Titanian Clinohumite in the Carbonatites of the Jacupiranga

American Mineralogist, Volume 77, pages 168-178, 1992

Titanian clinohumite in the carbonatitesof the JacupirangaComplex, Brazil: Mineral chemistry and comparisonwith titanian clinohumite from other environments Josf C. Glspan Departamento de Mineralogia e Petrologia, Instituto de Geoci€ncias,Universidade de Brasilia, 70910 Brasilia DF, Brazil

Ansrucr Titanian clinohumite (TiCl) occurs in the carbonatites of the Jacupiranga Complex, Brazil, either as the result of alteration of olivine (Mitchell, 1978) or in the reaction rock developed in the contact ofthe carbonatitesand the host magnetite pyroxenite, where it may not be related to olivine. Lamellae of titanian chondrodite (TiCh) are found, and the presenceof lamellae of polysomesrepresenting n > 4 is inferred. The TiCl samplescontain TiO, up to 3.78 wto/0,FeO up to 3.09 wto/0,and F up to 2.34 wto/o.Fe and Ti show a positive correlation. Complex zoning is always present. The only consistenttrend is that of increasingTiO, toward the margins, but even this trend is only observedin the crystals of the reaction rock. The partitioning of Fe and Mg between TiCl and olivine indicates enrichment of Mg in the TiCl and is different for samples from the three sovites. The characteristicsof the TiCl are thought to derive from a changing metasomatic environment, and all TiCl so far reported from carbonatite complexesis possibly of metasomatic origin. The JacupirangaTiCl is similar in composition to TiCl from marbles, an observation that cannot be extendedto TiCl from carbonatitesin general.Samples of TiCl from kimberlites have compositions similar to those from some Alpine peridotites. Fluorine TiCl is certainly stable in the upper mantle, but becauseof its rarity it is considered a mineral of no importance to mantle geology.

INrnonucrrou and the host magnetite pyroxenite (Gaspar and Wyllie, presented Titanian clinohumite (TiCl) is not an abundant min- 198.3).Microprobe analysesare in this contri- petrological implications eral in the Earth's crust but is the most abundant species bution, and some structural and of the humite group. Specialattention was paid to it'*tren are discussed'Analytical results are compared with those geological in 1976 Aoki et al. suggestedthat it coulcl be a storage from other environments, and some consid- location of Hro in the mantle. However, tatei studils erations^regardingTicl occrurencein the upper mantle presented' generatedcontrary evidence(Smith, 1977;Mitchell, 1978; are briefly Engiand Lindsley, 1980;Trommsdorffand Evans,1980). Dymek et al. (1988) returned to the problem, sr ggesting that titanian chondrodite (Tich) and not Ticl "ought to Tnn JAcUpIRANGA cOprpr,ex be evaluated further as a possible upper mantle mineral, Melcher (1966) presentedthe geologyof the Jacupiran- as it is the expectedhigh-pressure phase." ga Complex including the carbonatites.Gaspar and Wyl- The crystal structureof the TiCl has been studied by lie(1983)summarizedtheavailableknowledgeaboutthe Robinson et al. (1973), Kocman and Rucklidge (1973), complex and presentednew data on the carbonatites.Five Fujino and Tak6uchi (1978), Ribbe (1979), and Abbott different carbonatite intrusions were recognized (Cl to et al. (1989), and TEM studies of this mineral were car- C5, from the oldest to the youngest).Cl, C3, and C4 are ried out by Muller and Wenk (1978) and White and Hyde sovites (calcite rock), C2 is a dolomitic sovite, and C5 is (1982). Experimental work on the stability of the hy- a rauhaugite (dolomite rock). Based on mineral chemis- droxyl-TiCl has been performed by Merrill et al. (1972), try, a sequenceof magmatic evolution from C2 to C5 was Yamamoto and Akimoto (1977), and Engi and Lindsley established,but Cl representsa chemically separatein- (1980). Reviews of the humite group minerals can be trusion (Gaspar and Wyllie, 1983, 1987). Roden et al. found in Ribbe (1980)and Deer et al. (1982). (1985) determinedthe Sr and Nd isotopic composition Clinohumite was reported in the Jacupirangacarbon- of the Jacupirangacarbonatites and found that Cl to C3 atites by Melcher (1966), and one analysis of it was pre- are isotopically different from C4 and C5, and magma sentedby Mitchell (1978). TiCl occurs in the Jacupiranga contamination is evoked. Gaspar (1989) presented a carbonatites as a product of alteration of olivine or in a comprehensive study of the geology and mineralogy of rock that resulted from the reaction of the carbonatites the complex. 0003-oo4x/92l0102-0168$02.00 168 GASPAR: TITANIAN CLINOHUMITE t69

The reaction rock

A reaction that developed between carbonatites and CLINOHUMITE pyroxenite a-cl the host magnetite formed a rock that is com- o-c3 posed of an alternation of carbonate and silicate-rich tr-cq bands, millimeters to centimeters in width. The whole X - REACTIONROCK CHONDRODITE banded rock may reach more than 2 m in width. The o -cl E na silicate bands are composedmainly of brown phlogopite, u "- 6 acicular alkali amphibole, and magrretitewith platy exso- = l lution lamella of ilmenite. Calcite and apatite are also z z Xx common. TiCl and serpentinepseudomorphs after oliv- o present. pyroxenite F ine may be Near the magnetite con- ooz tact, the grain size is submicroscopic, becoming larger oo 1*"" near the carbonatite bands. Locally, meters-widepockets F of pegmatitic recrystallization occur where phlogopite crystals may reach up to 40 cm in diameter. In these portions, many other mineral speciesmay be found, but augite, perovskite, and severalzeolites are the most common. The carbonate bands are composed of calcite with accessorybrown phlogopite, apatite, and amphibole.

Clinohumite 02 03 The JacupirangaTiCl is orange to red-brown in color and displays a light yellow to colorless pleochroism in Fe CATIONNUMBER thin sections.It has two different types ofoccurrence.The Fig. l. TiCl and TiCh analyses(Table l) showinga general first is as reaction rims or patches in olivine crystals in positivecorrelation of Ti and Fe. Diferent total cation numbers the three sovites of the complex where some olivine crys- usedfor normalizations(see discussion in the text) make com- parisons valuesmeaningless. tals may be completely replaced (Mitchell, 1978). Rare basedon absolute coarse-grainedsamples ofthe Cl sovite contain no relicts of olivine. It was in these crystals that rims of chondrodite have been found during the analysis, and no optical TiCl is the variable metal to Si ratios (Mri/Si) as dis- discontinuity has been observed. In C4, the olivines are cussedbelow. The highest FeO content is 3.09 wto/0,and strongly serpentinizedand only small patchesof TiCl are the highest TiO, is 3.78 wto/ofor the TiCl and 4.43 wto/o found. The second type of occurrenceis in the reaction for the TiCh. The TilSi ratio in TiCh is higher than that rock. In this case, the crystals are well developed and of coexisting TiCl but not by a factor of 2 as would be locally very abundant, and they have diameters up to 2 expectedin coexisting TiCh and TiCl (Aoki et al., 1976; mm. Habits vary from anhedral to euhedralshort prisms, Ehlersand Hoikens, 1987;Dymek et al., 1988).This fac- and inclusions of magnetite, ilmenite, and calcite are tor of 2 results because chondrodite contains twice as common. In a few casesthe TiCl is poikilitic to these many Mg(OHF)O "layers" as does clinohumite. The minerals. Mg(OHF)O "layer" is the part of the structure that con- lains the M3 site where Ti replacesMg (Ribbe, 1979). h ANll.yrrc,ll, PRoCEDURE is apparent from Figure l, despite some scatteringof the EDS analyseswere carried out using an ARL/EMX au- data, that Ti and Fe show a positive correlation in the tomated microprobe. WDS analyseswere perfbrmed in a TiCl ofthe Jacupirangacarbonatites, a behavior already Camebax Microbeam automated microprobe. Standard observed by Evans and Trommsdortr(1983) in the TiCl analysisconditions were usedfor both methods.Identical from an Alpine recrystallized garnet peridotite. In detail, results were obtained by the two methods for major ele- however, TiCl from the reaction rock does not display ments (Si, Ti, Fe, and Mg), with some differencesfor Mn this positive correlation: increasingTi is accompaniedby and Ca. As would be expected,the EDS analysesfailed almost constant Fe. to detect Mn and Ca or gave inaccurate results for them The F contents (2.34-1.13 wto/o)are in reasonable becauseof their low concentrations (Table l). Data col- agreement with observations by I-apin (1982) that cli- lection was done on crystal cores and rims. Four to six nohumite from carbonatite environments reach 1.5-2.0 analyseswere carried out on each crystal except in the wto/0.The highest F contents of the TiCh (3.0 and 4.43 chemical profiles. Averages were made according to the wt%) do not represent higher X. contents (=0.35) than homogeneity of M,,/Si in each crystal (seebelow). those of the TiCl (0.3-0.6). X. does not have a well- defrned and constant negative correlation with Ti as would ANlr.yrrcll. RESULTS be expected from a substitution of the type MgF, : TiO, The microprobe results of the Jacupiranga TiCl are (Evans and Trommsdorff, 1983; Dymek et al., 1988). presentedin Table l. The most remarkable feature of the Mg(OHF), : TiO, is certainly the substitution mecha- 170 GASPAR: TITANIAN CLINOHUMITE

Tnsle 1. Microprobe analytical results of clinohumiteand chondrodite samples c1 12345C 5R 6C, 6C" 6R 78C8R (4) (4) (4) (4) (s) (2) (21 (3) (2) (3) (2t 0) sio, 37.35 37.55 37.85 37.80 38.74 35.30 38.85 38.47 35.00 38.47 38.60 38.10 Tio, 3.4s 3.26 2.05 2.08 3.36 4.43 3.55 3.78 2.93 2.54 2.93 2-97 Mgo 53.20 53.20 54.40 54.35 53.84 55.10 53.00 53.63 56.50 53.40 53.35 53.50 FeO' 3.04 2.93 2.73 2.67 2.70 1.25 2.88 2.58 0.76 2.67 2.88 3.09 MnO o.1o 0.28 0.46 0.32 0.18 0.24 0.26 0.26 0.24 CaO 0.06 0.16 o.o9 o.o3 o.o9 0.o2 0.03 0.24 0.06 0.07 F nd nd nd nd 1.37 3.05 1.13 1.25 3.46 1.70 1.75 2.33 HrO" 0.91 2]3 0.60 0.87 2.92 0.92 0.81 1.04 Total 97.20 96.95 97.19 96.99 101.24 102.42 100.37 100.78 102.07 100.02 100.59 101.34 O:F 0.58 1.28 0.48 0.52 1.46 0.71 0.74 0.98 Total 97.20 96.9s 97.19 97.00 100.66 101.14 99.90 100.26 100.61 99.30 99.85 100.36 Si 3.982 4.009 4.002 4.OO4 4.996 2.020 5.989 4.979 1.999 5.024 5.022 4.028 Ti 0.277 0.262 0.163 0.166 0.326 0.191 0.412 0.368 0.126 0.249 0.287 0.236 Mg 8.455 8.467 8.575 8.583 10.351 4.701 12.181 10.350 4.811 10.397 10.348 8.433 Fe2* 0.271 0.262 0.241 0.237 0.292 0.060 0.372 0.279 0.036 0.291 0.313 0.273 Mn 0.009 0.031 0.022 0.042 0.020 0.012 0.029 0.029 0.021 Ca o.oo7 0.018 o.o1o o.oo4 0.006 0.004 0.004 0.01s 0.009 0.008 Total 13.000 13.000 13.OOO 13.OOO 16.000 7.OOO 19.000 16.000 7.000 16.000 16.000 13.000 F 0.560 0.556 0.558 0.516 0.627 0.701 0.722 0.785 oHt 0.788 1.062 0.619 0]48 1.121 0.801 0.704 0.743 MI/Si 2.265 2.243 2.248 2.247 2.202 2.465 2.172 2.213 2.501 2.185 2.186 2.227

c3 c4 12C2R34 12 1C lR 2C 2R 3C (4) (2) (2) (4) (4) (2t (3) (2) (21 (21 (2t (2) sio, 38.75 39.00 39.30 38.90 38.47 38.50 37.77 37.90 37.90 38.30 37.70 38.10 Tio, o.72 0.81 0.76 0.73 0.91 2.28 2.32 2.46 3.21 2.45 3.19 1.64 Mgo 55.90 56.00 56.00 56.40 55.55 55.50 55.10 55.20 54.40 55.00 s4.30 55.70 FeO- 1.17 1.47 1.30 1.17 1.25 0.78 0.81 1.65 1.89 1.89 2.20 1.90 MnO 0.33 0.18 0.22 CaO 0.38 0.35 0.19 0.25 0.03 0.05 0.02 o.12 F nd nd nd nd 2.34 1.23 2.24 nd nd nd nd nd Hro"- 1.01 1.23 1.24 Total 96.92 97.63 97.55 97.45 99.89 99.76 99.72 97.21 97.52 97.64 97.39 97.34 O:F 0.98 o.52 0.94 Total 96.92 97.63 97.55 97.45 98.90 99.24 98.78 97.21 97.52 97.64 97.39 97.34 Si 4.999 5.004 5.991 4.988 4.992 4.975 4.007 3.992 4.OO2 4.025 3.988 3.996 Ti 0.070 0.078 0.087 0.070 0.089 0.221 0.185 0.19s 0.255 0.194 0.254 0.129 Mg '1o.752 10.712 12.726 10.781 10.744 10.691 8.714 8.668 8.563 8.616 8.563 8.708 Fe2* o.127 0.158 0.166 0.125 0.136 0.085 0.072 0.145 0.167 0.166 0.195 0.167 Mn 0.036 0.020 0.020 Ca 0.0s2 0.048 0.031 0.035 0.005 0.008 0.002 0.014 Total 16.000 16.000 19.000 16.000 16.000 16.000 13.000 13.000 13.000 13.000 13.000 13.000 F 0.963 0.507 0.755 oHt 0.860 1.051 0.874 Mn/Si 2.200 2.197 2.172 2.208 2.205 2.216 2.245 2.256 2.249 2.230 2.260 2.253

3R 4C 4R 5C, 5G, 5R 6C 6R 7C 7R (21 (2) (2) (10) (21 (4) (2) (2t (2t (21 sio, 38.10 38.40 38.10 38.70 38.50 38.52 38.40 38.55 38.50 37.85 Tio, 2.45 1.67 2.36 2.50 2.59 3.32 2.52 3.37 1.79 2.41 Mgo 54.90 55.30 55.40 54.90 55.60 54.22 54.90 53.65 54.55 53.20 FeO' 1.95 1.86 1.81 1.65 1.64 1.77 1.63 1.66 2.0't 1.68 MnO o.22 0.30 0.09 0.25 0.32 0.18 0.26 CaO 0.02 0.01 0.01 0.01 0.02 0.00 0.03 F nd nd no 2.27 1.88 1.24 1.47 1.20 1.49 1.56 H"O.. 0.69 1.38 0.99 1.58 0.97 1.20 1.00 Total 97.40 97.23 97.67 100.95 101.92 100.17 100.75 99.75 99.73 97.99 O:F 0.95 0.79 o.52 0.62 0.51 0.63 0.66 Total 97.40 97.23 97.67 99.99 101.13 99.65 100.13 99.24 99.10 97.33 Si 4.013 4.969 3.995 4.993 4.005 4.994 4.035 5.025 s.008 5.020 0.330 0.175 0.240 Ti 0.194 0.163 0.186 0.243 0.203 0.323 0.199 't0.577 Mg 8.621 10.667 8.660 10.559 8.621 10.479 8.600 10.425 10.519 Fe- o172 0.201 0.159 0.178 0.1r13 0.192 0.143 0.181 0.219 0.186 Mn 0.024 o.027 0.010 0.022 0.035 0.020 0.029 Ca 0.003 0.002 0.002 0.001 0.003 0.001 0.0(M Total 13.000 16.000 13.000 16.000 13.000 16.000 13.000 16.000 16.000 16.000 GASPAR: TITANIAN CLINOHUMITE t7l

Taete l-Continued

3R 4C 4R 5C, cuz 5R 6C 6R 7C 7R (2t (2t (2t 00) (21 (4) (2t (2) (2) (2) F 0.926 0.621 0.508 0.488 0.496 0.620 0.6s6 oHt 0.588 0.973 0.845 1.115 0.8,13 1.030 0.863 M-/Si 2.239 2.220 2.254 2.205 2.246 2.204 2.222 2.1U 2.195 2.',187 Note.'C1, C3, C4 : carbonatites; R : reaction rock; (r) : number of analyses; nd : not determined(EDS); XC : core composition; XR : rim composition. 'Total Fe as FeO. " Calculated. f Calculatedby oH : 2 - 2Ti - F. nism in the JacupirangaTiCl as well as in other occur- Si changedthroughout the crystal. In thesecases, similar rences(Jones et d., 1969). Cl could not be detectedwith Mri/Si analyseswere averagedseparately. Two casesof the microprobe. the zoning pattern present in these minerals will be ex- Despite the scatter of data in Figure 2, a roughly in- amined in detail below but only after some discussionof versevariation of TiO, with MgO/(MgO + FeO + MnO) Mri/Si. may be observed. This is caused by the substitution of both Ti and Fe for Mg. The analysis of a Jacupiranga The Mr,/Si ratio TiCl presentedby Mitchell (1978) is also plotred in Fig- Jones et al. (1969) showed that the general formula ure 2. According to Mitchell's description, it is a Cl TiCl. for the Ti-bearing humite minerals is n [MrSiOo]' The TiO, content of this TiCl is similar to that from the [M,-,Ti,(OH,Or,r,O',] where M includes the octahe- Cl carbonatite, but it has more FeO than any of the Cl drally coordinated cations Mg, Fe, Mn, Ca, Zn, etc., x < TiCl presentedin this work. Observethat TiCl from each l, and n is an integer that varies from I to 4, representing carbonatite and from the reaction rock has characteristic respectively the mineral speciesnorbergite, chondrodite, compositions that separatefields for each group. As ob- humite, and clinohumite. Ribbe (1979) observed that x servedin Figure 2, zoning is another feature of this TiCl. never appearsto exceedapproximately 0.5 Ti atoms per TiCl from carbonatites presents random variations, formula unit as a result of chargeimbalances. Following whereasthat from the reaction rock shows a consistent the general formula above and using Mri as the sum of increaseof Ti toward the margins. Crystal averageswere all octahedral cations including Ti (to follow the Joneset always made for TiCl from carbonatites except when Mr'/ al., 1969, nomenclature),Mri/Si for each member of the group is different. These values are, for z from L to 4, 3.0, 2.5, 2.3, and 2.25, respectively. Analyses of the Jacupiranga TiCl showed variable M.t/ wr7. Si. Table I shows Mrt/Si for the averages, and Figure 3 4

onoly s es 3

F CLINOHUMITE a - ct o_c3 o-cq X - REACTIONROCK o - RIM COMPOSITIONS o - MTTCHELL( 1978) CHONDRODITE o

MgO/(MgO+FoO+MnO) 2)6 2.24 2.3? Fig. 2. Analyses of cores, rims, and coexisting TiCl and TiCh. Observe the constant TiO, enrichment toward the rims of TiCl Mr,/ Si from the reaction rock. Analysis from Mitchell (1978) is a TiCl from the Cl carbonatite. TiCl from the same origin tend to plot Fig. 3. Histogram representing5l individual WDS analyses next to each other, delimiting a separate compositional field for and 22 averagesoftwo EDS analyseseach. TiCh analysesare each carbonatite and for the reaction rock. not included. GASPAR: TITANIAN CLINOHUMITE

Fe(0O6 ) Ti{o27 ) 250 Mri /Si Mq(85l ) xF(o88)

240 Fe(o17) xF(o to)

230 3 4 s 6 7 ro rt 12't'o*,f :"^: "'.J 220 Fig. 5. Electronmicroprobe profile for a TiCl from the reactionrock. Scales and normalizationsas in Figure4. t coRE L-UO- Rlr\4 a largerange of M''/Si would be found, mainly Fig. 4. Electronmicroprobe profile for a crystalfrom the Cl doubt that (the values (for n carbonatitewhere no olivinerelict is present.The scalepresent- from 2.25 clinohumite ratio) to lower : is 2.125).The ed for Mr,/Si is the samefor the otherelements Out the absolute 6 Mr,/Si is2-167 and for n: 8 M''/Si numbersare different) exceptfor X", which is half that scale. lower possible limit is 2.00, corresponding to the value Different values of ,? correspondingto Mn/Si are shown. All for olivine, or n: @. analyseswere normalized to 13 cationsfor the sakeof compar- The great majority of the TiCl analysesfrom Jacupi- ison, exceptX, that was calculatedaccording to assignedn val- ranga are in the range of Mr,/Si from 2.17 to 2.26 (Table ues. I and Fig. 3). The first interpretation ofthese data indi- catesthat theseTiCl are composed of a mixture of n: 4 is a histogram ofall the data except the TiCh. It is clear (M,,/Si : 2.25) and n : 6 (Mrt/Si : 2.167), similar to from these datathat a kind of nonstoichiometry must be those describedby White and Hyde (1982). We have seen present in the Ticl. Most of the data fall below the ex- above that TiCh also occursin the JacupirangaTiCl. This pected value of 2.25 that would be the minimum for the kind ofintergrowth results in larger areasof n: 2 in the humite group, i.e., the clinohumite Mri/Si. Theoretically, TiCl as few analysesseem to indicate a mixture of the it indicates that n > 4 is present in theseminerals. two phases(M,t/Si > 2.25) (Fig. 3). Intergrowths of TiCl Thompson (1978) considers the humite minerals as and TiCh were also reported by White and Hyde (1982) members of a polysomatic serieswhere norbergite is taken in synthetic crystals,in natural crystalsfrom the kimber- as the limit of the series and the other members as an litic diatremesof the Colorado Plateau(Aoki etal., 1976), interlayering of olivine and norbergite modules. Muller and from the Isua dunite (Dymek et al., 1988).Based on and Wenk (1978) found in a preliminary electron mi- X-ray observations, Aoki et al. (1976) determined that croscopy study that structural faults in two samples of their intergrowth was parallel to (001). clinohumite corresponded to humite and chondrodite profiles layers. White and Hyde (1982) carried out a study of Interpretation of chemical synthetic and natural humite minerals using electron mi- The interpretation of the profiles is not simple because croscopywith the purpose of looking for upper members four variables may contribute to the observedvariations. (n > 4) in the group and coherent intergrowths between The first variable is the already-discussedsuperstructures them. They found that clinohumite is a phase likely to and disordered faults inferred to be present, and the oth- contain lamellae parallel to (001) composed of n > 4, ers are the chemical substitutions inherent to these min- where n : odd is almost nonexistent. No perfectly or- erals. The three principal mechanisms of substitution, dered superstructureswere found, but "some crystals gave modified from Evans and Trommsdorff (1983), are (l) lattice images with almost regular groups of faults cor- Mg by Fe (and Mn,Ca), (2) OH by F, and (3) Mg(OHD, responding to higher members of the series" (White and by TiO,. Hyde, 1982, p. 60). These crystalscorrespond, according Figure 4 shows a profile acrossa coarse-grainedTiCl to their figures,to clinohumite sampleswhere the stack- from the Cl carbonatitewhere no relict olivine is present. ing sequenceis mainly made up of n : 4 and n: 6 in Analyses I and 2 correspondto TiCh, whereasthe others different proportions for each superstructure.Highly dis- correspond to intergrowths of TiCl and higher n mem- ordered regions may present lamellae with 14up to 24. If bers. Accordingly, the large variation observed between electron microprobe analyses were carried out on the cli- analyses3 and2 is a reflection ofa changein the relative nohumite samplesof White and Hyde (1982), there is no proportion of the elementspresent or, in other words, the GASPAR: TITANIAN CLINOHUMITE t73 result ofa changein the proportions ofolivine and nor- "titaniferous mineral" that occurs associatedwith the TiCl beryite modules in the structures. In Figure 5 (a TiCl from and olivine in a metamorphosedperidotite from the Vol- the reaction rock), it is also possible to see that for the tri Massif, Italy. The Mri/Si values in their analysesare inner part of the crystal the observed variation in Mg is low (2.065 and 2.005) and would representvery high n also a reflection ofthe changein the relative proportion numbers, chiefly for the lowest value that is close to that of olivine and norbergite modules that is easily observed of olivine. However, the high content of Ti in their anal- by the similar variations of Mg and Mr,/Si. However, at yses (3.95 and 3.93 wt9o) suggeststhat these phasesdo point 9, Mg begins to diverge from Mr,/Si and at the not belong to the humite group. For very high n values, margin has an opposite behavior in relation to it. Such the number of the M(F,OH)O layers,where Ti entersthe behavior is causedby the substitution of Mg by Fe and structure of the humite minerals (Ribbe, 1979), will be Ti, which can be seenin the profile (Fig. 5). so small that such a high TiO, content is not allowed, The H-H repulsion in the humite minerals plays a very unlessTi entersthe olivine layers, which seemsunlikely. important role in their stability (Fujino and Tak6uchi, 1978),and this repulsion becomesstronger with decreas- PrrnocBxnuc coNsrDERATroNs ABour rHE ing n (Ribbe, 1979).Jones et al. (1969) found that X. Jlcuprn-lNc.c' TiCl decreasesfrom norbergite to clinohumite, i.e., with in- Clinohumite is reported to contain the same Mg-Fe creasing n, which is in agreement with Ribbe's (1979) ratio as coexisting olivine (Cimmino et al., 1979; observation. The traversesin the TiCl from Jacupiranga Trommsdorff and Evans, 1980; Dymek et al., 1988). (Figs. 4 and 5) show a similar behavior, which is evi- Evans and Trommsdorff(I983) found that the Ku for Fe- denced by the same general variation of Xo and Mrt/Si. Mg partitioning between TiCl and olivine from an Alpine This fact is more evident in the larger variations of Mr,/ recrystallizedgarnet peridotite is a function of the Ti con- Si, from TiCh to TiCl as in Figure 4 and from TiCl to tent and X, in the TiCl. In Figure 10 of Dymek et al. TiCl intergrown with higher n members as in Figure 5. (1988), two casesdo not fall near the line representinga In detail, the substitution of MgF, by TiO, causesdevi- Ko: I for coexistingclinohumite and olivine. One is the ations in this behavior ofX., as in analysis l3 ofFigure 5. analysis ofthe JacupirangaTiCl and olivine reported by Mitchell (1978). This pair gives rli6 : 0.64. Our data give Mr,/Si and z values Kd: 0.62 for a Cl sample,Ka : 0.49 for a C3 sample, As seenfrom Figures 4 and 5, the variation of Mr,/Si and Ko :0.22for aC4 sample.Note that Mitchell's sam- inside each crystal is significant and the normalization of ple is also from the Cl carbonatite and is in good agree- the analysesbased on a unique number of cations for all ment with our data. The equations relating Ko and Ti or the TiCl (as 13 for example) would result in cation num- F from Evans and Trommsdorff(1983) do not apply to bers with little meaning. The normalization based on a the Jacupiranga pairs, and we were not able to find any fixed number of Si atoms would not solve the problem correlation between the observed values ofKu and other becausethe Si cation number per formula unit is equal chemical variables in the TiCl or the olivine. It is appar- to n and Mri/Si is a function of n. It doesnot matter what ent from thesecalculated coefrcients that the Jacupiranga normalization number is used, Mri/Si is always the same TiCl is enriched in Mg in relation to coexisting olivine. for the same analysis.So, a first normalization was made As already mentioned, TiCl in the Jacupiranga carbon- based on 13 cations, and Mr,/Si was calculated. New n atites occurs as reaction rims or patches in the olivine values were then assignedto different analysesor aver- crystals. It is obvious from the compositions of the two agesaccording to Mr,/Si, even if n needed to be an odd minerals that the replacementof olivine by TiCl implies number. The final cation normalization was made based the entry of Ti, F, and OH in the olivine crystals. The on the corresponding stoichiometric formula of the as- above calculated Fe-Mg partition coefrcients point to a signed n values (see Appendix l). The results are pre- replacementmechanism that includes also a loss of Fe, a sented in Table l, where for each analysis Si (as an in- gain of Mg, or both. The observed differencesin Ko for teger)is equal to the r value used and Total is the cation the pairs from the three carbonatites show that the final number used in the normalization. When we use n : 5 compositions of the TiCl are not a function of the initial or 6, it does not mean that a new mineral of the humite Fe-Mg ratio of the olivines. This fact, combined with the group exists corresponding to such n values, but it cer- marked chemical differencesof the TiCl from the three tainly representsintergrowths of the kind reported by carbonatites (Fig. 2), is evidence that the major control White and Hyde (1982). Only TEM studies are able to on their composition is external variables like tempera- reveal the real structures that are related to these num- ture, pressure,and the chemical composition of the re- bers. acting fluids. The chaotic zoning and the variability in A nonstoichiometric humite mineral has been reported the patterns of structural faults in each crystal (Figs. 4 by Ehlers and Hoikens (1987). It occurs as lamellae in and 5) are certainly a reflection of a rapidly changing Tifree clinohumite that is adjacentto olivine in marbles. environment and probably a reflection of variations in The published analysis has a Mr,/Si : 2.155, which is fluid compositions. The previously mentioned unexpect- not far from the value of 2.167 that correspondsto r?: ed relation of the Ti-Si ratios of coexistingTiCl and TiCh 6. Cimmino et al. (1979) presented two analyses of a is certainly also the result ofrapidly changingconditions t74 GASPAR: TITANIAN CLINOHUMITE of crystallization where equilibrium was not always at- currence is not very different from those of Sokli and tained. Catalio I and probably also has a metasomatic origin. TiCl that occursin the reaction rocks is associatedwith Nielsen and Johnsen (1978) regard the Gardiner TiCl to brown phlogopite, calcite, serpentine,magnetite, and il- be a magmatic occurrenceas in "certain kimberlites and menite. The serpentine forms pseudomorphs after oliv- carbonatites." J. Gittins (personalcommunication) found ine. Alkali amphibole, which is a common phasein these a sympathetic range of composition of TiCl with amphi- rocks, has not been found so far in associationwith TiCl. boles and micas in some Cargill carbonatites,which may The textures presentin these paragenesesshow that TiCl representthe first actual evidenceof a possiblemagmatic may be the alteration product of preexisting olivine grains, crystallization of TiCl. but in many others the TiCl seemsto have no relation at all with olivine, as is the case when TiCl is present as CoNrp,c,RrsoNoF cLrNoHUMrrE coMprosrrroNs FROM subhedral to euhedral prismatic crystals reaching up to DIFFERENT ENVIRONMENTS 20o/oof a silicate band and no olivine or pseudomorphs Clinohumite occursin five different environments: car- of it are present.Despite a similar variability in stacking bonatites, kimberlites, Alpine peridotites, Archean ultra- faults as in the TiCl from the carbonatites (Figs. 4 and mafics,and marbles.Their mode of occurrenceand origin 5), the TiCl in the reaction rocks presentsa constant zon- in carbonatite complexes have been discussedabove. TiCl ing trend of increasingTiO, toward the margins (Table I is reported from the Moses Rock, Green Knobs, and Bu- and Fig. 2), indicating a relatively more constant crystal- ell Park kimberlitic diatremes of the Colorado Plateau lization environment. (McGetchin et al., 1970;Aoki et al., 1976;Smith, 1977, Analyses were obtained on coexisting TiCl and phlog- 1979) and from the Ruslovaya kimberlite, Siberia (Vos- opite in two samplesof the reaction rock. The values of kresenkayaet al., 1965,in Deer et al., 1982).Menzies et Ko for F-OH gave 7.6 and 12.8, which is much higher al. (1987) describedthe diatremesfrom the Colorado Pla- than the partition coefficient of 4 for F-OH between cli- teau as composedof serpentinizedmicrobreccias that for nohumite and phlogopite in marbles (Rice, 1980). De- the case of Green Knobs and Buell Park are closely as- spite this high fractionation of F into TiCl relative to sociated with minette magmatism. According to the de- phlogopite, the F contents in clinohumite should not be tailed study of Smith (1979), the TiCl from the Green usedas an indication of high F environments, as in Lapin Knobs and Buell Park diatremes (and the associatedhy- ( I 982); the half percentof F normally presentin the much drous assemblageamphibole, chlorite, magnesite, and more abundant phlogopite is actually more significant. antigorite) is the product of hydration of upper mantle peridotite before the onset of the magmatic activity. TiCl Huvrrrr MTNERALSFRoM orHER CARBoNATTTES crystals dispersedin the rock resulted from the commi- The humite minerals are classifiedby Heinrich (1963) nution ofthe peridotite inclusions during transport to the as of very rare occurrencein carbonatites. Chondrodite surface(Smith, 1977, 1979). is reported in the Glenover and Nooitgedacht carbona- In Alpine peridotites, TiCl occurs in different gradesof tites in South Africa (Verwoerd, 1966)and from Palabora metamorphism depending on its F content (Evans and (Verwoerd, 1966; Palabora stafi 1976). Minerals of the Trommsdofi I 978, I 983; Trommsdorff and Evans, I 980; humite group occur sporadically in the Cargill Iake car- Cimmino etal.,1919), where its breakdown with increas- bonatite, Canada, according to Gasparrini et al. (1971); ing metamorphism results in an intergrowth of olivine the crystal structure of one TiCl from this carbonatitewas and geikielite (Trommsdorffand Evans, 1980).In Arche- determinedby Kocman and Rucklidge(1973).In the Sokli an ultramafics, TiCl may occur as lamellae in TiCh in carbonatite, the first minerals that form as alteration stratiform ultramafic rocks in the volcano-sedimentary products of olivinites are clinohumite and ferri-phlogo- sequencesof the Isua belt, westernGreenland (Dymek et pite (Vartiainen, 1980). A similar type of occurrenceof al., 1988) or as discrete crystals in the ultramafic bodies clinohumite is found in the Catalio I carbonatite, Brazll, of Wolf Creekand Ruby Range,Montana (Heinrich, 1963; where carbonate-rich veins intruding the dunites cause Desmarais,l98l). the transformation of olivine to clinohumite (Gasparand TiCl occurs both in contact and regionally metamor- Adusumilli, 1976). These last two occrurencesof clino- phosedlimestones, where it can form in the higher grades humite together with the above description of the TiCl (Rice, 1980).In the marbles of the Otztal crystalline base- and TiCh in the JacupirangaComplex point to a common ment, Ti-rich clinohumite and chondrodite often break metasomatic origin for these minerals in carbonatites,a down to a symplectite of Ti-poor clinohumite (or olivine) conclusionalso expressedby Mitchell (1978) and by Lap- and magnesiumilmenite, whereasF-rich clinohumite re- in (1982). mains stable. Nielsen and Johnsen (1978) described veins intruded The compositions of the clinohumite from the above in dunite that contain TiCl-rich centers in the Gardiner geologicalenvironments are compared in Figure 6. This Plateau Complex (a complex where carbonatite is pres- showsa plot ofMgO/(MgO + FeO + MnO) againstTiOr. ent) as well as the occurrence of TiCl in joints of this These two variables take into account the two most ef- dunite. The other minerals in the veins are phlogopite, fective substitution processesin clinohumite: Mg by Fe apatite, antigorite, and minor diopside. This kind of oc- (and Mn) and Mg(FOH), by TiOr. As a generalbehavior, GASPAR: TITANIAN CLINOHUMITE t75 we may say that F has an inverse correlation with Ti and JAC '^.- CAREONATITES that, as a consequence,F decreasesfrom the top to the + X - XENOLITHSIN KINIBERLITES? RGPtrU N-ALPINE PERIOOTITES bottom of Figure 6. EI.ARCHEAN ULTRAMAFICS Clinohumite from marbles delimits a well-defined O- IV]ARBLES composition field (except for one analysis) representing 6 the most Mg-rich clinohumite of all types of occurrence. Marbles contain, as well, some of the most TiOr-poor clinohumite samples.The compositional field ofTiCl from 4 the Jacupirangacarbonatites (Mitchell, 1978; this study) a falls inside that of clinohumite from marbles and presents F the same trend of increasingTiO, with decreasingMgO/ (MgO + FeO + MnO). Clinohumite from veins in dunites from the Gardiner Plateau is at the limit of the marble field. However, TiCl from the Cargill Lake car- o 76 o 80 bonatite is by far the most Fe-rich clinohumite so far Mgo/(Mgo+ Feo+ Mno) analyzed. geo- Clinohumite from the Colorado Plateau has lower MgO/ Fig. 6. Comparison of TiCl compositions from different logical environments. Data referencesfollow. Carbonatites: JAC, (MgO + FeO + MnO) and higher TiO, than those from Jacupiranga from this work; filled diamond, Jacupiranga from marbles and from Jacupiranga.The TiCl richest in TiO, Mitchell (1978); filled triangle, Cargill Lake from Kocman and comesfrom the Moses Rock diatreme (Fig. 6) and is rep- Rucklidge (1973); filled circles,Gardiner Complex from Nielsen resentedby a unique occurrenceas inclusions with oliv- and Johnsen (1978). Xenoliths in kimberlites?: crosses,Green ine and geikielite in garnet (McGetchin et al., 1970).This Knobs, Buell Park, and Moses Rock (Colorado Plateau) from TiCl contains 0.55 Ti per formula unit, which is slightly Smith (1977),McGetchin et d. (1970),Aoki et al. (1976),and above the upper Ti limit, approximately 0.5, as reported Fujino and Tak6uchi (1978); x, Ruslovayakimberlite from Vos- by Ribbe (1979). The humite mineral from Siberia has kresenskayaet al. (1965, in Deer et al., 1982).Alpine peridotites: lower TiO, but similar MgO/(MgO + FeO + MnO) to RGP, recrystallized garnet peridotite (Cima di Gagrone) from with right those from the American diatremes. The TiCl from the Evans and Trommsdorff(1983); square diagonal at top, Cima di Gagrronefrom Evans and Trommsdorff (1978); Colorado Plateau is similar in composition to that from squares,Val Malenco from Trommsdorffand Evans (1980) and the Alpine peridotites, except the sample of TiCl from Jones et al. (1969) in Robinson et al. (1973); squareswith di- the recrystallized garnet peridotite (RGP) reported by agonal at left top, Voltri Massif from Cimmino et al. (1979). Evans and Trommsdorff(1983). However, TiCl from the Archean ultrarnafics: square with x, metadunite from the Isua RGP defines a particular composition field and is not, belt, westernGreenland, from Dymek et al. (1988); filled square, itself, coincident with other Alpine TiCl. The RGP TiCl Ruby Range,Montana, from Desmarais(1981). Marbles: circles, contains the highest F contents of all reported clinohum- Ross Lake, AIta aureola, westem Tauern, Otztal basement, and ite samples associatedwith ultramafic rocks of any en- Hameenkyla (Finland) from Rice (1980), Moore and Kerrick vironment. The only two analysesfrom Archean ultra- (1976\,Fnnz and Ackermand (1980), and Ehlers and Hoikens (1987), (1969)in Robinson (1973), mafics have very different compositions from each other and from Joneset al. et al. respectively. for all chemical variables considered(Fig. 6). TiCl from a metadunite of the Isua belt has a particular composition characteized, bV hreh TiO, and high MeO/(MgO + FeO compositional fields (Fig. 6). Such similarity in compo- + MnO). sition is certainly a reflection of the similarity of the para- In the RGP, Evans and Trommsdorff(1983) found a genesis in both rocks represented by the abundance of large variation in the chemical composition of TiCl crys- carbonate and the presence of at least some of the follow- tals not only as a whole (Fig. 6) but also on a very small ing minerals: olivine, phlogopite, serpentine, magnetite, scaleas, for example, in the same thin section. This fea- and Mg-rich ilmenite. The similarity in composition of ture shows that chemical variations on a millimeter scale TiCl in marbles and the Jacupiranga carbonatites cannot may causea significant shift in the final composition of be extrapolated to carbonatites in general, as illustrated clinohumite. Such characteristics and the above-men- by the composition of the TiCl from the Cargill Lake tioned large diference in composition of two TiCl from carbonatite (Fig. 6). TiCl from the Colorado Plateau and the same environment, those from Archean ultramafics, TiCl from the Alpine peridotites have similar composi- are an indication that TiCl compositions are not a reflec- tions (Fig. 6). The two types crystallized in similar envi- tion of the macrogeologicalenvironment from which they ronments (peridotites with the presence of a fluid phase). crystallize. It is important, however, to remember that this statement is entirely based on major elements and TiCl arn TiCh rN THE UPPER MANTLE? that there are no data available regarding the trace ele- Using the topological analysis of the reactions in the ments or the isotope compositions of TiCl from the dif- ternary subsystem forsterite-geikielite-Hro and the ex- ferent geologicalenvironments. TiCl from the Jacupiran- perimental work of Engi and Lindsley (1980) and Duft ga carbonatites and TiCl from marbles have coincident and Greenwood (1979), Dymek et al. (1988) srggested 176 GASPAR: TITANIAN CLINOHUMITE that TiCh is a mantle mineral. They suggestedthat the etal. (1972).Evans and Trommsdorff(I983) reportedon intergrowths of TiCh and TiCl, such as those from the the occurrenceof fluorine-hydroxyl-TiCl in an Alpine re- xenoliths of the Colorado Plateau and from the Isua crystallized garnet peridotite from Cima di Gagnone, metadunite, "may represent an arrested decompression Switzerland.The X. content of theseclinohumite samples reaction." Despite the fact that the observedintergrowths varies from a little above 0.1 to approximately 0.45 and of TiCh in the JacupirangaTiCl do not seem to be can- the number of Ti cations per formula unit from a little didates for such a proc€ss,there follows a discussion about above zero to approximately 0.4 (from their Fig. a). X. the existenceof thesehumite minerals in the upper man- and Ti show inverse correlation. The paragenesisin- tle. cludesolivine, chlorite, tremolite, cummingtonite, ortho- BecauseMcGerchin et al. (1970), Aoki et al. (1976), pyroxene, and magrresite,and the estimated conditions and Aoki (1977) observedTiCl and TiCh in the xenoliths for the Central Alpine garnet lherzolite (800 "C and 25 from the diatremes of the Colorado Plateau (regardedby kbar) representthe conditions through which the clino- them as kimberlites), they supported the idea that clinohumite sampleswere stable. The Xu present is large enough humite (and chondrodite) may be componentsof the up- for these conditions if the calculated curves from Engi per mantle, being consequentlya site for HrO. Aoki et al. and Lindsley (1980) are taken into account. (1976), based on the experimental work of Yamamoto Cimmino et al. (1979), after studying the TiCl-bearing and Akimoto (1977), proposed that TiCl and TiCh crys- serpentinites from the Voltri Massil Italy, questioned lallrzed,directly from the kimberlite magma. Smith (1977, whether the HrO liberated by the subsidingoceanic litho- 1979)demonstrated that the TiCl-bearing xenoliths from spherewould produce clinohumite and chondrodite after the Colorado diatremes were the result of hydration of olivine in the overlain peridotite edge, a suggestional- upper-mantle peridotites "in a range of water fugacities ready made by Aoki et al. (1976).Smith (1979)also sug- at temperaturesbelow 700 "C and depths from 45 to 60 gested this as a hypothesis to explain the source of the plus km," eliminating the possibility of a direct crystal- HrO that causedthe hydration of the Ticl-bearing xeno- lization of TiCl from a kimberlitic magma. Mitchell liths from the Colorado Plateau diatremes. Helmstaedt (1978), after observing the metasomatic origin of TiCl in and Schulze(1979), based on the similarity of the ultra- the Jacupirangacarbonatites, questioned the suggestion mafic xenolith suite from the Colorado Plateau "kimber- that TiCl is ever a liquidus phase in kimberlites or is lites" to metamorphosed ophiolite complexes of high- important in the mantle. Engi and Lindsley (1980) car- pressuremetamorphic belts, stated that a shallow slab of ried out reversedhydrothermal experimentson a natural subductedoceanic lithosphere existed under the Colora- hydroxyl-TiCl [previous experiments by Merrill et al. do Plateau at the time of the eruptions from which the (1972) and Yamamoto and Akimoto (1977) were not able xenoliths were derived; this theory is more in agreement to reverse the reactionsl and found that it breaks down with recent analysis of the geotectonic implications of at 475 T. (3.5 kbar), 6209 "C (14 kbar), and 625 "C (21 xenoliths in the southwesternarea of the U.S.A. (Menzies kbar) to the assemblageolivine + ilmenite + vapor. et al., 1987). Trommsdorffand Evans (1980) describedthe breakdown Disregarding the suggestion that clinohumite and of hydroxyl-TiCl to olivine * magnesian ilmenite + chondrodite may crystallize from a kimberlite magma, magnetite in the Malenco serpentinite at approximately we are essentiallyleft with two possibilities for the exis- 520'C and 3 kbar. Considering that the TiCl from the tence of clinohumite and chondrodite in the upper man- xenoliths of the Colorado diatremes is also hydroxyl-TiCl, tle. The first is that these minerals crystallize in the hy- it is possible that the experimental data and the condi- drous oceaniclithosphere and are carried down with the tions of stability of the hydroxyl-TiCl deducedfrom nat- slab, and the secondis that HrO releasedfrom dehydra- ural assemblagesare consistent. tion reactionsin the descendingoceanic slab producesthe The entry of Ti in the structure of humite minerals has alteration of olivine to clinohumite or chondrodite in the a destabilizing effectthat limits the Ti content of the nat- peridotite wedgeabove it. ural minerals to approximately 0.5 cations per formula The suggestionof crystallization of TiCl or TiCh in the unit (Ribbe, 1979).It is the substirution of oH by F rhat oceaniclithosphere in subduction zonesis consistentwith increasesthe stability ofthese minerals becauseit reduces the occurrenceof these minerals in ophiolites. The prob- the number of neighboring H atoms, diminishing the lem is the survival of the humite minerals as the slab amount of proton-proton repulsion (Ribbe, 1979). Engi sinks. We have seenthat most of the TiCl that occurs in and Lindsley (1980) calculated the shifts caused in the the ultramafic rocks is Ti rich and F poor, and this type upper stability limit of TiCl containing different mole of clinohumite breaks down at temperaturesbelow that fractions of F and found significant displacementstoward of the upper mantle. As noted by Trommsdorffand Evans higher temperatures.The breakdown reaction (TiCl : ol- (1980), the typical product ofthe breakdown ofTiCl, the ivine * geikielite) for a clinohumite sample containing olivine-ilmenite intergrowths, has never been observed Xr:0.6, accordingto Figure I from Engi and Lindsley in xenoliths or as megacrysts.The only exception is the (1980),would occur at about 1000 "C for a pressureof occrurenceof olivine, ilmenite, and TiCl as inclusions in 3.5 kbar. These calculated curves seem to be in reason- garnet from the Moses Rock diatreme (McGetchin et al., able agreementwith the unreversed reactions of Merrill 1970). No textural detail is given, but this olivine and GASPAR: TITANIAN CLINOHUMITE 177 ilmenite could have originated by the breakdown of TiCl Volri Massif (Westem Liguria): Inferences on the geodynamic evolu- becauseof pressure releasecaused by the ascent of the tion ofpiemontese ultramafic sections.Ofioliti, 4(2), 97-120. W.A., R.A., Zussman, (1982) Rock-forming miner- xenolith. Despite fact Deer, Howie, and J. the that most of the TiCl in ultra- als, vol. lA: Orthosilicates (2nd edirion). Ilngman, London. mafic rocks is F poor, F-bearing TiCl has been found in Desmarais, N.R. (1981) Metamorphosed precambrian ultramafrc rocks in the Alpine peridotites (Evans and Trommsdorfr 1983) the Ruby Range,Montana. PrecambrianResearch, 16, 67-101. and was stable under 800 "C and 25 kbar. Another pos- Duft, C.J., and Greenwood, H.J. (1979) Phaseequilibria in the system sible exception would be the clinohumite sample from MgO-MgFr-SiOr-HrO. American Mineralogist, 64, | 156-l 172. Dymek, R.F., Boak, J.L., and Brothers,S.C. (1988) Titanian chondrodite- the Ruslovaya kimberlite (Voskresenkayaet al., 1965, in and titanian clinohumite-bearing metadunite from the 3800 Ma Isua Deer et al., 1982) (see Fig. 6) that contains 1.8 wto/oF, supracrustal belt, West Greenland: Chemistry, petrology, and origin. but it is not possible to know from their description if American Mineralogist, 73, 547-5 58. this TiCl comes from comminuted xenoliths or is the Ehlers, IC, and Hoikens, G. (1987) Titanian chondrodite and clinohumite product in marbles from the Otztal crystalline basement. Mineralogy and Pe- alteration of olivine from the kimberlite. The fact trology,36, 13-25. is that F-bearing clinohumite seems to be stable under Engr, M., and Lindsley, D.H. (1980) Stability of titanian clinohumite: upper mantle conditions (Engi and Lindsley, 1980). The Experiments and thermodynamic analysis. Contributions to Mineral- suggestionregarding the alteration of olivine to TiCl or ogy and Petrology,72, 415-424. gamet TiCh in the peridotite wedge above the subduction zone Evans, 8.W., and Trommsdofl Y. (197E)Petrogenesis of lherzolite, Cima di Gagnone, I*ponrine Alps. Earth and Planerary Science seemstheoretically possible,but no direct or indirect ev- Letters,40,333-348. idenceofit exists. - (1983) Fluorine hydroxyl titanian clinohumite in Alpine recrys- To the question of whether TiCl or TiCh exists in the tallized garnet peridotite: Compositional controls and petrological sig- upper mantle, the answeris probably yes;but to the ques- nificance. American Journal of Science, 28t4, 355-369. (1980) Phase relations metamorphic tion whether Franz, G., and Ackermand, D. and of they are important to the upper mantle history of a clinohumite-chlorite-s€rpentine-marble from the Western petrology, the answer is certainly no. The rarity of cli- Tauem Area (Austria). Contributions to Mineralogy and Petrology, 75, nohumite as a whole in ultramafic rocks combined with 97-rr0. the fact that fluorine clinohumite (which is the stable phase Fujino, K., and Tak6uchi, Y. (1978) Crystal chemistry of titanian chon- under typical upper mantle conditions) is even rarer in drodite and titanian clinohumire of high-pressure origin. American Mineralogist, 63, 535-543. these rocks and the practical absenceof these minerals Gaspar, J.C. (1989) G€ologie et min6ralogie du cornplexecarbonatitique from world-wide collected xenoliths make their impor- de Jacupiranga,Bresil, 343 p. Ph.D. lhesis, University ofOrl6ans, France. tance to mangle geology negligible. Engi and Lindsley Gaspar, J.C., and Adusumilli, M.S. (1976) Sobre os carbonatitosde Ca- (1980) observedthat if l0loof the upper mantle consisted talao I, GO. Boletim de Mineralogia, 4,5-22. Gaspar, and Wyllie, P.J. (19E3) Magnetite in the carbonatites from of fluorine clinohumite. the calculationsof J.C., the abundance the Jacupirangacomplex, Brazil. American Mineralogist, 68, 195-213. of F in the Earth would be significantly different. This - (1987) The phlogopites from the Jacupiranga carbonatite intru- observation is relevant, but to define the actual abun- sions. Mineralogy aDd Petrology, 36, l2l-134. dance of a certainly very rare mineral in the upper mantle Gasparrini, 8.L., Gitrins, J., and Rucklidge, J.C. (1971) Chemical varia- is not a simple task. tion among the non-carbonate minerals ofthe Cargill Iake carbonatite, Ontario. Absfaas, Canadian Mineralogist, 10, 913. Heinrich, E.Wm. (1963) Paragenesisof clinohumite and associated min- AcxNowr,rncuBxrs eralsfrom Wolf Creek,Montana. American Mineralogist,48, 597-613. I thank Serrana S/A de MineragSo, G.C. Melcher, and V.A.V. Girardi Helmstaedt, H., and Schulze, D.J. ( | 979) Garnet clinopyroxenite-chlorite for making possible the 6eld work; I.M. Steele,J.V. Srnith, and D. Ohnen- eclogite in a xenolith from Moses Rock Further evidence for meta- stetter for assistance with microprobe analyses; J. Roux for invaluable morphosed ophiolites under the C,olorado Pliateau. In F.R. Boyd and computer assistance; Othon H. konardos for English corrections, and H.O.A. Meyer, Eds., The mantle sample: Inclusions in kimberlites and Bernard W. Evans for the general review of the paper. Special thanks to other volcanics, p. 357-365. American Geophysical Union, Washing- the Centre de Recherches sur la Synthese et la Chimie des Mineraux ton, DC. (CRSCM/CNRS), Orleans, France, which, mainly by the assistancegiven Jones, N.H., Ribbe, P.H., and Gibbs, G.V. (1969) Crystal chemistry of by Philippe Rossi, provided me the best work conditions while I was the humite minerals. American Mineralogist, 54,391411. living in Orleans and Peter J. Wyllie, without whom this work would Kocman, V., and Rucklidge, J. (1973) The crystal structure of a titanif- have not been possible. Financial support was provided by Conselho Na- erous clinohumite. Canadian Mineralogist, 12, 39-45. cional de Desenvolvimento Cientifico e Tecnologico (CNPQ Proc. 20lll5/ Iapin, A.V. (1982) Carbonatite differentiation process. International Ge- 87.5. ology Review, 24(9), 1079-1089. McGetchin, T.R., Silver, L.T., and Chodos,A.A. (1970)Tiranoclinohum- Rnrrnnxcns crrED ite: A possible site for water in the upper mantle. Journal of Geophys- Abbotl, R.N., Jr., Burnham, C.H., and Post, J.E. (1989) Hydrogen in ical Research, 7 5(2), 255-259. humite-group rninerals: Structure-energy calculations. American Min- Melcher, G.C. (1966) The carbonatitesofJacupiranga, Sio Paulo, Brazil. eralogist,74, I 300-1306. In O.F. Tuttle and J. Gittins, Eds., Carbonatites,p. 169-181. Intersci- Aoki, K. (1977) Titanochondrodite and titanoclinohumite derived from ence, New York. the upper mantle in the Buell Park kimberlite, Arizona, USA. A Reply. Menzies, M.A., Arculus, R.J., Best, M.G., Bergman, S.C., Ehremberg, Contributions to Mineralogy and Petrology, 61,217-218. S.N., Irving, A.J., Roden, M.F., and Schulze,D.J. (1987) A record of Aoki, K., Fujino, K., and Akaogi, M. (1976) Titanochondrodite and ti- subduction processesand within-plate volcanism in lithospheric xeno- tanoclinohumite derived from the upper mantle in the Buell Park kim- liths of the southwestern USA. In P.H. Nixon, Ed., Manrle xenoliths, berlite, Arizona, USA. Conrributions to Mineralogy and Petrology, 56, p.59-74. Wiley, New York. 243-253. Merrill, R.B., Robertson, J.K., and Wyllie, P.J. (1972) Dehydration re- Cimmino, F., Messiga,B., Piccardo, 8., all,dZada, O. (1979) Tiranian action oftitanoclinohumite: Reconnaissanceto 30 kilobars. Earth and clinohumite-bearing assemblageswithin antigoritic serpentinites of the Planetary Science Letters, 14, 259-262. 178 GASPAR: TITANIAN CLINOHUMITE

Mitchell, R.H. (1978) Manganoan magnesianilmenite and clinohurnite drodite, hydroxyl-clinohumite and l0 A-phase. American Joumal of from the Jacupiranga carbonatite, 56o Paulo, Brazil. American Min- Scierce,277, 288-312. eralogist, 63, 544-547. MaNuscnrsr RECEIVEDSep'reunen 6, 1990 Moore, J.N., and Kerrick, D.M. (1976) fuuilibria in siliceousdolomites I I, l99l ofthe Alta aureole,Utah. American Joumal ofScience, 276,502-524- MrNuscnrp'r AccEprEDSemerdsen Muller, W.F., and Wenk, H.R. (1978) MixedJayer characteristicsin real humite structures. Acta Crysrallographica, A34, 607-609. AppnNnrx 1. THr us^lcr oF Mr/Si To ASSTcNN VALUES TO HUMITE MINERALS Nielsen, T.F.D., and Johnsen,O. (1978) Titaniferous clinohumite from Gardiner Plateau Complex, East Greenland. Mineralogical y[qBazine, From the stoichiometry of the humite minerals, the relation 42, 99-t0t. of Mrt/Si with n is Palabora Mining Company Limited Mine Geological and Mineralogical staff (1976) The geology and the economic deposits of copper, iron, Mr, 2n+l and vermiculite in the Palabora igneous complex: A briefreview. Eco- nomic Geology, 7 1, 177-192. Si: , Ribbe, P.H. (1979) Titanium, fluorine, and hydroxyl in the humite min- cations erals.American Mineralogist, 64, lO27-1O35. where Mri is the sum of all octahedrally coordinated (1969). Appendix Figure - (1980) The including Ti, as defined by Jones et al. humite seriesand Mn-analogs. In Mineralogical So- : ciety of America Reviews of Mineralogy, 5,231-274. I shows the curve of this function that is limited by n I and by Mr,/Si : 2, which is the olivine ratio (n: oo). It is obvious Rice, J.M. (1980) Phaseequilibria involving humite minerals in impure gets progres- dolomitic limestones, Part L Calculated srability of clinohumite. Con- from the curve that the assignmentof n by Mtr/Si tributions to Mineralogy and Petrology, 71,219-235. sively more difficult and inaccurate as /? increases and that for procedure give results, par- Robinson, K., Gibbs, G.V., and Ribbe, P.H. (1973) The crystal structure high z values this will meaningless of the humite minerals. IV. Clinohumite and titanoclinohumite. Amer- ticularly ifanalytical errors are taken into account. However, for ican Mineralogist, 58, 4!-49. the most abundant cases,which are represented by low t?phases, Roden, M.F., Rama Murthy, V., and Gaspar, J.C. (1985) Sr and Nd Mr,/Si is usefrr!. isotopic composition of the Jacupiranga Carbonatite. Journal of Ge- ology, 93, 212-220. Smith, D. (1977) TitanochoDdroditeand titanoclinohumite derived from the upper mantle in the Buell Park kimberlite, Arizona, USA. A discussion.Contributions to Mineralogy and Petrology, 61,2I3Jl5. - (1979) Hydrous minerals and carbonatesin peridotite intrusions from the Green Knobs and Buell Park kimberlitic diatremes on the C-oloradoPlateau. In F.R. Boyd and H.O.A. Meyer, Eds., The mantle p. sample: Inclusions in kimberlites and other volcanics, 345-356. 6 American Geophysical Union, Washin4on, DC. - 25 Thompson, J.B. (1978) Biopyriboles and polysornatic series.American = Mineralogist, 63, 239-249. Trommsdofi V., and Evans, B.W. (1980) Titanian hydroxyl-clinohumite: Formation and breakdown in antigorire rocks (Malenco, Italy). Contributions to Mineralogy and Petrology, 72,229-242. Yartiainen, H. (1980) The petrography, mineralogy and petrochemistry ofthe Sokli carbonatite massif, northern Finland. Geological Survey of Finland Bulletin, 313, l -126. Verwoen( W.J. (1966) South African carbonatites and their probable mde of origin. Annale Universiteit Stellenbosch, 4 l, ser. A, no. 2, 234 p. White, T.J., and Hyde, B.G. (1982) Electron microscopy study of the humite minerals: I. Mg-rich specimens. Physics and Chemistry of Min- n erals,8, 55-63. Yamamoto, K., and Akimoto, S. (1977) The system MgO-SiOlHrO at Appendix Fig. l. Curve representingthe variation of Mrt/Si high pressures and ternperatures-Stability fields for hydroxyl-chon- with n.