HRTEM Characterization of Scapolite Solid Solutions Isnnrlor- Hlssln

American Mineralogist, Volume 73, pages 119-134, 1988

HRTEM characterization of scapolite solid solutions

Isnnrlor- HlsslN, Pnrnn R. Busrcr Departments of Geology and Chemistry, Aizona State University, Tempe, Aizona,85287, U.S.A.

Ansrn,lcr The scapolite composition varies between that of the end members marialite Nao[AlrSinOro]Cland meionite Cao[AluSiuOro]COr.Two binary solid-solution seriesexist betweenthe end members: series(a) between0 and 750lomeionite is formed by a coupled .COr]Alr, replacement[Nao .Cl]Si, + [NaCa, and series(b) between7 5 and 1000/omeionite is formed by a coupledreplacement [NaCar.COr]Si = [Cao'COJAl. Thesereactions behave independently. High-resolution transmission-electronmicroscopy indicates that series (a) scapolitesbelong to spacegroup P4 or P4/m, the result of ordering of the clusters (in the squarebrackets) and that series(b) scapolitesbelong to spacegroup 14/m, the result of disordered clusters. The ordering of the clusters in series (a) gives rise to antiphase domains; suchdomains are not observablein series(b). Although crystalsfrom both series may be intergrown, the order-disorder relationships resulting from the net chargedi.ffer- encesamong the clustersindicate that they give rise to chemical domains, despite the fact that the framework is continuous. The observedcompositional variations in scapolitecan be explainedadequately in terms of the two independentsubstitution schemesgiven above. Al-O-Al bonds, which tend to be unfavorable in aluminosilicates,are more numerous in series(b) scapolite; it seemstikely that the highly charged [Cao'COr]6* clusters stabilize such bonds.

INrnonucrroN (b) are disordered.This differencein ordering arisesfrom clusters and pro- Scapolitesare a fairly common group of rock-forming the net charge differencesamong the aluminosilicate minerals that occur in a wide variety of vides an adequateexplanation for the unusual substitu- preliminary report appeared metamorphic and altered igneous rocks. They are of in- tion schemesin scapolite.A (Hassan 1986). terest becauseof their unusual stoichiometry, relative as an abstract and Buseck, structural complexity, and the possibility that they could act as storage sites for volatiles in the lower crust and Pnnvrous woRK uppermantle (Newtonand Goldsmith, 1975,1976;Lov- There is no sharp structural changeacross the scapolite ering and White, 1964). series,so scapolite commonly is consideredas a contin- This work forms part of an on-going study to syste- uous solid-solution series between the end members matize the role of volatile components that are of pet- Nao[AlrSinOro]Cl(marialite : Ma) and Cao[AluSiuOro]CO' rologic importance and that also give rise to important (meionite : Me) (e.g.,Deer et al., 1963;Lin, 1975;Lin featuresin framework alumininosilicate mineral groups. and Burley,1973a,1973b,1973c,1975; Levien and Pa- These include, for example, the sodalite group (Cl-, plke, 1976; Strunz, 1978). The meionite percentage[o/o SO;-, 5'-, OH-, HrO; e.g.,Hassan et al., 1985),helvite Me : l00Ca/(Na + Ca)l is used as the chemical index group (S'z-;e.g., Hassan and Grundy, 1985),cancrinite to indicate the compositions of scapolite (Shaw, 1960; group (COl-, SO? , Cl , OH-, HrO; e.g., Hassanand Deer et al., 1963).Evans et al. (1969)proposed two sub- Grundy, 1984),and scapolitegroup (Cl , CO3-, SO?-). stitution schemesfor scapolites:between pure Ma and In all these related mineral goups, large anions are in NaCar[Al,Si?O,o]CO3(Me,r), the substitution is cagesor channelsformed by AlOo and SiOo tetrahedra. Na.SirCl + Ca.AlrCOr,and betweenMe- and Me'oo,the Scapolitesare interesting minerals to examine with high- substitution is NaSi + CaAl, as in plagioclasefeldspars. resolution transmission-electronmicroscopy (nnrrrvl) be- From Me- to Me,oo,the anion site is theoretically filled causethey provide an opportunity to study order-disor- wirh CO3-. der among clustersof atoms. Becauseof the dual substitution schemesin scapolite, This study indicates that scapolite forms two solid so- the resultsofEvans et al. (1969) have been interpreted lutions: series(a) between0 and 750lomeionite and series as indicating two binary solid-solution series that meet (b) between 75 and l00o/omeionite. Ordering of [Nao. at Me,, (Ulbrich, 1973a Oterdoom and Wenk, 1983). Cll3+and [NaCa.'COr]5+clusters occurs in series(a), but Aitken et al. (1984) also suggestedthe presenceof two the clustersof [NaCar.COr]s+and [Ca.'CO.]6* in series seriesbecause ofa changein the spacegroup. However, 0003-004x/88/0I 02-0 I I 9$02.00 119 120 HASSAN AND BUSECK: SCAPOLITE SOLID SOLUTIONS previous studies on the compositional variations across work. Scapolite in space group P4r/n contains three tet- the scapolite series, particularly around Merr, show no rahedralsites (T1, T2,T3 Fig. l), eachof multiplicity 8, sharp structural changethat would suggestdivision ofthe whereasin spacegroup 14/m, T2 and T3 combine (: T2') series,so the presenceoftwo binary solid-solution series to give multiplicity 16. Scapolite contains two formula was still debatableand is addressedin this study. units per cell, so the composition of Mer^ can be written Lin (1975) suggestedthat the unusual stoichiometric as NarCa,[Al8 Si,6Oo8]Cl' CO,. A structurecorresponding variations in scapolitearise from the chemical and phys- to this composition, as far as the framework atoms are ical differencesbetween the Cl- and COr2-anions. Ac- concerned,is best described by space group P4'/n (Lin cording to Lin (1975),tilting of the COI group causes and Burley, 1973a, 1973c; Levien and Papike, 1976). The displacement of Ca2+and Na+ cations along the c axis Tl and T3 sites are fully occupied by Si, and T2 is oc- and ordering of Al3* and Sio* cations. This tilting has not cupied by Al. This occupancy gives rise to an ordered been detected, however, in other structure refinements distribution of Al and Si (Fig. l) with no A1-O-AI bonds; (Levienand Papike,1976; Papike and Stephenson,1966; such bonds tend to be unstable in aluninosilicates Aitken et al., 1984). The compositionalvariations are (Loewenstein,1954). further explained in terms of local neutralization of elec- For scapolites that contain more Si than Merrr, the trostatic chargesbetween the tilted COr2-groups, the dis- excessSi replacesAl in the T2 site, and the T2 and T3 placed Na and Ca atoms, and the tetrahedral framework sitesare disordered.In pure Ma, the T2 and T3 sites are cations(Lin, 1975).Chamberlain et al. (1985)interpreted equally occupied by 7r Si and 3/eAl (: T2'), and Tl is the compositional variations as achieving local charge fully occupiedby Si; the symmetry, thereforeincreases to balancebetween the Na+ and Ca'* cations, and Cl- and I4/m (Lin and Burley, 1973a).This disordered arrange- COI- anions. They further suggestedthat the two possible ment gives rise statistically to aft' Al-O-Al bonds per exchangereactions, NaCl + CaCO, and NaSi + CaAl, formula unit in pure Ma (Oterdoomand Wenk, 1983). do not behave independently. For scapolitesthat contain more Al than in Me.rr, it Basedon X-ray diffraction studies,the two established appearsthat neither the T2 nor T3 site can contain more spacegroups for scapolite are I4/m (Papike and Zoltai, than 0.50 Al and that therefore Al replacesSi in the Tl 1965;Papike and Stephenson,19661,Ulbrich, 1973b;Pe- sites.The T2 and T3 sites also attain a disordered Al-Si terson et al., 1979;Aitken et al., 1984) and P4r/n (Lin distribution so as to reduce the number of Al-O--Al bonds and Burley,1973a,1973b,1973c,1975; Levien and Pa- formed (Lin and Burley, 1973a).In pure Me, each of the pike, 1976).However, basedon electron-diffraction stud- three tetrahedral sites contains 7z Si + YzAl, and the ies, the spacegroup is P4 or P4/m for intermediate mem- symmetry therefore increasesto 14/m. This disordering bers ofthe scapolitegroup (Phakeyand Ghose,1972). gives rise statistically to % Al-O-Al bonds per formula This conclusionis supportedby Buseckand Iijima (1974) unit in pure Me (Oterdoom and Wenk, 1983). Klee but not by Oterdoom and Wenk (1983). (1974a, 1974b)has shown that it is theoretically possible Scapoliteconsists of two types of four-memberedrings, to avoid Al-O-Al bonds up to an AllSi ratio of 5/7 (i.e., each of which is made up of AlOo and SiOo tetrahedra Me,,) by having 500/oAI on Tl and 37.5o/o Al on T2 (: (Fig. la). The type I ring consistsonly of Tl tetrahedra T3). that point in the same direction, whereasthe type 2 ring The Al-O-Al avoidance rule holds in framework alu- consists of both T2 and T3 tetrahedra that point alter- minosilicates that have an Al/Si ratio of l: I , as in anor- nately up and down. The two types ofrings are connected thite, sodalite-group, and cancrinite-group minerals. Al to form five-membered rings along columns parallel to and Si are presumably also distributed in a fully ordered the c axis (Fig. 1b). Continuous oval-shaped channels mannerin pure Me, which also has an Al:Si ratio of l:1. contain the Na+ and Ca2+cations (Fig. la). Each Cl and The problem is that scapolite contains five-membered CO3 anion is surrounded by four (Na,Ca) cations. The rings. In pure Me, one-half of these rings must contain 3 resulting anion-cation clustersare in large cagescentered Al and 2 Si tetrahedra, so the Al-O-Al avoidance rule at (Va,t/t, Ya)and (%, Te,T+) (Fie. 1b). has to be broken. The space groups I4/m and P4r/n reflect almost no Lin and Burley (1973c)have shown that the intensities difference in the scapolite structure as far as the distri- ofthe type b (h + k + l: odd) reflectionsare strongest bution of the interframework (Ca2+,Na+, Cl , CO3-) ions for Me.r, and decrease(extrapolated) to zero at the end are concerned.The (Na,Ca) atoms are on the same set of membersMa and Me (Fig. 2). They attribute thesechanges equivalent sites (multiplicity 8). The Cl and COI- an- to Al-Si ordering on the T2 and T3 sites. In Merrr, the ions also shareequivalent sites (multiplicity 2) if the po- CI:CO. ratio is l:1, and they suggestedthat distortions sitional disorder of the planar CO.2 group is ignored and arising from substitutions of these anions add to the in- if it is consideredas a sphere.However, if the interframe- tensities of the type b reflections.Antiphase domains are work ions in intermediate scapoliteswere ordered, then observableonly with type b reflections. These domains neither spacegroup I4/m nor P4r/n would be adequate are interpreted as arising from either ordering ofCl- and for describing their structural roles. Cq @hakey and Ghose, 1972) or from Al-Si ordering The structural differencebetween I4/m and P4'/n for in samplesthat contain little Cl- (Oterdoom and Wenk, scapolitearises from the Al/Si distribution in the frame- l 983). a t001l

Fig. I . The structure of marialite scapolite, with the unit cell outlined. The framework tetrahedral sitesare labeled.The Cl atoms are drawn largest,followed by Na, then oxygen atoms. (a) Projection along c axis showing the two types of four-memberedrings and the continuous, oval-shapedchannels b that contain the Na atoms. (b) StereoscopicD-axis projection (c-axis vertical) showing a Iarge cage 10101that contains [Nao.Cl]3* types of clustersand columns that contain the frve-memberedrings in which AI-O-A1 bonds can be formed. Data from Levien and Papike (1976) and drawn using onrEr (Stewart, r976). r22 HASSAN AND BUSECK: SCAPOLITE SOLID SOLUTIONS

as [Na..Cl]Si, + [NaCar.CO3]Alr.In series(a), the clusters are expectedto order becauseof the overall charge differenceof +2 valenceunits (v.u.) (+3 v.u. for [Nao' Cll3* clustersand +5 v.u. for [NaCar'COr]s+clusters). Such cluster ordering may also be coupled to Al-Si ordering. In series(b), the coupled substitution is [NaCar' CO3lSi+ [Cao.COr]Al;however, because of the net charge difference of + I v.u. between the clusters there is less tendency for them to order. In series (b), we consider ordering of Na* and Ca2+cations since both anion sites are filled with CO3-, and the compositional difference between the clusters can be expressedas a Na+ = Ca2+ substitution. These two cations have similar radii and for each other, so Na-Ca disorder- Fig. 2. The relationshipbetween the intensityratio r of type commonly substitute b to typea reflections(r : >I(h + k + I : odd)/>I(h+ k + I : ing may be expected. even)and percentMe, and also r versusend-member cluster Al and Si ordering is possible acrossthe scapolite se- percentagederived from percentMe. The dashedline showsthe ries; however, HRTEMcannot distinguish Alr* from Si4* division of the compositionalseries at Merr. Note that from becauseof their similar scattering powers. This type of Merr,there is a changein scaleon thepercent Me axis.However, ordering is currently being studied by the uas Nun tech- the scaleon the cluster-axisis linear.In eachseries, the scaleis nique, and preliminary results were reported by Sherriff given in termsof percentageof the end-membercluster that is et al. (1986). on the right. The largestr valueoccurs in the Mer^ scapolite, wherethe [Nao C1]3+: [NaCar'COr]s+ ratio is l:1. If theclusters MarsnrA.r,s usnn aredisordered, the r valueshould be zeroin pureMa and also Evanset al. (1969)were made in Me,, to Me,oo.The solid trianglesrepresent theoretical data Samplesfrom the study of We a crystalwith com- points.The experimentaldata points(solid dots) are from Lin availableto us by D. M. Shaw. chose position (sample0N70 in Evanset al., 1969)because its and Burley(1973c), and their variation was representedby a Me.r, Mer,, compositionthat showsthe curve.Solid lines serve as reference to indicatethe deviationof compositionis closeto the reflections(Fig. 2). The sampleis from Mpwap- the datafrom linearity.The two samplesused in this studyare strongesttype b gneiss. indicatedby straightarrows. wa,Tanzania, where it occursin quartzofeldspathic The crystalsare transparent pale-yellow and ofgem quality. The sec- ond samplehas composition Merru (sample 0N47 in Evanset al., 1969)from Slyudyanka,Siberia, USSR and is of unknown Electrostaticenergy calculations for Mer- indicate that occurrence.The crystalsare white, brittle, and bladed.Of the the energyis minimized if Na+ is adjacent to Cl and if samplesthat wereavailable to us,this sampleis closestin com- Ca2+ is adjacent to COI- (Chamberlainet al., 1985). positionto the Me endmember. Moreover, thesecalculations suggestthat short-rangeor- deringof [Nao.Cl1.*and [Cao'COr]6+clusters is energet- ElncrnoN MrcRoscoPY ically favorable and might give rise to antiphasedomains. Electronmicroscopy was performed using a JEoLzoocx elec- Scapor,rrr STRUCTURALMoDELS tron microscopeoperated at 200 keV with a LaBufilament. A +12' double-tilt,top-entry goniometer stage was used. The According to the X-ray structuresof scapolite,the two sphericalaberration coefrcient, C", ofthe objectivelens was 1.2 cagesper cell are identical as far as the Al-Si framework mm. A 40-pm(i.e., radius of 0.76A r) objectiveaperture was is concerned.In the end members Ma and Me, the cages centeredabout the incidentbeam, and a 400-pmcondenser ap- contain [Nao.Cl]3* and [Cao'CO.]6+ clusters, respective- erturewas used to recordimages at 530000 magnification. ly. If theseclusters were the only typesin the intermediate Fragmentsof scapolitesuitable for study wereobtained by members, then for any ratio of theseclusters, we should crushingthe mineral in acetoneusing an agatemortar and pestle. get a valid composition for scapolite.For example,a ratio The crystalsin suspensionwere deposited on holeycarbon sup- port wererecorded from regions. of [Nao.Cl]3+to [Cao.COr]6*of l:l impliesthat in Mero films,and nnrrlr data thin the CI:CO. ratio is l:1. However, this ratio occurs in INa.lcB sTMULATToNS Me.rr. Thus, the occurrenceof only [Nao.C1]3+and [Cao. COrlu* clusterscannot give rise to the observed compo- Prior to the present study it was unclear whether it sitional variations in scapolite.The additional occurrence would be possibleto distinguishclusters of [Nao.Cl]3* 'CO,1t* of at least [NaCa. clustersis necessaryto explain from [NaCa.'COr]'* usingHRrEM imaging. Owing to the scapolite compositional variations, as is shown in this instability of scapolite in the electron beam (Buseckand study. The electrostaticenergy calculations are based on Iijima, 1974), it was decided to by-pass the painstaking Merrr, so the calculations also indicate that ordering of task of working with such materials until we were con- [Nao. Cl]3* and [NaCa, . CO.], * clustersis energeticallyfa- vinced by image simulations that the different clusters vorable (Chamberlainet al., 1985). could be distinguished from each other. The coupled substitution in series(a) may be written Simulations of ftigh-resolution /attice lmageswere ob- HASSAN AND BUSECK: SCAPOLITE SOLID SOLUTIONS I z5

ORDER tained with the snnr-r set of programs (O'Keefe and Bu- DISORDER seck,1979; O'Keefe et al., 1978),which are basedon the multislice method (Goodman and Moodie, 19741,Cowley Ioot and Moodie, 1957). Images were calculated for several projection directions over a range of d,efocus(M from -+-OOto - 1400 A (increments of 100 A) and for crystal thicknesses(fO from about 3 to 200 A (increments cor- respondingto the repeat distancein the projection direction). Calculationswere performed using the electron optical parameters given above for the JBoI- zoocx microscope, together with a divergence angle of 0.001 radian (estimated from the width of a diffraction spot under imaging conditions) and a depth of focus of 50 A. Image simulationswere made for two structural models TIool basedon the Mer^ composition (with tetragonal cell pa- rametersa : b : 12.069A and c : 7.581 A;. A aisor- dered model is based on the average X-ray structure (P4,/n) of Levien and Papike (1976), and an ordered model is basedon the sameframework, but with the clusters [Na.'C1]3+and [CarNa'COr]s+ordered in the cages at (r/t, t/r, t/e)and (1/t,3/e,3/t),respectively. These structural models were input in spacegroup Pl for the purpose of image calculations.The simulated imageswere all output on the same relative contrast scale. Through-thickness,through-focus seriesof images for itt rol all the projections show systematictrends. Although the images are nearly thickness independent for the range calculated,the objectivelens defocusis critical. Figure 3 showsa comparison of a set of calculatedimages for var- ious zone axesin the two structural models.These images are selectedbecause they provide the maximum contrast differences.Differences between the two structural models are seenin all the projections ofFigure 3, but with varied degreesof clarity. The arrows mark rows that are identical in the disorderedmodel but that differ in the ordered t model. For example,the [001] zone showsthe positions Ito of the clustersas having the same contrast for the disordered model and different contrast for the ordered model. In the [001] zone, the clusterscorrespond to the positions of the large white spots with dark centers. [Nao'Cl]3+ is lighter and [NaCa.'COr]s+ is darker, but the differenceis not great. This small differencewould be difrcult to de- tect in experimentalimages. Similar small differencesalso occur for the [ 10] zone. However, the calculatedimages for zonestl00l, tOl ll, and [012] clearlyshow differences betweenthe two models that could allow them to be dis- [CI1zl tinguished from each other in experimental images. The imagesare strongly defocusdependent, so the two models are only distinguishableat specificdefocus values; these values changewith thickness and projection direc-

F- Fig. 3. Calculated images for ordered and disordered struc- turaimodels. In the following zones,the values for thickness(A) -800; and defocus(A), respectively,are (a) [001]: 22.7, O) [100]: '12.4, -800; -800; -1200; (c) [110]:76.8, (d) [0ll]:51.4, (e) - l0l2l:75.9, 1200. t24 HASSAN AND BUSECK: SCAPOLITE SOLID SOLUTIONS

tion. This point is illustrated in Figure 4 forthe projection along the l0l2l zone for the two structural models. The calculatedimages show that the two models are only clear- ly distinguishableat the defocusvalue of -1200 A. Sim- ilar features occur for the other projections. Therefore, extreme care is necessaryin recording and interpreting the experimental images; otherwise the results could be misleading.

Er,BcInoN-IIFFRACTIoN PATTERNS Reciprocal-spaceobservations of scapolite show that the reflections can be divided into four types: l. Type a: (h + k + l): even.These occur as strong diffraction maxima. 2. Type b: (h + k + l): odd. Theseoccur as sharp spots and are weaker than type a reflections. They are forbidden in spacegroup 14/m. 3. Type c: (00/), /: odd. Although forbiddenin both spacegroups 14/m and P4r/ n, they appear as sharp spots in some scapolite diffraction patterns. They violate the existenceof the 4, screw axis in spacegroup P4r/n. 4. Type d: (hk}), h + k: odd. Although forbiddenin both spacegroups I4/m and,P4r/n, they are present as sharp spots in some scapolite diffraction patterns. They violate the existenceof the r glide plane in spacegroup P4r/n. The type c and type d reflections do not disappearon tilting about systematicrows containing these spots, and therefore they are unlikely to be causedby multiple diffraction. Moreover, they were recorded both at 200 keV and with a Phillips 400T microscope operating at 100 keV, and they show no difference. A [001]'zone selectedarca electron diffraction (seeo) pattern for Mern, is shown in Figure 5a. Similar patterns were obtained for the Merru. In over-exposedsAED patterns for the [001] zone (not shown), type d reflections occur as diffuse spots in the zero-order Laue zone, and sharp but weak type b reflections occur in the first-order Laue zone. These reflections, however, appeat stronger in other zones (Fig. 5b). Moreover, the ordering reflections cannot occur by multiple diffraction in the [001] zone. The Mern, crystals(Figs. 6a, 6d) and someof the Me,u crystals(Figs. 6b, 6e) have similar reflections.These senp patterns show reflections of types b, c, and d that are absent for some of the Merru crystals (Figs. 6c, 6f), indicating that there is local variation in composition or ordering within the Merru sample. Reflectionsof types b and c for the [00] zone are weak comparedto type a reflections.However, in the [0ll] zone, type b and type d reflections are strong compared to the type a reflections,and they do not decreasein intensity as rapidly with beam exposureas they do in the 75.9 A THICKNESS [100] zone. Therefore, different results may be obtained [otzl zoNE for different orientations, dependingon the time taken to Fig.4. Comparisonofdefocus effectfor a75.9-4, thickness recordthe imagesbecause scapolite damages more in some in the [012] zone for the ordered (right) and disordered (left) orientations than in others. The reflections oftypes b, c, models. and d can be made to disappearwithout losing all of the HASSAN AND BUSECK: SCAPOLITE SOLID SOLUTIONS 125

Fig. 5. (a) A [001]-zone sAEDpattern showing type a reflections.(b) A [012]-zone sAEDpattern for the Mern, sample showing strongtype b (e.g.,021, 142,221, 142) and type d (e.g.,100, 300, 500) reflections. intensities in the main type a reflections. This effect is indicatecluster ordering. The framework atoms are shown particularlyevident for the [001] and [100] zones.Levien as small dots with the sizeof the tetrahedralcations being and Papike (1976) observed that the reflections that vi- smaller. The unit cells in the structure projections serve olate 14/m symmetry decreaseconsiderably in intensity as scales. with temperature up to 1000 "C but do not disappear. The [001] sAEDpattern showsonly type a reflectionsin Throughout this temperaturerange, the constancyofthe this orientation (Fig. 5a), and therefore ordering ofclus- Al-O and Si-O bonds indicate practically no Al-Si dis- ters cannot be seenin the [001] image (not shown). A order. These observations, therefore, suggestthat Al-Si [001] imagewas shownby Buseckand Iijima (1974),and order cannot solely be related to the reflections that vi- some imagesare given below but discussedin a different olale I4/m symmetry. Moreover, beam-damage and context(Fig. 1l). heating seemto causedisordering of the interframework The serp pattern for the [00] zone (Fig. 6a) shows 10ns. that the extra reflections are weak compared to the main The electron-diffraction results suggestthat either P4 type a reflections.An optical diffractogram of a [100] im- or P4/m is the likely spacegroup for intermediate scap- age does not show the 001 and 010 reflections,but it olites that contain Cl . Crystals that tend to be Cl free shows weak 021 reflections: however, the indicated or- only show type a reflections(some crystal fragmentsfrom dering is not evident in the image (not shown). Mernu,this study; Merr, Petersonet al., 1979; synthetic The sneo pattern for the [012] zone (insert in Fig. 7) Mero,Aitken et al., 1984;Menr, Lin and Burley, 1973c; was photographedbefore and after imaging and did not Ulbrich, 1973b). Published chemical analysesfor these change significantly, thus indicating little beam damage samplesalso indicate that they contain little or no Cl to the crystal. An optical diffractogram of the image (in- The Cl-free crystals and the end member Ma, therefore, sert in Fig. 7) showsthat the crystal is well oriented and, belong to spacegrorry I4/m. in particular, that the 001 and 021 reflections (arrows) are present.The experimental image (Fig. 7) shows con- ' trast that arisesfrom ordering of [Nao'Cl]3* and [NaCa, Hrcrr-nnsor-urroN IMAGESoF scApoLITE COrlt* clusters.These contrast differencesare seenalong In order to estimatethe effectof beam-damage,we rou- alternatingplanes {021} and {042} (long arrows), and tinely recorded sAEDpatterns before and after imaging. {100} and {200} (short arrows). Optical diffractogramswere made from the experimental The experimental and the calculated images (Fig. 7) imagesto determine which reflectionscontributed to the suggestthat the Na+ and the Ca2+positions, while difr- images.All suitableimages were matched by detailed im- cult to see,are distinguishable.The Ca2* positions appear age simulations. darker on both sides of the dark rows of spots, whereas In all the imagesgiven below, structure projections are the Na+ positions appearlighter on both sidesof the white scaledand oriented so that they superimposedirectly on rows of spots. the electron micrographs. Their positioning is based on The contrast of equivalent spots varies considerably comparison betweenthe simulated and experimental im- acrossthe experimental image (Fig. 7) and is even more ages.In the structure projections, the [Nao'Cl]3+ clusters pronounced in some images for the Me, u sample (not are unshadedand [NaCar'CO.]5* clustersare shadedto shown). Although this effect may arise from beam dam- 126 HASSAN AND BUSECK: SCAPOLITE SOLID SOLUTIONS It 001 [0tll

Fig. 6. Comparisonof [100] (Figs.6a, 6b, 6c) and [01I ] (Figs.6d, 6e, 6f) SAEDpatterns for Me,n, @igs.6a, 6d) and Me,, u (Figs. 6b, 6c, 6e, 6f). Reflectionsof the following types are shown:type b (Figs.6a, 6b contain 021,012,014, 032, etc.; Figs. 6d, 6e contain lll,122, etc),type c (Figs.6a, 6d contain001, 003, etc.),and type d (Figs.6d, 6e contain 100,300, etc.).These reflections are absentin some crystal fragmentsof the Merru sample (Figs. 6c, 6f). HASSAN AND BUSECK: SCAPOLITE SOLID SOLUTIONS r27 Iot2l

Fig. 7. A t0121image of the Mer, , sample. The seep pattern and the optical difractogram are shown as inserts at the upper- -1200 and iower-left, respectively.The simulated image in the central insert is basedon the ordered model (I1 :75.9 L, Af : A; cf. Fig. 3e). The projected structure (lower right; the a axis is horizontal and the [02I] axis is vertical) shows ordering of [NaCa,' CO3l5*and [Nao.Cl]3* clustersthat is evident in the image. t28 HASSAN AND BUSECK: SCAPOLITE SOLID SOLUTIONS

Fig. 8. A [0] U Hnreu image (top) and low- magnification image (bottom) of the Me'n ' sample.Inserts ofthe sAEDpattern and optical diffractogramare given in the upper- and lower- left corners, respectively, in the image at the top. The calculatedimage in the central insert is basedon the orderedmodel (H : 51.4 A. A/: -800 Ar cf. Fig. 3d). Ordering of [Nao Cll3* and [NaCar'COr]5*clusters is indicated in the projected structure (lower-right corner; the a axis is horizontal and [0ll] is vertical) and is more evident in the image shown at the bottom in which white and black arrows point out rows that have light and dark contrast, respectively. age, it is interpreted as disorder, reflecting the excess (Fig. 7) nearly as well as the calculated image for the amount of one cluster over the other since their ratio is ordered model. Nevertheless,calculated images based on not l:l in the sample. Moreover, this randomnessof the disordered model were simulated with both crystal equivalent spots showing different contrasts cannot give tilt and beam tilt introduced as variables, as these vari- rise to the sharp spots seenon the sAEDpatterns. ablescan give rise to imageartifacts (Smith et al., 1983). The calculated [012] image based on the disordered However, it was not possible to reproduce the experi- model (Fig. 3e) does not match the experimental image mental image (Fig. 7) by using the disordered model and HASSAN AND BUSECK: SCAPOLITE SOLID SOLUTIONS r29

I02|

Fig. 9. A [021] image of the Mer* sample showing a sharp APB (arrows). these additional variables. Therefore, we conclude that ing may be easierto seein the low-magnification image the observedvariation in contrast in Figure 7 arisesfrom shown at the bottom of Figure 8, where it is separated ordering of [Nao.Cl]3*and [NaCa..COr]'* clustersand from the HRTEMimage at the top. not from experimental variables. The ordering of these clustersis also shown in Figure 8. The sAEDpattern and Axrrpnl.sn DoMATNSrN scAPoLrrE optical diffractogram (inserts in Fig. 8) show the I lI and The observedordering of [Na4.CU3+ and [NaCar.COr]s* 100 reflections (arrows); alternating {lll} and {222} clusterscan give rise to antiphasedomains. If the cavity planes(long arrows)and tl00) and {200} planes(short at (Va,Vt,7+) is occupiedby [Na..Cl]3+clusters, then the arrows) show different contrast in the image. This order- (3/a,3/t,3/+)cavity would be occupiedby [NaCa..COr]s* 130 HASSAN AND BUSECK: SCAPOLITE SOLID SOLUTIONS

Fig. 10. A [012] image of the Mer* sample showing a wide APB (arrows). clusters; in the adjacent domain the occupancy of the age, and the sAEDpattern are also shown as inserts in cavities could be reversed.Antiphase domain boundaries Figure 9. Figure l0 shows a wider APB (arrows). A saep (APBs) in scapoliteare difficult to image in nnrev mode; pattern, simulated images, and a HRrEMimage for this they cannot be seenon the rEv screenbut only on pho- projection are given in Figures 5b, 3e and 4, and 7, re- tographic prints. spectively. Shifts in fringes can also arise from a change Figure 9 showsan APB having the shapeof a question in thickness, but simulated images for thicknessesup to mark. The fringes are shifted by t/zaoacross the sharp 165.2A ([021] zone) and 177.2A (Otzl zone),respec- APB (anows). The projected structure, the simulated im- tively, do not show any shift or reversalofcontrast ofthe HASSAN AND BUSECK: SCAPOLITE SOLID SOLUTIONS l3t lattice fringes. We interpret these APBs as arising from break between the Ma and Me end members indicates ordering of [Nao.CL]3+and [NaCar.COr]'* clustersin that all the clustersgive rise to similar geometricaleffects scapolite. in the structure, although the clusters are chemically di- verse. Crmvrrcar, DoMATNSrN THE Mnrn.u scApoLITE Several plots illustrating the variation in chemical The electron-diffraction results, particularly Figure 6, composition of scapolite show breaks at Mer, (Evans et suggestthat there is local variation in composition or al., 1969).Figure l2 showsthe compositionalvariations ordering within the Mern.usample and that Cl-free scap- of natural scapolites;it is the same as that observed by olites generally show only type a reflections. Moreover, Evanset al. (1969).[In plotting Fig. 12 (and Fig. 2), we the nnrru images show ordering of [Nao.Cl]3+ and had the choice of making either the percent Me axis or [NaCar.CO,]s+clusters in series(a) scapolite(Mern , and the cluster axis linear, and the latter was chosen. If the some crystals of Me, u) and lack of ordering of [NaCar. percent Me axis is made linear, the inflection point at CO,l'* and [Cao.CO3]6*clusters in series(b) scapolite Me- is still presentbut appearsmore subtle.l Our results (somecrystals of Mernu). show that the observed linear compositional variations Figureslla, llb, and llc were taken at about l5-s are the result of substitution between [Nao'Cl]Si, and intervals from one region of a Me?e6 crystal. This series [NaCa,.COr]Al, in series (a) and substitution between ofexposuresshows the resultsofreactions that are caused [NaCa,.COr]Siand [Ca,''COr]Alin series(b), thus giving by the electron beam. The images show a continuous rise to the inflection at Merr. These two substitution crystal that is made up largely of series(b), with inclu- mechanismsdo behaveindependently. The composition- sions of many smaller domains that are presumably Cl- al variations in scapolite are, therefore, explained ade- bearing from series(a). Ifthese figuresare viewed at low quately in terms of the two independent substitution angle along the direction of the large arrow [i.e., along mechanisms given above, and they require the occur- the (110) planesl, it can be seenthat the (l l0) planesare rence of only three types of clusters.However, we cannot shifted by Vz(ll0) in the many small (Cl-rich?) domains exclude the possibility that other types of clusters may in Figures I la and I lb, but not in Figure I lc. This shift occur at different compositions. in fringes occurs in a direction associatedwith type a Antiphase domains are observedonly in scapolitecon- reflectionsand cannot arise from APBs. taining Cl-, and they arise from the observedordering of Optical diffractograms were taken from all images in [Nao.Cl]3*and [NaCa.'CO']'* clusters.Although Oter- Figure 11, and they show only type a reflections with doom and Wenk (1983) observedAPBs with type b re- minor changein intensities from one image to another. flections and attribute these to Al-Si ordering, it is pos- This observation does not rule out the possibility of or- sible that the small amount of Cl in their sampleformed dering of clusters in the many small (Cl-rich?) domains sufficient clusters that are ordered, as in the Mernu sca- becausethey may be too small to give additional reflec- polite, and thesegave rise to the APBs. tions. The maximum intensity ratio r of type b to type a The sizesof the (Cl-rich?) domains decreasefrom Fig- reflections occurs for the composition Merr, (Fig. 2) be- ure I la to I lb, until in Figure I lc the entire crystal ap- causeordering ofclusters is greatestwhen there are equal 'COr]s* pearshomogeneous. These observations indicate that the numbersof [Nao.Cl]3* and [NaCa, clustersin the electronbeam inducesdisordering among all the clusters. structure. The Mernusample (Fig. 2) has a small r value, More importantly, the results suggestthat crystals from and this can be explained by ordering of [Nao'Cl]3* and series(a) and (b) form chemical domains becausethe two [NaCar'COr]s+clusters in someparts of the crystal,which types are immiscible, despite the fact that the overall is reasonablebecause chemical analysesof this sampleby crystal is continuous with regardsto the framework. Evanset al. (1969)indicate that it containssome Cl-. As searchesdid not reveal type b reflections in Cl-free sam- Dtscussrou ples, series(b) samplesand the end member Ma belong Although there is no sharp structural break between to space group I4/m. Series(a) samplesbelong to space Ma and Me, scapoliteshould be consideredas two binary group P4 or P4/m. solid-solution seriesthat meet at Merr. The Meruis therefore an end member and should be consideredas an in- CoNcr-usroNrs dependentmineral species,as was also suggestedby Ait- Scapolite consists of two binary solid-solution series ken et al. (1984).The [Nao.Cl]'* and [NaCar.COr]5+ that meet at Merr. Therefore, a mineral having the Me^ clusters occur in series(a) scapolite,and they are gener- composition is a valid end member and should be con- ally ordered.[NaCar.COr]s+ and [Cao.COr]6*clusters in sidered as an independent mineral species.The compo- series(b) scapolite,and they are disordered.Clusters from sition varies by the coupled replacement [Nao'Cl]Si, + the two seriesprobably cannot be mixed homogeneously [NaCa..CO.]AI, in series(a) and by the independentcou- becauseof net chargedifferences, and so they give rise to pled replacement[NaCa.'CO3]Si + [Cao'COr]Al in se- chemical domains. These domains seem to occur in the ries (b). HRrEMimages indicate that clustersin series(a) Mernusample. The fact that there is no sharp structural are ordered and give rise to APBs, and that clusters in F F c c H

on o o

O il - (g F o 'to cf o H o

bo IJ. e]I$ oo @ts

I t.

o +

o o o ,"1eoc.tec] .rfoc'eeceNl ,ettc.teNl

* P

bb IJT t34 HASSAN AND BUSECK: SCAPOLITE SOLID SOLUTIONS series(b) are disorderedand so do not give rise to APBs. minurn-silicon distributions, carbonate group disorder, and thermal Series(a) scapolite belongs to spacegroup P4 or P4/m, expansion American Mineralogist, 61, 864-877. (1975) ofthe scapolitegroup. (b) group Lin, S.B. Crystalchemistry and stoichiometry and series scapolite belongs to space I4/m, as Acta GeologicaTaiwanica, 18, 36-48. does the end member Ma. Clusters from the two series Lin, S.B, and Burley, B.J (1973a)Crystal structureof a sodium and cannot be mixed homogeneouslybecause of order arising chlorine-nch scapolite.Acta Crystallographica,829, 1272-1278. from net chargedifference and give rise to chemical do- -(1973b) The crystal structureof meionite. Acta Crystallographica, mains, despite the fact that the framework is continuous. B29,2024-2026. - (l 973c)On the weak reflectionsviolating body-centeredsymmetry Finally, Al-O-Al bonds, which tend to be unfavorable in in scapolites.Tschermaks Mineralogische und Petro$aphischeMitteil- aluminosilicates,are more numerous in series(b) scapo- ungen,20,28-44. lite; it seemslikely that the highly charged [Cao.COr]6* - (l 975) The crystalstructure ofan intermediatescapolite-wernerite. clustersstabilize these bonds. Acta Crystallographica,831, 1806-1814. Llambias, E.J., Gordillo, C.E , and Bedlioy, Dora. (1977) Scapoliteveins in a quartz monzodiorite stock from Los Molles, Mendoza, Argentina. AcrNowr,oocMENTS American Mineralogist, 62, 132-135. Loewenstein,W (1954) The distribution ofaluminum in the tetrahedra provided The scapolitesamples were by courtesyof Dennis M. Shaw of silicatesand aluminates.American Mineralogist, 39, 92-96. We thank John C Barry and Fred Allen for helpful suggestionsand also Lovering, J.F., and White, A.J.R. (1964)The significanceof primary scap- the two anonymous refereesand JeffreyE Post, an unofficial referee,for olite in granulitic inclusions from deep-seatedpipes Journal ofPetrol- useful comments on this manuscript. This work was supported by NSF ogy,5,195-218. EAR-8408169; Grant microscopy was done at the ASU nnerrl facility, Newlon, R C., and Goldsmith, J.R. (1975) Stability of scapolitemeionite which is supported by NSF and ASU. (3CaAlrSirO8CaCO) at high pressuresand storageof CO2in the deep crust. Contribution to Mineralogy and Petrology, 49,4942. -(1976\ RnrnnrNcns cITED Stability of the end-memberscapolites: 3NaAlSiO. NaCl, 3CaAlrSirO, CaCO:, 3CaAlrSirO. CaSOo.Zeitschrift fiir Kristalloga- Aitken, BG. (1983) f-Xco, stability relations and phase equilibria of a phie,143,333-353 calciccarbonate scapolite. Geochimica et CosmochimicaActa, 47, 35I- O'Keefe, M A., and Buseck,P R. (1979) Computation of high resolution )ol. reu imagesof minerals.American CrystallographicAssociation Trans- Aitken, B.G., Evans,H.T , Jr , and Konnert, J.A. ( I 984) The crystal struc- actions.1 5. 27-44 ture of a syntheticmeionite. NeuesJahrbuch fiir Mineralogie Abhand- O'Keefe,M.A., Buseck,P.R., and lijima, Sumio (1978)Computed crystal lungen,149,309-324. structureimages for high resolution electron microscopy.Nature, 274, Buseck,P R., and Iijima, Sumio. (1974) High resolution electronmicros- 322-324. copyof srlicates.American Mineralogist, 59,l-21. Oterdoom,W.H., and Wenk, H.R. (1983)Ordering and compositionof Chamberlain,C.P, Docka, J.A., Post,J.E., and Burnham,C.W. (1985) scapolite:Field observationsand structural interpretations.Contribu- Scapolite:Alkali atom configurations,antiphase domains, and com- tion to Mineralogyand Petrology,83, 330-341. positional variations. American Mineralogist, 70, 134-140. Papike, J J., and Stephenson,N.C. (1966) The crystal structlre of miz- Cowley,J.M., and Moodie, A.F. (1957)The scatteringof electronsby zonite, a calcium- and carbonate-richscapolite. American Mineralo- atoms and crystals.I. A new theoretical approach. Acta Crystallogra- gist,51,l0l4-1027. phica,10,609-619. Papike,J J., and Zoltai, Tibor. (1965) The crystal structureofa marialite Deer, W.A., Howie, R A, and Zussman,J (1963)Rock-forming min- scapolite.American Mineralogist, 50, 64 1-655. erals:4. Framework silicates.Longmans, London. Peterson,R.C., Donnay, Gabrielle, and LePage,Yvon. (l 979) Sulfatedis- Evans,B W, Shaw,D.M., and Haughton,D.R. (1969)Scapolite stoichi- order in scapolite.Canadian Mineralogist, 17, 53*61. ometry Contributions to Mineralogy and Petrology,24, 293-305 Phakey, P.P., and Ghose, Subrata.(1972) Scapolite:Observation ofan- Goodman, P., and Moodie, A F. (1974)Numerical evaluation of N-beam tiphasedomain structure.Nature Physical Science,238, 78-80. wave functions in electron scatteringby the multi-slice method. Acta Shaw, D.M. (1960) The geochemistryof scapolite.Part I. Previous work Crystallographica,A30, 280-290. and generalmineralogy Journal of Petrology, 1, 218-260. Grazian| Giorgio, and Lucchesi, Sergio.(1982) The thermal behaviour Sherrifi B.L., Grundy, H D., and Hartman, J S. (1986) Al-Si ordering of of scapolites.American Mineralogist, 67, 1229-1241. scapolitesfrom'eSi, and "Al, MAsNMR (abs.). International Mineral- Hassan,I , and Buseck,P.R (1986)A Hnreu characterizationofscapolite ogical AssociationAbstracts with Program, 229. solid-solutionseries (abs.) International MineralogicalAssociation Ab- Smith,D.J , Saxton,W.O., O'Keefe,M.A., Wood,G.J., and Stobbs,W.M. stractswith Program,120. (1983) The importance of beam alignment and crystal tilt in high res- Hassan, I , and Grundy, H.D (1984) The character of the cancrinite- olution electron microscopy.Ultramicroscopy, I l, 263-282. vishnevite solid-solution series.Canadian Mineralogist, 22, 333-340. Stewart, J.M., Ed. (1976) The xnav system. Technical Report TR-446. -(1985) The crystal structures of helvite group minerals, University of Maryland, CollegePark, Maryland, U S.A. (Mn,Fe,Zn)r(BeuSiuO.o)S,American Mineralogist, 70, 186-192. Strunz, H. (1978) MineralogischeTabellen Geestand Porting, kipzig. Hassan,I., Peterson,RC., and Grundy, H.D. (1985)The structureof Ulbrich, H.H. (1973a) Crystallographic data and refractive indices in lazurite, ideally Na.Car(AlSi"O,o)S,, a member of the sodalite group scapolites American Mineralogist, 58, 8l-92. Acta Crystallographica,C4 l, 827-832. -(1973b) Structural refinement of the Monte Somma scapolite,a Klee, W.E. (l 974a)Al/Si distribution in tectosilicates:A graph-theoretical 930,6meionite SchweizerMineralogrsche und PetrographischeMitteil- approach Zeitschrift fur Kristallographie, 140, 154-162. ungen,53, 385-393. - (1974b) AllSi distribution in tectosilicates:Scapolites. Zeitschrift Iiir Kriqtallographie,140, 163-168 M.tNuscnrpr RECETvEDDecsl,rssn 5, 1986 Levien, Louise, and Papike, J.J. (1976) Scapolitecrystal chemistry: Alu- Mewuscnrrr AccEprEDSegrsMssr 23. 1987