analysis of microbial biomassfrom the Rainbow hydrothermal plume at 36815’N on the Mid-Atlantic Ridge. Earth and Planetary Science Letters 157, 223^231.

Palandri, J. L. & Reed, M. H. (2004). Geochemical models of metaso-matism in ultramafic systems: serpentinization, rodingitization,and sea floor carbonate chimney precipitation. Geochimica et

Cosmochimica Acta 68, 1115^1133.Parise, J. B. (1980). Structure of heazelwoodite (Ni3S2). Acta

Crystallographica 36B, 1179^1180.Paulick, H., Bach, W., Godard, M., de Hoog, J. C. M., Suhr, G. &

Harvey, J. (2006). Geochemistry of abyssal peridotites (Mid-Atlantic Ridge, 15820’N, ODP Leg 209); implications for fluid^rockinteraction in slow spreading environments. Chemical Geology 234,179^210.

Proskurowski, G., Lilley, M. D., Kelley, D. S. & Olson, E. J. (2006).Low temperature volatile production at the Lost City hydrother-mal field, evidence from a hydrogen stable isotope geotherm-ometer. Chemical Geology 229, 331^343.

Rajamani, V. & Prewitt, C. T. (1975). Refinement of the structure ofCo9S8. Canadian Mineralogist 13, 75^78.

Robie, R. A. & Hemingway, B. S. (1995). Thermodynamic propertiesof minerals and related substances at 298�15K and 1bar (105

Pascals) pressure and at higher temperatures. US Geological Survey

Bulletin 2131.Rona, P. A.,Widenfalk, L. & Bostrom, K. (1987). Serpentinized ultra-

mafics and hydrothermal activity at the Mid-Atlantic Ridge crestnear 158N. Journal of Geophysical Research 92, 1417^1427.

Rosenqvist, T. (1954). A thermodynamic study of iron, cobalt andnickel sulfides. Journal of the Iron and Steel Institute 176, 37^57.

JOURNAL OF PETROLOGY VOLUME 50 NUMBER 1 JANUARY 2009

58

Salters, V. J. M. & Stracke, A. (2004). Composition of the depletedmantle. Geochemistry, Geophysics, Geosystems 5, Q05004, doi:10.1029/2003GC000597.

Schmidt, K., Koschinsky, A., Garbe, S. D., de Carvalho, L. M. &Seifert, R. (2007). Geochemistry of hydrothermal fluids from theultramafic-hosted Logatchev hydrothermal field, 158N on the Mid-Atlantic Ridge; temporal and spatial investigation. Chemical Geology242, 1^21.

Seyfried,W. E. & Dibble,W. E. J. (1980). Sea water^peridotite interac-tion at 3008C and 500 bars: implications for the origin of oceanicserpentinites. Geochimica et Cosmochimica Acta 44, 309^321.

Seyfried, W. E. & Ding, K. (1995). Phase equilibria in subseafloorhydrothermal systems: a review of the role of redox, temperature,pH and dissolved Cl on the chemistry of hot spring fluids on mid-ocean ridges. In: Humphris, S. E., Zierenberg, R. A.,Mullineaux, L. S. & Thomson, R. E. (eds.) Seafloor Hydrothermal

Systems: Physical, Chemical Biological and Geological Interactions.

Geophysical Monograph, American Geophysical Union 91, 248^272.Seyfried, W. E., Jr, Foustoukos, D. I. & Allen, D. E. (2004).

Ultramafic-hosted hydrothermal systems at mid-ocean ridges: che-mical and physical controls on pH, redox and carbon reductionreactions. In: German, C. R., Lin, J. & Parson, L. M. (eds.) Mid-

Ocean Ridges: Hydrothermal Interactions Between the Lithosphere and

Oceans 148, American Geophysical Union: GeophysicalMonograph267^284.

Seyfried,W. E., Jr, Foustoukos, D. I. & Fu, Q. (2007). Redox evolutionand mass transfer during serpentinization: An experimental andtheoretical study at 2008C, 500 bar with implications for ultrama-fic-hosted hydrothermal systems at mid-ocean ridges. Geochimica etCosmochimica Acta 71, 3872^3886.

Seyler, M., Lorand, J. P., Dick, H. J. B. & Drouin, M. (2007).Pervasive melt percolation reactions in ultra-depleted refractoryharzburgites at the Mid-Atlantic Ridge 15820’N; ODP Hole 1274A.Contributions to Mineralogy and Petrology 153, 303^319.

Shiga,Y. (1987). Behavior of iron, nickel, cobalt and sulfur during ser-pentinization, with reference to the Hayachine ultramafic rocks ofthe Kamaishi mining district, northeastern Japan. Canadian

Mineralogist 25, 611^624.Sleep, N. H., Meibom, A., Fridiksson,T., Coleman, R. G. & Bird, D. K.

(2004). H2-rich fluids from serpentinization: geochemical andbiotic implications. Proceedings of the National Academy of Sciences of

the USA 101, 12818^12823.Smyth, J. R. & McCormick, T. C. (1995). Crystallographic data for

minerals. In: Ahrens, T. J. (ed.) Mineral Physics and Crystallography;

a Handbook of Physical Constants. Washington, DC: AmericanGeophysical Union, pp. 1^17.

Snow, J. E. & Dick, H. J. B. (1995). Pervasive magnesium loss bymarine weathering of peridotite. Geochimica et Cosmochimica Acta 59,4219^4235.

St�len, S., Gr�nvold, F., Westrum, J. E. F. & Kolonin, G. R. (1991).Heat capacity and thermodynamic properties of synthetic heazle-woodite, Ni3S2, and of the high-temperature phase Ni3�xS2.The

Journal of ChemicalThermodynamics 23, 77^93.St�len, S., Fjellvafi g, H., Gr�nvold, F., Seim, H. & Westrum, E. F.

(1994). Phase stability and structural properties of Ni7��S6 andNi9S8çHeat capacity and thermodynamic properties of Ni7S6 attemperatures from 5K to 970K and of Ni9S8 from 5K to 673K.Journal of ChemicalThermodynamics 26, 987^1000.

Sturchio, N. C., Abrajano, T. A., Jr., Murowchick, J. B. &Muehlenbachs, K. (1989). Serpentinization of the Acoje Massif,Zambales Ophiolite, Philippines; hydrogen and oxygen isotopegeochemistry.Tectonophysics 168, 101^107.

Thayer,T. P. (1966). Serpentinization considered as a constant-volumemetasomatic process. American Mineralogist 51, 685^710.

Thompson, G. & Melson, W. G. (1970). Boron contents ofserpentinites and metabasalts in the oceanic crust: Implicationsfor the boron cycle in the oceans. Earth and Planetary Science Letters

8, 61^65.Tucholke, B. E., Lin, J. & Kleinrock, M. C. (1998). Megamullions and

mullion structure defining oceanic metamorphic core complexes onthe Mid-Atlantic Ridge. Journal of Geophysical Research 103,9857^9866.

Wetzel, L. R. & Shock, E. L. (2000). Distinguishing ultramafic- frombasalt-hosted submarine hydrothermal systems by comparing cal-culated vent fluid compositions. Journal of Geophysical Research 105,8319^8340.

Wolery, T. J. (1992). EQ3/6, A software package for geochemical modeling of

aqueous systems: Package overview and Installation guide (version 7.0).Livermore, CA: Lawrence Livermore National Laboratory.

Wolery, T. J. & Jove-Colon, C. F. (2004). Qualification of thermody-namic data for geochemical modeling of mineral^water interac-tions in dilute systems. In: US Department of Energy (ed.) Energy.Bechtel SAIC Company, LasVegas, Nevada, USA.

Yurkova, R. M., Slonimskaya, M. V., Daynyak, B. A. & Drits, V. A.(1982). Hydrogen and methane in genetically different serpentinetypes from Sakhalin at the Koryak Mountains. Doklady AkademiiNauk SSSR 263, 420^425.

KLEIN & BACH KLEIN AND BACH PERIDOTITE^SEAWATER INTERACTIONS

59