I Present address: Mineralogisch-Petrographisches Institutund Museum, Universitat Kiel, 2300-Kiel, West Germany.

0003-mxx84/09 I 0-0834$02. 00

American Mineralogist, Volume 69, pages 834-847, lg$4

Experimental methods

Starting materials were spectroscopically pure (Johnson andMatthey) SiO2, CaCO3, BaCOg, SrCO3, MgO and Fe2O3. Ap-proximately l0Vo of theFe2O3 was added as isotopically enriched

E34

Redox e_quilibria of iron in alkaline earth silicate melts: relationshipsbetween melt structure, oxygen fugacity, temperature

and properties of iron-bearing silicate liquids

BrlnN O. Mvsnu, Devro Vrnco AND FRTEDRTcH A. SBrrenrl

Geophysical Laboratory, Carnegie Institution of WashingtonWashington, D. C. 20008

Abstract

M<issbauer and Raman spectroscopy have been used to relate redox equilibria of iron tomelt structure in MO-SiO2 systems (M : Ba2+, Sr2*, Ca2*, Mgr*). Log Fe2*/Fe3* is alinear function of log/e,, although the slope of the lines depends on bulk composition. TheFe2*/Fe3* is a linear function of NBO/T (nonbridging oxygens per tetrahedral cation) foreach type of divalent metal cation. The ferrous-ferric ratio is also proportionalto Zlf of themetal cation at constant M/Si (or NBO/Si) of the melts.

With Fe2+/Fe'1 4 1, ferric iron is in tetrahedral coordination. The spectroscopic dataindicate that further reduction may result in a coordination transformation of the remainingFe3+, probably to six-fold coord-ination. The Fe2+/Fe3+ of this transformation appearsinsensitive to the type of metal cation and the NBO/T of the melt. Its value increasessomewhat as the total iron content of the system is increased.

Liquidus phase equilibria for compositions with Calsi : 0.7 and Me/Si : 0.7'with l0 wt.Vo iron oxide added were calculated as a function of Fe2*/Fe3*. An increase inFe2*/Fe3* results in a decrease of 35'C in the liquidus temperature in the system CaO-SiO2-Fe2O3-FeO (at Fe2*/Fe3* - 0.25). In the analogous Mg-bearing system, themaximum temperature decrease is 200"C (at Fe2+/Fe3+ - 4.0). Silica polymorphs (and,over a limited ferric/ferrous range, two liquids in the Mg-bearing system) occur on theliquidus until the temperature minimum is reached. With additional increase in Fe2+/Fe3+,metasilicate minerals (pseudowollastonite and clinoenstatite, respectively) are stable. Inthe system MgO-SiO2-Fe2O3-FeO, a silica polymorph (cristobalite) reappears on theliquidus with Fe2+/Fe3* - 9.0, whereas in the analogous Ca system, pseudowollastoniteremains on the liquidus until all ferric iron has been reduced to ferrous iron.

Introduction type ofmetal cations available for charge-balance ofFe3+Iron oxide redox equilibria in magmatic liquids are of in tetrahedral coordination and the proportions of coex-

interest because p"z+1p"r+ influences melt structure and isting anionic units (Dickinson and Hess, l98l; Mysen etderivative properties (Mysen et al., l9g2b) and because al".l982b). These proportions are related to the network-the distribution of iron oxides between liquidus minerals modifying cations in the melts (Mysen et al., 1982b). Inand melt may be used as an indicator of magmatic history natural magmatic liquids, alkaline earths commonly are(e.g., Bowen et a1., 1930; Mysen et al., l9g0; Hageerty, the most abundant network-modifying cations and may1978; Sato and Valenza, l9g0). preliminurv outu iiii.ui" also potentially serve to charge-balance Fe3* in tetrahe-that ferric iron may be both a network former and a dral coordination in analogy with observations on alka-network modifier (6.9., Mysen et al., l9g0), whereas l ine-earth-bearingaluminosil icatesystems(e.g.,Seifertet

ferrous iron commonly is considered to be a network al., 1982:Navrotskyetal., 1982). Inordertodescribethemodifier in silicate melis (e.g., Mao et al., 1973; Nolet et interrelations between redox equilibria of iron oxides anda1., lg79). Thus melt properties thar depend on melt the structure of magmatic liquids, we decided, therefore,polymerization may also depend on iron content and to study the efect oftemperature and oxygen fugacity onFe3*DFe. the redox equilibria of iron oxide in simple alkaline earth

The redox ratio of iron oxides and the structural silicate melts.position of Fe3+ and Fe2* in melts may depend on the

57Fe2O3 e0-95% 57Fe1 to enhance the sensitivity of M6ssbauerspectra for the determination of Fe2*/Fe3*2 and the structuralpositions of Fe2* and Fe3*. Batches of 200-300 mg were groundunder alcohol for I hr or more, dried at 400'C and stored at 1 l0"Cuntil used. The samples (30-70 mg) were suspended on 0.1-mmPt loops (Presnall and Brenner, 1974) it vertical Pt- and MoSi2-heated, quench furnaces. All samples were quenched in eitherliquid nitrogen or water. The oxygen fugacity in the furnace wascontrolled with CO-CO2 gas mixing and monitored with anY2O3-doped ZrO2 oxygen sensor (Sato, 1972). The accuracy ofthe sensor was checked periodically against the equilibria 2Ni +O, = 2NiO and 2Fe + 02 = 2FeO (Chou, 1978; Deines et al.,1974). The accuracy of the oxygen fugacity, based on thiscalibration, is better than 0. I log unit offllrat the temperature ofthese experiments (1550'C). The precision is 0.01-0.02 log unit.The composition of the glasses, within analytical uncertainty(2%,relative), accords with the nominal values. Electron micro-probe analyses are available from the authors upon request. Thecompositions of the starting materials, as used throughout thisreport, are defined with their symbols as described in Table l.

Raman and Mrissbauer soectra were obtained with the auto-

2 It is recalled that Fe3+/)Fe, obtained directly from theMossbauer spectra, is related to Fe2*/Fe3* as Fe2*/Fe3* : (l -

Fe3*DFe)/(Fe3*DFe).

E35

mated Raman and Mtissbauer systems at the Geophysical Labo-

ratory (e.g., Mysen et al., 1982a)' The Raman data were obtained

by excitation of the sample with the 514-nm line of an Ar*-ion

laser with 2-4 watts at the sample with the same configurations

as in Seifert et al. (1982). The samples were examined optically

before and after exposure to the laser to evaluate possible

radiation damage. On occasion, discoloration and sometimesprecipitation of an iron oxide were observed. In such cases, the

Raman spectrum was discarded. Mtissbauer spectra were ob-

tained on the same samples, occasionally both before and after

exposure to the laser beam. In optically clear glasses, no effects

of laser exposure could be observed in the Mossbauer spectra'

All other aspects of the Raman methods have been described

elsewhere (Mysen et al., 1982a Virgo et al.' 1981, 1982; Seifert

et al.. 1982). Curves of Lorentzian and Gaussian line shapes

were fitted to the digitized output from the Mtissbauer and

Raman spectrometers, respectively, as summarized elsewhere(Mysen and Virgo, 1978; Virgo et al., 1981, 1982, 1983a; Mysen

et al. , 1980, 1982a).Alternative fitting methods of the Mossbauer spectra with

different lines-shapes (e.g., Eibschutz, et al., 1980; Eibschutz

and Lines, 1982), or with least-squares fitting of an assumed

correlation of hyperfine parameters of a large number of elemen-

tary doublets (e.g., Wivel and Morup, l98l) were attempted'

These methods did not yield results for Fe3*/)Fe or hyperfine

MYSEN ET AL,: IRON IN ALKALINE EARTH SILICATE MELTS

Table l . Mossbauer results

Conposltion - los f ^ F e - ( I ) Fe- ( I I )

IS QS IS QS

cs2E2.1cs2E4.6cs2F9. 1cs2F5cs2F5cs1. 4F5csl . 4F5cs1 .4F5cs l . 4F5csl . 2F5cs1 . 2F5csF5csF5csF5cs .8F5cs . 8F5cs. 7F5cs1.4Fl_0cs1. 4F10cs1. 4Fl0

MS1 .4F5MS1. 4F5MSl. 4F5MSl. 4F5MS l .2F5MS1 .2F5

SrSF5

BS4F5BS3F5BS2F5BS1. 5F5BS1. 2F5BS1. 1F5BSF5

L7 25t725L725I ) ) U

155015501550155015501550155015501 q q n

1550155015501550155015501550

1550155015501 550I ) ) U

1550

1550

1550155015501550155015501550

0 . 6 80 . 6 80 . 6 80 . 6 82 . 5 00 . 6 8

4 . 0 06 . 0 00 . 6 82 . 5 00 . 6 82 . 5 04 . 0 00 . 6 82 . 5 00 . 6 80 . 6 82 . 5 04 . 0 0

0 . 6 82 . 5 04 . 0 06 , 0 00 . 6 82 . 5 0

0 . 6 8

0 . 6 80 . 6 80 . 6 80 . 6 80 . 6 80 , 6 80 . 6 8

0 . 7 0 0 . 7 10 . 5 6 0 . 7 80 . 3 1 r . L 7o . 3 2 L , L 10 . 5 3 0 . 8 50 . 3 2 1 . 1 30 . 5 7 0 . 7 70 . 5 7 0 . 5 5

o . 3 2 1 . 1 4Q . 5 2 0 . 7 90 . 3 1 1 . 1 50 . 4 5 0 . 9 10 . 5 5 0 . 6 90 . 3 2 1 . 1 90 . 3 4 1 . 1 6o . 3 2 1 . 2 0o . 3 2 1 . 1 60 . 3 3 r . l 20 . 5 5 0 . 7 1

0 . 3 4 1 , 2 40 . 7 0 0 . 7 10 . 8 5 r . 0 2

0 . 3 7 1 . 3 30 . 7 0 0 . 9 0

0 . 3 0 1 . 0 9

0 . 3 0 0 . 9 70 . 3 0 0 . 9 80 . 2 9 0 . 9 1o . 2 9 0 . 9 6o . 2 7 0 . 9 50 . 2 8 0 . 9 7o , 2 9 0 . 9 0

1 . 0 5 2 . L 91 . 0 1 2 . 2 61 . 1 0 2 . o 21 . 0 8 2 . O 21 . 0 7 2 . L 61 . 1 1 2 . l OL . 0 7 2 . 1 81 .05 2 .O31 .04 2 .21 ,1 . 0 6 2 . 0 s0 . 9 6 2 . 2 71 1 1 ? 1 5

1 . 0 5 2 . 2 01 . 0 0 2 . 2 91 , 0 0 t , 9 91 . 0 8 2 . 1 21 . 0 0 2 . o 81 . 0 0 2 . o o1 . 1 1 2 . 0 51 . 1 1 2 , 0 6

1 . 1 5 2 . L 31 1 ' 7 t 1

1 . 1 - 3 2 . L 41 . 0 5 2 . ! 81 . t 6 2 . r O1 . 1 1 2 . L 9

1 . 0 0 L . 8 2

0 . 9 4 2 . 0 00 . 9 5 1 . 9 60 . 9 7 L . 9 70 . 9 5 1 . 9 30 . 9 5 1 . 9 r0 . 9 3 1 . 9 30 , 7 5 2 . L 8

0 . 8 5 l - . 8 70 . 8 5 1 . 9 3o . 9 2 1 . 6 s0 . 9 1 L . 6 70 . 9 3 1 . 8 80 . 9 5 1 . 7 8o . 9 2 1 . 8 90 . 8 6 L . 6 51 , 0 1 1 . 6 50 . 9 1 L . 1 60 . 7 5 1 . 6 80 . 9 6 1 . 8 50 . 8 9 1 . 8 8o . 9 7 l . 7 r

0 . 9 1 1 . 7 1

: : : : : :0 . 9 s L . 7 50 . 9 5 1 . 9 1

o .99 1 .81 -0 . 9 5 1 , 8 60 . 9 6 1 . 8 0o . 9 2 L . 5 21 . 0 0 r . 7 70 . 9 5 r . 8 9

0 .2490 . 3960 .4430 . 584o . 2 4 70 . 6140 . 2 8 80 . 1 1 70 .0000 . 6 3 80 . 3560 . 6 7 50 .4L70 .2280 . 7100 . 4 7 20 . 8 2 3o , 6 7 70 .3970 . 2 3 1

0 .5230 . 1880 . 0650 .0000 . 4 8 20 . 2 2 2

o . 7 4 6

0 . 6840 . 6 8 40 . 6 9 90 . 7314 . 7 7 L0 . 7 8 40 .833

*Isomer shlft (IS) and quadrupole splitt lng (QS) (m/sec) relative to Fe oetal'

Abbrev ia t lons : c , Cao; l . { , MgO; Sr ' s ro i B , Bao; S ' S iO2; F ' fe2o3. CS2 ' e tc . ,

CS2F5, e tc . , re fe r to v t .7 . Ee2O3 added to € ta r t lng mter ia l .refeE to nolar Sio2/netal oxlde.

E36 MYSEN ET AL.: IRON IN ALKALINE EARTH SILICATE MELTS

MS1.4F5 CSt.4Fs

Vd@ity, [n/s

Fig. l. Comparison of Mdssbauer spectra of melts (obtained at 298 K) with composition CSl.4F5 (SiO2/CaO : l.4,with5 wt.%FezOr added) and MSl.4F5 (SiOr/MgO : 1.4, with 5 wt.Vo Fe2Os added) at 1550"C as a funcrion of oxygen fugacity.

6A

tig

tA

I

I

6

B

II

tc

6

t

.l

I5IE

T

5IaE

I

o

B

=It8

2

o

-2

lF.8+/:F.. O tl

-24 O,0

parameters that were significantly ditrerent from those reportedhere. Thus, the fitting methods employed in earlier reports (e.g.,Virgo et al., 1981, 1982, 1983a) were employed. In these fits, theareas ofthe individual doublets were constrained to be equal, aswell as the line-width of the ferric doublets. The latter constraintwas not imposed on the Fe2*-doublet(s) as Mtissbauer spectrafor highly reduced glasses typical ly indicate that Fe2'-doubletsare asymmetric (e.g., Mao et al., 1973; Mysen and Virgo, 1978;Nolet et al.. 1979 Levitz et al.. 1980). In the most oxidizedsamples, the spectra can be resolved into one ferric doublet andtwo high-velocity components of two Fe2* doublets. The twocorresponding low-velocity components nearly coincide andcould not be resolved statistically. With the resultant fiveline fit(without the area constraints for the Fe2* components), the areaof the single Fe2*-low velocity component is similar to the sumof the two high-velocity components within 5%o relative. Inreduced samples two asymmetric ferrous doublets result in thestatistically best fit (e.g., Fig. l). These two Fe2*-doublets aredenoted Fe2*(I) and Fe2*(II) in the text. Interested readers maycontact the authors for additional details and program listings forthe alternative and current fittine routines.

Results

Mossbauer data

Relationships between Mossbauer data and bulk meltcomposition, temperature and oxygen fugacity are sum-marized in Table I with selected Mossbauer spectra inFigure 1. Decreasing/62 results in a rapid decrease in theintensity of the resonant absorption between 0.5 and 0.8mrn/sec in the M<issbauer spectra and a slight shift tohigher velocity of the high-velocity component of theferric doublet (Fig. l). This spectroscopic change isassociated with an increase in absorption in the highest-and lowest-velocity portions of the spectra. These spec-troscopic features result from the increase in Fe2*/Fe3*of the melts and a slight increase in the velocity of thelow-velocity component of the Fe3* doublet.

At the same oxygen fugacity, temperaturg and degreeof polymerization of the melt, the Fe2+/Fe3+ increases inthe order Ba2* I Sr2+ < Caz* < tlgz* (zlP increases)(Fie. 2), as also observed by Douglas et al. (1965), Lauerand Morris (1977) and Mysen et al. (1980). The Fe2*/Fe3*is linearly correlated with the degree of polymerization ofthe melt (NBO/T) at the same temperature and /o. (Fig.3). The dependence of Fe2*/Fe3* on NBO/T3 (or NBO/Si)

3 The NBO/T may be calculated from bulk chemical composi-tion after recasting to chemical formula and assignment oftetrahedrally coordinated cations (from structure determina-t ions). Two equivalent expressions,

NBO/T: l /T ) l /nMi .i : l

or NBO/T = (2O - 4T)/T, can be used. In these expressions n isthe electrical charge of network-modifying cations M1, T is thetotal number of tetrahedrally coordinated cations and O is thenumber of oxygens. Metal cations required for charge-balance oftetrahedral cations such as Al3* and Fe3* are subtracted from

837

Fe3*/ EFe

1-0 0.9 0.8 0.7 0.6 0.5

mg2'

Fo2+ r6s*=o 44 (L)-ou, at= o ss

ca2'

i r2*

8.2*

0.4 0.6

Fe2*/Fe3+

Fig. 2. Relationship between Fe2*/Fe3* and type of metalcation (Ba2+, St'*, Ca2*, Mg2*), expressed as Zlr2 (Z :

electrical charge, r = ionic radius for six-fold coordination;Whittaker and Muntus, 1970) for metasilicate melt compositionsequilibrated at 1550"C in air.

is also a function of the type of network-modifying cation,the oxygen fugacity and the anionic structure of the melt(Table 2; Figs. I and 3). The slopes ofthe lines increase asthe metal cation is changed from Ca2* to Ba2+ , and forthe same metal cation, with decreasing/o, (Fig. 3).

As is commonly observed in silicate systems (e'g.,

Fudali, 1965; Lauer and Morris, 1977; Mysen et al., 1980;Goldman, 1983), there is a nearly linear relationshipbetween log/s" and log (Fe2*/Fe3*) (Fig. 4). The slope ofthe lines decreases, however, from about -0.4 for thecalcium disilicate composition to about -0.25 as theNBO/T of the melt increases toward that of calciummetasilicate (Fig. a). The slope is also greater for Mg-bearing systems than for Ca-bearing systems and maydecrease (in an absolute sense) with increasing ironcontent ofthe system (Fig. a). This behavior differs fromthat of Fe2*/Fe3* in alkali silicate systems, where theslope does not deviate significantly from -0'25 (Mysen etal., 1980; Virgo et al., 1983a). Thus, the activity coeffi-cient ratio, yFe3* lyFez* , decreases with increasing NBO/T and with decreasing Zlf of metal cations in alkalineearth silicate melts, whereas this ratio is independent ofNBO/T in alkali silicate systems.

Regardless of the degree of polymerization of thesilicate solvent (expressed as NBO/Si), the ISo"r+ remainsat about 0.3 mm/sec until Fe3+DFe decreases below 0.5(Fie. 5). With Fe3*DFe between 0.5 and 0.4, the ISo":+and QS"-r+ change rapidly as a function of decreasing

their total abundance to obtain M1. In this paper, the NBO/T and

NBO/Si are used as expressions indicating structural and compo-sitional features of the melts in the sense that, if the tetrahedrallycoordinated cations are known, the value of NBO/T can becalculated from bulk composition.

MYSEN ET AL.: IRON IN ALKALINE EARTH SILICATE MELTS

2.6

2.2N

1 . 0

E3E MYSEN ET AL.: IRON IN ALKALINE EARTH SILICATE MELTS

Table 2. Average values of hyperfine parameters* (relative to Fe metal)

F"3* F e - ' ( I )Systm

Qs

0 .318 r 0 .004

0 .57 ! 0 . 06

0 . 3 5 f 0 . 0 2

0 . 7 0 t 0 . 0 0

0 . 30

0 .29 t 0 . 01 -

1 . 1 6 I 0 , 0 3

0 . 7 3 r 0 . 0 9

L . 3 2 L O . 0 7

0 . 8 4 ! 0 . 0 9

1 . 09

0 . 9 6 I 0 . 0 3

1 . 0 6 ! 0 . 0 5

1 . 0 4 r 0 . 0 5

1 . 1 9 ! 0 . 0 6

1 . 1 2 ! 0 . 0 1

1 .00

0 . 9 2 ! 0 . 0 8

2 . 0 6 r 0 . 0 6

0 . 9 0 I 0 . 1 1

2 . 1 8 r 0 . 1 2

2 . 2 0 ! O . O L

L . 8 2

1 . 9 8 t 0 . 0 9

0 . 9 3 I 0 . 0 2 1 . 7 4 t 0 . 0 8

1 . 8 2 I 0 . 1 1

1 . 0 1 i 0 . 0 2 1 . 8 2 i 0 . 0 6

0 , 9 4 r 0 . 0 2 1 . 8 8 i 0 . 0 2

*Isomer shift (IS) and quadrupole splirt lng (QS) in m/sec.f(IV) and (VI) refer to lnferred coordLnarlon of Fe$.

redox ratio (oxygen fugacity). This ferric/ferrous range isinsensitive to the type of metal cation (Fig. 5; see alsoVirgo et al., 1983a, for data on alkali and alkaline earthaluminosilicate systems). It is noted, however, that inCSl.4 melt (NBO/Si of calcium silicate solvent equals1.4) the change in IS,.l+ and QS"-31 occurs at somewhatlower values of Fe3+l)Fe for l0' wt.Vo than for 5 wt.VoFe2O3 added (cf. CSl.4F5 and CSl.4Fl0, Fig. 5).

Raman data

Raman spectra were obtained on three compositionswith NBO/Si of the iron-free end-member compositionequal to I (CS2), 1.4 (CSl.4) and 2.38 (CS.8). The spectraof the end members and other relevant compositions havebeen reported and interpreted elsewhere (Verweij, 1979a,b; Verweij and Konijnendlk, 1976; Brawer and White,1977; Furukawa et al., l98l; Virgo et al., 1980; Mysen etal., 1982b; Domine and Periou, 1983) and only oneexample is included here for reference (Fig. 6).

The spectra are interpreted on the basis (see Mysen etal., 1982b; Domine and Piriou, 1983, for reviews; andFurukawa et al., 1981, for theoretical calcula[ions) thatthe anionic structure of melts on binary metal oxidesilicate joins may be described as combinations of struc-tural units with NBO/Si values corresponding to thestoichiometric expressions, SiOX-, Si2O9-, SiO3-, SirOS-and SiO2. For the compositions in the present study, theassignments of the most important Raman bands associat-ed with the various structural units are summarized inTable 3.

For quenched melts equilibrated in air, a distinct difer-ence between the spectra of iron-bearing and iron-freesamples is the depolarized band near 990 cm-l in thespectrum of CS2F5. For this melt, as well as those ofCS1.4F5 and CS.8F5, there is also an increase in Ramanscattering intensity near 900 cm-r lTable 3). Further-more, the relative intensity of the band near llfi) cm-r inCS2F5 and CSI.4F5 melts (assigned to Si-O- stretchvibrations in SizO3- units) is lowered as Fe2O3 is added tothe melt and equilibrated with air at 1550"C. There may bea slight relative increase in intensity near I150 cm-r (Si-Oo stretching in three-dimensional network units) and a

more pronounced relative increase in the 950-cm-' band(Si-O- stretching in SiO3- units; Furukawa et al., lgEl).The addition of l0 wt.Vo Fe2O3 to CSl.4 melts results infurther evolution of these trends (Table 3). AIso, theasymmetry of the band near 600 cm-r (mixed stretch andbending mode of Si-O-Si bridging bonds in depolymer-ized structural units) becomes very pronounced. At oxy-gen fugacity below about l0-2 5 (and decreasing Fe3"/)Fe), melts of composition CSl.4F5 (Fie. 6) have Ramanspectra with only those bands observed in the iron-freesystem. The intensities of the bands (Table 3), in particu-lar the ratio between the l100- and 950-cm-r bands, havechanged, however, relative to those of iron-free CaO-SiO2 melts with similar NBO/Si (Fig. 6; see also Fig. 3 inMysen et al., 1982b).

Structural interpretation

The 900- and 990-cm-l bands in the Raman spectra aregenerally assigned to stretch vibration of T-O' bonds (T

H

EEo.7

0.6

0.5

0.4

0.3

o.2

0.1

1550'C

a ELO . SlO2

o C.O - SiO2

NBOIT

Fig. 3. Comparison of relations between Fe2*/Fe3* andNBO/T for melts on the joins CaO-SiOz and BaO-SiO2 with 5wt.Vo Fe2Ot added and equilibrated at 1550'C in air and for CaO-SiO2 also atfoz = l0-2 5. In these compositions, the NBO/T wascalculated with Fe3* as a network former, as discussed further intext.

MYSEN ET AL.: IRON IN ALKALINE EARTH SILICATE MELTS 839

log foz

Fig. 4. Log (Fe2+/Fe3+) vs. log/62 for melrs equilibrated at1550'C as a function of melt composition, type of metal cationand amount ofiron added. For the expression, log (Fe2*/Fe3+; =alogft12 + b,the coefficients a and b are, respectively, CS2F5,-0.35 and -0.3E; CSl.4F5, -0.33 and -0.42: CSl.4Fl0, -0.26and -0.48; MSl.4F5. -0.36 and -0.2E: CSl.2F5. -0.28 and-0.43: CSFS. -0.26 and -0.49: CS.8F5. -0.24 and -0.55.

= tetrahedral cation) where T is probably predominantlyFe3* (Fox et al., 1982). In some of the glasses where theMrissbauer spectra suggest that tetrahedrally-coordinatedferric iron is absent, there remains a 900 cm-t band (seeTable 3). Most likely, this band results from the existenceof Si2O?- units in the melts. Whether such units also existin Fe3*(Iv)-bearing samples cannot be ascertained. Thisinterpretation is consistent with that of Virgo et al. (1981,1983b) for spectra of quenched melts (glasses) in thesystem Na2G-SiO1Fe2O3-FeO where Fe3* is in tetrahe-dral coordination. The values of ISr"3* and QS.":+ (Fig. 5,Table l) from the Mrissbauer spectra of these samples aresimilar to those found for tetrahedrally coordinated Fe3+in crystalline ferrisilicates (Annersten and Halenius,1976; Waychunas and Rossman, 19E3; Amthauer et al.,1977). Thus, both the Mossbauer and the Raman datafrom the melts equilibrated in air at 1550"C are consistentwith tetrahedrally coordinated ferric iron. The small butsignificant differences in IS..r+ and QS."r+ for tetrahe-drally coordinated ferric iron with different cations forelectric charge-balance (Table 2) may result from slightlydifferent Fe3*-O distances in the tetrahedra and possiblysome changes in polyhedral distortion.

The hyperfine parameters for ferric iron of quenchedmelts with Fe3*DFe lebs than 0.5 (corresponding to logfo, : -r.t at 1550"C) change rapidly from ISo.:+ - 0.3mm/sec and QS..r+ - 1.1 mm/sec at Fe3+DF6'> O.S toISr"l+ = 0.5 mm/sec and QSr"r+ = 0.8 mm/sec at FE3+/

)Fe - 0.4 or less (Fig. 5; Table 1). These changes inISr.r+ with decreasing Fe3*DFe, as well as the tempera-ture dependence of the hyperfine parameters for ferriciron obtained from spectra taken at 298 and 77 K with therespective low and high values of Fe3*/)Fe (see Virgo etal., 1982, 1 983a), are consistent with octahedrally coordi-nated ferric iron with Fe3*DFe < 0.4 and with tetrahe-drally coordinated ferric iron with Fe3*DFe > 0.5. In theFe3*DFe range between 0.5 and 0.4, the Mdssbauer dataare consistent with a transition from tetrahedral to octa-hedral ferric iron where the two types of Fe3*-O polyhe-dra may coexist (see also Virgo et al., 1983a, for furtherdiscussion of the Mcissbauer spectra of silicate melts withcoexisting tetrahedral and octahedral ferric iron). Addi-tional support for this interpretation of the ivlcissbauerdata is the observation that the 900- and 990-cm-' depo-laized Fe-Oo symmetric stretch bands (generally as-signed to Fe3+-O" in tetrahedral complexes; see also Foxet al., 1982) diminish in intensity with decreasing/6, (anddecreasing Fe3*DFe; Table 3, Fig. 6; see also footnote inTable 3 relating to ambiguities in this interpretation), thusalso indicating the disappearance of Fe3*(IVFO bonds inthe melts as the Fe3+DFe decreases.

The two ferrous doublets (e.g., Fie. I ; Table 1) differ in

QS."z* by more than lUVo,with insignificant differences inIS,.z+. Such differences in quadrupole splitting for thetwo Fe2+-O octahedra may be related to their degree ofdistortion (Annersten and Olesch, 1978; Mysen and Vir-go, 1978; Calas and Petiau, 1983; Waychunas and Ross-man, 1983). With this interpretation, the Fe2+(ID-O poly-hedron is the most distorted of the two.

Interaction between anionic units and ferric and

ferrous iron

Some of the structural changes from Fe2+/Fe3* varia-tions and coordination transformation of Fer* in themelts may be evaluated with the Raman spectroscopicdata (Fie. 6; Table 3). In the spectra, the relative abun-dance of anionic units is reflected in the relative intensi-ties of the Raman bands that result from T-O bonds (T =

tetrahedrally coordinated cation) in these units. Seifert etal. (1981) (see also Mysen et al., 1982a) devised a methodby which the relative abundance of anionic units inindividual MO-SiO2 or M2G-SiO2 (M : a specific univa-lent or divalent metal cation) may be calculated as afunction of IvI/Si of each system. This same method isused here for the system CaO-SiOrFe-O. Results ofthese calculations are shown for composition CS1.4F5 asafunction of its Fe3+DFe (Fig. 7). In Figure 7 the relativeabundance oftetrahedral Fe3* units is omitted because oftheir low abundance. For the melt equilibrated in air at1550'C, approximately 1.0% of the tetrahedral structureconsists of Fe3*1IV;-O polyhedra. For lower values ofFe3*DFe, Fe3+ is a network modifier, and the anionicnetwork can be described entirelv in terms of silicateunits.

^ 0.8

r

*, O.Or

- 0.4

840 MYSEN ET AL.: /RON 1N ALKALINE EARTH SILICATE MELTS

sl

q

o

g

E 0.6

; 0.s

rssooc. s w* Fe-o- o cs2a c s l 4r c s r 2o c sa c s 8

0.8 ' . -o o 'o '

v MSI 4F5

a cs l 4F5x csl 4F10

0 4 0 . 6

Fe3+ | LFe

Fig' 5. Variations in hyperfine parameters for ferric iron (isomer shifts relative to Fe metal) (ISp.r* and QS"".*) as a function ofFe3*/)Fe along the join CaO-SiO2, as a function of type of metal cation (Ca2* and Mg2*) and iron content.

0.4 06

Fe3+/ EFe

There are several features of the data in Figure 7indicating that Fe3+ and Fe2+ show different structuralbehavior than Sia+ and Ca2+ in melts in the system CaO-SiO2-Fe-O. (Compare Fig. 3 in Mysen et al., 1982b, inwhich composition SW40 is equivalent to CSl.4 in thepresent study.) The composition CS1.4F5 atlo. : l0-0 68(air) and 1550"C has NBO/T : 1.36. Compared with meltswith the same NBO/Si on the join CaO-SiO2, the iron-bearing melt has a larger proportion of three-dimensionalnetwork units, less Si2O3-, less SiO3- and slightly moreSiOX-. The increase in relative abundance of three-dimensional network units follows from the increasedpolymerization of composition CSl.4F5 compared withiron-free CSl.4 caused by tetrahedrally coordinatedFe3*. This increase in SiO2 units, coupled with an in-crease in SiO]- and a decrease in SizO3-, can also berelated to the structural position ofFe2+. The increase inthe proportion of SiO3 units indicates that Fe2*-Opolyhedra are formed with nonbridging oxygens fromSiO]- units rather than from nonbridging oxygens fromSi2O3-. At constant NBO/Si, the SiO2 + SiO3 increases

relative to SizO3 with increasing electronegativity of themetal cation (Liebau, l98l; Mysen et al., 1982b). Addi-tional reduction offerric to ferrous iron has only a verysmall effect on the overall degree of polymerization.

Inasmuch as Fe(IV)3+-O polyhedra appear to be three-dimensionally interconnected (Fox et al., 1982; Virgo etal., 1982, 1983b), this polyhedron may be described asFeO2 , as also suggested elsewhere (Paul and Douglas,1965; Douglas et al., 1965; Goldman, 1983). One may thenexpress a more complete redox equilibrium involving thesilicate anionic structure, ferrous and ferric iron,

4SiO2 + Si2O3 + 4FeOz = 5SiO3+ siox- + 4Fe2+(vD + 02, (l)

with the equilibrium constant

- . l re ' -<vt) \4 /s io3-\4 ls io| s io3-\-K , - l l l -'

\ Feoi / \ sio,-/ ls',o-l'' (2)

This expression is also consistent with the relationshipsbetween Fe2*/Fe3*, oxygen fugacity, NBO/T of the meltand Zlf of the network-modifying cations (Figs. 2-4). In

1.0

1550 oc,5 wg Fe2o3

F e 3 * / l F e

v MSr 4F5a csl 4F5

1550 "c x cst 4Fto

Fe3+ /zFe

E41MYSEN ET AL,: IRON IN ALKALINE EARTH SILICATE MELTS

|rE.'@tFt'q d'FF4

i . pt 5t _| ( n

''=' a

o

oo

i FI EI 6a &' E

o. !

c!

(g

oN

o

i :l ol l Lr 4

t

o

q

c)

ob0

x

: p6 . 6 nI . q iI E

I r r t El o

\ O -

b ; $

ra i;(J

4=

o6

iF

i . ' " 9| l o li ; ot o r| ; . f i

t >

6 ' .O E

tT

RI. l z

P o ir df , ri hi * -

8 E

Ea

: 8 E'L z.; R

d 0€ r

dEi \

ii t

q

!I

F

i E8: . ? z! ?e .I ! :

rt l+t . -6t

o

;. Fq o -9 @2 t 2

; id d9' ,iE Ht >

5r

EI

iD

; 8-ti 2

4

842

Table 3. Summary of relative areas (intensities) of Raman bandsand their assisnments

-to8fo, ffi r.3+/rr"

MYSEN ET AL,: IRON IN ALKALINE EARTH SILICATE MELTS

0 . 6 8

c s . 8 F 5

0 . 6 8 3 0 . 8 2 i . 9 + 4 1 . 3 o . 7 t

xAreas are re la t i ve (Z) to to ta l in tegre ted area o f S1-O s t re tch bandsin the 8OO-1200 cm- I f reqren .y range, In , th is h lgh- f requency enve lope,the in tens i t les o f the band near 1050 cm-1 (S i -o . s t re tch f ron br ida incoxygen, e .g . t Lasaga, 1982) and Ie -Oo s t re rch band6 (900 and 990 cn :1)are exc luded.

+ 8 5 0 , e t c . , d e n o t e r e l a t t v g a r e a s I i n t e n s i t i e 6 ) o f b a n d s ( S i - O - s t r e t c hin un i ts o f 5104" - , S i2o7o- , s lo jz - and s i ro .z - type , respecr ive ly ) o fRaMn bends w l th f requen i ies app;ox iMted 5y- these nuDbers . The 1150represent the re la t l ve a rea o f the S1-Oo s t re tch f ron lh ree-d lnens lona lneEwork un i ts ,

tTh ls^ inEend l ty p robab ly a lso ln t ludes e cont r ibu l ion f rm an unre€o lvedFe-Oo s t re tch band near 9OO cE

the range of bulk melt compositions of the present study,increasing NBO/T of the melt results in an increase in therelative abundance of SiO3- and SiOX- units and adecrease in SiO2 and Si2O?- (Furukawa et al., 1981;Mysen et a1., 1982b). The relative increase in SiOX- is

Fe3+/t Fe

Fig. 7. Relative abundance of coexisting anionic units(recalculated to l00Va; FeO! units < l%) as a function of Fe3*/)Fe for composition CSl.4F5. "Fe-free" denotes the relativeabundances for the iron-free end member (compare also withSW40; Mysen et al., 1982b).

. _. _. _.91F.:ll:.:yll1.

5 4 0.3 0.2 0.1 0.0

Fe3+ | EFa

Fig. 8. Calculated degree of polymerization (NBO/T) of meltswith 5 and l0 wt.Vo Fe2O3 added as a function of their Fe3*DFe.Compositions CSl.4 and CS have SiO2/CaO : 1.4 and 1.0,respectively, and MS 1.4 has SiO2/MgO = 1.4. F5 and FlO denote5 and l0 wt.Vo Fe2O3 added to the starting material. The NBO/Tis calculated from Fe3*DFe (Table l) after structural assignmentof Fe3*. The l ines csF5 [Fe3*(vD], csF5 [Fe3+(IV)], CSl.4F5[Fe3*(VD] and CSl.4[Fe3+(IV)] are calculated examples of thechange in bulk melt NBO/T of these compositions as a functionof Fe3*DFe if Fe3* is a network modifier or a network former,respectively, over the entire Fe3*DFe range from 1.0 to 0.0.

small compared with that of SiO3-. At constant/6., the(Sio3-/SiOt4 (Sio3-SioX-/Si2o3-) rherefore increases.For constant Kl this increase requires that Fe2+(VI)/FeOtdecrease. Thus, decreasing melt polymerization results inoxidation offerrous to ferric iron as observed here (Fig. 4)and elsewhere (e.g., Douglas et al., 1965; Larson andChipman, 1953).

Increasing Zlf of the metal cation is positively corre-lated with Fe2+/Fe3+ Gig. 2; see also Douglas et al., 1965;Paul and Douglas, 1965; Mysen et al., 1980). This correla-tion may be understood by considering equations (l) and(2) together with the influence of different metal cationson the proportions ofcoexisting anionic units. IncreasingZlf results in an increase in the abundance (SiOr +SiO3-) relative to SizO3- units for melts with similarmetaVsilicon (Mysen et al., 1982b; see also Liebau, 1981).As a result of this abundance change, the ratio product(Sio3-/SiOt4 (SioX-Sio3-/Si2o3-) in equation (2) is pos-itively correlated with increasing Zlf of the metal cation.In order to maintain the value of K1 in equation (2), theFe2*/Fe3* must decrease, in accord with observation.

Redox equilibria, melt structure and properties

Redox equilibria and melt polymerization

The redox ratio of iron oxides and the stmcturalpositions of Fe3+ and Fe2+ afect the overall degree of

cs2

3 , 0 4 4 , 2 4 5 . 2 7 . 6

c s , 8

1 4 . 8 5 3 . 2

cs2F5

c s l . 4

3 6 . 6

c s r . 4 F 5

4 0 . 4

4 1 . 8

c s t . 4 F 1 0

4 2 . a

0 . 5 8

0 . 6 8

4 . 06 . 0

0 . 6 8

4 , 0

4 . 56 , 5

3 . 0

3 r . 7

5 5 . 8

5 3 . 25 0 . 85 1 . 0) f . J

5r.2

0 . 3

0 . 6 10 . 2 90 . 1 20 . 0 0

0 . 6 8

l- 5 wr% Fe^O^Moleular Weisht 56{C.O) |

l - - 1 0 m % F e 2 O 3

40lM9O) -. - l0 wr% Feroa

l5il (BaOl---- 10wt% Feroa

\ i '

MYSEN ET AL.: IRON IN ALKALINE EARTH SILICATE MELTS E43

Fig. 9. Relative increase (%) in NBO/T of iron-bearing meltswith network-modifying Fe3* [NBO/T(VI)] compared with thecalculated NBO/T of the melt if Fe3* is a network former [NBO/T(IV)I as a function of total amount of iron, Fe3*DFe, molecularweight of the metal oxide and degree of polymerization of theiron-free end member ("NBO/T").

polymerization of silicate melts (NBO/T). The NBO/T ofseveral calcium silicate and magnesium silicate meltshave been calculated as a function of their iron contentsand Fe3+DFe (Fig. E). For all compositions, the additionof 5 wt.Vo Fe2O3 or more to the starting materials resultsin polymerization relative to the iron-free end membersbecause ferric iron is tetrahedrally coordinated in samplesequilibrated with air at 1550"C. Under all conditionsferrous iron is in octahedral coordination (Mao et al.,1973; Bell and Mao, 1974; Nolet etal.,1979; Seifert et al.,1979; Mysen et al., 1980).

In Figure 8 the observed changes in NBOiT are com-pared with the hypothetical and calculated variations ofNBO/T with Fe3+ as a network modifier [Fe3+(VI)] andnetwork former [Fe3+(IV)] throughout the Fe3*DFerange from 1.0 to 0.0. Because Fe3* undergoes a coordi-nation transformation with Fe3+DFe ( 0.5, there isan initial continuous increase in NBO/T with decreasingFe3+DFe and then a rapid change as Fe3* is transformedto octahedral coordination (Fig. 8). There is thus a largediference in the degree of polymerization of iron-bearingsilicate melts depending on Fe3+DFe and on whetherFe3* is in tetrahedral coordination as illustrated with thecalculated results shown in Figure 9. The extent of thisdifference is greater the smaller the NBO/T, the greaterthe total iron content and the larger the molecular weightof the metal cation.

Phase equilibriaIn binary metal oxide-silica systems the degree of

polymerization of liquidus minerals can be correlated

with the polymerization of the melt (see, e.g., Greig,1927; Phillips and Muan, 1959, for relevant phase-equilib-rium data on alkaline earth silicate systems). Such simplecorrelations no longer hold when Sia+ is replaced with acharge-balanced M3+ cation such as Fe3+ (or Al3+) asindicated in Figure 10. In this figure the bulk melt NBO/Tof the tridymite-pseudowollastonite liquidus boundaryhas been calculated for the system CaO-FeO-FezOr-SiO2 (Osborn and Muan, 1960a) and compared with themelt compositions (expressed as NBO/T) for the sameliquidus boundary in the system CaGAl2O3-SiOz (Os-born and Muan, 1960b). The liquidus phase relations inthe system CaO-FeO-FezOr-SiOz (Osborn and Muan,1960a) were determined in equilibrium with air, where thetemperatures along the tridymite-pseudowollastoniteboundary increase from 12(X'C at the invariant pointtridymite + pseudowollastonite + hematite + liquid to1436"C on the binary join CaO-SiO2. Thus, the Fe3*DFeof the melts is less than 1. The redox ratios of iron used tocompute the NBO/T of the melts on the liquidus bound-ary shown in Figure 10 were calculated from the presentdata (Table 2, Fig. 8). For all compositions, exceptperhaps those within l-2 wI.Vo iron oxide, the ferric ironis tetrahedrally coordinated.

In view of the commonly observed correlation betweenliquidus phase equilibria and important features of themelt structure (see, e.9., Kushiro, 1975; Burnham, 1981;Fraser et al., 1983), the expansion of the pseudowollas-tonite liquidus volume to more polymerized compositionswith increasing M3*/1M3*+Si4*) (M3* : Al3* or Fe3+)indicates diferent structural roles of the tetrahedrallycoordinated Fe3* and Al3* compared with that of Si4*. A

o

zF

6z

F

o6z

I

F

o,z

a

3 0.m=

=! 0 .

lridymis-Fddomll&nit- CrO. Al2O3 - SiO2-- CaO - F€O - F62O3 . SiO2

-10 -20 -30 -40 -50 -50

Percht d6@. NBO/T

Fig. 10. Change (Vo) of degree of polymerization of liquids atthe pseudowollastonite-tridymite liquidus boundary in thesystems CaO-Al2O3-SiO2 and CaO-FeO-FezO3-SiO2 as afunction of M3*/(M3* + Sia*) of the liquid (M3* refers to Fe3*and Al3*). The change is relative to the composition of theliquidus boundary in the system CaO-SiO2. Phase-equilibriumdata are from Osborn and Muan (1960a,b).

u4 MYSEN ET AL.: IRON IN ALKALINE EARTH SILICATE MELTS

0.9 0.8 07 0.6 0.5 0.4 0.3 0.2 0.1

Fo3*/EFe

FG3+/ EFo

Fig. ll. Calculated liquidus phase equilibria of compositionsCSl.4Fl0 and MSl.4Fl0 as a function of Fe3*/!Fe.

simple polymerization by increasing Si4+/Ca2+ results ina contraction of the tridymite liquidus volume.

It can be seen (Fig. l0) that the liquidus volume ofpseudowollastonite has expanded by slightly more than50% (decrease in NBO/T of the melt relative to the valueat the CaO-SiOz join) by the time hematite replaces themetasilicate liquidus phase. This expansion indicates thatthe activity of SiO]- units in the melts has increased. Thesituation is similar in the system CaO-Al2O3-SiOz Gig.l0). Thus, one may infer similar structural roles of Al3+and Fe3+ with Ca2+ for charge-balance. Available Ramandata indicate that both Al3* (Mysen et al., 1981, 1982c)and Fe3+ (Fig. 6; see also Table 3 and Virgo et al., 1981,1982) occur principally in three-dimensional networkstructural units in the melts. Solution of Fe2O3 (or Al2O3),therefore, tends to enhance the abundance of such units.This tendency requires, however, an increased abun-dance of depolymerized structural units in order to main-tain mass balance, as illustrated by the schematic expres-sion

si2o3- + M2o?- = 2sio3- + 2Mo2. (3)

In this expression, M could be either Fe3+ or Al3+ withsuitable charge-balance for their tetrahedral coordination

together with Sia+. Available data indicate that expres-sions such as (3) are shifted strongly to the right, thusleading to an enhanced proportion of three-dimensionalnetwork units together with depolymerized silicate unitssuch as SiOS-. With the assumption of ideal mixing of theM-cations in the MO2 unit, the activity of SiOz willincrease at a slower rate than that of SiO3- because of theselective substitution of A13* and Fe3* (or both) for Sia*in the MOz units in the melts. As a result, increasing M3*/(M3+ +Si4+) (M3+ : Al3* or Fe3*) causes an expansion ofthe liquidus volume of depolymerized silicate minerals.

Because the melt structure depends on Fe3*/)Fe,liquidus phase equilibria of iron-free silicate minerals canbe shown to vary with variable redox ratio of iron. Someconsequences of these relationships can be pursued withcalculated phase equilibria in portions of the systemsCaO-SiOrFe2O3-FeO and MgGSiOrFezOrFeO (Fig.I l). In Figure I I the liquidus phase equilibria of composi-tions CS1.4 and MSl.4 (iron-free NBO/Si is 1.4 for bothCa- and Mg-bearing compositions) with l0 wt.Vo Fe2O3added have been estimated as a function of Fe3*/)Fe.The values of Fe3*DFe are, of course, simple functionsof oxygen fugacity (see Fig. 4). Thus, oxygen fugacitycould replace Fe3*/)Fe as the x axis in both figures. Thecalculated topologies of the phase diagrams depend solelyon the chosen Fe3+/)Fe and the coordination polyhedraof Fe3* and Fe2+ in the melts. The data reported abovesupport these conclusions, but the liquidus phase equilib-ria shown in Figure I I can be calculated without thatinformation.

The phase equilibria have been obtained by calculatingCaO/SiOz and MgO/SiO2 of iron-free melts with NBO/Sicorresponding to the NBO/T values obtained from thevarious values of Fe3*DFe and the structural position ofFe3+ in the melts. In these calculations, it is assumed,therefore, that Fe3+(IV) occurs in separate complexeswhose composition is not affected by the extensive andintensive variables. The abundance of these complexesmerely affects the Si/O of the silicate network. It is alsoassumed that iron oxide minerals or solid solution of ironin the liquidus minerals is negligible, as indicated by theavailable phase-equilibrium data on the systems CaO-SiO2-Fe-O and MgO-SiO2-Fe-O (Muan and Osborn,1965). Finally, it is assumed that the indicated preferenceof Fe2+(VD for SiOl- and Fe3*1V9 for SiOX- units (seeabove) does not greatly affect the liquidus equilibria in theiron-concentration ranges of interest. For Fe2*(VD, thisassumption is probably justified for the Mg-bearing sys-tem in view of the crystal-chemical similarities of Fe2*and Mg2*. For the Ca-bearing system, the assumptionprobably leads to an underestimation of the activity ofSiO]- units. Thus, it is possible that the compositionalrange calculated as the liquidus stability field of metasili-cate (a-CaSiOl) may be too narrow.

Octahedrally coordinated Fe3+ enhances the SiOX-abundance in melts compared with both MgO-SiO2 andCaO-SiO2. The maximum Fe3*/(M2*+Fe3*) is, howev-er, so small (0.049 at Fe3*DFe = 0.3 in CSl.4Fl0) that a

P

tF

possible affinity of Fe3* for specific anionic units such asSiOf- has a negligible effect on the activities of otherunits in the melt. Inasmuch as the units of particularinterest are meta- and tectosilicates, the neglect of theFer'(VI) association with SiOf is justif ied.

In the system CaO-SiO2 @hillips and Muan, 1959), themelt polymerization (decrease in NBO/Si) resulting fromaddition of l0 wt.Vo Fe2O3 results in a change frompseudowollastonite at 1500'C to tridymite on the liquidusat 1470"C (Fig. I l). The tridymite liquidus field extends toFe3*/)Fe - 0.8 at the minimum liquidus temperature inthe system (-1435"C). Additional reduction of ferric toferrous iron is associated with pseudowollastonite on theliquidus and an increase in liquidus temperatures. Thesephase equilibria result from the increased liquidus tem-perature along the join CaO-SiO2 in the CalSi range (orNBO/Si : 2calsi) corresponding to the increased NBO/Si resulting from decreasing Fe3*DFe. A maximum liqui-dus temperature occurs between Fel*DFe = 0.2 and 0.1.

With less than l0 wt.%oFe2O3 added to CS1.4 composi-tion, reduction of ferric to ferrous iron may not result in asufficient NBO/Si increase (to 1.7) to reach the thermalmaximum. For end-member CaO-SiO2 compositionswith higher CalSi than that of CSl.4, the maximumliquidus temperature occurs with higher values of Fe3+/)Fe, or for the same Fe3*/)Fe, with lower total ironcontents. Finally, there is a narrow Fe3*/)Fe range withdiscontinuous changes in the slope of the pseudowollas-tonite liquidus curve. This Fel*DFe range correspondsto that where Fe3* undergoes transformation from net-work former to network modifier. For lower iron con-tents, this behavior of the liquidus curves will occur athigher Fe3+DFe (higher oxygen fugacity).

The composition MSl.4 has the same M/Si (and, there-fore, similar NBO/Si in the molten form) in the systemMgO-SiO2 as CSl.4 in the system CaO-SiO2. The phaseequilibria in MgO-SiO2 are, however, considerably dif-ferent from those of CaO-SiO2 (Greig, 1927; Phillips andMuan, 1959). First, MSl.4 has cristobalite on the liqui-dus, whereas the l iquidus phase of the CSl.4 compositionis a metasilicate (pseudowollastonite). This differencetranscends into the system MgO-SiO2-Fe2O3-FeO(Muan and Osborn, 1956; see also Muan and Osborn,1965, for summaries of relevant phase-equilibrium data).In equil ibrium with air, MSl.4 + l0 wt.To Fe2O3(MSl.4Fl0) has two immiscible l iquids coexisting at1780'C, whereas for CS1.4FI0, tridymite is a l iquidusphase at l470"C.In the magnesium-bearing system, as inthe calcium-bearing system, decreasing Fe3+/>Fe 1de-creasing/6) results in a rapid calculated decrease in theliquidus temperature to a minimum near 1570'C for Fe3+/)Fe - 0.3 where cristobalite coexists with clinoenstatite(Fig. ll). The coexisting tectosilicate and metasilicate atthe temperature minimum were also observed forCSl.4Fl0 but at higher values of Fe3+DFe (-0.8). Fur-thermore, the lowering of the liquidus temperature byreduction of ferric to ferrous iron is much greater in thesystem MgO-SiOz-Fe2O3-FeO (-200'C) than in the sys-

E45

tem CaO-SiO2-Fe2O3-FeO (-35"C). For the MSl.4Fl0composition, a further decrease in Fe3*DFe results in atemperature increase along which liquidus clinoenstatiteremains over only a narrow composition interval beforecristobalite again becomes a liquidus phase. This trenddiffers from that of CSl.4Fl0, where the metasilicateliquidus mineral remains unti l Fe3*/)Fe : 0 (Fig. l l).

Summary

From the Raman and Mossbauer spectroscopic datacoupled with published phase equilibria it is concludedthat:

l. The Fe3* is tetrahedrally coordinated in silicatemelts with alkaline earths for charge-balance and Fe3*/)Fe > 0.4-0.5, whereas for melts with lower Fer*/)Fe,ferric iron is a network modifier. Ferrous iron is always anetwork modifier.

2. The polymerization (NBO/T) of silicate melts is afunction of Fe3*DFe. All melt properties that depend onNBO/T are therefore dependent on Fe3'/)Fe.

3. The Fe3*/)Fe of alkaline earth silicate melts hassimple functional relations to oxygen fugacity, tempera-ture and the type of alkaline earth metal. Thus, the NBO/T of the melts depend on the same intensive and exten-sive variables.

4. Liquidus volumes of iron-free silicate minerals (e.g.,meta- and tectosilicate minerals) can then be estimated asa function of iron contents and Fe3*/)Fe of the liquid.

Acknowledgments

Critical reviews by L. W. Finger, P. McMillan, E. F. Osbornand H. S. Yoder, Jr., are appreciated.

References

Amthauer, G., Annersten, H., and Hafner, S. S. (1977) TheMrissbauer spectrum of 57Fe in titanium-bearing andradites.Physics and Chemistry of Minerals, 1,399-413.

Annersten, H. and Halenius, U. (1976) Ion distribution in pinkmuscovite: a discussion. American Mineralogist, 61, 1045-t050.

Annersten. H. and Olesch, M. (1978) Distribution of ferrous andferric iron in clintonite and the M6ssbauer characteristics offerric iron in tetrahedral coordination. Canadian Mineralogist,r8. 199-204.

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

Bowen, N. L., Schairer, J. F., and Wil lems, H. W. V. (1930) Theternary system: Na2Si03-Fe2O3-SiO2. American Journal ofScience, 20,405-455.

Brawer, S. A. and White, W. B. (1977) Raman spectroscopicinvestigation of the structure of silicate glasses. IL Soda-alkaline earth-alumina ternary and quaternary glasses. Journalof Non-Crystalline Solids, 23, 261-278.

Burnham, C. W. (1981) The nature of multicomponent alumino-silicate melts. Physics and Chemistry of the Earth, 13-14, 197-227.

Calas, G. and Petiau, J. (1983) Structure of oxide glasses:

MYSEN ET AL.: IRON IN ALKALINE EARTH SILICATE MELTS

8,16

spectroscopic studies of local order and crystallochemistry:geochemical implications. Bulletin de Mindralogie, 106, 33-55.

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

Deines, P., Nafziger, R. H., Ulmer, G. C., and Woermann, E.(1974) Temperature-oxygen fugacity tables for selected gas-mixtures in the system C-H-O at one atmosphere total pres-sure. Bulletin ofthe Earth and Mineral Sciences, PennsylvaniaState University, 88.

Dickinson, M. P., and Hess, P. C. (1981) Redox equilibria andthe structural role of iron in aluminosilicate melts. Contribu-tions to Mineralogy and Petrology, 78,352-358.

Domine, F. and Periou, B. (1983) Study of sodium silicate meltsand glasses with infrared reflectance spectroscopy. Journal ofNon-Crystalline Solids, 55, 125-131.

Douglas, E. W., North, P., and Paul, A. (1965) Oxygen ionactivity and its influence on the redox equilibrium in glasses.Physics and Chemistry of Glasses, 6,216-233.

Eibschutz, M. and Lines, M. E. (1982) Observation of second-order quadrupole shift in Mossbauer spectrum of amorphousYAG (yttrium garnet). Physical Reviews B,25,4256-4259.

Eibschutz, M., Lines, M. E., and Nassau, K. (1980) Electricfield-gradient distribution in vitreous yttrium garnet. PhysicalReview B. 21. 3767-3770.

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, 21, 169-178.

Fraser, D. G., Rammensee, W., and Jones, R. H. (1983) Themixing properties of melts in the system NaAlSi2O6-KAlSi2O6determined by Knudsen cell mass spectrometry. Bulletin deMindralogie, 106, II l-l17.

Fudali, F. (1965) Oxygen fugacity of basaltic and andesiticmagmas. Geochimica et Cosmochimica Acta, 29, 1063-1075.

Furukawa, T., Fox, K. E., and White, W. B. (1981) Ramanspectroscopic investigation of the structure of silicate glasses.III. Raman intensities and structural units in sodium silicateglasses. Journal of Chemical Physics, 75,3226-3237.

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

Greig, J. W. (1927) Immiscibility in silicate melrs, Pt. II. Ameri-can Journal of Science, 13, 133-155.

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

Kushiro, I. (1975) On the nature of silicate melt and its signifi-cance in magma genesis: regularities in the shift of liquidusboundaries involving olivine, pyroxene and silica minerals.American Journal of Science. 275. 4ll-431.

Larson, H. and Chipman, J. (1953) Oxygen activity in iron oxideslags. Transactions of the American Institute of Mining andMetallurgical Engineers, 196, 1089-1096.

Lasaga, A. C. (1982) Optimization of CNDO for molecularorbital calculation on silicates. Physics and Chemistry ofMinerals. 8. 36-46.

Lauer, H. V. and Morris, R. V. (1977) Redox equilibria ofmultivalent ions in silicate melts. Journal of the AmericanCeramic Society, 60, 443-451.

Levitz,P., Bonnin, D., Calas, G., and Legrpnd, A. P. (1980) Atwo-parameter distribution analysis of Mrissbauer spectra innon-crystalline solids using general inversion method. Journalof Physics E: Scientific Instruments. 13.427-437.

MYSEN ET AL.: IRON IN ALKALINE EARTH SILICATE MELTS

Liebau, F. (1981) The influence of cation properties on theconformation of silicate and phosphate anions. In A. Nav-rotsky and M. O'Keetre, Eds., Structure and Bonding inCrystals, Vol. 2, p. 197-232. Academic Press, New York.

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

Muan, A. and Osborn, E. F. (1956) Phase equilibria at liquidustemperatures in the system MgO-FeO-FezO3-SiO2. Journal ofthe American Ceramic Society, 39, l2l-140.

Muan, A. and Osborn, E. F. (1965) Phase Equilibria AmongOxides in Steelmaking. Addison-Wesley, Reading, Massachu-setts.

Mysen, B. O. and Virgo, D. (197E) Influence of pressure,temperature and bulk composition on melt structures in thesystem NaAlSi2O6-NaFe3+Si2O6. American Journal of Sci-ence.278. 1307-1122.

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

Mysen, B. O., Virgo, D., and Kushiro, I. (1981) The structuralrole of aluminum in silicate melts-a Raman spectroscopicstudy at I atmosphere. American Mineralogist, 66,678:701.

Mysen, B. O., Finger, L. W., Seifert, F. A., and Virgo, D.(1982a) Curve-fitting of Raman spectra of amorphous materi-als. American Mineralogist, 67, 686-696.

Mysen, B. O., Virgo, D., and Seifert, F. A. (1982b) The structureof silicate melts: implications for chemical and physical prop-erties ofnatural magma. Reviews ofGeophysics, 20, 353-383.

Mysen, B. O., Virgo, D., and Seifert, F. A. (1982c) Distributionof aluminum between anionic units in depolymerized silicatemelts as a function of pressure and temperature. CarnegieInstitution of Washingxon Year Book, 81, 360-366.

Navrotsky, A., Peraudeau, P., McMillan, P., and Coutoures, J.-P. (19E2) A thermochemical study of glasses and crystals alongthe joins silica-calcium aluminate and silica-sodium aluminate.Geochimica et Cosmochimica Acta. 46. 2039-2049.

Nolet, D. A., Burns, R. G., Flamm, S. L., and Besancon, J. R.(1979) Spectra of Fe-Ti silicates: implications to remote sens-ing of planetary surfaces. Proceedings of the l0th Lunar andPlanetary Science Conference, 177 5-1786.

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

Osborn, E. F. and Muan, A. (1960b) Phase equilibrium diagramsof oxide systems. Plate 2. The system CaO-AlzOr-SiOz.American Ceramic Society, Columbus, Ohio.

Paul, A. and Douglas, R. W. (1965) Ferrous-ferric equilibrium inbinary alkali silicate glasses. Physics and Chemistry of Glass-es, 6, 207-211.

Phillips, B. and Muan, A. (1959) Phase equilibria in the systemCaO-iron oxide-SiO2 in air. Journal of the American CeramicSociety, 42,413423.

Presnall, D. C. and Brenner, N. L. (1974) A method for studyingiron silicate liquids under reducing conditions with negligibleiron loss. Geochimica et Cosmochimica Acta. 38. 1785-178E.

Sato, M. (1972) Electrochemical measurements and control ofoxygen fugacity and other gaseous fugacities with solid elec-trolyte systems. In G. C. Ulmer, Ed., Research Techniquesfor High Pressure and High Temperature, p. 43-99. SpringerVerlag, New York.

Sato, M. and Valenza, M. (1980) Oxygen fugacities of the

MYSEN ET AL.: IRON IN ALKALINE EARTH SILICATE MELTS u7

layered series of the Skaergaard intrusion, East Greenland,American Journal of Science, 280-,4,. 134-158.

Seifert, F. A., Virgo, D., and Mysen, B. O. (1979) Melt struc-tures and the redox equilibria in the system NA2O-FeO-Fe2O3-Al2O3-SiO2. Carnegie Institution of Washington yearBook, 78, 5l l -519.

Seifert, F. A., Mysen, B. O., and Virgo, D. (1981) Quantitativedetermination of proportions of anionic units in silicate melts.Carnegie Institution of Washington Year Book, E0, 301-302.

Seifert, F. A., Mysen, B. O., and Virgo, D. (19E2) Three-dimensional network melt structure in the systems SiOrNaAlO2, SiO"-CaAl2O4 and SiO2-MgAl2Oa. American Miner-alogist,67, 696:71E.

Verweij, H. (1979a) Raman study of the structure of alkaligermanosilicate glasses. I. Sodium and potassium metager-manosilicate glasses. Journal of Non-Crystalline Solids, 33,4l-53.

Verweij, H. (1979b) Raman srudy of the structure of alkaligermanosilicate glasses. II. Lithium, sodium and potassiumdigermanosilicate glasses. Journal of Non-Crystalline Sotids,33 ,55-69.

Verweij, H. and Konijnendijk, W. L. (1976) Strucrural units inK2O-PbO-SiO, glasses by Raman spectroscopy. Journal ofthe American Ceramic Society, 59, 517-521.

Virgo, D., Mysen, B. O., and Kushiro, I . (1980) Anionicconstitution of silicate melts quenched at I atm from Ramanspectroscopy: implications for the structure of igneous melts.Science, 208, 137 l-1373.

Virgo, D., Mysen, B. O., and Seifert, F. A. (19E1) Relationshipbetween the oxidation state of iron and structure of silicatemelts. Carnegie Institution of Washington Year Book, 80, 30E-3 l l .

Virgo, D., Mysen, B. O., Danckwerth, P. A., and Seifert F. A.(t982) Speciation of Fe3+ in l-atm Na2O-SiO2-Fe-O melts.Carnegie Institution of Washington Year Book, 81,349-353.

Virgo, D., Mysen, B. O., and Danckwerth, P. A. (l9E3a) Thecoordination ofFe3* in oxidized vs. reduced sodium alumino-silicate glasses: a 57Fe Mdssbauer study. Carnegie Institutionof Washington Year Book, E2, 309-313.

Virgo, D., Mysen, B. O., and Danckwerth, P. A. (1983b) Redoxequilibria and the anionic structure of Na2O .xSiO2-Fe-O

melts: effects of oxygen fugacity. Carnegie Institution ofWashington Year Book, E2, 305-309.

Waychunas, G. A. and Rossman, G. R. (19E3) Spectroscopicstandard for tetrahedrally coordinated ferric-iron: yLiAlOz:Fe3*. Physics and Chemistry of Minerals, 9,212-215.

Whittaker, E. J. W. and Muntus, R. (1970) Ionic radii for use ingeochemistry. Geochimica et Cosmochimica Acta, 34, 945-957.

Wivel, C. and Morup, S. (lgEl) Improved computational proce-dure for evaluation ofoverlapping hyperfine parameter distri-butions in M<issbauer spectra. Journal ofPhysics E: ScientificInstruments. 14. 605-610.

Manuscript received, October 21, 1983;accepted for publication, May I, 1984.