American Mineralogist, Volume 67, pages 1144-1154, 1982

Techniques for using iron crucibles in experimental igneous petrology

Cenl R. TnoRNssn exo J. SrepHr,N HuBsNEn

U.S. Geological Survey959 National Center

Reston, Virginia 22092

Abstract

Some iron metal used to fabricate crucibles contains impurities comprising manganese,titanium and other elements. Under closed-system conditions these impurities impose an

fo, that can differ from that of a silicate charge in equilibrium with pure metallic iron.Equilibration of the impurities in the crucible with the charge can reduce the iron oxidecomponent of the charge and contaminate it with manganese and titanium. Pretreatment ofimpure iron crucibles in a CO2-CO atmosphere at 1050'C, under conditions slightly morereducing than Fe-Fer-*O, minimizes undesirable changes in the bulk composition of thecharge.

Introduction

Bowen and Schairer (1932), in their study of theiron-oxide-SiO2 system, pioneered the use of ironcrucibles to "control and define" iron oxide in itsferrous state at high temperature. However, Bowenand coworkers, and others using metallic iron tocontain iron oxide-bearing materials, found appar-ently unpredictable variations in the ferrous-ironcontent of the run products. Iron-bearing silicatecharges in iron crucibles open to an atmosphere ofpurified nitrogen are gradually oxidized, as reflect-ed by the progressive oxidation of the metallic ironcapsule and a steady increase in ferrous iron con-tent of the charge with time. This change wasascribed to the small but significant proportion ofoxygen present in the "inert" gas and was avoidedby keeping run times to "not more than fifteenminutes" (Bowen and Schairer,1932, p. 179 1935;Bowen et al., 1933), a run duration supposedlysufficient for achievement of crystal/melt equilibri-um in their system, but one which is probably notadequate for all systems. In systems open to anatmosphere of mixtures of CO2 and H2 or CO2 andCO gases, accumulation or depletion of ferrous ironin the charge is the result of either oxidation ofmetallic iron in contact with the charge or reductionof iron oxide within the charge. The extent ofchange is primarily controlled by run duration andthe oxygen fugacity (/o,) of the gas mixture (e.g.,Roeder, 1974, Fig. 1). An additional problem withthe open-system technique is volatilization and lossof alkalis from the charge, resulting in changes in

bulk composition which affect changes in the oxida-tion state of iron (Thornber et al. 1980).

Experimental research on the petrogenesis oflunar samples necessitated devising methods tocontain silicate melts at low/6, that are in equilibri-um with metallic iron (disseminated throughoutsome lunar rocks) while maintaining the bulk com-position of the charge for a time sufficient to reachequilibrium-a period of hours to days or evenweeks. Toward this end, Muan and Schairer (1970)revived a method, originally explored (and discard-ed) by Bowen and Schairer (1932), of sealingcharges in iron crucibles, which in turn are sealed inevacuated silica-glass tubes. Muan and Schairerreported significant ferrous-iron gain in charges andattributed it to a reaction between the metal cruci-ble and ferric iron in the starting mixture.

Subsequent experimental work on natural andsynthetic lunar materials, in similar closed systems,has been sporadically plagued by analogous prob-lems of iron gain or loss. The implications of suchundesired, perhaps unrecognized, bulk-composi-tion changes and explanations for their cause arediscussed by Kesson (1975) and Walker et al. (1975,1976) and are summarized by O'Hara and Hum-phries (1977). Because of the successful use ofultra-high-purity iron crucibles (Johnson-MatheyGrade 1,99.998Vo Fe)t by Kesson (1975), Walker el

rAny use of trade names in this publication is for descriptivepurposes only and does not constitute endorsement by the U.S.Geological Survey.

0003-004x/82/l I l2-l 144$02.00

THORNBER AND HUEBNER: IRON CRUCIBLES IN EXPERIMENTAL PETROLOGY l 145

al. (1977), Longhi et al. (1974), and Green et al.(1975), recent workers have generally assumed thatless pure iron crucible material, such as "ingot"grade iron (Grove et al., 1973; Walker et al., 1972)or electromagnet-grade iron (Muan and Schairer,1970; Thornber and Huebner, 1980), may impartundesirable changes in the bulk composition of thecharge.

Problems of ferrous-iron loss encountered by thepresent authors in their experimental study of thethermal history of lunar fragment-laden basaltTTll5(Thornber and Huebner, 1980) prompted a system-atic investigation of the use of iron crucibles inclosed systems. We wanted to understand the na-ture of impurities in crucible materials, their role ingoverning redox equilibria, and the effect of theimpurities upon the bulk composition of thecharges. Despite the obvious disadvantages of usingimpure iron crucibles, this material is, by virtue ofbeing less expensive, more expendable. Polishedthin sections can be cut through the crucible with-out concern for its reclamation, allowing observa-tion of an entire cross-section through the chargeand thus a more complete evaluation of its chemicaland textural variability. In particular, observationof the crucible-charge interface textures is oftencritical for the recognition of crystal nucleation oncrucible walls. For this reason, we sought methodsof using relatively impure iron crucibles by mini-mizing or eliminating undesirable effects of cruci-ble-charge interaction.

Characterization of iron metals investigated

Previous investigators considered the relevantimpurities in iron capsules to be carbon (Biggar eral., 1974; Longhi et al., 1974) and phosphorus(Walker et al.,1976,1977).We also initially consid-ered the possibility that sulfur, arsenic and hydro-gen (the last occluded during refining and manufac-turing of the iron rod stock) might be undesirableimpurities. However, typical mill analyses of im-pure iron do not report hydrogen and show carbon,arsenic, phosphorus, and sulfur at levels too low toaccount for the observed iron gain or loss.

Petrographic examination of iron from differentsources reveals characteristics that might be used toselect the most suitable iron for different experi-ments. We can distinguish different lots of impureiron on the basis of their inclusions (Figs. 1a, b, andTable 1). For example, different lots of electromag-net-grade iron contain at least I volume percent ofimpurities in the form of I to -10 micrometer

medium- and dark-grey blebs, yellow prisms andminor plum-pink and tan-colored inclusions. Quali-tative chemical analysis by energy-dispersive elec-tron microprobe and scanning-electron-microscopetechniques reveals that the inclusions contain vari-ous combinations of manganese, titanium, iron,nickel, chromium, aluminum, silicon and sulfur(Table 1). The yellow prismatic phase has the colorand morphological characteristics of titanous oxide(TiO). Materials Research Corp. uenz grade andJohnson-Mathey specpure foil do not show anyinclusions or precipitates when examined with apetrographic microscope at 800x magnification.

The bulk chemistry of some lots of iron metal isnot known. Clearly, however, at least some analy-ses supplied by manufacturers fail to report poten-tial reducing agents such as hydrogen, and showtitanium and manganese values (Table 2) that aretoo low to account for the observed inclusions(much less any additional titanium and manganesealloyed with the iron). Analyses of high-purity ironshow negligible amounts of manganese and titanium(Table 2), consistent with the absence of inclusions(Table 1) in iron.

Experimental methods

Our experiments used crucibles of iron (CoreySteel Corp. lot #92474, Armco electromagnet-gradeand Materials Research Corp. lot #2611926 VP-grade, Table 2) and a sintered glass having thecomposition of lunar basalt 77115 (Thornber andHuebner. 1980). The crucibles were 0.250 inch indiameter and 0.30 inch long. The inner bore was0.205 inch in diameter and finished with a conicalbottom, giving a maximum depth of 0.26 inch. Atypical capsule weighed -1 gm. The crucibles weremachined and stored in low sulfur oil, a procedureproven effective by Walker et al. (1977), cleaned inspectrographic grade N-Heptane or Freon, andthen stored in a vacuum desiccator.

The sintered glass was prepared by thrice fusingat -1300'C (in air) a manganese-free reagent-gradeoxide-and-carbonate mix, in a platinum cruciblepretreated with the mix under similar conditions.To assure homogeneity, the oxidized glass wascrushed and ground in reagent-grade isopropyl alco-hol before each remelting. The composition of theglass measured after the third fusion is presented inTable 3. The oxidized glass was then placed in apresaturated AgTsPd3e alloy container and reducedat 1050"C and I bar pressure in a CO2 and H2 gasmixture with an fo, of 1g-l: s bar. Particular care

I l,16 THORNBER AND HUEBNER: IRON CRUCIBLES IN EXPERIMENTAL PETROLOGY

Fig. I . Reflected-light photomicrographs (oil immersion) showing inclusions in different lots of Armco electromagnet-grade iron; (a)yellow prisms (y) and plum-colored blebs (p) in Corey Steel Corp. loI #92474, (b) medium-grey elongate blebs m-g in LipiniMuanRod. Bar scale in each photograph represents 25 pm.

was taken to preserve the composition of the re-duced and sintered glass by quenching it in liquidnitrogen, then storing it in a vacuum desiccator.The sintered product was tested with a magnet andcharacterized by X-ray diffraction and optical tech-niques, all of which showed that it contained anegligible amount of metallic iron. The sinteredstarting material was ground in alcohol, dried in airand loaded into iron crucibles. The same batch ofreduced starting material was used in all experi-ments reported in this investigation.

Hydrogen treatment of crucibles was accom-plished by heating them at -825"C in a stream ofhydrogen which flowed through a silica tube in ahorizontal furnace. Crucibles in this assembly werebrought up to temperature (- l/2 hour), heated for Ihour, and then cooled to room temperature (-1l2hour) before removal. In the presence of the hydro-gen stream, iron, manganese, and titanium are allstable as metals. Several such capsules were bakedin a vacuum before being filled with silicate charge.Our intention in including this extra step was to

determine whether or not H2 could be removedfrom a-iron (at 800'C) or "y-iron (at 1100'C).

Some crucibles were annealed at 1050"C and Ibar in a flowing gas mixture of COz and H2 or CO2and CO. The latter mixture was used when ahydrogen-free atmosphere was needed. Ideally, themixing ratios 9-21 for CO2-H2 or 7-23 for CO2-COdefine the same furnace /6, values (10-146 bar,compared with l0-ra'0 for Fe-Fe1-*O). Practically,the temperature may have been too low for theachievement of furnace-gas equilibrium (Huebner,1975); if so, the prevailing"fo, would have been morereducing. Nevertheless, at all .fo, values believedencountered, iron metal, MnO, and TiO2 are stablephases. The crucibles were quenched in a mannerthat did not cause oxidation or otherwise alter theircomposition. Drop-quenching risks contaminationwith carbon, a stable and ever-present phase in thecool, upstream portion of the flowing COz-CO gasmixture. The best procedure proved to be rapidremoval from the top (downstream) end of thefurnace and movement through a N2-rich atmo-

THORNBER AND HUEBNER: IRON CRUCIBLES IN EXPERIMENTAL PETROLOGY rt47

Table l. Description and qualitative analyses ofinclusions in iron-crucible materials

I r on Ma te r i a l Inclusion Type

Yel I ow( pr i sms )

Pl um(b l ebs )

Tan Medium-grey l {edium-grey Dark-grey(b l ebs ) ( b l ebs ) ( e l onqa te b l ebs ) { b l ebs )

Armcot rod (Corey Steellot #92474)

Armcot rod (Corey Steell o t # 9 1 9 1 3 )

Armcot rod (Corey Steellot f91 368)

Batten cruci b l es0 . 2 5 " d i a m .

Ba t t en c ruc i b l es0 . 1 0 " d i a m .

Armcot rod (L ip in/Muan )

Armcot crucib leRoede r ( 1974 )

MRC VP rod('tot #26/1926)

I'IRC Marz rod(lot #26/2226)

Johnson -Mathey(specpure fo i I )

Armcot rod 92474:C02lH2 t reatedC02lC0 t reatedHZ treated

Anncot rod (L ip in/ l {uan)H2 t reated

Elements presentl n t n c t u s l o n

AI

si

J

Ca

t l

Cr

Mn

Fe*

Ni

t race

no i nc l us l ons obse rved

no i nc l us i ons obse rved

T Xxx

no l nc l us l ons obse rved

no l nc l us i ons obse rved

X

x

x

I

x

x

X

x

)(xx

x

x

x' €2ec.ttormg net- gna.de .iltt na

Fe deteded no.q be &n to uttvuudi.ttg ilon latt.

sphere (eminating from the dewer with liquid nitro-gen) and into the liquid nitrogen to minimize contactwith air.

All capsules were filled with sintered glass suchthat the crucible-to-sample weight-ratio was 16.0t

0.5. After the capsules were loaded, their lids wereloosely fitted, and charge and crucible were heated(825'C) in a vacuum for 10 minutes to dry the chargeand remove any hydrocarbons introduced duringcharge or crucible preparation. Iron sponge, previ-

I 148

Table 2. Chemical analyses of iron crucible materials (ppm)

Armco el ectro-l Batten2 MRC vP3 MRC Marz4r c r r o n c r u c i b l e s

THORNBER AND HUEBNER: IRON CRUCIBLES IN EXPERIMENTAL PETROLOGY

1 0 0

NA

80

c

s i

P

AI

S

T i 7 0 0 N A < 1 0 . 0 1 . 4 0

Cr NA NA 30.0 1 .60

Ii'ln 800 28O NA NA

the development of the manganese-titanium richphases (Fig. 2); conversely, reduction of the cap-sule material in hydrogen gas diminishes the occur-rence of inclusions as they are reduced and alloywith the metallic iron. This result can be explainedby considering the values of/6, used in the capsulepretreatment. At relatively high fo,the manganeseand titanium, originally present in iron-rich alloy,are oxidized and form included precipitates. At theextremely low /s, values characteristic of the hy-drogen-rich atmosphere, the metallic state is stableand both manganese and titanium tend to alloy withthe iron. We have no evidence that the pretreatmentchanges the bulk composition with respect to man-ganese and titanium; rather, these elements aresimply redistributed among the phases present.

Oxidation halos formed when impure crucibleswere annealed in the CO2-H1 or COZIO gasmixtures. On oxidation, the prismatic, yellow TiOprecipitates are altered to grey manganese-titaniumoxide. The width of the halo is defined as the depthof this oxidation from the crucible surface, and isdistinctly different in capsules treated in COz-Hzand COrCO gas mixtures, but at the same nominalvalues of oxygen fugacity and temperatures. Incrucibles treated in COrHz for 143 hours, rindwidth is -200 microns, yet CO2-CO treatments for

Table 3. Composition *tt.T,:t:tit material (fused three times

Average o f l0 ana lysesox ide @

Si 02

I i02

Al 203

Fe0*

oz**Mn0

Mgo

Ca0

NaZ0

Keo

Tota 1

2 . 8 0 . 1 3

I

M2 iton neealula,ted a,a Fe1i l

"ExcoAltt oxqgen alunvlng all iaan il (etuic

l 8 .0 12 .0

50.0

THORNBER AND HUEBNER: IRON CRUCIBLES IN EXPERIMENTAL PETROLOGY

;::i::!q93i;i,:

Fig. 2. ReflectedJight photomicrograph (oil immersion) showing oxidation halo in CO2-Co-treated electromagnet-grade ironcrucible. Arrow points toward edge of crucible. Note medium-grey oxidation rim around yellow prisms (y) and development ofmedium and dark-grey blebs (m-g and d-g). Bar scale represents 25 /rm.

ir::!'d:t;!;:r;:,

only 96 hours resulted in a rind which is -270microns wide. This difference could be due to thepreferential permeation of hydrogen from the COz-H2 furnance gas deep into the metal, where it wouldtend to retard oxidation. (As we will see later, thisconclusion is supported by the relative difference iniron loss with time in runs using CO2-H2 pretreatedcrucibles as compared with similar experiments inCO2-CO treated crucibles.)

The effect of H2 treatment on impure iron cruci-bles is most readily observed in the electromagnet-grade material which contains abundant grey man-ganiferous inclusions (Lipin/Muan Rod), presum-ably a manganous oxide. Exposure to the H2atmosphere at high temperature results in develop-ment of numerous vugs, some of which surroundremnant oxides at the outer extremities of the rod.

In contrast to the behavior of manganese andtitanium, crucible pretreatment appears to changethe bulk volatile content of the metal. Treatment atlow /e, effectively removes oxide tarnish, leavingthe crucibles shiny and bright. We also found thatannealing in hydrogen reduces the carbon contentof electromagnet-grade iron from - 100 ppm (Table2) to

I 150 THORNBER AND HUEBNER: IRON CRUCIBLES IN EXPERIMENTAL PETROLOGY

Table 4. Run summary

RunnumDer

Cruc i b l e*materi al Run data

tem!---imeResul tswtl I ronas FeO

4 .82 . 0I . 30 . 7

1 . 84 .8z . l

' l I

1 . 6

6 . 7E 2

3 . 2'10 .7

1 0 . 79 . 39 . 5

9 . 41 0 . 3I 0 . 0

9 . 71 1 . 4

1 0 . 3

l l . 41 l . 9

46 EM22 EM

EMI O E M6 E M7 E M8 E M

1 4 E M1 3 E M1 5 E M

34 EM

36 EM

Cruc ib le p re t rea tmentatmo3p-h-erTffi !f Tf TTfi E(hrs)

untreateduntreated

6 Z f

82s8?5825

82582582s

825

825

| 050I 050I 050

C0z-C0 1 050c0;-c0 r I 00cot-co 1o5oC02-C0 1100

untreatedunt reatedunt rea ted

no yesno yes

nononon o

yesyesyes

I 100"c /l / 4 h rI 1 00 .c /

l h rn 0n0no

I 250t252

1250125012501250

1250I 25312 50

1250

1250

1252t25lI 250

I 2501 250I 2501250

t252t2s2t252

I 250I 250

t252

1 250I 250

2 3 . 548

1 2 0 . 5

48

48

2448

r 4 3

4868

I 2 0t 2 ? . 5

?47 7

l s l

24I 6 2 . 5

24t 5 l

I2448

t42

96

2492

t . 02 . 1

H2H2H2H2

H2H2H2

H2

H 2

C0rHzcoi-HiC02-H2

yesyesyesyes

yesyesyes

yes

yes

yesyesyes

yesyesyesyes

yesyesyes

yesyes

n0

yesyes

non on ono

n onono

nono

no

nono

?2

2

H2 825Hi 82s

2I

22

I

2?

?

2

1 6 3r 6 01 6 0

96489648

t 742l 843

EMEMEM

EMEMEMEM

VPVPVP

VPVP

VP

VPVP

q

' l o

2021

4l38

44

3940

H2 825

Clt-Hz I I00coi-Hi I loo

1 1 61 1 6

I

EItrt Cory Steel Lot 192474, laneo eleotronagnet-grsdt.ihontWr Moteti.aLt Ruet*Jt Cotponabion Lot 126/1926, VP-gadde 'iton

the FeO content of the melt decreases (Fig. 3). Webelieve that the oxidation of manganese (and per-haps titanium) alloyed in the iron, and the conver-sion of reduced titanium oxide to higher oxides, canreduce FeO initially present in the silicate melt.

Iron loss from melts contained in hydrogen-treat-ed electromagnet-grade crucibles, annealed in avacuum before being loaded, is sporadic and com-parable to that observed in runs using untreatedelectromagnet-grade crucibles (Table 4). Thus, hy-drogen and vacuum heat-treatment of impure iron,despite removing carbon, is unsatisfactory. Therewere no inclusions observed in crucibles of VP ironpretreated in hydrogen or in a COr}Iz gas mixture,and no substantial changes in FeO concentration

were observed in runs using either treated or un-treated high-purity iron containers. Furthermore,no inclusions formed in the crucibles during theseruns.

Electron-microprobe analyses of charges run inimpure crucibles demonstrate that transition metalimpurities diffuse from the capsule and into thesilicate melt. The concentrations of MnO and TiO2within the charge are shown to increase with time(Fig. 5), indicating movement of manganese andtitanium from the crucible and into the charge. Thisobservation, in conjunction with that of an in-creased abundance of oxide impurities at the cruci-ble-charge interface, is suggestive of mutual diffu-sion of manganese and titanium into the charge and

Eo s o E

o

0 4 0 coo

0 3 0 5q)

u0 2 0 f

o

0 r 0 a0)

q)

9 z oo

o 4 0

o

oor 6 0

s

THORNBER AND HUEBNER: IRON CRUCIBLES IN EXPERIMENTAL PETROLOGY l t 5 l

t-J

t 0

l _ -l ' - - - - - l '

. t t l6 7

- 0 0 0! 5 0

Run t ime (hours)

Fig. 3. Percent FeO loss in charge (solid line/left ordinate) and depth ofcrucible oxidation halo (dashed line/right ordinate) ys. timein charges run in H2-treated electromagnet-grade iron crucibles. Length of bars represents total variation of l0 depth measurements.Run numbers correspond to those given in Table 4,

1 4 0r 3 01 2 0I l 0t 0

of oxygen into the crucible. Thus, it appears thatthese impurities must in part be responsible for theobserved ferrous iron reduction.

That substantial manganese accumulated in runs

reported by Roeder (1974) is shown by a compari-son of his original starting compositions with thoseof the run products (Roeder, personal communica-tion, 1981). We find that polished sections of the

Fig. 4. ReflectedJight photomicrographs (air) showing oxidation halo developed in H2-treated electromagnet-grade iron cruciblescontaining charges and run for 8 hours (left) and 24 hours (right) (runs l0 and 6 in Table 4). The crucible-charge interface is at thebottom of each photograph. Bar scale represents 50 pm.

tt52 THORNBER AND HUEBNER: IRON CRUCIBLES IN EXPERIMENTAL PETROLOGY

crucibles used by Roeder contain a substantialamount of manganese-rich inclusions (Table 1). Wehave also noted that in long runs at subsolidustemperatures, manganese-bearing capsules contam-inate pyroxene by causing an anomalously highMnO content (as much as 5.5 wt.%) at the surfaceof originally manganese-poor (0.2% MnO) orthopy-roxene plates.

Both FeO depletion and MnO and TiO2 accumu-lation in silicate charges are alleviated by pretreat-ing the electromagnet-grade crucibles in atmos-pheres more oxidizing than pure hydrogen (Figs. 5and 6). Contamination of the charge by manganeseand titanium and FeO loss was observed in runsusing H2-treated crucibles but was less notable inCO'r}Iztreated crucibles. Increased concentrationof manganese and titanium was barely detectable incharges contained in CO2-CO-treated electromag-net-grade crucibles and FeO loss did not occur inruns of up to 68 hours duration. Minor FeO lossafter heating for -120 hours may have resultedfrom run times that exceeded those of crucible CO2lCO pretreatment (this observation should be born

in mind when making long runs in pretreated cruci-bles). Comparison of the COz-Ifz and the CO2-COresults indicates that hydrogen occluded in theCoz-H2treated crucibles may be partly responsiblefor bulk-composition changes of the charges, aspreviously suggested from the relative differencesin crucible halo development as a function of pre-treatment. Also, in runs using H2-treated crucibles,minor devitrification of the inner wall of the silicatube is observed and may be attributed to thepresence of minor H2O generated through crucible-charge interaction. Pretreatment in CO2-CO gasappears to affect immobilization of the manganeseand titanium by converting these elements from themetallic state (in which they would diffuse rapidly)to oxides (which are relatively immobile in theiron). These observations are also consistent withresults of open system experiments using electro-magnet-grade iron by Schwerdtfeger and Muan(1966) using CO2-CO gas mixtures and those ofRoeder (1974) using COz-H2 mixtures. Schwerdt-feger and Muan reported no appreciable gain ofMnO or TiOz by Mn-Fe-Si-O compositions equili-

4 0

3 0

oO)

N

oF

s=

0 6

o 4

o 2

o d

os)o

c=s3

2 0

0 8

, _ 8

22ltar-' --- 'a7 coz /Hz Pret'eotment -aU

ot^ 4 2\r- coz /9ofiGot'n"nt

Run t ime (hours)

Fig. 5. Wt.Vo TiO2 (above) and wt.Vo MnO (below) vs. time in charges run in electromagnet-grade iron crucibles pretreated in H2(squares), COZ-H2 (triangles), COrCO (circles) and untreated (diamonds). The composition of the starting mixture (Table 3) isrepresented by a hexagon. Each point represents an average of three to five analyses and error bars define one standard deviationabout the mean. Run numbers correspond to those in Table 4.

r 5 01 4 0r 3 01 2 0I t 0

H 2 Preireotmenr_ 8

A'ra6rx;_ 7t a -

ri) 43v

22

#----X:

brated in atmospheres of CO-CO whereas somemanganese accumulation is evident in basaltic-melt-ing experiments of Roeder (1974).

The lack of serious iron loss problems in experi-ments of Lipin and Muan (1975) and Lipin (1978)using electromagnet-grade iron crucibles in evacu-ated silica tubes could be explained by the intrinsicfo, of the specific lot of iron used. As shown inFigure lb and Table l, the "Lipin/Muan" materialcontains abundant elongate grey oxide blebs whichappar to have a composition of mangano-wiistite.Crucibles of this material should have an intrinsic,fo, relatively close to that of the silicate charge inequilibrium with metallic iron. Thus, there would beno tendency for substantial redox interaction be-tween the charge and its container.

Conclusions

It is apparent that the ferrous oxide content ofcharges in iron crucibles varies in response to thelocal redox potential of the system. This redoxpotential can be defined by the "fo" of the ambientatmosphere (in open systems), the incipient fo. ofthe starting material, and the,fo, imposed by theiron crucible and its impurities. It is also dependentupon the temperature and duration of the run, aswell as the presence of other components in thesystem which affect changes in the oxidation stateof iron dependent of "fo, (Thornber et al., 1980;Roeder. 1974).

I 153

As suggested by previous investigators and asindicated by our results, high-purity iron cruciblescan be used to contain iron-bearing silicate in closedsystems without causing substantial bulk-composi-tion changes. However, it is also apparent that thedegree of charge-crucible redox interaction usingimpure iron crucibles can be highly variable. Itappears that oxidation of crucible impurities islargely responsible for ferrous iron loss in iron-bearing oxide materials heated in impure iron cruci-bles and that the extent of iron-loss is dependentupon the intrinsic fo, of the crucible material asregulated by the nature of its impurities. The mostprominent of the complement of reduction reac-tions, written to emphasize the direction of masstransfer, appear to be:

l. Mnlslucible) * FeO1slr..g.;-+ MnO * F€lcharge+crucibte)

2a. Tilcrucu"; * 2FeO1"ha1gs;-+ TiOz * 2F€1"hr.g.+crucible)

or2b. TiO1".'"iore) * F€O1"6u.*",

--+ TiOz * Felgharge+crucibte)

3. Hz(crucibte) * FeOlgharge)--; F€(charee) + HzO

The technique for pretreatment of electromagnet-grade iron crucibles is a critical factor affectingsubsequent redox equilibria in these experiments.

THORNBER AND HUEBNER: IRON CRUCIBLES IN EXPERIMENTAL PETROLOGY

6 0

,t.0

2 0

0.0

oo,oI

c

oor

s; n J

2 2 ,

H 2 prelreotmenl- 8

r501 4 01 2 0

Run i ime (hours)

Fig. 6. Wt.Vo FeO vs. time in charges run in electromagnet-grade iron crucibles pretreated in H2 (squares), C.Or-Hr (triangles) andCOr-CO (circles) and untreated (diamonds). The composition of the starting mixture (Table 3) is represented by a hexagon. Eachpoint represents an average of three to five analyses and symbol size is considerably larger than one standard deviation about themean for each data point. Run numbers correspond to those in Table 4.

THORNBER AND HUEBNER: IRON CRUCIBLES IN EXPERIMENTAL PETROLOGY

Results obtained using crucibles treated at 1050"Cin atmospheres of variable hydrogen content sug-gest that the amount of occluded hydrogen affectsthe degree of exchange of the reducing impuritiesand oxygen between crucible and charge and ispartly responsible for charge reduction. The rateand magnitude of interdiffusion of titanium, manga-nese, and oxygen between crucible and charge isdependent upon the "fo, of capsule pretreatmentand, to a lesser extent, the presence or absence ofhydrogen in the system.

It appears that pretreatment of electromagnet-grade iron crucibles in CO7-CO gas satisfactorilyminimizes ferrous-iron loss by immobilizing themanganese and titanium impurities as oxide blebs(thereby hindering movement of manganese and/ortitanium into the charge and of oxygen into thecrucible) and by removing carbon (and hydrogen)present in untreated crucible material. Such a treat-ment may not be necessary for impure crucibleshaving an intrinsic,fo, which is relatively close tothe Fe-Fe1-*O buffer.

AcknowledgmentsWe want to thank Mary Woodrutr for making many of the

polished thin sections of the crucibles and James McGee formasterful maintenance of our moribund microprobe. Gordon L.Nord, Jr. provided the Mn analysis ofthe surface ofthe annealedpyroxene plate. Donald Lindsley (SUNY-Stony Brook) taughtone of us (JSH) the usage of iron sponge. We appreciate theconstructive reviews of this paper by David Walker (HarvardUniv), Ian Duncan (SMU), Peter Roeder (Queen's Univ) and I-Ming Chou and Bruce Lipin (U.S. Geol. Survey). This effortnucleated from a study of the crystallization of a lunar highlandsfragmental basalt. We are grateful to the National Aeronauticsand Space Administration for providing partial support throughContract T-781H.

References

Biggar, G. M., Humphries, D. J., and O'Hara, M. J. (1974)Experimental crystallization and geochemical interpretation oflow titanium basalts. Lunar Science V,60-62.

Bowen, N. L., and Schairer, J. F. (1932) The system FeO-SiO2.American Journal of Science. 24. 177-213.

Bowen, N. L., Schairer, J. F. and Posnjak, E. (1933) Thesystem, CaO-FeO-SiO2. American Journal of Science, 26,193-284.

Bowen, N. L. and Schairer, J. F. (1935) The system, MgO-FeO-SiOz. American Journal of Science, 29, l5l-217.

Green, D. H., Ringwood, A. 8., Ware, N. G., and Hibberson,W. O. (1975) Experimental petrology and petrogenesis ofApollo 17 mare basalts (abstract). Lunar Science VI, 3ll-313.

Grove, T. L., Walker, D., Longhi, J., Stolper, E. and Hays,J. F. (1973) Petrology of rock 12002 and origin of picriticbasalts at Oceanus Procellarum. Proceedings of the LunarScience Conference 4th, 995-l0ll.

Huebner, J. S. (1975) Oxygen fugacity values of furnace gasmixtures. American Mineralogist, 60, 815-823.

Jarosewich, E., Nelen, J. A., and Norberg, J. A. (1979) Electronmicroprobe reference samples for mineral analyses. Smithso-nian Contributions to the Earth Sciences no. 22 (R. F. Fudali,ed.) 68-72.

Kesson, S. E. (1975) Mare basalts: Melting experiments andpetrogenetic interpretations. Proceedings of the Lunar ScienceConference 6th, 921 -9M.

Lipin, B. R. and Muan, A. (1975) Equilibrium relations amongiron-titanium oxides in silicate melts: The system CaMgSizOe-"FeO"-TiOz in equilibrium with metallic iron. Proceedings ofthe Lunar Science Conference 6th, 945-95E.

Lipin, B. R. (1978) The system Mg2SiOa-Fe2SiOa{aAl2Si2Os-SiO2 and the origin of Fra Mauro basalts. American Mineral-ogist, 63, 350-364.

Longhi, J., Walker, D., Grove, T. L., Stolper, E. M., and Hays,I.F. (1974) The petrology of the Apollo l7 mare basalts.Proceedings of the Lunar Science Conference 5th,447469.

Muan, A. and Schairer, J. F. (1970) Melting relations of materi-als of lunar composition. Carnegie Institute of WashingtonYearbook, 69, 243-245.

O'Hara, M. J. and Humphries, D. I. (1977) Problems of iron gainand loss during experimentation on natural rocks: the experi-mental crystallization of five lunar basalts at low pressures.Royal Society of London Philosophical Transactions, A 286,3 13-330.

Roeder, P.L. (1974) Activity of iron and olivine solubility inbasaltic liquids. Earth and Planetary Science Letters, 23,397-410.

Schwerdtfeger, K. and Muan, A. (1966) Activities in olivine andpyroxenoid solid solution of the system Fe-Mn-Si-O atll50'C. Metallurgical Society AIME Transactions, 236, 201-210.

Thornber, C. R., Roeder, P. L., and Foster, J. R. (1980) Theeffect of composition on the ferric-ferrous ratio in basalticliquids at atmospheric pressure. Geochimica et CosmochimicaActa. 44. 525-532.

Thornber, C. R. and Huebner, J. S. (1980) An experimentalstudy ofthe thermal history offragmentladen "basalt" 77115.Proceedings of the Conference of the Lunar Highlands Crust(J. J. Papike and R. B. Merrill, eds.) p. 233-252.

Walker, D., Longhi, J., and Hayes, J. F. (1972) Experimentalpetrology and origin ofFra Mauro rocks and soil. Proceedingsof the Third Lunar Science Conference, |, 797-81'1 .

Walker, D., Longhi, J., Stolper, E. M., Grove, T. L. and Hays,J. F. (1975) Origin of titaniferous lunar basalts. Geochimica etCosmochimica Acta. 39. 1219-1235.

Walker, D., Kirkpatrick, R. J., Longhi, J., and Hays, J. F.(1976) Crystallization history of lunar picritic basalt sample12002: phase equilibria and cooling rate studies. GeologicalSociety of America Bulletin, 87, 646-656.

Walker, D., Longhi, J., Lasaga, A. C., Stolper, E. M., Grove,T. L. and Hays, J. F. (1977) Slowly cooled microgabbros15555 and 15065. Proceedings of the Lunar Science Confer-ence.8th. 152l-1547.

Wiggins, L. B. and Huebner, J. S. (1981) Combined X-raywavelength and energy dispersive analysis to increase micro-probe eftciency. U.S. Ceological Survey Open File Report#8t-104r.

Manuscript received, January 28, 1982;acceptedfor publication, July 22, 1982.