American Mineralogist; February 2008; v. 93; no. 2-3;
p. 283-299; DOI: 10.2138/am.2008.2571
© 2008 Mineralogical Society of America
Boralsilite, Al16B6Si2O37, and "boron-mullite:" Compositional variations and associated phases in experiment and nature
Edward S. Grew1,*,
Heribert A. Graetsch2,
Birgit Pöter2,
Martin G. Yates1,
Ian Buick3,
Heinz-Jürgen Bernhardt4,
Werner Schreyer2,
,
Günter Werding2,
Christopher J. Carson3 and
Geoffrey L. Clarke6
1 Department of Earth Sciences, University of Maine, 5790 Bryand Research Center, Orono, Maine 04469-5790, U.S.A.
2 Institut für Geologie, Mineralogie und Geophysik, Ruhr-Universität Bochum, D-44780 Bochum, Germany
3 Research School of Earth Sciences, Australian National University, Canberra, ACT 0200, Australia
4 Zentrale Elektronen-Mikrosonde, Institut für Geologie, Mineralogie und Geophysik, Ruhr-Universität Bochum, D-44801 Bochum, Germany
6 School of Geosciences, University of Sydney, NSW 2006, Australia
Correspondence: * E-mail: esgrew{at}maine.edu
 |
ABSTRACT
|
|---|
Boralsilite, the only natural anhydrous ternary B2O3-Al2O3-SiO2 (BAS) phase, has been synthesized from BASH gels with Al/Si ratios of 8:1 and 4:1 but variable B2O3 and H2O contents at 700–800 °C, 1–4 kbar, close to the conditions estimated for natural boralsilite (600–700 °C, 3–4 kbar). Rietveld refinement gives monoclinic symmetry, C2/m, a = 14.797(1), b = 5.5800(3), c = 15.095(2) Å, β = 91.750(4)°, and V = 1245.8(2) Å3. Boron replaces 14% of the Si at the Si site, and Si or Al replaces ca. 12% of the B at the tetrahedral B2 site. A relatively well-ordered boralsilite was also synthesized at 450 °C, 10 kbar with dumortierite and the OH analogue of jeremejevite. An orthorhombic phase ("boron-mullite") synthesized at 750 °C, 2 kbar has mullite-like cell parameters a = 7.505(1), b = 7.640(2), c = 2.8330(4) Å, and V = 162.44(6) Å3. "Boron-mullite" also accompanied disordered boralsilite at 750–800 °C, 1–2 kbar.
A possible natural analogue of "boron-mullite" is replacing the Fe-dominant analogue of werdingite in B-rich metapelites at Mount Stafford, central Australia; its composition extends from close to stoichiometric Al2SiO5 to Al2.06B0.26Si0.76O5, i.e., almost halfway to Al5BO9. Boralsilite is a minor constituent of pegmatites cutting granulite-facies rocks in the Larsemann Hills, Prydz Bay, East Antarctica, and at Almgotheii, Rogaland, Norway. Electron-microprobe analyses (including B) gave two distinct types: (1) a limited solid solution in which Si varies inversely with B over a narrow range, and (2) a more extensive solid solution containing up to 30% (Mg,Fe)2Al14B4Si4O37 (werdingite). Boralsilite in the Larsemann Hills is commonly associated with graphic tourmaline-quartz intergrowths, which could be the products of rapid growth due to oversaturation, leaving a residual melt thoroughly depleted in Fe and Mg, but not in Al and B. The combination of a B-rich source and relatively low water content, together with limited fractionation, resulted in an unusual buildup of B, but not of Li, Be, and other elements normally concentrated in pegmatites. The resulting conditions are favorable in the late stages of pegmatite crystallization for precipitation of boralsilite, werdingite, and grandidierite instead of elbaite and B minerals characteristic of the later stages in more fractionated pegmatites.
Key Words: Boralsilite • "boron-mullite" • werdingite • boron • pegmatite • Rietveld refinement • electron microprobe • Larsemann Hills • Antarctica • Almgjotheii • Norway
|
INTRODUCTION
|
|---|
The first anhydrous ternary phases to be synthesized in the system B2O3-Al2O3-SiO2 (BAS) were poorly ordered high-temperature (930–1600 °C) orthorhombic materials related to mullite and restricted to a band of solid solution between 3:2 mullite (Al6Si2O13) and Al18B4O33 (Fig. 1
; Letort 1952; Dietzel and Scholze 1955; Scholze 1956; Gelsdorf et al. 1958; Gielisse and Foster 1961). G. Werding, W. Schreyer, and their colleagues synthesized mullite-like orthorhombic compounds hydrothermally at geologically accessible temperatures (e.g., Werding and Schreyer 1996). The compositions of most of these materials could not be determined; an exception was the compound having the formula Al8B2Si2O19 and doubled a and c parameters compared to mullite (Werding and Schreyer 1992). However, the first ternary BAS phase to be discovered in nature, boralsilite, Al16B6Si2O37, is ordered and monoclinic, with all three cell parameters doubled (Grew et al. 1998a; Peacor et al. 1999). Armed with knowledge that an ordered compound had been discovered, Pöter et al. (1998) successfully synthesized under hydrothermal conditions a monoclinic compound having both the degree of order and the composition of natural boralsilite.
Nonetheless, the stability relationships and conditions of formation of boralsilite and the mullite-like compounds remain poorly understood. The objective of the present study is to re-examine synthetic boralsilite and mullite-like aluminoborosilicates and compare the synthetic compounds with naturally occurring analogues to better constrain their origin in nature and the conditions under which they formed. We refined the crystal structure of the synthetic boralsilite to determine whether it is as ordered as the natural material. During the 2003–2004 Antarctic field season, we collected specimens from nine boralsilite-bearing pegmatites and mapped the distribution of these pegmatites and associated B-enriched host rocks in the Larsemann Hills, the type locality for boralsilite. Additional insight on the origin of boralsilite and related compounds is provided by the discovery of a mullite-like boroaluminosilicate as a breakdown product of werdingite at Mount Stafford, central Australia (Buick et al. 2006).
Following Werding and Schreyer (1984, 1996), and for want of a better term, we refer to the synthetic and natural mullite-like orthorhombic ternary aluminum-borosilicates as "boron-mullite." The ferromagnesian solid solutions cordierite-sekaninaite, grandidierite-ominelite, and werdingite-Fe analogue of werdingite are referred to simply by the names of the Mg end-member throughout the paper, although these minerals are, for the most part, Fe-dominant in the B-rich metapelites from Mount Stafford and in the pegmatite from Almgjotheii.
|
ANALYTICAL METHODS
|
|---|
The synthetic and natural borosilicate compounds were analyzed with a Cameca SX-100 electron microprobe at the University of Maine using wavelength-dispersive spectroscopy (WDS). Analytical conditions were twofold at each spot: 5 kV accelerating voltage and 40 nA beam current for B; 15 kV accelerating voltage and 10 nA beam current for Mg, Al, Si, Ca, Ti, Mn, Fe, and As—the only constituents found above detection limits (except Ca) in these four borosilicates and dumortierite. The spot size for all elements ranged from 5 to 20 µm depending on grain size; a larger spot size was used if at all possible. Cameca’s PeakSight software allowed for the cycling from one set of conditions to another. Samples were analyzed for B in peak-area mode with a Mo-B4C (200 Å) synthetic crystal because the B peak is broad and asymmetrical (Fig. 2a); peak integration was carried out over 1000 steps, spanning wavelengths 60 to 78 Å. The upper and lower portions of this peak area were background and were regressed to determine the background contribution under the peak. The other constituents were measured in peak-height mode. Backgrounds for all constituents were measured at each analytical spot. We used the following minerals as standards: elbaite [BK
, sample 98144, Dyar et al. (2001) using B from the crystal-structure refinement and the University of Maine electron microprobe analysis for the other constituents], diopside (MgK, CaK), kyanite (AlK, SiK), rutile (TiK), rhodonite (MnK), almandine (FeK), and skutterudite (AsL). Data were processed using the X-Phi correction of Merlet (1994).
View larger version (28K):
[in this window]
[in a new window]
|
FIGURE 2. Traces of WDS scan of wavelengths where BK peak is located. (a) Scan of boralsilite (sample 010602a1; grain 3). The measured peak area lies between the two vertical bars, beyond which is background. (b) Scans of minerals containing no B showing increase of background with increasing SiO2 content.
|
|
BK X-rays are typically difficult to quantify with the electron microprobe due to absorption, peak shifts, and interference with high-order spectra (McGee and Anovitz 1996). BK measured on boralsilite using a 200 Å, Mo-B4C crystal is asymmetric having 2.5 times higher background on the short-wavelength side of the BK peak compared to the long-wavelength side. The background in the BK interval increases markedly with SiO2 content at wavelengths that would also affect the low wavelength side of the BK peak (Fig. 2b
). This interference is likely due to seventh- and eighth-order SiK peaks and would affect the both the estimated background and the area under BK peak in both standards and unknowns. Higher SiO2 content of the standard compared to boralsilite and "boron-mullite" may lead to an over estimation of B2O3, but quantifying these interferences is difficult. Equally difficult to quantify is the impact of the peak at approximately 67 Å (Fig. 2b). This peak is unrelated to any specific interference and may be caused by fluorescence of B from the Mo-B4C crystal (McGee and Anovitz 1996).
Prior to the analyses at the University of Maine, H.-J. Bernhardt carried out electron-microprobe analyses (EMPA) of the synthetic products for SiO2 and Al2O3 using a Cameca SX-50 at the Ruhr-Universität Bochum. The data were corrected using a PAP matrix-correction procedure assuming that B2O3 is equal to the difference between the total and 100 wt%. The EMPA at the University of Maine gave SiO2 and Al2O3 contents that fall largely within the ranges obtained at the Ruhr-Universität Bochum, but measurement of B2O3 brought the totals only to 67.1–80.5 wt% for disordered boralsilite and "boron-mullite." One explanation for this deficit could be water in the run products, but the analyzed grains did not decompose as expected if large amounts of H2O had been present, although some showed a burn mark after long exposure to the electron beam. Instead, the appearance of the grains in the SEM images suggested that the analyzed grains are porous aggregates. In light of the study by Sorbier et al. (2004) showing that signal loss in mesoporous alumina (pore diameter <50 nm) resulted from epoxy soaked up during sample preparation, we suspected that the low totals resulted from an analogous contamination during preparation of the plug mounts. We performed the following three tests on one grain each of "boron-mullite" (B4) and ordered boralsilite (W2) at 15kV, 10 nA, with a focused beam: (1) counts per second (cps) of CK for "boron-mullite" were five times greater than cps of CK for ordered boralsilite (67 vs. 13 cps) and decreased with time after repeated analyses at one spot; (2) wt% totals for "boron-mullite" increased from 72 to 84% after repeated analyses at one spot, whereas totals for boralsilite remained unchanged; and (3) a trace of Cl was detected in WDS and energy dispersive spectroscopic (EDS) scans of "boron-mullite;" minor Cl was found in the epoxy in the plug. These observations can be explained by the presence of C in the contaminated grain in addition to that in the carbon coat and by volatilization of Cl-bearing epoxy under the electron beam. The repeated analysis at one spot on "boron-mullite" B4 resulted in a decrease in the proportion of B2O3 from 0.169 to 0.149, which exceeds the range in B2O3 obtained by repeated analyses of boralsilite W2, 0.236–0.243. This drift suggests contamination from the epoxy, but no B was detected in the epoxy used in the plug mount (detection limit estimated to be 0.25 wt% B2O3), and no B was reported in a chemical analysis supplied by the manufacturer of the epoxy. Whatever the cause of the drift in B2O3 contents, its effect appears not to be significantly greater than other uncertainties in the EMPA, and the data can provide some useful information on the compositions of disordered boralsilite and "boron-mullite."
The scanning electron microscope (SEM) images were obtained on a Cambridge stereoscan 250 MK3 instrument operated by Rolf Neuser at the Ruhr-Universität Bochum. The samples were sputtered-coated with gold.
Starting materials for the experiments in most cases were gels having an Al:Si ratio of 8:1 or 4:1 prepared from tetraethyl orthosilicate (TEOS) and Al powder dissolved in HNO3 (diluted 1:1), then gelatinized with ammonia and heated to thoroughly decompose the nitrate. Boron was added to the resulting gel as B2O3 or H3BO3 in stoichiometric amounts corresponding to Al16B6Si2O37 or Al8B2Si2O19, respectively, or in amounts in excess of these stoichiometries (Table 1; preliminary report in Pöter et al. 1998). Hydrothermal syntheses were carried out in 1996 and 1997 at 700–800 °C, 1–4 kbar (15–57 days duration) and 450–800 °C, 10 kbar (48–430 h duration); the stability of boralsilite was explored in a few runs at pressures to 30 kbar and temperatures to 1000 °C. Grains from runs B1, B3, B4, W1, and W4 were set in epoxy and mounted in plugs for EMPA.
Pöter et al. (1998) used an automated Siemens D500 powder diffractometer and a Philips PW1050 goniometer to examine run products (5° < 2
< 65°, Si standard). Several run products were re-examined (Table 1), at the Ruhr-Universität Bochum, using X-ray powder diffractograms recorded with a Siemens D5000 diffractometer and an image plate Guinier camera (G670 Huber). The D5000 diffractometer has a modified Debye-Scherrer geometry with the samples enclosed in 0.5 mm glass capillaries. CuK1 radiation was obtained from a focusing Ge(111) monochromator. A position sensitive detector with an entrance window of 50 mm was used.
For the Rietveld refinement of the boralsilite structure, VIAl8VAl8O10(BO3)4(BO4)2(Si2O7), six scans were measured in the range from 10 to 100 °2 with a step size of 0.0078° and a counting rate of 4 s per step. The scans were summed up in order to obtain a single data set. The Guinier camera is also equipped with a Ge monochromator, providing CuK1 radiation. The 2 range was 10 to 100° with a step size of 0.005°. Exposure time was 1 day. The crystal structure of boralsilite was refined according to the Rietveld method (Rietveld 1969) using the Jana2000 program package (Pet
í
ek and Du
ek 2000; Duek et al. 2001). Lattice parameters, four peak-shape parameters of the pseudo-Voigt profile function, and one parameter for a zero-point correction were refined. A Legendre polynomial with 22 parameters was used for the description of the background. The reflections are slightly broadened. The structure was initially refined with the atomic coordinates and the ideal formula Al16B6Si2O37 of natural boralsilite (Peacor et al. 1999). Common isotropic thermal displacement parameters were refined for all O atoms and for the cations in tetrahedral and threefold coordination. The interatomic distances, tetrahedral angles and O-B-O angles in the BO3 units were restrained to remain close to the values of Peacor et al. (1999). A difference-Fourier analysis showed excess electron density at the Si, O10A, and O4 positions. The site-occupancy factors were refined in such a way that combined occupancies of the tetrahedral Si and B2 sites by B and Si (or Al) and vacancies on the O10A and O4 sites were allowed; this refinement yielded a significant reduction of the R values (from Rwp = 4.70 to 4.36) indicating mixed occupancy at the Si and B2 sites as well as incomplete occupancy of the O10A and O4 sites. A new difference-Fourier synthesis showed no residual electron density larger than 0.3 e–, i.e., it gave no clear indication of electron density at the interstitial O10B site. In the final stages of the refinements, all atomic, profile, and background parameters were refined simultaneously. Corrections for absorption and extinction were found to be unnecessary. Preferred orientation was not observed. A Berar correction for serial correlations was carried out by multiplying the standard uncertainties with an appropriate factor (4.3) provided by the Jana2000 program to obtain statistically more realistic values (cf. Hill and Flack 1987). The quality of the refinement is only moderate, i.e., the goodness of fit is 2.44, but the Bragg R values are small enough that there should be no serious errors in the occupancies.
|
SYNTHETIC BORALSILITE AND "BORON-MULLITE"
|
|---|
Werding and Schreyer (1984) were the first to report the hydrothermal synthesis of a phase giving an X-ray diffraction (XRD) pattern similar to the orthorhombic phase Al18B4O33 that Scholze (1956) synthesized at 1100 °C and 1 bar in the anhydrous Al2O3-B2O3 system and reported to have cell parameters close to those of mullite, but with two of them doubled. A similar "boron-mullite" was found as a breakdown product of alkali-free tourmaline at 600–850 °C, 2–8 kbar (e.g., Fig. 3) in the MgO-Al2O3-B2O3-SiO2-H2O system (Werding and Schreyer 1984) and dumortierite in the Al2O3-B2O3-SiO2-H2O system (Werding and Schreyer 1996). "Boron-mullite" could also be synthesized in the latter system at 800–1050 °C, 1 bar–4 kbar (Werding and Schreyer 1992; Pöter et al. 1998); in one case, the phase was reported to be Al8B2Si2O19 (Fig. 1). In addition to orthorhombic "boron-mullite," Pöter et al. (1998) succeeded in synthesizing a monoclinic phase identical to natural boralsilite and suggested that the previously synthesized samples of "boron-mullite" are metastable precursor phases to boralsilite.
|
XRD PROPERTIES AND CHEMICAL COMPOSITION OF ORDERED BORALSILITE
|
|---|
Using newer XRD instrumentation, we re-examined 14 of the 1997 experimental products reported by Pöter et al. (1998) and found that only four are ordered boralsilite (Table 1, Fig. 4a). Ordered boralsilite forms individual prismatic crystals up to 50 µm long and 20 µm wide partly encrusted with very fine prisms (e.g., Figs. 5a and 5b) or aggregates of finer prisms. The powder XRD pattern shows numerous sharp reflections characteristic of the pattern for ideal boralsilite and cell parameters could be calculated (Table 2). A fifth sample (B25) is better ordered than the disordered boralsilite described below, and could be almost as ordered as the four listed in Table 1, but critical peaks are obscured by the reflections from dumortierite and the OH-analogue of jeremejevite.
View larger version (22K):
[in this window]
[in a new window]
|
FIGURE 4. XRD patterns of experimental products. (a) Ordered boralsilite with a few reflections indexed. (b) Disordered boralsilite and "boron-mullite." m = reflections of "boron-mullite" not overlapping with boralsilite reflections. (c) Disordered boralsilite with minor H3BO3 impurity. (d) "Boron-mullite" with a few reflections indexed, minor quartz and H3BO3 impurities. The 101 quartz reflection is a shoulder on the 210 "boron-mullite" reflection.
|
|
View larger version (84K):
[in this window]
[in a new window]
|
FIGURE 5. Images obtained by scanning electron microscopy of run products. (a) Prism of ordered boralsilite among grains of the amorphous phase. (b) Very fine-grained prisms on the surface of a prism of ordered boralsilite. (c) An aggregate of disordered boralsilite. (d) A close-up of an aggregate of disordered boralsilite. (e) Aggregate of "boron-mullite." (f) Same aggregate as e but magnified.
|
|
Electron-microprobe analyses of two relatively coarse-grained samples of ordered boralsilite (B1; W2) show that it contains less Si than natural boralsilite and that Al varies inversely with Si at constant B (Table 3, Fig. 6a). A Rietveld refinement of a sample of ordered boralsilite gave a structure (Table 4; Appendix Table 11) similar to that of natural boralsilite from the Larsemann Hills (Peacor et al. 1999). Significant differences are substitutions at two cation sites (Si, B2) and vacancies at the O4 site, as well as at the O10A site. Substituting B for Si at the Si site and Al (not Si) for B at the B2 site is consistent with the EMPA results, i.e., these individual substitutions sum to Al substitution for Si (Fig. 6a). Partial occupancy at O4 could be related to incorporation of B at Si, but the resulting decrease in negative charge is not offset by the decrease of positive charge. The overall charge also is not balanced: +73.72 vs. –72.16. This discrepancy could be due either to undetected Al substitution for Si at the Si site, cation vacancies or to overlooked O. Detecting Al substitution for Si is difficult because the number of electrons is the same for Si4+ and Al3+, and discernable lengthening of T-O bonds requires a relatively large amount of Al substitution. In boralsilite, the effects of simultaneous B substitution would offset any lengthening of the T-O bond. In addition, soft constraints were set for the interatomic distances to obtain regular tetrahedra. We tested for vacancies at the Al sites. The site-occupancy factors of some of the Al atoms were found to refine to lower values, but the deviations did not significantly exceed 1
. There is no residual electron density larger than 0.38 e–; these peaks are normal noise within the margin of error, and thus too low to be assigned to interstitial O. Peacor et al. (1999) reported occupancy at O10B in natural boralsilite, but the displacement parameter was found to be unusually large, suggesting that the O atom is not at a definite lattice position but smeared over a limited region. We doubt that O could have been found at O10B with Rietveld refinement. In summary, the lack of electron-density peaks shows that there is no charge at regular lattice sites, although this does not exclude the possibility of O atoms in defects. Thus, we have no explanation for the charge imbalance; a single-crystal structure study will be necessary to resolve the issue.
View larger version (18K):
[in this window]
[in a new window]
|
FIGURE 6. (a) Compositions of synthetic boralsilite from runs B1 (ordered, filled symbols) and W2 (ordered, unfilled symbols). Gray-filled symbols represent the composition of W2 refined by the Rietveld method (assuming that only Al replaces B at the B2 site). (b) Compositions of boralsilite in samples 121502E, 020602A, 121102B, 121102C, and holotype (8812905-1 and -2) from the Larsemann Hills (Type I). Open squares indicate ideal composition Al16B6Si2O37. The lines indicate the ideal substitutions.
|
|
|
XRD PROPERTIES AND CHEMICAL COMPOSITION OF DISORDERED BORALSILITE AND "BORON-MULLITE"
|
|---|
The XRD patterns of seven run products differ from that of ordered boralsilite in that reflections are broadened and the intensity of the reflections is modified (Fig. 4c). In five cases, these products were originally considered to be "boron-mullite" similar to the compound Al8B2Si2O19 reported by Werding and Schreyer (1992). However, we believe that these run products more likely consist of disordered boralsilite because weak reflections of boralsilite can be found. Broadening could be due to either a very small size of coherently scattering domains or strain associated with lattice imperfections. The disordered boralsilite is neither microcrystalline boralsilite nor a mechanical mixture of boralsilite and an orthorhombic phase; testing the latter assumption did not give meaningful results for the lattice parameters. If we assumed that the products are disordered boralsilite, the lattice parameters (Table 2) could be refined with LeBail fits, but some anisotropic broadening had to be neglected. Simulations indicate that the intensity distribution cannot be explained by occupancy at O10B instead of at O10A.
Reflections of an orthorhombic phase in addition to those for boralsilite are definitely present in the XRD pattern of B26 and W7, which results in a set of doublets (e.g., Fig. 4b); the extra reflections could be indexed as mullite, as could those in and the pattern for the single phase mixed with minor quartz in W6 (Fig. 4d). Cell volumes of these orthorhombic phases (Table 2) are 2–3% smaller than those of 3:2 mullite [e.g., 167.335(2) Å3, Balzar and Ledbetter 1993] and 2:1 mullite [e.g., 168.04(3) Å3, Angel and Prewitt 1986], most likely due to incorporation of B (see below). A very broad, low hump between 20.0 and 20.4 °2 in W6 is close to the peak at 20.3 or 20.5 °2 in the patterns for disordered boralsilite, itself corresponding to the 
12 and 112 reflections in ordered boralsilite, and thus could be due to incipient development of a boralsilite-like structure. The patterns reported for Al8B2Si2O19 (Werding and Schreyer 1992) and for syntheses at 1050 °C and 1 bar (B4 and W4; Table 1) are similar to Figure 4d except that this reflection, which is found at 20.301, 20.433, and 20.395 °2, respectively, is more prominent, i.e., the development of boralsilite apparently has proceeded further. The presence of this single reflection made it necessary for Werding and Schreyer (1992) to index its powder XRD pattern with a supercell having two doubled cell parameters; we indexed its powder XRD pattern with a mullite cell to give cell parameters very similar to those for "boron-mullite" W6 (Table 2).
SEM images of the run product of B3 show irregular masses with a knobby surface (Fig. 5c) in places with fine prisms (Fig. 5d); we presume that these masses are aggregates of disordered boralsilite. In contrast, grains of "boron-mullite" in W4 show discrete edges and relatively smooth, quasi-planar faces (Figs. 5e and 5f), but these grains also seem to be highly porous aggregates. The grain of analyzed disordered boralsilite in epoxy plug B3 has an irregular outline, whereas the analyzed "boron-mullite" grains in epoxy plugs B4 and W4 have a polygonal outline; i.e., the analyzed grains in the plug mount appear to be the same material as the grains imaged with the SEM. Compositions of disordered boralsilite plot in a roughly linear trend passing about midway between Al8B2Si2O19 and Al16B6Si2O37, i.e., disordered boralsilite differs in composition from ordered boralsilite (Fig. 7). "Boron-mullite" compositions plot a relatively tight trend between Al8B2Si2O19 and the Al borate Al18B4O33, with the material synthesized from the 4:1 gel (W4) being more siliceous than material synthesized from the 8:1 gel (B4). This result is consistent with the inference of Werding and Schreyer (1992) that the composition of their "boron-mullite" is Al8B2Si2O19, because they obtained a single-phase product from the 4:1 gel composition and a B content (11.54 wt% B2O3) consistent with this formula. The composition of the "boron-mullite" synthesized from a 3:1 gel + melted B2O3 by Letort (1952) plots along this trend, but at higher Al2O3/SiO2 ratio despite the lower Al2O3/SiO2 ratio of the starting gel. The distinct trends for disordered boralsilite and "boron-mullite," together with simultaneous presence of both in some run products, e.g., B26, W7 (Table 1, Fig. 4b), suggests the absence of a gradation between the two phases.
View larger version (22K):
[in this window]
[in a new window]
|
FIGURE 7. Compositions of "boron-mullite" synthesized from gels at 1050 °C and 1 bar (B4, W4, this study) and at 930 °C and 1 bar (3:1 gel, Letort 1952) plus compositions of disordered boralsilite (B3) in mol% oxide. Trend line is based on least-squares fit to the nine points for "boron-mullite" B4 and W4.
|
|
The "amorphous" phase
SEM images of the run product of B1 and W2 showed that ordered boralsilite is associated with irregular knobby masses (e.g., Fig. 5a), which we presume is amorphous because it did not give an XRD pattern. Similar appearing material in the plugs for B1 and W2 is extremely heterogeneous; energy-dispersive and wavelength-dispersive scans revealed that Al is far more abundant than Si and that substantial B is present, that is, the composition is similar to disordered boralsilite and "boron-mullite."
Formation conditions of synthetic boralsilite and "boron-mullite"
Boralsilite and "boron-mullite" have been synthesized hydrothermally in the B2O3-Al2O3-SiO2-H2O (±MgO) system over a wide range of temperatures at pressures up to 10 kbar (Fig. 3). Boralsilite (ordered and disordered) was found in syntheses between 450 and 800 °C. The only synthesis below 700 °C, i.e., B25 at 450 °C and 10 kbar, also yielded substantial amounts of dumortierite and the OH-analogue of jeremejevite, which have been synthesized at similar temperatures in other experiments (Stachowiak and Schreyer 1998; Wodara and Schreyer 2001). Boralsilite has broken down by 1000 °C (Fig. 3) and at 20 kbar (Table 1). It is likely that disordered boralsilite is metastable, whereas ordered boralsilite could be stable at 700–800 °C, 1–4 kbar, as Pöter et al. (1998) suggested. Nonetheless, it remains unclear what physical-chemical conditions favor transformation of disordered to ordered boralsilite. It involves a change of composition, at least in some cases (Figs. 1 and 7). There is no evidence that duration, gel composition, or proportion of H3BO3 played a critical role. One run (B7) was seeded with ordered boralsilite from B1, but this did lead to the formation of ordered boralsilite. Of the 14 run products that were re-examined, five contained an apparently amorphous phase and four of these yielded ordered boralsilite. This observation suggests that another factor could play a role, for example, the chance seeding by an unknown impurity in some runs resulted in a portion of the material crystallizing to ordered boralsilite that coarsened rapidly amid a residue that failed to crystallize sufficiently to give reflections in an XRD pattern, even in runs as long as 57 days (run B27; Table 1).
|
A NATURAL "BORON-MULLITE"?
|
|---|
A mineral provisionally identified as "boron-mullite" has been found in three specimens of SiO2-undersaturated, B-rich metapelites from Mount Stafford, central Australia. Specimens MSTgran and 95–75D originate from a stratiform grandidierite-bearing lens in metapelitic migmatite and cordierite granofels in metamorphic zone 4 (granulite facies), close to the boundary with zone 3 (transitional between the upper amphibolite facies and the lower granulite facies). Specimen MST1002 was found in float within zone 4 (Buick et al. 2006 and unpublished manuscript).
All three specimens are dominated by cordierite, hercynite, and K-feldspar (Buick et al. 2006 and in preparation). Biotite and andalusite porphyroblasts are also major constituents in MST1002, but the andalusite is replaced by sillimanite aggregates in the other two specimens. Small amounts of relict tourmaline are present in MST1002. Relatively coarse prisms of grandidierite are characteristic of MSTgran and 95-175D, whereas in MST1002, grandidierite enclosing abundant fine ilmenite is found only as an overgrowth on werdingite. Prisms of werdingite (XFe = 0.49–0.62, e.g., MST1002, Table 5) most typically form a fringe on andalusite or sillimanite bundles replacing it; there also are loose aggregates of subparallel prisms in cordierite.
"Boron-mullite" is visible only in back-scattered electron images (Fig. 8). This identification is based solely on chemical composition (see below); we have no XRD data to confirm that the analyzed phase is related to mullite. The mineral occurs exclusively as a replacement of werdingite, generally in the vicinity of andalusite or sillimanite, which it could also be replacing; it is also commonly in contact with cordierite, and locally with hercynite, biotite, and K-feldspar (as an inclusion). The "boron-mullite" is heterogeneous at a very fine scale; sections viewed roughly down the c axis appear as a patchwork of different shades of gray, whereas sections viewed at a high angle to it show banding (Figs. 8b, 8c, and 8d), in both cases due to compositional variation, and suggesting that "boron-mullite" is fibrous.
View larger version (59K):
[in this window]
[in a new window]
|
FIGURE 8. Back-scattered electron images of borosilicates in metapelites from Mount Stafford, central Australia. (a) "Boron-mullite" replacing werdingite in a cordierite matrix. Light grains are hercynite (subequant) and ilmenite (tabular). Sample 95-175D. (b) Banded "boron-mullite" has replaced werdingite adjacent to andalusite. Werdingite shows indistinct zoning. Sample MST1002. (c) Banded "boron-mullite" has replaced werdingite. Sample MSTgran. (d) Banded "boron-mullite" has replaced werdingite. Lightest bands (S) approach sillimanite in composition. Sample MST1002. Abbreviations: A = andalusite, BM = "boron-mullite", C = cordierite, H = hercynite, S = sillimanite-like, W = werdingite.
|
|
Attempts to analyze a given area of apparently homogeneous electron-scattering intensity with the electron microprobe failed to give consistent results, requiring us to treat each spot analysis individually (Fig. 9, Table 6) instead of averaging them. Its composition extends without break from nearly stoichiometric Al2SiO5 almost halfway across the BAS triangle toward the aluminoborate Al5BO9 (Fig. 1). A least-squares fit to the data gives SiO2 16.5, Al2O3 76.2, B2O3 7.1, Sum 99.8 wt% as the end point of the compositional range. Although the range of compositions overlaps with those of the three samples of "boron-mullite" synthesized at high temperatures by Dietzel and Scholze (1955), the natural "boron-mullite" contains stoichiometric Al2SiO5 as one component, not Al6Si2O13 (3:2 mullite) as inferred by investigators of the high-temperature phases. The trend for the Mount Stafford "boron-mullite" is also distinct from the trend for "boron-mullite" synthesized from gels (Fig. 1).
View larger version (21K):
[in this window]
[in a new window]
|
FIGURE 9. Compositions of "boron-mullite" in B-rich metapelitic rocks from Mount Stafford, central Australia. Plotted data are the results of individual analyses on three or four grains in each of the three sections; only analyses totaling 98–102 wt% and containing less than 0.5 wt% MgO and 1.2 wt% FeO, i.e., negligible werdingite impurity, are plotted. Ratios refer to Al2O3: SiO2 (mol) in mullite. Sil = sillimanite. Lines are least-squares fits to the data.
|
|
Metapelites, metapsammite, and metabasites from the Mount Stafford area experienced low-pressure/high-temperature metamorphism (e.g., Vernon et al. 1990; Greenfield et al. 1998; White et al. 2003), and zone 4 reached granulite-facies conditions (a minimum temperature of 775–785 °C at 3.3–4 kbar, White et al. 2003). Within this zone, metapsammites and moderate-Al metapelites commonly contain the assemblage quartz + biotite + cordierite + K-feldspar ± orthopyroxene ± garnet, and high-Al metapelites, quartz + K-feldspar + biotite + cordierite + sillimanite + hercynite (e.g., Greenfield et al. 1998; White et al. 2003). The formation of grandidierite, werdingite, and "boron-mullite" is closely tied to partial melting of B-rich metapelite during which tourmaline, relics of which remain in MST1002, is a reactant, and werdingite and grandidierite are products in samples 95-175D and MSTgran, and werdingite alone in MST1002, at close to peak temperatures (780 °C, 3.3–4 kbar for the boundary between zones 3 and 4, White et al. 2003). "Boron-mullite" could have resulted from either retrograde or prograde breakdown of werdingite. Werding and Schreyer (1992) reported the incongruent melting of werdingite to a "boron-mullite" at 1200 °C, 1 bar under nearly anhydrous conditions. Although this temperature far exceeds those estimated for Mount Stafford, we cannot rule out the possibility that at higher pressures in the presence of granitic melt, werdingite could have melted incongruently at much lower temperatures.
|
CURRENT STATUS OF "BORON-MULLITE"
|
|---|
Werding and Schreyer (1996) suggested that the ternary compositional range of "boron-mullite" occupies a quadrilateral with vertices at 2:1 mullite, 3:2 mullite, AlBO3, and Al5BO9 (Fig. 1). However, solid solution between 3:2 mullite and Al18B4O33 (dashed line in Fig. 1) and their own phase Al8B2Si2O19 (Werding and Schreyer 1992) constituted the only evidence at the time for ternary phases in the quadrilateral. No naturally occurring ternary phases were known until boralsilite was discovered.
The results of the present study suggest the presence of at least two solid-solution series extending into the "boron-mullite" quadrilateral, one extending from sillimanite toward Al5BO9 in what is presumed to be a natural "boron-mullite," and the other from Al8B2Si2O19 toward Al18B4O33 in material synthesized from gels. The three compositions of "boron-mullite" synthesized from melt plot close to the trend for the presumed natural "boron-mullite," raising the possibility that the synthetic solid solution proposed by Gielisse and Foster (1961) extends to sillimanite rather than to mullite.
Werding and Schreyer (1996) suspected that "boron-mullite" would be largely metastable over the proposed compositional range, a suspicion borne out by restriction of "boron-mullite" compositions to two narrow bands. Many of the hydrothermal syntheses of "boron-mullite" are within the stability field of dumortierite (Fig. 3) and could be metastable, especially in the presence of quartz. However, the compositions plotted in Figures 1 and 7 are those of "boron-mullite" synthesized at temperatures well above the stability of dumortierite. In addition, the rocks containing the presumed natural "boron-mullite" were affected by temperatures above the stability of dumortierite at <4 kbar. In summary, there is good reason to think that "boron-mullite" has a stability field at high temperatures and low pressures in rocks sufficiently rich in B.
Natural boralsilite
Boralsilite is a relatively widespread, albeit minor, constituent of pegmatites belonging to two generations in the Larsemann Hills, one associated with the D2-D3 deformation and the second with D4 (deformation scheme of Carson et al. 1995). The earlier generation typically forms irregular pods or veins ranging from a few decimeters to a meter or so in thickness; larger bodies extend 100 m or more and are mapped as Rumdoodle pegmatite (Carson and Grew 2007; this paper, Fig. 10). The bodies are either roughly concordant or crosscut S2 and S3 fabrics; in places, they are folded and show an axial-planar fabric. The later generation consists of undeformed, generally planar, cross-cutting veins a few centimeters to a few decimeters in thickness. Characteristic of both generations of pegmatites are graphic tourmaline-quartz intergrowths, commonly associated with bright red microcline (Fig. 11a), but also occurring in plagioclase (Figs. 12a and 12b). Seven of the nine pegmatites in which boralsilite has been identified belong to the D2-D3 generation and six intersect one of the four B-enriched units containing prismatine, grandidierite, and tourmaline; the other three pegmatites cut nearby units instead (Fig. 10). Pegmatites containing prismatine or grandidierite, but not boralsilite, are D2-D3 pegmatites also cutting the four B-enriched units. Early dumortierite is present only in D4 pegmatite veins at locality 121502. In contrast, tourmaline, including graphic intergrowths with quartz, occurs in pegmatites further from the borosilicate-bearing metamorphic units, e.g., between the northern and southern areas of borosilicate-rich rocks on Stornes Peninsula shown in Figure 10.
View larger version (56K):
[in this window]
[in a new window]
|
FIGURE 10. Geologic maps of B-rich metamorphic rocks in the Stornes Peninsula (modified from Carson and Grew 2007) showing localities for boralsilite collected during the 2003–2004 season (open circles; GPS coordinates are given in Appendix Table 3). The holotype sample was collected near locality 121102 (Douglas Thost, personal communication, 1991). Boralsilite was identified using optical microscopy or the electron microprobe except 122005. D4 indicates the one locality where borasilite occurs in second-generation pegmatite veins; the veins at 122005 could also be D4. Prs = prismatine; Tur = tourmaline, Qtz-Fsp = quartzofeldspathic. The order in which the metasedimentary units are listed does not correspond to relative depositional age, which remains indeterminate.
|
|
View larger version (145K):
[in this window]
[in a new window]
|
FIGURE 11. Photographs of mineral associations in pegmatites from the Larsemann Hills, East Antarctica. (a) Graphic tourmaline-quartz intergrowth (black and white) in coarse-grained red microcline, D2-D3 pegmatite at locality 113001, about 100 m SW of locality 112906. Lens cap is 5 cm in diameter. (b) Photomicrograph of twinned prisms of boralsilite overgrown by secondary tourmaline, which is at extinction. Cross-polarized light, sample 120902G. (c) Radiating fibrous bundle of boralsilite (stained yellow above pointer) adjacent to graphic tourmaline-quartz intergrowth in a D2-D3 pegmatite at locality 121102. (d) Photomicrograph of acicular, partially altered boralsilite adjacent to a graphic tourmaline-quartz intergrowth. Boralsilite is largely engulfed in olive and blue secondary tourmaline. Diameter of section is 2.5 cm. Plane polarized light, sample 120405D. (e) Photomicrograph of aggregates of boralsilite in quartz near grandidierite. Secondary andalusite encloses boralsilite (A+B). Pyrite is partially replaced by oxide. Plane-polarized light, sample 112906B. Abbreviations: A = andalusite, B = boralsilite, G = grandidierite, P = pyrite, Q = quartz, T = tourmaline.
|
|
View larger version (163K):
[in this window]
[in a new window]
|
FIGURE 12. Images of mineral associations in the pegmatite from locality 121102. (a) Photomicrograph of graphic tourmaline-quartz intergrowth, prismatine and boralsilite with quartz (not marked) in section 121102C2, plane light. Matrix is plagioclase. (b) Enlargement of the center of (a), but taken with section in a different orientation, which resulted in a difference in the colors of prismatine and tourmaline. (c) Photomicrograph of sillimanite with fringes of werdingite and boralsilite in section 121102B, plane-polarized light. Dark-blue tourmaline is replacing grandidierite. Matrix is plagioclase. (d) Photomicrograph of sillimanite with fringes containing werdingite, boralsilite, and grandidierite in section 121102B, plane light. Matrix is plagioclase (f); area in back-scattered electron image. (e) Back-scattered electron image of a fringe of boralsilite around sillimanite in section 121102B. Matrix is plagioclase with minor quartz (not marked). (f) Back-scattered electron image of a portion of fringe in (d) showing intergrowth of boralsilite, werdingite and grandidierite, and a tiny grain of apatite. Abbreviations: A = apatite, B = boralsilite, G = grandidierite, Pl = plagioclase, Prs = prismatine, Q = quartz, T = tourmaline, W = werdingite.
|
|
Typically, boralsilite forms bundles and sprays of prisms, in places euhedral, in quartz or rarely K-feldspar, most commonly near the graphic tourmaline-quartz intergrowths (Grew et al. 1998a; this paper, Figs. 11c and 11d). Minerals closely associated with boralsilite (Appendix Table 21) in the samples collected in 2003–2004 include plagioclase, from which it is commonly separated by quartz (Fig. 13b), primary dumortierite, grandidierite (e.g., Fig. 11e), werdingite, sillimanite (121102B only), and prismatine (121102C only, Figs. 12a and 12b). Primary dumortierite forms gray bundles of subparallel prisms up to 2 cm long; a few bundles exceed 1 cm in width. Grandidierite forms early, relatively coarse-grained prisms in a few cases (e.g., Fig. 11e), but in other it is a later phase (e.g., Figs. 12b and 12d). Werdingite is present in a single section from locality 121102 (Table 5), which is the sixth known locality worldwide (Grew et al. 1998b; Buick et al. 2006) and the first for Antarctica. Apatite is present in every boralsilite-bearing section (e.g., Fig. 12f); monazite, zircon, and rutile are less common. Grew et al. (1998a) reported boralsilite being replaced by foititic-olenitic tourmaline, diaspore, and kaolinite(?); other secondary minerals in the newly collected samples are andalusite (Fig. 11e) and bright-blue dumortierite, which is found in five of the nine pegmatites. Biotite, cordierite, and wagnerite are also present in boralsilite-bearing pegmatites, but not in close proximity to boralsilite.
View larger version (48K):
[in this window]
[in a new window]
|
FIGURE 13. Back-scattered electron images of boralsilite in pegmatites from the Larsemann Hills, East Antarctica. (a) Boralsilite (type I) shows faint banding parallel to prism length. Sample 010602A. (b) Acicular boralsilite (type II) enclosed in quartz surrounded by feldspars. Sample 010205J4. (c) Aggregate of boralsilite prisms (type I) showing indistinctly lighter cores. Sample 121502E. (d) Aggregate of boralsilite prisms (type 2) showing patchy light areas. Sample 112906B. Abbreviations: B = boralsilite, K = K-feldspar, Pl = plagioclase, Q = quartz, T = tourmaline.
|
|
Sections 121102B and 121102C offer particular insight into the sequence of crystallization of the borosilicates. Primary tourmaline is limited to the graphic tourmaline-quartz intergrowths (Figs. 12a and 12b), in some of which coarse-grained biotite is also found. Prismatine in prisms up to 1 cm in length is present in coarse-grained plagioclase surrounding the graphic tourmaline-quartz intergrowths in 121102C; terminations of the prisms appear to have been embayed, suggesting the prisms might be xenocrysts. Boralsilite is present with quartz in the plagioclase, whereas a few fine grains of grandidierite occur around the prismatine (e.g., Fig. 12b) or in plagioclase. Grandidierite is also later than prismatine in the pegmatite at locality 010205, but in this case, grandidierite is relatively coarse-grained and appears to be a relatively early phase. In section 121102B, grandidierite (Table 5), werdingite, and boralsilite form fringes on aggregates of sillimanite prisms (Figs. 12c–12f). Boralsilite also coarsens away from the aggregates to form independent prisms, e.g., the prism to the left in Figure 12e. In summary, tourmaline in graphic intergrowth with quartz, rare dumortierite, and some grandidierite crystallized early in the Larsemann Hills pegmatites, whereas boralsilite, rare werdingite, and most grandidierite crystallized later, followed by a second generation of tourmaline and dumortierite. Prismatine is an early phase, but we cannot exclude the possibility that it is xenocrystic in the pegmatites, having been plucked from segregations in the source rocks.
In terms of composition, there are two types of boralsilite in the Larsemann Hills pegmatites (Table 3). Type 1, including the holotype specimen, is the more common; prisms are heterogeneous (Figs. 13a and 13c). Electron scattering increases with Si, which varies inversely with B over a narrow range of Si (Fig. 6b). Variation of Al is roughly inverse with Si for Si < 2 atoms per 37 O, but is nearly independent of Si for Si > 2 atoms per 37 O, consistent with very limited solid solution toward sillimanite (Fig. 1).
Type-II boralsilite has been found in two samples (112906, 010205, Fig. 14a) also containing grandidierite (e.g., Fig. 11e); it also appears heterogeneous in BSE images. Variations of Si, (Fe + Mg), Al and B are consistent with a solid solution of up to 30% of an ideal werdingite component, (Fe,Mg)2Al14B4Si4O37, in boralsilite (Fig. 14a). Boralsilite associated with werdingite and grandidierite in sample 121102B is Type I, but with higher MgO and FeO than other Type I (Table 3).
View larger version (18K):
[in this window]
[in a new window]
|
FIGURE 14. (a) Compositions of boralsilite in samples 112906B and 010205J4 from the Larsemann Hills (Type II). (b) Compositions of boralsilite in samples HE138B2 and HE138B3 from Almgjotheii, Rogaland Complex, SW Norway. Squares indicate ideal composition Al16B6Si2O37. The lines indicate ideal substitutions. Three Si per 37 O corresponds to 50% solid solution of an ideal werdingite component (Fe,Mg)2Al14B4Si4O37.
|
|
Boralsilite from Almgjotheii, SW Norway, the only other known locality of the mineral (Grew et al. 1998a), was also re-analyzed using the method described above (e.g., Table 3). Boron decreases with Si, but an increase of (Fe + Mg + Mn) with Si is less evident than in type II from 
Top
Abstract
Introduction
