The Partitioning of Fluorine1 between Granitic Melt and Mn-rich Garnet 1 1 JAMES L. MANER IV *, DAVID LONDON , GEORGE B. MORGAN VI
1
ABSTRACT
INTRODUCTION Garnet has been used extensively to interpret the thermobarometric history of Earth because of its refractory nature, resistance to re-equilibration, and its presence in a wide range of rock types. Garnet could also be used to estimate the F content of granitic melt. A complete solid solution based on the substitution 4F- for 1SiO44(Grew et al. 2013) between spessartine and the hypothetical Mn2+, F end-member katoite (Mn3Al2[]3F12: where [] is site vacancy) has not been assessed, however garnets with 8mol% Mn2+, F katoite are found in nature (Smyth et al. 1990). S-type granites are typically F-poor and contain spessartine-rich garnets whereas A-type granites are F-rich and rarely contain spessartine-rich garnets. Because spessartine-rich garnets have been known to incorporate weight percent levels of F and are found in S-type granites as a solid solution of almandine-spessartine, it is possible that garnet could be used to decipher the F budget of S-type granitic melt prior to volatilization of F.
Experiments began by combining muscovite (Spruce Pine, NC), albite (Copelinha, Brazil), orthoclase (synthetic mixture of reagents), and quartz (GE Ultra-high purity) in a mixture that represents a peraluminous, haplogranite minimum composition (Ab36.48Or31.44Qtz27.76Crn4.31). Spessartine (Sps95Alm5) and almandine (Alm46Prp44Sps6Grs4) were added as sources of Fe, Mn, and Mg. Silver fluoride was added as a source of fluorine. Because silver alloys with the precious metal capsule, dissolves easily in hot water, and doesn’t add other elements such as Na and Al, other fluorine bearing materials were avoided. Homogenization of starting material mixtures was completed by grinding the mixture dry in an agate mortar and pestle.
The forward direction experiment to 800oC produced garnet rims on relict spessartine and almandine cores, biotite and oxides (Figure 1). Garnet cores are resorbed and the rims have welldeveloped faces (Figure 2). Biotite crystals are approximately 5-10µm wide and 20-40µm long (Figure 3). Oxide crystals are less than 20µm (Figure 2). The average composition of rims on relict almandine grains are Sps84Alm6Prp10 and rims on relict spessartine grains are Sps77Alm11Prp12. The F content of garnet rims is 0.57 wt.% (1σSD: 0.13) for rims on relict spessartine grains and 0.46 wt.% (1σSD: 0.18) for rims on relict almandine grains. Biotite crystals are phlogopite-rich (Phl87Ann13) with less than 10% IVAl. The F content of biotites is 6.53 wt.% (1σSD: 0.34). Resulting partitioning coefficients (defined here as DFmin/melt, where ‘min’ is the weight percent of F in the mineral and ‘melt’ is the weight percent of F in the glass) are DFSps/melt=0.23(0.05, 1σSD), DFAlm/melt=0.18(0.07, 1σSD), and DFBt/melt=2.59(0.14,1σSD).
One experiment at 700oC yielded the same products as the 800oC step but with the addition of corundum (Figure 4). Corundum crystals are less than 1µm wide (Figure 4). EMPA shows that the Starting material mixtures were added to Au capsules in addition to 7 to 8 weight percent deionized, ultra filtered water. biotite is compositionally identical to the 800oC products. Garnets have a new outer rim that is only a few microns wide and are not analysable by EMPA methods. The composition of the glass at The capsules were sealed using a TIG (Ar-gas) welder. The mass of each capsule was measured before and after welding to o o check for water loss during welding. Capsules were then placed in a drying oven set at 110oC, removed after resting in the oven 700 C is distinctly different from the glass at 800 C in that the ASI, MnO, FeO, MgO, and F contents decreased and SiO2 increased. for an hour and re-weighed to check for leaks from poor welding. Capsules that didn’t show signs of leakage were used for experiments. Capsules were loaded into NIMONIC-105 cold-seal reaction vessels then placed into furnaces, pressurized to 200 MPa and heated to the temperature of interest. Pressure was monitored during each experiment with a factory-calibrated Heise bourdon tube gauge. Temperature was monitored using Chromel-Alumel thermocouples. Uncertainty in pressure and temperature is 10MPa and ±5oC. Oxygen fugacity was not monitored internally during each experiment by the use of a oxygen buffer; however, the f(O2) of the reaction vessel system is known to be approximately 0.5 log unit below the Ni-NiO oxygen buffer Two experiments were ran to 800oC for 8 days. After 8 days, both experiments were quenched isobarically in a jet of compressed air at a rate of ~300oC/min until they fell below 100oC. One of the two experiments was then reran to 700OC directly and allowed to rest for 14 days. At the conclusion of each experiment, the capsules were removed from their vessels and reweighed to check for a gain or loss of mass. It is common for capsules to gain a few milligrams during the experiment because Ni from the cold-seal system alloys with the precious metal capsules. After weighing each capsule, the capsules were punctured to check for fluid and pH. Experimental products were examined optically and mounted in epoxy for EMPA.
DISCUSSION
Figure 4 0.300
y = 0.0467x0.4302 R² = 0.6817 0.200
0.150
0.100
0.050
ANALYTICAL METHODS
Figure 5 0.200
0.000 0
Compositions and identification of phases and back-scattered electron images were collected using the Cameca SX-50 electron microprobe at the University of Oklahoma. Energy dispersive x-ray analysis (EDXA) was used to identify phases. Wavelength-dispersive x-ray analysis (WDS) was used to determine the composition of glasses and crystals. Glass analyses used a two part analytical routine at 15 kV accelerating voltage: an initial 2 nA beam current and a 20µm spot was used for Na, K, Ca, and Al, and Si, and all other elements were analysed subsequently with a 20 nA beam current and a 20µm spot. The compositions of crystalline products were determined using a 15 kV (for garnet) or 20 kV (for biotite) accelerating voltage, 20 nA beam current and a 2µm spot. Elements analysed include Si, Al, K, Na, Ca, Mg, F, Ti, Mn, and Fe. Minimum detection limits for all elements analysed are below 0.05 wt.% except for F at 0.09 wt%.
5
10
15
20
25
30
Mn/Fe
Figure 6
BACKGROUND
y = 4x
0.180 0.160 0.140
y = 0.9642x + 0.0327 R² = 0.3741
0.120 0.100 0.080 0.060 0.040 0.020
0.800
0.000 0.000
0.010
0.020
0.030
0.040
0.050
0.060
0.070
1/4 F (apfu)
0.700
0.600
Fe (apfu)
Glass
Biotite
0.100
The partitioning of F between garnet and melt and the incorporation of F into the garnet crystal structure deserves more attention. The data presented in this work is preliminary and more experiments need to be conducted to ensure equilibrium is met. The previously accepted incorporation mechanism of F into the garnet crystal structure deserves more attention. It is possible that the substitution of F anions for SiO44– anions doesn’t operate independent of other substitution mechanisms.
Figure 7
Garnet 0.000 0.000
3.5 0.050
0.100
0.150
0.200
0.250
0.300
F (apfu)
3
The partitioning of F between garnet and granitic melt could help unravel the problem of determining the amount of this volatile element in melt. Other monitors of F exist, however they use a rarer mineralogy than garnet and are vulnerable to re-equilibration during slow cooling or upon contact with secondary fluids. Potentially, the Mn/Fe ratio of garnet or total F in garnet could be used to determine the amount of F in a granitic melt at the time of crystallization.
Mn (apfu)
2.5
Several experimental studies have been conducted with the goal of using mineral compositions to determine the amount of F in melt at the time of crystallization. London et al. (2001) determined the partitioning of F between melt and the amblygonitemontebrasite solid solution. Price et al. (1999) measured the concentration of F in granitic melt in equilibrium with a common mineral assemblage in A-type granites: titanite+fluorite+quartz+rutile. Icenhower and London (1997) experimentally determined a relationship between the Mg’ (Mg/Mg+Fe) of biotite and DFBt/melt. Because amblygonite-montebrasite and the assemblage titanite+fluorite+quartz+rutile are not common in S-type granites, these mineral monitors can’t be used to determine the F budget of many S-type granites. Biotite is a common accessory mineral in both Stype and A-type granites and can be used successfully to determine the amount of F in melt at the time of crystallization; however, biotite is known to react with late-stage fluids and it is possible that some of F in biotite is from late-stage fluid. It is possible that the F content of garnets can be used to decipher the F content of melt.
Figure 3
2
y = 7.164x + 1.6622 R² = 0.7268
1.5
ACKNOWLEDGEMENTS
1
Garnet Fe-Garnet w/ Rim
Figure 8
0.5
8.0000
0
0
7.0000
Biotite
y = 0.0835x R² = 0.9586
5.0000
Mn-Garnet w/ Rim
0.1
0.15
0.2
0.25
0.3
I’d like to thank Dr. David London and Dr. George B. Morgan VI for their support and advice thus far into this project. Dr. London has provided gold for my experiments and funding for my EMPA work. Dr. Morgan has assisted me with proper EMPA techniques from the start. I’d like to thank the ConocoPhillips School of Geology and Geophysics at the University of Oklahoma for sponsoring part of this presentation.
REFERENCES
4.0000 3.0000
Oxides
0.05
F (apfu) 6.0000
Corundum
The partitioning of F between spessartine-rich garnet and melt could be dependent on the Mn/Fe ratio of the garnet; however, because of large standard deviations on F, the partitioning could be independent of Mn/Fe ratio. The partitioning of F between biotite and melt has been show by Wolf and London (1997) to be a function of the Mg’ (Mg/(Mg+Fe)) in biotite. Figure 8 shows a plot of their data with one data point (blue diamond) representing the average from this work. The discrepancy could be due to a difference in the starting F content of the two glasses or to disequilibrium in the present work.
CONCLUSIONS
y = -0.372ln(x) - 0.4305 R² = 0.6741
0.400
0.200
Figure 1
Another possible mechanism for the incorporation and monitoring of fluorine could be the increasing the Mn/Fe ratio of garnet. As the Mn/Fe ratio in garnet increases, the F content increases. This could be due, in part, to the Fe-F avoidance such that with increasing Mn and decreasing Fe, more F can be incorporated into the structure (Figures 4, 6 and 7). However, even at an Mn/Fe ratio of 100, the maximum amount of F incorporation is 0.34 apfu. Figure 7 also exhibits the saturation of Mn in garnet (unfilled diamonds) and that the amount of F in garnet is independent of Mn content at Mn saturation.
0.500
0.300
Biotite
DFBt/melt
Fluorine has unique chemical characteristics in granitic melt. Unlike other halogens, fluorine preferentially partitions into granitic melt rather than a coexisting vapor phase (Webster 1990). As much as 8 wt.% F can be incorporated into a metaluminous to peraluminous granitic melt (Webster 1990). Above 8 wt.%, F will partition into a vapor phase, if a vapor phase is present (Webster 1990). The addition of fluorine shifts the haplogranite minimum toward the feldspar binary, reduces the temperature of the eutectic, reduces melt viscosity thereby increasing diffusion of other elements in melt, and enhances the partitioning of high-field strength elements into melt (Manning 1981, Dingwell and Scarfe 1985, Webster 1990, Dingwell and Webb 1992). Because granites are composed predominantly of minerals whose structures don’t accept F, most or all of the original F in a granitic melt will eventually partition into a vapor phase. Because most or all of the F available in the original melt will eventually partition into a vapor phase, whole-rock chemical analysis provides a minimum estimate of the original melt’s F budget.
The incorporation of fluorine into the garnet structure seems to be controlled mostly by the Mn/Fe ratio of the garnet and/or Si deficiency (Figures 4 and 5). Smyth et al. (1990) proposed a substitution of one 4F– for SiO4– tetrahedral. This substitution mechanism, when fully operated, leads to a solid solution between spessartine (Mn3Al2Si3O12) and and Mn2+,F analogue of katoite (Mn3Al2[]3F12; where [] is site deficiency). In a natural spessartine, the fluorine content is 3.68 wt.% F (Smyth et al., 1990). The hypothetical end-member Mn2+,F analogue of katotite contains 51 wt.% F. For ~4 wt.% F to be incorporated into the garnet structure, the 4F—>SiO44– operation must be completed once. It’s possible that there is no solid solution between spessartine and the F-rich analogue and that 1 apfu F (or ~4wt.%) is the maximum amount of F permissible in spessartine-rich garnets. It is also possible that more F could be incorporated into the garnet structure when coupled with H+ via a coupled substitution.
0.250
Z site deficiency (3-Si apfu)
The partitioning of F between Grt-melt seems to be controlled by the Mn/Fe ratio of Grt and/or T-site deficiency. The average F content of new garnet is 0.57 wt.% (1σ=0.13) in overgrowths with XSps=0.83 and 0.46 wt.% (1σ=0.18) with XSps=0.76. The corresponding average F content of glass is 2.52 wt.% (1σ=0.10), yielding crystal/melt partition coefficients in the range of 0.18-0.23. Further experiments are being conducted to test the dependence on F concentration and temperature and to reduce the variability of F content among products.
RESULTS
EXPERIMENTAL METHODS
F (apfu)
As a potential monitor of fluorine concentration in evolved granitic liquids, the partitioning of F between garnet and peraluminous Mnbearing granitic melt has been assessed by cold-seal experimental techniques and electron microprobe analysis. Garnets and glasses (melts) were synthesized from a mixture of minerals and reagents at 800°C, 200 MPa, and an f(O2) near NNO-0.5 log units. Relicts of garnets (Sps95Alm5 and Alm46Prp44Sps6Grs4) added as sources of Fe, Mg, and Mn have Sps-rich overgrowths. The average compositions of overgrowths on Alm-rich relicts are Sps76Alm12Prp11-Grs1 and those on Sps-rich relicts are Sps83Alm6Prp10Grs1.
School of Geology and Geophysics, University of Oklahoma, 100 East Boyd Street, SEC 710, Norman, OK 73019-1009 (*correspondence:
[email protected])
Dingwell, D.B. and Scarfe, C.M. (1985) Chemical diffusion of fluorine in melts in the system Na 2O-Al2O3-SiO2. Earth and Planetary Science Letters, 73, 377-384 Dingwell, D.B. and Webb, S.L. (1992) The fluxing effect of fluorine at magmatic temperatures (600-800oC): A scanning calorimetric study. American Mineralogist, 77, 3033
2.0000
Icenhower, P.J. and London, D. (1997) Partitioning of fluorine and chlorine between biotite and granitic melt: experimental calibration at 200 MPa H2O. Contributions to Mineralogy and Petrology, 127, 17-29
1.0000
Glass
0.0000 0.0000
London, D., Morgan VI, G.B., and Wolf, M.B. (2001) Amblygonite-montebrasite solid solutions as monitors of fluorine in evolved granitic and pegmatitic melts. American Mineralogist, 86, 225-233
20.0000
40.0000
60.0000
Mg'
80.0000
100.0000
Manning, D.A.C. (1981) The effect of fluorine on the liquidus phase relationships in the system Qz-Ab-Or with excess water at 1 kb. Contributions to Mineralogy and Petrology, 76, 206-215 Price, J.D., Hogan, J.P., Gilbert, M.C., London, D., and Morgan VI, G.B. (1999) Experimental study of titanite-fluorite equilibria in the A-type Mount Scott Granite: Implications for assessing F contents of felsic magma. Geology, 27, 951-954 Smyth, J.R., Madel, R.E., McCormick, T.C., and Munoz, J.L. (1990) Crystal-structure refinement of a F-bearing spessartine garnet. American Mineralogist, 75, 314-318
Figure 2
Figure 4
Webster, J.D. (1990) Partitioning of F between H2O and CO2 fluids and topaz rhyolite melt. Contributions to Mineralogy and Petrology, 104, 424, 438
