6t3

Conadian Mineralogist Vol. 30, pp. 613-631(1992)

THE WODGINITEGROUP. II. CRYSTALCHEMISTRY

T. SCOTTERCIT', PETREERNf ANDFRANK C. HAWTHORNE Department of Geological Sciences,University of Manitoba, Winnipeg, Manitoba R3T 2N2

CATHERINE A. MCCAMMON2 Department of Physics, University of Manitoba, Winnipeg, Manitoba R3T 2N2

ABSTRACT

The wodginite group of minerals has been studied by electronmicroprobe, X-ray powder diffractometry, X-ray single-crystalmethods, Mcissbauer spectroscopy and atomic absgrption spectroscopy.Ideal wodginite is MnSnTa2O6; however, wodginite also can host major amounts of Fe2+, Fd*, Ti, M and Li, and minor to trace amounts of Snz+, Zr, Sc, W and Ca. Vacanciesoccur at the.4 (Mn) site and seemlikely for the I (Sn) site. Three homovalent and four heterovalentmechanisms of substitution dominatewodginite chemistry,resulting in the four known species of the wodginite group plus syntheticMn2Fe3+Ta5Ol5, which is not yet recognizedas a natural species.A procedure for the calculationof formulae has beendeveloped for electron-rnicroprobedata; it permits calculationof Li contents and the Fe2+:Fe3+ratio. There is a strong correlation betweenstructural and compositionalproperties of ordered membersof the wodginite group. By multiple regressionmethods, a seriesof equationshave been developedfor the calculation of unit-cell parametersfrom compositional data for ordered members of the wodginite group; these equationsgive unit-cell parameterswith root-mean-squareerrors averaging0.007 A on cell edgesand 0.04o on the B angle. Natural membersof the wodginite group are commonly partiany disordered.The B angle, cell volume and the ratio I(021)/1Q20)are most sensitiveto cation order. An index of order has been derived by comparing measured and calculatedI angles,and it givesthe degreeof order to within 590. Keywords:wodginite group, order-disorder, crystal chemistry,.

SoMMAIRE

Nous avons 6tudi6 les min6raux du groupe de la wodginite par microsonde6lectronique, diffraction X (m6thode despoudres et sur cristauxuniques), spectroscopie de Mrissbaueret spectroscopiepar absorptionatomique. La formule id6ale de la wodginite est MnSnTa2Os;toutefois, la wodginite acceptecour:rmment des proportions imponantes de Fd+, Fd+, Ti, Nb et Li, et des quantitdsmoindres de Sn2+, Zr, Sc, W et Ca. Des lacunespeuvent aussi impliquer la position.4 (Mn), et semblentprobables dans la position B (Sn). Trois m€canismesde substitution homovalenteet quatre de substitutionh6t6rovalente r6gissent le champ de compositionde la wodginite, et m&nentaux quatre espdces connuesdans ce groupe, en plus de Mn2Fe3+Ta5O15,,eui n'a toujours pas 6t6 trouv6 dans la nature. Nous ddcrivons un protocole de calcul de la formule pour les compositionsd6termin6es d la microsonde6lectronique, afin d'obtenir la teneur en Li et le rapport Fez+:Fe3*. Il existe une forte corrdlation entre les propri6tds structurales et compositionnellesdes membres ordonn€s du groupe de la wodginite. Nous nous sommesservis de rdgressionmultiple pour obtenir une s6ried'dquations visant d calculerles parambtresr6ticulaires i partir de donndessur la composition des membresordonn€s du groupe. Ces €quationsp.r.itt.tt de calculer les parambtresavec des erreurs'de0.007 A et 0.04osur lesdimensions et I'angleB, respectivement.Les membresnaturels de cegroupe montrent assezcouramment un d6sordrepartiel. L'angle B, le volume de la maille et le rapport I(O2L)/I(220)sont les plus sensiblesau degrd d'ordre cationique.Un indice du degr6d'ordre, bon i 590 prbs, ddcoulede la diffdrencedans les angles0 mesur6et calcul6.

(Traduit par la R6daction) Mots-cl4s:groupe de la wodginite, ordre-ddsordre,chimie cristalline,tantale.

rPresentaddress: Mineral SciencesSection, Canadian Museum of Nature, Ottawa, Ontario KIP 6P4. 2Presentaddress: Bayerisches Geoinstitut, Universit?itBayreuth, Postfach l0l25l, 8580Bayreuth, Germany.

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 6t4 THE CANADIAN MINERALOGIST

INTRoDUCTIoN ture. is devoid of order of its cations; however, cations in the wodginite structure order Work on the chemistry of wodginite has into three distinct sites, causinga quadrupling of principally appeared in descriptive reports of its the unit-cell volume of the parent ixiolite sut'cell: occurrence;there are few studieswith the chemistry a* : ?-oi,b* : 2bi, c* = ci, where the subscripts of the mineral as a primary objective. Given past w and i stand for wodginite and ixiolite, respective- confusion regarding the crystal structure of the ly. This schemeof cation order resultsin the ideal : mineral, its now recognizedstatus as a group of formula ABC2O1, with Z 4 and spacegroup populated B, Sn, and minerals @rcit el al. 1992a),its economic value as A/c; ideallyA is by Mn, by an ore mineral of tantalum, and its relatively C by Ta (Fergusonet al, 1976).Owing to cation common occurrence (Table l), we decided to order, there are two chemically distinct types of examine the crystal chemistry of the wodginite ng-zag chains in the wodginite structure (Fig. l). group. One type consists of alternations of .,4 and B Coctahedra. Unlike studies of natural material, synthesis octahedra;the other consistssolely of given level in work has been directed toward understandingthe Only one type of chain occupiesany crystal chemistryof the group. Turnock (1966)was the closest-packedarray; this results in two types the first to synthesizewodginite; he showed that of layers in the structure, one with chains of C polyhedra, A B ferric iron plays a role similar to that of tin in the the other with chains sf ard polyhedra. of structure. Komkov (1970) grew tin-bearing A more complete discussion the wodginite group wodginitesimilar in chemistryto natural wodginig- structural crystallographyof the given (1992b). group minerals.Sidorenko et ql. (1974)showed that is in Ercit et al. syntheticwodginite could be nonstoichiometricby growing sampleswith variableSn:(Mn+Ta) ratios. EXPERIMENTAL Gatehouseet ql. (1976) showed that Li could be incorporatedinto the wodginite structure in major Most of the electron-microprobeanalyses were amounts.Komkov & Dubik (1983)proved Sn = done at the University of Manitoba using energy Ti isomorphisrnin syntheticwodginite. dispersion(ED); the main data-setwas augmented (WD) Ferguson et al. (1976) showed the wodginite slightly with wavelength-dispersion data structure to be derived from an ixiolite substruc- obtained at the CanadianMuseum of Nature. ED X-ray spectrawere obtained with a MAC 5 electron microprobe using a Kevex Micro-X 7000 spectrometer.Spectra were collected for 200 live TABLE1. LoCALITIES0F U0DCINITE-GRoIPI|INERALS secondswith an operating voltage of 15 kV and a Local l t)t Saeple Saoplo ilo. obtalnsd sample current of 5 nA (measuredon synthetic l. Alto do clz pegDatite, Brull Y A.15 fayalite), and were corrected for current and 2. Ankole, thanda Y A.17 3. Annle Clalm pe$natits, llanltoba Y AC2-79,A-3 voltage drift. Line overlaps, such as TaMd and 4. Bensonp€gioatlte, ZiEbabre tl TaMq on NbZa, SnZB on TiKcu, and MnKB on 5. Coo3aCounty, AlabaDa Y CX-l 6. Cross Lake, llanltoba Y Ct-l FeKa were dealt with by non-iterative techniques 7. EraJarvl pognatlte fleld, Fln]and tf 8. Greenbushsspegnstlte, lf. Australla I C8-l to -4 of spectral stripping involving library spectra of 9. H€rbbNo. 2 p€gEatlt€, Vlrglnla I H2-1 individual elementstaken from standardsused in 10. Kalbln Range,tSSR N ll. Kariblb, ilanlbla nc-l the analysis of wodginite. Manganotantalite 12. , USSR KZ-l 13. Red Cmss Lak€, tlanltobs (MnKcr, TaMa), fayalite and chromite (FeKo), 14. Kola Penlnsula, USSR K0-l rutile and titanite (Tii(o), cassiterite (SnZcu), 15. Krasonlc€, l,loravla, Czechoslovakla cz-l 16. Lavra Jabutl, lllnas Gerals, Brazll LJ-l to -3 CaNb2O6(NbIa), Sc metal (ScKa), ZrO2 (ZrLa) 17. l'larble Bar. ll. Australla t{B-t (CaKcr)were Data 18. ltlt. llatther, I. Australla tst-l and usedas standards. 19. iluhsnba River, Rranda A-26 were reduced with Kevex software using the 20. llanplng, Chlna cH-l 21. Odd lest peguatlt€, lilanltoba 0I-102 MAGIC V program (Colby 1980). WD X-ray 22. Peerlsss psgnatlte, 5. Dakota A-29,-30 analyses were done with a JEOL 733 electron 23. Eastern Sayan, USSR t{ 24. Serldozlnho, Paralba, Bfazll Y BP.I microprobe using Tracor Northern 5500 and 5600 25. Eastsm Slbarla, USSR N 26. Strlckland Quarry, Connectlcut Y sQ-l automation. The operating potential was 15 kV; a 27. Kohero. NaDlbia Y ovK-8 25 nA beam current was used. Data for standards 28. Sulula pegnatite, Tamola, F.lnland Y sK-l 29. Tabba labba, l. Auslralla Y TT.I were collected for 25 s or 0.2590 precision per 30. Tahara, Japan Y TJ.I 31. TancopegEatlt€, llanltoba Y (069-,StitP-, element, whichever was attained first. Data for TSE-sorlss) sampleswere collectedfor 50 s or 0.590 precision 32. Lorar Ianco poglatlte, Fanltoba Y 33. Tuscany, Italy N at the lo level, whicheverwas attained first. The 34. Transbalk8l rsglon, tSSR Y rl-l following standardswere used: NiTa2O6(TaMa), 35. Vlltanlml p€gmtlte, Flnland Y VII-T 36. Iodglna, l. Australla Y m-t MnNb2Ou (MnKo, NbZo), cassiterite(SnZo), 37. Ygllo*nlf€, NUT 1 YKF-soriss rutile (Ti/(a), almandine (FeKa), R.EE-bearing

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 CRYSTAL CHEMISTRY OF THE WODGINITE GROUP 615

'\

\. t.'

_\. ::.;2

Y FIc. l. The ideal structure of wodginite projectedalong X. Polyhedraof one level alongXare shownin solid rule; polyhedraof adjacentlevels are shownin broken rule. .A and B octahedraoccupy one level of the structure;their pattern of order is shown. C cationsoccupy the octahedrashown in broken rule. The octahedra sharecis-edges along Z, which produceso-PbO2-like zig-zagchains.

glass (ScKcu),CoWOa CY,/Ma)and synthetic ZrO, tion data, exceptfor a few, very small samples,for (ZrLoi. Overlap between WMa and TaM$ was which either a conventional 114.6-mm-diameter manually correctedprior to data reduction. Data Gandolfi camera or a Nicolet R3m single-crystal reduction was done with a conventional ZAF diffractometer was used. For powder diffrac- routine in the Tracor Northern TASK series of tometry, Ni-filtered or monochromated CuKcu programs. Lithium determinationswere done by radiation was us€d;annealed CaF2Ia : 5.46379(4) atomic absorption spectroscopyby Mr. V. Kubat A1 was the internal standard. Indexing and of the Ecole Polytechnique. 57Fe Mdssbauer refinement of unit-cell parameterswere done with spectra were collected ai the University of the CELREF progrirm of Appleman & Evans Manitoba. Sampleswere powdered and weremixed (1973).Weights were assigned to eachmeasurernent with benzophenoneto ensure random orientation accordingto the following scheme.Sharp peaksof of the grains. Becauseof their high Ta contents, normal width at half height were assignedweights wodginite-group minerals are strong 7-ray absor- of l; all ragged or abnormally broad peaks were bers; the optimum samplethickness was calculated assignedweights of 0. I to 0.5, dependingupon the according to the procedureof Long et ol. (1983). quality of the peak. Peaksthat were obviously in Spectra were obtained at room temperature, an overlappingrelationship were omitted fronr the collectedwith a 512-registermultichannel analyzer, refinement. and were folded to give 245 data points over an Most heating experimentswere done in air with approximaterange of 4 to + 4 mm/s (relative to a Fisher Isotemp muffle furnace, Model 186. The cu-Fefoil). furnace temperature was periodically calibrated by Philips PWl050 and PWl7lO powder diffrac- monitoring the melting point of NaCl. Temperature tometers were used to collect the powder-diffrac- regulation by the furnace is precise to a 20"C'

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 6t6 THE CANADIAN MINERALOCIST

Some heatingswere done in controlled (CO:CO2) TABLE3. VARIATIONSIN CHE.IISTRYOF THE I'IAJOR EIEI'IENTS atmospheres with a vertical quench furnace; l.lean l,lulmun tilinlnuo ldeal temperaturecontrol with precise l.ln 7.1 9.1 2.r this furnace is to Fe t.7 8.7 0.0 0.0 t l0'C. Mineral syntheseswere done anhydrously TI 1.4 6.2 0.0 0.0 5n 15.1 t7 0 at atmosphericpressures using the Fisher muffle Nb 4.0 il.7 0.5 0.0 furnace.Reagent-grade pure oxidesand carbonates 54.9 43.6 54.5 were used as starting materials. Starting powders All flgures refer to ut. % ol the elerent were weighedto produce0.3 to 0.5 g samples,and were mixed by light hand-grinding in acetonefor Fergusonet al. (1976)gave the ideal formula of 20 minutes. The mixed powders were then pressed wodginite as MnSnTa2Os;as the data in Table 2 into pellets under a lCX)00kPa load to optimize attest,wodginite straysfar from ideality. The main reactivity. After 4 to l0 hours of heating, samples substituentsfor Mn, Sna+ and Ta are Fe, Ti and were removed from the furnace, reground, and Nb; trace amounts of Sn2+ (E.E. Foord, pers. subjected to continued heating. In most cases, comm.), Zr, Sc, Ca, and W have been detected, reactions were complete after 4 to 20 hours; and Li can be present. Maximum variations for however,some samplesrequired repeatedgrinding eachelement are summarizedin Table 3. and heating (up to 40 hours). For both the heating Cation-cation correlation plots @ig. 2) attest to and synthesisexperiments, Ag-Pd foil was usedas some of the dominant mechanismsof substitution a sample holder for runs at 1050"C or lower; Pt for wodginite-groupminerals. Figure 2a is a plot foil was used in all runs at higher temperatures, of total Fe versas Mn. The strong, linear and someruns at 1000"C.The foil was iron-soaked correlation betweentotal Fe and Mn (R : -0.85), wherenecessary. with a slope of -1, indicates that most Fe in wodginite-group minerals is incorporated by Fe- CHstvttsrny for-Mn substitution; this also suggeststhat the ferrous state of Fe is the dominant one. However, Resultsof electron-microprobe analyses severalof the data points fall off the main trend, particularly at very low Fe contents. This finding Results of electron-microprobe analyses of indicates that significant mechanismsof substitu- selected wodginite-group minerals are given in tion not involving Fe also operateat the,4 site. Table 2: other results can be found in Ercit et al. A good negative corelation exists betweenTi (1992a,b),and a tabulation of all 214 compositions and Sn (Fig. 2b); however, only 52t/o of the obtainedfor the presentstudy is availablefrom the variation in Sn can be accountedfor by Ti-for-Sn Depository of UnpublishedData, CISTI, National substitution. Figure 2b shows that substantial ResearchCouncil of Canada,Ottawa, Ontario KIA Ti-for-Sn substitution is possiblein natural mem- 0s2. bers of the wodginite group: some samplesshow up to 75Voof their Sn replacedby Ti. Despitethe linear trend, all data points are displacedbelow the TABLE2. CHTJiICALCOMPOSITION OF SETECTED IIODGINITE.GROUP I4INERALS ideal curve for Sn + Ti isomorphism (bold line in A-3 A-26 CH-l K0-2 KZ-1 ti|C-l T,J-l TSE-16 Fig. LlzO, vt.% 0.00 0.20 0.17 0.00 0.35 0.00 0.00 0.14 2b), showing that constituentsother than Sn 14n0 10.7 4.3 9.0 9.9 7.1 11.0 9,6 8.9 and Ti occur at the B site. The deviation of the Fe0 0.0 6.3 1.3 0.0 0.0 0.0 1.6 2.3 sc2q 0.0 0.0 0.0 0.0 0.0 0.0 1.7 0.0 data points from the trend increasesas the Ti F€z0l 1.1 t.'t 2.0 0.0 0.0 0.0 . 0.1 t.2 Tt0: 0.2 0.9 0.1 0.0 0.0 0.0 4.7 5.5 content decreases;some samples with no detectable lr02 0.0 0.0 0.0 0.0 0.0 0.0 0.2 0.0 Ti deviate from ideality by almost 10090. snq 15.8 14,5 14.0 17.8 7.8 14.5 7.4 8.8 Nb20, 6.6 10.9 5.9 4.3 10,8 4.4 0.8 11.1 Ta = Nb isomorphism is extensive(Fig. 2c); TaPs 64.5 6l.t 67.0 66.3 73.7 68.9 71.0 60,4 l,toj ---.:---0,-0- wodginite-group minerals show Nb:Ta as high as 99.t 99.9 99.6 98.4 99,7 98.8 97.8 98.3 0.35, although the norm is much lower. Ideal Catlons per j2(0) wodginite should have a maximum of 8(Ta + Nb) Ll 0.00 0.32 0.2a 0.00 0.61 0.00 0.00 0.23 per unit cell; however, as Figure 2c shows, this liln 3.28 l.5r 3.27 3.71 2.54 4.08 3.48 3.0i fe& 0.44 2.17 0.45 0.00 0.00 0.00 0.56 0.76 number is always significantly exceeded,with a 0.00 0.00 0.00 0.00 0.00 0.00 0.62 0.00 maximum deviation of 3.7 surplus (Ta + Nb) Fel 0.61 0.51 0.64 0.00 0.00 0.00 0.05 0.35 Tl 0.04 0,29 0.05 0.00 0.00 0.00 1.52 1.6'l cations per unit cell, and a mean deviation in the Zr 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.00 Sn 2.45 2.36 2.39 3.14 1.32 2,55 1.26 1.40 vicinity of I cation per unit cell. The sampleswith Nb 1.09 2.02 1.14 0.85 2.07 0.88 0.15 2.01 Ta + Nb greatly in excessof 8 cationsper unit cell Ta 7.81 6.82 7.78 7.96 8.54 8.25 8.28 6.57 t{ nn7 correspondto data points that deviatesignificantly 16.00 16.00 16.00 15.65 15.07 15.89 16.05 15.95 from ideal Sn + Ti isomorphism (Fig. 2b), - Fe&:Fe} and $20 are calculated accordlngto stoichlometric constralnts (see text), except for sanplo KZ-l (1120msasured). indicating that excessTa or Nb is incorporated at - data for silples CH-I, KZ-l and TJ-l are averagesof two anatyseseacn. the B site, as shown in the structure analysis of - o0o: not detected, '---': not m€asured Ferguson et al. (1976). Figure 3 shows that it is

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 CRYSTAL CHEMISTRY OF THE WODGINITE GROUP 617

E tro 2.1 a tr a tr tr o tr tr n q oo = n tr Otr tr = Ro n d q r.6 o "F*o- go g2 " o. En* o .o tr8tr IL 8""o z 0.8 tr d" snE o o - -n "%'-L*frofl"f,U-tr-6 fu o - u+rn 0.0 23 2 4 Mn (p.f.u.) Sn + Tt (p.f.u.)

oo 12 o@ b o \ b @\ tr a2 tr0 nnu a 10 rh qU Eh\ o. cq- qfi o.F BrDr|0 = n tr ottrrF -tt- o o Gl 8 F "im. t tr-tr-tr tr _.o_ O tr_tt O ntrtr: tro 2 Sn (p.f.u.) 123 Sn + Tt (p.f.u.) tu c + j2 Frc. 3. Relationshipsbetween ltf andtrl+ cations. (a) Nb versasi.la+ cations; (b) excessTa (= 3-rir. 1u, : versus trtf+ cations. The lack of any (negative) IL correlationbetween Nb and trtf+, and the presenceof + a good negativecorrelation between Ta and /tla , show .o that Nb is not involved in B-site chemistry. z1

plots: (a) I 10 12 Frc. 2. Plots of cation-cation correlations total Fe versasMn; (b) Ti verszsSn (the line showsthe ideal ta (p.f.u.) 1:l slope);(c) Ta versasNb.

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 618 THE CANADIAN MINERALOGIST

TABTE5. coilmSITI0NALDATA (wt'%) roR UoD0INITEsAl'lPLEs USED IN only excessTa and not Nb that is incorporated at I{OSSBAUIREXPERIIIENTS the B site. It is interesting to note that to date, A-17 CB-z GB-4 LJ-z TSt-126 wodginite has the only known structure for which m rln core rlm core rln Nb and Ta show differencesin ordering behavior. Mno 5.4 4.'I 6.4 6.? 7.5 7.1 to.2 10.0 Feo 5.8 6.4 4.7 4.6 3.5 3.7 te2uj l.J l.a 1t 1.3 1.4 0.8 0.8 Ti02 2,2 3.3 0.7 t.4 0.0 0.0 7.7 4.6 Mdssb auer spect roscopy 5n02 ll.0 10.4 14.6 13.2 15.3 i5.6 7.9 10.6 Nbr0j 7,5 8.0 o.J 6.9 7.2 9.5 7.9 Taz05 66,9 64.8 64.5 65.0 64.2 61,I 62.6 Although preliminary cation-cation plots sug- 100.1 99.0 98.8 99.4 99.8 99.2 98.6 98.0 gest that much of the Fe in wodginite-group Catlons per 32(0) minerals is ferrous and resides at the A site (Fig. l{n 1.89 1.66 ?.33 2.24 2.69 2.57 3.42 3.49 Fe?. 2.01 2.22 1.69 t6K tin l1l 0.49 0.53 2a), the evidenceis indirect; consequently,direct Fel 0.41 0.45 0.51 0.58 0.43 0.43 0.23 0.25 examination of the valenceof iron in wodginite- Ti 0.69 1.03 0.23 0.44 0.00 0.00 2.29 t.42 Sn r.82 l.7l 2.49 2.24 2.61 2,66 1.25 1.75 group minerals was done by M6ssbauerspectros- Nb 1.41 1.49 1.21 1.08 1.32 1.39 1.69 t.47 Ta 7.57 7.30 7.50 7.68 7.54 7,47 6.57 7.03 copy. Measurementswere made on sevensamples; 15.81 15.85 15.95 r5.90 15.89 15.84 15.95 15.95 five of the$e are naturally occurring membersof - Fez':Fe* estlmatedfrom llossbauerspectrat fomula calculatlons the wodginite group, including the inferred Fd*- wefedone lgnorlng the possiblllty of Ll. rich sampleA-17 from Ankole, Uganda, and two are synthetic Fe3*-rich samplesthat fall close to Mn(Fd+Ta)Ta2O3in composition (synthesizedby group minerals as a whole show variable valences Turnock 1966). of iron, but confirms that they tend to have Fd+ The 5?FeM6ssbauer spectra are shown in Figure dominant over Fe3+. The fact that the Sn-free 4. The synthetic samples(TRl-l and TR3-l) are wodginite samplessynthesized by Turnock (1966) characterizedby a single absorption doublet, with are unstableat low/(OJ indicatesthat Fd+ cannot an isomer shift of about 0.4 mm/s, and a be stabilized at the B site of ordered wodginite. quadrupolesplitting of 0.5 mm/s. This is undoub- This finding implies strong order of Fd* at tbe A tedly dueto Fe3+;the absenceof any other doublets site and Fe3+ at the .B site. lteSn in thesespectra confirms that all the iron is ferric, E.E. Foord (pers.comm.) has carriedout a as Turnock (1966)inferred from the /(O, condi- experiment on wodginite CX-l from Coosa tions of the synthesis. County, Alabama. A very small proportion oi the All naturally occurring members of the tin in the sampleis divalent Q x. 3t/o), and most wodginite group have5TFeM6ssbauer spectra that probably substitutesfor Mn in the structuie, on the differ from those of the synthetic samples. The basisof its chargeand inferred size(effective ionic vtJSn2*l absorption envelopenear the Fe3+doublet is more radiusof = l.l0 A from calculationsby complex, and there is a substantial additional the authors). Sn2* is scarce in the highrlO2) component with an absorption maximum at environmentsin which wodginite-group minerals approximatelyZmm/s. Both featuresare consistent occur, as indicated by the common associationof with a doublet having an isomer shift of about I .l wodginite and cassiterite, and very rare association mm/s and a quadrupolesplitting of about 2mm/s, of wodginite and stannomicrolite @rcit et ql. 1987). characteristicof Fd+. ConSequently,Sn2* contents of all natural mem- Individual 57Fe M6ssbauer parameters and bers of the wodginite group are expectedto be Fd*:Fd+ ratios are given in Table 4; composition- extremelylow, and in this study, all Sn determined al information for the samples used in the by microprobe is assumedto be tetravalent. Mdssbauerexperiments is given in Table 5. For the natural iron-rich samplesstudied here, the propor- Lit hium dete r m inat ions tion of Fd+ varies from 6890 to 8390 of the total Fe. This finding indicatesthat natural wodginite- In their study of the polymorphs of LiTa3Os, Gatehouseet ol. (1976) found that the medium- temperature polymorph, M-LiTa3O6' has the a/. (1990) found TABLE4. 5?Fel4ossMutR PAMI4ETERS wodginite structure. Voloshin et le F& FeY Fe! the first natural occurrenceof this compound, and -Fe I aIo aEo t-17 ).66(7) .08(13)2.16(7)1.328) 0.3714) 0.15(ll, 0.i7(7) iB-2 ).57(6) .0e(4)2.05(7),.J5 5) 0.4214) 0.17(10,0.26(3) '.54(6) ,r0(4) l.e2(7),.27t2) 0.4714) 0.18(9) 0.23(4) lJ-2 ).64(11) .06(14)2.05(25)).34r.46)0.381 ll) o.l5(7810.25(7) rsE-l ':l91tu) t:lllt t) ).4415) 0.15rr3) 0.s4(24, 0.32(9) tRl-l :l:16) 1.52t6) 0.3714) 0.53(6) 1.0 rrc-l ).49r5) 0.3915) 0.54(6) 1.0 Frc. 4. 57Fe-Mdssbauerspectra of wodginite-group - all unlts afe m/s, * relative to Fe-foll minerals.

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 CRYSTAL CHEMISTRY OF THE WODGINITE GROUP

4.276

4.263

?.186

7.156

7.t32

?.105 ro O 8.491 x

F z. 8.464 = o L) 10.84

10.81

7 .55?

7 .536

I I .54

11.50

-3 -2 -l 0 | 2 VELOCITY(I1N S-']

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 620 THE CANADIAN MINERALOGIST

TABLE6. tITHIIJI,ICONCENTMTION IN I4ODGINITT.GROUP IiIINERALS group minerals were analyzedfor Li by atomic SamDfe: A-17 G69-23 KZ-l LJ-3 TP-l TSE-76 VII-1 absorption spectroscopy.The results are given in Ll (ppn): 138 275 1643 47 Table 6. Two of the samplesare rich in Li; five are poor in Li. The two samplesrich in Li are also enrichedin Ta, which supports the mechanismof coupled substitution describedabove.

named it lithiowodginite. In M-LiTa3O6 and CRYSTALLOCRAPHY lithiowodginite, Li occupies the ,4 site, and Ta occupies the B and C sites; hence the coupled Wodginite-group minerals typically occur as substitution Mn + Sn :: Li * Ta is responsible irregular aggregatesof crystals or as parallel to for the incorporation of Li in the structure. subparallelgroups of crystals.Yirtually all crystals Consequently,significant amounts of lithium are are twinned; penetration twins with {001}or {100} to be found only in Ta-rich members of the as the composition planes are the most common. wodginite group. Predominanceof the form {lll} in combination To assessthe range, limits and behavior of with this style of twinning producesa characteristic lithium in the wodginitestructure, seven wodginite- flattened bipyramidal morphology; where present,

z

-Y

a

b G Frc. 5. The morphology of penetration-twinnedwodginite-group minerals: (a) lithiowodginite, Tanco p€gmatite, Manitoba; (b) ferrowodginite, Ankole, Uganda; (c) wodginite, Muhembe, Rwanda. Exaggeratedirregular lines representcomposition planes. Forms are a [@], d UOl], r [310],p Iltf, 4 {l3ll.

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 CRYSTAL CHEMISTRY OF THE WODGINITE GROUP 621

TABTE7. UNIT.CELLPARAI'IETERS FOR SPECII{ENS OF HOOCINITE.GROUP 31.O 14INEMtS smple a(A) b(A) c(A) e(t v(Ar) A-17 5.121(1)90.51(2) 553,5(1) A-25 5.r26( 1) e0.3r(2) 554,6(r) A-29 5.r17(l) 91.17(i) 559.4(l) tr A-30 5.r17(r)9r.21(2) 559.4(r) AC2-79-t 5.r42 (r ) e0.78(3) 56r.e(2) trtr Btit-lA 5.112(2)90.ee(3) 558.6(2) G69-2 5.10e(r) 91.07(3) 55s,6(1) G69-13 5.u6(2) 90.94(3) 556.4(2) G69-14 5.i14(r ) 90.95(r) 555.9(1) G69-17 5.112(2)91.10(3)557.0(l) c G69-23 5.106(l) eI.04(3) 554.5(1) @.3o.5 G69-39 5.118(1) 90.93(3) 556.7(1) G69-42 5.113(l) 90.e7(1)555.7(l) G69-464 5.118(1) e0.e0(1) 556.2(1) 5.r08(l) 90.78(2) 553.2(1) GB.2 5.110(1) 90.80(2) 554.3(1) cB-3 5.u3(l) e0.97(l) 557.0(l) GB-4 s.120(3)90.66(4) 554.4(2) H2-l 5.r09(r) 90.75(l) 553.0(r) KZ-1 5.0e4( r ) 91.08(l) 555.3(1) r,J-l 5.138(1) 90.60(l) 559.1(1) LJ-2 5.126(r) 90.82(2) 558.1(r) t4c-1 5.1le(1) 91.14(2)5s9.5(r) ltM-l 5.0e6(4) e0.83(4) 552.4(4) otl-102 5.114(1)el.20(1) 558.5(r) 30.0 st{P-12 5.105(1) e1.le(l) 557.4(1) 0.1 sQ-t s.115(1)9r.18(2) 558.3(l) TL-45 s.r13(1)90.90(3) 554.7(2) r(o21)/ t(22o) TSE.5 s.102(l) 91.12(3) 556.6(r) TSE-7 s.112(1)9i.2r(1) s57.9(r) TSE-16 5.106 (r) 91.01(2)554.4(l) Frc, 6. Correlation of 0 and the relative intensity of TSE-28 5.110(1)9r.11(2) 558.3(l) supercellreflections; (021) is a supercellreflection, and lSE-30 5.116(1)90.77(2) 55s.0(l) TSE-31 5.103(1) 90.91(2) 553.4(1) (220) is a subcell reflection. The ratio I(021)/I(?20) is s.r11 (r ) 90.99(3) 555.0(r) measure of the relative intensity of supercell TsE-40 5.l09 (r) 91.06(2) 556.2(l) a TSE-45 5.r08(l ) 90.86(?) 553.4(l) reflections.All intensitiesare peakheights; all intensity lSE-48 5.r09(l)9r.03(2) 555.2(l) data were collected with a Philips PWl7lO powder tJL-O5 5.114(2)90.94(4) 554.7(4) TSE-76 5.082(1 ) 91.04(2) 553.3(r) diffractometer using an automatic divergenceslit and TSE-77 5.116(r)90.82(3) s5s.0(r) TSt-90 5.r25(l)e0.40(6) 554.e(3) monochromatedCuKo radiation. All sampleslisted in TSE-97 5.r08(r) e0.87(2) s54.5(l) Table 12were used in this plot, exceptfor H2-l, which TSE-100 5.u6(3) 90.8e(9) 555.1(5) TSE-107 e0.71(2) s50.5(l) was not examinedwith the above model of diffrac- TSE-109 90.62(5) 552.r(3) tometer. The displacementof the data point shown as TSE-122 ll 90.i8(2) 548.s(5) TSE-r26 E 90.13(3) 552.40) a filled symbol from the generaltrend is due to the lSt-130 446 11.413 551.s(I inaccuracyof B measurementfor samplesin which 0 TsE-132 485 11.446(1 554.5(t TSE-134 462 11.429(1 is approximately90o. TSE-136 475 ll.44l(3 TSE-137 489 556.I TSE-138 507 557.2 TT.1 Eno 5J/.J vlI a5l uD-1 All unlt cell parmeters uere obtalnedwlth Phillps pof,derdif- of wodginite compositions, the X-ray-diffraction fractoneters (Ni-filtered or graphite-monochronatlzed Cu(o radi a- tlon, CaFzjniernal standard)or a Nlcolet R3mslngle-crystal propefiies of some samples (e.9., A-17) are dlffractonoter (graphlte-nonochromatlzedl,lo(o radlatlon). intermediatebetween those of ixiolite and of type wodginite. Specifically, their supercellreflections have lower relative intensities than in type wodginite, and their B anglesapproach 90' as the this readily distinguisheswodginite-group minerals relative intensities of the supercell reflections from -groupminerals and (Fig. approach 0 (Fig. 6). These features imply that 5). Wodginite-groupminerals can also be polysyn- wodginite-groupminerals as a whole show variable thetically twinned across{l@l (e.9. sampleSK-l); degreesof order. however, this twin relationship is much less It is important at this stageto clarify the nature common than the penetrationtwinning. of the order-disorder phenomena.In our descrip- All refined unit-cell parametersfor wodginite- tions, we use the following terminology @rcit er group minerals of this study are given in Table 7. al. 1992b). Wodginite-group minerals that show The cell parametersshow considerablevariability: crystallographic properties fully compatible with 4 range! from 9.417to 9.533 A, b from 5.082to the ideal structure of wodginite are termed (fully) 5.142 A, c from 5.082to 5.142 A, and 0 from ordered, However, cation disorder involving nearly 90o to 9l.2lo. We would expect from the mixing among the A, B and C sites modifies the broad chemical variations shown by wodginite- symmetryof wodginite-groupminerals, resulting in group minerals that a large part of the variation in diffraction properties that are intermediate to X-ray-diffraction propertiesis merely symptomatic ixiolite and wodginite; we describethese as partiolly of the chemicalvariations. However, for a variety ordered.

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 622 THE CANADIAN MINERALOGIST

Order can be induced in disorderedto partially orderedmembers of thg columbitegroup by heating at high temperaturesQernf & Turnock l97l). On the basisof the structural and chemicalsimilarities of columbite-groupand wodginite-groupminerals, it would seem likely that heating of partially ordered members of the wodginite group at conditions similar to those used for columbite- group minerals should similarly induce order. Heating experimentswere done under the same conditions .$tandardized for columbite-group mineralsbyternf & Turnock (1971),j.e., l000oC for 16 hours. Most heatingswere done in air rather than in/(O)-controlled atmospheresbecause at the time of the experiment,all iron was assumedto be ferric on the basisof results of the synthesiswork of Turnock (1966). This early oversight did not seemto introduce significant errors: l. All heatedsamples were crystal fragments, not o2o powders, minimizing the surface area available for (cuKo) oxidation. 2. For compositions shown by wodginite-group minerals, oxidation-reduction reactionsat l000oC are slow (Turnock 1966).Cern! & Turnock (1971) showedthat unit-cell dimensionsof Fe-rich colum- bite-group minerals are statistically identical for bil granular samples heated in air versus similar samplesheated in CO2:COcontrolled atmospheres. For the structural and chemical reasonsoutlined above, similar behavior is expectedfor wodginite- group minerals. 3. Most samplesare Mn-rich. For these samples, oxidation of a minor proportion of iron during heatingshould not significantly modify patternsof lt order. 4. Heating experimentsin atmospheresof differing /(O) were done on the only abundant sampleof a ferrous-iron-rich member of the wodginite group (A-25). A single crystal was ground to a powder and was then divided into two batches for the

TABI.E8. UI{IT CELI PAMI.IE'TERSFOR HEATTD SAi{PLES OF l sample a(A) b(A) c(A) I (') l/ (A3) A-r7H 9.4s3(l) 1r.4?9(r) s.098(l 90. 550.8(1) A-25H1 9.456(2) ll.4?9(2) 5.094(l on 550.4(l) A-?5H2(C0) 9.457(3) 1r.431(3) 5.093(r 91.01 550.5(2) A-25H2(C0r)9.458(2) 11.426(2) s.093(1 550.3( r ) /\rsE-126 A-29H 9. s31(3) r1.465(2)5.118(l ot 559.2(1) A-30H 9.526(2) 11.474(2)5.r16(1 559.1( 1) ACz-79-lH 9.527(2) u.471(2) 5.119(l 559.4(1) G69-2H(C0) e.4e2(3) 11.452(2) 5.108(2 91.07 555.r (1) G69-2H(C02)9.491(2) 1r.449(2) 5.106(r 9l 5s4.8(l) G59-23H 9.485(l) u.449(1) 5.108(l 91.05(I 554.6(l) GB-2H 9.46112') 11.434(l) 5.105(1 90. 5s2.5(l) Hz-rH 9.457(2) 11.4?7(3) 5.098(2 90,92 550.8{2) KZ-IH 9.484(2) 1r.493(?) 5.095(I 9t,02 555.3(1 ) LJ-IH 9.507(l) 1r.463(l) 5.u6(r 90, 557.4(1) )[ 0H-102H 9.s3r(l) 11.475(1)5.125(l 91.l2 560.4(1 ) TSE-7H 9.517(1) r1.469(t) 5.114(l 91. 558.r ( r) rsE-4oH 9.498(2) r1.457(2)5.110(1 ol I I 555.9(l) I-L-J TSE-45H 9.461(2) 1!.432(l) 5.r01(1 55r.6(1) 31 30 29 31 30 29 TSE-76H 9.458(1) u.506(l) 5.083(l 91.04(I 553.0(l) TsE-126H 9.441(2) 1r.414(2) 5.088(l 90 548.2( l ) l,{D-tH 9.529(2) 11.477(l) 5.t21(2 560.0(2) o2e All heatlngs are in alr unless othemlse ind.icated. (cuKc)

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 CRYSTAL CHEMISTRY OF THE WODGINITE GROUP 623

same behavior for the cell parametersof samples Frc. 7. Changesin X-ray powder-diffraction pattems of of orderedFe-poor wodginite. partially orderedmembers of the wodginitegroup with Upon heating, many of the samples showed heating, as typified by sample TSE-126. (a) In the significantlyincreased intensities of supercellreflec- range I 8" to 26o?i (CuKo), intensitiesof the supercell tions relative to subcell reflections (Fig. 7a), (200),(lll), (lll), (021) (130)increase reflections and increased angles(Fig. 7b), and altered unit-cell to the subcell reflection (220); B upon heating relative parameters (Table 8). However, many samples (b) the separationof (221) and increasesupon Q2l) significant changesupon heating. heating, indicative of an increasein 0. N: natural, .F/: failed to show heated. This latter group must be fully ordered;the samples that showed change upon heating must be only partially ordered. The unit-cell parametersI and V showed the most changeupon heating;they are the parameters that are most sensitiveto order-disorder effects. experiment.Powders were used instead of crystal Figure 8 is a plot of these two parameters for fragmentsto ensurehomogeneity and to facilitate samplesused in the heating experiments.The data oxidation-reduction reactions.One batch of pow- for any given sample are representedas a vector der washeated in CO L(O, : 10-17atml, the other (hereafter designated"heating vector"); the tail in CO2 V(O) = lO2'5 atml. Heatings resulted in end of eachvector representsthe unheatedsample, different color$ of powder (CO2: brown' CO: and the tip of each vector representsthe heated brown-black); however,no significant differences sample. Most of the vectors are subparallel, and in relativepeak intensitiesor in unit-cell parameters indicate that 6 increases and V decreases with could be detected(Table 8, sampleA-25H2), nor increasingorder among the cations. In two cases did extra phasesform. Grice (1970)observed the (WD-l and OW-102), Z was actually observedto

(rt

r, 555

90.o 9o.2 90.4 90.6 90.8 El.O 91.2 I (o) FIc. 8. Plot of unit-cell volume (lz) versusp for natural and heated pairs of wodginite-groupminerals. The tail end of eachvector marks the natural sample; the tip marks the heatedsample.

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 624 THE CANADIAN MINERALOGIST

increase slightly upon heating, most probably allow for the calculationof fully orderedformulae owing to reactionwith minor amountsof associated for wodginite-groupminerals. By assigningall iron cassiteritedetected in the patterns of the unheated initially to the ferrous state, and iteratively but not the heatedsamples. The heatingvectors for adjusting the valence ratio to match A and B the Li-rich samplesKZ-l and TSE-76 slightly but site-occupanciesof iron, formulae were calculated significantly run oppositeto the generaltrend. We for all sevensamples for which Li determinations interpret this as being due to a loss of Li during weremade. The formulae for severalof the samples heating and not to changesin the degreeof order are nonideal; their B- and C-site sums conform to (the relative intensitiesof supercellreflections for wodginite stoichiometry, but their.A-site sums are these samples do not change significantly on lower than expected.This is best shown by sample heating). KZ-l; its formula is (Mne.6al-io.t5)sj.7e(Tao.5rSno.rr) s6.e6(Ta1.a6Nbo.srrzOs,which translatesto a 2l9o Fonvula CelcularroN eNo Strr CHEMISTRy deficiency in the /-site sum. Such deficiencies cannot be ascribed to a systematic error in Li The calculationof formulae for wodginite-group determination; the Li-rich sample TSE-76 shows minerals has the potential of being problematic. nearly ideal stoichiometry: (Lin.53Mr5.ar)>o.x * The majority of compositions reported in the (Ta0.75Snn.26Fe3o.o,)ro.ge(Ta1.76Nb6.2a) g2Os. literature were obtained by electron-microprobe The only possible explanation for the .A-site analysis.Thus determinationsof Fe2+:Fe3+ratios deficienciesin some samplesis that they repre$ent and Li contents are almost nonexistent. Most vacancies incorporated at the site for charge- wodginite-groupminerals are very fine grainedand balancereasons. This is further supported by the inhomogeneous;consequently, it is doubtful that behavior of samples KZ-l and TSE-76 upon F*+:Fd+ and Li determinationswill be made in heating (see Crystallography), wherein heating future studies unless microchemical methods are seemsto have resulted in a loss of Li from the readily available. Another obstacleto calculation structurewithout producing any significant change of a formula is the variable degreeof cation order in diffraction properties indicative of changesin shown by wodginite-group minerals as a whole, the degreeof crystallinity or in the degreeof cation making site assignmentsfor individual samples order. impossiblewithout some quantitative measureof It is presumablycharge imbalance at the .B site the degreeof order. that regulates the location and abundance of The initial question to be answeredis whether vacanciesat the A site. The following mechanisms there are sufficient constraints on the crystal are responsiblefor balancing much of the excess chemistryof wodginite-groupminerals to allow for positive chargeintroduced by B-site Ta: the calculation of the two undeterminedchemical 2tlstto*l .. B[Fe3*]+ B[Ta5+] variables,namely Li and Fe2+:Fe3+. A[Mn2*] + B[Sn4+]= A[Li*] + B[Ta5+] Chemically,the simplestsite is C, which in fully Presumably, vacanqes are rncorporated in the orderedmembers of the wodginitegroup hostsonly wodginite structure when these mechanismsfail, Ta and Nb. As Figure 3 and the refinement by r.e., when p(Li) and f(O2) are too low for the Fergusonet ol. (1976)show, Nb is excludedfrom mechanismsto operateefficiently. From what has the ,B site; it is concentratedat the C site, and Ta so far beenestablished about the crystal chemistry makes up the remainder of the site. On the basis of the wodginite group, cation-site vacanciescan of both size and charge, tungsten is expectedto only be introduced if Ta5* substitutesfor Sna+at prefer the C site to the B site. the B site; hence the incorporation of vacancies For fully ordered members of the wodginite G) at the .,4site takes place by the mechanism: group, the B site hosts Sn, Ti, and all Ta in excess AI + 2B[Tas+]e a[Mn2+] + 2B[Sn4+] of the C-site requirements.The synthesiswork of At first, considerationthis extra complication Turnock (1966)shows that Fe3+is localizedat the would seem to make the calculation of ordered B site. On the basisof size and chargesimilarities, formulae for wodginite-groupminerals a difficult it is expectedthat Sc and Zr also prefer this site. task, if not an impossibleone when only electron- All Mn and Li are assignedto the A site in fully microprobe data are available. However, Li, ordered members of the wodginite group. By Fd*:Fd+ and ordered-formula calculation are deduction from the work of Turnock (1966), and possibleif the following constraintsare applied: asFigure 2a attests,Fe2* showsa strongpreference l. The sum of the B plus C sitesis 12 cations per for the .4 site. On the basis of argumentsalready unit cell. presentedhere, Sn2+ is expectedto prefer the .4 2. All iron at the A site is ferrous; all iron at the site. B site is ferric. Initially all iron is assumedto be Ignoring variabledegrees of cation order for the Fd+, and iron is introducedto the.B site to meet present, the above constraints are sufficient to the conditions of constraint number l.

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 CRYSTAL CHEMISTRY OF THE WODGINITE GROUP 625

3. Excesspositive charge introduced at the B site by Ta is balancedby the introduction of Fe3* at the B site and of Liatthe A site. As the proportion of Fe3+is already dictated by constraintsI and 2, Li can be calculatedby charge-balanceconstraints (number of positive chargesper unit cell = 64). o 6 As a necessary consequenceof the above, c, vacanciesin cationic siteswill be generatedonly if .13 os calculatedFe3* and Li fail to balance all of the -l'l,Pl! excess charge introduced by B-site Ta. The calculated number of Li cations represents a maximum, and the number of vacanciesrepresents a minimum; the inherent assumptionis that during growth, the wodginitestructure attempts to balance excesspositive chargewith available cations first, and incorporates vacanciesonly if such cations (i.e., Li, Fe3+)are not available.This would seem to be responsiblefor the composition of sample oo". KZ-I, which has the highest number of /-site ffi vacanciesto date for any memberof the wodginite Frc. 9. Calculatedversus observed Fe3+/(Fe3+ +Fe2*) group, and is known to have originated in an for selectedwodginite-group minerals. The calculated .4-cation-poorenvironment @rcit 1986). value comes from ordered formulae derived from Using the above approach, ordered formulae electron-microprobedata; the observedvalue comes The referenceline marks werecalculated for wodginite-groupminerals (e.g., from M

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 626 THE CANADIAN MINERALOGIST

differencesin atomic weight and charge between Li+! Sn, Nb, Ta versus Lii i.e., slight errors in the measurementof the concentrationsof heavy,highly chargedelements will translate into large errors in the calculation of Li+. The vertical bars in Figure l0 representvariations in the Li calculation due solelyto intracrystallinezonation; it is possiblethat a large component of the scatter is due also to inadequate estimation of the bulk composition from the one or two microprobe analysesused in the calculation. To be more conclusiveregarding Li, more analyses of Li-rich members of the wodginitegroup are needed;unfortunately, known Li-rich samples are too few in number and are insufficient in bulk to allow Li determinationsby conventionalmethods. It shouldbe noted that the calculationprocedure concentrate$vacancies at the A site in preference Mn to the .B site. The crystal-chemicalreason for this approach is as follows. If vacanciesare incor- porated to balanceexcess positive chargeat the B To site, then the.4 site is the preferred locus if charge balanceis to be achievedlocally: in the wodginite structure(Fig. l), there are two ,4-cation polyhedra adjacentto eachB-cation polyhedron; however,no two B-cation polyhedra are mutually adjacent. Figure I I illustratesthe variation in,4- and B-site chemistry on the basis of fully ordered formulae. The chemistry of the ,4 site is dominated by Mn, Fd+, Li and vacancies.In Figure llA, data for Mn, Fe2+and [Li+t]l are shown; compositionsof wodginite-groupminerals cluster near the Mn apex, E-{ .in Eh - ;-,\t and scatter along the Mn-Fe or Mn-[Li+E] lo TliSdj$b^ ;_..:tal!|'tr_-^tTttr - ---bbF,l^EF Lo*. fr o sidelines. Wodginite-group minerals with low ffiF t9 -n ufiu n- Mn:Fd+ ratios tend to be impoverishedin Li and uu- .,4-sitevacancies, and thus in B-site Ta. There is no crystal-chemicalreason for such behavior. How- Sn Ti ever, when this behavior is consideredalong with Fro. I l. Main-elementchemistry of the/4 andI sites.For (l) sympathy between Mn and Li in pegmatitic plot (bottom), symbolsrepresent - the B-site filled elbaite "tsilaisite'o (Schmetzer& Bank 1984),and sampleswith Fe3+less than 0.3 cationsper unit cell. (2) the enrichmentin Mn shown by oxide minerals in the lepidolite classof (Cernf & Ercit '1985), quenceof the substitution mechanism2Sn - Fe3+ it seemsthat a common geochemicalcontrol + Ta. Figure l lB showsthat Sn dominatesin most exists for the associationof Mn and Li in observed; however, natural minerals in general. In addition, wodginite-group of the compositions achieve end-member MnSnTa2Og mineralswith high contentsof Ta at the I site tend samples never As the Ti content of wodginite-group to form in environmentscharacterized by high p1u composition. proportion Ta relative to the chemicalpotentials of other cations; minerals decreases,the of B-site geochemical these environments develop very late in pegmatite increases.This is expectedto be a granitic fractionation, when Mn:Fd+ is typically very high control; as fractionation progressesin (Ercir 1986). pegmatites,&ru: (pr.ru * prJ typically increases,hence Figure llB is a plot of the main B-site granitic pegmatites hosting wodginite-group constituents Sn, Ti and Ta. Hollow symbols minerals typically become enriched in Sn with representsamples with high Fe3+(>0.3 cationsper respect to Ti prior to significant increasein the unit cell), filled symbolsrepresent samples with low chemicalpotential of Ta. Fd*. The fields for the two subgroups overlap The chemistry of the C site is not shown, as it extensively, except in the region where Ta involves only M-for-Ta substitutions. Concentra- dominatesthe B-site chemistry, a necessaryconse- tions of Nb in the C site are typically low: for the

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 '627 CRYSTAL CHEMISTRY OF THE WODGINITE CROUP

presentmodel of fully orderedformulae, the mean betweenthese sidelines. There is no crystal-chemical Nb occupancyis 1590,with a maximum of 3590. reasonfor this trend; the answerlies in pegmatite geochemistry. Extreme Ta-enrichment in granitic CoMposITroNAL- STRUCTURALCoRRELATToNS pegmatites tends to occur where the bulk of Mn-for-ne substitution is nearly complete (tern! Because of the relatively large number of & Ercit 1985,Ercit 1986).The low degreeof scatter substitution mechanismsinvolved in wodginite- in Figure 12 also showsthat severalsubstitutions, group minerals, coupled with the variable degrees e.g., F*+ + Mn and Ti .. Sn, havesimilar effects of cation order, dala on mineral chemistryand the upon the a and b cell parameters. degree of order cannot be easily derived from A multiple regressionanalysis using the SAS unit-cell parameters. computer program GLM (GeneralLinear Models) The use of heating veqtors,such as in Figure 8, was done on 25 fully orderedsynthetic, natural and allows for semiquantitative assessmentsof the heatedwodginite-group minerals to assesscomposi- degreeof order; the larger the magnitude of the tional controls on unit-cell parameters. Run vector, the lessordered the mineral is in its natural conditionsand unit-cell parametersfor the samples state. In Figure 8, individual heating-vectorscover of synthetic wodginite are given in Table 9. The a large range of the spaceoccupied by the data; independentvariables of the analysiswere chosen obviously, plots like this emphasize structural on the basisof inferred mechanismsof substitution. variability within the group. Conversely, the The six variablesused were Li, Fd+, Fd+, Ti, Nb magnitudesof heatingvectors in a-b spaceare less and the sum of ,4- and.B-sitevacancies. The results than in V-B space;a and b are relatively insensitive of the analysis are given in Table l0; statistical to cation disorder, so that a plot ofa versusb (Fig. information is given in Table ll. The coefficients 12) might contain useful compositional informa- and interceptsof the equations for each unit-cell tion. This is the case;Fe- and Ti-rich compositions parameterare given acrosseach row of the table. plot in the lower left-hand corner, Mn- and Sn-rich The order of each equation was determined by compositionsplot in the upper parts, and composi- stepwiseaddition with a maximum R2 improve- tions rich in B-siteTa (plus Li and. -sitevacancies) ment, using the SAS progxam STEPWISE. With plot closestto the lower right-hand corner of Figure the exception of the model for 0, Nb did not 12. Concerningmineral chemistry, the low degree contribute significantly to the functions; as the of scatter in Figure 12 shows that Mn + Fe and presenceof Nb in wodginite-groupminerals is due Sn + Ta isomorphismare decoupledin wodginite- solely to Nb-for-Ta substitution, and the effective group minerals: samplesplot along the (Fe,Ti)- ionic radii of Ta and Nb are identical (Shannon (Mn,Sn) sideline and the (Mn,Sn)-(B-site Ta) 1976), Nb-for-Ta substitution is not expectedto sideline, but there is no spread of data points affect the cell edgessignificantly. Five composiiion-

.FJI. o l.:'; t rr fit'

9.4{) I 1.99 1'1.41 t1.4tt 11.45 11.47 11.49 11.61 b (A) Frc. 12. Compositionaltrends in a-b space.

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 628 THE CANADIAN MINERALOGIST

TABLE9. SYNTHITICIIOOGINITE-GROUP PHASES USED IN CELLPARA}IETTR microprobe analysisvel,sa,t powder diffractometry. CALCULATIONS the bulk cornpositionsof the powdersmay Composltlon Startlng l4lxtufe T Mediun Run As such, ('c) Tine (h) deviate significantly from the microprobe-derived MnsnTazos MnC03r Snoa+ Ta?or 1050 alr, I atm 10+10 composition. MnTiTazos l'lncoj + Tl02 r Ta20j 1050 alr, I atm 10+10 Mn2Fe3'Ta50162t'1nTa206 + Felaoa 1225 alr, I atn* 3. The multiple regressionmodel is a linear one, CEtL PARAI.IETERS whereassolid solution in minerals may result in ItnSnIaz0a MnTlfaz'B l4n2Fe?Tar016 nonlinear correspondencebetween compositional a 9.s37(r)A e.428(2)A 9.450(1)A data and unit-cell parameters (Newton & Wood b 11.482(l) 11.410(1) 11.426(1) c 5.124(1) 5.084(1) s.094(r) 1980). p 9r.24(r) ' er.20(r)' 90.89(l)' v 561.0(l)A! 546.7(1)A! 549.7(1)A! Despitethe above considerations,the equations ' from Turnock(1966). of Table l0 predict the unit-cell parameters of wodginite-groupminerals reasonably precisely; the root mean square (RMS) errors in a, b and c are TABLEIO. RESULTSOF THEI'IULTIPLE REGRESSION ANALYSIS 0.009,0.007 and 0.0M A respectively,in 0 the RMS Cell Inter- is 0.04' (comparable to the precision of B Para- cept Coeffici ents &2 merer (7,) measuremeltfrom precessionphotographs), and in Li TI ilb V it is 1.2 A3. Figure13 shows the correspondence a 9.539 -0.014 -0.034 -0.032 -0.031 -0.044 0 betweenobserved and calculatedvalues of a, b, c b 11.475 -0.008 0.012 0 -0.019 -0.028 0 of the data and c 5.123 0 -0.017 -0.024 -0.011 -0.014 0 and 0. The low degreeof scatter p 91.28 -0,10 -0.04 0 -0.03 -0.21 -0.11 lack of any systematicdeviation from ideality for -1.5 -3.2 -4 [ 560.5 0 any of the three types of sample (natural, heated Unlts: a, b, ct At Ot'i Vt 43, and synthetic) further attest to the validity of the calculationprocedure. of Table 10, unit-cell parameters TABLEII. STAT1STICALDA]A FORTABTE IO With the data fully ordered formulae Inter- Fe2r Li Ti FeY Nb can be calculatedfrom the cepl of wodginite-groupminerals; the result of such a a 5(0.0) 4(0.3) 2(0.0)l0(0.s) 2(0.0) 6(0.0) necessarilygives cell parametersdevoid b 2(0.0) 3(r.e) 2(0.0) - 2(0.0) 4(0.0) calculation c 2(0.0) - l(0.0) 5(0.0) 1(0.0) 3(0.0) of the effects of cation disorder, i.e., comparable p 2(0.0) 2(0.0) l(0.1) l(0.3) 2(0.0) eto.ol v 5(0.0) 5(r.2) 3(0.0) l(0.4) 3(0.0) 7(0.0) to the cell parametersof heated wodginite-group Leadlngnumbers are standarddeviatlons on the last s.lgnificant minerals. The advantageof such a calculation is flgure of the corespondlngparareters of Table 10. Numbersln parenthesesare % probabllltles for Holtl.0. that information on degree of cation order in wodginite-groupminerals can be obtained without having to resort to heating experiments,which is especially useful where samples are in- al variableswere necessaryto model a, and V; B homogeneous. only four were neededto model b and c. Because The angle is best to quantify the degreeof ofthe way the variableswere selected, the intercepts B order, as the sensitivityof0 to disorderhas already of the five equationsnecessarily represent the cell been established, and unlike the other cell parametersof MnSnTa2Os.Input to the equations parameters, the value of for fully disordered per B consistsof the number of atoms unit cell of wodginite (= ixiolite, sensustricto) is known (90'). each independentvariable, based on the calcula- An index of order (A) for wodginite-group tions of the fully orderedformulae. As an example, minerals, defined as [0(obs) - 90"]/[0(ord) - 90o], the calculateda cell edgeof hypotheticalFeTiTa2O6 should serye as a measureof the degreeof order - (0.014* - (0.031, : would be 9.539 4) 4) 9.359 of the sample.B(obs) refers to the unheatedsample, A. and B(ord), to either 0 calculatedfrom Table l0 or Problems with the set of calibration data and obtained from heating experiments.Samples for regressionmodel are not trivial: which A is equal to 0 are fully disordered,samples l. Powder-diffraction data represent structural with A equal to I are fully ordered, and samples data for a bulk sample; estimation of the bulk with intermediatevalues of A are partially ordered. composition of the sample from electron- On the basisof the precisionof measurementand microprobe data is difficult owing to strong calculation of B, the relative standard deviation compositional zoning of some samples. In cases associatedwith the calculationof A is estimatedto where core and rim compositionsof a crystal are be 5Vo. available, the assignedbulk composition is taken The assumptioninherent in this approachis that to be a weighted mean of 2/3 times the core the transformation from ixiolite to wodginiie is composition plus l/3 of the rim composition. continuous,i.e.,the relationshipbetween A and the 2. In nearly all cases,separate parts of crystalsor degree of order is taken to be continuous and separate batches of crystals were used for linear. The possibility of the transformation being

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 CRYSTAL CHEMISTRY OF THE WODGINITE GROUP 629

9.58

t t.t2

3.52 't tr g o g 1.48 o |! (, 8.48 (, o o I1.44 a 9.44

.A 11.40 ^ 9.40 ru 9.40 9.44 S.48 9.62 11.40 11.44 | 1.48 I 1.52 a (obs) b (obs)

6.12 91.2

5.10 g o o o t! (.) (, e 1.0 5.08 (, @

5.08

5.O4u- 5.04 6.06 5.08 6.tO 6.12 e0.8 e 1.o s1.2 c (obs) F (obs) Fto. 13. Plots of calculatedverszs observed unit-cell pruameters(a, b, c, B) for samplesused in the regressionanalysis. Squaresdenote natural samples,filled circlesdenote heated samples, and filled trianglesdenote synthetic wodginite.

discontinuouscannot be ruled out; however,plots like Figure 6 suggestthat at the current level of TABLE12. DTGREEOF ORDERFOR SELECTED SAI.IPLTS OF I{ODGINI]E.GROIJP precision,the assumptionof a continuoustransfor- I'IINERALS smple p(calc)' A % orderr mation is probably acceptable. TSt-130 90.00 90.90 0.00 0 Estimates of the degree of order for selected | 5E- IZO 90.t3 90.92 0.14 15 TSE-t22 90.18 90.87 0.2i 20 wodginite-group minerals are given in Table 12. A-2s 90,31 90.88 0.35 35 TSE-90 90.40 90.93 0.43 45 The validity of the definition of the ordering index TSE-109 90.62 90.99 u.o5 03 A-17 90.sr 90.78 0.65 65 is shownby the excellentagreement between A(calc) tJ- 1 90.60 90.91 0.65 65 TSE-137 90.71 9t.07 0.66 65 from Table 12 and A(meas) from the structural TSE-134 90.'t2 91.00 0.12 70 study et al. 1992b) for the partially ordered 6B-4 90.66 90.85 0.77 75 @rcit TSE-30 90.77 90.97 o,79 80 samplesCX-l and A-17. Sample CX-I, Ta-rich TSE-77 90.82 91.01 0.8t 80 H2-1 90.75 90.93 0.81 80 wodginite, has a A(calc) of 0.60 and a A(meas)of C69-46a 90.90 91.02 0.88 90 LJ.2 90:82 90.88 0.93 95 0.59(l); sampleA-17, Fd+-rich wodginite,has a 069-17 91.10 9t.14 0.96 95 A(calc) of 0.65 and a A(meas) of 0.66(l). t rNnded to the nearest 5% . Furthermore, all samplesfrom the structural study

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 630 THE CANADIAN MINERALOGIST

that were inferred to be fully ordered have fellowships to TSE and a Canadian Museum of calculatedvalues of A statisticallvidentical to 1. Nature researchstipend to TSE. We appreciatethe careful editing by R.F. Martin. SUMMARY REFERENCES l. Wodginite, sensustricto, hasthe generalformula ABC2Oy,where,A = M[,8 = Sn4+,C: Ta, AppLEMAN,D.E. & Ever.rs, H.T., Jn' (1973):Job 2. Other cation constituents (minor ones in 9214:indexing and least-squaresrefinement of powder parentheses)are: .,4 site: Fe2*, Li, (Sn2+, ca); I diffraction data. U.,S.Geol. Surv., Comput. Contrib. ( Doc. PB2-16188). site:Ti, Fe3+,Ta, (Sc, Zr); C site: Ta, Nb, (W). 20 mS seem 3. Vacanciesoccur at the.4 cation site and 6enN9, P. & Enctt, T.S. (1985):Some recent advances likely for the B site. in the mineralogyand geochemistryof Nb and Ta in 4. Taking the sitepreferences of Fd+ (A site)versus rare-element granitic pegmatites. Bull. Mindral. l0E, Fe3* (B site) into consideration, cation-ordered 499-532. formulae can be calculated for wodginite-group minerals.As a consequenceof the formula-calcula- - & - (1990): Mineralogy of and tion procedure,Fe2+:Fe3* and the Li contentscan tantalum: crystal chemical relationships, paragenetic be calculated. aspects and their economic implications. In oLan' (P. 5. Multiple-regressionanalysis of compositional thanides,Tantalum and Niobium M

AcrNowlencEMENTS -& Suvola, J. (1987):The compositionof stannomicrolite. Neues Jahrb. Mineral. Monatsh., The authorsthank A.J. Anderson,E.E. Foord, 249-252. A.-M. Fransolet,R.V. Gaines,R.I. Gait, O. von F.C. & denNV, P. (1992b):The Lahti, Luna, I. HAwTHoRNE, Knorring, R. Kristiansen, S. J. wodginite group. I. Structural crystallography.C-ar. Nakai, Ni Yung Xiang, the late W.L. Roberts, the Mineral.3l). 597-611. late T.G. Sahama,R. Schooner,J. Siivola,A.C. Turnock, A.V. Voloshin,S. White and M.A. Wise FERcusoN,R.B., HAwTHoRNE,F.C. & GnIce, J.D. for the loan of samples.E.E. Foord generously (1976):The crystal structuresof , ixiolite and suppliedinformation about Sn2+in wodginite. We wodginite from Bernic Lake, Manitoba. II. are indebtedto the Tantalum Mining Corporation Wodginite. Can. Mineral. 14, 550-560. of Canada Limited for accessto the workings of the Tanco pegmatite during the period of study, GArEHousE,B.M., NEcAs,T, & RorH, R.S. (1976):The and to Tanco mine geologistsR.A. Crouseand P. crystal structureof M-LiTa3Os and its relationshipto the mineral wodginite. J, Solid Stote Chem. 18, l-7. Vanstone for their help with our underground was in activities. Financial support for the study Gnrcr, J.D. (1970): The Nature ond Distribution of the the form of a NSERC operating grant, DEMR Tantolum Minerals in the Tonco Mine Pegmatite at ResearchAgreement, and Tantalum-.Mining Cor- Bernic Lake, Manitoba. M.Sc. thesis, Univ. poration of Canada contract to *, a NSERC Manitoba, Winnipeg, Manitoba. operating grant to FCH, and a NSERC PostgraduateScholarship, University of Manitoba Kourov, A.L (1970): Relationship between the X-ray

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021 CRYSTAL CHEMISTRY OF'THE WODGINITE GROUP 631

constantsof columbites and composition. Dokl. Acad. SHANNoN,R.D. (1976):Revised effective ionic radii and Sci. USSR195, 117-119. systematicstudies of interatomic distancesin halides and chalcogenides.Acta Crystallogr. A.32,7 5l-i 67. - & DUBIK,O. Yu. (1983): Chemical composition of wodg:rnites and possible conditions of their formation in nature. .lz Problemy Iftistallokhimii i SDoRENKo,G.S., SoLNTsEva,L.S. & Gonzurvsreva, Genezisa Mineralov (A.V. Sidorenko & D.V. S.A. (1974): Minerals with wodginite structure and Rundkvist, eds.). Izdatel'stvo Nauka, Leningrad their syntheticanalogues, Dokl. Acad, Sci, USSR216, (12+1291'in Russ.). t27-t30.

LoNa, G.L., CRANSHAW,T.E. & Lor.lcwonrH,G. (1983): TuRNocK,A.C. (1966):,Synthethic wodginite, tapiolite The ideal Mdssbauer effect absorber thicknesses. and tantalite. Can. Mineral. 8,461470. MdssbauerEffect Ref. Data J. 6,4249. VoLosHrN, A.V., PAKHoMovsKn, Ye.A. & MAKsrMovA,N.V. & Knvosrova, V,A. (1970):Mineral- BAKHcHrsARArrsEv,A.Yu. (1990):Lithiowodginite - ogy and crystal chemistry of some tantaloniobates. a new mineral of the wodginite group from granitic Dokl. Acad. Sci. USSR 193. ll9-122. pegmatitesin easternKazakhstan. Mineral. Zh. 12, 94-100(in Russ.). NEwroN, R.C. & Woon, B.J. (1980):Volume behavior of silicatesolid solutions.Am. Mineral. 65,733-745.

ScHtrmtzrn, K. & BaNr, H. (1984):Crystal chemistryof tsilaisite (manganesetourmaline) from Zambia. Neues ReceivedJune 26, 1991, revised manuscript acceptedMay Jorhb. Mineral. Monatsh,, 6l-69. 21,1992.

Downloaded from http://pubs.geoscienceworld.org/canmin/article-pdf/30/3/613/3420475/613.pdf by guest on 03 October 2021