An Electron Microscope Study of Leucophoenicite

American Mineralogist, Volume 68, pages 04, l9E3

An electronmicroscope study of leucophoenicite

Triuorny J. Wnrrer AND BRUcEG. Hvoe Research School of Chemistry, Australian National University P.O. Box 4, Canberra,A.C.T. 2600,Australia

Abstract

Three leucophoenicitespecimens were examinedin the transmissionelectron micro- scopeand by X-ray powder diffractionand electron-probemicroanalysis. A comparisonof observedand computer-simulatedstructure images supports Moore's proposalsthat the leucophoenicitestructure possessesedge-shared pairs of 04 tetrahedra that are half- occupiedby Si, and that the crystalscontain planarfaults parallelto (001).The latter are interpretedas lamellaeof other membersof a proposedleucophoenicite series (analogous to the well-known humite series).Such faults frequentlycause microtwinning, often on a scaleof less than 1004. Humite seriesmembers were also very occasionallyintergrown with leucophoenicite,and one specimen(from Pajsberg,Sweden) was associatedwith chrysotile. In the electron beam leucophoenicitereadily decomposesto tephroite, Mn2SiOa.

Introduction to a simplealloy structure(White and Hyde, 1983a).The Leucophoeniciteis dimorphicwith manganhumite,ide- cation array is twinned, cubic-close-packed(c.c.p.) Mn ally 3Mn2SiO+'Mn(OH)2, although the former always with Si in three-quartersof the Mn6 trigonalprisms at the containsseveral weight percent of CaO (in the present compositionplanes. Two protons occupy each of the specimens-6%). A morphological study by palache remainingprisms. Oxygenatoms are in SiMn3tetrahedra (1928) indicated that it is crystallographicallydistinct and Mn3 triangles (HMnr tetrahedra). from the humite minerals, and a preliminary power X-ray The cation array is a regular intergrowth of the Ni2In examination(Moore, 1967)suggested the existenceof alloy type (twinnedc.c.p. Ni . . . 2,2. . . with In in Ni6 severalleucophoenicite types with monoclinicor ortho- trigonalprisms) and the CrB alloy type (twinnedc.c.p. Cr rhombic symmetry. Finally, a singlecrystal X-ray struc- . . . I,l . . . with B in Cr6 trigonal prisms) (Hyde et al., ture analysisby Moore (1970)showed that leucophoeni- 1979). Thus, leucophoenicite is twinned c.c.p. cite is monoclinic P2/a, with the crystallochemical Mn . . . 1,2,2,2,1,2,2,2. . .or (1,23)withSior2Hineach formula Mnz[SiO+]z[(SiO4)(OH)2].We have reorientedto of the Mn6 trigonal prisms. The twin bandsare one or two P2/b, a-axisunique in order to facilitate a direct compari- atoms wide. This approach yields simple relations be- son with the structures of the humite minerals (pbnm or tweenthe structuresof a proposedleucophoenicite family P2rlb) and olivine (Pbnm). and those of the humite family and olivine. Figure I However, following Moore's studiesmany incomplete- shows leucophoenicitein both descriptions.Here we ly identifiedspecimens remained; in particularthe ques- concentrateon the cation array. tion of orthorhombic polymorphs was not settled (see Electron microscopy also Cook, 1969). Three leucophoenicitespecimens were examined;de- Structural considerations tails are given in Table 1. For examinationin the trans- Moore (1970)described leucophoenicite as a hexago- mission electron microscope, fragments were ground nally-close-packed(h.c.p.) anion array with the larger under ethanol in an agate mortar and then mounted on cations (mainly Mn) in octahedral interstices and the holey-carbonfilms supportedon coppergrids. They were smaller(Si) in tetrahedralinterstices. We havepreviously studiedat 100kV in a rr,ol l00CX microscopefitted wirh proposedan alternativedescription in which anions are a high-resolution,side-entry stage or at 200kV in a rrol insertedinto the intersticesof a cation array, which is 200CX microscope with an ultra-high-resolution,top- rather more regular than the anion array and corresponds entry stage. Both instruments had standard hair-pin filaments.Correction of objective-lensastigmatism was car- rPresent ried out on the carbon support film by Address:School of Science,Criffith University, Bris- minimizing contrast.High-resolution bane,Queensland 4lll, Australia imageswere recordedat magni 1009 1010 WHITE AND HYDE: LEUCOPHOENICITE

Table l: Leucophoenicitespecimens used in the study

Plefixes for Musem Nutrels:

BM British Musem of Natural HisEory sI Smithsonian Institute (Natural History) Su Swedish MuseN of Natural History

Sample and Source constituents Found Observatlons

Almost all crystals showed a considerable FrankIj.n Mine, sussex county, (1,23) feucophoenicite, ^ (221 N.J-. U.S.A. r. rilFr6 t-^hr^ire degree of fau1tin9. APproximately half (i.e- (BM 1946, 185) and sonolite (2r,3). were twinnedomicroscoPicafty twin bands < I00 A wide) .

fragments Franklin Mine, sussex county, (1, 23) leucophoenicite Most crys!a1s were "Perfcct". fuo N.J., U.S.A. were heavily faulted, and one fragnent (sr c6800) exhibited twinning.

perfect; Pajsberg, sweden (.?, 23) leucophoenicite, fi\e (1,23) leucophoenicite was nainly (sM 740214) + a little manqanhwite (22,3) two crystals wele twinned microscoPically. wj.th another + some orthochrysotsile The orthochrysotale was intergrown phase, possibly ta1c.

fications of 380,000x, 550,000x or 810,000x. Linear Image interpretation imaging conditions (i.e., crystals less than 804 thick) were alwaysutilized. Imagesfrom the l00CX instrumentconsisted of (001) Cleavagenormal to [00] (the principal zone of interest) fringes. The 200CX yielded higher-resolution lattice im- was well defined,and large thin crystalswere abundant. agesand structure images which could be matched with Beam damage was severe and rapid, particularly for computer-simulated images. The program used for the heavilv-faultedcrvstals. image calculations was purchased from Arizona State

o o \/\ 'j ir o, oo 'i a o a ,t a a !i.u

o aa

toa

si O

Mn (o,5O)O,O

oo Fig. L Moore's (1979)stnicture of leucophoeniciteprojected on (100).a. Structurein conventionalterms: h.c.p. anions with cations in octahedral and tetrahedral sites. Also emphasizedare sheetsof twinned c.c.p. cations, the repeat unit consisting of one band one atom wide (1) followed by three bandstwo atoms wide (2), which leadsto the twin formula (1, 23).b. Structure emphasizing the alloy-type structure of the cation array. (Note that this is more regular than the anion array above). Rows of edge-connected SiMn6 trigonal prisms are of the Ni2In type; the pairs of face-sharing trigonal prisms are of the CrB type. In the proposed leucophoeniciteseries various widths of the Ni2In type are inserted betweenCrB-like elements.(The olivine structure is anion-stufed Ni2In, with no CrBlike layers.) WHITE AND HYDE: LEUCOPHOENICITE l0lI

and experimentalimages (Figs. 5 and 6) supports the structure proposedby Moore, at least over the thick- nessesstudied.

Lamellar faults

As expectedfrom Moore's earlierX-ray studiesfaults parallel to (001) were found, especially in one .of the samplesfrom Franklin, N.J. (BM 1946,185). (The other sampleswere less imperfect.) A low-resolutionlattice imageof an inhomogeneouscrystal is shownin Figure7. High-resolution lattice imagessupport the notion that the variations are very thin lamellae (one or two unit cells wide) of highermembers of a leucophoeniciteseries, (1, 2^)with x > 3 (Figs. 5 and 8). No lower members(x < 3) were observed.

tG ttati3rkol 4l dirtriburkrn , of $i Fig. 2. orderingof theff:XTit 23)leucophoenicite can producemany polymorphs. The four simplest, designated A, B, C, andD, areshown projected down [100]. (Si-free Mnu trigonal prismsare outlined by dashes.)Note that the (1,23) cation array is maintained.Manganhumite (22,3)2 is alsoshown.

University and adapted to run on the A.N.U. Univac computerby Dr. J. R. Sellarand Dr. R. A. Eggleton.The following instrumental parameterswere employed for the imagecalculations: incident beam convergence: 0.2 or 0.4 mrad, objective aperture radius 0.39 or 0.79A-r, coefficientof sphericalaberration, C" : L5 mm, depthof gaussianfocus = 150A,slice thickness = 44.

Results

The distribution of Si atoms Moore's (1970)determination of the leucophoenicite(1, ,tt*t'*pq,* 23)structure showed two crystallographicailydistinct Si trltl atoms.The first SiOatetrahedron (in our descriptionthe * ,#'* S # i{{{{{t alta edge-sharingMn6Si trigonal prisms) is isolatedand the Si lr*tr {{{{{{{ .-'S sitefully occupied.The secondSi is statisticallydistribut- le 4 , I *,**.{{ I ed between two tetrahedra of an edge-sharedpair , i.e ., in an [(SiOa)(OH)z]group (or alternativelya pair face- of I sharedMn6 trigonal prisms). I If the latter Si arrangement is ordered a number of polymorphsare possible:four simple ones emphasizing i{t:{ff{tt{fi the trigonal prism array are shown in Figure 2. Structure r ff{lffffft imageswere calculatedfor both [100] and [010] projec- Fig. 3. Calculatedthrough-focus series of imagesfor the tions of Moore's statisticalmodel and polymorphsA [100] and zoneusing the statisticalmodel of leucophoenicite,and alsothe B (Figs. 3 and 4). (The calculatedimages for polymorphs ordered polymorphs A and B. (Computationalparameters were: C and D are readily deducedby consideringthose of the thickness: 404, objective aperture radius = 0.39A-r, C, = other ordered polymorphs, particularly around the opti- 1.5A, beam divergence: 0.2 mrad, chromatic abberation: mum defocusof -6004.) A comparisonof the calculated 150A,and defocusAf = -300 to -700A). t0t2 WHITE AND HYDE: LEUCOPHOENICITE

trfi r'.-.r'SrdiirlIG{|: Microtwinning :r.,:.., : :.disttibiiioa Potlllph Perymprph .''::,.:., ol.5i.r'-. s The electron diffraction pattern of a twinned crystal containsreflections from both twins. For leucophoenicite crystalstwinned on (001)the two sets of reflectionswill (almost with h even. -3S superimpose exactly)in lft10]zones For t/ri0l zones with h : 2n + I the reflectionsin ft/

Fig. 4. Simulatedcomputer images for the [010]zones for the statisticalmodel of (1, 23)leucophoenicite and also for its two simplest ordered polymorphs. (Computational parameters are the sameas for Fig. 3.)

Partly-ordered,long-range sequences were sometimes observed.An exampleis givenin Figure9 where(mostly) (1,25)elements are dispersedin a (1,23)matrix. As well as simpleintergrowths, more complexfaults [not parallelto (001)lwere also observed(Fig. 10).Higher-magnification imagesof the latter showthat in someinstances the lattice fringes either side of the fault appear to have reversed contrast(Fig. llA) suggestinga twin relation acrossthe fault surface(c/. below), while in other cases(Fie. llB) interpretationis impossiblebecause of changesin contrast which may be due to strainfields. The lattice image at either end of the fault then becomesvery complex indeed. (See also Fig. 15a.) It seemslikely that these Fig. 5. A high-resolutionstructural image ofleucophoenicite jogs faults are connectingterminating faults of the com- (BM 1946,185). The calculatedimage most closelymatching the mon type [(1,2*) with anomalousx] on widely spaced micrograph is that obtained from Moore's statistical model (c/. (001)planes, but this hypothesishas not beenadequately Fig. 3). Note the intergrowth with higher members of the checked. leucophoeniciteseries, (1, 24)and (1, 26). F F *: * Fig. 6. A micrographof (1, 23) leucophoenicitein rhe [010] * zone. (BM 1946, 185).The most closely matchingcalculated image is obtained from Moore's statisticalmodel for fl. 23) * leucophoenicitetcf. Fig. 41. p s

,tl.l,U,l,lrl

Fig. 7. An inhomogeneousregion of a predominantly(1, 23)leucophoenicite crystal (BM 1946,185). A. Low magnificationimage showingextensive disorder. B. Correspondingselected-area diffraction pattern: note the weak additionalspots indicative oftwinned leucophoenicite.C, D. Higher-magnificationimages interpreted as membersof the proposedleucophoenicite-type series (1,2*). 1014 WHITE AND HYDE: LEUCOPHOENICITE

Fig. 8. A higher-resolution lattice image of leucophoenicite (BM 1946,185). A. Selectedarea diffraction pattern. B. Coherent Fig. 10. A. A twinned (1, 23) leucophoenicitecrystal (BM intergrowth between membersof a "leucophoeniciteseries." 1946, 185); note the light and dark bands indicative of the two Imageswere calculatedonly for (1, 23);however it is likely that twin orientations. The rectangles outline areas with faults not darker bands are Ni2In (olivine) lamellae and lighter bands are parallelto (001).B. CorrespondingSAD pattern.The apparent CrB lamellae. QzL)-t arisesfrom twinning on (001).

dJJJ ;i: " 5" ..1 . Fig. 9. A. Selected-areadiffraction pattern. B. Correspondinghigh-resolution lattice image showing periodic insertion of a (1, 25) memberin a marrix of (1, 23)(BM 1946,185). (Notice that the numberof faults is quite smallif the parentstructure is takento be Il, 2s,(1,23)71,a complex superstructure with d(001): 1854.) WHITE AND HYDE: LEUCOPHOENICITE

Fig. I l. Enlargementsof areasA andB in Figure10. A. Note the reversalof contrastin the main lattice fringesacross the diagonalfault, suggesting twinning. B. Changesin contrastin the regionabove the diagonalfault makesinterpretation difficult. Twinningdislocations (lateral shifts in unifcell twin planes)are apparentin bothcases; and perhaps also an edgedislocation in B.

Of courseit is not surprisingthat twinnedleucophoeni- Fig. 13. A leucophoenicitecrystal (repeatedly)twinned on cite crystals occur. Twin boundaries are readily intro- (001)(BM 1946,185). A. SAD pattern.The reflectionsbelonging (1,23) duced in leucophoeniciteby intergrowth with an to one of the two twin orientationsare distinguishedby arrows. odd numberof lamellaeof any orthorhombicmember of Note that the "doubled" rows of spotsare not quite straight.B. the proposedseries (1,2^) with x even. An example of Low-resolution image showing the alternating dark and light sucha boundaryis shown schematicallyin Figure 14. bandswhich were characteristicof twinned crystals.

: rt ..I ll tr- IT ao--.1.t"@ \i, ',i,' G-{-{--91-P=+€ \i,' \ll 'l'

,,i" --.---J.#* 4' ili I il triol # I

[2To]

Fig. 12. Schematicillustration of the Url0l zoneaxis diffractionpatterns expected for leucophoenicitecrystals twinned on (001): zoneaxes [l l0] and [210].The upperdiagrams are for untwinnedcrystals, open and closed circles for twins I and II respectively.The lower diagramsare for twinnedcrystals, half-fitled circles being superimposed reflections from both twins. Note the doublingof the numberof reflectionsin alternaterows of the [l l0] zone. l0l6 WHITE AND HYDE: LEUCOPHOENICITE

, ,l , , , , , , 1 ) 1 'I ' ' 'l'l , , ? ? ? 2 I'l I I I t I t I l,r L L r'l I I t I lit I r lrl | ?,------JL- I 2r -Jt- 1 2,------JL- 1 2'------J

Fig. 14. Schematic[100] projection oftrigonal prism sheetsshowing the insertionofa(1,24) bandand a(1,2s) bandinto a matrix of (1, 23).Note that the former resultsin two regionsof (l , 23)leucophoenicite related by a mirror plane(twin boundary)while the latter doesnot. (Unit-celltwin planesare indicatedby arrow heads;the "microscopic" twin planeby a heavy arrow.)

The decomposition of leucophoenicite to tephroite beam,the decompositionwas acceleratedand tephroite, (2)2,was produced.The reactionis presumably The major difficulty experiencedin this study of leucophoenicite was decompositionof the specimenin the MnzlsiOelz[(SiO4XOH)2]--+ 3Mn2SiO4+ Mn(OH)2. electronbeam. Under normal operatingconditions, particularly for heavily faulted and/or micro-twinnedcrys- Under the experimentalconditions Mn(OH)z would prob- tals, damageresulted in the rapid and severediminution ably further decomposeto give MnO and H2O, but only of imagecontrast. Defocussing the beam allowedlonger tephroite,MnzSiOa, could be detected. working times, but also necessitatedlonger (>20 second) An exampleof a twinned leucophoenicitecrystal and photographicexposures, making high-resolutionmicros- the tephroitedecomposition product is shown in Figures copy impossible. Slightly longer working times were 15aand b respectively;the micrographsare from precise- availablewhen the specimenswere left in the microscope ly the samearea. The tephroiteconsists of smalldomains, under vacuum overnight. This suggeststhat adsorbed eachapparently perfect in structure,but slightly misori- waterfacilitates the reaction. entedwith respectto adjacentdomains. A higher-magnifi- By removingthe condenseraperture and focussingthe cation image of this crystal (Fig. 16) suggeststhat the

Fig. 15. A. A twinned(1, 23)leucophoenicite crystal in the tl00l projectiondirection (BM 1946,185). B. The samecrystal after it has beentransformed to tephroiteunder the electronbeam. Note the slightly mismatcheddomains. WHITE AND HYDE: LEUCOPHOENICITE l0t7

domains are tilted with respect to the electron beam and/or that there is a variation in thicknessacross the transformed crystal. Figure 17 shows an example of an incompletely transformed leucophoenicitecrystal. In addition to the fea- tures already mentioned it contains a few lamellar faults which appearto be similar to those encounteredin the alleghanyite(Mn-humite) series (White and Hyde, 1982b) i.e., sellaite,Mn(OH)z layers (in Mn-norbergitelamellae).

X-ray powder dffiaction All three leucophoenicite specimenswere poorly crys- talline; they gave powder patterns with very broad reflections. Moore's structural refinement of a Franklin specimen yielded the unit cell dimensions a = 4.826(6), b = 10.842(19),c : 11.324(\4, : 103.93'(9).A typical " Guinier camera powder pattern for our Franklin leucophoenicite(SI C6800)is given in Table 2. The refinedunit cell parametersare a = 4.857(5),b : 10.82(2),c = I 1.36(l)A, a : 103.85"(12).Electron-probe microanalysis (epue) gave a cation ratio of Mn : Ca : Mg : 0.88 : 0.09 : 0.03 (+ trace Zn). The observedlines are in Fig. 16. A higher-magnificationimage of a domain boundary good agreementwith previously publishedpowder data in Figure l5b (indicatedby arrows).Also note the slightbending of the zig-zagrows of white dots on the left-hand side of the micrograph.

Table 2: Guinier powder data for leucophoenicitefrom Franklin, N.J. (SI C6800)

d(obs) q(calc) hk1 r/7^

4 -403 4.408 110 I5 3.959 1r1 l4 3.623 3.626 r12 22 3.288 3.286 lr2 18 2.969 2.975 113 T4 2-Aa4 2 .449 13r 69b 2 -158 o04 2.155 | I 34 2 -752| t3zl

2 - 7251 024l 40 2-7221 o23l

2.695 2.694 1r3 100 2.63r 2.630 131 2 .499 2.494 74

2.451 2-452 114 3l 2.430 2.429 200 20 2- 3981 I04l 2.342 23 2 -3771 15al

2. 366 2 .366 | 20 2.363t iilt 2 - 2Ial 2.2II 272r 2 -207t oosI 1.817 036 1.815 | 1 60b r-8f3t zlal r.8rrI 111t 1-807 1.807 I 1- 799'

I-577 | 16I L.51'7 I Fig. 17. A leucophoenicitecrystal which has been r-srs I oo7 t incompletely transformed to tephroite. The lamellar faults are r.557 065 probably of the sort found in the alleghanyiteminerals, i.e., Mn(OH), boundaries(BM 1946.185). b bload lines l0l8 WHITE AND HYDE: LEUCOPHOENICITE

Table 3: Guinier powder data for leucophoenicitefrom exclusivelyas chrysotile(which containedno Mn), whose Pajsberg,Sweden (SM 74024) characteristicfilaments were intimately associatedwith platy leucophoenicite(Figs. 18 and l9). Although Mg d(obs) d(calc) madeup only a very smallproportion of the bulk analysis, the extreme morphology of chrysotile (fragmentswere a

4.4r 4,38 rt0 26 few hundred to 10004,in diameter) resulted in very many 4.344 29 chrysotile crystallites. (This effect was exaggeratedin the 3.950 1.94 r11 32 EM work since the serpentine filaments remained in 3.896 52 suspensionlonger than the larger leucophoenicitecrys- 23 tals, and were therefore selectivelydeposited upon the 3.6341 l0 21 26 l - etzl l12 t holey carbon films.) Attempts to correlate all the extra 35 reflections on the powder patterns to known chrysotile 3.423 unit cells were unsuccessful.A comparisonof our elec- 3,284 3. 28r 1t2 56i tron diffraction patterns with those of Zussman et al. 3. lo2 100 (1957) suggestedthat crystals were orthochrysotile or 2.940 2 .969 1r3 mixed ortho- and clino-polymorphs. 2.88r 2.881 l3r 8l vb 2 .83I 42

2.166y 004 | 2.755 29 structural 2 -745' tlzl A leucophoenicitemanganhumite

2.73I 0231 relationshiP 2.729 I 74 b 2 -1241 olnl 't4 2.691 2.694 1r3 b In Figure 2 the trigonal prism array of leucophoenicite 2 .628 13L 26 is compared to that of manganhumite; the similarity 2.453 2.45I tr4 35 betweenthe structures is obvious. Figure 20 shows that if 2.4311 04I 1 2.424 in mangan- 2 - 4291 oizl alternatelamellae, three trigonal prisms wide, : 2.374 r24 35 humite are rotated 180'about an axis at x xll4 along 2.359 141 the trigonalprism edgesparallel to c*, the (1,22'1,24)array 2.724 2.323 211 29 2.2L3 oo5, | --^l tt!2 94' 2 .2tL 2.2oe I I 2.204', 2t2'

2. l18 2.t22 2t2 8rr 2.106 65

2 .O57 2.O59 r,856 23 1.8171 2O4l 1. 814 65 vb t- gl4l tla I

b broad lines rs@e intensity may be due to phases olher thd leucophoenicite

(Cook, 1969;Moore, 1970),although the relativeintensi- ties vary considerably between the three patterns. The powder data for the Pajsberg(Sweden) specimen (SM 740214)are listed in Table 3. The refined unit cell parametersfrom the (1,23)leucophoenicite reflections are a = 4.819(5),b : 10.82(l),c = 11.39(l)A,a : 103.67"(22);EpMA gave a cation ratio of Mn : Ca : Mg : Ti : Al : 0.85 : 0.09 : 0.05 : 0.004 : 0.002.In agreement with Moore (1970)several additonal reflections were also present and, like Moore, we could attribute these to an orthorhombic unit cell with parameterssimilar to those of manganlumite (a : 4.870(4),b : 10.794(8),c : Fig. 18. An electron diffraction pattern and low-magnification 22.61(l)li; but selectedarea electrondiffraction did not imageofchrysotile (SM 740214).The difraction pattern indexing revealthe presenceof this phase. is for orthochrysotileaccording to Zussmanet al. (1957),but A major differencebetween the Swedishspecimen and additional spots indicate that clino-chrysotileis also present. those from Franklin was the presence of chrysotile, Note that the serpentinecrystal is not thick enough to obscure MgrSizOs(OH)rin the former. Analyticalelectron micros- the underlying carbon film. Also present is a large, platy copy showed that the Mg was accommodated almost leucophoenicitecrystal. WHITE AND HYDE: LEUCOPHOENICITE 1019

orthorhombic reflections in X-ray powder diffraction patterns could not be explained. Our work is similarly inconclusive;both HREM and SAD failed to elucidate the origin of the apparent orthorhombic symmetry. Moore (1970)suggested that the extra reflectionscould be due to polysynthetic twinning on a unit cell scale but althoughwe find frequenttwinning (on a scaleof lessthan l00A) it is not periodic. Superlatticeswere not detected, although partly ordered intergrowths were observed (Figs. 7, 9 and 2l). (In the related humites regular intergrowth sometimes occurs, yielding long c-axes (Whiteand Hyde, l982a,b).) Moore (1967)also reported that Weissenbergphoto- graphs for some leucophoenicite specimens showed streakingparallel to c*. This he attributed to "semi- random structure." Some of our electron diffraction patternswere similarly streaked,and this was seento be due to coherentintergrowth of various leucophoenicite- like structures(1,2*, x > 3)-a phenomenonobserved in all the specimensused in this study. (It is interestingthat (1,2*)structures with x ( 3 were never seen.) Complex faulting occurs extensivelyin one specimen

Fig. 19. A higher-resolution image of an orthochrysotile filament.The -7.3A fringescorrespond to d(002);the -4A t central fringes are probably d(020).(The relevant unit cell parametersarei a:5.322, b:9.219, c: 14.534for orthochrysotile.)

is obtained.A translationof alternate(1,22,1,24) blocks by al2 thenyields (1,23)leucophoenicite. Hence there is a MAfVGANHUMITE(2','" simple, direct "unit-cell twinning" relation betweenthe two structures (in addition to the crystallographic shear relationshippreviously noted (White and Hyde, 1983)). Humite-likeboundaries [i.e., twinned c.c.p. (3) bands] were observedin leucophoenicite,but only rarely. An exampleis given in Figures2l and22.The Re3Bportion of the manganhumiteregion is a singlethin lamella(white in the image,due ro its lower chargedensity). All other white lamellae occur as consecutivepairs which are characteristicof the statisticaldistribution of Si over the face-sharingtrigonal prismatic (edge-sharingtetrahedral) sitesthrough the thicknessofthe crystal.The result is an averagecharge density over theseinterstices which is half that of the adjacent olivine portions, whose trigonal prismsare fully occupiedby Si. The boundaryis shown schematicallyin Figure 22b. Note that neither the man- LEUCOPHOENICITE(1,2) ganhumitepoftion nor the (1,24)lamellae (because these Fig. 20. A schematic drawing of the way in which alwaysoccur in pairs) introducea microtwin into the (1, manganhumitemay be structurally related to leucophoenicite.A. 23) leucophoenicite.This is borne out by the electron The manganhumitestructure, (23,3)2 projected on (100).B. The diffractionpattern which showsno doublingof the num- structure obtained by rotating alternate lamellae of trigonal ber of spotsin the (13/)row. prisms of width d(002) in manganhumiteabout an axis at x : + l/4 hasa unit cell equivalentto that of manganhumite.The new Conclusion trigonal prisms produced by the operation are left unshaded.The twinformula is (1,22,1 ,24). C. Theleucophoenicite structure (1 , Previousstudies of leucophoeniciteyielded interesting 2r)is formedby translatingalternate (1,22,1,24) blocks through though perplexing results. In particular, the additional al2. 1020 WHITE AND HYDE: LEUCOPHOENICITE

.uluwlJs$,uliu,|J,w$,uru$Jjru$J.wlujJjjjj,lffilllw#{[tLl,r ' r?snganhumlfe(n=3)t2:3 '. ii Fig. 21. A disorderedleucophoenicite crystal (BM 1946,185). A. SAD pattern.B. Low magnificationmicrograph showing the extentof disorderin the crystal. Variationsin latticespacing have beeninterpreted as mainly (1, 2*) members;but a thin lamellaof manganhumiteis also present.

(BM 1946,185; Figs. l0 and l5a). Similarfaultsobserved tetrahedra in edge-sharingpairs (i.e., in face-sharing in the humite series(White and Hyde, 1982b)were rather trigonalprisms of cations)was a little surprising;at least rare, not nearly as common as in leucophoenicite.Mi- someshort-range order was expected.Naturally, in terms crotwinningwas also lessfrequent in the humitesthan in of the numberof unit cells tle specimensexamined were leucophoenicite. quite thick: 8.5 x a : aOA and 4 x b : 40A in the The apparentlystatistical distribution of Si betweenOa micrographsshown in Figures5 and 6 respectively.The possibilityof orderingin the [00] direction was consid- ered,but rejectedbecause it would result in a doubleda- axis for which there was no evidencein any X-ray or electrondiffraction pattern. Furthermore,comparison of observed and computer-simulatedimages in the [010] zone, which would reveal ordering along a, instead showed good agreementwith the "statistical model."

I ttz\ | Q'zl I i e"3) | L2") | A B

Fig. 22. A. An enlargementof the manganhumiteregion in Figure21. Imageswere calculatedfor manganhumiteand for 11,2r) leucophoenicite.B. A schematicillustration of the boundary. WHITE AND HYDE: LEUCOPHOENICITE t02l

(Nevertheless, the possibility of electron beam effects Hyde, B. G., Andersson,S., Bakker, M., Plug, C. M., and inducing disorder in ordered leucophoenicite samples O'Keeffe, M. (1979) The (twin) composition plane as an should not be discounted.) extendeddefect and structure-buildingentity in crystals.Pro- gressin Solid StateChemistry, 12,273-327. Moore, P. B. (1967)On leucophoenicites:I A note on form Acknowledgments developments.American Mineralogist, 52, 1226-1232. The authorsare gratefulto Mr. P. C. Tandy (British Museum, Moore, P. B. (1970) Edge-sharingsilicate tetrahedrain the NaturalHistory), Mr. PeteJ. Dunn (SmithsonianInstitution) and crystal structureof leucophoenicite.American Mineralogist, Dr. B. Lindqvist (Swedish Museum of Natural History) for 55,1146-1166. providingleucophoenicite specimens. We are also indebtedto Palache,C. (1928)Mineralogical notes on Franklin and Sterling Dr. J. R. Sellarand Dr. R. A. Eggletonfor preparingthe image Hill, New Jersey.American Mineralogist, 13,297-329. calculationprogram, Mr. N. Ware for carrying out electron White, T. J. and Hyde, B. c. (1983)A description of the probeanalyses ofthe samples,Dr. D. Goodchildfor the useofan leucophoenicite family of structures and its relation to the analyticalTEM, and Dr. Joyce Wilkie for the preparationof humitefamily. Acta Crystallographica,B39, l0-17. petrographicthin sections and providing the computer program White, T. J. and Hyde, B. G. (1982a)An electronmicroscope for refiningGuinier powder data. Many usefuldiscussions with study of the humite minerals: I Mg-humites. Physics and Dr. E. Makovicky clarifiedseveral problems in the early stages Chemistryof Minerals,8, 55-63. of this work. T. J. White acknowledgesthe financialassistance White, T. J. and Hyde, B. G. (1982b)An electronmicroscope providedby the CommonwealthDepartment of Educationand study of the humite minerals: II Mn-humites. Physics and theA.N.U. Chemistryof Minerals,8, 167-174. Zussman,J., Brindley, G. W. and Comer, J. J. (1957)Electron difraction studies of serpentineminerals. American Mineral- References ogist,42,133-153. Cook,D. (1969)Leucophoenicite, alleghanyite and sonolitefrom Franklin and Sterling Hill, New Jersey.American Mineral- Manuscript received,May 26, 1982; ogist, 58, 807-824. acceptedfor publication,February 10, 1983.