Titanium, Fluorine, and Hydroxyl in the Hrrmite Minerals

AmericanMineralogist, Volume 64, pages 1027-1035, 1979

Titanium, fluorine, and hydroxyl in the hrrmiteminerals

Paul H. RBns

Department of Geological Sciences Virginia Polytechnic Institute and State University Blacksburg, Virginia 24061

Abstract

In the humite homologousseries, nMrSiO 4' M F,TL,(F,OH)2-2,O2,(where n : I for norbergite, 2 for chondrodite, 3 for humite, and 4 for clinohumite, M : Mg >> Fe > Mn > Ca,Zn,Ni and 0 < x < l), x appearsnever to exceed-0.5 Ti atomsper formula unit, because local electrostaticcharge imbalances at the 3-coordinated(F,OH,O) anion and the 4-coordinated oxygen atoms increasevery rapidly as Tia* substitutesfor trP*, even with the con- comitant substitutionof 3-coordinated02- for (F,OH)'-. Both electrostaticand geometricar- gumentssuggest that Ti ordersinto the M(F,OH)O "layer" of the humite structures,t.e., into the M(3)OIV(F,OH,O)|I octahedron,which is the smallestof all octahedrain all the humite minerals(cl Fujino and Tak€uchi, 1978). Refractive indices, density, unit-cell dimensions, and volume are dramatically affected by the substitution of OH for F, and this has been studied as chemistry varies from 1.0 < (F,oH)'ill[(F,oH)"' + o''] = 0.0and 0.0< silv/2(F,oH,o) < l,/8 from sellaiteMg(F'oH), (n : 0) through the humitesto forsterite(r : o1. The volume per anion increasesmuch more rapidly as OH substitutesfor F in theseminerals than would be expectedon the basisof ob- serveddifferences in individual M-(F,OH) distances.As Yamamoto (1977)noted, the effective radius of OHIII would be -0.06A larger than that of Fttt based on individual bond lengths.However, if the grand mean M-(F,OH,O) distanceminus the effectiveradius of the M cationis used,OHIII would have to be -0.15A larger, and if the efective anion radius calculated from the observedvolume per anion is used,values -0.10A larger are expected.Pro- ton-proton repulsions,which increasedramatically both within a given mineral series(say F- to OH-chondrodite)and as n goesfrom 4 to 0, expandthe unoccupiedpolyhedra for the hcp anion array very much more than the occupiedoctahedra; physical parametersare accord- ingly affected.

Introduction crystal structuresof no less than three titanian clinohumites(Ti-CD and one titanian chondrodite(Ti- The formula for the homologous seriesof minerals Ch) have been examined. Least-squaressite refine- collectively called "humites" is nMrSiOo' Mr-,Ti, ments have produced somewhat conflicting results, (F,OH)r-r,Or, where n : I for norbergite, 2 for leaving us with two questionsregarding titanium: (l) chondrodite, 3 for humite, 4 for clinohumite, M is Why doesx appearnever to exceed0.5 Ti atoms per Mg >> Fe > Mn,Ca,Zn,Ni, and 0 < x < l. A seriesof formula unit in natural specimens?(2) Is Ti really or- Mn-isotypes for n : 2 (alleghanyite), 3 (mangan- deredinto the M3 octahedralsite, and if so,why? Fu- humite), and 4 (sonolite) also exists in nature, and jino and Tak6uchi (1978) implied answersto these these exhibit a substantial range of solid solution questions with vague referencesto Pauling's elec- with Mg; most are OH-rich. trostatic valencerule. One purposeof this paper is to Titanium-rich clinohumites and chondrodite (x > addressthat matter more specifically. 0.25) have recently attracted attention as possible Another crystal-chemicalproblem has to do with mineralogical sites for water in the earth's upper the nature of the OH e F substitutionin all the hu- mantle (McGetchin et al., l97O; Merrill et al., 1972; mites.The recentsynthesis of OH-Ch and OH-CI by Aoki er al.,1976;Mitchell, 1978),and as a resultthe Yamamoto and Akimoto (1974, 1977),the determi- 0nn3-{0/.X/79/09 l0- 1027$02.00 1028 RIBBE: HUMITE MINERALS

nation of the structure of OH-Ch by Yamamoto panciesin Ti-Cl Robinson et al. (1973ab) took dif- (1977),and the synthesisof many F,OH intermediate ferent approachesthan Fujino and Tak6uchi (1978); compounds by Dutry (1977) have provided a data the former report disorderedTi, the latter ordered. basefrom which the systematicsof the OH ? F sub- Kocman and Rucklidge (1973)-perhaps most stitution has beendeduced. wisely-simply assignedTi to M3 and refined Mgl Fe distributions within the constraintsof their chem- Titanium in natural humites ical analysis. Fujino and Tak6uchi's (1978) referenceto Paul- Titanium content ing's electrostaticvalence principle holds the key to Table I lists, anong others, the compositions of both the restrictedTi substitutionand Ti distribution the most Ti-rich humite minerals analyzed to date. among the nonequivalentoctahedral sites in humites. Note that x does not exceed 0.5 Ti atoms per formula Let us consider the hypothetical homologues, unit in any of the humites. Norbergite (No; n: l) is nMg,SiOo'Mg,-"Ti,(F,OH)2_2.O,.with 0 S x S l. rare in nature and apparently is relatively Ti-free. Note that one half of the octahedral sites are filled Ctnohumite (Cl; n :4) commonly is Ti-rich (Jones with cations in the proportion (x)Ti + (2n + l - et al.,1969, Fig. 4), but only one titanian chondrodite x)Mel. (Ch; n: 2) has ever been reported (Aoki er al., 1976), Charge imbalance with Ti disordered.With Ti dis- and it is epitaxially associated with a Ti-Cl which ordered there are only two types of anions as far as contains the same number of Ti atoms per formula local chargebalance is concerned,regardless of min- unit as Ti-Ch. Notably absent from this intergrowth eral species.The averagecharge on all M cations is is the intermediate phase, humite (Hu; n : 3). the same:(zl : alx/(2n + l)l + 2l(2n+ | - x)/(2n + l)1. The oxygensare 4-coordinatedto lSi + 3M and Titanium distribution are all overbondedby [ + 3(zl/61- 2.00esu. The The structures of both Ti-Ch and Ti-Cl have been (F,OH,O) anion is bonded to 3M atoms and is al- determined by Fujino and Tak6uchi (1978) who ways underbondedby an amount 3(zl/6 - (l + x) found by least-squares site refinement that most, if esu, where x : Or,,/(F + OH + Or,). Figure la in- not all, of the Ti is ordered in the M(OH)O layers, dicatesthe imbalancesas a function of x for each of i.e., in the M3 octahedral sites. Of course there are the four humitesin which Ti is disordered. problems associated with determining unambi- Charge imbalance with Ti ordered. If all the tita- guously the cation distribution among the octahedral nium is ordered into M3, there are collectively sites when three species of atom are present, namely among the humite homologuesthree types of 4- : -26) Me 0 l2), Fe * minor Mn,Ni (f and Ti 6f : coordinatedoxygens, Oo, O,, Or, the subscriptsrefer- 22) (J is the atomic scattering factor at snfl/)r: 0.0). ring to the number of M3 atoms to which they are Some basically unsatisfying assumptions must be bonded, plus one (F,OH,O) anion. From the multi- made to obtain convergence: in refining site occu- plicities of M sitesin the humites (Table 2) it can be

Table l. Compositional data for humite minerals whose structures have been refined

z Mineral (space group) r Ms Fe Mn Nt ca sio4 ri F oH* ori ,#.trTtl]r,

I NorbergiEe (Pbm) xx O.42 (36.2) 0.oil 1.99 <.01 <.01 <.01 1 0.01 1.81 0,17 0.02 Jones et aL. (L969)

2 chond,rodite (p21'/b) 9.36 (22.L) ' o.424 3.99 0.57 0.0r 0.01 o.42 0 1.15 0.85 Fujino & Tak6uchi (1978) o.o3 (22.1) <.0r 4.95 0.03 0.01 0.0r 2 <,oL L.27 0.73 <.0r cibbs e, aL. (I91O)

3 Homite (Pbm) 3.25 (15.9) 0. 201 6,49 0.22 0.09 <.01 3 0.20 0.79 0.81 0.40 Joles et aL. (1969)*** 0.10 (15.9) 0.01 6.60 0.35 0.05 <,01 3 0.01 1.06 0.93 0.01 Ribbe & Gibbs (1971)

4 Cllno}rtnite (P2r,/b) 5.59 (L2.4) 0.468 7.33 1.04 0.05 <.01 4 o.47 0 1.06 0.94 Robinson et aL. (1973) " 3.\4 (r2.4) 0.26 7.34 1.36 0.08 4 o,26 0.40 1.08 0.52 KocDao & Rucklidge (1973) 5.O1 (r2.4) o.429 7.44 r.09 0.02 0.02 4 o.43 0 1.14 0.86 Fujino 6 Tak6uchi (1978) o.22 (r2.4) o.o2 8.42 0,s0 0,06 < .01 0.02 1 . 04 0. 93 0. 03 Robinson et aL. (L973)

* Calculated by difference: OH = 2 - 1f + Orr), where OT. = O2x. ** Space group convention of Jones (1969), ***Most Ti-rich hunite analysis reported; the structure of this specinea has not been refined. RIBBE: HUMITE MINERALS

Ti DISORDEREDIN M Ti ORDEREDINTO M3 dered humites; the proportions of O0, O,, O, and ) o5 (F,OH,O) anionsper formula unit are listed in Table o o2 2 for eachof the structuretypes. C .9 It is clear from Figure I that the more Ti, the c o o vo greater the local charge imbalances, regardlessof o 0 whether Ti is ordered or disordered in any of the o c 0 homologues.Quite likely these imbalancesplace a o o - practical upper limit on Ti substitution. Comparing E o.5 o the two graphs,it is also evident that chargebalance o o is somewhatimproved when Ti is ordered into M3, E o -t.o exceptfor norbergite in which all anions are bonded o.o o.5 r.o r.o o.o o.5 Thus may argue that Ti prefers x = Ti per formulounil x = Ti per formulounil to two M3 sites. one x: frociionot O in (F,OH,O) r = froclionof O in (F.OH,O) M3 for reasonsof chargebalance, but it should also be observedthat M3 is alwayssmaller than the other Fig. l. Plots of .x yr. charge imbalanc€ at the anion for the humite minerals assuming (a) Ti disordered in all the M octahedra octahedra(even in the Mn-isotypes),although it has and (b) Ti ordered \nto M3. Symbols are explained in the text and fewer sharededges than Ml (seeTable 2). This is in Table 2. The charge imbalance was calculated in the following part explained by the fact that M3 is bonded to two manner for the (F,OH,O) site: S - (l + x), where S is the sum of monovalent (F,OH) atoms when no Ti is present. the Pauling bond strengths to the anion and x is the fraction of O Ti (r - 0.61,4.)is smaller than Mg'* (r in that site. For the oxygens, charge imbalance is simply S - 2. Quadrivalent - O.'72A)and the other W* cations found in humites, and quite likely prefers t}lreM3 site for that seenthat if all the Ti is orderedinto the M3 sites.the reasonas well. maximum Ti occupancyof M3 is x/2.,becausex/(2n Thus the crystal-chemicalarguments all point to + l) is the fraction of Ti in the (2n + l) M sitesand an ordered distribution of Tio* in M3, the somewhat 2/(2n + l) is the fraction of M sites which are M3 questionable least-squaressite refinements (Robin- sites. The Oo oxygens are charge-balancedbecause son et al., l973a,b) notwithstanding. Probably the they are not bonded to M3, the O, oxygensare over- most compelling argument for Tio* order is that bondedby x/6, the Or's by 2x/6, and the (F,OH,O) madeby Fujino and Tak6uchi(1978, p. 537-539).In anions are always underbondedby 4x/6. Figure lb addition to least-squaressite refinementsof Ti-Ch showsthe imbalancesfor any of the hypothetical or- and Ti-Cl, which both converged with relatively

Table 2. Fraction of tetrahedral sites filled, number and types of octahedral sites (per formula unit), and number and types of oxygen atoms and (F,OH,O) sites (per unit cell) for the humite minerals and forsterite. In italics is listed the number of octahedral edges shared with other octahedra (o) and with tetrahedra (t).

Number of 4-coord, n z tqinera! (symbol) Number of octahedral sites of tyDe:* of tvoe: Number of :l:::r:1-::CeE-L,saces 01 /F nH n)III oo .oz

L4 Norbergite (No) IlLz t2 tb 6

22 Chondrodite (Ch) 1/10 22 4844

34 Humite (Hu) 3/28 2 t22 30L268

4 2 Clinohunlte (C1) I/9 2l 222 20844

@ 4 Forsterite (Fo) L/8 1 I 16 0

Shared octrl edges: 4o,2t 4o,2t 20,It 20,It 20,Lt 30,It

* Subscripts z and c in the M(L)O, octahedra signify 1 and 1 site symetry, respectively. Subscripts 6, 5 and 4 refer to the number of ligands to t}:.e M2.octahedra which are 4-coordinated oxygens, e,9., M25= uofvta,ou,ollrr. 1030 RIBBE: HUMITE MINERALS

large estimatedstandard errors (+0.08 Ti) to indicate ation from Sel + No -+ Ch -+ Hu --+ Cl --+ Fo is all Ti in M3, they found that in theseepitaxially in- more complex than just (F,OH) for O. The charge- tergrown specimensTi-Ch had a TilSi ratio of 0.212, balanced substitution is [J" + 4(F,OH)III --+ Si'' * atmostexactly twice as great as that in Ti-Cl (0.107). 4O'', where D" is a tetrahedralvoid. The fraction of This is precisely the relationship expectedif Ti is tetrahedral sites occupiedby Si in the hcp array de- preferentially ordered into M3 in both structures:the creasesfrom l/8 in Fo to l/l2in No to zero in Sel M3/Si ratio is 2 in chondrodite,I in clinohumite. (seeFig. 2 andTable2). Epitaxial intergrowthssuch as the one reported by For the sake of comparing these structurally re- Tilley (1951),but more particularly those containing lated compounds;Ribbe et al. (1968) suggestedthat substantialTi, should be carefully analyzedin the fu- it would be convenient to normalize unit-cell vol- ture for additional confirmation of theseconclusions umesto a reducedcell of volume V : labcsinaf+ q, regardingthe ordering of Ti. where 4 is 4 for Fo, 6 for No, 5 for Ch, 14 for Hu, 9 for Cl. The parameterq waschosen to give a third di- Fluorine and hydroxyl mensiondslj + q - 1.5A,1.e., approximately equal Van Valkenburg (1961)synthesized F-end mem- to the radius of an averageanion in the hcp array. bers of the humite seriesfrom melts and by solid- Another perhapsmore physically useful volume term statereactions, but OH-end membershave been syn- is Vo: lt /4, the volume per anion, from which may thesizedonly at high temperaturesand pressures- OH-Ch and OH-CI were found to be stablebetween Froctionof tetrohedrolsites filled 700-1000"C and 29-77 kbar (Yamamoto and Aki- o.0 0 025 0 050 0 075 0.roo o t25 moto, 1974, 1977).Duffy (1977) synthesizedF_No, No ChHu Cl F-Ch and F-Cl and many inrermediate(F,OH)-hu- o mites, sellaite (MgFr), and "intermediate sellaite," MgFOH,' as part of his investigation of the system MgO-MgFr-SiOr-HrO. He also calculated least- squaresexpressions for molar volumes as a fgnction of compositionfor the phasesof variable F-OH con- c | 380 tent in this system,and found that with the exception c c, (r6)y predictedfor / ... of brucite "all the volume functions are linear in - .,' OH selloite ,, y'.!:.' | 360 L rO composition.For sellaite and humites the difference o -i) f //." in volume T 3 between the hydroxyl- and fluoro-end E q 0 membersis 2.95+0.19 cm, mol-'of (OH)r" (p.6; see sr o r 340 his Table VI, p. 5l). Duffy gives otr also equations for o o variationsof interplanar d-spacingswith F € OH for o) sellaite(Sel), No, Ch, and Cl (his Table V, p. 50). His | 320 ()o ottr'-')x data and some of those listed in Tables I and 3 are ll xx' tl c used in the following more detailed discussionof the L< x' o c relative effectsof F ? OH and other substitutionson o the unit-cell volumesof sellaite,the humites,and for- c) /..t--" --'lzll E sterite. In all these structuresthe octahedralsites in 5 ({ })"1 o the hexagonal closest-packedanion array are half- ll filled with Mg, and to all intents and purposesthe No Ch Hu Ct Fo [SiOo] tetrahedra (where present) are the same size to 08 06 04 02 0.o (mean Si-O - 1.63.!t')and eachshares the samethree (F+oH)E/(F*oH)E * edgeswith M oclahedra.But the compositionalvari- oE Fig. 2. The averageradius of anions (r^) plotted against(F + Oryrtt/[€ + OH)ril + Orv] and the fraction of tetrahedralsites I Tetragonal s€llaite, MgF2, has the hcp rutile structure (Baur, filled for the seriesof hcp compoundsSel + No + Ch + Hu + 1976): : a:4.6213, c:3.0519A, ll a2c. "Inrermediare sellaite." Cl + Fo. A : alleghanyite,T : t€phroite. Dots at the tails of MgFOH, is orthorhombic and is similar in two dimensions and c/ arrows represent (ra)y for natural Mg-equivalent homologues q to the humites and olivine. Its cell parameters (Duffy, 1977) arc (Table 3) correctedfor effectsofM site substituents;arrows point a:4.686, b: 10.123,c:3.O78A; q:2 (seedefnition of 4 be- to open circles which are values of (ro). predicted for OH_end low). membersby simpleproponionation. RIBBE: HUMITE MINEMLS 103I

be calculated a mean radius for the anion, (r^1 , : value determined by Ribbe and Gibbs (1971, p. : 0.51r'1',assuming Vois a cubeof dimensions2 (rol r. I166).In fact, somesimple arithmetic'?using (ro)' 1.370Apredicted for F-Ch and (ro)' : 1.400ob- Efective radii of F and OH servedfor OH-Ch showsthat the effective "radius" In Figure 2 there are three lines representinganion of OH"'must be 0.150Alarger than F"'. radii calculated in diferent ways (see Table 3); all Considernow the third line, \r^1, : 1.316- are plotted againstthe ratio (F + oH)rIrl[(F + oH)IIr 0.048[F/(F+ O)], nearthe bottom of Figure 2. Here * O'"1, which is of courserelated to the fraction of the effectiveanion radius ( ro) nis calculatedfrom Z^, tetrahedral sites filled with Si (cl upper and lower the volumeper anion: \rol r:0.5V1'. This "radius" abscissas).The first is a line labelled "Expected reflectsnot only changesin mean M-O,F,OH bond u, (ro)", whichjoins the 1.3004value for the radiusof lengthsbut alsothe effectsof Si'" substitutingfor E Ft" and the L378A value for the radius of O'" (Shan- which are normally thought to be bypassedin the de- non and Prewitt, 1969),and assumesthat a propor- termination of an anion radius basedstrictly on M' tional mix of F"'and O'' will producethe line (r") : O,F,OH bond-length considerations.Thus the effec- 1.378- 0.078[F/(F + O)]. Of coursethere are no tive "radius" of OHIII is 0.064,0.075 and 0.097A structures to which such a line may apply, but it larger than F"', as calculatedfrom (r^)' values for servesas reference. the pairs F-CllOH-Cl, F-ChlOH-Ch and F- An effective"radius" of Fttt in Sel was calculated SellFOH-Sel, taking into account the proportion of by subtracting r'* : 0.72A from the grand mean anionsin eachstructure which are O''. The dash-dot Mg-F distance(1.988A; Baur, 1976),and an effective line, representingthe substitution n'u + 4oHrrr + "radius" of O'" in Fo was determinedsimilarly using Si'u + 4O'' joins a valueof (ro) ' predictedfor a hy- ((Mg-O)) :2.ll4A (Smythand Hazen, 1973). Since pothetical OH-Sel with that observedfor Fo, passing the F"',/(F"' + O'") ratio decreasesand the fraction of near the observedvalues for OH-Ch and OH-CI and filled tetrahedral sites increaseslinearly from left to having nearly the sameslope (0.048vr. 0.044)as the right in Figure 2, one might expect that the dashed line A-T joining (r^)n for OH-Mn-Ch (alle- straight line, (ro), : 1.394- o.lz3lF/(F + o)1, ghanyite)and (ro) , for Mn-Fo (tephroite).Although joining Sel and Fo would contain points representing the data base is small, this is certainly no coinci- all the F-Mg-humite minerals. Unfortunately no dence. Furthermore the slopes of these lines are structuresof F-Mg-end membershave been refined, equal but opposite in sign to that representingthe -> but if the grand mean M-O,F,OH distances,((M- substitution!'' + 4F'' Si'' + 4OI". O,F,OH)), minus the mean radii of the M cations, Just what value should be assignedto the radius of (r"n), are plotted (dotted tails of arrows) for each of OH"' is somethingof a puzzle.If the effect of the E the refined natural specimens,they all fall above the <+ Si substitutioncould be ignored,we would say the dashedline. This is mainly due to OH"' substituting radius of OH"' would have to be as much larger than - :2 for F"'in the hcp anion array. On that assumption O'' as F"' is smaller(i.e., r.,' rp x 0.078tr'),but (ro)- valuesof (r^), (shownas centeredcircles) have been this is probably not the case,even though the predicted for OH-end members by simple propor- calculations indicated a difference of 2 x 0.075A. tionation of the OH/(F + OH) ratios observed Yamamoto(1917, p. 1483)noted that eventhe mean (Table 3) and the differencebetween observed radii Mg-O distancesin OH-Ch were slightly larger than and thoseexpected for pure F-end rnembers.This as- corresponding ones in the F,OH-Ch refined by sumption is more fully supportedin later discussion, Gibbs et al. (1970),and he said "The replacementof but attention is called here to the fact that the line F by OH causesan expansionof the whole anion ar- joining the (ro), valuesfor Fo and OH-Ch is close ray." A study of the unoccupiedpolyhedra in the hu- to the predicted values for OH-end members (the mites would be most enlightening. Surely proton- point for No is subjectto large errors due to the long proton interactions(see Fujino and Tak6uchi, 1978, extrapolation). Furthermore it becomesobvious that Fig. 3) serveto expandthe structuresof OH-rich hu- the net effectof OH substitutionfor F (r: 1.300,4')is mites considerablymore than is indicated by com- to expand the structure as characterizedby ((M- paring M-O,F,OH bond lengths. Perhaps that ex- rather more than expected O,F,OH)) basedon the 2 Since 807oof the M-O,F,OH bonds in Ch are to OvI, 0.8ro + value of 1.358Asuggested by Yamamoto (1977)for O.2r.: 1.3799 and 0.816 * O.2r6s: 1.400A. Subtracting the two the radius of OH"'. and much more than the 1.344, equationsgives res - re:0.030/0.2:0.15A. ro32 RIBBE: HUMITE MINERALS

Table 3. Numbers and ratios of three-coordinatedanions in syntheticand natural humites,forsterite, sellaitc, and brucite togetherwith normalizedvolumes, Z, correctionsto volume due to transitionmetal content,Att : Ilrul - rylv /raa",and calculateddensities.

Naoe, g, Symbols, No. of 3-coord. anions F+0.5 ot1 (F*oH)III ,,, r^er ^rrr /^j\ calc'd v \4J,, av (4J., reference* Fig. 3 F oH ori G{EFT-r+orv densiry

Norbergite 1.0 O 2.00 1.00 0.33 70.25 3.199 q=6 0.9 a 1.80 0.20 0.90 70.56 0.7 a 1.40 0.60 0.70 7I.L9 VO 2.00 1.00 70,37 10 I .81 0.17 0.02 0.91 70.54 0.00 Chondrodite 1.0 o 2.00 1.00 0.20 7r.23 3.205 q = 5 1.0 o 2.00 1.00 7L,27 0.8 o 1.60 0.40 0. 80 7r.68 0.8 o 1,60 0.40 0.80 7L.67 0.7 o 1.40 0.60 0.70 7t.82 u.o o r.20 0.80 0. 60 72.L4 2 o L,27 0,73 0.00 0. 63 7r.86 -0.03 9 o 1.86 0,11 0.03 0.94 71. 50 -0. l0 24 o 1.06 0.90 0.04 0. 54 72.24 -0.22 (OH-ch) Y o 2.00 0, 00 0.20 73.7 4 3.060 (Ti-ch) F&T a 0.00 1.11 0.89 0.22 72.76 +0.17 Humite V tr 2.00 1.00 0. 14 7r.78 q=r4 3 tr . 0.84 0,80 0.36 0.51 72,40 0.00 7 d 1,06 0.93 0,01 0.53 72.44 -0.22 Clinohumite 1.0 a 2.00 1.00 0. 11 17.83 3.21r q = 9 0.5 A 1.00 1.00 0.50 72,36 0.4 A 0.80 L.20 0.40 72.44 7 l r. 04 o.s2 o.04 0.53 72.52 -0.29 (oH-cl) Y&A 4 2.00 0. 00 0.11 73.02 3.139 (Ti-cl) K&R 0.42 1.05 0.53 0.34 73.08 -0.45 10 l o.oo 1.05 o.9s 0.24 12.98 -0.L2 F&T A o.oo 1.10 o.9o 0.22 72.89 -0.16 Forsterite, MSrSioO (yoder and Sahama, 1957) q = 4 0. 00 72,50 3.222 Sellaite, MgF2 (Duffey, 1977) q = I 1.00 65.18 5.rt4 Mg0HF (Duffey, 1977) q = 2 1.00 73.10 L. T+) Mg(OH)2 hypotheticaL Q = 1 i.00 81,03 2.312 Bruclte, Mg(0H)Z 1.00 47.22 2.05r

x Specimen numbers with decimls were synthesized by Duffey (1977)'. O.9 = nyIgZSiO4.Mg(F,90H.1)2 Those with integral numbers are frorn Joles et aL, (1969), V = Van Valkenburg (1961), Y = iamanoto (1917), F&T = Fujino and Tak6uchi (1978), Y&A = Yamamotoand Akimoto (L977), and K&R = Kocmanand Rucklidge (1973). plains why OH-No, which would have an octahedral cation radius from one structure to the abundanceof both shared and unshared OH...OH next, all the humite mineralsmay be systematizedac- octahedral edges,has not yet been synthesizedand cording to the averageoccupancy of the 3-coordi- may be unstableexcept at very ligh pressures(>150 nated (F,OH,O',) site. Pertinentdata are compiled in kbar?). Fujino and Takduchi (1978, p. 541-542) Table 3 and plotted in Figure 3, where curves have make a strong argument for H...H repulsion as the beendrawn through the points representingsynthetic ..under causeof structuralinstability and suggestthat specimens.Because the substitution of Ti into M3 is a F-free condition, OH-Ch and OH-CI appearto be accompaniedby a charge-balancingsubstitution of 5gfilized through incorporation of Ti in place of Oi.'ifor (F,OH)"', Or, had to be apportionedproperly Mg." Obviously it is the concomitantsubstitution of basedon its effectiveionic radius in order to model O. for OH that "stabilizes" the structure,because Ti the volume relationships established in Figure 3. itself "destabilizes"the structure by increasinglocal Sincethere is a great deal ofuncertainty about anion electrostaticchdrge imbalance (Fig. l). radii in any case, it was guessedthat OII (r : 1.357+O.OIOA;Shannon and Prewitt, 1969)is some- Unit-cellyolumes and ionic radil where nearly half as much larger than F"' (r : Using the normalized volumes Z, correctedby an 1.30010.012,{),as OH"' is (1.36< r < 1.454,de- amount A,V to account for differencesin the mean pending on how it is calculated----seeprevious para- RIBBE: HUMITE MINERALS

Effects of F = OH on physicalproperties iA A Optical properties Misplaced emphasisupon the isolated [SiOn] tet- 740 rahedron as the key structural unit in orthosilicates €A has diverted attention from the fact that most ortho- U silicateshave prismatic habits and cleavageand are for the most part optically positive (Ribbe, 1976). The Mg-humites are optically positive becausein all F of them serratededge-sharing octahedral chains are in fact the dominant structuralunits. tt .e Sahama(1953) calculated mean refractive indices, (r) (, er (R.t.) : (oQy)"' or (to'e)'/',for the F- and OH-mem- tt o bers of the humites basedon molar refractivity, plot- ,ot,"o tl 720 ,/ ting them against l/(n + l), with the end members ; "" being Fo (n : oo) and the F- and OH-sellaites(n : or" 0), MgF, and Mg(OH), (hypothetical).His curves(cl "6/ , Deer et al., 1962,p. 55, Fig. 18) becomeessentially straight lines when the data are plotted against (F,OH)/[(F,OH) + O], as in Figure 4. Refractive indices of the F-series (Van Valkenburg, 1955) and OH-CH and OH-CI (Yamamoto and Akimoto, 1974, 1977)are representedby *'s in Figure 4, and they fall very close to Sahama'scalculated values. 700 ro 08 06 04 02 00 The effect of substituting Fe + Mn for Mg is to in- ( F+ O.5gr.'/( F+OH +Ori) crease(R.I.) by -0.02 per molVoFe * Mn; Ti is ex- Fig. 3. A plot of normalized volumes, :Yol pected to have a somewhat gr€ater efect. But of cell lt - 4 vs. (F + --+ 0.5 Ori)/(F + OH + O1). Tails of vertical lines represcnt data courseOH F and Or, + F also affect (R.I.) and points uncorrected for the deviation of (r1,a) from 0.?2A, the 2V,leadill.g to the conclusionthat, exceptfor norber- radius of Mg2*. Symbols and labels are listed in Table 3. gite, which is generally very F-rich and Fe- and Ti- poor, naturally occurring members of the homolo- graphs).So the abscissais labelled (F + 0.5Ori)/G + gous seriescannot be distinguishedon the basis of OH + Or,), and this schemeis apparentlysuccessful. refractive indices or 2V (Sahama, 1953;Deer et al., Only five of the 27 humite specimenslisted in Table 1962).1 3 have sufficientTi to be significantlyaffected. Density Severalobservations on Figure 3 are pertinent. (l) -+ --+ The humites and clinohumites are apparently in- Densitiesof the F- and OH-seriesSel No Ch --+ --+ distinguishablein this sort of plot, although there are Hu + Cl Fo (and brucite for comparison) were no synthetichumites other than the F-end member to calculated from Mg-end member chemistries and : test this observation.(2) The approximate slopesof unit-cell volumes (see Table 3: Vol. lz x q). ln the lines vary from -3.5 for No to -2.25 for Ch to Figure 4 the F-humites are barely distinguishable, -I.2 for Hu-Cl, suggestingthat the relative effect of but OH has a profound effect on molar volume, as OH"' on the normalized volume increaseswith the noted earlier (cf. Fig. 2). Counteracting effects of proportion of OH in the total (F,OH,O)"'OrV anion heavier transition metals substituting for Mg and array. This lends further support to the idea that in- 3Sahama creasing numbers of proton-proton interactions in (1953) showed that only the extinction angle c:a on the "unoccupied" polyhedra decreasethe stability of (100) [axial conventionas in Jones(1969)] could be usedto differ- the humite homologues,accounting for the extremely entiate betweenchondrodite (22"-3lo) and clinohumite (9"-l5o), and thus of course distinguish phases high pressuresrequired to synthesizeOH-Ch and the monoclinic from their orthorhombic homologues. These extinction angles reflect the OH-CI and for the fact that OH-No hasnot vet been angle between c* and the serrated octahedral chains parallel to made. c: - l9o in chondrodite,- I lo in clinohumite. r034 RIBBE: HUMITE MINERALS

3.4? anion and decreasein density that is observed.This Sel No ChHu Cl Fo o II repulsionexplains the high pressurestability field for - DensitY,F members- _^ E - --.-t7' 5.2 >\ the OH-humites and the fact that OH-norbergite has .- o a- yet ..!9 E not beensynthesized. I o 3.o3 Acknowledgments 3 (, L too (, o I am indebtedto the Earth SciencesScction ofthe National Sci- 28 c, ence Foundation for partial financial support of this work under j x 164 .= grant EAR77-23114. Dr. ClarenceDuffy of Los Alamos Scientific ae. o Laboratory permitted me to use data from his dissertationin ad- 3 268 vanceof its completepublication. x 162 !t o !, o o c -"24 - E @ ,^^ t0v = > l: References o o 2-23 e '""tqa o Brucile Aoki, K., K. Fujino and M. Akaogi (1976)Titanochondrodite and 1 c 2.O titanoclinohumite derived from the upper mantle in the Buell o I rso Park kimberlite, Arizona, U.S.A. Contrib. Mineral. Petrol., 56, E =ll" 243-2sf. t)zv tl 'il;tJ"3I Baur, W. H. (1976) Rutile-type compounds.V. Refinement of - 1.54 MnO2 and MgF2.Acta Crystallogr.,832,2200-22M. E. No ChHu Cl Fo Deer, W. A., R. A. Howie and J. Zussman(1962) Rock-forming ro o.8 0.6 04 0.2 00 Minerals, VoL I: Ortho and Ring Silicates.Wiley, New York. (F,oH)/ftF,oH) Duffy, C. J. (197'l) Phase Equilibria in the System MgO-MgF2- + ol S|O2-H2O.Ph.D. Dissertation,University of British Columbia, Fig. 4. Calculated densities and mean refractive indices of the Vancouver,B.C., Canada. F- and OH-series of synthetic compounds, Sel + No + Ch + Hu Fujino, K. and Y. Tak€uchi (1978)Crystal chemistry of titanian + Cl + Fo. Volume data used in density calculations are from chondrodite and titanian clinohumite of high-pressureorigin. Table 3; refractive indices from Sahama (1953), and observed Am. Mineral., 63, 535-543. values (+) for F-humites from Van Valkenburg (1955) and for Gibbs, G. V., P. H. Ribbe and C. W. Anderson(1970) The crystal OH-Ch and OH-CI from Yamamoto and Akimoto (19'74, 1977). structuresof the humite minerals. II. Chondrodite.Am. Min- eral.,55,1182-1194. Jones,N. W. (1969)Crystallographic nomenclature and twinning larger OH (and Or,) substituting for F lead to the in the humite minerals.lnr. Mineral., 54,309-313. contlusion that density, much like refractive indices, P. H. Ribbe and G. V. Gibbs (1969)Crystal chemistry of is uselessas a determinativemethod. X-ray difracto- thehumite minerals Am. Mineral.,54,39l4ll. metry and electron probe microanalysis are obvi- Kocman, V. and J. Rucklidge (1973)The crystal structureof a ti- ously the most reliable techniquesfor identifying and taniferousclinohumite. Can.Mineral., I 2, 3945. McGetchin, T. R., L. T. Silver and A. A. Chodos(1970) Titanocli- characterizinghumite minerals. nohumite: a possiblemineralogical site for water in the upper mantle.,/. Geophys.Res., 75,255-259. Conclusions Merrill, R. B., J. K. Robertson and P. J. Wyllie (1972) Dehy- All crystal-chemicalarguments and most chemical dration reactionoftitanoclinohumite: reconnaissanceto 30 kilo- and X-ray data point to an ordereddistribution of Ti bars.Earth Planet. Sci.Lett., 14,259-262. Mitchell, R. H. (1978) Manganoanmagnesian ilmenite and tita- into the M3 sitein the M(F,OH)O "layers" of all the nian clinohumite from the Jacupirangacarbonatite, Sio Paulo, humite structures.The limited substitution (x < 0.5) Brazil. Am. Mineral., 63, 544-547. of Ti + O --+ M2* + (F,OH) Ribbe, P. H. (1976)Polyhedral chains in the structuresofgem or- is explicableon the basis '25th of the fact that local electrostaticcharge imbalance thosilicates:relationships to physical properties. Int. Geol. increasessignificantly with increasingTi + Congr. A bstracts,2, 591-594. O. -and G. V. Gibbs (1971) Crystal structuresof the humite From M-OH, F bond lengths,OH"'is found to be minerals:III. MglFe orderingin humite and its relation to other -0.06A larger in its effectiveradius than FII| (Yam- ferromagnesiansilicates. Am. Mineral., 56, 1155-1173. amoto, 1977).But as n goesfrom 4 to zero in the OH - and N. w. Jones (1968) Cation and anion sub- humite-sellaite seriesor from (say) F-Ch to OH-Ch, stitutionsin the humite minerals.Mineral. Mag., 37,966-975. it can only be the increasein proton-proton repul- Robinson, K., G. V. Gibbs and P. H. Ribbe (1973a)The crystal structuresof the humite minerals. IV. Clinohumite and tita- sion in the unoccupied polyhedra in the hcp anion noclinohumite.Am. Mineral., 58,4349. array that explains the large increasein volume per _ and _ (1973b)Correction: The crystal struc_ RIBBE: HUMITE MINEMLS 1035

.Mg(F,OH)r. tures of the humite minerals. IV. Clinohumite and titanocli- - (1961)Synthesis of the humites zMg2SiOa "/. nohumite.,4r2. Mineral., 58,146. Res.Nat Bun Stand.,A. Phys.Chem.,65A,415428. Sahama, Th.G. (1953) Mineralogy of the humite grotp. Ann. Yamamoto, K. ( 1977) Hydroxyl-chondrodite. A cta Crystallogr., Acad. Sci.Fennicae, III. Geol.Geogr., jI, l-50. B3J,t48t-1485. Shannon,R. D. and C. T. Prewitt (1969)Effective ionic radii in - and S. Akimoto (1974)High pressur€and high temperature oxides and fl uorides. A ct a Cry st all ogr., B 25, 925-946. investigations in the system MgO-SiO2-H2O. J. Solid State Smyth, J. R. and R. M. Hazen(1973) The crystalstructures of for- Chem.,9,187-195. steriteand hortonolite at severaltemperatures up to 900"C.lm. - and - (1977) The system MgO-H2O-SiO2 at high Mineral.,54 588-593. pressures and temperatures---stability field for hydroxyl- Tilley, C. E. (1951) The zoned contact-skarnsof thc Broadford chondrodite,hydroxyl-clinohumite and l0A-phase.Am. J. Sci., area, Skye: a study of boron-fluorine m€tasomatismin dolo- 277,288-312. mites.Mineral. Mag., 29,621-665. Van Valkenburg, A. (1955)Synthesis of the humites (abstr.).,4rr. Manuscript received, December 18, 1978; Mineral.,40,339. acceptedfor publication, April 3, 1979.