The effect of composition on Cr2+/Cr3+ in silicate melts

American Mineralogist, Volume 91, pages 1901–1908, 2006

0003-004X/06/1112–1901$05.00/DOI: 10.2138/am.2006.2097 1901

INTRODUCTION

Chromium occurs exclusively as Cr3+ in minerals crystallized at mantle oxygen fugacities (fO2

) values (Burns and Burns 1975). If Cr were to occur as both Cr3+ and Cr2+ in mantle melts, the crystal-melt partition coeffi cients would be expected to exhibit anomalous behavior relative to those of elements that occur in a single valence state; which may account for the early crystal-lization of chromite from basalts (Li et al. 1995). The presence of Cr2+ at values of terrestrial fO2

in Fe-free analogs of natural silicate glasses (Cr2+/ΣCr ~ 0.5, where ΣCr = Cr2+ + Cr3+) has been known for some time (Schreiber and Haskin 1976). The importance of Cr2+ in Fe-bearing compositions, however, has been less clear. Although crystal-melt partitioning experiments strongly indicate that Cr2+ does occur in Fe-bearing melts (e.g., Barnes 1986; Hanson and Jones 1998; Poustovetov and Roeder 2000), this oxidation state has never been observed directly in a terrestrial material. Indeed, it has been argued that Cr2+ does not occur in basaltic glasses and melts due to the difference in the redox potentials of Cr2+/3+ and Fe2+/3+, which suggests that only Fe2+, Fe3+, and Cr3+ should coexist at values of terrestrial fO2

(Schreiber and Haskin 1976; Schreiber et al. 1987). Recently, it has been shown by in situ X-ray absorption near-edge structure (XANES) spectroscopy of a synthetic mid-ocean ridge basalt

(MORB) composition, as a function of variable fO2 (terrestrial

values) at 1200–1400 °C, that Cr2+ is a signifi cant oxidation state in the melt (Berry et al. 2003). Cr2+ is not, however, retained on quenching to a glass due to an electron exchange reaction with Fe3+ (Cr2+ + Fe3+ → Cr3+ + Fe2+), suggesting that this reaction, or the difference in redox potentials, is temperature dependent. The abundance of Fe relative to Cr indicates there will always be suffi cient Fe3+, at terrestrial fO2

values, to completely react with, and remove, Cr2+ on cooling.

Recently a XANES method for quantifying Cr2+/ΣCr in glasses has been presented (Berry and O’Neill 2004). A shoulder on the main absorption edge systematically changes intensity with fO2

and must relate to, and can be used to quantify, the oxi-dation state of Cr. The intensity of this shoulder was determined for fi ve CaO-MgO-Al2O3-SiO2 (CMAS) compositions (plus one containing TiO2) equilibrated at ~20 values of logfO2

(between –0.68 and –16.13 at 1400 °C) and fi t to a thermodynamically derived expression (see section on XANES analysis), allowing Cr2+/ΣCr to be evaluated for each sample. The good agreement between the data and the fi t suggests that it should be possible to determine the dependence of Cr2+/ΣCr on logfO2

from a more limited number of data points. Indeed, samples prepared under conditions where Cr2+/ΣCr ≈ 0, 0.5, and 1 should be suffi cient, making the investigation of a large range of melt compositions feasible.

In the present study, Cr K-edge XANES spectra were recorded for glasses of 27 compositions quenched from melts equilibrated at 1400 °C and logfO2

values of –2 (Cr2+/ΣCr ~ 0), –7, –9 (Cr2+/ΣCr

* Present Address: Department of Earth Science and Engineering, Imperial College London, South Kensington, SW7 2AZ, U.K. E-mail: [email protected]

The effect of composition on Cr2+/Cr3+ in silicate melts

ANDREW J. BERRY,1,* HUGH ST.C. O’NEILL,1 DEAN R. SCOTT,1 GARRY J. FORAN,2,3 AND J.M.G. SHELLEY1

1Research School of Earth Sciences, Australian National University, Canberra, ACT 0200, Australia2Australian Synchrotron Research Program, Australian Nuclear Science and Technology Organisation, PMB 1, Menai, NSW 2234, Australia

3Australian National Beamline Facility, KEK, Photon Factory, Oho 1-1, Tsukuba-shi, Ibaraki-ken 305-0801, Japan

ABSTRACT

Chromium K-edge X-ray absorption near-edge structure (XANES) spectra were recorded at room temperature for 27 CaO-MgO-Al2O3-SiO2 (CMAS) glass compositions quenched from melts equilibrated at various oxygen fugacities (fO2

) at 1400 °C. Values of Cr2+/ΣCr were determined from the intensity of a shoulder on the main absorption edge, attributed to the 1s → 4s transition, which is characteristic of Cr2+ in these glasses. For each composition, Cr2+/ΣCr could be quantifi ed as a function of fO2

, using a theoretical expression, from as few as three samples (Cr2+/ΣCr ≈ 0, 0.5, and 1). This allowed logK', or the reduction potential of the Cr3+/2+ half-reaction, and hence the relative change in the ratio γCr

melt3+O1.5

/ γCrmelt

2+O, to be determined for each composition. At constant fO2, log[Cr2+/Cr3+] was

found to decrease linearly with increasing optical basicity. The variation in logK' with composition is controlled by γCr

melt3+O1.5

, corresponding to the capacity of the melt to stabilize both the charge and the preferred solvation site of Cr3+. The method was then applied to spectra recorded in situ at 1400 °C for a synthetic mid-ocean ridge basalt (MORB) composition, allowing Cr2+/ΣCr to be quantifi ed in a Fe-bearing melt for the fi rst time. Cr2+/ΣCr was found to vary from ~0.45 at the nickel-nickel oxide (NNO) fO2

buffer to ~0.90 at iron-wüstite (IW). This indicates that Cr2+ is likely to be the dominant oxidation state in terrestrial basaltic melts.

Keywords: XANES spectroscopy, chromium oxidation states, silicate melts, MORB

BERRY ET AL.: Cr2+/Cr3+ IN SILICATE MELTS1902

~ 0.5), and –16.1 (Cr2+/ΣCr ~ 1). All the compositions are Fe-free to enable samples to be studied as quenched glasses that preserve the high-temperature Cr oxidation state ratio. This study should provide insight into the compositional parameters that stabilize Cr2+ and/or Cr3+ and allow Cr2+/ΣCr for natural compositions to be predicted. This was tested using available 1400 °C in situ data for an Fe-bearing MORB melt.

EXPERIMENTAL METHODS

Glasses were prepared with 27 different compositions from reagent grade SiO2, Al2O3, MgO, and CaCO3 to which 0.5 wt% Cr2O3 was added. The glasses comprise compositional series previously described and referred to as AD6, CMAS7, CAS/MAS, Longhi6, and Osborn6 (O’Neill and Eggins 2002; O’Neill and Mavrogenes 2002; O’Neill and Berry 2006). The compositions were sus-pended on loops of either Pt or Re in a gas-mixing vertical tube furnace at 1400 °C and equilibrated with mixtures of O2/CO2 (logfO2

= –2) or CO2/CO (logfO2 =

–7, –9). Cr-free oxides were also equilibrated under CO between layers of Cr3C2 in a graphite capsule, resulting in logfO2

= –16.13 and a Cr2O3 solubility of ~0.5 wt% in the melt (O’Neill and Berry 2006). After at least 24 h (72 h for those in equilibrium with Cr3C2) all samples were quenched in water to produce glasses, mounted in epoxy, and polished. For the samples equilibrated at logfO2

= –16.13, any Cr3C2 associated with the glass was identifi ed readily by optical microscopy and masked with Pb tape. Further experimental details are given in Berry and O’Neill (2004). The averages of the logfO2

= –2, –7, and –9 samples for the 27 compositions, determined by energy-dispersive electron microprobe analysis, are given in Table 1. An Fe-bearing MORB composition (in wt%: SiO2 = 51.8, Al2O3 = 16.0, CaO = 12.3, FeO = 10.0, MgO = 8.2, TiO2 = 1.3, Cr2O3 = 0.5) was studied

at 1400 °C using a furnace designed for in situ XANES spectroscopy of silicate melts (Berry et al. 2003).

Cr K-edge XANES spectra were recorded in fl uorescence mode at the Aus-tralian National Beamline Facility; Beamline 20B (bending magnet) at the 2.5 GeV Photon Factory (KEK, Tsukuba, Japan). The excitation energy was selected using a Si(111) channel cut crystal monochromator, resulting in a spectral energy resolution (when coupled with the CrKα core-hole width) of 2.4 eV. Fluorescence was recorded using a Canberra GL0110S 10 element Ge array detector. The energy was calibrated by defi ning the fi rst derivative peak of Cr in stainless steel foil to be 5989.2 eV. Spectra were recorded from 5985–6025 eV with a step size of 0.25 eV. For further details see Berry and O’Neill (2004).

XANES ANALYSIS

A shoulder on the K absorption edge of Cr in CMAS glasses increases in intensity with decreasing fO2

. This shoulder is attrib-uted to the symmetry forbidden 1s → 4s transition of Cr2+ (Sutton et al. 1993), although the exact assignment is irrelevant for this work. The 1s → 4s transition is not observed for Cr3+ due to the strong preference for high-symmetry octahedral coordination (readily accommodated by the coordination freedom available in a melt), but is allowed for Cr2+ that occurs in a Jahn-Teller distorted environment (so as to remove the d4 electronic degen-eracy); the shoulder is thus diagnostic of Cr2+. The intensity of the 1s → 4s transition is most easily quantifi ed from the derivative spectrum in which the shoulder appears as a peak. The area of this peak (from 5991.3–5996.3 eV), denoted as A1s/4s, is linearly

Comp. SiO2 Al2O3 MgO CaO NBO/T Basicity logfO2 = –2 logfO2

= –7 logfO2 = –9

A1s/4s Cr2+/ΣCr A1s/4s Cr2+/ΣCr A1s/4s Cr2+/ΣCr 1 AD 50.6(6) 15.1(3) 10.3(2) 24.0(7) 0.941 0.605 0.024* 0.038 0.207§ 0.475 0.296 0.688 0.094 0.507 0.709 2 AD+Fo 49.4(3) 13.0(2) 17.1(2) 20.5(5) 1.230 0.609 0.020* 0.040 0.151 0.341 0.297 0.676 0.075 0.368 0.697 3 AD+Wo 51.5(12) 6.2(2) 4.0(2) 38.3(12) 1.474 0.632 0.020* 0.027 0.123 0.298 0.234 0.590 0.063 0.325 0.606 4 AD+En 54.6(3) 9.3(2) 21.3(5) 14.8(4) 1.286 0.594 0.021* 0.045 0.190 0.420 0.328 0.726 0.065 0.434 0.736 5 AD+Qz 67.8(10) 9.9(2) 6.7(2) 15.6(7) 0.525 0.557 0.022* 0.060 0.263§ 0.597 0.348 0.786 0.096 0.614 0.798 6 Longhi1 63.5(2) 7.2(1) 13.3(1) 16.0(1) 0.909 0.573 0.015 0.062 0.234 0.533 0.355 0.795 7 Longhi2 49.9(2) 13.4(1) 5.2(1) 31.5(2) 1.022 0.618 0.013* 0.033 0.151 0.368 0.269 0.654 8 Longhi3 63.6(1) 17.4(1) 9.0(1) 10.0(1) 0.330 0.553 0.018 0.100 0.306 0.656 0.418 0.871 9 Longhi4 63.4(2) 14.2(1) 4.7(1) 17.7(1) 0.440 0.564 0.011 0.070 0.244 0.567 0.360 0.813 10 Longhi5 69.9(7) 16.3(4) 8.5(2) 5.3(1) 0.197 0.535 0.032† 0.103 † † 0.433 0.866 11 Longhi6 63.1(1) 14.6(1) 8.2(1) 14.1(1) 0.464 0.561 0.010 0.079 0.268 0.603 0.380 0.830 12 Osborn1 47.1(2) 9.7(1) 8.5(1) 34.7(1) 1.508 0.636 0.007* 0.021 0.122 0.309 0.226 0.570 13 Osborn2 51.2(2) 14.6(1) 3.6(1) 30.6(2) 0.867 0.611 0.016 0.035 0.165 0.402 0.274 0.673 14 Osborn3 48.4(1) 14.6(1) 13.4(1) 23.6(1) 1.117 0.612 0.013 0.035 † † 0.284 0.673 15 Osborn4 45.5(1) 14.5(1) 10.5(1) 29.5(1) 1.238 0.628 0.001* 0.027 † † 0.247 0.607 16 Osborn5 42.4(2) 19.4(1) 8.5(1) 29.7(1) 1.011 0.630 0.008 0.019 † † 0.213 0.525 17 Osborn6 50.6(2) 19.5(1) 14.4(1) 15.5(1) 0.726 0.591 0.019 0.052 † † 0.340 0.756 18 CMAS–A 56.1(5) 15.7(5) 12.1(2) 16.1(1) 0.695 0.581 0.016 0.064 0.243 0.545 0.361 0.796 19 CMAS–B 57.0(9) 14.2(4) 20.1(5) 8.7(1) 0.838 0.574 0.019 0.078 0.251 0.566 0.390 0.859 20 CMAS–C 57.5(7) 6.6(3) 19.5(4) 16.4(2) 1.305 0.591 0.017 0.050 0.214 0.481 0.338 0.750 21 CMAS–E 49.3(6) 10.3(4) 19.0(3) 21.4(1) 1.463 0.614 0.008 0.036 0.143 0.339 0.282 0.654 22 CMAS–F 61.9(8) 6.1(2) 13.5(4) 18.5(1) 1.057 0.581 0.017 0.083 0.219 0.582 0.335 0.871 23 CMAS–G 59.3(9) 12.6(6) 2.2(2) 25.9(2) 0.635 0.585 0.012 0.049 0.194 0.471 0.311 0.742 24 CAS–2 40.9(4) 12.1(3) 0.0(0) 47.0(3) 1.571 0.667 –0.007* 0.012 0.080 0.237 0.172 0.474 25 CAS–3 69.0(4) 20.5(2) 0.0(0) 10.5(1) –0.017 0.535 0.034† 0.090 † † 0.407 0.847 26 CAS–4 53.2(5) 12.7(1) 0.0(0) 34.1(4) 0.852 0.611 0.010 0.034 0.156 0.396 0.266 0.669 27 MAS–1 61.5(4) 18.2(2) 20.3(2) 0.0(0) 0.473 0.550 0.043 0.118 0.351 0.704 – – Notes: Uncertainties are one standard deviation; σ(logfO2) = ±0.03; σ(A1s/4s) = ±0.005; σ(Cr2+/ΣCr) = ±0.03; σ(a0) = ±0.03 (±0.04), σ(a1) = ±0.06 (±0.09), σ(logK’) = ±0.04 (±0.06), σ(logγ#

CrO1.5) = ±0.04 (±0.06); values in parentheses are uncertainties resulting from fi ts to only three data points. For compositions 1–5, the values in italics are those determined in Berry and O’Neill (2004).* Samples containing Cr6+. † Samples containing a Cr3+ crystalline phase (typically spinel); values omitted for spinel-rich samples. ‡ Samples containing Cr3C2. § logfO2

= –7.5. || Fixed at 0.029.

TABLE 1. Average glass compositions expressed as wt% oxide, excluding ~0.5 wt% Cr2O3; the structural parameter NBO/T and the optical basicity; the 1s → 4s derivative peak areas (A1s/4s) and corresponding Cr2+/ΣCr for four values of logfO2

; parameters obtained from the fi t to Equation 6 (a0, a1, and logK’); and activity coeffi cients γ#

CrO1.5 (see text for defi nition)

BERRY ET AL.: Cr2+/Cr3+ IN SILICATE MELTS 1903

correlated with Cr2+/ΣCr resulting in the expression,

Cr2+/ΣCr = a0 + a1A1s/4s (1)

where a0 and a1 are constants that depend on the melt composi-tion. Cr2+/ΣCr is related to fO2

by the reaction:

Cr2+O + 1/4O2 = Cr3+O1.5. (2)

From the equilibrium constant, K, for this reaction we ob-tain:

log log . .KX

X= +Cr O

melt

Cr O

melt

Cr O

me3+

2+

3+1 5 1 5γ llt

Cr O

melt O2+γ+1

4 2log f (3)

where X is the mole fraction and γ the activity coeffi cient. For low ΣCr, as in this study, it may be assumed that γCr

melt3+O1.5

and γCr

melt2+O are constant for a given composition (Henry’s Law), and

that Cr3+-Cr2+ interactions are insignifi cant. It is then convenient to defi ne a modifi ed equilibrium constant, K', for the reaction referenced to a standard state of Cr2+O and Cr3+O1.5 at infi nite dilution, such that:

log log log .′= −K Kγ

γCr O

melt

Cr O

melt

3+

2+

1 5 (4)

Rearrangement of Equation 3 then gives (Berry and O’Neill 2004),

Cr

Cr

+

O

2

1 4

1

1 10 2Σ=+ + ′( )log logf K

.

(5)

This function has a sigmoidal form when Cr2+/ΣCr is plotted against logfO2

, i.e., at low values of fO2 Cr2+/ΣCr is constant and

equal to 1, at high values of fO2 Cr2+/ΣCr is constant and equal

to 0, with the two extremes being connected by a transitional region where both Cr2+ and Cr3+ coexist. The logfO2

range over which the oxidation of Cr2+ to Cr3+ occurs, or the “slope” of the sigmoidal curve, is controlled by the number of electrons involved in the reaction (one in this case) and is described by the 1/4 coeffi cient in Equation 5. The position of the curve relative to the logfO2

axis is given by the logK' term. –logK' is defi ned as the electrochemical reduction potential (EM') of a metal in a melt (Schreiber 1987). Combining Equations 1 and 5 gives

Aa

af K1s/4sO

=+

−⎛

⎝⎜⎜⎜⎜

⎞

⎠⎟

+ ′( )1 1

1 1011 4 0

2log log

⎟⎟⎟⎟⎟. (6)

Using this equation it is possible to determine logK', a0, and a1 (and hence Cr2+/ΣCr) from a series of XANES spectra pre-pared over a range of logfO2

. In our previous work (Berry and O’Neill 2004), A1s/4s values were determined for fi ve composi-tions at ~20 values of logfO2

. The resulting fi ts to Equation 6 were excellent and verifi ed the assumptions used in deriving this equation, including both Henry’s Law behavior and the linear relationship between Cr2+/ΣCr and A1s/4s. The relative stability of Cr2+ in a particular melt composition is given by the logK' value, which displayed substantial variability: i.e., Cr2+/ΣCr varied between ~0.3 and 0.7 at NNO-2 for the compositions studied. This indicates a signifi cant compositional effect on either γCr

melt3+O1.5

or γCrmelt

2+O.The small errors in logfO2

and A1s/4s, and the resulting good agreement between the data and fi t [resulting in σ(Cr2+/ΣCr) ± 0.015], indicate that the sigmoidal curve, and hence logK', may be defi ned from samples prepared under conditions where Cr2+/ΣCr ≈ 0, 0.5, and 1; the values of A1s/4s at Cr2+/ΣCr ≈ 0 and 1 effectively defi ne a0 and a1, whereas that at ≈ 0.5 will constrain logK'. An extra point in the region where Cr2+/ΣCr is changing rapidly will improve the accuracy of the method.

RESULTS

The samples are typically optically transparent and homoge-neous, and vary in color from blue (reduced) to green (oxidized). Some samples, most commonly those equilibrated at logfO2

= –7, contain crystals of a Cr3+ phase (usually spinel). Glasses equilibrated at logfO2

= –2 and –16.1 lost variable amounts of Cr (which will not affect the results since we are in the Henry’s Law region) and SiO2 (up to 2 wt%; see O’Neill and Berry 2006) respectively, by volatility.

Chromium K-edge XANES spectra for several glasses are shown in Figure 1. For all spectra a constant baseline was sub-tracted followed by normalization to the edge crest (maximum intensity). Although it is more common to normalize spectra to fl uorescence above the edge, crest normalization is part of an established procedure for quantifying Cr2+/ΣCr in which the predicted relationship between Cr2+/Cr3+ and fO2

is obtained within extremely precise experimental error (Berry and O’Neill 2004). Figure 1a shows spectra for six compositions equilibrated at

logfO2 = –16 a0 a1 logK‘ log γCrO1.5

A1s/4s Cr2+/ΣCr 0.423 0.992 –0.020 2.39 1.91 –1.91

0.997 0.039(10) 2.265(24) 1.871(17) –1.870.431 0.983 –0.006 2.30 1.99 –1.95

0.998 0.029|| 2.247(14) 1.921(12) –1.880.385 0.987 –0.025 2.63 2.11 –2.19

0.989 0.013(10) 2.534(26) 2.056(18) –2.140.446 0.987 –0.002 2.22 1.86 –1.94

0.992 0.020(10) 2.181(23) 1.844(16) –1.920.441 0.994 0.011 2.23 1.70 –1.71

0.999 0.048(12) 2.154(27) 1.650(19) –1.660.447 0.992 0.031 2.15 1.68 –1.830.408 0.990 0.002 2.42 1.98 –1.910.481 0.992 0.065 1.93 1.46 –1.440.445 0.994 0.047 2.12 1.63 –1.590.502 0.997 0.042 1.90 1.44 –1.050.462 0.996 0.060 2.03 1.57 –1.510.393 0.990 0.004 2.50 2.12 –2.050.404 0.993 –0.004 2.47 1.93 –1.950.419 0.992 0.004 2.36 1.94 –1.890.409 0.989 0.023 2.36 2.06 –1.920.400 0.985 0.001 2.46 2.21 –2.020.449 0.995 0.010 2.19 1.76 –1.760.455 0.994 0.030 2.12 1.67 –1.830.447‡ 0.979 0.039 2.10 1.59 –1.810.450 0.993 0.014 2.18 1.78 –1.970.429 0.986 0.017 2.26 2.01 –1.990.379‡ 0.978 0.041 2.47 1.56 –1.760.419 0.993 0.020 2.32 1.80 –1.860.371 0.984 0.031 2.57 2.28 –2.070.481 0.997 0.020 2.03 1.51 –1.220.397 0.993 0.010 2.48 1.94 –1.990.505 0.998 0.036 1.90 1.37 –1.50

TABLE 1.—Extended


logfO2 = –9. The 1s → 4s transition of Cr2+ appears as a shoulder

on the absorption edge (Fig. 1a) and a peak in the derivative spec-trum (Fig. 1b). The area of this peak, A1s/4s, (determined between the dashed lines) is linearly correlated with Cr2+/ΣCr. Although A1s/4s varies for a given value of Cr2+/ΣCr with composition, to fi rst order, the spectra in Figures 1a and 1b suggest signifi cant differences in Cr2+/ΣCr, at constant fO2

, for each glass.Samples were prepared at four values of logfO2

(–2, –7, –9, and –16.1) and a typical set of spectra for one composition is given in Figure 1c. For increasing fO2

the spectra exhibit a de-crease in 1s → 4s transition intensity and an increase in edge energy, as expected for an increase in Cr3+ relative to Cr2+. The corresponding derivative spectra are shown in Figure 1d, and the areas of the 1s → 4s peaks at all values of logfO2

, for all compositions, are given in Table 1. The uncertainty in A1s/4s was estimated previously (Berry and O’Neill 2004), for spectra re-corded and processed in an identical fashion, as ±0.005 and that in logfO2

as ±0.03 (all errors reported in this paper are quoted as one standard deviation). The presence of Cr6+ or Cr-rich spinel in some samples was readily identifi ed by characteristic spectral features; a 1s → 3d pre-edge peak for Cr6+, and a shoulder and crest shift for Cr-spinel (Berry and O’Neill 2004). These glasses are indicated in Table 1. The presence of spinel should have no effect on the Cr2+/Cr3+ ratio of the melt since the equilibrium between the melt and oxygen buffer (Eq. 2) is independent of this phase. The decrease in the Cr content of the melt resulting from spinel crystallization also has no effect because of Henry’s Law. The presence of a spinel contribution to the XANES spectrum, however, will result in an A1s/4s value inconsistent with Cr2+/ΣCr of the melt. For spinel containing samples prepared at logfO2

= –2 (Cr2+/ΣCr ~ 0) the A1s/4s values are given in Table 1 since the

effect on Cr2+/ΣCr is likely to be small. Spinel is rare at this fO2,

presumably because crystallization is suppressed by the vola-tile loss of Cr from the melt. The A1s/4s values are omitted for spinel contaminated samples prepared at logfO2

= –7. The Cr6+ contribution is weak in all cases and is not expected to affect A1s/4s signifi cantly (Berry and O’Neill 2004). The spectra of two samples equilibrated at logfO2

= –16.1 also indicate a contribution from Cr3C2 (19 and 22).

The A1s/4s values were fi t to Equation 6. In most cases, the fi t is excellent and an example is shown in Figure 2a (solid symbols). The equation is only unable to fi t the data adequately when one of the points is for a sample containing Cr-spinel (Fig. 2a; open symbols); if that point is excluded, then the fi ts are consistent with those of the other data sets. The fi t parameters a0, a1, and logK', together with the resulting values of Cr2+/ΣCr determined using Equation 1, are given in Table 1. The relationship between Cr2+/ΣCr and logfO2

defi ned by the value of logK' (Eq. 5) is shown in Figure 2b for several compositions. The uncertainty in Cr2+/ΣCr was estimated as ±0.03 by scaling the value of ±0.015 determined in Berry and O’Neill (2004) for the reduced number of data. If this uncertainty is propagated through Equation 5, the expected error in logK' is ±0.04 for fi ts to four data points and ±0.06 for fi ts to three points.

Cr K-edge XANES spectra for a synthetic MORB recorded in situ at 1400 °C and logfO2

values of –6, –7, –8, –9, and –10 have been reported previously (see Fig. 3b of Berry et al. 2003). The spectra were treated in an identical manner as for the glasses described here. The derivative peak areas for each value of logfO2

are given in Table 2 and were fi t to Equation 6. The resulting parameters have large uncertainties (e.g., logK' = 1.59 ± 0.12). This is due to the absence of data at very high and low fO2

values,

0

1

Nor

mal

ised

Inte

nsity

a

1s 4s

5990 6000 6010

0.0

0.1

Der

ivat

ive

b

Energy/eV

1s 4s

c

-16

-9

-7

-2

5990 6000 6010

d

Energy/eV

FIGURE 1. (a) Cr K-edge XANES spectra for compositions (in order of increasing edge energy) 10, 19, 17, 14, 12, and 24, quenched from 1400 °C and logfO2

= –9, and (b) the corresponding derivative spectra. The feature attributed to the 1s → 4s transition of Cr2+ is indicated. The integral range used to determine the area of the 1s → 4s peaks in the derivative spectra (A1s/4s) is shown by the dashed lines. (c) Cr K-edge XANES spectra for composition 18 equilibrated at the values of logfO2

indicated, and (d) the derivative spectra.


which constrain a0 and a1, resulting in all three parameters being highly correlated. This correlation means that the uncertainty in logK' may be greatly reduced by constraining a0 to a physically

reasonable value. If a0 is set to 0.019 (the average value in Table 1), then logK' = 1.559 ± 0.019 and a1 = 2.319 ± 0.028. Although a0 for this composition may be slightly different from the average value, we suggest that a realistic uncertainty in logK' is ±0.05, similar to those determined for the Fe-free glasses. The resulting Cr2+/ΣCr values and fi t are plotted in Figure 2c.

The activity coeffi cients γ*CrO, calculated relative to composi-tion AD (i.e., γ*CrO = γCrO/γAD

CrO), have been determined at 1400 °C for the melt compositions used in this study (O’Neill and Berry 2006). From Equation 4, we defi ne logγ*CrO1.5

= logγ*CrO – logK' + C, where C = logK, and then logγ#

CrO1.5 = γ*CrO1.5

– C = logγ*CrO – log K'. The calculated activity coeffi cients logγ #CrO1.5

are given in Table 1. These relative coeffi cients can be quantitatively used to address the effect of melt composition on reaction 2.

DISCUSSION

Equation 6 provides excellent fi ts to the A1s/4s values obtained for the four logfO2

conditions at which each composition was equilibrated. The equation, which is based on the fO2 dependence of Cr2+/Cr3+ in a melt, reassuringly cannot fi t the data when one of the points is “contaminated” by a contribution from Cr in spinel or Cr3C2 (Fig. 2a). This suggests that the method is relatively robust. The fi t parameters (logK', a0, and a1) are given in Table 1, together with those reported in Berry and O’Neill (2004) for compositions 1–5 obtained from fi ts to up to 16 points (italics in Table 1). Cr2+/ΣCr values calculated for each glass using Equation 1 are also given in Table 1. The four-point fi ts result in larger logK' values, but the difference is approximately constant and relatively small (0.045 ± 0.015). Thus, although there may be a small error in the accuracy of the logK values determined in this way, we do not expect signifi cant errors in the relative precision. The data were also fi t by fi xing a0 = 0.029 and evaluating a1 as (1 – a0)/A1s/4s(–16) (from Eq. 1 assuming Cr2+/ΣCr = 1 at logfO2

= –16), as suggested by Berry and O’Neill (2004), to reduce the number of variables. The average difference in logK' is only 0.02 (less than the average uncertainty) suggesting that this approximation method also yields satisfactory results.

Figure 2b shows how Cr2+/ΣCr varies with logfO2 for several

compositions. Clearly, at a constant logfO2 there is signifi cant

variability in Cr2+/ΣCr arising from the different compositions of each glass, e.g., at logfO2

= –7, Cr2+/ΣCr ranges from 0.23 to 0.70. The effect of melt composition on stabilizing one oxidation state relative to another for various elements, in particular Fe2+/Fe3+, is well known (e.g., Morris and Haskin 1974; Kilinc et al. 1983; Dickenson and Hess 1986; Schreiber et al. 1994; Jayasuriya et al. 2004). In general terms, more basic compositions (less polymer-ized due to the presence of network modifi ers such as Ca, Mg, Na, and K) stabilize higher oxidation states (Schreiber et al. 1994). This is attributed to the increase in the average negative charge

0.0

0.5A

1s/

4s

a

0.0

0.5

1.0

Cr2

+/

Cr

IW NNO b

0.0

0.5

1.0

-15 -10 -5 0

Cr2

+/

Cr

logfO2

IW NNO c

FIGURE 2. (a) The area of the 1s → 4s peak in the XANES derivative spectrum (A1s/4s) at logfO2

values of –16, –9, –7, and –2 for compositions 18 (solid symbols) and 17 (open symbols), and fi ts to the data using Equation 6. For composition 17 the sample at logfO2

= –7 contains Cr-spinel. Errors are less than the size of the symbols. (b) Cr2+/ΣCr, determined from A1s/4s using Equation 1 and the constants in Table 1, as a function of logfO2

for compositions, in order of increasing Cr2+/ΣCr at constant logfO2

, 24, 12, 13, 20, 11, and 27. The solid lines are fi ts to the data points shown. Error bars are comparable to those in c, but are not shown for clarity. (c) Cr2+/ΣCr values for MORB determined from spectra recorded in situ at 1400 °C. The dashed lines correspond to values of the iron-wüstite (IW) and nickel-nickel oxide (NNO) fO2

buffers at 1400 °C.

TABLE 2. The area of the 1s → 4s derivative peak (A1s/4s), and calcu-lated Cr2+/ΣCr values, determined from Cr K-edge XANES spectra of MORB recorded at 1400 °C and various values of logfO2; σ(A1s/4s = ±0.005); σ(Cr2+/ΣCr) = ±0.03

logfO2 A1s/4s Cr2+/ΣCr–6 0.194 0.452–7 0.251 0.587–8 0.313 0.733–9 0.346 0.812–10 0.381 0.895


on each oxygen arising from largely ionic bonding between O and network modifi ers (Ca, Mg, etc.) and more covalent bond-ing between O and network formers (e.g., Si, Al). The ionically bonded O atoms are better able to donate an electron pair (Lewis base), or electron density, and hence stabilize a highly charged cation. When electron donation to a cation exceeds that which can be accommodated, the cation either oxidizes to increase its charge (by reaction with the fO2

buffer) or decreases its coordina-tion number to reduce the number of electron donors (Baucke and Duffy 1991). The general concept has been described and quantifi ed as optical basicity, in which the amount of negative charge received by a probe ion (such as Tl+ or Pb2+) in a glass composition is determined by optical absorption spectroscopy (Baucke and Duffy 1991). Studies of a large number of glasses have enabled optical basicity values for component oxides to be defi ned (Duffy 1993), allowing theoretical basicity values to be calculated for each composition (given in Table 1). The optical basicity description is successful for silicate slags and strong correlations with redox ratios have been found for alkali silicate glasses (Baucke and Duffy 1991; Duffy 1993). The correlation between log[Cr2+/Cr3+] at logfO2

= –9 determined in this work and the optical basicity of the compositions is shown in Figure 3. The points for those compositions for which the fi ts (and hence evaluation of Cr2+/ΣCr) were compromised by the presence of Cr3C2 deviate from the trend and were omitted.

Figure 3 indicates that an increase in the basicity of the melt stabilizes the higher oxidation state (Cr3+). Such linear relationships have been reported for several elements (Duffy 1993). There is a similarly strong correlation between logK' and optical basicity (expected from the relationship between logK' and Cr2+/ΣCr; Eq. 5). The data suggests that the optical basicity model provides a fi rst order explanation for the stabilization of Cr3+ relative to Cr2+, at constant temperature and fO2

, in melts. Other parameterizations of the melt, including NBO/T and Φ [related to [O2–] and proportional to (Al+Si)/O; Lauer and Morris 1977], do not correlate as strongly with log[Cr2+/Cr3+]. The basicity trend is also consistent with the relative stabili-ties of Cr3+ and Cr2+ reported previously for two compositions (Schreiber and Haskin 1976). Furthermore, a logK' value of 1.6 was quoted by Schreiber (1987) for a forsterite-anorthite-silica composition at 1500 °C. For the calculated optical basicity of this glass (0.585) we would predict a logK' of 1.8 at 1400 °C. The two values are in good agreement noting the difference in temperature (an increase in temperature produces an increase in Cr2+/ΣCr and a corresponding decrease in logK'; Schreiber and Haskin 1976).

The melt structure, which is defi ned by the melt composition (at constant temperature and pressure) and hence linked to the ba-sicity through the degree of polymerization, also infl uences redox equilibria due to its ability to provide the preferred solvation sites of each oxidation state. Cr3+ strongly favors regular octahedral coordination, whereas Cr2+ typically occurs in a tetragonally distorted octahedral or square planar geometry (e.g., Belsky et al. 1984; Miletich et al. 1999), although there is evidence to suggest that it may be tetrahedrally coordinated in CMAS melts (O’Neill and Berry 2006). This distorted or low-symmetry environment allows the 1s → 4s transition characteristic of Cr2+ in a glass to gain intensity. The intensity of the transition when

Cr2+/ΣCr = 1 [i.e., at A1s/4s(max)] was investigated as a function of the concentration of the component oxides. The correlation between A1s/4s(max) and XCaO is shown in Figure 4a. There is no correlation with MgO or Al2O3 and only a vague positive trend with SiO2 (which probably mirrors the effect of changing CaO). The correlation in Figure 4a indicates that at low XCaO contents the Cr2+ site is more highly distorted than at high XCaO. Given that a distorted site favors Cr2+ relative to Cr3+, we would expect large values of A1s/4s(max) to be associated with compositions that stabilize Cr2+. This should be refl ected by a strong correlation between A1s/4s(max) and γ*CrO, values of which have been deter-mined by O’Neill and Berry (2006). However, no correlation exists, indicating that the Cr2+ site distortion is a property of the quenched glass rather than the melt. Thus, although quenching preserves the high-temperature oxidation state ratio (in the ab-sence of other redox variable species), it does not maintain the melt coordination environment. Such changes in melt structure with temperature are not unexpected. The internal consistency of the quantifi cation method indicates that the quench rates and thus glass-transition temperatures for our samples must be suf-fi ciently uniform that the same Cr2+ average site occurs in each glass of a given composition. If the site were to vary between samples, then A1s/4s would comprise contributions from differ-ences in the site, in addition to variations in Cr2+/ΣCr, resulting in scatter from the relatively smooth dependence of A1s/4s on fO2

observed here and in Berry and O’Neill (2004).

The correlation between logγ#Cr3+O1.5

and XCaO is shown in Figure 4b. There is a general trend indicating stabilization of Cr3+ with increasing XCaO. This fi nding is consistent with CaO-rich melts containing abundant octahedral sites and the strong pref-erence of Cr3+ for octahedral coordination; Si, Al, and Mg may all be tetrahedrally coordinated. The “bulge,” or deviation from the linear trend with XCaO, corresponds to those compositions rich in Mg, with the magnitude of this deviation being linearly correlated (r = 0.91) with XMgO. This fi gure, therefore, suggests that two processes are contributing to the stabilization of Cr3+ in the melt: the availability of a suitable solvation site (related to the XCaO), and the optical basicity, which for a given XCaO,

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

0.52 0.56 0.60 0.64 0.68

log[

Cr2

+/C

r3+]

Optical Basicity

FIGURE 3. Correlation between log[Cr2+/Cr3+] determined at logfO2 =

–9 and the optical basicity of each glass. The solid symbols are for the Fe-free compositions and the × is for MORB.


will increase with increasing XMgO. The large value and range of logγ#

CrO1.5 compared to logγ*CrO, and the lack of a correlation

between γ*CrO and any melt compositional or structural parameter (O’Neill and Berry 2006), indicates that logγ#

CrO1.5 —or the capac-

ity of the melt to accommodate both the charge and preferred coordination environment of Cr3+—determines the variability in logK', and hence Cr2+/ΣCr as a function of composition.

Cr2+ is not found in “natural” Fe-bearing glasses quenched from terrestrial fO2

values, but does occur in the melt (Berry et al. 2003). Cr2+ is removed on cooling by the electron exchange reaction Cr2+ + Fe3+ → Cr3+ + Fe2+. Thus, Cr2+/ΣCr of Fe-bearing compositions can only be determined by in situ experiments on the melt. The Cr2+/ΣCr values obtained from spectra of MORB recorded at 1400 °C are shown in Figure 2c and vary from 0.45 at NNO to almost 0.90 at IW. This is the fi rst time that Cr2+/ΣCr has been quantifi ed for an Fe-bearing melt and indicates that Cr2+ is a major oxidation state in basaltic magmas. The fact that the high-temperature data can be described by Equation 6 suggests that there is no signifi cant change in Cr2+/ΣCr from that expected by Equation 5 and that any interactions between Cr2+/Cr3+ and Fe2+/Fe3+ due to electron exchange must be thermodynamically insignifi cant at 1400 °C. Any effect would presumably appear as non-ideal behavior, however, no deviation from the theoretical value of the fO2

exponent (1/4) was observed when this term was allowed to vary during the fi t.

The A1s/4s(max) value calculated for MORB is plotted in Figure 4a and is inconsistent with the trend found for the Fe-free com-positions. This is not unexpected as A1s/4s quantifi es the transition intensity, which will vary with changes in Cr2+ coordination between the melt and glass. Thus values for spectra recorded at room temperature and 1400 °C will not be comparable. How-ever, the logK' and log[Cr2+/Cr3+] values determined for MORB are consistent with trends defi ned by the Fe-free glasses, such as the optical basicity (Fig. 3). This observation suggests that Cr2+/ΣCr for Fe-bearing compositions can be predicted from the composition of the melt. Due to experimental limitations none of the compositions studied contain Na2O, K2O, or H2O, all of which will affect the melt structure and basicity. In particular, Na2O and K2O are more “basic” than CaO and may have a strong destabilizing effect on Cr2+, although the relative abundance of these elements suggests that their contribution may be limited.

The XANES method for quantifying Cr2+/ΣCr described here has allowed the compositional dependence of Cr2+/Cr3+ in Fe-free glasses to be determined. The results are in good agreement with the optical basicity description of how melts stabilize oxidation states. The apparent success of this method for quenched glasses has prompted its application to the in situ study of melts, which has allowed Cr2+/ΣCr in an Fe-bearing melt to be quantifi ed for the fi rst time. Cr2+ is an important and possibly dominant oxida-tion state in basaltic volcanism at terrestrial fO2

values.

ACKNOWLEDGMENTSPart of this work was performed at the Australian National Beamline Facility

with support from the Australian Synchrotron Research Program, which is funded by the Commonwealth of Australia under the Major National Research Facilities Program. We thank the Australian Research Council for fi nancial support, N.G. Ware for electron microprobe analyses, Steve Sutton for a constructive review, and Mike Toplis for his editorial handling.

REFERENCES CITEDBarnes, S.J. (1986) The distribution of chromium among orthopyroxene, spinel,

and silicate liquid at atmospheric pressure. Geochimica et Cosmochimica Acta, 50, 1889–1909.

Baucke, F.G.K. and Duffy, J.A. (1991) The effect of basicity on redox equilibria in molten glasses. Physics and Chemistry of Glasses, 32, 211–218.

Belsky, H.L., Rossman, G.R., Prewitt, C.T., and Gasparik, T. (1984) Crystal structure and optical spectroscopy (300 to 2200 nm) of CaCrSi4O10. American Miner-alogist, 69, 771–776.

Berry, A.J. and O’Neill, H.St.C. (2004) A XANES determination of the oxidation state of chromium in silicate glasses. American Mineralogist, 89, 790–798.

Berry, A.J., Shelley, J.M.G., Foran, G.J., O’Neill, H.St.C., and Scott, D.R. (2003) A furnace design for XANES spectroscopy of silicate melts under controlled oxygen fugacities and temperatures to 1773 K. Journal of Synchrotron Radiation, 10, 332–336.

Burns, V.M. and Burns, R.G. (1975) Mineralogy of chromium. Geochimica et Cosmochimica Acta, 39, 903–910.

Dickenson, M.P. and Hess, P.C. (1986) The structural role and homogeneous redox equilibria of iron in peraluminous, metaluminous, and peralkaline silicate melts. Contributions to Mineralogy and Petrology, 92, 207–217.

Duffy, J.A. (1993) A review of optical basicity and its applications to oxidic systems. Geochimica et Cosmochimica Acta, 57, 3961–3970.

Hanson, B. and Jones, J.H. (1998) The systematics of Cr3+ and Cr2+ partitioning between olivine and liquid in the presence of spinel. American Mineralo-gist, 83, 669–684.

Jayasuriya, K.D., O’Neill, H.St.C., Berry, A.J., and Campbell, S.J. (2004) A Mössbauer study of the oxidation state of Fe in silicate melts. American Mineralogist, 89, 1597–1609.

Kilinc, A., Carmichael, I.S.E., Rivers, M.L., and Sack, R.O. (1983) The ferric-ferrous ratio of natural silicate liquids equilibrated in air. Contributions to Mineralogy and Petrology, 83, 136–140.

Lauer, H.V. and Morris, R.V. (1977) Redox equilibria of multivalent ions in silicate glasses. Journal of the American Ceramic Society, 60, 443–451.

Li, J.-P., O’Neill, H.St.C., and Seifert, F. (1995) Subsolidus phase relations in the system MgO-SiO2-Cr-O in equilibrium with metallic Cr, and their signifi cance

0.36

0.40

0.44

0.48

0.52A

1s/

4s

(ma

x)

a

-2.2

-2.0

-1.8

-1.6

-1.4

-1.2

-1.0

0.0 0.1 0.2 0.3 0.4 0.5

log

#C

rO1.

5

b

XCaO

FIGURE 4. Correlations between XCaO and (a) A1s/4s(max) (i.e., when Cr2+/ΣCr = 1) and (b) logγ#

CrO1.5, for each glass composition (solid symbols).

The × in a corresponds to MORB at 1400 °C.


for the petrochemistry of chromium. Journal of Petrology, 36, 107–132. Miletich, R., Nowak, M., Seifert, F., Angel, R.J., and Brandstätter, G. (1999) High-

pressure crystal chemistry of chromous orthosilicate, Cr2SiO4. A single-crystal X-ray diffraction and electronic absorption spectroscopy study. Physics and Chemistry of Minerals, 26, 446–459.

Morris, R.V. and Haskin, L.A. (1974) EPR measurement of the effect of glass composition on the oxidation states of europium. Geochimica et Cosmochimica Acta, 38, 1435–1445.

O’Neill, H.St.C. and Berry, A.J. (2006) Activity coeffi cients at low dilution of CrO, NiO, and CoO in silicate melts in the system CaO-MgO-Al2O3-SiO2 at 1400 °C: using the thermodynamic behavior of transition metal oxides in silicate melts to probe their structure. Chemical Geology, 231, 77–89.

O’Neill, H.St.C. and Eggins, S.M. (2002) The effect of melt composition on trace element partitioning: an experimental investigation of the activity coeffi cients of FeO, NiO, CoO, MoO2, and MoO3 in silicate melts. Chemical Geology, 186, 151–181.

O’Neill, H.St.C. and Mavrogenes, J.A. (2002) The sulfi de capacity and the sulfur content at sulfi de saturation of silicate melts at 1400 °C and 1 bar. Journal of Petrology, 43, 1049–1087.

Poustovetov, A.A. and Roeder, P.L. (2000) The distribution of Cr between basaltic melt and chromium spinel as an oxygen geobarometer. Canadian Mineralo-gist, 39, 309–317.

Schreiber, H.D. (1987) An electrochemical series of redox couples in silicate melts: a review and applications to geochemistry. Journal of Geophysical

Research, 92, 9225–9232.Schreiber, H.D. and Haskin, L.A. (1976) Chromium in basalts: experimental de-

termination of redox states and partitioning among synthetic silicate phases. Proceedings of the 7th Lunar Science Conference, p. 1221–1259. Pergamon Press, Inc., New York.

Schreiber, H.D., Merkel, R.C., Jr., Schreiber, V.L., and Balazs, G.B. (1987) Mutual interactions of redox couples via electron exchange in silicate melts: models for geochemical melt systems. Journal of Geophysical Research, 92, 9233–9245.

Schreiber, H.D., Kochanowski, B.K., Schreiber, C.W., Morgan, A.B., Coolbaugh, M.T., and Dunlap, T.G. (1994) Compositional dependence of redox equilibria in sodium silicate glasses. Journal of Non-Crystalline Solids, 177, 340–346.

Sutton, S.R., Jones, K.W., Gordon, B., Rivers, M.L., Bajt, S., and Smith, J.V. (1993) Reduced chromium in olivine grains from lunar basalt 15555: X-ray absorp-tion near edge structure (XANES). Geochimica et Cosmochimica Acta, 57, 461–468.

MANUSCRIPT RECEIVED SEPTEMBER 20, 2005MANUSCRIPT ACCEPTED APRIL 20, 2006MANUSCRIPT HANDLED BY MICHAEL TOPLIS

Documents

The effect of composition on Cr2+/Cr3+ in silicate melts