American Mineralogist, Volume 70, pages 317-331, 1985

Redox equilibria and the structural states of ferric and ferrous iron in meltsin the system CaO-MgGAl2O3-,SiO2-FeO: relationships between

redox equilibria, melt structure and liquidus phase equilibria

BrsnN O. MvseN, DevIo Vlnco, Erse-RA.cNHno Neunre,uN,t AND FnmonlcH A. Surrnr2

G eophy sical Lab or ator y, C arne gie I nstitution of W ashingtonWashington, D.C. 20008

Abstract

Relationships between melt structure and redox equilibria of iron inCaO-AlrOr-SiOr-Fe-O and MgO-AlrOr-SiOr-Fe-O melts with CalAl and Mg/Al > 0.5have been determined with M<issbauer spectroscopy at 1 atm pressure. These data andpublished phase equilibria in iron-free systems were used to calculate liquidus equilibriainvolving iron-bearing melts and iron-free minerals.

Ferrous iron is a network modifler (probably in octahedral coordination) in all compo-sitions studied. Ferric iron is tetrahedrally coordinated in melts with Fe3+pFe > 0.5, andundergoes a gradual coordination transformation in the Fe3*pFe range between 0.5 add 0.3.In this Fe3+pFe-range, tetrahedrally- and octahedrally-coordinated ferric iron may coexist.The temperature-dependence of the Mcissbauer hyperfine parameters and the temperature-independence of the intensity ofthe absorption envelope are consistent with a local structuralunit that may be stoichiometrically similar to Fe.On. The Fe2+/Fe3+ is linearly correlatedwith polymerization (nonbridging oxygens per tetrahedrally coordinated cations, NBOru)and Al/(Al + Si) of the melt. There are linear relationships between log (Fe2*7Fe3+) and logfo, and between log (Fe2+/Fe3*) and 1/T (absolute temperature). The standard-state freeenergy of reduction of ferric to ferrous iron, calculated from these lines, decreases withincreasing Zlrz (ionization potential) of the alkaline earth metal cation, with decreasing bulkmelt NBO/T (more polymerized melts) and with decreasing AV(AI + Si) of the melt.

In magnesium aluminosilicate melts with NBO/[:0.6 and AV(AI + Si):0.2 (typicalvalues for quartz tholeiite and basaltic andesite) with 10 wt.% iron oxide added as FerO, theliquidus phase is tridymite when equilibrated with air. Calculations indicate that at fo" be-tween 10-2 and 10-3 atm the liquidus phase changes to protoenstatite, and then to forsteriteat even lower oxygen fugacities. Substitution of Ca (or Na) for Mg results in expansion of themetasilicate (pyroxene) liquidus field and contraction of that of tridymite. Fractional crys-tallization trends of magmatic liquids are, therefore, significantly dependent on oxygen fu-gacity, degree of polymerization of the magma (NBOA), Al/(Al + Si) and the relative abun-dance of alkali metals and alkaline earths.

Introduction

Crystal-liquid equilibria of natural magmatic liquidsdepend on the structures of both the melt and the minerals.The liquidus phase equilibria in the systemsCaO-AlrO.-SiO, (Osborn and Muan, 1960a) andCaO-FerOr-SiO, (Osborn and Muan, 1960b) indicatethat the influence of Fe3* and Al3+ on the phase relationsat 1 atm pressure are similar (Fig. 1). These available datasuggest that the influence ofAl3* and Fe3* on the anionicstructure of the melts in these two ternary systems mayresemble each other. With the same degree of poly-

r Present address : Mineralogisk-Geologisk Museum, Universityof Oslo, Sars gt. 1, Oslo 5, Norway.

2 Present address: Mineralogisch-Petrographisches Institut,Universitat Kiel, Olshaussenstr. 40-60, 2300 Kiel, Federal Repub-lic of Germany.

merization of the melts (recalculated as nonbridging oxy-gens per tetrahedrally coordinated cations, NBOff3), re-

3 The use of NBO/[ as a compositional indicator requires meltstructural information. The NBO/T can be calculated from chemi-cal compositions as shown by Mysen et al. (1982a, 1984a) afterassignment of tetrahedrally-coordinated cations. As shown else-where (for review of information, see Mysen et al., 1982a, 1984b;McMillan and Piriou, 1983), Al3+, Tia+ and P5+ most likely aretetrahedrally-coordinated in silicate melts compositionally rele-vant to magmatic processes. The structural role of Fe3+ is morecomplex, and is the subject of the present report. The NBO[,when calculated from bulk compositions in this paper, rncludesferric iron only after its structural position in a given rnelt hasbeen specifically discussed. As summarized by Mysen et al.(1982a), free oxygen (that is oxygen that is not bonded to anytetrahedrally-coordinated cation) is not likely in silicate melts aspolymerized as natural magmatic liquids. With this information,the NBO/T can be used as an expression of overall degree ofpolymerization of silicate melts as well as a structural indicator ofindividual anionic units in the melts.

0003-004X/8 5i030,|{)3 I 7 $02.00

3r8 MYSEN ET AL.: FERRIC AND FERROUS IRON IN MELTS

o.o or

"4 r , r , f

, 0 .3

Fig. 1. Liquidus phase equilibria in the systemsCaO-AlrO.-SiOr, CaO-FerO.-SiO, and CaO-FeO-SiO, as afunction of All(Al + Si), Fe3+/(Fe3* + Si; and Fe2+/(Fe2* + Si)for melts with bulk NOB/I: 1. (Data from Osborn and Muan,1960a,b,c.)

placement of Si4+ by Al3* or Fe3+ leads to a rapid re-duction of the silica-polymorph liquidus temperature and atransition from tridymite to pseudowollastonite on theliquidus at approximately the same temperature and thesame M3+/(M3* + Sia*) (M'* : Al3+ or Fe3+) (Fig. 1).These trends indicate that the activity of the SiO2 compo-nent decreases and that of the metasilicate increases atnearly the same rate as the M3+/(M3+ + Sia*) increasesboth for M3+ : Al3* and Fe3*. Aluminum is in tetra-hedral coordination provided that there is suflicient alkalior alkaline earths for charge-balance of Al3+ in the melts(Taylor and Brown, l979a,b; Hess and Wood, 1982; Seifertet al., 1982). Ferric iron may also be in tetrahedral coordi-nation at least in iron-bearing melts equilibrated with air(see also Mysen and Yirgo, 1978; Brown et al., 1978; Dick-enson and Hess, 1981; Calas and Petiau, 1983). These liqui-dus phase equilibria in the systems CaO-AlrO.-SiO, andCaO-FerOr-SiO, differ dramatically, however, from thosewith all iron as Fe2+ (Fig. 1; see also Osborn and Muan,1960c), where replacement of Sia* by Fe2+ (and a con-comitant decrease in Ca2* to maintain a constant NBO/I)results in an increase in the temperature of the cristobaliteliquidus and the appearance of two liquids as the compo-sition approaches the join FeO-SiOr. It is suggested, there-fore, that both ferric and ferrous iron have an importantinfluence on the melt structure but that the effect of the twocations is quite different.

In addition to the seemingly important impact of ironoxides on the structure of silicate melts relevant to magma-tic liquids (and, thus, properties that are affected by meltstructure), redox equilibria of iron are also of intrinsic in-terest because such data may be used to deducetemperature-oxygen fugacity histories of magma (e.9., Hag-gerty, 1978; Haggerty and Tompkins, 1983; Sato and Val-enza, 1980). It is necessary, therefore, to calibrateFe3*/EFe as a function ofintensive and extensive variables

(see, e.g., Sack et al., 1980; Thornber et al., 1980; Kilinc etal., 1983, for empirical calibration of redox ratios of mag-matic liquids). The physicochemical principles that governthe redox equilibria must be established before such datawill have general applicability to natural magmatic pro-cesses.

The present study was conceived to address some of theprincipal features ofredox equilibria ofiron oxides that areinterrelated with the structure of silicate melts and equilib-ria between melts and minerals. In this report, relationshipsbetween All(Al + Si), NBO/T, types of charge-balancingmetal cations and oxygen fugacity have been evaluated.Inasmuch as peraluminous magmatic liquids are compara-tively rare, the compositions of the melts were chosen sothat (Al + Fe3+) is always less than the sum of alkalineearth metal cations.

Experimental methods

Starting materials in the systerns CaO-AlrOr-SiOr-FerO. andMgO-AlrO.-SiOr-FerO. were prepared from spectroscopicallypure SiOr, AlrOr, Fe2Or, MgO and CaCOr in batches of 200 mgand ground under alcohol for I hr before use. The FerO. wasisotopically enriched with 57Fe to about 5?FeTEFe:0.1 to facilitate Mcissbauer spectroscopic analysis. The aluminosilicate com-positions are denoted MAS and CAS for the MgO- and CaO-bearing systems, respectively. The weight percentage of FerO.added to the starting marterials is indicated by the symbols F5and F10 (5 and 10 wt.Vo FerOr, respectively). Roman numerals (Ito XIII) refer to iron-free compositions where NBOI and All(Al + Si) are similar (Fig. 2). In each system there is a series withvariable All(Al + Si) (0.14-{.43) and constant NBOru (0.67) and aseries with constant Al/(Al + Si) (0.334) and changing NBO/[(0.17-1.33). In the system CAS, there is an additional compositionline with different but constant NBO/T (0.86) and changing Af(Al + Si) (0.0}{.43). Finally, there is a composition line thatcovers a range of compositions where both AV(AI + Si) andNBOA are variable (in the ranges 0.08--0.43 and 0.16{.86, respec'tively). These compositional ranges were chosen to cover thosecommonly observed in natural magrratic liquids.

Sintered 20-25 ng pellets of the oxide starting materials weresuspended on single loops of 0.004-in.diameter Pt wire with theweight ratio of the sample to Pt about 100. These pellets weretransformed to a melt in MoSir-heated, vertical quench furnaces.The samples were quenched in water at a quenching rate exceed-ing 500'C/sec. The oxygen fugacity was controlled with CO{O2gas mixtures where the /o, was monitored with a YrordopedZrO, oxygen sensor (Sato, 1972), supplied by Oxide Fabricators,Inc., Victoria, Australia. The oxygen fugacity, as calibrated againstnickel-nickel oxide and iron-wustite oxide buffers (Chou, 1978)and calculated COICO2 (Deines et al., 1974), is accurate to within0.05 log unit. Precision is better than 0.02 log unit. The temper-atures were monitored with one thermocouple 1 cm above thesample and one within the oxygen sensor (displaced I crn horizon-tally from the sample). The temperatures, as checked against themelting point of Au (1062.5'C), are accurate to within *5"C andprecise to ! 1"C.

Electron microprobe analyses show that the compositions ofthe run products are in accord with nominal compositions withinanalytical uncertainty (approximately 2%o, relative) except for iron.As much as a 5o/o iron-loss to the Pt wireJoop (relative to thetotal amount added) was encountered in experimental chargesunder the most reducing /o, conditions. Electron microprobe

7. /t*o tiquias

-/./ '''-inl*,n"

\ /Tr idymia

Y/ /Piddowolle3roniteL\==:1f---s__U:F:3:

1 200

MYSEN ET AL.: FERRIC AND FERROUS IRON IN MELTS

2 0 m 4 o $

in the systems MgO-AlrO.-SiO, and CaO-AlrO.-SiOr. See text for

3t9

Fig. 2. Composition of iron-free end-member starting materialsadditional discussion of bulk compositional relationships.

analyses of the run products are available from the authors uponrequest.

The redox ratio ofiron and structural information on Fe2+ andFe3t in the quenched samples were obtained with 5?Fe resonantabsorption Mrissbauer spectroscopy. The spectra are fitted withleast-squares minimization (Davidon, 1959) to one ferric doubletand one or two ferrous doublets. Several alternative fitting meth-ods, using quenched melts in the systems NarG-SiOr-Fe-O,BaO-SiOr-Fe-O, CaO-SiOr-Fe.O, MgO-SiOr-F+0, NarGAl2O3-SiO2-Fe-O, CaO-AlrO3-SiO2-Fe-O and MgO-AlrOr-SiO, have been evaluated and will be discussed in detail in aseparate paper (Virgo and Mysen, 1985). Only a brief summary ofthe principles and the most important conclusions will, therefore,be presented here.

These fitting routines include those in which it is assumed thatthe cumulative envelope of the absorption spectra consists of anumber of overlapping elementary doublets of Lorentzian shape,and phose in which it is assumed that the hyperfine parameters(isomer shift and quadrupole splitting) are linearly correlated(Wivel and Morup, 1981; Danckwerth and Virgo, 1982; Danck-werth et al., 1982; Virgo and Mysen, 1985). Another method, inwhich a variable number of doublets, of pure Lorentzian lineshape, are assumed for either Fe2+ or Fe3*, has been predomi-nantly used in the present analysis of the spectra. This method ofdeconvolution results in fits that on statistical grounds are com-parable to those obtained by fitting hyperfine parameter distri-butions. As discussed in more detail elsewhere (Virgo and Mysen,1985), the velocities of the component peaks of the quadrupolesplit doublets of Fe3* and Fe2* and the Fe3+/EFe derived withthe two fitting methods are within 5% (relative) of each other.Furthermore, in choosing between, for example, single or multipleFe3* and Fe2+ doublets when fitting separate doublets of Lorent-zian line shape (Virgo and Mysen, 1985, see also Mysen et al.,1984), the average values of the hyperfine parameters (isomer shiftand quadrupole splitting) in the simplest twodoublet (one ferricand one ferrous) fits are within 5% of the weighted averages fromfits of multiple doublets.

It may be argued that the Mtissbauer spectra of silicate glassestheoretically are best described in terms of a distribution of thehyperfine field. The simpler fitting routine, using a limited numberof ferrous and ferric doublets of Lorentzian line shape was, how-ever, employed. This decision was made partly because of theexcessive CPU time required for the least-squares fitting of thehyperfine parameters, and partly (and for the present purposemore important) because the structural information discussed inthe present paper and the Fe3+/EFe obtained from these fits areconsistent with those obtained with the more complex fits of thespectra (Danckwerih et al.,1982; Virgo and Mysen, 19E5) and alsowith independently obtained information.

It is proposed here that the melt structural correlations withdoublets of pure Lorcntzian shape are also compatible with amore realistic model of melt structure in which information fromRaman (Brawer, 1975; Furukawa et al., 1981; Mysen et al.,1980a,b, 1982a), infrared (Domine and Piriou, 1983), transmissionelectron microscopy (Gaskell, 1975) and 2esi NMR spectroscopictechniques (Kirkpatrick et al., 1982) indicate the existence of dis-tinct (20-100A) chemical and structural units in silicate melt struc-tures.

In using this fitting method, all doublets were constrained tohave equal areas. For ferric iron, the halfwidths were also con-strained to be equal. The cumulative envelope of highly reducedglasses (no Fe3+ present) is, however, typically asymmetric (e.g.,Mao et al., 1973; Mysen and Virgo, 1978; Levitz et al., 1980).Thus, in the fitting procedure, the half-widths were not con-strained for the Fe2 +doublets.

The Fe3+BFe [or Fe2*/Fe3* : (1 - Fe3+/EFe)/(Fe3+BFe)]is obtained from relative areas of doublets. The redox ratio thusobtained is within 7'/" (telative) of the results from wet chemicalanalysis (Table 1). With an approximately 3-5% relative uncer-tainty in the Mcissbauer data (as indicated by replicate analyses,replicate experiments and measurements at two different absorbertemperatures, 7'l and 298 K) and a 6% uncertainty reported (Sacket al., 1980) for the wet chemical method, it is considered that a7% difference is within the cumulative analvtical uncertaintv of

320 MYSEN ET AL,: FERRIC AND FERROUS /RON IN MELTS

M d s s b a u e r s p e c t l o s c o p y

Table 1 Comparison of Fe3+/DFe determined by wet-chemicaland Mcissbauer spectroscopic methods

Wet chemistry

ISr"r*. Similar relationships between ISr".*, QSr".* arrdFe3+pFe have been observed in the analogous sodiumaluminosilicate system (Virgo et al., 1983; Mysen andVirgo, 1983a; Virgo and Mysen, 1985) and in thealuminum-free Ca- and Mg-silicate systems (Mysen andVirgo, 1983b; Mysen et al., 1984). These results may indi-cate the existence of a second ferric iron doublet. This hy-pothetical doublet could not be inserted in the fits withoutimposed constraints on its quadrupole splitting. Withoutindependent information on the values of the quadrupolesplitting, insertion of such a doublet with the quadrupolesplitting constrained was not attempted. It is noted, how-ever, that in the analogous Fe3*pFe transition range inthe system NarO-AlrOr-SiOr-Fe-O, it has been found(Virgo and Mysen, 1985) that with those somewhat moreresolved spectra, a second ferric doublet could be included.The present spectra are, however, merely consistent withsuch a possibility.

In the more oxidized samples (Fe3+pFe>0.5) thevalues of ISp".* and QSp.:* are generally similar to thoseobserved for tetrahedrally coordinated ferric iron in crys-talline silicates (e.g., Hafner and Huckenholz, L97l; Anner-sten and Halenius, 1976; Mysen et al., 1980a; Amthauer etal., 1977; Waychunas and Rossman, 1983). Moreover, thesevalues are similar to those found for iron-bearing glasseswhere other spectroscopic data (nxens, nrn, fluorescenceand Raman spectroscopy) also suggest that Fe3+(IV) exists(Brown et al., 1978; Calas et al., 1980; Fox et al., 1982;Virgo et al., 1981, 1982; Calas and Petiau, 1983; Mysenand Virgo, 1983b; Mysen et al., 1980a, 1984).

The gradual increase in ISp.s* from values near 0.3mm/sec (298K) with Fe3+pFe 2 0.5 to values between 0.6and 0.7 mm/sec at Fe3*pFe < 0.4 (Fig. 4) (with a con-comitant decrease in QSp"s*) may indicate a signiflcantchange in the structural position of ferric iron in thesealuminosilicate melts as a function of Fe3*pFe. Similarchanges in hyperfine parameters for ferric iron as a func-tion of Fe3*pFe have been observed in the systemsNarG-SiOr-Fe-O, CaO-SiOr-Fe-O, MgO-SiOr-Fe-Oand Na2O-Al2O3-SiO2-Fe-O (Mysen et al., 1984a; Virgoand Mysen, 1985). Raman spectra of quenched melts in thesystem CaO-SiO2-Fe-O (Mysen et ol, 1984) andNarO-SiOr-Fe-O (Virgo et al., 1982, 1983; Mysen andVirgo, 1983c) indicate that this increase in ISp""* was as-sociated with a systematic decrease in the intensity ofFe3*(IVFO stretch bands in those spectra. This decreasedintensity was interpreted to reflect the disappearance oftetrahedrally-coordinated ferric iron in a ferric iron-concentration range that was considerably greater than thesensitivity of the Raman spectra to Fe3+(IVIO bonds inthe quenched melts (Virgo et al., 1983; Mysen et al., 1984).Further support for this suggestion is found in the values ofisomer shifts of octahedrally coordinated ferric iron(ISp"r* : 0.4--0.65 mm/sec) in crystalline materials (Anner-sten and Halenius, 1976; Annersten and Olesch, 1978;Nolet and Burns, 1979; Evans and Amthauer, 1980; Hug-gins et al., 1975; Annersten et al., 1978; Amthauer et al.,1980). This ISp"r+-raogo is similar to that of the present

001*002*004*010 *FeAb*MV78f

0 . 8 00 , 8 80 . 6 00 . 8 90 . 7 50 . 6 8

0 . 0 50 . 0 50 . 0 40 . 0 50 . 0 50 . 0 4

0 . 8 6 + 0 0 40 . 9 6 + 0 . 0 50 . 6 5 + 0 . 0 30 . 9 6 + 0 . 0 50 . 8 2 + 0 . 0 40 . 7 1 + 0 . 0 4

* S a m p l e p r o v i d e d b y D r . I . S . E . C a t m i c h a e l , U n i v e t s l t yo f C a l i f o r n l a , B e r k e l e y . R e d o x d a t a a r e f r o m M o e t a 1 .( 1 9 8 2 ) a n d C a r m i c h a e l ( p e r s o n a l c o m m u n i c a t i o n , 1 9 8 3 ) .Uncertainty ln these uet-chemlca1 analyses is 6"1. Forbulk chemlcal analyses of these samples, see Table 1 ofM o e t a 1 . ( 1 9 8 2 ) .

tFrom Mysen and Virgo (1978). Bulk composlt lon (byw e l g h t ) : A n 4 6 I o 1 5 . 2 ( S i 0 2 ) 2 9 . j ( F e 2 O 3 ) 8 . t ( o s b o r n , p e r -sonal connunicat ion, 1977; as quoted by Mysen andV l r g o , 1 9 7 8 ) .

the two methods (wet chemistry and Mdssbauer resonant absorp-tron).

Results

Oxygen coordination around ferric and ferrous ironRepresentative sTFe Mijssbauer spectra are shown in

Figure 3, and a complete set of calculated Fe3*pFe andhyperfine parameters from all spectra is given in Table 2.The hyperfine parameters in Table 2 are calculated fromthe statistically best fits (whether with one or two ferrousdoublets). Except for the most oxidized samples, the in-clusion of the Fe2+(I) doublet results in a 2V25% im-provement in the values of y2, and the residual distributionbecomes more random. When its inclusion was not justifiedstatistically, the second doublet was not included.

Topologically, changes in the spectra as a function ofAll(Al + Si), NBOA and fo, are similar in both systems.The spectra have been interpreted to consist of one ferricand generally two ferrous doublets except under the mostreducing conditions (fo,: l}-a - 10-e atm), where thereis no indication of ferric iron. When either one or twoferrous doublets were statistically permissible, the velocitiesof the ferric iron component peaks were independent of thenumber of ferrous doublets.

The high-velocity component of ferric iron shifts from-0.7 mm/sec to 0.9-1.0 mm/sec as Fe3+/EFe is reducedbelow 0.3 (Fig. 3, Table 2). The low-velocity component ofthe Fe3+ doublet is shifted from - -0.5 mm/sec to -0

mm/sec in the same Fe3*/>Fe range. A transition occurs inthe Fe3+pFe range between 0.5 and 0.3, within whichrange intermediate velocities were obtained. Within thistransition range there is an increase in line width of theFe3* component peaks (from -0.7 mm/sec to -1.0mm/sec). The velocity changes are reflected in an increasein the isomer shift (IS) ffrom -0.3 mm/sec (relative to Femetal) in the most oxidized samples to between 0.6 and0.65 mm/sec for samples with Fe3+pFe < 0.3{.41. Theincrease in ISp".* is associated with a general decrease inquadrupole splitting (QS) (Fig. 4) although the QSp""*trend is considerably more scattered than that of the

MYSEN ET AL: FERRIC AND FERROUS IRON IN MELTS

MASIVF5. lO.o'c6 rA$([F5, t55dc cAsxilrF5. r56dc

3

o

4

Fig. 3. Selected 57Fe resonant absorption Mcissbauer spectra (at 298 K) in the systems MAS and CAS. See Table 2 for details ofFe3*BFe and hyperfine parameters.

321

.!

tI

I

I

IITE

!

I

IT

It

I

E

I

I

I

!

E

FI

II

I

II

melts with Fe3*pFe < 0.5. Thus, the correlation of thesystematic changes of isomer shifts of ferric iron with pub-lished Raman data as well as the correlation with hyperfineparameters of octahedrally-coordinated ferric iron in crys-talline materials lead to the conclusion that the gradualand systematic increase in ISp..* (and decrease in QS."r*)

as the Fe3+/>Fe is lowered below about 0.5 results from acoordination change of Fe3 * from four-fold to six-fold.

There is a ferricferrous range where the isomer shiftsand quadrupole splitting are intermediate between about0.3 and 0.65 mm/sec, and with continuous change in thisvalue as a function of decreasing Fe3*pFe. This behavior

-L--!i:!lJ

322 MYSEN ET AL: FERRIC AND FERROUS IRON IN MELTS

Table 2. Experimental results

F e - E e - ( I )

1S* qS rS QS

F e - ( I T )

I s Q s F e - / I F e F e ' / F e -

cAsrF5 1625cAsIIFs 1625CASIIIF5 1625cAsIvF5 1625cAsvFs 1625CASVAF5 1625cAsvB 1625cASVcF5 7625cAsvrF5 1625CASIF1O 7625CASIIF1O 7625c4sII IFl0 1625cAsvFlo \625c A s v I F 5 1 5 7 5cAsrrF5 1550c A s T I l P 5 1 5 5 0cASrvFs 1550cAsvF5 1550cAsvBF5 1550cAsvcF5 1550cAsvIF5 1550cAsvrIF5 1550CASVIIF5i I55OCASVIIIF5 1550CASIXFs 1550CASXF5 1550CASXIF5 1550CASXIIF5 1550CASXTIIF5 1550c A s l v F 5 1 5 5 0cAsIxF5 1550cAsxlrrF5 1550cAsrvF5 1550cAslxF5 1550CASXIIIF5 1550CASIVFs 1550c A s I V F 5 1 5 5 0cAsTXFs 1550CASXIIIF5 1550cAsIt IF5 1500cAsTvF5 1500CASVF5 15OOcAsvAF5 1500cAstvFlo 1550cASVFfo 1550cAsvFlo 1550CASIVFs 1450cAsrvFs 1450cAsTXF5 1400CASXTTIF5 14OOCASM5 1400cAsrvF5 1400CASM5 1400

usrF5 1650M S I T F 5 1 6 5 0M S I ] I F s 1 6 5 0W S I V F 5 1 6 5 0usIVFI0 1650msvrlrs 1650MSVIIIFs 1650lhsrxF5 1650MSXF5 1650MSXIF5 1650msxI lF5 1650usxII lF5 1650M S I I I F s 1 5 5 0ruSIVF5 1550M S V I I F s 1 5 5 0ruSVTTIF5 1550MSIXFs 1550MSXF5 1550rusxrF5 1550MSXrrF5 1550USXIIIF5 I55OruSIVF5 I55OMSIXF5 I55OruSIXFs 1550rusxlrrr5 1550MSIVI5 I55OUSIVE5 1550MSIXFs 1550M S X I I I F 5 1 5 5 0u s l v F 5 1 5 5 0h s I V F s 1 5 5 0USIXF5 1550MSXIIIF5 1550MSTVF5 1450u s I v F l o 1 4 5 0

0 6 80 . 6 80 , 6 80 . 6 80 . 6 80 6 80 . 6 80 . 6 80 6 80 . 6 80 . 6 80 6 80 6 80 6 80 6 80 6 80 . 6 80 . 6 80 . 6 80 . 6 80 . 6 80 6 80 5 80 6 80 6 80 6 80 . 6 80 6 80 , 6 83 . 0 03 , 0 03 0 06 , 0 06 0 06 . 0 09 0 09 . 0 09 . 0 09 . 0 00 . 6 80 6 80 . 6 80 6 80 6 80 6 80 6 80 6 80 6 80 . 6 80 . 6 83 . 0 06 0 09 . 0 0

0 6 80 6 80 . 6 80 . 6 80 . 6 80 . 6 80 . 6 80 . 6 80 . 6 80 6 80 . 6 80 . 6 80 . 6 80 , 6 80 . 6 80 6 80 6 80 6 80 . 6 80 . 6 80 6 83 0 03 . 0 03 0 03 0 06 0 06 , 0 06 0 06 . 0 09 , 0 09 . 0 09 0 09 0 00 , 6 80 . 6 8

cao-A12O3-S i02-Fe-o

0 . 2 9 1 1 . 3 r 90 . 2 9 r 1 . 1 3 70 . 3 0 1 | 3 2 1o . 2 9 2 1 . 1 4 8o . 2 8 9 1 . 3 6 30 . 2 9 9 1 3 0 70 . 3 2 7 | 2 6 5o 329 I 173o 2 7 5 I 3 0 2o 2 3 6 L 4 5 Ao 26I L7A2o , 2 4 9 1 . 2 9 40 , 2 4 7 7 . 2 4 0o 2 7 6 r . 4 1 6o 2 6 3 7 . 4 7 40 , 2 9 5 I 3 r 9o 299 t .3070 , 2 1 5 L 3 1 10 . 2 9 4 I 2 5 8o 2 7 4 r . ] 2 7o - 2 1 8 I . 2 a 2o 293 r 2630 3 9 8 1 . 3 2 00 . 2 7 6 1 3 4 3o 269 I 408o 2 7 3 I 3 3 5o 2 1 8 1 , 2 9 7o 1 0 2 r . 2 2 10 2 8 5 7 . 2 6 2o 6 3 6 0 . 5 7 1o 4 8 7 1 . 0 3 00 5 7 0 0 , 8 2 0o 4 9 9 1 . 5 0 5o 3 8 2 1 . 3 3 0o 7 7 2 0 . 6 6 3

0 305 I 2A70 2 8 3 1 3 0 20 . 2 7 0 1 3 8 1o 2 9 5 7 . 2 8 60 , 2 9 0 1 3 r 1o 2 6 4 I - 3 7 3o , 2 4 1 1 , 3 3 1o 2 7 7 1 . 3 1 6o 3 9 2 1 . 3 s 1o 2 6 5 r . 3 1 9o 2 1 5 I . 2 8 20 . 3 7 0 1 . 1 7 1o . 7 4 2 0 . 6 3 2

1 . 0 7 9 2 7 4 21 . 0 9 5 2 2 4 6r.068 2 r280 . 9 4 1 2 0 1 60 . 9 9 7 r 9 2 41 . 0 4 9 2 I 3 31 . 0 9 4 2 2 0 51 . 0 8 7 2 , O 9 5r . o 2 7 2 , 1 7 41 0 4 0 2 . 2 0 01 . 0 0 4 1 , 9 5 41 . 0 2 1 2 . 0 4 8o 9 r 2 L , 1 2 3r 0 6 2 2 - 2 4 2I 040 2 180I 0 8 7 2 . 0 9 5r oL2 2-135r , r 7 2 2 . 0 1 01 . 0 6 8 2 . r 2 81 . 0 5 7 2 2 2 40 . 9 7 0 1 9 1 41 . 0 0 0 1 . 9 2 31 239 2 4091 0 2 8 2 - 1 6 11 . 0 4 r 2 , 3 0 30 9 8 5 1 . 9 0 61 0 5 0 2 . 2 3 17 t 4 2 2 - 0 5 41 1 8 7 r 8 7 7o 947 2 343o 962 2 4460 9 8 5 2 - 0 5 3r 046 2-193r o49 2 1961 , 0 3 8 2 - 2 7 4r . 0 3 5 2 . 2 5 01 . 1 7 3 2 , 4 0 01 . 0 4 5 2 , 3 3 0r 0 4 1 2 - 1 7 4I o 2 L t 9 0 90 9 9 8 I 9 6 8o 9 1 4 r . 9 2 11 . 0 1 8 1 9 4 10 , 9 9 3 1 . 8 9 70 , 9 8 2 1 . 9 4 90 . 9 3 8 1 . 9 1 60 . 9 8 8 1 . 9 6 17 . r 3 7 2 . 1 4 71 001 I-8911 0 5 4 2 . 1 4 4I 162 7 9621 . 0 2 8 2 . 2 9 31 . 0 2 1 2 . 1 6 2

Mao-Al^0^-SiO^-Fe-0

0 . 4 1 9 7 3 2 3 0 8 3 7 2 - 9 9 8o . 6 1 0 0 , 8 3 0 0 . 8 5 0 2 , 7 t O0 . 4 7 0 | 2 3 4 0 - 9 0 0 2 . 1 r O0 5 9 0 0 . 7 6 0 0 . 8 2 0 2 . 5 6 0o 2 9 0 I . 3 2 0 1 . 0 1 0 2 0 8 00 . 2 9 1 | 4 3 6 1 . 0 5 8 2 . 1 8 10 . 6 s 0 0 . 7 0 0 0 . 8 4 0 2 6 6 00 . 6 4 2 0 . 7 4 5 0 . 9 1 3 2 , 6 3 4o 3 2 2 1 . 4 7 4 1 . 0 9 1 2 3 6 1o 434 1 139 0 830 2 62I0 4 5 0 1 . 1 7 0 0 8 7 0 2 6 8 00 . 4 4 0 1 , 1 4 0 0 8 5 0 2 , 6 L 00 . 4 3 0 7 2 3 0 0 8 I 0 2 - J 7 0o 4 2 3 \ , 3 0 7 A 8 9 2 2 . 9 0 60 3 1 0 1 . 4 5 0 I 2 1 0 2 . 0 1 04 , 4 4 9 I . 2 2 4 0 8 3 9 2 . 8 0 50 , 5 5 2 0 . 9 1 0 0 , 9 5 3 2 . 5 2 30 1 1 3 I - 4 2 9 1 . 0 7 6 2 . 2 9 90 4 4 0 I . 2 4 5 0 . 8 t 4 2 - a 7 4

0 5 5 6 0 6 7 80 . 5 0 0 1 0 0 00 5 2 5 0 9 0 70 5 6 1 0 . ] 4 20 . 7 1 1 0 . 4 0 50 6 1 0 0 . 6 3 90 . 5 6 4 0 . 1 1 30 - 5 1 4 0 . 8 3 90 . 5 5 9 0 . 7 8 90 . 4 4 6 r - 2 4 50 . 5 8 s 0 - 7 0 90 . 5 5 8 0 , 1 9 20 - 7 4 6 0 . 3 4 r0 . 5 6 8 0 . 7 5 90 . 5 6 4 0 . 7 1 4o 5 4 4 0 8 3 90 6 4 1 0 . 5 6 0o 759 0-3770 . 6 1 5 0 . 6 2 70 591 0 6750 . 1 4 9 0 3 t 60 694 0 4410 . 6 5 0 0 . 5 3 80 603 0 6580 . 5 1 8 0 - 9 2 90 . 6 9 9 0 . 4 3 10 , 6 1 9 0 4 7 20 . 6 s 0 0 . 5 3 00 . 6 1 8 0 . 6 1 9o . 2 9 r 2 4 3 60 . 3 2 2 2 , 1 1 00 . 3 1 0 2 . 2 2 50 , 0 1 9 5 1 . 6 30 . 1 5 5 5 . 4 7 2o , o 7 2 7 2 a 90 0000 0000 0000 . 0 0 00 . 6 4 7 0 . 5 4 5o 1 4 2 0 . 1 4 10 . 8 1 5 0 - 2 2 70 . 5 8 s 0 , 4 3 90 . 7 2 2 0 . 3 8 60 . 8 1 7 0 . 2 2 40 , 8 6 3 0 . 1 5 80 . 7 8 1 0 . 2 8 00 . 1 5 4 0 . 3 2 70 . 7 2 4 0 . 3 8 10 . 7 5 8 0 . 3 1 90 - 4 7 0 1 . 1 2 80 . 0 4 4 2 1 . 8 30 . 0 0 0

0 4 4 20 . 3 2 10 . 3 5 00 . 3 1 30 . 4 9 0o . 4 5 70 . 3 2 80 . 1 8 0o . 4 7 30 4 0 10 4 0 10 4 3 50 . 4 8 20 , 5 1 10 5 5 30 . 4 s 20 . 3 0 00 . 5 3 90 . 4 7 8

7 2602. II91 8 5 01 . 8 5 9L . 0 2 61 . 1 9 11 - 9 5 9

I . 1 1 6r . 4 9 51 , . 4 9 2\ . 2 9 9L 0 7 50 . 9 5 80 , 8 0 8L . 2 I 42 . 3 3 20 . 8 5 51 . 0 9 3

0 4 8 3 1 . 0 7 00 4 6 9 1 1 3 10 168 4 9700 0 5 6 1 6 . 8 50 . 0 7 8 1 1 . 8 30 188 4 -3790 . 0 0 00 0000 0000 0000 , 0 0 00 . 0 0 00 . 0 0 00 . 0 0 00 . 5 7 1 0 , 7 4 40 6 1 4 0 . 4 8 3

| , o 2 2 1 . 5 5 61 02r r .5540 . 9 7 8 r 1 0 61 . 0 3 4 1 . 4 3 50 . 9 2 3 I 8 0 7r . 0 2 0 r . 6 2 5r . 0 2 6 I 4 9 8

r . 0 2 6 r . 5 9 41 . 0 3 7 I 6 6 61 . 0 2 0 7 6 2 50 9 1 1 r , 4 9 8

1 0 3 4 r 4 1 5r o 2 4 1 , 6 7 40 . 9 7 8 L , 1 0 60 . 9 5 8 r . 6 0 41 . 0 0 5 r . 2 4 4o , 9 4 9 1 , 4 1 3I 0 1 7 1 . 3 5 5

t 1 0 4 I 9 3 00 9 7 7 1 5 4 91 0 1 2 t . 6 2 7

0 , 9 8 9 1 . 5 1 40 , 9 3 3 r , 1 9 7o - 9 2 a 1 , 8 0 80 . 9 3 6 1 . 7 9 80 9 8 5 1 . 6 5 70 9 9 5 1 . 8 1 01 . 0 r 4 1 . 6 1 0r . o r 2 1 . 5 5 31 0 1 6 r . 6 4 4o . 9 9 5 1 . 4 9 07 . 1 2 7 1 , 7 6 71 . 0 2 0 1 5 6 00 986 r 612

0 . 9 9 6 1 . 6 5 80 . 8 6 4 l - 9 1 41 . 0 2 3 1 . 7 0 6o . 9 9 4 1 . 6 0 0

r 054 r 806o 920 1 880r . o 2 0 1 . 7 5 00 . 8 2 0 1 . 7 3 00 . 8 5 0 1 . 4 9 0o . 9 9 6 I . 4 9 ]0 8 9 0 r . 8 5 00 . 9 4 1 1 8 3 31 0 3 1 r . 7 2 2I 0 3 r I 8 0 20 9 8 0 1 7 1 00 910 1 7000 . 9 9 0 L . 1 2 01 . 0 2 5 1 . 8 r 90 9 1 0 1 . 8 7 01 , 0 1 9 t , ' t 3 50 . 9 9 0 r . 1 4 rr . 0 2 4 r . 5 6 81 . 0 2 8 1 . 7 7 4

o 4 3 8 7 2 3 20 4 4 0 I t 9 00 . 6 0 0 1 . 0 8 00 , 8 1 0 0 1 6 00 9 0 0 0 . 8 7 00 . 5 8 9 I 0 5 8

0 . 3 0 0 r . 4 9 00 . 3 0 0 1 . 4 0 0

0 , 8 2 6 2 . 8 2 8 1 . 0 1 1 1 7 5 30 . 8 5 0 2 . 6 8 0 0 9 9 0 r 1 4 01 0 2 0 2 , 4 7 0 1 0 3 0 | 7 2 01 0 2 0 2 5 2 0 1 . 0 1 0 1 7 8 01 r 8 0 2 . 6 5 0 1 . 1 4 0 L 7 2 01 0 4 0 2 . 3 4 1 1 . 0 1 4 L . 6 1 1I 0 7 0 2 . 2 4 0 1 . 0 3 0 1 . 5 7 01 - 2 1 0 2 . 4 4 0 1 . 1 6 0 I 7 7 01 . 0 8 0 2 . 4 0 0 1 0 5 0 1 . 6 4 01 . 0 7 0 2 , 2 2 0 1 . 0 4 0 1 . 6 1 01 0 8 0 2 . 2 a O 0 9 9 0 I - 6 2 01 2 L O 2 - 4 6 0 1 . 1 6 0 1 7 7 01 , 1 0 0 2 . 2 1 0 1 . 1 6 0 1 - 6 2 01 . 0 6 5 1 . 9 7 0 0 . 8 8 0 1 8 3 0I . 2 1 0 2 . 0 1 0 0 . 9 1 0 I 8 9 01 . 0 4 0 r 9 7 0

* Isoner sh i f t s ( IS) a t quadrupo le sp l i t r lng (QS) in m/sec .e la r ive ro Fe neEat .t u n d e r l i n i n c d e n o l e s s p e c t r u n t a k e n a t 7 7 K .

MYSEN ET AL.: FERRIC AND FERROUS /RON IN MELTS

Fez+ lFe3+

323

0 0 . 1 0 2 0 . 3 0 . 4 0 5 0 6 0 7 0 8

Fa3+/tFo

could be the result ofcoexisting tetrahedral and octahedralferric iron, or it could be the result of localized electronhopping between adjacent "sites" for Fe2+ and Fe3+ asobserved, for example, in the Miissbauer spectra of someiron-bearing silicate minerals (e.g., Nolet and Burns, 1979;Amthauer et al., 1980; Coey et al., 1982). Localized electronhopping is, however, inconsistent with. the observedtemperature-independence of the ferric doublet (see Table2, additional discussion will be presented by Virgo andMysen, 1985). Moreover, as noted above, in melts in thesystem NarO-Al2O3-SiO2-Fe-O, two ferric doubletscould be fitted in the spectra from samples in this inter-mediate Fe3*/EFe-range. One doublet resulted in hyper-fine parameters consistent with tetrahedrally-coordinatedFe3+ and one with octahedrally-coordinated Fe3* (Virgoand Mysen, 1985). In the alkaline earth aluminosilicatesamples discussed here, there is a significant broadening ofthe component peaks of the fitted ferric iron doublet in theFe3*/EFe range between 0.5 and 0.3, a broadening that isconsistent with the existence of at least two ferric doublets.The resolution of the Miissbauer spectra of these samplesdoes not appear, however, to permit a fit without severeconstraints (quadrupole split).

It is concluded, therefore, that in highly oxidized(Fe3*pFe > 0.5) aluminosilicate melts with Ca2+ orMg2* for electrical charge-balance, ferric iron is in tetra-hedral coordination. Similar obsefvations have been madefor melts in the systems NarO-SiO2-Fe-O (Virgo et al.,1982, 1983), CaO-SiOr-Fe-O, MgO-SiOr-Fe-O (Mysenet al., 1984) and NarO-Al2O3-SiO2-Fe-O (Mysen andVirgo, 1983a; Virgo and Mysen, 1985). From the Raman

F62+ lFe3+

MgO-Al203-SiO2

cao-Al 2 ()3 -Sio2

0 0.1 02 0.3 0.4 0,5 0 6 01 0,8

Fo3+/tF6

spectra of the alkali and alkaline earth silicate melts, theRaman spectra indicate that this ferric iron was highlyordered (either constant Si/Fe3*, or no Si in those tetra-hedra). Similar conclusions may apply to the alumi-nosilicate systems.

In all systems and with different iron contents, a gradualincrease in ISp,"r* (and a decrease in QSp".*) began to takeplace with Fe3+BFe decreasing below 0.5. This latter ob-servation leads to the conclusion that the coordinationtransformation of ferric iron appears tied to the relativeproportion of ferric and ferrous iron in the melts. Thus, amelt complex involving an association between these twocations (e.g., Fe.On) may be suggested. Magnetic propertiesof glasses in the system CaO-SiO2-Fe-O as a function ofFe3+pFe (O'Horo and Levy, 1978) are also consistentwith this suggestion. As additional ferrous iron is formedwith further reduction of Fe3+/EFe, this additional Fe2*most likely occurs as a network-modifying cation linkingthe non-bridging oxygens in the various coexisting units.

Ferrous iron is commonly considered to be a networkmodifier in silicate melts. Coordination polyhedra with sixoxygens are consistent with the values of isomer shifts andquadrupole splitting of both ferrous iron doublets (e.g.,Mao et al., 1973; Bell and Mao, 1974; Nolet et al., 1979;Mysen and Virgo, 1978; Calas and Petiau, 1983; Mysen etal., 1980a).

Redox relations

The Fe2+/Fe3* values are positively correlated with de-creasing NBO/T and Al/(Al + Si) at the same temperatureand oxygen fugacity (Fig. 5). Similar correlations have been

o

Fig. 4 Variations in hyperfine parameters (quadrupole splitting and isomer shift) in the systems MgO-AlrO.-SiOr-Fe-O (opencircles) and CaO-AlrO.-SiOr-Fe-O (closed circles) as a function of measured Fe3 *BFe. Isomer shifts are relative to Fe metal.

324 MvS.EN ET AL.: FERRIC AND FERROUS IRON lN MELTS

{@[ A/tA + Sl

Fig. 5. The Fe2+4i6:+ of quenched melts in the systems CaO-Al2O3-SiO2-Fe-O (CAS) and MgO-AlrOr-SiO2-Fe-O (MAS) as afunction of NBOI at Af(Al + Si) : 9.334 and as a function of Af(Al + Si) at indicated NBOA *,ith 5 wt.% iron oxidc added as FerO.and equilibrated in air at 1550"C.

a €t , !

reported from several Al-free alkali and alkaline earth sili-cate systems (e.g., Larson and Chipman, 1953; Paul andDouglas, 1965; Goldman, 1983; Mysen et al., 1984). Thereis a near linear relation between redox ratio and NBOI inthe NBOI range between about 0.3 and 1.0 (see also Fig.5) in both the system CaO-AlrOr-SiOr-Fe-O and thesystem NarO-Al2O3-SiO2-FeO (data also from Mysenand Virgo, 1983a; Virgon and Mysen, 1985). In theMgO-AlrOr-SiOr-Fe-O system the linear correlation be-tween Fe2*/Fe3+ and NBO/I obtained for Ca- and Na-aluminosilicate glasses no longer exists (Fig. 5). This differ-ent behavior of the Fe2+/Fe3+ versus NBO/T results fromthe fact that even for MAS melts with NBOA < 0.4equilibrated with air at 1550'C, the Fe3+pFe is less than0.5, and some of the ferric iron no longer is in tetrahedralcoordination. Inasmuch as the activity coeflicients ofFe3+(IV) and Fe3*(V! most likely differ from each other,the Fe2+/Fe3+ will depend on Fe3*1Iv)7Fe3*1Vg as alsoindicated by the data in Figure 5.

In the Al-free alkali- and alkaline-earth-bearing end-member systems the Fe2+/Fe3+ is a linear function of Zlr2(ionization potential) of the dkaline earth and alkali metalcation (Mysen et al., 1984), and NBOI has no apparenteffect on this relationship. In the present, aluminous sys-tems, the p.z+/pe3+ also increases with increasing Zlr,(Fig. 6), but the relationship is distinctly nonlinear and thedependence of Fe2+7Fer+ on Zfr2 is more pronounced themore polymerized the melt (smaller NBO/D.

With ferric iron in tetrahedral coordination, log(Fe2+Ae3*) is linearly correlated with l/T (absolute tem-perature) andlogfo,Figs. 7 and 8). The nonlinearity of theMASM5 curve (Fig. 7c) results from a coordinationtransformation of Fe3+ in this composition as the temper-ature is increased from 1550" to 1650'C (see Table 2).

For melts with Fe3+(IV) and Fe2+(vI), the Fe2+/Fe3+

decreases with decreasing degree of polymerization (in-creasing NBO/D of the melts (e.g., Larson and Chipman,1953; Douglas et al., 1965; Virgo et al., 1982; Goldman,1983; Mysen et al., 1984). As a result, it has been suggested(Holmquist, 1966; see also Goldman, 1983) that the redoxequilibria may be illustrated with a simplified expression ofthe form:

4FqIUO; :4Fe2*(VD + 602- + 02. (1)

z?

Fig. 6. The p"2+Fs3+ of melts in the systemsNarO-AlrOr-SiOr-Fe-O, CaO-AlrOr-SiO2-Fe-O and MgG-Al2O3-SiO2-Fe-O expressed as Zfr2 of the alkali and alkalineearth cation as a function of bulk melt NBOff at 1550"C and theoxygen fugacity of air. (Data for the system NAS from Mysen andVirgo, 1983a; Virgo and Mysen, 1985.)

TsmFratom,'C

1350 1450 1550

casvFl0

MYSEN ET AL.: FERRIC AND FERROUS /RON IN MELTS

TemtsrduB,'c

1350 1450 1550 1650

325

TmF6tore, 'c

1350 1i150 l5S t650

- 0 1

i - { 5

9 - 0 6

-07

o 2

o.l

- 0 1

4 2

-o,3

-0.7 r-0.. L-r.0

A . 9I

0.6.-

With the assumption of a linear relationship the standard-state free energy of reduction of ferric to ferrous iron ex-pressed with equation (1) can be estimated as the lines (Fig.7) conform to the expression

ln K: -LGolRT, (2\

Fig. 8. Relationships between redox ratios and oxygen fugacityat 1550"C as a function of the same bulk compositional variablesas the data in Fig. 7. Dashed lines for systemNarO-AlrOr-SiOr-Fe-O are from Mysen and Virgo (1983a) andVirgo and Mysen (1985).

6 5 6 0 5 5 5 0

t/T x to.(K-r l

where the equilibrium constant, K, is that of equation (1).Strictly speaking, the concentration ratio, Fe2+/Fe3*should be the activity ratio, ap"r*fap".*. The linear relation-ship between log (Fe2+/Fe3+) and llT indicates, however,that the activity coefficient ratio, yp,r,f.!p.3*, is temperatureindependent in the temperature-range studied here. Thestandard free energies of reduction are given in Table 3. Itis evident that at the same NBO/T and Af(Al + Si), theAGo of reduction is greater for Ca-aluminosilicate meltsthan for Na-aluminosilicate melts with Fe3 + in tetrahedralcoordination and Fez+ in octahedral coordination. Fur-thermore, the free-energy changes decrease with decreasingbulk melt NBO/[ and with decreasing AV(AI + Si) forboth Na and Ca systems. The relationship of AGo to Af(Al + Si) is less pronounced for higher values of bulk meltNBOA (Table 3).

There is a near linear relation between log (Fe2*/Fe3+)

Table 3. Standard free energies ofreduction (kcal/mol) offerric toferrous iron

C o n p o s i r i o n N B O / ? * A 1 l ( A l + S i ) A c o

6 0 5 5

r/T x r04{x- l)6 0 5 5 5 0

1/T r 1o4lK-l)

Fig. 7. Log (Fe2*//Fe3*) vs. 1lT (K) as a function of type of metal cation at constant NBOff and Al/(Al + Si); as a function ofAl/(Al + Si) with the same metal cation and NBO/T; and as a function of NBOI with constant AV(AI + Si) and metal cation. Dashedlines for NAS obtained from data for the system NarO-AlrO.-SiOr-Fe-O(Mysen and Virgo, 1983a; Virgo and Mysen, in prep.).

I

CASIVF5CASXIIIF5CASIXT5cAsvF5cAsvAF5cAsvBF5cAsvcF5cAsvFl0

NASIVF5+NASIXF5TNASXIIIF5

0 . 3 3 40 . 1 3 80 . 3 3 4o . 4 2 90 . 2 6 L0 . 1 4 80 . 0 5 00 . 5 2 9

0 . 3 3 40 . 3 3 40 . 1 3 8

0 . 6 50 . 1 70 . 8 50 . 8 50 . 8 50 . 8 50 . 8 5

0 . 6 50 . 1 70 . 6 5

7 . 7

6 . 9

3 . 8

3 . 87 . 7

3 , 9

3 . 4

*Nonbrldglng oxygens per Eetrahedral cat ions ln theiron-free end-member sysEen.

tData froD Mysen and Virgo (1983a).

-{),

not to2

t26 MYSEN ET AL; FERRIC AND FERROUS IRON IN MELTS

and log /", (Fig. 8). Similar relationships have been ob-tained in complex natural magmatic liquids (e.g., Thornberet al., 1980) as well as in numerous simpler binary andternary silicate systems (e.g., Seifert et al., 1979; Goldman,1983; Virgo et al., 1983; Mysen et al., 1984). In contrast toNa-bearing systems, where the slopes of the log /o, vs. log(Fe2+/Fe3*) are about -0.25 for a range of bulk meltNBO/T values, in the alkaline earth aluminosilicate meltsystems the slopes range from -0.15 to -0.37. The slopesgenerally become steeper with increasing Zlr2 of the alka-line earth cation. No clear trend emerges for the relationsbetween NBOII or Al/(Al + Si) and the slopes of the redoxcurves. This observation differs from the data reported inthe Al-free end-member system, where, for example, theabsolute value of the slope increases with decreasing bulkmelt NBO/T (Mysen et al., 1984). Evidently, this simplerelationship is affected by the presence of significant pro-portions of Al3 + in the melts.

Discussion

Polymerization and redox equilibria

Ferrous iron is a network modifier in the silicate meltsstudied here. Ferric iron, on the other hand, can be both anetwork former and a network modifier. Its coordinationpolyhedron is principally a function of Fe3+pFe. Thedegree of polymerization (NBO/T) of the iron-bearing sili-cate melt depends, therefore, on the same intensive andextensive variables that affect Fe3*pFe. A possible excep-tion to this rule may be peralkaline aluminosilicate meltswhere Dickenson and Hess (1981) found Fe2+/Fe3* tobe essentially independent of NBO/T. Some calculationsused to illustrate the extent of changes of NBOI in thesystems NarO-AlrO.-SiO, (Mysen and Virgo, 1983a;Virgo and Mysen, 1985), CaO-AlrO.-SiO, andMgO-AlrO.-SiO, are shown in Figure 9.

The Fe3+pFe is a simple function of temperature (Fig.7). Thus, the NBO/T (degree of polymerization) of melts inthe systems NAS, CAS and MAS at the same Al/(Al + Si)and oxygen fugacity increases with increasing temperature.Because the bulk melt NBO/T increases with decreasingFe3+/EFe as long as Fe3+ is in four-fold and Fe2* is insix-fold coordination, the NBOA will increase with in-creasing temperature, with decreasin g fo" and with decreas-ing Al/(Al + Si). Inasmuch as the temperature dependenceof Fe3*/EFe is greater the greater the Zlr2 of the alkalineearth or alkali metal cation, the NBO/T of iron-bearingmelts becomes more distinctly dependent on the type ofmetal cation in the order Na < Ca < Mg.

Lowering of the oxygen fugacity reduces the Fe3*pFeand thus results in enhancement of NBO/T. For the threesystems MgO-AlrO.-SiOr-Fe-O, CaO-AlrO.-SiOr-Fe-O and NarO-AlrO.-SiOr-Fe-O at 1550"C, Fe3* willundergo a coordination transformation at some ./o, lessthan that of air (Figs. 4 and 8). It can be seen from theresults in Figure 98 that the NBO/T changes by as muchas l5-25'lo as a function of decreasing for.'lhe change ofslopes of the curves in Figure 9 and eventual transition

c

T . l $ c

Fig. 9. Variations in melt n"rrr*n"",,." (NBOfD as a func-tion of Fe3+BFe calculated as a function of temperature (A),oxygen fugacity (B) and Al/(Al + Si) (C). Reference points,(NBO/T)o, are at 1450'C (A), -log./o, : 0.68 (air) (B) and Al/(Al + Si) : 0.0 (c).

from an increase to a decrease in NBO/T with decreasing

.fo, stem from the gradual coordination transformation offerric iron from IV-fold to Vl-fold coordination. This ap-parent contrast in the functional relations between NBO/Tand Fe3+pFe for Fe3+(IV) and Fe3+(VI) results from thefact that with Fe3* in octahedral coordination, the redoxequilibrium may be illustrated with the formalized ex-pressron;

Fe3+(VD + O- : Fe2+(V! + U2O2, (3)

where O- is nonbridging oxygen in the silicate network.Consequently, increasing Fe3 +/EFe at constant /e, requiresa decreasing activity of the nonbridging oxygen. That is tosay, the bulk melt NBO/T will, in fact, decrease as theresult of reducing Fe3+(V! to Fe2+(VI). This transition ismore rapid in the system MAS than in CAS and morerapid in CAS than in NAS.

The negative (and generally linear) correlation betweenFe2*/Fe3+ and Al/(Al + Si) also results in increasing bulkmelt NBO/T as a function of decreasing All(Al + Si).Consequently, even though the structural positions ofAl3+and Sia* in aluminosilicate melts most probably remainunaltered by the changes in Af(Al + Si) (divalent metalcations charge-balance Al3* in four-fold coordination), thebulk melt NBO/T of the iron-bearing MAS, CAS and NASsamples decreases systematically with increasing AV(AI

t

I

I

E

I

t

+ Si). Inasmuch as Fe3+pFe is more sensitive to AV(AI+ Si) in the iron-bearing magnesium aluminosilicatesystem than in the other systems (Fig. 5), the NBOI vari-ations resulting from the change in Fe3+/EFe are alsomore sensitive to Al/(Al + Si) in the systemMgO-AlrO.-SiOr-Fe-O than in the systemsCaO-AlrO.-SiOr-Fe-O and NarO-AlrOr-SiOr-Fe-O(Fig. e).

Liquidus equilibria and the structure ofiron-bearing aluminosilicate melts

Liquidus phase-equilibrium relations and melt-mineralelement partitioning depend on the bulk compositions(structure) of both the mineral and the coexisting liquid.For example, a survey of REE partition coeflicients be-tween garnet and magmatic liquids (Irving and Frey, 1978)indicates that the partition coeffrcients may vary by morethan an order of magnitude as a function of magma com-position alone. Compositional variables, recast to atomicratios such as Si/O of the melt, have been used to obtain acorrelation between this ratio and partition coeflicients be-tween olivine and such melts (Watson, 1977: Hart andDavis, 1978). The Si/O is a simplified expression of meltpolymerization, but does not take into account other tetra-hedrally coordinated cations (e.g., Al3*, Fe3+, p5* andTi4*), distribution of such cations between different struc-tural units in the melt (e.g., Mysen et al., 1981, 1982b, 1985)and the possible complexing of the trace element with spe-cific elements in the melt (Watson, 1976; Ryerson andHess, 1980). It has been found for Ni, Mn and several REEthat the polymerization parameter, NBO/T, which doesnot, however, consider cation distribution or complexing, islinearly correlated with olivine-liquid and diopside-liquidpartition coeflicients in aluminosilicate systems such asCaMgSirOu-NaAlSiOn-SiO, and MgrSiOn-NaAlSi.Or-CaAlrSirO, (Mysen and Virgo, 1980). The linear corre-lations at constant temperature are valid whether or notCa2+, Mg2+ or Na+ acts as a charge-balancing cation oftetrahedrally coordinated Al3+ at least in the NBO/Trange from 0.2-1.0.

One may therefore relate changes in melt-mineral parti-tion coeffrcients that result from the relationship betweenredox ratio and NBO/T of the melt to intensive and exten-sive variables that govern the values of Fe3+pFe (Fig. 9).For example, the variations in NBO/T as a function of/o,or Al/(Al + Si) with 5 wt.Vo iron added as FerO. to a meltwith NBO/T:0.6 are well within the NBOA ranges oflinear correlation between partition coeflicients andNBO/T in the systems summarized by Mysen and Virgo(1980). As mentioned briefly above [see equation (l)], ferriciron as a network former most likely forms separate ferritecomplexes (Fox et al., 1982; Virgo et al., 1982; Mysen et al.,1984). These complexes may be described with the gener-alized expression FeOf -2")* (Holmquist, 1966; Goldman,1983). Virgo et al. (1982) and Mysen and Virgo (19S3b)concluded from Mcissbauer and Raman data that mostlikely n : 2 in alkali and alkaline earth silicate melts. Virgoet al. (1982) concluded from the Raman data, for example,

327

that the character of this complex does not depend on theamount of tetrahedrally-coordinated Fe3+. Only the rela-tive abundance of the ferrite complex changes with ferriciron concentration. The relative abundance of FeO, com-plexes (which are analogous to the three-dimensionally-interconnected aluminate or silicate complexes found inthese and other melts) in a melt will affect the bulk meltNBO/T but not the character of the bridging and non-bridging oxygens in the silicate anionic units. Equation (1)may be combined with any polymerization reaction for thesilicate such as, for example;

SiO3- : SiO2 + O2-, (4)

to illustrate the relationship between Fe3+(I9rFe2+(VDand the NBO/T of the melt:

6SiO3-(2) + 4Fe2+(VI) + 02

:6sior(0) + 4Fe(IV)Ol. (5)

In equation (5), the numbers in parentheses represent theNBO/T (or NBO/Si) of that particular melt structural unit.The Roman numerals represent the oxygen coordinationnumber around Fe3+ and Fe2*.

The changes in olivine-liquid Ni and Mn partition coef-ficients in the system MgrSiOo-NaAlSi.Or-CaAlrSirO, at1550"C have been calculated (Fig. 10). The Ni and Mn dataare from Hart and Davis (1978) and Watson (1977), respec-tively. The 1550"C results are calculated by extrapolationof partition coeffrcients that are linearly correlated (r is acorrelation coeflicient) with NBO/T at constant temper-ature from 1250"C to 1450'C. The correlations at the dif-ferent temperatures are as follows:

K&tl'iq (Hart and Davis, 1978):

1250'C: K : -24.1(NBO/T) + 23.6, r :0.95, (6)

MgO-A|2O3-SiO2-FFO 5n % F.2O3 d&d

MYSEN ET AL.: FERRIC AND FERROUS IRON IN MELTS

N B O n = 0 . 6 i l A l / l A l + S ) - Ot o 2 ' 1 O - 0 5 3 , T - l S ' C

0 1 0 2 0 - 3

N B o f r = 0 6 d l o ! f o 2 - 4 6 8

A/ {A l + S i l= OgT = tffi"C

NBo/T = o6 rt fo2 = to-o 53

a r / t a r + $ ) = 0 3 sl - l f f c

1 2 3 4 5_ tq roz

* . 6

ti 1

Fig. 10. Calculated changes in hypothetical olivine-liquidmanganese and nickel partition coellicients in the systemMgO-AlrO.-SiOr-Fe-O with 5 wt.% iron oxide added as FerO,relative to values at Al/(Al + Si) :6.6 and -log./o, : 0.68, re-spectively. Base element partitioning data from Watson (1977\ andHart and Davis (1978). See text for details ofdata extrapolation.

328 MYSEN ET AL.: FERRIC AND FERROUS /RON IN MELTS

1350'C: K: - l2.4G'fBO/rD + 15.9, r:0.97, (7)

1450"C: K: -5.11(NBO/T) + 0.77, r :0 .98, (8)

which at 1550"C yields

1550'C: K: -4.52(NBO/T) + 8.72,r :O.99. (9)

K$;tte (watson, 1977):

1300'C: K : -1.02(NBOA) + 1.59, r : 0.91, (10)

1350"C: K: -1.10(NBOA) + r .62, r :0 .96, (11)

1450"C: K: --0.59(NBO/T) + 1.23,r:0.98, (12)

which at 1550" yields

1550'C: K: -0.44(NBO/D + 1.08, r :0 .99, (13)The calculated examples (Fig. 10) are for the end-

member system MgO-Al2O3-SiC2 with NBO/T about 0.6.Five wt.% iron oxide as FerO, was used in the calcula-tions. The relationships between Fe3+BFe, fo. and All(Al + Si) (Figs. 5 and 8) were used to calculate theNBO/T and thus the changes in 61t"t-tio as a function ofAl/(Al + Si) and/o,. For this amount of total iron, changesin Af(Al * Si) alone have a real but relatively small effecton the partition coeftcients in the range All(Al + Si) : g-0.5. At lower temperature, or with more iron added to thesystem (e.g., as found in natural basaltic liquid), the influ-ence of Al/(Al + Si) on the partition coeffrcients will beenhanced because the partition coeflicients are more de-pendent on NBOI at lower temperature.

The changes in partition coeffrcients with/o, were calcu-lated relative to the value of Fe3*pFe in melts equili-brated with air at 1550'C. A reduction of the oxygen fu-gacity by approximately two orders of magnitude results ina transformation of all Fe3+ from tetrahedral to octahedralcoordination, under which conditions the NBO/T of themelt reaches a maximum value. This maximum corre-sponds to about an 8%o decrease in the value of the parti-tion coellicients. At lower temperatures, where Fe3+pFemay be more sensitive to fo, and the partition coellicientsare more strongly temperature dependent, with iron con-tents comparable to those of basalt, the effect of/o, on thepartition coeffrcients may be more than twice that shown inFigure 10. Thus, oxygen fugacity significantly affects thecrystal-liquid element partitioning of transition elementssuch as Mn and Ni even though the oxidation states of thetransition elements themselves may not be affected.

The liquidus phase equilibria of Fe-free minerals in iron-bearing aluminosilicate systems can be correlated with thestructure of the coexisting iron-bearing aluminosilicatemelt. A few calculated examples of liquidus equilibria in thesystems CaO-AlrO.-SiO, and MgO-Al2O3-SiO2 with 10wt.o% iron oxide added (in the calculations) as FerO, areshown in Figure 11. The composition of the iron-free meltused in this calculation is defined by the intersection of theline Al/(Al + Si) : 0.2 with the metasilicatrtectosilicateliquidus boundary (Fig. 11). The NBOI of melts along theline All(Al + Si) : 0.2 (labeled c in Fig. l1E and F) as afunction of Fe3 + DFe and coordination of Fe3 + was calcu-

lated. The NBO/T at Al/(Al + Si) : 0.2 was then projectedinto the ternary composition plane, and the phase equilib-ria were derived from the published data (Osborn andMuan, 1960a,d). The compositions thus projected willmove along the line Al/(Al + Si):0.2 as defined by thetotal iron content and Fe3 +pFe.

In both systems the addition of 10 wt.% FerO. shifts thecomposition from the metasilicate-tectosilicate liquidusboundary into the liquidus field of a silica polymorph. Thetemperatures are between 70'and 100'C higher than thoseat the liquidus boundaries of the iron-free compositions.The metasilicate-tebtosilicate liquidus boundary is reachedat Fe3+BFe between 0.65 and 0.45. This liquidus bound-ary is encountered at a greater value of NBOru and, thus,at higher Fe3*pFe in the system CAS than in the systemMAS. This difference probably results from the greater dis-persion of anionic units in magnesium aluminosilicate thanin calcium aluminosilicate melts (see Liebau, 1981; Mysenet al., l982a,b). That is, even at the same bulk melt NBOA,the abundance of SiO!- and SiO, units relative to Si2O3-units is greater in the magnesium system than in the cal-cium system. Such differences are consistent with the meltstructure data in the systems CaO-SiO, and MgO-SiO,(Mysen et al., 1982a). This liquidus boundary represents aminimum temperature as further lowering of Fe3+BFe ismanifested by metasilicate on the liquidus with increasingthermal stability. The steeper slope of the liquidus surfacein the Ca system is a direct consequence of the topology ofthe liquidus surface of pseudowollastonite in the systemCaO-AlrO.-SiO, (Osborn and Muan, 1960ad). The in-flection of the metasilicate liquidus surfaces and thermalmaximum at Fe3+pFe -0.3 (Fig. llA and B) is due to thecoordination transformation of Fe3+. This transformationresults in a maximum value of NBO/T at this Fe3+pFeand thus the topological features shown in Figure 11. Thedashed lines fdenoted Fe3+(IV]) represent the slopes ofthemetasilicate liquidi calculated for a hypothetical situationwith Fe3 + in tetrahedral coordination in the entireFe3 */EFe range.

The liquidus phase equilibria in these systems may alsobe displayed as a function of recalculated All(Al + Si)along lines of constant Fe3+pFe and, therefore, constantNBO/T. The shaded areas in Figure l1 (E and F) representthe compositional ranges obtained by adding 10 wt.% ironoxide as FerO, with Fe3*pFe between 1.0 and 0.0 toaluminosilicate melts with NBO/T : 0.6. The topologies ofthe liquidus surfaces along composition lines withFe3+BFe : 1.0, 0.5 and 0.0 are also shown in Figure 11 (Cand D).

The addition of ferric iron, and a concomitant increasein polymerization (from NBO/T:0.6 to -0.42), result inan increase in the liquidus temperature relative to the iron-free solvent across most of the Al/(Al * Si) range (0.G{.5)(Fig. l1C and D). A silica polymorph is on the liquidusto Al/(Al + SD > 0.1. For a given Fe3*pFe, this range inAV(AI + Si) is 30-60% wider in the magnesium alumi-nosilicate system than in the calcium aluminosilicatesvstem.

l0s* F.zO3 dd.d

OO-AI2Or -gO2-F.--O

Ar/{Ar+r$}.0.2 / ^..9"I t "

Trilyhil /./

FllmiD

AN@/T.0.108

FJ'tDF.

MySEN ET AL: FERRIC AND FERROUS IRON IN MELTS

Fig. f 1. Calculated liquidus phase equilibria at I atm pressure in the systems CaG-AlrOr-SiO2-Fc-O and MgO-AlrO3-SiO2-Fc-Owith 10 wt.7o iron oxide added as FerO.. In C and D the solid line denotes Fe3+pFe : 1.0; dashed line, Fe3+72Fe :0.5; dotdashedline, Fe3*pFe : 0.0. In E and F the line labeled a denotes AI/(AI + Si) : 6.2. See text for additional details.

329

r20 L1,0

' . 12S0

IE 1260

I

o9IB

I

o

Ett.a

0.5

F.3+/2 F.

The value of Af(Al * Si) at the tectosilicate-metasilicateliquidus boundary decreases and the temp€rature increasesas ferric iron is reduced to ferrous iron (Fig. llC and D).These changes result from the increased bulk melt NBOruwith reduction of Fe3+ to Fe2*. This same reduction inFe3+pFe is also accompanied by rapid expansion of themetasilicate liquidus volume in the calcium system and bythe appearance of orthosilicate (forsterite) in the mag-nesium system. There is no aluminosilicate liquidus phasein the magnesium system, and spinel is a liquidus phase asthe AV(AI + Si) is increased further. In the calcium alumi-nosilicate system, a further increase in Af(Al * Si) resultsin the appearance of anorthite with a thermal maximum onthe liquidus surface. Thus, merely increasing Al/(Al + Si) atconstant NBOA of the melt results in a transition firstfrom a tectosilicate to a metasilicate liquidus phase (de-creased polymerization of the liquidus phase), then to acompletely polymerized aluminosilicate phase. The liquidusvolume of anorthite shrinks with further reduction of ferriciron. Finally, the highly depolymerized pyrosilicate, gehle-nite, instead of anorthtite, occurs on the liquidus for com-positions with nearly all iron as Fe2* and Af(Al + Si)about 0.5. Mysen et al. (1982a,b, 1985) noted the existenceof analogous liquidus phase relations in the systemNarO-AlrO.-SiO, (see also Osborn and Muan, 1960e).They related these changes in liquidus phase equilibria to astrong preference of Al3+ for three-dimensional networkunits in the melt. As a result, increasing Af(Al + Si) isaccompanied by greater relative abundance of TO,

(T: Al + Si) units. In order to maintain mass-balance(constant NBO/T), the relative abundance of silicate anion-ic units with their NBO/I greater than that of the meltbulk must also increase. These abundance changes in themelts are, then, reflecled in the changes in liquidus phaseequilibria as a function of Af(Al + Si) of the system.

It is evident from the above discussion that in the rangesof Al/(Al + Sil, NBOft, Fe3*pFe and total iron contentof natural magmatic liquids, major changes in the liquidusequilibria of iron-free minerals take place simply as a func-tion of the oxidation state of iron. The transition fromanorthite to gehlenite as a liquidus phase is the most ex-treme example (NBO[ contrast of 3 across the liquidusboundary).In a tholeiitic liquid, the NBOA typically is inthe range 0.H.8 and the AV(AI + Si) is 0.2-O.3 (see Mysenet al., 198?a" for calculations). From the model system cal-culations (Fig. 11) one may suggest, therefore, that a highlyoxidized basalt could have tridymite (or quartz) on theliquidus. Reduction of ferdc iron would result in the appearance of a metasilicate phase (pyroxene) and completereduction in the stabilization of an orthosilicate (olivine)liquidus phase. It is clear, therefore, that the relationshipbetween redox equilibria and melt structure is of majorimportance to the understanding of the liquidus equilibriaof magmatic systerns.

AcknowledgmentsCritical reviews by H. S. Yoder, Jr. and two anonymous re-

viewers are appreciated.

l0 n t F.2O!

ofi

c

tro42OE--SO2-F.-{

Al l lA+gl '0.20Tddynib

/ iFo..ruit -/ ,l

I L.''t ,/ /

330 MYSEN ET AL: FERRIC AND FERROUS IRON IN MELTS

References

Amthauer, G., Annersten, H., and Hafner, S. S. (1977) TheM<issbauer spectrum of sTFe in titanium-bearing andradites.Physics and Chemistry of Minerals, l,339413.

Amthauer, G., Langer K. and Schliestedt, M. (1980) Thermallyactivated electron delocalization in deerite. Journal of Physicsand Chemistry of Minerals, 6, 19-30.

Annersten, H. and Halenius, U. (1976) Ion distribution in pinkmuscovite: a discussion. American Mineralogist, 61, 1045-1050.

Annersten, H. and Olesch, M. (1978) Distribution of ferrous andferric iron in clintonite and the M<issbauer characteristics offerric iron in tetrahedral coordination. Canadian Mineralogist,16,199--204.

Annersten, H., Olesch, M. and Seifert, F. A. (1978) Ferric iron inorthopyroxene: a Miissbauer spectroscopic study. Lithos, 11,301-310.

Bell, P. M. and Mao, H. K. (1974) Crystal-field spectra of Fe2*and Fe3* in synthetic basalt glass as a function of oxygen fu-gacity. Carnegie Institution of Washington Year Book, 73,496-497.

Brawer, S. A. (1975) Theory of the vibrational spectra of somenetwork and molecular glasses. Physical Review B, ll, 3173-3194.

Brown, G. E'., Keefer, K. D., and Fenn, P. M. (1978) ExtendedX-ray fine structure (rnrs) study of iron-bearing silicate glass.(abstr.) Geological Society of America Abstracts with Programs,10, 373.

Calas, G. and Petiau, J. (1983) Structure of oxide glasses: spec-troscopic studies of local order and crystallochemistry: geo-chemical implications. Bulletin Mineralogique, 106, 33-55.

Calas, G., Levitz,P., Petiau, J., Bondot, P., and Loupias, G. (1980)Etude de I'order local autoudu fer dans les verres silicates natu-rels et synthetiques a I'aide de la spectrometrie d'absorption. X.Revue de Physique Appliquee, 15, 116l-1167.

Chou, I. M. (1978) Calibration of oxygen buffers at high P and Tusing the hydrogen fugacity sensor. American Mineralogist, 63,690-703.

Coey, J. M. D., Moukarik, A. and McDonagh, C. M. (1982) EIec-tron hopping in cronstedite. Solid State Communications, 41,797-800.

Danckwerth, P. A. and Virgo, D. (1982) Structural state of Fe inthe system NarO-SiOr-Fe-O. Carnegie Institution of Wash-ington Year Book, 81, 342-344.

Danckwerth, P. A., Virgo, D., and Mysen, B. O. (1982) The struc-tural state of Fe in alkaline vs. alkaline-earth silicate systems.Carnegie Institution of Washington Year Book, 81,34-347 .

Davidon, W. C. (1959) Variable metric method for minimization.Argonne National Laboratory, ANL 5990.

Deines, P., Nafziger, R. H., Ulmer, G. C., and Woermann, E.(1974) Temperature-oxygen fugacity tables for selected gas mix-tures in the system C-H-O at one atmosphere total pressure.Bulletin, Earth and Mineral Sciences, The Pennsylvania StateUniversity,88.

Dickenson, M. P. and Hess, P. C. (1981) Redox equilibria and thestructural role of iron in aluminosilicate melts. Contributions toMineralogy and Petrology, 78, 352-358.

Domine, F. and Piriou, B. (1983) Study of sodium silicate melt andglass by infrared reflectance spectroscopy. Journal of Non-Crystalline Solids, 55, 125-131.

Douglas, E. W., North, P. and Paul, A. (1965) Oxygen ion activityon redox equilibrium in glasses. Physics and Chemistry ofGlasses, 6,216-223.

Evans, B. J. and Amthauer, G. (1980) The electronic structure ofilvaite and the pressure and temperature dependence ofits s?Fe

Mcissbauer spectrum. Journal of Physics and Chemistry ofSolids,41,98!1001.

Fox, K. E., Furukawa, T., and White, W. B. (1982) Transitionmetal ions in silicate melts, Part 2, Iron in sodium silicateglasses. Physics and Chemistry of Glasses, 23, 169-178.

Furukawa, T., Fox, K. E., and White, W. B. (1981) Raman spec-troscopic investigation of the structure of silicate glasses. III.Raman intensities and structural units in sodium silicate glasses.Journal of Chemical Physics, 7 5, 3226-3237.

Gaskell, P. H. (1975) Construction of a model for ordered amor-phous tetrahedral materials using ordered units. PhilosophicalMagazine,32,2ll-229.

Goldman, D. S. (1983) Oxidation equilibrium of iron in borosili-cate glass. Journal of the American Ceramic Society, 66,205-209.

Hafner, S. S. and Huckenholz, H. G. (1971) Mdssbauer spectrumof synthetic ferridiopside. Nature, Physical Science, 233,255-261.

Haggerty, S. E. (1978) The redox state of planetary basalts. Geo-physical Research Letters, 5, 443-446.

Haggerty, S. E. and Thompkins, L. A. (1983) Redox state of theearth's upper mantle from kimberlitic ilmenites. Nature, 303,295-300.

Hart, S. E. and Davis, K. E. (1978) Nickel partitioning betweenolivine and silicate melt. Earth and Planetary Science Letters,40,203-220.

Hess, P. C. and Wood, M. L (1982) Aluminum coordination inmetaaluminous and peralkaline silicate melts. Contributions toMineralogy and Petrology, 81, 103-112.

Holmquist, S. (1966) Ionic formulation of redox equilibria in glassmelts. Journal ofthe Arnerican Ceramic Society, 49,228-229.

Huggins, F. E., Mao, H. K. and Virgo, D. (1975) Mossbauer stud-ies at high pressure. Carnegie Institution of Washington, YearBook 74,40H10.

Irving, A. J. and Frey, F. A. (1978) Distribution of trace elementsbetween garnet megacrysts and host volcanic liquids of kimber-litic to rhyolitic composition. Geochimica et CosmochimicaActa, 42,'l7l-789.

Kilinc, A., Carmichael, L S. E., Rivers, M. L., and Sack, R. O.(1983) The ferric-ferrous ratio of natural silicate liquids equili-brated in air. Contributions to Mineralogy and Petrology, 83,136-14r.

Kirkpatrick, R. J., Smith, K. A., Kinsey, R. A., and Oldfield, E.(1982) High-resolution 2eNMR of glasses and crystals in thesystem CaO-MgO-SiOr. (abstr.) EOS,63, 1140.

Larson, H. and Chipman, J. (1953) Oxygen activity in iron oxideslags, Transactions AIME, 196, 1089-1096.

Levitz, P., Bonnin, D., Calas, G., and Legrand, A. P. (1980) Atwo-parameter distribution analysis of Mcissbauer spectra innon-crystalline solids using general inversion method. Journalof Physics E, Scientific Instruments, 13,427437.

Liebau, F. (1981) The influence of cation properties on the confor-mation of silicate and phosphate anions. In M. O'Keeffe and A.Navrotsky, Eds., Structure and Bonding in Crystals, Vol. 2, p.197-232. Academic Press, New York.

Mao, H. K., Virgo, D., and Bell, P. M. (1973) Analytical study ofthe orange soil returned by the Apollo 17 astronauts. CarnegieInstitution of Washington Year Book, 72, 631-638.

Mo, X.-X., Carmichael, I. S. 8., Rivers, M., and Stebbins, J. (1982)The partial molar volurne of FerO. in multicomponent silicateliquids and the pressure dependence of oxygen fugacity inmagrnas. Mineralogical Magazine, 45, 237 -245.

Mysen, B. O. and Virgo, D. (1978) Influence of pressure, temper-ature and bulk composition on melt structures in the system

MYSEN ET AL.: FERRIC AND FERROUS IRON IN MELTS 331

NaAlSirOu-NaFe3+SirOu. American Journal of Science, 278,1307-1322.

Mysen, B. O. and Virgo, D. (1980) Trace element partitioning andmelt structure: an experimental study at 1 atrn pressure. Ge-ochimica et Cosmochimica Acta, 44,1917-1930.

Mysen, B. O. and Virgo, D. (1984a) Redox equilibria, structureand melt properties in the system NarO-AlrO.-SiOr-Fe-O.Carnegie Institution of Washington Year Book, 82,313-317.

Mysen, B. O. and Virgo, D. (1983b) Iron-bearing alkaline earthsilicate melts: relations between redox equilibria of iron, meltstructure and liquidus phase equilibria. Carnegie Institution ofWashington Year Book, 82, 317-321.

Mysen, B. O. and Virgo, D. (1983c) Effect of pressure on thestructure of iron-bearing silicate melts. Carnegie Institution ofWashington, Year Book; 82, 321-325.

Mysen, B. O., Seifert, F. A. and Virgo, D. (1980a) Structure andredox equilibria of iron-bearing silicate melts. American Miner-alogist,65, 867-884.

Mysen, B. O., Virgo, D., and Scarfe, C. M. (1980b) Relations be-tween the anionic structure and viscosity of silicate melts-aRaman spectroscopic study. American Mineralogist, 65, 690-7tt.

Mysen, B. O., Virgo, D., and Kushiro, I. (1981) The structural roleof aluminum in silicate melts-a Raman spectroscopic study at1 atm. American Mineralogist, 66,678-701.

Mysen, B. O., Virgo, D., and Seifert, F. A. (1982a) The structure ofsilicate melts: implications for chemical and physical propertiesof natural magma. Reviews of Geophysics and Space Physics,20. 353-383.

Mysen, B. O., Virgo, D., and Seifert, F. A. (1982b) Distribution ofaluminum between anionic units in depolymerized silicate meltsas a function of pressure and ternperature. Carnegie Institutionof Washington Year Book, 81, 360-366.

Mysen, B. O., Virgo, D., and Seifert, F. A. (1984) Redox equilibriaof iron in alkaline earth silicate melts: relationships betweenmelt structure, oxygen fugacity, temperature and properties ofiron-bearing silicate liquids. American Mineralogist, 69, 83+847.

Mysen, B. O., Virgo, D. and Seifert, F. A. (1985) Relationshipsbetween properties and structure of aluminosilicate melts.American Mineralogist, 70, 88-105.

Nolet, D. A. and Burns, R. G. (1979) Ilvaite: a study of temper-ature dependent delocalization by the Miissbauer effect. Physicsand Chemistry of Minerals, 4,221-234.

Nolet, D. A., Burns, R. G., Flarnm, S. L., and Besancon, J. R.(1979) Spectra of Fe-Ti silicate glasses: implications to remote-sensing of planetary surfaces. Proceedings, 10th Lunar ScienceConference, l7'l 5-17 86.

O'Horo, M. P. and Levy, R. A. (1978) Effect of melt atmosphereon the magnetic properties of a (SiOr)or(CaO)rrl.r[FerO.].,glass. Journal of Applied Physics, 49, 1635-163?.

Osborn, E. F. and Muan, A. (1960a) Phase equilibriurn diagramsof oxide systems. Plate 2, The system CaO-Al2O3-SiOr. Ameri-can Ceramic Society, Columbus, Ohio.

Osborn, E. F. and Muan, A. (1960b) Phase equilibrium diagramsof oxide systems. Plate 10. The system CaO-"FerO."-SiOr.American Ceramic Society, Columbus, Ohio.

Osborn, E. F. and Muan, A. (1960c) Phase equilibrium diagramsof oxide systems. Plate 7. The system CaO-"FeO"-SiOr. Ameri-can Ceramic Society, Columbus, Ohio.

Osborn, E. F. and Muan, A. (1960d) Phase diagrams of oxidesystems. Plate 3. The system MgO-Al2O3-SiO2. American Cer-amic Society, Columbus, Ohio.

Osborn, E. F. and Muan, A. (1960e) Phase diagrams of oxide

systems. Plate 4. The system NarO-AlrO.-SiOr. AmericanCeramic Society, Columbus, Ohio.

Paul, A. and Douglas, R. W. (1965) Ferrous-ferric equilibrium inbinary alkali silicate glasses. Physics and Chemistry of Glasses,6,207-21t.

Ryerson, F. J. and Hess, P. C. (1980) The role of PrOr in silicatemelts. Geochimica et Cosmochimica Acta, 4,,61L-425.

Sack, R. O., Carniichael, L S. E., Rivers, M., and Ghiorso, M. S.(1980) Ferrous-ferric equilibria in natural silicate liquids at Ibar. Contributions to Mineralogy and Petrology, 1 5, 369-377.

Sato, M. (1972) Electrochemical measurements and control ofoxygen fugacities with solid electrolyte systems. In G. C. Ulmer,Ed., Research Techniques for High Pressure and High Temper-ature, p. 43-99. Springer Verlag, New York.

Sato, M. and Yalenza, M. (1980) Oxygen fugacities of the layeredseries of the Skaergaard intrusion, East Greenland. AmericanJournal of Science, 280A, 13+158.

Seifert, F. A., Virgo, D., and Mysen, B. O. (1979) Melt structureand redox equilibria in the system NarO-FeO-Fe2O3-Al2O3-SiOr. Carnegie Institution of Washington Year Book, 78, 511519.

Seifert, F. A., Mysen, B. O., and Virgo, D. (1982) Three-dimensional network melt structure in the systemsSiOr-NaAlOr, SiOr{aAlrOn and SiOr-MgAlrOo. AmericanM ineralogist. 67 . 696-7 18.

Taylor, M. and Brown, G. E. (1979a) Structure of mineral glasses.I. The feldspar glasses NaAlSi.OB, KAlSi3O8, CaAlrSirOr. Ge-ochimica et Cosrnochimica Acta, 43, 6l-77.

Taylor, M. and Brown, G. E. (1979b) Structure of mineral glasses.II. The SiOr-NaAlSiOo join. Geochimica et CosmochimicaActa, 41, 1467-1475.

Thornber, C. R., Roeder, P. L., and Foster, J. R. (1980) The effectof composition of the ferric-ferrous ratio in basaltic liquids atatmospheric pressure. Geochimica et Cosmochimica Act4 44,s2s-533.

Virgo, D. and Mysen, B. O. (1985) The structural state of iron in

oxidized vs. reduced glasses at I atm: a s?Fe Mcissbauer study.

Physics and Chemistry of Minerals, in press.

Virgo, D., Mysen, B. O., and Seifert, F. A. (1981) Relationship

between oxidation state of iron and structure of silicate melts.

Carnegie Institution of Washington Year Book, 80, 308-31 1.

Virgo, D., Mysen, B. O., and Danckwerth, P. A. (1982) Speciation

of Fe3 t in l-atm NarO-SiOr-Fe O melts. Carnegie Institution

of Washington Year Book, 81, 349-353.Virgo, D., Mysen, B. O., and Danckwerth, P. A. (1983) Redox

equilibria and the anionic structure of NarO ' xSiOr-Fe-O

melts: effect of oxygen fugacity. Carnegie Institution of Wash-

ington Year Book,82, 305-309.Watson, E. B. (1976) TwoJiquid partition coelficients: experi-

mental data and geochmemical implications. Contributions to

Mineralogy and Petrology, 56, ll9-134.Watson, E. B. (19'77) Partitioning of manganese between forsterite

and silicate tiquid. Geochimica et Cosmochimica Lcta, 41,

136Tt374.Waychunas, G. A. and Rossman, G. R. (1983) Spectroscopic stan-

dard for tetrahedrally coordinated ferdc iron:yLiAlOr:Fe3+.

Physics and Chemistry of Minerals, 9,212'215.

Wivel, C. and Morup, S. (1981) Improved computational pro-

cedure for evaluation of overlapping hyperfine parameter distri-

butions in M<issbauer spectra. Journal of Physics E, Scientific

Instruments, 14, 605-610.

Manuscript receitsed, February 7, 1984;

accepteilfor publication, Notember 19, 1984.