Journal of Petrology Advance Access originally published online on November 13, 2006
Journal of Petrology 2007 48(2):219-230; doi:10.1093/petrology/egl054
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Ferric Iron in CaTiO3 Perovskite as an Oxygen Barometer for Kimberlitic Magmas I: Experimental Calibration
School of Earth and Ocean Sciences, University of Victoria, 3800 Finnerty Road, Victoria, BC, V8W 3P6, Canada
RECEIVED OCTOBER 27, 2005; ACCEPTED AUGUST 30, 2006
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONCALIBRATION OF AN OXYGEN...SUPPLEMENTARY DATAREFERENCES |
|---|
A method to estimate the oxygen fugacity (fO2) during the crystallization of kimberlites is developed using the Fe content of CaTiO3 perovskite (Pv), a common groundmass phase in these rocks. With increasing fO2, more Fe exists in the kimberlitic liquid as Fe3+, and thus partitions into Pv. Experiments to study the partitioning of Fe between Pv and kimberlite liquid were conducted at 100 kPa on simple and complex anhydrous kimberlite bulk compositions from 1130 to 1300°C over a range of fO2 from NNO – 5 to NNO + 4 (where NNO is the nickel–nickel oxide buffer), and at Nb and rare earth element (REE) contents in the starting materials of 0–5 wt % and 1500 ppm, respectively. The partitioning of Fe between Pv and kimberlite liquid is influenced mostly by fO2, although the presence of Nb increases the partition of Fe3+ into perovskite at a given T and fO2. Multiple linear regression (MLR) of all the experimental data produces a relationship that describes the variation of Fe and Nb in Pv with fO2 relative to the NNO buffer:
|
|
, and Nb and Fe as cations per three oxygens). Over the range of conditions of our experiments, this relationship shows no temperature (T) dependence, is not affected by the bulk Fe content of the kimberlite starting material and reproduces experimental data to within 1 log fO2 unit. KEY WORDS: kimberlites; oxygen fugacity; perovskite; ferric iron; magma
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCALIBRATION OF AN OXYGEN...SUPPLEMENTARY DATAREFERENCES |
|---|
Iron exists in terrestrial magmas as Fe2+ and Fe3+, with their relative proportions depending on the oxygen fugacity (fO2). Kennedy (1948
) first recognized the utility of Fe3+/Fe2+ as a record of fO2 in magmas, and as a general petrogenetic indicator. Fudali (1965) further documented how Fe3+/Fe2+ increases with acidity in igneous rocks and emphasized how fO2 controls igneous differentiation, by dictating how iron partitions between the liquidus minerals and coexisting liquid.
The Fe3+/Fe2+ in melts is governed by the redox reaction
|
| (1) |
; Thornber et al., 1980; Kress & Carmichael, 1988). Application of these expressions permits a measure of the fO2 along the entire liquid line of descent of magmas preserving glass (Carmichael, 1991; Haggerty, 1994). Kimberlites are unique rocks whose composition and intensive variables (e.g. T, fO2) elucidate the conditions that occur in their mantle source region, or during their ascent to the surface. Kimberlites often carry diamond or precipitate carbonate, and the stability of these minerals is directly influenced by the fO2 of the magma during its ascent. As the main primary source of diamonds, the fO2 of kimberlites may thereby determine the likely presence or quality of diamonds in this magma.
Unfortunately, kimberlites are complex rocks and never preserve glass, so that the empirical expressions relating Fe3+/Fe2+ of glass, T, fO2 and composition cannot be applied. Previous estimates of the fO2 of kimberlite have been mainly deduced by theoretical calculation, and remain rigorously untested by the lack of fresh material in these rocks (Mitchell, 1973, 1986; LeRoex et al., 2003). The discovery of fresh kimberlites in the Lac de Gras region (Pell, 1995) affords some constraints on their fO2. For example, Fedortchouk & Canil (2004) examined spinel inclusions in fresh olivine phenocrysts and showed that kimberlites co-precipitate these minerals between 1030°C and 1170°C at a maximum fO2 of NNO – 2 (where NNO is the nickel–nickel oxide buffer). The certainty of their fO2 estimates, however, depends on the activity of SiO2 (aSiO2) of the magma, which is only broadly constrained, as shown by the work of Mitchell (1973, 1986).
To address this issue, we develop an empirical oxygen barometer for kimberlites based on the partitioning of Fe between silicate liquid and CaTiO3 perovskite, an accessory mineral common in the groundmass of kimberlites. Mössbauer spectroscopy of synthetic perovskites (Burns, 1989; Waerenborgh et al., 2001; Mitchell, 2002) shows that Fe can be present as Fe3+, Fe2+ and even Fe4+ in the structures of synthetic CaTiO3-based perovskites, but that the latter two species occur in only minor amounts (<7%). Natural CaTiO3-based perovskites contain only Fe3+ (Muir et al., 1984; Mitchell et al., 1998). We follow the summary of Mitchell (2002) and assume that the principal valence state and substitution is as Fe3+. We hypothesize that with increasing fO2, more Fe exists in the kimberlitic liquid as Fe3+, and thus partitions more easily into perovskite. Ultimately, our assumption of only Fe3+ in perovskite does not affect our empirical calibration, as we measure and model only bulk Fe (as Fe2O3) in perovskite from our experiments.
Some kimberlite perovskites can contain significant amounts of Nb2O5 and rare earth elements (REE) (up to 15 wt %), with some compositions showing correlations between Fe3+ and Nb or REE (Chakhmouradian & Mitchell, 2000). This suggests the potential for coupled substitutions at high concentrations of Nb and/or REE in perovskites, potentially complicating a simple empirical oxybarometer derived from the Fe3+ content alone. We compiled literature data (n = 131), which show that 85% of kimberlite perovskites contain <1 wt % Nb2O5 and that no relationship between Fe3+ and Nb exists at these concentrations. In the same database, 95% of the perovskites contain <5 wt % REE and there is no clear relationship with Fe3+. None the less, to be certain, and to make our study more applicable to natural kimberlites containing Nb and REE, we also investigated the effects of these elements on the partitioning of Fe3+ into perovskite.
Using our experimental data we examine the effects of composition, T and fO2 on the partitioning of Fe between perovskite and kimberlite liquid and use the experimental data to devise an empirical oxybarometer. In a companion paper (Canil & Bellis, in press) we apply this oxybarometer to examine the fO2 during the crystallization of perovskite in some well-characterized natural kimberlites.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCALIBRATION OF AN OXYGEN...SUPPLEMENTARY DATAREFERENCES |
|---|
Starting materials
The first challenge was to establish starting compositions for our experiments that are representative of the wide compositional variation in kimberlites. The abundance of xenoliths that contaminate almost all kimberlites makes identification of a primary magma difficult. Examples of compositions that approximate primary magma have been described for the Mayeng (Apter et al., 1984
), Benfontein (Mitchell, 1997), Koidu (Taylor et al., 1994), and Aries kimberlites (Edwards et al., 1992). It has been argued that the scarcity of xenoliths and xenocrysts, the low SiO2 content and the fine-grained nature of the aphanitic kimberlite at Wesselton, South Africa indicate that it is a primary magma composition (Shee, 1986; Edgar et al. 1988). Price et al. (2000) contended that the aphanitic margins of thin, hypabyssal kimberlite dykes from the Jericho kimberlite are primary magmas, based on the lack of macrocrysts, the fine-grained and homogeneous nature of the groundmass and its bulk composition. Thus, we based the starting compositions for our experiments on both the Wesselton and Jericho compositions, which are richer in Ca and poorer in Mg than most kimberlites (Fig. 1). The Jericho compo sition has a higher CaO and lower MgO and TiO2 content relative to that of Wesselton.
|
Natural kimberlites contain between 5 and 20 wt % H2O and/or CO2 (see caption to Fig. 1 for references). Because the primary objective of our work was to study the effect of fO2 and T on the partitioning of Fe between perovskite and kimberlite liquid, all experiments were conducted at 100 kPa in a gas-mixing furnace where fO2 is controlled precisely. Under these conditions H2O and/or CO2 cannot be maintained in solution, and, as a result, these volatiles were excluded from the starting compositions. The Wesselton and Jericho kimberlite analyses were re-calculated on a volatile-free basis and synthetic kimberlite starting compositions were then constructed from these modified base compositions. The lack of H2O will increase the crystallization T relative to the natural hydrous case, but we show below that this does not affect the phase assemblages.
Kimberlites have high MgO contents in part as a result of accumulation of mantle olivine xenocrysts, estimated to be of the order of 
40 wt % in some of the Lac de Gras kimberlites (Fedortchouk & Canil, 2004). To increase the amount of liquid present in the experiments, we chose to remove the xenolithic olivine component from our synthetic starting compositions and thereby reduce the amount of olivine in run products. We subtracted 60 wt % and 50 wt % mantle olivine component (Fo90) from the Wesselton and Jericho base compositions, respectively, to reduce their MgO concentrations (Fig. 1), and therefore crystallization T, and increase the quantity of melt present at lower T. We also reduced the amount of CaCO3 component in the Jericho composition by the removal of all CaO assigned to its CO2 content, because as noted above, the experiments were by design without CO2.
The resultant kimberlite starting compositions (Table 1) were synthesized from reagent grade oxides SiO2, TiO2, Al2O3, Fe2O3, MnO, MgO, CaCO3, Na2CO3, K2CO3 and P2O5 according to the normalized anhydrous compositions of the Wesselton and Jericho kimberlites. In addition, simpler analogs (WC1, WC2, WC3, JC2) were created by mixing fewer component oxides (SiO2, TiO2, Fe2O3, MgO and CaCO3). Kimberlites also contain high concentrations of rare earth elements (REE), Sr and Nb, which are known to substitute up to 10 mol% as CeFeO3, CeNbO3, Th0·5TiO3, SrTiO3 and NaNbO3 components in perovskite (Nickel & McAdam, 1963; Carmichael, 1967; Smith, 1970; Mitchell, 1972; Chakhmouradian & Mitchell, 2000). Two of the simpler system compositions (WC2 and WC4) were doped with 300 ppm Ce and a ‘cocktail’ of 1000 ppm of each of Ce, La, Sr, Nb, respectively—concentrations similar to those found in natural kimberlites (Chakhmouradian & Mitchell, 2000; Eccles et al., 2004). The latter trace elements were added to the oxide mixtures as standard element solutions and dried before grinding and decarbonation. Finally, subsets of the simple system composition (WC) were doped with 0·5 wt %, 2 and 5 wt % reagent grade Nb2O5 to examine the effects of Nb on Fe partitioning.
|
To further enhance the amount of liquid present in experiments at low T (1130°C), we followed the approach of Toplis & Carrol (1995
), and synthesized a second set of kimberlite analogs (WC3, WC4, WD3 and JC2), based on the composition of the residual melt formed after 85% crystallization in experiments on the first set of analogs (WC1, WD2, JC2). The use of these ‘lower T’ analogs allowed for easier phase description and identification along the entire liquid line of descent of our synthetic kimberlites from above the liquidus to as much as 98% crystallization.
Reagent grade oxides and carbonates were dried for 24 h and then weighed and mixed with a mortar and pestle. P2O5 is hydroscopic and was therefore weighed and added shortly before mixing. Each mixture was ground under acetone in an agate mortar, decarbonated in a Pt crucible in a box furnace for 24 h at 950°C, and then fused for 24 h at 1525°C and poured onto a metal plate. These compositions all quenched to glass that was then crushed and ground under acetone in an agate mortar to a grain size of less than 100 µm. Starting glass compositions were analyzed by electron microprobe for major and minor elements and by laser ablation–inductively coupled plasma mass spectrometry (LA-ICPMS) for trace elements (Table 1). Analytical methods are discussed below.
Experiments
Experiments were performed in a Deltech DT-31 vertical tube furnace over a range of fO2 from five log units below the NNO buffer to four units above this buffer at T of 1130–1300°C (Table 2). For each experiment, sample powders were mixed with acetone to form a slurry, then loaded onto loops of 0·15 mm diameter Pt wire and were sintered using a torch. Iron loss to Pt loops was avoided by running each experiment in duplicate, with pre-saturation of the Pt loop under the desired T and fO2 conditions for a minimum of 24 h. When pre-saturation was complete, the glass bead was dissolved in HF and a new slurry was sintered to the Fe pre-saturated Pt loop and run at identical T–fO2 conditions for the final experiment. Temperature was measured with a Pt–Pt90Rh10 thermocouple calibrated at the melting point of gold. Oxygen fugacity was controlled by CO–CO2 gas mixing at a total flow rate of 200 cm3/min and measured in each experiment with a solid zirconia electrolyte cell. Fluctuations in the e.m.f. were less than ±5 mV during each experiment, which corresponds to ±0·05 log fO2 units. The zirconia cell was not calibrated but resulting fO2 values were within 5% of the values calculated by JANAF thermochemical tables. The experiments with the highest fO2 were performed in pure CO2 gas. Run durations varied from 24 h at 1300°C to 72 h at 1130°C.
|
For each experiment, the samples were introduced to the vertical tube furnace and the desired fO2 was achieved after
5 min, once all air was purged from the tube by the CO–CO2 gas mixture. For six experiments, samples were introduced to the furnace at a lower fO2 for 5 min and then raised to a higher fO2. In some experiments, samples were introduced to the furnace above the run T, held there for 10 min and cooled slowly (10°C/h) to the run T, to grow larger perovskite crystals (Table 2). At the end of each experiment, samples were removed from the furnace and quenched to a glass within seconds in a stream of air. Run products were crushed, mounted in oil and examined under transmitted light. The remaining chips were mounted in epoxy, and polished.
Analytical methods
Coexisting phases were examined qualitatively by backscatter SEM (Hitachi S-3500N Scanning Electron Microscope) at the University of Victoria using a 15 kV, 20 nA primary beam. Back-scattered electron (BSE) images of run products were acquired with a Philips XL30 electron microscope at the University of British Columbia. Electron microprobe analyses of run products were carried out with a CAMECA SX50 electron microprobe at the University of British Columbia. Major and minor elements were determined at 15·0 kV acceleration voltage and a beam current of 20·1 nA on the Faraday cup with a 1 µm beam. Analytical conditions were 20 s counting time on peaks and 10 s on backgrounds for all major elements except for Fe, Ce and La (60 s), Sr (120 s) and Nb (160 s). Standards were albite (Na), olivine (Mg, Si), orthoclase (Al, K), apatite (P), diopside (Ca), rutile (Ti), synthetic rhodonite (Mn) and fayalite (Fe). The synthetic glasses of Drake & Weill (1972) were used as standards for Ce, La, Sr and Nb in glass and perovskite phases. Between five and 20 analysis points were measured for phases in each run product and were averaged (Electronic Appendix 1; available at http://www.petrology.oxfordjournals.org). Data reduction was done with the ‘PAP’ 
(
Z) method (Pouchou & Pichoir, 1985).
LA-ICPMS analyses were carried out at the University of Victoria, using a VG Elemental PQ II S + ICP-MS and a Merchantek solid-state, frequency quadrupled 266 nm Nd:YAG UV laser ablation system in a He atmosphere. Trace elements in starting glasses were determined using a laser spot size of 50 µm at a frequency of 10 Hz with an energy of 1·8 mJ. 43Ca was used as the internal standard for NIST 613 and starting glasses. Each block of analyses consisted of measuring the NIST glass twice followed by fewer than 10 unknowns then remeasuring the NIST glass twice. Data were recorded as time-resolved spectra collected over 90 s with 30 s allotted to collecting backgrounds. The individual measurements were normalized and reduced to concentrations using PlasmaLab© (Isomass). Accuracy and precision on standard glasses are better than 10% (Chen et al., 2000; Canil et al., 2003).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCALIBRATION OF AN OXYGEN...SUPPLEMENTARY DATAREFERENCES |
|---|
Phase equilibria
Phase proportions were calculated by mass balance of compositional data for all phases and the bulk composition. All experiments produced a mass balance for all elements and phases totaling to within 1% of the bulk composition. Glass is present in all experiments. Olivine (<200 µm) and spinel (10–50 µm) saturate in the higher MgO starting materials (WC1, WC2, WD2), followed at decreasing T by monticellite (<200 µm), perovskite (10–50 µm) (Fig. 2) and in some cases melilite group minerals (melilite, åkermanite, gehlenite) (10–50 µm). Pyroxene (<200 µm), ulvöspinel (10–50 µm) and kirschteinite (<200 µm) appear in lower MgO compositions (WC3, WC4, WC5, WD3, JC2).
|
The details of phase equilibria and composition in all starting compositions are complex and for the sake of brevity are not described here. The only compositions relevant to this study are those saturated in perovskite (Electronic Appendix 1). Detailed investigation of the two simple system compositions (WC, JC) over a large T interval (1130–1295°C) produced varying amounts of perovskite and up to
75% melt. Experiments below 1130°C were not attempted, as the small amount of melt present would have precluded high-quality glass analyses. The solidus was not determined as part of this study. In addition to the phases listed in Table 2, minute Fe-oxide grains were observed in many run products but were too small to be analyzed quantitatively by electron microprobe (Fig. 2). The WC composition saturates in perovskite at 1300°C, followed by the WD composition at 1180°C and the JC composition at 1130°C. Perovskite is subhedral to anhedral and ranges in size from 10 to 50 µm depending on cooling history. No zoning was encountered in perovskites from experimental runs (Fig. 2). Abundances of perovskite range from 4 to 16%, depending on T and bulk composition.
Perovskite and liquid compositions
With changing fO2, perovskites show variance in their Fe content as well as in Nb, Sr and REE in strongly doped compositions (Electronic Appendix 1). Using the equation of Kress & Carmichael (1988) we calculate a large change in Fe3+/
Fe of the liquid saturated in perovskite above NNO – 3·5, whereas below this fO2 there is little change in this ratio (Fig. 3). With the exception of a few experiments on ultrabasic compositions (katungite and madupite), the liquids used to calibrate the Kress & Carmichael (1988) equation are outside the composition range of liquids in this study. None the less, we use their calculation here only as a guide, and to show that less variability in the Fe2O3 content of perovskite is expected at conditions below NNO – 3·5, because at this fO2 there is probably very little Fe3+ available in the liquid.
|
The Wesselton-based (‘W’) series of compositions that saturate in perovskite were investigated at variable fO2 and constant T (1130°C or 1255°C) and at constant fO2 (NNO – 2) and variable T (Figs 4–6
). A positive correlation was found between fO2 and the Fe2O3 content of perovskites in the Wesselton-based compositions (WC, WC2—Fig. 4a) and T variations of 200°C have no observable effect on the Fe2O3 content of perovskites in the WC or WD starting compositions (Fig. 6a). The bulk Fe content of the starting material also does not affect the amount of Fe partitioned into Pv at a given T and fO2 (Fig. 7). Increasing amounts of Nb in the starting material result in an increase in the amount of Fe in perovskite at any particular fO2 condition, although the effect of fO2 on Fe in Pv remains similar for a given level of Nb in the starting material (Fig. 8).
|
NNO) after Frost (1991
|
|
|
|
Equilibrium and reversals
Time series experiments conducted on the WC compositions (Fig. 9) demonstrate no change in perovskite composition over run times varying from 24 to 96 h, suggesting that equilibrium was attained in 24 h and that no Fe loss to the platinum loop occurred. In five of six experiments, the Fe3+ content of perovskites from experiments that approached run conditions from more reducing conditions are similar to those that approached equilibrium from higher fO2 (shown by arrows in Fig. 4b). We have no explanation for the one experiment that has a lower Fe3+ content in perovskite than expected when the fO2 was increased to the final run conditions (Fig. 4b). Experiments run with different thermal histories, starting either at the run T or cooling slowly to the final run T, produced similar results, although crystals were larger with slow cooling.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCALIBRATION OF AN OXYGEN...SUPPLEMENTARY DATAREFERENCES |
|---|
Perovskite saturation
With the possible exception of ulvöspinel, fO2 is not a strong variable in controlling phase appearance in our experiments. The appearance of perovskite is controlled primarily by T and liquid composition. Liquids must have sufficient Ti and the appropriate aSiO2 for perovskite to crystallize. For example, WC with 7·8 wt % TiO2 saturates in perovskite at 1295°C, followed by WD with 6·1 wt % TiO2 at 1180°C and finally JC with 2·2 wt % TiO2 at 1130°C (Table 2). The bulk Fe content of the starting materials varies from 11· 6 to 21 wt % FeO* (Table 1) and has no influence on the saturation T of perovskite. There is a slight increase in the saturation T of perovskite with increased Nb content of the liquid.
Substitution mechanism for Fe3+ in perovskite
The correlation of Ti and Fe3+ cations in perovskite from undoped starting compositions shows that the latter substitutes on the octahedral B site (ABO3) (Fig. 10 inset). The substitution of Fe3+ for Ti4+ in perovskite, however, requires a coupled substitution for charge balance (e.g. Ca2+Ti4+ 
Me3+Fe3+, where Me is a metal cation). In the five experiments doped with Ce or La, no correlation was found between Fe3+ and Ce3+ and La3+. Furthermore, the latter cations are in too low a concentration in the perovskites from those experiments (when compared with natural kimberlites) to satisfy the coupled substitution.
|
A plot of Nb and Fe cations plotted against Ti cations for all experiments doped with Nb shows a 1:1 correlation across a range of fO2 from NNO – 3 to NNO + 3·5 (Fig. 10). For perovskites in our experiments, Nb, Fe and Ti can be the only cations that occupy the octahedral (B) site. Data plotted in Fig. 10b and c show that Fe and Nb substitute for Ti in equal amounts, that each has a 1:1 substitution ratio, and that both Fe and Nb account for all Ti substitutions (up to 20% of the site), within error. Therefore, we conclude that for Nb-doped experiments the substitution reaction
|
| (2) |
It has been suggested that oxygen vacancies in the perovskite structure increase with Fe concentration (McCammon et al., 2000). The results from this study show that Fe concentration increases with fO2. Because more Fe substitutes for Ti with increasing fO2 (Figs 10 and 11) more oxygen vacancies could exist in the perovskite structure with increasing fO2 (i.e. increasing Fe concentration). To examine further how the coupled substitution of Fe and Nb is affected by fO2, Fig. 11 shows Fe and Nb cations plotted against Ti cations for a range of fO2, with Fe cations serving as a proxy for fO2, and a limited range of Nb doping (1·5 wt %). At high fO2, the number of Fe cations substituting for Ti lies above the 1:1 substitution ratio, whereas at low fO2 the number of Fe cations lies below this ratio (Fig. 11). This relationship suggests that at high fO2 conditions, some proportion of Fe substitutes for Ti according to a vacancy mechanism:
|
| (3) |
), it is possible that defects control the substitution of Fe3+ in perovskites that do not contain significant amounts of Nb. Whether this is true requires some measurement of defect populations, and the exact ratio of all potential Fe species (Fe2+, Fe3+, Fe4+) in perovskite at the conditions of our experiments.
|
Effect of temperature and volatiles
The crystallization temperatures of perovskites in our experiments are higher than those of natural kimberlites, which are deduced to be between 700 and 850°C (Mitchell, 1986
). The higher temperatures in our experiments are attributed to the lack of volatiles in the starting materials. The crystallization temperature of perovksite in a Wesselton kimberlite containing 6·20 wt % H2O and 4·77 wt % CO2 at pressures of 1·0–4·0 GPa (Edgar et al., 1988) was extrapolated to 100 kPa to provide an estimate of the crystallization temperature for perovskite in volatile-bearing kimberlite at the pressure of our experiments (Fig. 12). The extrapolated crystallization temperature is only 100°C lower than that for the equivalent anhydrous Wesselton compositions in this study. This difference is attributed to the effect of H2O, which depresses crystallization temperature. The presence or absence of H2O, however, does not affect observed phase assemblages. Over the crystallization interval, we observe the identical phase assemblage (Ol + Mo + Px + Pv + L) in volatile-free Wesselton kimberlite (WC, WD) as was observed in the same kimberlite by Edgar et al. (1988) in the presence of H2O and CO2.
|
The partitioning of (bulk) Fe between perovskite and liquid, expressed as DFeO* Pv/Liq shows a barely discernible variation with T over
200°C (Fig. 6b). The slight change in DFeO* Pv/Liq, however, is mainly attributed to T-dependent changes (of 10% relative) in the Fe3+ content of the liquid (Kress & Carmichael, 1988), rather than in the Fe2O3 content of the perovskite, which remains constant over a significant T range (Fig. 6a). Thus, at least in our experiments, T has no observable effect on the amount of Fe that substitutes into perovskite (Fig. 6a).
Effect of fO2
For a given bulk composition, fO2 is by far the largest and most significant of all variables for the Fe2O3 content of perovskite. For all series of bulk compositions (WC, WD, JC) the Fe3+ content of perovskite (expressed as wt % Fe2O3) was found to show a strong correlation with fO2 over a range of T from 1100°C to 1300°C at fO2 above NNO – 3. At fO2 conditions below NNO – 3·5, perovskites from the WC and JC compositions show nearly constant Fe2O3 (Fig. 5a) suggesting that this relationship may break down. At these low fO2 conditions, however, the Fe3+/Fe in the liquid coexisting with perovskite shows little change (Fig. 3), and may reach the point where differences in the Fe2O3 content of the perovskites cannot be measured and appear essentially constant.
|
CALIBRATION OF AN OXYGEN BAROMETER |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCALIBRATION OF AN OXYGEN...SUPPLEMENTARY DATAREFERENCES |
|---|
Weighted least-squares regression of data from perovskites that crystallized in the Nb-free Wesselton (WC) compositions (n = 16) between 1130°C and 1300°C over a range of fO2 from NNO + 4 to NNO – 4 (Fig. 4a) gives the relationship
|
| (4) |
. Each data point in the regression was weighted by 1/, where is one standard error of the mean Fe2O3 content of perovskite. Perovskites from the WD and JC series compositions were excluded from this regression because they only precipitated at 1130°C, and at low fO2, where Fe3+/Fe in the liquid is so low (Fig. 3) as to result in a nearly constant Fe content of the perovskites (Fig. 6a). At fO2 
NNO – 3·5, equation (4) reproduces 75% of all experimental data from all Nb-free starting compositions in this study (n = 31) to within 0·5 log fO2 units and 85% of all data to within 0·7 log fO2 units. We consider these to be conservative estimates of the uncertainty in equation (4) as an oxygen barometer.
At a given fO2, Nb levels clearly have an effect on the Fe content of perovskite (Fig. 8). We used multiple linear regression to simultaneously describe the relationship of Fe, Nb and fO2 in all experimental data, defining an oxygen barometer:
|
| (5) |
.
The multiple linear regression fit (5) reproduces the majority of data over six orders of magnitude in fO2 to within one log fO2 unit (Fig. 13). Most of the outliers from equation (5) are at low T (1130°C) and where Fe3+/Fe in the liquid is low. Because of the range of fO2 and Nb contents over which the multiple linear regression method (5) agrees with observed data it is recommended as the preferred oxybarometer. Because the majority of kimberlite perovskites do not contain more that 1 wt % Nb2O5, the simpler, Nb-free oxybarometer (4) could also be suitable in such cases. The divergence of calculated from observed values occurs mostly at the extremes in fO2 of the experimental calibration, which could potentially be attributed to the presence of Fe2+ or Fe4+ in perovskites at very low and high fO2, respectively. More experimental work on the crystal chemistry and valence state of Fe in CaTiO3 perovskites is required at these conditions to address this issue. For this reason, we caution that the oxybarometer is likely to be suspect when fO2 is high (i.e. near air) or low (near or below the stability of Fe metal), but such conditions are probably never realized in any terrestrial magma (Carmichael, 1991).
|
|
SUPPLEMENTARY DATA |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCALIBRATION OF AN OXYGEN...SUPPLEMENTARY DATAREFERENCES |
|---|
Supplementary data for this paper are available at Journal of Petrology online.
|
ACKNOWLEDGEMENTS |
|---|
We thank M. Raudsepp and J. Spence for assistance with EMP and ICP-MS analysis, respectively, and J. MacKenzie and L. Coogan for helpful discussions. We especially appreciate the thorough and insightful reviews by C. Ballhaus, B. R. Frost, R. Luth and C. McCammon. This research was supported by an NSERC of Canada Discovery Grant to D.C.
*Corresponding author. E-mail: dcanil{at}uvic.ca
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCALIBRATION OF AN OXYGEN...SUPPLEMENTARY DATAREFERENCES |
|---|
Apter DB, Harper FJ, Wyatt BA, Scott Smith BH. (1984) The geology of the Mayeng kimberlite sill complex, South Africa. In Kornprobst J (Ed.). The International Kimberlite Conference(Elsevier, Amsterdam) pp. 43–57.
Berg GW and Carlson JA. (1998) The Leslie kimberlite pipe of Lac de Gras, Northwest Territories, Canada: evidence of near surface hypabyssal emplacement. Extended Abstracts, 7th International Kimberlite Conference(University of Cape Town, Cape Town) pp. 81–83.
Burns RG. (1989) Mössbauer spectra of 57Fe in rare-earth perovskites: applications to the electronic states of Fe in the lower mantle. In Navrotsky A and Weidner DJ (Eds.). Perovskite: a Structure of Great Interest to Geophysics and Material Science. Geophysical Monograph, American Geophysical Union 45: pp. 81–90.
Canil D and Bellis A. (2007) Ferric iron in CaTiO3 perovskite as an oxygen barometer for kimberlite magmas 2: applications. Jouranal of Petrology (in press).
Canil D, Schulze DJ, Hearn BCJ, Hall DC, Milliken SM. (2003) Lithospheric roots beneath western Laurentia: the geochemical signal in mantle garnets. Canadian Journal of Earth Sciences 40:1027–1051.
Carmichael ISE. (1967) The iron–titanium oxides of salic volcanic rocks and their associated ferromagnesian silicates. Contributions to Mineralogy and Petrology 14:36–64.[CrossRef]
Carmichael ISE. (1991) The redox states of basic and silicic magmas: a reflection of their source regions? Contributions to Mineralogy and Petrology 106:129–141.[CrossRef][Web of Science]
Chakhmouradian AR and Mitchell RH. (2000) Occurrence, alteration patterns and compositional variation of perovskite in kimberlites. Canadian Mineralogist 38:975–994.
Chen Z, Canil D, Longerich HP. (2000) Automated in situ trace element analysis of silicate materials by laser ablation inductively coupled plasma mass spectrometry. Fresenius Journal of Analytical Chemistry 368:73–78.[CrossRef][Medline]
Clement CR. (1982) A comparative geological study of some major kimberlite pipes in the Northern Cape and Orange Free State. Ph.D. thesis University of Cape Town.
Dawson JB. (1967) Geochemistry and origin of kimberlite. In Wyllie PJ (Ed.). Ultramafic and Related Rocks(John Wiley, New York) pp. 241–251.
Dawson JB. (1972) Kimberlites and their relationship to the upper mantle. Philosophical Transactions of the Royal Society of London, Series A 271:297–311.
Dawson JB and Hawthorne JB. (1973) Magmatic sedimentation and carbonatitic differentiation in kimberlite sills at Benfontein, South Africa. Journal of the Geological Society, London 129:61–85.
Drake MJ and Weill DF. (1972) New rare earth element standards for electron microprobe analysis. Chemical Geology 10:179–181.[CrossRef][Web of Science]
Eccles DR, Heaman LM, Luth RW, Creaser RA. (2004) Petrogenesis of the late Cretaceous northern Alberta kimberlite province. Lithos 76:435–459.[CrossRef][Web of Science]
Edgar AD, Arima M, Baldwin DK, Bell DR, Shee SR, Skinner EMW, Walker EC. (1988) High pressure–high temperature melting experiments on a SiO2-poor, CaO-rich aphanitic kimberlite from the Wesselton mine, Kimberley, South Africa. American Mineralogist 73:524–533.[Abstract]
Edwards D, Rock NMS, Taylor WR, Griffin BJ, Ramsay RR. (1992) Mineralogy and petrology of the Aries diamondiferous kimberlite pipe, Central Kimberley Block, Western Australia. Journal of Petrology 33:1157–1191.
Fedortchouk Y and Canil D. (2004) Intensive variables in kimberlite magmas, Lac de Gras, Canada, and implications for diamond survival. Journal of Petrology 45:1725–1745.
Frost BR. (1991) Introduction to oxygen fugacity and its petrologic significance. In Lindsley DH (Ed.). Oxide Minerals: Petrologic and Magnetic Significance. Mineralogical Society of America, Reviews in Mineralogy 25: pp. 1–10.
Fudali RF. (1965) Oxygen fugacities of basaltic and andesitic magmas. Geochimica et Cosmochimica Acta 29:1063–1075.[CrossRef][Web of Science]
Gurney JJ and Switzer GS. (1973) The discovery of garnets closely related to diamonds in the Finsch pipe, South Africa. Contributions to Mineralogy and Petrology 39:103–116.[CrossRef][Web of Science]
Haggerty SE. (1994) Superkimberlites: a geodynamic diamond window to the Earth's core. Earth and Planetary Science Letters 122:57–69.[CrossRef][Web of Science]
Illupin IP, Sobolev SF, Zolotarev BP, Lebedev-Zinovyev AA. (1974) Geochemical specialization of kimberlites from various parts of Yakutia. Geochemistry International 11:357–370.
Kennedy GC. (1948) Equilibrium between volatiles and iron oxides in igneous rocks. American Journal of Science 246:529–548.
Kopylova MG, Russell JK, Cookenboo H. (1998) Petrography and chemistry of the Jericho kimberlite (Slave Craton, Northern Canada). In Gurney J (Ed.). Extended Abstracts, 7th International Kimberlite Conference(University of Cape Town, Cape Town) pp. 449–451.
Kress VC and Carmichael ISE. (1988) Stoichiometry of the iron-oxidation reaction in silicate melts. American Mineralogist 73:1267–1274.[Abstract]
LeRoex AP, Bell DR, Davis P. (2003) Petrogenesis of Group I kimberlites from Kimberley, South Africa: evidence from bulk-rock geochemistry. Journal of Petrology 44:122261–2286.
McCammon CA, Becerro AI, Langenhorst F, Angel RJ, Marion S, Seifert F. (2000) Short-range ordering of oxygen vacancies in CaFe(x)Ti(1 – x)O(3 – x/2) perovskites (0 < x < 0·4). Journal of Physics: Condensed Matter 12:2969–2984.[CrossRef]
Mitchell RH. (1972) Composition of perovskite in kimberlite. American Mineralogist 57:1748–1753.[Web of Science]
Mitchell RH. (1973) Composition of olivine, silica activity and oxygen fugacity in kimberlite. Lithos 6:65–81.[CrossRef]
Mitchell RH. (1986) Kimberlites: Mineralogy, Geochemistry, and Petrology(Plenum, New York).
Mitchell RH. (1997) Kimberlites, Orangeites, Lamproites, Melilitites, and Minettes: a Petrological Atlas(Almaz Press, Thunder Bay, Ont).
Mitchell RJ. (2002) Perovskites Modern and Ancient(Almarez Press, Thunder Bay, Ont) pp. 314.
Mitchell RH, Choi JB, Hawthorne FC, McCammon CA, Burns PC. (1998) Latrappite: a re-investigation. Canadian Mineralogist 36:107–116.
Muir IJ, Metson JB, Bancroft GM. (1984) 57Fe Mössbauer spectra of perovskite and titanite. Canadian Mineralogist 22:689–694.
Muramatsu Y. (1983) Geochemical investigations of kimberlites from the Kimberley area, South Africa. Geochemical Journal 17:71–86.[Web of Science]
Nickel EH and McAdam RC. (1963) Niobium perovskite from Oka, Quebec: a new classification for minerals from the perovskite group. Canadian Mineralogist 7:683–697.
Pell JA. (1995) Kimberlites in the Slave structural province, Northwest Territories: a preliminary review. EGS 1995-12 Northwest Territories Geological Mapping Division, Department of Indian and Northern Affairs, Yellowknife, Canada.
Pouchou JL and Pichoir F. (1985) PAP (Z) procedure for improved quantitative microanalysis. Microbeam Analysis 32:104–106.
Price SE, Russell JK, Kopylova MG. (2000) Primitive magma from the Jericho pipe, N.W.T, Canada: constraints on primary kimberlite melt chemistry. Journal of Petrology 41:6789–808.
Robinson DN. (1975) Magnetite–serpentine–calcite dikes at Premier Mine and aspects of their relationship to kimberlite and to carbonatite of alkalic carbonatite complexes. Physics and Chemistry of the Earth 9:61–70.[CrossRef]
Sack RO, Carmichael ISE, Rivers M, Ghiorso MS. (1980) Ferric–ferrous equilibria in natural silicate liquids at 1 bar. Contributions to Mineralogy and Petrology 75:369–376.[CrossRef]
Schmidberger SS and Francis D. (2001) Constraints on the trace element composition of the Archean mantle root beneath Somerset Island, Arctic Canada. Journal of Petrology 42:1095–1117.
Scott BH. (1979) Petrogenesis of kimberlites and associated potassic lamprophyres from Central West Greenland. In Boyd FR and Meyer HOA (Eds.). Proceedings of the Second International Kimberlite Conference(American Geophysical Union, Washington, DC) pp. 190–205.
Scott Smith BH, Danchin RV, Harris JW, Stracke KJ. (1984) Kimberlites near Orroroo, South Australia. In Kornprobst J (Ed.). Proceedings of the Third International Kimberlite Conference(Elsevier, Amsterdam) pp. 121–142.
Shee SR. (1986) The petrogenesis of the Wesselton mine kimberlites, Kimberley, South Africa. Ph.D. thesis University of Cape Town.
Skinner EMW and Scott BH. (1979) Petrography, mineralogy and geochemistry of kimberlites and associated lamprophyre dykes near Swartruggens, Western Transvaal, R.S.A. In Boyd FR and Meyer HOA (Eds.). Proceedings of the Second International Kimberlite Conference(American Geophysical Union, Washington, DC).
Smith AL. (1970) Sphene, perovskite and co-existing Fe–Ti oxide minerals. American Mineralogist 55:264–269.
Smith CB, McCallum ME, Coopersmith HG, Eggler DH. Boyd FR and Meyer HOA. (1979) Petrochemistry and structure of kimberlites in the Front Range and Laramie Range, Colorado Wyoming. Proceedings of the Second International Kimberlite Conference(American Geophysical Union, Washington, DC) pp. 186–189.
Smith CB, Gurney JJ, Skinner EMW, Clement CR, Ebrahim N. (1985) Geochemical character of southern African kimberlites: a new approach based upon isotopic constraints. Transactions of the Geological Society of South Africa 88:267–280.
Taylor WR, Tompkins LA, Haggerty SE. (1994) Comparative geochemistry of West African kimberlites: evidence for a micaceous kimberlite endmember of sublithospheric origin. Geochimica et Cosmochimica Acta 58:4017–4037.[CrossRef][Web of Science]
Thornber CR, Roeder PL, Foster JR. (1980) The effect of composition on the ferric–ferrous ratio in basaltic liquids at atmospherc pressure. Geochimica et Cosmochimica Acta 44:525–532.[CrossRef][Web of Science]
Toplis MJ and Carrol MR. (1995) An experimental study of the influence of oxygen fugacity on Fe–Ti oxide stability, phase relations, and mineral–melt equilibria in ferro-basaltic systems. Journal of Petrology 36:51137–1170.
Waerenborgh JC, Figueiredo FM, Frade JR, Colomer MT, Jurado JR. (2001) Fe4+ content and ordering of anion vacancies in partially reduced AFexTi1–xO3–y (A = Ca, Sr: x < 0·6) perovskites. An 57Fe Mössbauer spectroscopy study. Journal of Physics: Condensed Matter 13:8171–8187.[CrossRef]
Zhang R and Liu B. (1983) Kimberlites from North China. Geochemical Journal 17:209–221.[Web of Science]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  Y. Fedortchouk and D. Canil Diamond oxidation at atmospheric pressure: development of surface features and the effect of oxygen fugacity European Journal of Mineralogy, June 1, 2009; 21(3): 623 - 635. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Melluso, M. Lustrino, E. Ruberti, P. Brotzu, C. de Barros Gomes, L. Morbidelli, V. Morra, D. P. Svisero, and F. d'Amelio MAJOR- AND TRACE-ELEMENT COMPOSITION OF OLIVINE, PEROVSKITE, CLINOPYROXENE, Cr-Fe-Ti OXIDES, PHLOGOPITE AND HOST KAMAFUGITES AND KIMBERLITES, ALTO PARANAIBA, BRAZIL Can Mineral, February 1, 2008; 46(1): 19 - 40. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||