characterization of synthetic pargasitic amphiboles ...scanning-electron micrographs of high-yield...

American Mineralogist, Volume 72, pages 580-593, 1987

Characterization of synthetic pargasitic amphiboles(NaCarMgoM3*Si6Al2O2r(OHrF)ri M'* : Alo Cr, Ga, Sco In) by

infrared spectroscopy, Rietveld structure refinemenf and27A1, 2esi, and reF MAS NMR spectroscopy

Marr Rluosnrr, Ar,r,.q.NI C. Tunhtocx, FuNr C. HewrnoRNEDepartment of Geological Sciences, University of Manitoba, Winnipeg, Manitoba R3T 2N2, Canada

Banrln-l L. Snrnnrrr'*Department of Geological Sciences, Brock University, St. Catharines, Ontario L2S 3Al, Canada

J. SrnprrnN HlnrrvraxDepartment of Chemistry, Brock University, St. Catharines, Ontario L2S 3A1, Canada

Ansrntcr

Natural and synthetic amphiboles can display long-range and short-range order-disorderthat must be characterized for proper interpretation of synthesis experiments. In this study,octahedral M3* (Al, Cr, Ga, Sc, In) analogues of pargasite and fluor-pargasite were syn-thesized and characterized by optical and scanning-electron microscopy; X-ray diffraction;infrared spectroscopy; 2741, 2eSi, and reF magic-angle spinning (vres) nuclear magneticresonance (Nvrn) spectroscopy; and Rietveld structure analysis. Although "end-member"pargasite was synthesized readily, yields were never quite l00o/o; cell dimensions wereconsistent with previous studies. Substitution of Cr, Ga, Sc, and In for octahedral Alreduced yields to less than 900/0. Variation in cell volume with the cube of the averageoctahedral cation radius for this isostructural series was not linear, indicating incompletesubstitution of these cations for Al. For the pargasites, the infrared spectra showed thatthe octahedrally coordinated trivalent cations were significantly disordered over the M(I,2,3)sites ofthe octahedral strip. In the fluor-pargasites, the trivalent cations tended to be morestrongly ordered at M(2). The ,eSi MAS NMR spectrum of scandium-fluor-pargasite wasincompatible with complete ordering of tetrahedral Al at the T(l) site.

INrnorucrtoN

One of the goals of experimental mineralogy is to growminerals with specific compositions and structures undercontrolled conditions. By characterizing the products,mineral compositions and structures may be relatedquantitatively to growh conditions such as temperatureand pressure. These data are vital to understanding theproperties and phase relations ofanalogous natural min-erals as a function of their environment.

Natural and synthetic amphiboles can display long-range and short-range order-disorder that must be char-acteized for proper interpretation of the results of syn-thesis experiments. Of specific importance here is cationorder-disorder and chain-width order-disorder. This pa-per reports on the synthesis and characterization of oc-tahedral M3* analogues of pargasite and fluor-pargasite,part of a systematic study to critically assess the resultsof amphibole synthesis and to more adequately charac-terize the products (Raudsepp, 1984; Raudsepp et al.,1982; Raudsepp et al., 1984; Hawthorne et al., 1984).

* Present address: Department of Geology, McMaster Uni-versity, Hamilton, Ontario L8S 4Ml, Canada.

0003-004x/87/0506-0580$02.00 580

Our study was restricted to Fe-free, monoclinic amphi-boles with the C2/m structure because previous studies(e.g., Veblen, 198 l; Maresch and Czank, 1983) show thatFe-Mg-Mn amphiboles are not only difficult to synthe-size, but also are subject to short-range chain-width dis-order, high densities of stacking faults, and other localstructural disorder. There is at present no indication thatsuch local structural disorder occurs in synthetic calcicand sodic amphiboles, but little work has been done onthis problem.

Pargasites are monoclinic calcic amphiboles with theC2/m structure (Hawthorne, 1983b). The ideal end-member formula of pargasite is NaCarMgoAl-SiuAlrOrr(OH)r. In this study, synthetic analogues withoctahedral Al replaced by Ga, Cr, Sc, and In were grownfor both the hydroxy- and fluor- end members. Run prod-ucts were characterized by optical and scanning-electronmicroscopy (morphology; detection of foreign phases);X-ray diffraction (cell dimensions; detection of foreignphases); infrared spectroscopy (ordering); 27,{1, 2eSi, andreF rrv{r{s NMR spectroscopy (ordering and site occupan-cies); and Rietveld structure analysis (ordering and siteoccupancies).

RAUDSEPP ET AL.: SYNTHETIC PARGASITIC AMPHIBOLES 5 8 1

Pnnvrous sruDIES

Many workers have synthesized (nominal) pargasite(Boyd, I 959; Gilbert, I 969; Holloway, 197 3; Semet, I 973;Hinrichsen and Schtrmann. l97l: Braue and Seck. 1977:Oba, 1980; Westrich and Holloway, l98l). Additionalphases were often reported as present in minor amounts,but the cell dimensions reported generally agree fairlyclosely with each other. Boyd (1954) and Westrich andNavrotsky (1981) synthesized fluor-pargasite, again withminor amounts of additional phases present in the runproducts.

ExpnmnrnNTAL METHoDS

Synthesis

Dry mixtures of the anhydrous amphibole stoichiometry wereprepared from commercial reagents. SiO, was added as CorningFused Silica Code 7940. The glass was crushed in a steel mortar,cleaned in aqua regia, washed in distilled water and dried at1000'C. SiO, was added to certain fluor-pargasite mixes as de-hydrated H,SiOr.zHrO. A(OH)3 was heated at 900"C for 48 hto yield 7-AlrO3. MgO was dried at 1000"C to constant weight.For hydroxy-pargasites, either Ca was added as CaCO, that hadbeen decarbonated to constant weight before weighing, or Cawas added to the mix as CaCO, and the whole mix was decar-bonated at 1000"C. In fluor-pargasites, Ca was added as CaFr.For hydroxy-pargasites, Na was added as NarSirO, prepared af-ter the method ofSchairer and Bowen (1955); in fluor-pargasites,Na was added as NaF. Cr, Ga, Sc, and In were added as oxides,dried at 400t. End-member pargasite was also prepared as agel after the method of Hamilton and Henderson (1968) usingNarCO., CaCO., Al metal, Mg metal, and SiOo(CrHr)..

Conventional or TZM-alloy cold-seal pressure vessels heatedin horizontal, split-tube furnaces were used for hydroxy-pargas-ite syntheses. About 30 to 90 mg of mix corresponding to theappropriate anhydrous pargasitic amphibole composition plus 5to 2O wo/o double-distilled and deionized water were sealed bywelding into Au capsules and reacted at temperatures between800 and 930qC under l- to 3-kbar water pressure. Pressure ves-sels were quenched under pressure by a jet of cold, compressedair followed by immersion in cold water.

Fluor-pargasite charges consisted ofabout 2Oto 40 mg ofmixwith the appropriate amphibole stoichiometry. These were re-acted isothermally in sealed Pt capsules suspended in vertical,quenching furnaces. Runs were quenched by dropping the cap-sule into a container of cold water below the furnace. Fluor-pargasite syntheses were also attempted at linear cooling rates of1 to 2'C h- ' .

Capsules were weighed before and after runs to check for weightchanges that would indicate capsule leaks. The capsules werethen examined under a binocular microscope for external sigrrsof leakage, especially if postrun and prerun weights were notsimilar. After opening, products were checked for contamina-tion, signs ofreaction with capsule material, texture, and grainsize. A small portion of the product was crushed and mountedon a glass slide for viewing at higher power with a polarizingmlcroscope.

X-ray powder difrraction

X-ray powder diffractograms were routinely taken for all runproducts at a scanning speed of (30"20 min '). These allowedrapid evaluation of the amphibole yield and of the nature and

approximate concentrations of nonamphibole phases. Subse-quently, all runs with high yields of amphibole were scanned at0.620 miri'' to obtain accurate d spacings for cell-dimensionmeasurement and identification of nonamphibole phases.

Powder difractograms were obtained on a Philips AutomatedPowder Difractometer System rwtzto usin-g monochromatizedCu radiation (CuKa, wavelength : 1.5418 A). The finely groundamphibole run product and a small amount of BaF, were blend-ed thoroughly by grinding under alcohol. The mixture was spreadon a glass slide with alcohol to form a thin smear. The BaFrla:6.19860(5) Al was calibrated against Si INBS Standard Refer-ence Material 640a, a: 5.43083(4) Al. Amphibole peaks wereindexed by comparison with published patterns of amphibolesof known structure and composition. Only those reflections thatcould be unambiguously indexed and did not overlap signifi-cantly with neighboring peaks were used in cell-dimension cal-culations. These requirements restricted usable amphibole re-flections to l0 to 12, between 9" and 45"20. Cell dimensions wererefined using the program ofAppleman and Evans (1973).

Scanning-electron microscopy

Scanning-electron micrographs of high-yield run products suchas scandium-fluor-pargasite show no trace ofextraneous phases(Fig. la); however, layer silicates are readily visible (Fig. lb) inscanning-electron micrographs of the run products (scandium-fluor-eckermannite; Raudsepp et al., unpub. ms.) but were notobserved optically or in the X-ray powder-diffraction pattern.Semiquantitative energy-dispersive spectroscopic capabilitywould have been invaluable in the identification ofthese phases.

Infrared spectroscopy

Powdered samples were prepared by grinding about 10 mg ofamphibole by hand in an alumina mortar with ethanol until thegrain size was generally less than 2 pm. This was achieved quick-ly because most of the synthetic amphiboles were initially lessthan 2 pm in grain size. After drying to evaporate the ethanol,3-5 mg of amphibole was mixed with 200 mg of KBr, either byhand grinding in an alumina mortar, or in a dentist's amalgam-ator (Wig-LBug). This mixture was dried under vacuum at 125"Cand was pressed in an evacuated, heated die into a I 3-mm pellet.

High-resolution (2.0 cm ') infrared spectra of hydroxy-am-phiboles in the fundamental O-H stretching region (3600-3800cm r) were recorded on a Nicolet Fourier transform interfero-metric infrared spectrophotometer, model rrlx-t, equipped witha Nicolet 1280 computer for signal processing. The samplechamber was purged with dry N, before and during spectrumcollection. Frequency measurements were calibrated internallyagainst a He-Ne laser and are accurate to 0.01 cm ' accordingto the manufacturer.

Rietveld method of crystal-structure refinement

Specimens were ground with ethanol in an alumina mortarfor at least 5 min and loaded either into an Al holder with aglass insert to support the powder, or into a HF-etched depres-sion in a glass slide. The slurry was worked with a probe so thatit was evenly distributed and dried with its surface precisely flushwith the top of the holder. Grain size was generally less than 5pm with some grains up to 20 pm long. Although most of theseamphiboles crystallized lrith prismatic to acicular habits, pre-ferred orientation was apparently not severe because ofthe ex-tremely fine grain size, and special precautions were not takento eliminate it during sample preparation. Attempts by otherworkers (e.g., Young and Wiles, 1981) to minimize preferredorientation by mixing the specimen with an equal amount of

582 RAUDSEPP ET AL.: SYNTHETIC PARGASITIC AMPHIBOLES

Fig. l. Scanning-electron micrographs: (a) scandium-fluor-pargasite; (b) scandium-fluor-eckermannite (Raudsepp et al., unpub.ms.). Arrow points to layer-silicate crystals. Scale bar represents I pm.

inert filler (ground glass) did not significantly improve refine-ments.

X-ray intensity data were collected on a Philips automateddiffraction system pwlTro equipped with a graphite-crystalmonochromator for CuKa radiation. Beam divergence was con-trolled with an automatic divergence slit so that a constant area(approximately 1.9 cm,) of the specimen was irradiated through-out the scanning range. Intensities were measured at 0.0220 stepswith counting times of either 8 or I 6 s per step; scanning rangeswere between 8' and 7320-

Structures were refined with a slightly modified version of theprogram DBw 2.e (Wiles and Young, l98l) using the modifiedLorentzian peak-shape function (mod2). Refinements were donein three stages. First, the scale factor, cell parameters, and zero-point were refined with atomic positions, site occupancies, andisotropic temperature factors for individual atoms fixed at esti-mated values approximately correct for amphiboles. In this stage,a background model based on inspection was used; other profileparameters were estimated, either from the intensity data or frompublished work, and were not refined. In the second stage, thehalf-width parameters, peak-asymmetry parameter, and pre-ferred-orientation parameter were included in the refinement. Inaddition, the background was refined as a second-order poly-nomial function in 20. Because most of the amphibole diffractionpatt€rn above about 2620 consists entirely of severely over-lapped peaks, background modeling was difrcult. Attempts tomodel the background by extrapolation between areas of lowintensity between peaks failed above 2620. In the third stage,the remaining structural parameters were added to the refine-ment, and the background model function was expanded to thethird order. Background refinements at higher orders failed toconverge. Parameter shifts in the final cycle of refinement weregenerally less than 0.1 to 0.2 sigma. Best refinement results (low-est R factors) were obtained by refining an overall "isotropic

temperature factor" as an approximate absorption-extinctioncorrection. Complete refinement of a typical amphibole structurerequires the simultaneous refinement of 45 to 50 parameters. Inorder to decrease computing time, the refinements were per-formed at 0.0420 steps rather than at the collected interval of0.02'. The results were essentially identical; standard errors wereslightly higher at the larger step interval.

MAS NMR

MAs NMR spectra were obtained on powdered minerals usinga magic-angle spinning probe (Fyfe et al., 1982) and Delrin ro-tors on Brucker wH-4oo and cxp-zoo multinuclear Fourier trans-form Nun spectrometers equipped with 9.4- and 4.7-T super-conducting magnets, respectively. The samples were spun atapproximately 3500 Hz at an angle of 54.7" to the magnetic fi eld.

2eSi uls NMR spectra were recorded at a frequency of 79.46MHz on the wn-+oo spectrometer with 8192 data points using aspectral width of 25 000 Hz and 30' pulses with a 5-s delaybetween pulses. Chemical shifts are reported with reference totetramethylsilane (TMS). "Si chemical shifts were calculated fromthe semi-empirical formula of Janes and Oldfield (1985) thatuses group electronegativities ofadjacent cations and takes intoconsideration variation in mean Si-O-Si angles.

27Al ves NMR spectra were recorded at 104.23 MHz on thewH-4oo spectrometer with 30' pulses, a spectral width of 62 500Hz and 0.3-s relaxation delay between pulses. Chemical shiftsare reported in parts per million with reference to saturatedaqueous [A(HrO)6]3-(CIO.);. The broad'z7Al resonances are ac-curate to within +0.5 ppm.

reF MAs NMR spectra were recorded at a frequency of 188.15MHz on the cxp-zoo spectrometer with 8192 data points usinga spectral width of 100000 Hz and 5-s delay between pulses.Chemical shifts were determined using CuFu e162.9 ppm) as asecondary standard and are reported in ppm to low field ofCFClr.

RAUDSEPP ET AL.: SYNTHETIC PARGASITIC AMPHIBOLES

TABLE 1. Synthesis conditions and products

Run no. Products

583

T P tCc) (bar) (h)

GPA-A2aPA-A2PA-A2A

PA-A8PA-A8A

PA.A1O

CrPA-A3CrPA-A4CrPA-A5

GaPA-A1GaPA-A2GaPA-A3

ScPA-AsScPA-A6

lnPA-44

FPA.NMRFPA-BULFPA-84FPA-85

FCrPA-A1FCrPA-43

FGaPA-A1FGaPA-A2FGaPA-A3

FSCPA-A1FScPA-A3

905 3000846 1000831 2100

758 1000846 1000817 2100

930 1000840 1000500 1000

840 1000s90 1000

801 1000

NaCa.MgoAlSiuAlrOrloH),65 >957o cam + gh + cPx + fo + ne + spl?

384 >95% cam + cpx + fo + pl + ne + spl?891 >95% cam + cpx + fo + pl + ne + spl?

(PA-42 annealed)384 >95% cam + cpx + fo + pl + ne + gh891 >95o/o cam + cpx + fo + pl + ne + gh

(PA-AB annealed)1'126 >957o cam + cpx + fo + pl + ne + spl?

NaCarMgoCrSi6AlrOrr(OH),45 807o cam + cpx + esk + spl + fo + pl + ne70 80%cam + esk + cpx+ p l + fo+ sp l + ne+ l y73 80%cam + esk+ cpx+ p l + fo + sp l + ne+ l y

NaCarM goGaSiuAlrOz(OH),49 30%cam +cpx + p l + fo+ ne + l y70 90%cam + cpx+ pl + fo+ ne + lY73 907o cam + cpx + fo + pl? + ne? + ly?

NaCarMg.ScSi6Al,O,r(OH),70 90% cam + cpx + ScrO3 + fo + ne + Pl70 90% cam + cpx + ScrO3 + to + ne + Pl

NaCa2M94lnSi6AlrOrr(OH),73 90% cam + cpx + gh + fo + InrO3

NaCarMgoAlS6Al,OzF,63 >90o/" cam + cpx + Pl + sPl + ne + fo48 >907o cam + cpx + Pl + spl + ne + fo

308 >90% cam + pl + fl + cpx + spl + fo + ne12O >90% cam + cpx + Pl + sPl + ne + fo

NaCarM goCrSi6AlrOrrF,71 807o cam + mchr + pl + cpx + fo90 80% cam + mchr + pl + cpx + fo

NaCa2MgoGaSi6AlrO,.F271 8'hcam + pl + fo + cpx

382 857o cam + fo + pl + cpx + fl + gl90 857o cam + Pl + fo + cPX

NaCa2M gnScSi6AlrOrrF"71 >957o cam + cox + fo + ne?75 >95% cam + cpx + fo + ne?

834 2000830 2000

838 2000

1000 11000 1

1193-799 1938 1

1006 11000 1

1006 11273-844 1

1000 1

1006 11073 1

Nofe: Abbreviations: cam : clinoamphibole; fo : forsterite; ly : layer silicate; esk : eskolaite; ne :

nepheline; cpx : clinopyroxene; gh : gehlenite; pl: plagioclase; spl : spinel; gl : glass; mchr:magnesio-chromite; fl : fluorite.

ExpnmvrpnrAr, RESULTS

Synthesis products

Synthesis conditions and products for typical pargasiticamphiboles grown in this study are given in Table 1. Onlythose phases that could be unambiguously identified frompowder X-ray patterns, or rarely, optically, are listed.Typical amphiboles grown in this study formed blockyto prismatic crystals I to 20 1rm wide and up to 40 pmlong. Cr-bearing varieties were pale green and slightlypleochroic; indium-pargasite was very pale yellow andfaintly pleochroic; all others were clear and colorless. Celldimensions for pargasites from selected runs are given inTable 2, and full run descriptions are given in Table 3.1

' To obtain a copy of Table 3, order Document AM-87-335from the Business Office, Mineralogical Society of America, 1625I Street, N.W., Suite 414, Washinglon, D.C. 20006, U.S.A. Pleaseremit $5.00 in advance for the microfiche.

Infrared spectroscopy

In the 3800-3300-cm' region, high-resolution in-frared spectra of amphiboles exhibit fine structure that issensitive to the cation occupancies of the M(l) and M(3)sites (Hawthorne, 1983a, 1983b). In binary solid solu-tions, there are eight possible ways of distributing twodifferent cations over the three M sites coordinating eachhydroxyl. In amphiboles, however, the three M sites co-ordinating the hydroxyl are in a pseudotrigonal arrange-ment that introduces an accidental degeneracy to somebands and reduces the number of resolvable bands tofour. In end members, 2M(l) + M(3) configurationsaround each hydroxyl are identical, and a single, sharphydroxyl-stretching band results (Fig. 2). Figure 2 showstypical natural amphibole spectra collected under the sameexperimental conditions as the synthetic amphibole spec-tra. Note the sharp, narrow peaks with band widths be-tween 6 and 9 cm-l, typical values for natural amphibole


TreLe 2. Cell dimensions of synthetic pargasitic amphiboles

Run no. a (A) b (A) c (A) Bf ) v(41

GPA-A2APA.A8APA-A2APA-A1O

CrPA-A3CrPA-A4CrPA-A5

GaPA-A1GaPA-A2GaPA-A3

ScPA-AsScPA-A6ScPA-As-

lnPA-A4

FPANMRFPA-BULFPA.B4FPA.85FPA-BUL"

FCrPA-A1FCrPA-A3FCrPA-A3-

FGaPA-A1FGaPA-A2FGaPA-A3.

FScPA-AlFScPA-As'

9.904(3)9.907(3)9.894(2)s.8e7(2)

9.914(2)9.917(3)9.909(2)

9.849(2)e.e23(3)9.910(4)

9.942(3)9.944(2)9.9405(7)

e.e37(3)

9.830(4)9.827(3)9.818(3)9.820(3)9.8277(7',)

s.834(3)9.845(8)9.8402(5)

9.846(2)9.852(3)e.8600(5)

e.881(2)9.8852(4)

NaCarM glAlSi6Al,Orr(OH),17.941(s) s.281(2117.929(6) 5.282(2)17.948(5) s.280(1)17.946(4) 5.284(11

N aCarM g1C rS i6Al 20 2JOHL17.993(4) 5.285(1)17.998(6) 5.287(2)17.989(5) 5.28s(2)

NaCarM g4GaSi6Al rO 2JOH)217.953(4) 5.297(1117.973(5) 5.292(1)17.976(71 5.289(2)

NaCa2Mg{ScSi6AlrOrr(OH),18.101(5) 5.297(1)18.096(5) 5.298(1)18.094(1) 5.2986(4)

NaCarM 94lnSi6AlrOrr(OH),18.030(4) s.289(2)

Fluor-pargasiteNaCarM g4AlSi6AlrO,,F,

17.919(6) s.294(2)17.927(7) 5.293(2)17.929(7) 5.295(2117.931(s) 5.293(2)17.930(1) 5.2931(4)

NaCa2MglCrSi6AlrOzF,17.971(s) 5.286(1)18.00s(13) 5.284(4)17.9800(9) s.2914(3)

NaCarMg4GaSi6AlrOrrF217.951(4) 5.296{1)17.945(4) 5.299(1)17.968s(9) 5.3027(3)

NaCa,M g4scSi6Al2O,2F,18.145(4) 5.317(1)18.1574(7) 5.3186(2)

105.54(3) 904.0(3)105.51(2) 904.q3)105.50(2) 903.s(3)10s.51(2) 904.4(21

105.44(2) 908.7(2)10s.41(2) 909.8(3)10s.42(21 908.2(3)

10s.17(2) 903.9(2)105.49(2) 909.7(3)105.54(3) 907.9(2)

105.37(2) 919.2(3)105.39(2) 919.2(3)105.364(5) 918.97

105.54(2) 912.9(3)

105.16(3) 900.0(3)105.19(3) 899.8(4)105.27(4) 899.2(4)105.20(3) 899.4(3)105.168(4) 900.21

105.07(2) 902.1(3)105.06(6) 904 4(7)105.108(3) 903.83

105.16(2) 903.5(2)105.25(2) 903.9(2)105.198(3) 906.62

105.17(2) 920.1(21105.213(2) 921 18

* Determined by Rietveld structure refinement.

spectra (Strens, 1974). Of particular interest is the peakshape of the single MgMgMg stretching band in the trem-olite spectrum (Fig. 2a). The shape corresponds to thatof a markedly skewed Gaussian distribution. Figure 2shows a typical natural actinolite spectrum with Mg andFe2* as the predominant octahedral cations. Infraredspectra of synthetic pargasites with M3* : Al, Cr, Ga, andSc are presented in Figure 3. For the ordered case, inwhich the M(2) site occupancy is 0.5Mg f 0.5M3+ andthe M(1,3) sites are occupied solely by Mg, the spectrumshould consist of a single band corresponding to theMgMgMg configuration. However, the spectrum of par-gasite (M3* : Al) consists of two major bands at 3709and3676 cm ', and a weak shoulder at about 3645 cm '.

Because the sample was from a high-yield run and there-fore close to the nominal composition, the two majorbands were assigned to the MgMgMg and MgMgAl con-figurations, respectively; the weak band is probably dueto minor A-site vacancies.

Band widths (Fig. 3) are about 25 to 33 cm ', whichare considerably larger than tlpical values ofabout 6 cm '

for natural amphibole spectra (Fig. 2). The origin of thelarge band width in synthetic amphiboles is not clear. Itis tempting to ascribe it to poor crystal quality (high mo-saicity) inducing positional disorder that resulted in a mi-nor shift of the principal stretching band; however, thepowder-difraction patterns suggest a high degree of crys-tallinity. Additionally, the degree of local disorder inintermediate natural amphibole solutions did not causebroadening of infrared peaks on this sort of scale (Fig.2b), and consequently it is difficult to envision local dis-order as the cause ofbroadening in scandium-pargasite,in which the Mg and Sc are of fairly similar size, and inwhich the amphibole seemed to have good crystallinity.Thus, we cannot propose a convincing reason for the widepeaks in the infrared spectra of these amphiboles.

The frequency shift of the MgMgAl band relative tothe MgMgMg band is -33 cm '. Semet (1973) also re-corded the frequency shift of the MgMgAl band in thesynthetic pargasite spectrum as -33 cm-'. Strens (1974)showed that the frequency shifts of individual bands (inreciprocal centimeters) are a function ofthe electronega-

3700 3650 .

3600

Frequency (cm')

Fig. 2. Typical infrared spectra of (a) natural tremolite and(b) natural actinolite.

tivities of the bonded cations according to the approxi-mate relationship

Lv -- 35n(X-" - X),

where 0 < n < 3 is the number of Mg ions replaced inany one Mgr(OH), cluster, and XMe and X" are the Allred-Rochow electronegativities of Mg and the substituent cat-ion, respectively. This gives a frequency shift of -9 cm-rfor the MgMgAl band relative to MgMgMg, which is sig-nificantly different from the observed shift of -33 cm-'.Clearly, the magnitude of the frequency shift is not asimple function of electronegativity. The frequency shiftsgiven by Strens (1974) were determined mainly from con-figurations involving Mg and divalent cations, so pre-sumably the shifts observed for trivalent cations cannotbe adequately described using this equation owing to thedifferent formal charge of the cations involved.

Infrared spectra of chromium-pargasite and gallium-pargasite (from this study) and magnesio-hastingsite (Se-met, 1973) all have shoulders near 3676 cm I (Fig. 4)corresponding to the MgMgAl configuration in end-member pargasite (Fig. 3). These spectra suggest thatsynthetic pargasites (M:* : Cr, Ga, or Sc) and magnesio-hastingsite (M:+ : Fe3t) are not of the nominal com-positions but contain minor octahedral Al. This discrep-ancy between the nominal composition and thecomposition of the synthetic amphiboles is particularlyevident in the spectrum of chromium-pargasite (M:* :

Cr), which consists of major bands at37l0 and 3659 cm '

and a weak shoulder at 367 4 cm-' (Fig. 3). The two majorbands were assigned to MgMgMg and MgMgCr configu-ratiorls; the shoulder at 3674 cm ' is probably due to theMgMgAl configuration. The spectrum of gallium-pargas-ite (M3t : Ga) is similar. Bands at 3706,3676, and 3665cm I were assigned to the MgMgMg, MgMgAl, andMgMgGa configurations, respectively. Both chromium-and gallium-pargasite spectra exhibit poorly resolved finestructure other than the three principal bands. These maybe due to inadequate resolution or perhaps to the presence

585

.requency (cm r)

Fig. 3. Infrared spectra ofsynthetic pargasites: (a) pargasite;(b) chromium-pargasite; (c) gallium-pargasite; (d) scandium-par-gasite.

of minor layer-silicate impurities. In contrast to chro-mium- and gallium-pargasites, the scandium-pargasite(I11:* : Sc) spectrum is straightforward. The three majorbands at 37 ll. 3679, and 3673 cm-r were assigned to theMgMgMg, MgMgAl, and MgMgSc configurations, re-spectively.

Rietveld crystal-structure refinernent

The Rietveld method (Rietveld, 1967, 1969) uses thewhole powder-diffraction pattern to characterize the

3780 3720 3660 3600

Frequency (cm' )

Fig. 4. Infrared spectra of synthetic pargasites and magnesio-hastingsite. Arrow points to MgMgAl bands additional toMgMgCr, MgMgGa, and MgMgFe3* bands. (a) Chromium-par-gasite; (b) gallium-pargasite; (c) magnesio-hastingsite (from Se-met. 1973).


o

o

oc6F

ooc(It

.=E(t,c(d


TneLe 4. Atomic positions for synthetic pargasitic amphiboles

SCPA-A5 FPA-BUL FCrPA-A3 FGaPA-A3 FSCPA-A3

TABLE 5. Interatomic distances (A) for synthetic pargasitic am-ohiboles

ScPA- FPA- FCrPA- FGaPA- FSCPA-A5 BUL A3 A3 A3

o(1)

o(2)

o(3)

o(4)

o(s)

o(6)

o(7)

r(1)

r(2)

M(1 )

M(2)

M(3)

M(4)

A

x

z

xvz

x

z

xvz

x

z

x

z

xvz

xvz

x

z

x

z

x

z

xvz

0.117(2) 0.1 1 9(1)0.0935(8) 0.0868(6)0.233(21 0.216(2)

0.108(2) 0.116(1)0.1741(6) 0.1676(6)0.712(2) 0.740(2)

0.109(2) 0.106(1)0 00.713(s) O.711(41

0.352(1) 0.366(2)0.2495(6) 0.2522(6)0.770(41 0.782(3)

0.354(1) 0.357(1)0.1401(5) 0.1380(5)0.119(3) 0.126(2)

0.344(2) 0.341(1)0.1091(6) 0.1091(5)0.638(3) 0.631(3)

0.348(2) 0.348(2)0 00.290(5) 0.286(4)

0.2824(81 0.2821(7)0.0820(3) 0.0831(3)0.303(2) 0.309(2)

0.288(1) 0.2953(9)0.1682(3) 0.1721(3)0.81 1(2) 0.809(1)

0 0

0.113(1) 0 .118(1) 0 .1108(9)0.0876(6) 0.0896(5) 0.0864(4)0.212(2) 0.221(21 0.219(2)

0.114(1) 0.106(1) 0.1138(9)0.1703(s) 0.1703(5) 0.1709(4)0.740(2) 0.730(2) 0.724(1)

0.100(1) 0.101(1) 0.1008(9)0 0 00.714(3) 0.709(3) 0.711(2)

0.359(1) 0.359(1) 0.3s98(8)0.2498(5) O2487(5) 0.2468(4)0.792(21 0.790(2) 0.788(2)

0.363(1) 0.350(1) 0.3535(9)0.1379(4) 0.1377(4) 0.1388(3)0.134(2) 0.116(2) 0122(2)

0.342(1) 0.3460) 0.3447(9)0.1116{4) 0.1093(4) 0.1140(3)0.622(3) 0.615(2) 0.620(2)

0.3s6(1) 0.344(1) 0.353(1)0 0 00.2s5(3) 0.278(3) 0.291(2)

0.2817(6) 0.2844(6) 0.2838(5)0.0824(3) 0.0842(2) 0.0833(2)0.301(1) 0.307(1) 0.304(1)

o.2s24(7) 0.2893(7) 0.2909(5)0.1703(3) 0.1707(21 0.1692(2)0.813(1) 0.809(1) 0.8090(9)

0 0 0

1 .600 1.554 1.6041 .709 1.681 1.6681.784 1.715 1.7261 .630 't.642 1 .656

1 .681 1.648 1.664

1.728 1.705 1.6941 .639 1.618 1.5881 .671 1.738 1 .7581 .596 1 .61 0 1.622

1 .6s9 1.668 1.666

r(1)-o(1)r(1)-o(s)r(1Fo(6)r(1)-o(7)(r(1)-o)r(2)-o(2)r(2)-o(4)r(2FO(5)r(2)-o(6)(r(2)-o)M(1)-o(1) x2M(1lO(2) x2M(1)-O(3) x2(M(1Fo)

M(2)-o(1) x2M(2Yo(2) x2M(2)-O(4) x2(M(2Fo)

M(3)-o(1) xaM(3FO(3) x2(M(3FO)

M(4)-O(2) x 2M(a)-O(4) x 2M(a)-o(s) x2M(4)-o(6) x 2(M(4FO)

2.129 2.1092.037 2.08'l2.064 2.0392.077 2.076

2.149 2.1012.011 1.9941.964 2.0222.041 2.039

1.586 1.6511.649 1 .6671.653 1.7251.646 1.6681.634 1.678

1.745 1.6901.575 1 5811.686 1 .7061.700 1 .60s1.677 1646

2.110 2-0722.031 2.0732.027 2.0402.056 2.062

2.088 2.1331.984 2-0752.062 2.0782.045 2.095

2.146 2.0852.042 2.0382.111 2.069

2.387 2.4162.379 2.3682.649 2.6152.686 2.6552.525 2.514

2.0612.0722.0612.065

2.0722.0882.1342.098

2.230 2.097 2.0812.087 2.063 2.0082.182 2.086 2.057

2.351 2.448 2.4202.374 2.296 2.3802.581 2.601 2.5542.740 2.768 2.7072.512 2.528 2.515

x 0 0 0y 0.2811(3) 0.2768(3) 0.2776(31z l h V 2 Y 2

x 0 0.024(3) 0.031(2)y Y 2 V 2 Y 2z 0 0.057(1) 0.072(5)

Nofej lsotropic temperature factors (8): O(1): O(2): O(3): O(a) =0 . 8 ; o ( 5 ) : 0 ( 6 ) : 1 . 1 ; o ( 7 ) : 1 . 2 ; r ( 1 ) : T ( 2 ) : 0 . 4 ; M ( 1 ) : M ( 2 ) :M(3) : 0.6; M(4): 0.9; A : 2.3

structure of a material. The structure parameters of themineral, atomic coordinates, site occupancies, and ther-mal parameters, together with various experimental pa-rameters affecting the pattern, are refined by least-squaresprocedures to minimize the difference between the wholecalculated and observed patterns.

Structures of five pargasitic amphiboles synthesized inthis study were refined with the Rietveld method. Resultsof the structure refinements are summarized in Tables 4to 7. The Rietveld agreement index R* varied betweenI T.4 and 14.60/o. A typical calculated and observed pow-der-diffraction pattern is given in Figure 5 for scandium-fl uor-pargasire (FScPA-A3).

Cell dimensions derived from the refinements are givenin Table 2. In general, they are up to an order of mag-

0.0865(5) 0.0888(s) 0.0879(4) 0.087s(4) 0.0870(3)V2 V2 Y2 V2 y2

0 0o.1749(4) 0.1763(3)0 0

0 00.2786(3) 0.2791(2)V2 Y2

0.017(3) 0.029(2)Y2 Y20.068(5) 0.068(4)

nitude more precise than those derived from normal least-squares refinement of powder-diffraction data (Table 2).With the exception of gallium-fluor-pargasite, cell dimen-sions are identical within 2 or 3 sigma for both methodsof determination for identical samples. Atomic positionsare given in Table 4. Interatomic distances (Table 5) werecalculated using RFINE (Finger, 1969). Individual tetra-hedral bond lengths (Table 5) exhibit ranges ofvariationthat are inconsistent with single-crystal structure data forboth pargasites and alkali amphiboles. Mean tetrahedralbond lengths are better behaved and are reasonable, ex-cept for gallium-fluor-pargasite (4) in which (T(2FO) ismuch larger than (T(IFO). Variation in individual (M-O) with mean ionic radii of constituent octahedral cat-ions is not consistent with trends for single-crystal struc-tures (Hawthorne, 1983b).

In the amphibole structure, there are pseudo4lideplanes ll(010) at y : +Vq, and thus the T(l) and T(2)tetrahedra are pseudosymmetrically related. The diffrac-tions containing information on the differences betweenthe T(l) and T(2) tetrahedra are very weak, and conse-quently the shifts of the atomic positions of the atomsconcerned are highly correlated during least-squares re-finement (Table 6). The derivative information, such asbond lengths, is thus rather imprecise. Similar effects oc-cur in the P2r/c pyroxenes examined by Hawthorne et al.(1984), Raudsepp et al. (1984), and Raudsepp et al. (un-pub. ms.). It the P2,/c pyroxenes, pseudo-C-centeringresults in high correlations of the tetrahedral positions

0.1751(4) 0.1775(5)0 0

0o.1767(4)0

000

000

000

00

000


Scand ium- f luor -pargas i te

587

1 . 0

0 . 5

and unrealistic bond lengths; however, site occupanciesdetermined by Rietveld refinement and Mdssbauer spec-troscopy are in good agreement. For structures in whichsuch pseudosymmetry does not occur (e.g., olivine), thereis good agreement between Rietveld and single-crystaldata for both atomic positions and site occupancies(Raudsepp et al., unpub. ms.). This suggests that the highcorrelations should not affect the site occupancies givenhere for the amphiboles, and hence they should be fairlyaccurate.

Synthetic amphibole site occupancies are given in Ta-ble 7. All negative values differ from zero by less than 3sigma and may be regarded as zero; similarly, occupan-cies greater than 1.0 may be set to 1.0. In general, powderstructure refinements of these synthetic amphiboles werecomparable with most powder refinements of other min-erals (Young, 1980). Rietveld refinement is still a tech-

TneLe 6. Selected correlations from Rietveldrefinement of scandium-fluor-pargasite

lr h'L-il[*0 . 0

i l | i lHi l t l i l t l ilililril lrilrilililnI

0 . 0

1 0 1 5 2 0 2 5 3 0 3 5 L t O L 1 5 5 0 5 5 6 0 6 5 1 0 0 2 e

Fig. 5. Calculated and observed X-ray ditrraction pattern of scandium-fluor-pargasite from Rietveld structure refinement. Opensquares are the observed data, solid line is the calculated pattern, and the vertical bars below the pattern mark all possible Braggreflections (Ka, and Kar). The residual pattern obtained by subtracting the observed and calculated patterns is shown below.

nique undergoing rapid development, and several aspectsare currently not satisfactory:

l. The peak profile function used in the Rietveld anal-ysis did not adequately model the observed peak shape.Diference plots (Fig. 5) show consistent sinusoidal fluc-

TneLe 7. M(1), M(2), M(3) site occupancies ofsynthetic pargasitic amphiboles

Scandium pargasiteMg 1.037(22) Sc -0.037(22)Ms 0.789(20) Sc 0.211(20)Mg 0.7921271 Sc 0.208(27)

Fluor-pargasiteM(1 )M(2) (occupancies indeterminate)M(3)

ChromiumJluor-pargasite0.920(10) Cr 0.080(10)0.840(11) Cr 0.160(11)1.061(14) Cr -0.061(14)

GalliumJluor-pargasite1.002(6) Ga -0.002(6)c.900(6) Ga 0.100(6)0.962(8) Ga 0.038(8)

ScandiumJluor-pargasite0.922(111 Sc 0.078(1 1)0.s97(12) Sc 0.403(12)1.00s(15) Sc -0.00s(15)

M(1)M(2)M(3)

o(4)-o(7)o(1)-o(2)o(5Fo(6)r(1)-r(2)

-0.25-0.54-0.60-0.69

M(1) MgM(2) MgM(3) Ms

M(1) MgM(2) MgM(3) Ms

M(1) MgM(2) MgM(3) Mg

0 -0.620.25 -O.370.18 -0.030.04 -0.66


'uAl

u t A l

1 5 0 1 0 0 5 0 0 - 5 0

ppmFig. 6. 2'Al r"res NMR spectrum of synthetic scandium-fluor-

pargasite; Y denotes spinning sidebands.

tuations that reflect inadequate fits at the peak base. Theseare best observed for the (020) and (l l0) peaks between9.5o and 10.6'20.

2. The single refinable asymmetry parameter did notadequately model peak asymmetry over the entire 20range. Best results (lowest R*o values) were obtained byrefining the asymmetry parameter only below about3220;refining this parameter for the whole patern gave higherR values.

3. All of the samples contained from I to l0o/o extra-neous phases that contributed to the overall pattern. Peaksdue to phases other than amphibole are marked by ar-rows in Figure 5 and listed in Table l. The refinementprogram allows regions ofextraneous intensity to be ex-cluded, but extra intensity under amphibole peaks couldnot be excluded in this way; this raises the R factors sig-nificantly.

Conversely, the residual pattern that remains (Fig. 5)after the observed diffraction pattern is subtracted fromthe calculated pattern is valuable because it comprises thediffraction patterns of phases other than amphibole in therun product. In the whole pattern, the scattering contri-butions of these phases are partly or completely maskedby the intense amphibole pattern; this may lead to thefalse assumption that the amphibole yield is near l00o/0.With the current profile function, the interpretation ofthis residual pattern is not straightfoward because the dif-fraction patterns of extraneous phases are mixed with re-sidual intensity derived from inadequate p€ak-shapemodels. However, this situation will improve when moreadequate peak-shape algorithms become available.

Use of the Rietveld method ensures that the amphibolepattern is correctly indexed and that non-amphibole peaksare not indexed as amphibole peaks. This is of more im-portance than one would initially think, as several pub-lished synthetic-amphibole diffraction patterns have in-

Fig. 7. '?esi \,r,{s NMR spectra: (a) natural tremolite; (b) scan-dium-fluor-pargasite.

correct indexing and wrong cell dimensions as a result of

the presence of non-amphibole peaks.

MAS NMR

The'??Al spectrum of scandium-fluor-pargasite is shownin Figure 6. The 'z7Al chemical shifts are known to be

sensitive to the coordination of Al (Mueller et al.' l98l);shifts around 0 ppm correspond to octahedrally coordi-nated Al, and shifts around 70 ppm correspond to tet-rahedrally coordinated Al. The major peak in Figure 6 is

at 68.5 ppm, corresponding to tetrahedrally coordinatedAl as expected for a pargasitic amphibole. This centralpeak is flanked by two pairs of spinning sidebands (marked

as such in Fig. 6) and an additional very small peak at-3.8 ppm. The latter corresponds to octahedrally coor-

dinated Al, and the relative intensities of the two peaks

indicate that it represents about 40lo of the total Al. Note

that the infrared spectrum of scandium-pargasite (the hy-

droxy analogue) also indicates the presence of a small

amount of octahedrally coordinated Al (Fig. 3).The 'zeSi spectrum of a natural tremolite is shown in

Figure 7a. There are two resolved peaks of approximatelyequal intensity at -87.2 and -90.2 ppm, respectively,corresponding to Si at the T(2) and T(l) sites in the struc-ture; a similar spectrum is given by Smith et al. (1983).

The peak at -87 .2 ppm is assigned to the T(2) site with

two adjacent silicate tetrahedra, and the peak at -90.2

ppm is assigned to the T(l) site with three adjacent tetra-hedra, in accordance with the shift to high field with con-densation (Lippmaa et al., 1980). This peak assignmentis corroborated by the calculations ofJanes and Oldfield(1985) of the chemical shift in tremolite (Table 8). The

diferences between our measured values and Janes andOldfield's calculated and measured values of chemicalshift are consistent with variations in composition and

order in the two natural samPles.The'zeSi spectrum of scandium-fluor-pargasite is shown

in Figure 7b. There is a broad resonance centered at about-85 ppm apparently consisting of several overlappingpeaks. This spectrum was resolved into a series of Lor-entzian peaks by the method of least-squares, using theconstraint that the half-widths (full-width at half-maxi-mum) of all peaks be equal. The minimum possible num-

ber ofcomponent peaks in the spectrum is four, and so-

- a 0 - 9 0 - 1 0 o - 1 1 oppm

RAUDSEPP ET AL.: SYNTHETIC PARGASITIC AMPHIBOLES 589

TneLe 8. Observed and calculated 2esi MAS NMRchemical shift for amphiboles

Calculated. Measured

r(1) (3si)r(2) (2Si)

r(1) (3sDr(1 ) (2sil ADr(1) (1Si2Ar)r(2) (2sDr(1) (3ADr(2) (1 SiI ADr(2) (2AD

Tremolite-93.8 -92.21 -90.2-83.9 -87.8f -87.2

Fluor-scandium-pargasite-90.0-86.8-82.7-82.5-79.3-78.4-74.3

- 8 0 - 9 0 - 1 0 0ppm

- a o - 9 0 - 1 o oppm* From the relationship of Janes and Oldfield (1 985);

shifts in parts per million relative to TMS.t Janes and Oldfield (1985), Table 1 1 (data from Magi

et al.. 1984).

lutions were calculated for four, five, six, and seven peaks.The calculated spectra are shown in Figure 8 and may becompared with the observed spectrum in Figure 7b. Thepositions and calculated intensities of the componentpeaks in these spectra are listed in Table 9.

The local arrangements of atoms around the T(l) andT(2) positions are shown in Figure 9. Using the labelingof Figure 9, the possible next-nearest-neighbor arrange-ments are shown in Table l0; there are eight possiblearrangements for T(l) and four possible arrangements forT(2), labeled I to 12, respectively. Peaks resulting fromarrangements 2,3,and 5 will closely overlap, as will thosefor arrangements 4, 6, and 7 and arrangements 10 andI 1, reducing the number of possible resolved peaks toseven. The number of peaks could be further reducedbecause of the energetic unfavorability of certain arrange-ments or by accidental overlap. The resolution of thespectrum is not really sufrcient to a priori decide on thenumber of peaks that actually occur. However, we mayconsider possible arrangement models, and their com-patibility with the observed spectmm may allow us toaccept or reject certain ofthese arrangements. Therefore,with regard to Al-Si ordering over the tetrahedral sites,we may recognize in general four possible models:

Model 1. All Al ordered at T(1), no Al-O-Al linkagesallowed. This model would produce four peaks, numbers5, 9, (10 + I l), and l2; as the amount of Si at T(l) wouldbe half that arT(2), the intensity of peak 5 would be one-third of the total intensity, assuming that the transitionprobability is the same for each arrangement.

Model 2. All Al ordered at T(1), Al-O-Al linkages al-lowed. This model would produce five peaks, numbers l,5, 9, (10 + I l), and 12;the sum ofthe intensities ofpeaksI and 5 would be one-third of the total intensity.

Model 3. Al disordered over T(1) and T(2), no Al-O-Al linkages allowed. This model would produce six peaks,numbers 5, (6 + 7), 8,9, (10 + I l), and 12; the sum ofthe intensities of 5 + (6 + 7) + 8 would be half the totalintensity for complete Al disorder.

Model 4. Al disordered over T(1) and T(2), Al-O-Al

Fig. 8. (a) Four, (b) five, (c) six, and (d) seven peak fits to the'?esi MAs NMR spectrum ofpargasite. Calculated spectrum is shown;experimental envelope shown in Fig. 7b.

linkages allowed. All seven peaks would occur, with thesumoftheintensitiesof I + (2 + 3 + 5) + (4 + 6 + 7)equal to half of the total intensity for complete Al dis-order.

First, we consider the case for which there is no acci-dental overlap of peaks, and the high- to low-field se-quence of peaks is as calculated in Table 8. Comparisonof the observed and ideal intensities for the various num-ber of peaks in each of the above models is shown inTable 9. By far the best agreement occurs for model 3 inwhich Al is disordered over T(l) and T(2) (but there areno Al-O-Al linkages) and six peaks result. Next, we con-sider the possibility of accidental positional degeneracy.Examination of Table 8 shows that arrangements (4 +

TneLe 9. Peak fits to the 2ssi MAS NMR spectrum of scandium-fluor-pargasite

Peak IntensityPeak intensity observed

position (arbitrary (% ot(ppm) units) total)

ldeal Differencemodel lobs.value ideall(/") (v"l

No. ofpeaksfitted

- 89.9 17 339-86.1 40328-83.0 23564

80.2 12 505-90 .3 15103-86.8 24865

84.6 2699381 .6 19 531

-78.7 4 991-90.5 13217

87 .7 14 89185.6 27222

-83.3 1624881.2 14243

-78 .3 4190-91 .1 1069-89.3 1871- 86.8 17 956-85.1 21 856-83.1 14481-81.0 12341

78.2 4228

19 33 14

'I

j o o 3 3 1 1

5049

1 0

T ( 2 ) "

1 ) - T ( 1 ) ^

T ( 2 ) "

T ( 1 )

IT ( 2 )

IT ( 1 )

T (

590

Fig. 9. Next-nearest-neighbor (tetrahedral site) arrangementsabout the T(1) and T(2) sites in clinoamphiboles.

6 + 7) tT(l) : lsi2All and 9 [T(2) : 2Si] have approx-imately the same calculated position and quite possiblycould result in overlapping peaks. This being the case,models I and 2 would contain the same number of peaks,as they do not have any arrangements involving peaks 4,6, ard/or 7. On the other hand, models 3 and 4 do haveboth peak (4 + 6 + 7) and peak 9, and overlap wouldreduce the number of resolved peaks to five and six, re-spectively. For model 2, the intensity of the first two re-solved peaks [5 + (6 + 7 + 9)l is ideally 500/o * intensityofpeak 9; thus the ideal intensity is significantly greaterthan 500/o; the observed value (Table 9, five-peak fit) is44o/o, and the deviation is thus significantly greater than6010. For model 4, the above-mentioned overlap results insix peaks, and the ideal intensity of the first three peaksmust be significantly greater than 500/o; the observed val-ue is 600/0, so model 4 also seems possible.

Summarizing the above results, the'zeSi MAS NMR spec-trum of scandium-fluor-pargasite shows much betteragreement for ordering models involving Al disorder atthe T(l) and T(2) sites; in addition, the agreement is bet-ter for a model involving no IvAl-O-tvAl linkages.

teF vras NMR spectra of tremolite (Fig. l0a) and fluor-scandium-pargasite (Fig. lOb) consist of one very broadpeak at - 17 | .7 and - 169 .6 ppm, respectively, with mul-tiple spinning sidebands. The F is in a more asymmetricenvironment in the pargasite than in tremolite as shownby the larger intensity of the spinning sidebands in thepargasite spectrum. This is probably due to the presence

Tnale 10. Possible next-nearest-neighbor ar-rangements in pargasite

Arrangement A B


- 1 0 0 - 2 5 0-200

1 S i2 S i3 S i4 S i5 A l6 A l7 A l8 A l

9 S i10 s i1 t A l12 A l

T(1) siteD I

AIJ I

AIJ I

AISiAI

siSiAIAISiSiAIAI

ppm

Fig. 10. reF MAs NMR spectra of (a) tremolite; (b) fluor-scan-dium-pargasite.

of Sc at some M sites and positionally disordered Na atthe A(2) and A(m) sites, locally destroying the mirrorsymmetry at the O(3) site. The principal peak in the par-gasite spectrum has some structure at the top; the shoul-ders on the inner sides ofthe spinning sidebands suggestthat there are two distinct bands of unequal intensity.These can be assigned to F coordinated by MgMgMg andMgMgSc respectively. Additional broadening in bothspectra could be caused by dipolar coupling between non-bonded adjacent F atoms.

DrscussroN

Synthesis

End-member pargasite was readily synthesized butyields were never quite 1000/o; minor phases could alwaysbe observed either in ser'r photographs or in powder-dif-fraction patterns. However, all pargasite studies give fair-Iy consistent results, and the cell dimensions reported bymost studies are fairly close. Ignoring the outlying values,the total spread on each of the axial lengths is

RAUDSEPP ET AL.: SYNTHETIC PARGASITIC AMPHIBOLES 5 9 1

phibole yields were about 990/0, there was always somematerial observable by srrr,r that appeared not to be am-phibole. It is possible that this small amount of otherphase(s) was due to starting-mix contamination duringthe preparation procedure or to the minor contaminantsalways present in reagent-grade materials.

Substitution of Cr3*, Ga, Sc, and In for vIAl reducedyields to 80-900/0, and additional phases were prominentin the powder-diffraction patterns. Deviation from nom-inal composition was demonstrated by the infrared spec-tra of this series, all of which showed MgMgAl bands inaddition to the MgMgM3* bands in the principal hydrox-yl-stretching region (Figs. 3, 4). In addition, the total Sccontent of the scandium-pargasite was derived from theRietveld structure refinement (0.65 atoms pfu); the nom-inal value is 1.0, indicating an vIAl content of 0.35 atomspfu. Another indication of compositional deviation fromnominality is given by the type I stability diagram forthis series. Prewitt and Shannon (1969) have shown thatin an isostructural series, the cell volume is a linear func-tion of the cube of the octahedral radius of the variablecation. Although Hawthorne (1978) has shown that thisrelationship is intrinsically nonlinear, the variation incation radius in this series is sufficiently small that thelinear model is adequate. Because the occupancy of thevariable cation was generally less than nominal, this plotwas modified to include the cubed average of all octa-hedral cation radii calculated from site occupancies. Fig-ure I I shows this diagram for the pargasites synthesizedhere; the ideal line is drawn through pargasite and throughthe datum point for the scandium-pargasite using itsRietveld-derived composition. As is apparent from Fig-ure I l, pargasites with Cr and Ga substituted for "'Al lieoffthe line, showing that they deviate from their nominalcomposition. This argument does assume that the A siteremains full (and hence the volume changes are not afunction ofA site occupancy). In the pargasites with Crand G4 there are no A-site vacant bands of any signifi-cance in the infrared spectra, indicating that the A sitesare almost or completely filled; the pargasite itself maybe an exception with its weak shoulder at about 3645cm-'. In addition, the deficiency of substituent cations inthe octahedral strip could not be compensated by vari-able alkalis at the A site, as the charge variation is in thewrong sense. Consequently, these pargasites should con-tain significant vIAl that is not indicated by their nominalcompositions.

The results of fluor-pargasite syntheses largely paral-leled those ofhydroxy-pargasites, except for higher yieldsof the scandium-fluor-pargasite compared to scandium-pargasite. The complete series of G4 Cr, and Sc substi-tutions for octahedral Al was characterized by Rietveldstructure refinement. Site occupancies (Table 7) show thatchromium-fluor-pargasite and gallium-fluor-pargasite aredeficient in Cr and Ga, respectively, compared to thenominal composition, but that scandium-fl uor-pargasitehas nearly the ideal Sc content. Variation of cell volumewith the cube ofthe average octahedral cation radius (Fig.

("r,.r,.,)t (At)

Fig. I 1. Cell volume vs. the cube ofaverage octahedral cat-ion radius for pargasites (squares) and fluor-pargasites (circles).The hollow triangle represents magnesio-hastingsite synthesizedon the cuprite-tenorite buffer (Semet, 1973). The hollow symbolsdenote nominal compositions, the filled symbols denote com-positions derived from the Rietveld site-occupancy refinements;pargasite and fluor-pargasite are presumed to be "on composi-tion."

1 1) is not linear, but is improved over hydroxy-pargas-ites.

Cation ordering

Much effort was expended on these pargasite series notonly because of the success of previous work, but alsobecause this composition is intrinsically interesting withrespect to M3* ordering at the octahedral sites and Alordering at the tetrahedral sites.

For the pargasites, the infrared spectra (Fig. 3) showthat the octahedrally coordinated trivalent cations weresignificantly disordered over the M(1,2,3) sites of the oc-tahedral strip. There was a significant change in the ratioof the MgMgM3*: MgMgMg band intensities across theseries. This could be caused by changes in both orderingand clustering ofcations coordinated to the hydroxyl and,in the absence of a specific model for clustering, cannotbe used to derive the degree of ordering (Whittaker, 1979).It is extremely interesting that in natural pargasites, theoctahedrally coordinated trivalent cations are nearlycompletely or completely ordered at M(2), whereas thesesynthetic pargasites were significantly disordered. We triedto determine if this was a kinetic effect by comparingsynthesis runs of 27 and 1126 h. There was no discernibledifference in the infrared spectra of these two run prod-ucts; the product of the longer run might have been slight-ly coarser in grain size, but even this was difficult to de-cide. In addition, infrared spectra of pargasites (PA-A2A,PA-A8A, Table l) annealed for 891 h at 500 and 590{,respectively, were identical to spectra collected before an-nealing. Consequently, if the ordering of Mr* at M(2) innatural amphiboles is a postcrystallization effect, the ki-netics were far too slow for it to be observed in our ex-

2 2 6 q)2

, ^ ot d 6

o)(t'14 ?I

' " t r

9 1 4

9 1 0

592 RAUDSEPP ET AL,: SYNTHETIC PARGASITIC AMPHIBOLES

periments. Conversely, if the natural pargasites crystal-lized in an ordered configuration, the reasons for thedifference in the ordering patterns ofthe natural and syn-thetic amphiboles are certainly not clear at present.

In the fluor-pargasites, the trivalent cations (with theexception of Al for which we have no information) tend-ed to be more strongly ordered at M(2) than in the par-gasites. Rosenberg and Foit (1977) observed similar be-havior for Fe2* in amphiboles and micas, whereby Fe2*and, apparently, Cr3*, Ga, and Sc avoid sites coordinatedby F anions.

The results for the possible order-disorder relation-ships for tetrahedrally coordinated Al are much moreambiguous. In general, the (T-O) distances (Table 5) fromthe Rietveld refinements indicate complete ordering ofI"Al at T(l) (except for gallium-fluor-pargasite). However,the individual T-O distances are known to be stronglyaffected during the refinement by local pseudosymmetry,and seem to be very imprecise. Thus the significance ofthe (T-O) values is questionable.

The '?eSi spectrum seems incompatible with completeordering of 'vAl at T(1), as this should give two or threewell-separated peaks. The component peak intensities aresensitive to both ordering and clustering, and calculationof site occupancies from component peak intensities mayrequire additional techniques sensitive to Al-Si ordering.

CoNcr,usroNs

The.present study has underlined the importance ofdetailed characteization of amphibole synthesis prod-ucts, particularly with regard to cation order-disorder re-lationships. Of particular interest are the following points:

l. Infrared spectroscopy in the hydroxyl-stretching re-gion is a rapid and versatile technique in the character-ization of cation order-disorder at the M(l) and M(3)sites in synthetic amphiboles. Furthermore, deviationsfrom nominal composition can also be revealed with somesensltrvrty.

2. Rietveld structure reflnement can be used to char-acleize cation order-disorder when there is a significantdifference in the scattering power of the constituent cat-ions. This is particularly useful for fluor-amphibole workin which infrared spectroscopy cannot be used for thispurpose.

3. The MAS NMR 2eSi spectrum shows good agreementwith models involving Al disorder over the T(l) and T(2)sites, with best agreement for a model involving no I"Al-

O-tuAl linkages. However, these results cannot be regard-ed as conclusive because ofthe complex spectrum over-lap and the unknown effect ofoctahedral-cation variationon the spectrum.

4. The characterization methods used in this work haveshown that the amphiboles synthesized have orderingpatterns distinctly different from their natural analogues.The proper characteizalion of these products assumesadditional importance if thermodynamic properties areto be measured. as the Dresent work shows that the state

of order in a synthetic amphibole cannot be assumed byanalogy with natural minerals.

AcrNowr,nncMENTS

Financial assistance was provided by the Natural Sciences and Engi-neering Research Council of Canada in the form ofoperating gants to A.C Turnock, F. C. Hawrhorne, and J. S. Hartman. B. L. Sherriff was

supported by an Ontario Graduate Fellowship. We thank C. A. Fyfe, G'

J Kennedy, and R. E. Lenkinski of the University of Guelph for assis-lance with instrumentation and the Southwestern Ontario High Field NMRCentre for instrument time. We thank David M. Jenkins and an anony-mous reviewer for thorough reviews ofthe original manuscript.

RnrnnrNctsAppleman, D.8., and Evans, H.T. (1973) Job 9214: Indexing and least-

squares refinement ofpowder diffraction data. U.S. National TechnicalInformation Service, Document PB 216 188.

Boyd, F.R. (1954) Amphiboles. Carnegie Institution of Washington YearBook 53, 108-l I I

- ( 1959) Hydrothermal investigations of amphiboles. In P.H. Abel-son, Ed., Researches in geochemistry, p. 377-396. Wiley, New York.

Braue, Wolfgang, and Seck, H.A. (1977) Stability of pargasite-richterite

solid solutions at I kb water vapor pressure. Neues Jahrbuch fiir Mi-neralogie, Abhandlungen, 130, 19-32.

Charles, R.W. (1980) Amphiboles on the join pargasite-ferropargasite.

American Mineralogist, 65, 99G100 IFinger, L.W. (1969) Determination of cation distribution by least-squares

refinement of singJe-crystal X-ray data. Carnegie Institution of Wash-ington Year Book 67, 216-217.

Fyfe, C.A., Gobbi, G.C., Hartman, J.S., knkinski, R.E., O'Brien, J'S.,Beange, E.R., and Smith, M.A.R. (1982) High resolution solid stateMAs spectra of'?esi, '??Al, "B and other nuclei in inorganic systems usrnga narrow bore 400 MHz high resolution NMR spectrometer. Journalof Magnetic Resonance, 47, 168-17 8

Gilbert, M.C. (1969) Reconnaissance study of the stability of amphibolesat high pressue. Carnegie Institution of Washington Year Book 67,167-170.

Hamilton, D.L., and Henderson, C.M.B. (1968) The preparation of sili-cate compositions by a gelling method. Mineralogqc'al Magazine, 36,832-8 38.

Hawthorne, F.C. (1978) The relationship between cell volume, meanbondlength and effective ionic radius. Acta Crystallographica, A34, 1 39-140.

- ( I 983a) Quantitative characterization of site-occupancies in min-erals. American Mineralogist, 68, 287 -306.

-(1983b) The crystal chemistry of the amphiboles. Canadian Min-eralogist, 21, 17 3-480.

Hawthorne, F.C., Raudsepp, Mati, Williarns, 8.L., and Hartman, J.S.(1984) Characterization ofcation ordering in synthetic amphiboles byRietveld structure refinement and 27Al and'e Si MAs NMR spectroscopyGeological Association of Canada/Mineralogical Association of Cana-da Program with Abstracts, 9, 72.

Hinrichsen, Thomas, and Schitu-mann, I

synthetic amphiboles on the join Mnl+Mg, "[Si8Orr](OH),.

AmericanMineralogist, 68, 7 44-7 53.

Mueller, D., Gessner, W., Behrens, H.J., and Scheller, G. (1981) Deter-mination of the aluminum coordination in aluminum-oxygen com-pounds by solid state high resolution 27 Al NMR. Chemical Physics I-et-ters,79, 59-62.

Oba, Takanobu. (1980) Phase relations on the tremolite-pargasite join.Contributions to Mineralogy and Petrology,

'l |, 247-256.Prewitt, C.T., and Shannon, R.D. (1969) Use of radii as an aid to under-

standing the crystal chemistry of high pressure phases. American Crys-tallographic Association Transactions, 5, 5 l-60.

Raudsepp, Mati (1984) Evaluation ofamphibole synthesis and productcharacterization. Ph.D. thesis, University of Manitoba, Winnipeg.

Raudsepp, Mati, Turnock, A.C, and Hawthorne, F.C. (1982) Synthesisand characterization of "iR3* analogues of pargasite and eckermannite.Geological Association of Canada/Mineralogical Association of Cana-da Program with Abstracts, 7, 75.

Raudsepp, Mati, Hawlhorne, F.C., and Turnock, A.C. (1984) Derivationof site-occupancies in synthetic pyroxenes and arnphiboles by Rietveldstructure refinement. Geological Society of America Abstracts withPrograms, 16,630.

Rietveld, H.M (1967) Line profiles ofneutron powder-diffraction peaksfor structure refinement. Acta Crystallographica, 22, I 5 l-152.

-(1969) A profile refinement method for nuclear and magneticstructures Joumal of Applied Crystallo$aphy,2, 65-71.

Rosenberg, P E., and Foit, F.F. (1977) Fe,t-F avoidance in silicates. Geo-chimica et Cosmochimica Acra. 41. 345-346.

Schairer, J.F., and Bowen, N.L (1955) The system KrO-Al2O3-SiOr.American Journal of Science. 245. 193-204.

Semet, M.P. (1973) A crystal-chemical study of synthetic magnesiohas-tingsite. Amencan Mineralogist, 58, 480-494.

593

Smith, K.A., Kirkpatrick, R.J, Oldfield, Eric, and Henderson, D.M. (1983)High-resolution silicon-29 nuclear magnetic resonance spectroscoplcstudy of rock-forming silicates. American Mineralogist, 68, 1206-1215.

Strens, R.G.J (1974) The common chain, ribbon, and ring silicates. InV.C. Farmer, Ed., Infrared spectra of minerals. Mineralogical Society,London.

Veblen, D.R. (1981) Non-classical pyriboles and polysomatic reactions inbiopyriboles. Mineralogical Society of America Reviews in Mineralogy,9A, t89-236.

Westrich, H.R., and Holloway, J.R. (1981) Experimental dehydration ofpargasite and calculation of its entropy and Gibbs energy. American

Joumal of Science, 281,922-934.westrich, H.R., and Navrotsky, Alexandra. (1981) Some thermodynamic

properties of fluorapatite, fluorpargasite, and fluorphlogopite. Ameri-can Journal ofScience, 281, l09l-l 103.

Whittaker, E J.W. (1979) Clustering of cations in amphiboles. Physics andChemistry of Minerals, 4, l-10.

Wiles, D B., and Young, R.A. ( 198 l) A new computer program for Riet-veld analysis ofX-ray powder diffraction patterns. Journal ofAppliedCrystallography, 14, 149-15 L

Young, R.A. (1980) Structural analysis from X-ray powder diffractionpatterns with the Rietveld method. Symposium on Accuracy in PowderDiffraction, National Bureau of Standards, Gaithersburg, Maryland,143-163.

Young, R.A., and Wiles, D.B. (1981) Application of the Rietveld methodfor structure refinement wrth powder diftaction data. Advances in X-rayAnalysis, 24, l-23.

MaNuscnrvr rucErvED Mev 29, 1986Meruscnrpr AccEprED Jm;lnv 17, 1987


characterization of synthetic pargasitic amphiboles ...scanning-electron micrographs of high-yield...

Documents