American Mineralogist, Volume 72, pages 89-101, 1987

Synchrotron radiation xANESspectroscopy of Ti in minerals: Effects of Ti bonding distances,Ti valence,and site geometryon absorption edgestructure

GlnNu A. WlvcnuNls Center for Materials Research,Stanford University, Stanford, California 94305, U.S.A.

AssrRAcr The near K-edge X-ray absorption spectraof Ti in a suite of silicate and oxide minerals have been examined in an effort to improve Ti characterization in solids. The spectra consistofelectronic excitations,occurring in the "pre-edge" region, and multiple scattering resonanceson the side and top ofthe edge.The pre-edgefeatures are insensitive to Ti-O bond length, but are sensitive to valence, occurring about 2.0 eV lower in energyin Ti3+ samples.The secondpre-edge feature is also sensitive to octahedral site distortion and to the presenceof tetrahedralTia* owing to intensification createdby d-p orbital mixing. The multiple scatteringfeatures shift to lower energy as Ti-O bond length increasesand also changein number, intensity, and energyin responseto varying ligand symmetry' Analysis ofthese spectralfeatures suggests that tetrahedral Ti4* probably occurs only at very small concentrationlevels in silicates,even when the total Sia* and Al3+ concentration is well below that necessaryto fill available tetrahedralsites, as in the melanite-schorlomite garnets.The levels of Ti3* are more difficult to determine becauseof an apparent sensi- tivity of the pre-edgeTio* features to bulk chemistry. However, the overall results are consistentwith only small amounts of Ti3+, even in samplesfrom localitieswhere other analyses(wet chemical and Mossbauer)have indicated significant Ti3*.

INt:noouc.ttoN transition metals to allow accuratesite-occupancy refine- Ti is an abundant element in the Earth's crust and a ment by X-ray diffraction methods. Thus, the first-row major constituent in many minerals. It also occurs as a transition elementsare commonly lumped together with trace or minor element in many mineral assemblages Ti in crystal-structurerefinement. Therefore, Ti has been where knowledge of the site partitioning and site occu- generallyincompletely characterizedin most mineral sys- pationswould be of value in geochemicalcalculations. tems. Unfortunately, owing to the limitations of most ana- The main questionsconcerning the crystal chemistry lytical techniques, Ti ordinarily can be studied only in of Ti in mineralsare (l) to what extentit actuallyenters terms of gross abundanceas by electron microprobe, or tetrahedralcoordination; (2) its site preferenceamong oc- by inference from details of optical spectroscopy and tahedral sites ofvarying size, chargebalance, and distor- M

(Lageret al., 1981),where the strongpreference of Ni2+ Wooley, 1968; Huggins et al., 1977a, 1977b)generally for octahedral sites dictates the site partitioning, in syn- indicate small amounts of Ti3* and in some casesTi3+/ thetic Ti-rich andradite (Weber et al., 1975), and in syn- Ti,o,ratios of 25o/o,but with considerableFe3+ present in thetic (ca,Y,Fe,Ti) garnets (Geller et al., 1965). These most cases.This is in strict conflict with l-atm aqueous casessuggest that tetrahedral Tia+ can be accommodated Ti3+ chemistry where Ti3* is oxidized to Ti4" by the pres- in considerableamounts by stable structures.Hence the enceofFe3*. Oxygenfugacities believed necessaryto sta- notion that small fractions of the available Tio" may re- bilize Ti3* are too low for terrestrial conditions (Ander- side on tetrahedral sites in natural species cannot be son et al., 1970). However, it may be possible that readily dismissed. unusually low oxygen fugacities are unnecessaryfor the In natural pyroxenes, Ti is generally refined on the stabilization of Ti3* in structuresthat may preferentially smaller octahedralMl site with no clear evidencefor any partition Ti3* and isolate it electronically by rapid crys- tetrahedraloccupancy (Peacor, 1967). Even in caseswhere tallization. This is by direct analogy to the stabilization considerableTi3+ is present,as in a meteoritic pyroxene of Fe speciesout of their customary stability rangesby (Dowty and Clark, 1973), all Ti appearsto reside in Ml occupation in certain structures(Verhoogen, 1962). T13* In amphiboles, Ti is usually assignedto the octahedral does occur in major concentrations in lunar and mete- strip sitesMl, M2, and M3, but someinvestigators have oritic minerals. assignedTi to the tetrahedralsites (Hawthorne, l98l). Another problem is the details ofcharge-coupled sub- The best relevant structural refinement is probably the stitutions that must occur in silicateswhen Tia+ replaces neutron-scatteringstudy of oxykaersutiteby Kitamura et divalent and trivalent ions. For example, in aegirine au- aI. (1975), who found that Ti is strongly ordered into the gites, Tio* presumably substitutesonly for Fe3+and Al3+ Ml site. An earlier X-ray refinement by Kitamura and on the Ml site. Charge balance can be maintained by Tokonami (1971)on the samespecimen had refinedFe * substitution of Fe3+or Al3+ for Sia+on an adjacent tet- Ti together and indicated a preferenceof these ions for rahedral site, by vacanciesin the M2 sites (one Na va- the Ml and M3 sites. In contrast, X-ray refinements of cancyfor eachMl tia+) or on the Ml sites(one vacancy an oxykaersutite by Hawthorne and Grundy (1973) in- for threeMl Ti4+ions), or by an excessof Na on the M2 dicated that all of the Ti resided on M2. A refinement by sites such that [Na] > trivalent Ml ions. In pure acmite, Robinsonet al. (1973)also placedall of the Ti in a high- only the tetrahedral-substitution or vacancy-generation Ti pargasiteinto M2. Since Ti has a distinctive neutron- schemeis possible. In minerals with other structural to- scattering amplitude compared with other elements of pologiesor more complex chemistry, the possibilities are similar atomic number, neutron-scattering refinements more numerous and may be quite complicated. should be superior to X-ray refinementsof Ti, given suf- In the presentwork, the X-ray near K-edge absorption ficient samplequantities. However, the differencesin these specta(xaues) of Ti in a suite of Ti-bearing minerals have refinements may also reflect changesin Ti partitioning been collectedin an attempt to correlate structural infor- during dehydration or oxidation reactions. mation on the Ti sites in these specieswith X-ray ab- In micas, Ti may be a major constituent mainly in the sorption spectral features. The xeNns region of the ab- biotites and oxybiotites (Bailey, 1984).X-ray refinements sorption spectnrm is rich in structure that is sensitive to of lM and 2M, coexistingoxybiotites by Ohta et al. (1982) the details of the Ti coordination environment and the suggestthat Ti preferentially occupiesthe M2 octahedral Ti valence(Waychunas et al., 198l). XANESspectroscopy site. A study ofTi-bearing synthetic phlogopite solid so- has already proved to be of value in the determination of lutions (Robert, 1976) indicated that Ti solubility in- Fe site occupanciesin minerals(Waychunas et al., 1983, creasedwith increasingtemperature and decreasedwith I 986),glasses (Brown et al., I 978),and aqueoussolutions increasingpressure, and that all Ti substitution occurred (Apted et al., 1981,1985). Ti xAr.rES,in combinationwith on octahedral sites according to the scheme 2sit" + analysis of the extended X-ray absorption fine structure Mg"t : 2AlIv + Ti"I. The pressure-temperaturebehavior (nxars) oscillations,have beenused recently to determine is consistent with Ti substitution in a site somewhat the Ti site occupancyin TiOr-SiO, glasses(Greegor et al., smaller than ideal, so that the much smaller tetrahedral 1982) and the coordination of Ti in metamict Ti-bearing site is greatly disfavored. Mansker et al. (19'79)have re- oxide minerals(Greegor et al., 1983). ported considerableTi (up to 14 wto/oTiOr) in a high-Ba XANESspectroscopy and EXAFSspectroscopy allow a di- biotite from nephelinite at Oahu, Hawaii. Despite a con- rect probe into the Ti-site environment, unlike most oth- siderable undersaturation of Si in the tetrahedral site, er analytical tools that must be used in an indirect man- analysisof chemical-substitutiontrends favors Tio* only ner. Most importantly, xaNrs and EXAFSspectra yield in octahedralcoordination. information only on a specificion's immediate environ- The presenceof Ti3" is more difficult to pin down. ment. Hence there is no averaging of structural infor- SinceTi usually occurswith Fe2* and Fe3+,the ostensibly mation as occurs with X-ray scatteringmeasurements or unique optical spectrum of Ti3+ is likely to be obscured rR spectroscopy.Further, thesetechniques can be applied by Fe2+-flr+ chargetransfer (Rossman, 1984; Rossman, to trace quantities of Ti, given sufficient X-ray flux. On 1980; Faye, 1968).Analyses of Ti garnets(Howie and the other hand, xarvesspectroscopy suffers from the cur- WAYCHUNAS:xares SPECTROSCOPYOF Ti 9T

Table1. Ti-bearingsamples examined in theinvestigation

Sample Chemistry Reference

Benitoite San Benito County, Cali- BalolTirolsi2e6NaoorO"oo;trace Al, Fe, Ca, Sr, Zr, V Lairdand Albee (1972) fornia Schorlomitegar- Magnet Cove, Arkansas CarroF4loo(Tioe@Fe6t76Fdir6Mgo i5rMnoqo)[Si, @Alo r'1- Ericksonand Blade (1963) net* (BM 32025a) FefhrlO,"* Schorlomitegar- livaara,Kuusamo, Fin- Ca2 @Nao,@Koo22Fqi€(Tio r@Fe6j1rFe8ir1Mgo@oMno ois)- Lehijarvi(1960) net* land(BM 50014) [si, KAlos5F€tr]01, o@ Schorlomitegarnet ChristmasMountains, Ca2 wFqhJZt o $Hf mTioqsF4.tBF4&EMgo @Mno d- Thisstudy Texas(Caltech 801 |) Cro @r)[Si, reAlo sF€3,.]O," * Melanitegarnet San Benito County, Cali- Ca3 @(Tio 6TrFqjsFeSiosMgo 6?Mno o@Crooo2zro @)[Si2 7so- Thisstudy fornia Alo 1@Fe8t61]Or2@o Neptunite San Benito County, Cali- Lio6ooNa2 ,aKo ezoBao01o(Fer s5Mgo 4eMno 1e6)Ti26osir osr Thisstudy fornia (singlecrystal; o.o*o Stanford collection) Titanite Pale-greencrystal (loca- Ca1 oro(TiossAlo osFeo ois)[Sio$oAlo @2]Os ooo Thisstudy tion unknown; Stan- ford collection) Acmitic clinopyrox- Norway (UCLA collec- Nao 75cao1, Fe6i5(Tio ooFe835Fe610Mgo o6Alo @Mno or- W. A. Dollase(UCLA; pers. ene tion;MS 2971) si, ero6 oo comm.) Kaersutite HooverDam, Arizona Ca1 ssNao FKo s2o(Tio6sFqSToFeet@Mgs o j2Alo 603- Campbelland Schenk (1950) MnooroXOH)o ,1sFo issi6lBAll ,27023M Barkevikite White Creek, California Ca, @Nao 7@l(or,5[fL *FdioFe?]*Mg2 6A10,sMno o'8F Thisstudy (UCLAcollection; MS (oH)r 17siser6Al,07202, om 28731 Biotite MolodezhnayaStation, K1enNao o2sBaoooe(Tio 58Fqi€FeitsoMg2 ssoAlo@Mno 06)- E. S. Grew(pers. comm.) Enderby Land, Antarc- si5317Al2@o22ooo tica (UCLA collection) Ti3*clinopyroxene G. R. Rossman(Califor- NaTi3*Si2O6 Prewittet al.(19721 nia Instituteof Tech- nology) llmenite'- HunakerCreek, Yukon FeTiO3 Thisstudy Territory, Canada (Stanford collection; 4711\ Rutile Macalno,Inyo County, TioseoFeo or3 Alo@r Mfi,p, o* Thisstudy California(Stanford collection; 4839) Anatase V. T. Baker (analyzedre- TiO,with onlytrace impurities agent) aTirO3 ALFA Products,Thiokol, Ti,o3 99.9% Ventron Division

- Material receivedas fine powder; quoted analysisis for representativesample from locality.Semiquantitative X-ray fluorescenceanalysis indicates that these samples have similar concentrationsof maior ions Ca, Ti, and Fe. -. Fine grains in hematite matrix. Hematite has only trace quantities of Ti.

rent weaknessoftheory to fully explain spectralfeatures. beam conditions have been available for the collection of Hence the analysisapproach is at presentlimited to spec- EXAFSdata for most samples. tral correlations between standard structures and un- The minerals examined in the present work are sum- knowns. maized in Table I along with relevant chemical data. Most of the data presentedin this work have been ac- quired with the use of synchrotron radiation generated ExpnnrlrnNTAI, PRoCEDURES during parasitic storage ring operation at the Stanford Synchrotron Radiation Laboratory (SSRL). In this mode The sample preparation and data collection procedures are the X-ray flux is sufficient for the acquisition of good- identical to those described in Waychunas et al. (1983) except was fashioned around the X-ray beam quality xANESdata, but problematical for similar-quality that a helium-filled bag path to minimize air absorption of the 4.9-5.1-keV radiation. ExAFsspectra since the signal to noise requirements for The entrance-beamslits were adjusted to give Ti near-edge ExAFsare about 20 times higher than those for acceptable structureof about I .0-eV resolution asjudged by pre-edgestruc- xANEs.Hence, only the Ti xnvps are analyzed in detail ture peak widths of anatase run over a range of slit spacings. here. The quantitative site coordination and bond dis- Energycalibration on the beam lines at SSRL was done using tancesobtainable from high-quality EXAFsspectra will be an iron-foil standard, taking the first inflection point on the K-edge presentedin a future work when dedicated(high-cunent) as7lll.2 eV. The anatasespectrum was collectedrepeatedly as 92 WAYCHUNAS: xaNnsSPECTROSCOPY OF Ti

BENITOITE z ILMEN ITE f,4 t o a Zr-Ti GARNET

kz MAGNETCOVE SCHORLOMITE J I SAN BENITOCO MELANITE

z o ANATASE F 0_ RUIILE t 4955 4960 4965 4970 497s 4980 4985 4990 o NEPIU N ITE ENERGY(eV) o Fig. 1. The near K-edge X-ray absorption spectrum of il- BIOTITE menite fitted with Gaussianfeatures. The dashedline is the ob- servedspectrum. The fit is of representativequality. BARKEVIKITE

KAERSUTITE an additional standard. This indicated a reproducibility of0.l TITANITE eV in the energyrange. Data analysisfor the xeNss spectrawas accomplishedby least- squaresfitting of Gaussiandistributions to the near-edgespectral envelopeafter subtraction of the polynomial-fitted photoelectric backgroundabsorption. Integratedpeak areasare usedas a mea- sure of feature intensity. A representativeexample of such a fit is shown in Figure 1, which depicts the ilmenite near-edgeab- 4940 4970 5OOO 5O3O sorptronspectrum. ENERGY(eV) Microprobe analysisof severalof the sampleswas performed Fig.2. Comparisonplot of nearK-edge Ti X-ray absorption on carbon-coatedmounted thin sectionswith a reol Model 733 The photoelectronbackground-absorption profile has microprobe utilizing the Bence-Albeedata-reduction scheme. A spectra. removed,and all spectrahave been rescaled in amplitude beam voltageof 15 kV, current of 15 nA, and diameterof l0 been to facilitatecomparison. /rm was used.The analyseswere initially computed taking all Fe as present in the ferrous state. Subsequentrecalculation using evaluation of the most probable chemical analysis by the pro- bital (MO) levels are observed.These are best exempli- cedureof Dollaseand Newman (1984)allowed recasting of the fied in the anatase and rutile spectra as shown in the Fe into ferrous and ferric components. For this purpose, all Ti comparisonof spectrain Figure 2. Of thesefeatures, it is was assumedto be tetravalent. the second absorption at about 4969 eY that undergoes Rnsur,rs the largest intensity change from compound to com- pound. In the lower-symmetryTi4* sites,such as in nep- General near-edgestructure tunite and kaersutite,this feature is quite strong, whereas In the silicatesand oxidesobserved, the near-edgeX-ray the other pre-edgefeatures are weak or absent. absorption spectra of Ti4+ have up to nine resolvable The existenceof three featuresin the pre-edgeregion, absorption features that can be separated into three rather than two as might be expected from straightfor- subgroups for further discussion. In all of the samples ward application of crystal-field or MO theory to an oc- examinedin the presentstudy, the Ti occupiesmainly tahedralTin* bonding arrangement,suggests that a sim- octahedrally coordinated sites. The samples most likely ple one-electrontransition interpretation of the Ti edge to have small amounts of tetrahedral Ti4+ or Ti3* are the may be insufficient.The separationof the secondand Ti garnet samples. Thus changesin edge structure are third featuresis about 2.7 eY in all spectra where these mainly due to variations in octahedral-sitegeometry. featurescan be fitted. This is roughly consistentwith cal- The edgeregion under considerationis between4960 culationsof the MO levels(Tossell et al., 1974)and the and 5020 eV. Above 5020 eV the featuresobserved are one-electrondensity of states(DOS) (Grunes et al., 1982) of much reducedamplitude and are generallyconsidered in a TiOu cluster.Correlation with the analogoussplitting due to single-scatteringEXAFS. in Ti L-[I and O K absorption spectraof TiO, (Fischer, The first three featuresappear in the "pre-edge" region 1972)is alsogood. However, the MO calculationreveals where,in compoundsof other transition metals,transi- no satisfactoryunfilled energylevel assignableto the first tions to unfilled, largely metal character,molecular or- pre-edgefeature, and the DOS calculation similarly pre- WAYCHUNAS: xeNrs SPECTROSCOPYOF Ti 93

Table2. Energiesof fitted Ti X-ray absorptionedge features

Pre-edgefeatures Main-edoefeatures Sample

Benitoite 4966 7(1) 4969 2(1) 4978.8(1) 4985.1(1 ) Schorlomitegarnet (MagnetCove, Arkansas) 4967 1 4969 0 4977 2 4983.1 4988.3 Schorlomitegarnet (Kuusamo,Finland) 4967.0 4969.0 4976 I 4983.1 4988 3 Schorlomitegarnet (ChristmasMountains. Texas) 4967.5 4969.3 4978.3 4983 5 4988.4 Melanitegarnet (San Benito,California) 4966 5 4969.1 4976.4 4982.6 4988.1 Neptunite 4966 I 4969 0 4975.6 4982.1 4988.0 Titanite 4966 5 4969 3 4973.9 4977 I 4982.0 4986.6 Acmiticclinopyroxene 4966.7 4968 9 4971.5 4977 I 4983.8 4987.5 Kaersutite 4966.6 4968.6 4975.4 4981.1 4987.0 Barkevikite 4966 9 4968.7 4976.0 4981.6 4987.2 Biotite 4VOO.J 4968.7 4976.5 4983.3 4988.8 NaTi3tSi,O6 4966.1 4967.3 4969 1 4976.5 4981 9 4987 3 llmenite 4VOO.O 4VOV.O 4972.1 4978 1 4983 7 4989.2 Rutile 4967.0 4969.7 4972 7 4978.6 4984.2 4993.4 Anatase 4966 7 4969 5 4972.0 4977.8 4984.1 4990.2 aTi"O3 4967 7 4976.0 4985.5 BarTiOo'- 4967.8 4974 0 4979.5 4983.8

" Featuretoo weak or ill-definedto fit. -. Analysisof spectrumcollected by Greegoret al., 1983

dicts no weak peak in this region. This failure of one- (1980),who arguedthat the first main shoulderon the electron theory suggeststhat the first peak is due to a absorptionedge (the fourth featureas definedin this work) multielectronexcitation process. is due to a ls to 4p plus shakedown transition, i.e., a Many investigators have observed the first weak pre- two-electrontransition process. edgepeak in the rutile absorption spectrum,but they usu- The last three features seen on the Ti edge of the ex- ally do not attempt to assignit to any transition (Fischer, amined speciesoccur about a secondmajor absorption 1972'.Balzarottiet al., 1980).However. the secondand crest at about 4998 eV. In anataseand benitoite, three third featuresare commonly assignedto the 2trr and 3e, clear featuresare evident, but in other samplesthese ap- "crystal-field" states, respectively. Differing interpreta- pear blurred into a single broadenedcrest. Since features tions for the first peak have been suggestedby Grunes on this crest are usuallv not resolved, there was no at- (1982),who assignedthe weakfirst featureto a core hole tempt made to fit them to individual peakswith the least- exciton state,and by Greegoret al. (1983)who assigned squaresprocedure. the first feature to the normally filled ll," state, which The xeNes spectnrm of Ti4* in tetrahedral coordina- gives rise to an absorption through a shake up or shake tion in the compoundBarTiOo (Greegor et al., 1983)was off transition during the formation of the core hole. also fitted by the least-squaresprocedure. In this spec- The next three absorption featuresoccur on the steeply trum only one strong pre-edgefeature is observed, and slopedside ofthe absorptionedge and at its first crest.In this occurs at somewhat lower energy than the corre- benitoite, only two of these three features are observed, sponding peak in the octahedralTi4* spectra.The rest of whereasin the schorlomite-melanitegarnets, the edgecrest the edgeis similar to that of benitoite, except that there appearsto have been split relative to the benitoite edge is an additional weak featureon the side of the main edge. and the fourth feature shifted to slightly lower energy. Past results with Fe compounds and minerals (Way- In neptunite and the other silicates where Ti presum- chunas et a1.. 1983) and other transition-metal com- ably occupies low-symmetry six-coordinated sites, the pounds(Wong et al., 1984;Bianconi et al., 1985)suggest edgesare all similar with two featuresconstituting shoul- that the edge and pre-edgefeatures yield somewhat dif- ders on the main edgein addition to a singlewell-defined fering but complementary information about the absorb- edgecrest. er site. The energiesofthe pre-edgefeatures are relatively These featureson the main part of the edgeare gener- insensitive to changesin mean metal-oxygen distance, ally believed to be due to multiple scatteringof the eject- but their intensity is coupled to the degreeof d-p mixing ed photoelectric wave by the atomic cageformed by the which accompaniesoctahedral-site distortion. Tetrahe- first few neighbor atomic shells.As such, their energyand dral coordination introduces considerable such mixing intensity should be sensitiveto detailsof this ligand ar- with dramatic intensification of pre-edgefeatures. In con- rangement(Kutzler et al., 1980).Other interpretations trast, the featureson the main part of the edgemay vary have been made, for example, that of Bair and Goddard markedly in energywith rather subtle changesin the site 94 WAYCHUNAS:xervrs SPECTROSCOPY OF Ti

o Ti++ SAI\4PLES e Ti3t SAMPLES

r NaTiSirOn t-- € \ :NEPTUNITE I 2 060 ^.J^ EI .a (N), N NB. N\ B NIB 6 BN BroTrrE(B) z ^\ \ I U l' I \ I 1o J 2040 Ti2O3 o ll I \ I \ ^-T z ^^rl.t .ft \ --Lti__.;CLINOPYROXENE 2020 oo DO IMCtc' 5b tT tl Genruets 6 ,l tvtc i bg z \ MC t\4cI TEXAS (T) u !MC I \I \ \ ilvAARA o 2 000 I \ \ I x \ MAGNETCOVE (MC) I \ \ sAN BENTTO(S8) J '1 I \ \ I 980 .l ?'! a\ lILI\4ENITE U I I \I I \ \ TITANITE(Ti) z I \ I r 1960 Ti'l'n Ti. ri. nl Tio R. -J nulle (n) U l.F .Ti \ I I \ \ I t' __.;ANATASE 1 940 -BENITOITE 4965 4967 4969 4971 4973 4975 4977 4979 4981 4983 4985 49a7 4989 4991 4993 TITANIUMEDGE FEATURE ENEHGY (evi Fig. 3. Effect ofmean metal-oxygen bond length on Ti near K-edge feature energies. geometry and are relatively sensitive to the mean metal- ilar to that observed for Fe2+edge features (Waychunas oxygen distance.This sensitivity is a consequenceofthe et al., 1983). multiple scatteringabout the atomic cagethat gives rise The absorption-feature positions for tetrahedral Ti4+ to thesefeatures. The energiesofall fitted featuresin the in BarTiOo do not conespond to the octahedral-feature compounds studied appear in Table 2. positions that would be expectedfor the Ti-O mean dis- tance in BarTiOo of 1.74 A and are more similar to oc- Shifts of x,tNBsfeature energi€swith metal-oxygen tahedral Ti4* features appropriate to a Ti-O distance of bonding distance 1.94 L. This is the effect of the reduced ligand field The positions of the first six identif,ed pre-edge and strength owing to the smaller coordination number and main edge featuresare plotted in Figure 3 as a function is also observedin Fe2+edge spectra (Waychunaset al., of the mean Ti-O distance. Although there is consider- 1983). However, in Fe2+ spectra, the main edge-crest able scatter,the trend ofdecreasing energy with increas- energiesfor the four-coordinated speciesare at energies ing Ti-O distanceis clear in the main-edgefeatures. Some higher than any of those observedin six-coordinatedFe2+ of the scatter may be due to the use of inappropriate samples. distances.For example, there is only a relatively small proportion of Tia+ substituting for Fe3+in the clinopy- Edge variations betweenTi4+ and Ti3+ roxene hence, it is not known how closely the Ti site in Two samples with high Ti3+ concentrations were ex- clinopyroxene is approximated by distances refined on amined, aTirO3 with the corundum structure (Newnham the basis of mainly Fe3* in Ml. Given the similarity of and DeHaan,1962),and NaTiSirOupresumed to be iso- Tia+ and Fe3+radii, it is reasonablethat these site dis- structural with acmite (Prewitt et al., 1972). The aTi2O3 tancesshould also be quite similar. The energy shift for edgewas quite smooth with only four resolvablefeatures the main-edgefeatures is about -20 eY/A,which is sim- Gig. a). In this respect,the edgeis similar to that of Fe in aFerO. (Calaset al., 1980). Gaussian line fitting of the aTirO, edge indicates an energy shift of analogous features relative to the Tia+ samples.The broad pre-edgepeak is shifted by -1.8, -1.9, and -2.0 eV relative to the main pre-edgepeaks in anatase,ilmenite, and rutile, respectively. Similarly, the side-edgepeak fitted at 4976 eV in aTirO' is shifted z o - | .8, -2.1 and -2.6 eV relative to thesespectra. F , L E The edge-crestenergies are not as easily compared be- o m cause only a single broad feature tops out the aTi2O, edge.However, if the position of this featureis compared with the averageposition ofthe edge-crestfeatures on the anatase, ilmenite, and rutile edges, the relative energy - - - 4960 4980 5000 5020 4960 4980 5000 5020 shifts observedare | .7, | .0, and 3.3 eV, respective- ENERGY(eV) ly. Thus there is an averageshift of about -2.0 eV for Fig. 4. Comparison ofthe near iK-edgespectra ofanalogous each feature or segmentof the edge.Examination of Fig- Ti3+ and Ti4+ structures. ure 3 revealsthat most, if not all, of the shifts in the main- WAYCHUNAS:xexes SPECTROSCOPY OF Ti 95 edgefeatures are due to changesin the Ti4 distance.In e pre-edge tI / contrast, the shift of the feature is much larger u^- NEPTU N tTE tz) uu than any Ti-O bondlength effect. fi KAERSUTITE The NaTiSirO6 spectrum is also depicted in Figure 4, U zt# where it is compared with that of Ti4* in the clinopyrox- fizo ene sample. The NaTiSirOu sample consistedof a finely F ground powder produced from a small (0.5 mm) single )U '. crystal. The best spectraobtained were noisier J than those U c from other samplesowing to the small amount of sample. rfiANtrE // cLINoPYRoxENE >10 Lfi Comparison of the fitted-feature positions with F those in o z the Tia+ clinopyroxene samplegave resultsin accordance u with those observedwith the oxide spectra.All three pre- 2) edgefeatures could be fitted and showed energy shifts of BENITOITE o a -0.6, -1.6, and -2.4 eY; the two side-edgefeatures, -1.3 and -1.9 eV; and the edgecrest, -1.7 eV. As 0102030405060 with the oxides,the main-edgefeatures appear to be shift- o2(d"g2) ed consistentwith the changein bond length, whereasthe Fig. 5. Is - 3d feature intensity as a function ofbond-angle pre-edgefeatures have a much larger valence effect. variance in octahedralTi siresin minerals.

Intensity of pre-edgefeatures as a function of site distortion DrscussroNl As noted above, the intensity of the second pre-edge The concentrationof tetrahedral Tia+ in feature on the Ti K-edge appearsto be sensitive to Ti- natural silicates site geometry. This is a consequenceof the nature of the The plot in Figure 5 suggeststhat the ls to 3dintensity transition process producing the feature, in this case a observed on the K absorption edgein silicates is largely core ls to mainly Ti-ionJocalized 3d trr-type "crystal- due to distortions in octahedrally coordinated sites, and field" state. Such a ls to 3d transition is Laporte forbid- that it is consistent with the degree of such distortion. den for a pure electric dipole interaction, but may be Since the presenceofa portion oftetrahedral Tia* would partially allowed by mixing noncentrosymmetric orbital also greatly intensify the ls to 3dfeature, the presenceof characterinto the excited state. This can be achieved by tetrahedralTia+ could be quantitatively estimatedif these static or dynamic symmetry reduction, or by electric two ls to 3d absorption intensification effects could be quadrupole interactions (Roe et al., 1984). The latter, in separated.The ls to 3d absorption in BarTiOo is quite fact, give rise to a significant proportion ofthe analogous strong, being 700/oof the intensity of the edge-crestmax- pre-edgefeature in several Cu compounds (Hahn et al., imum. If it is assumedthat the silicate ls to 3d feature 1982). However, a correlation can be shown with site intensifies in direct proportion to the tetrahedral Tia+ distortion from perfect octahedralgeometry and the ls to content and that BarTiO* is a suitable l00o/otetrahedral 3d intensity in silicates (Fie. 5). In this plot, o2 is the standard, then rough constraints can be placed on the variance ofthe l2 bonding anglesin the TiO6 octahedron possible tetrahedral Ti4+ concentration in the observed (Robinson etaI., l97l). The error bars indicate the range specres. ofsite variancesobserved for each mineral except in the For example, the best candidatesfor significant amounts caseofthe garnet error bar, which indicates the range of of tetrahedral Tio* are the melanite-schorlomite garnets. observed ls to 3d intensities for all four samples.The Ti- The chemical analyses(Table l) indicate that all have site occupancyresults of Kitamura et al. (1975) on oxy- largetetrahedral-site Si deficienciesand high Ti contents, kaersutite and Ohta et al. (1982) on oxybiotite were used but also have large concentrationsofFe3+ that substitute to selectthe Ti octahedralsite in the kaersutiteand biotite on the tetrahedral sites.The effect of the small andradite structures for the variance calculation (Table 3). structure site distortion is estimated from Figure 5 at 3o/o Although correlation is evident, it is imperfect proba- of the edge-crestabsorption. The observed mean inten- bly becauseorbital mixing is not a simple function of sity for all four garnet samplesis 7.50/0. bond-anglevariance. In contrast, plots of ls to 3d inten- If one relatesthe total ls to 3d intensity in terms of the sity versus bondJength variance show much less cone- octahedraland tetrahedral Tio+ derived intensities using lation. This is as expected,because the radial part ofthe the 3o/oand 70o/ointensification factors and the propor- atomic orbital wave functions is largely unimportant in tions of octahedral and tetrahedral Tia+, a mean tetra- transition probability (Cotton, 197l). Neither type of plot hedral Tia* concentration of 6.70/oof all Ti4* is obtained. for the oxides shows much correlation. The pre-edgere- The presenceofTi3* is neglected.In eachindividual case, gion in all Ti oxide spectra is different from that of the the respective tetrahedral proportions are San Benito Ti-containing silicates, suggestingthat the factors con- melanite 10.10/0,Texas schorlomite 4.lo/o,lvaarire trolling the pre-edgefeature intensity in oxides may be schorlomite 6.20/o,and Magnet Cove schorlomite 6.20/0. different from those in the silicates. However, other factors can also lead to ls to 3d inten- 95 WAYCHUNAS: xrNEs SPECTROSCOPY OF Ti

metry with a single Ti-O distance.Accordingly, a single featureis seenon the main edgewithout subsidiary shoul- ders. Multiple features are resolved about the second Hzo maxima near 4998 eV. The garnetspectra all have similar be- 2ts edgesinasmuch as there is no shoulder immediately AFNET F "split" 2 low the first edgecrest, but the main edgehas been o1o into two featwes that vary in relative intensity among the + j25 (rvrc) samples.The second major edge maximum is at an en- T ergy similar to that in the benitoite spectrum, but has -BENlrotrE unresolvedfeatures. The garnet Ti site symmetry is nom- inally 3 with a center of symmetry and six equivalent Ti- Fig.6. ,, r:)HJ,il:;1TJ;' a runctionormurtipre O distances.However, the garnets in this study are ac- scattering(main-edge) feature energy. tually solid solutions of several components and have intrinsic site-occupationdisorder. Thus, there is actually some range of Ti-O distancesand site symmetries. It is sification. In solid solutions there are always local distor- thus possiblethat the "split" edgeobserved in the garnets tions produced by neighbor-cation substitutions. In the is due to symmetry breaking by chemical disorder. caseof the Ti garnets,the new analysesin this study as In order to consider this further, the energyofthe fea- well as the literature analysesindicate mainly Ca-filled tures on the main-edge maximum (the fourth, fifth, and eight-coordinatedsites. Because the octahedralsites share sixth features depicted in Fig. 3) can be plotted against edgesonly with the eight-coordinated sites, the largest the ls to 3d intensity (Fig. 6). The featuresgenerally shift geometric effectsthat would occur from Ca substitution to lower energies,and featuresfive and six progressively should not be present. However, charge-coupledsubsti- separateas the I s to 3d intensity increases.The downshift tutions such as Fer+ for Sia+ in adjacent corner-sharing in energyis expecteddue to the generalcorrelation of site tetrahedracould have considerableeffect. Since the bond- size and distortion in silicates (Goldman and Rossman, angle variance for the andradite structure refined by No- ms.). The energiesof the benitoite featuresform a smooth vak and Gibbs (197l) is so small, almost any such vari- trend with the feature energiesof the other samples,sug- ations may increasethe site distortion and thus produce gesting that the fourth and fifth garnet features are the greater ls to 3d intensity. Hence, the ls to 3d intensity sameas thoseobserved on the benitoite edge.These move observedmay not require the presenceofany significant to lower energiesas the site is increasinglydistorted, and amount of tetrahedral Tia+ for its explanation. the sixth feature appears at about 4988 eV. As seen in Another consideration is any energy shift in the ls to Figure 2, this new feature increasesin intensity with the 3d transition due to the coordination changefrom sixfold ls to 3d feature. Fitting of these features shows that the to fourfold. In BarTiO., the ls to 3d transition is shifted increasein the sixth-featureintensity is only a small frac- some 1.6 eV down from its position in the three Tia* tion ofthe increasein the ls to 3d intensity. This is con- octahedral oxides examined in this study. If a similar sistent with the sixth feature being a multiple scattering shift occurs in the ls to 3d absorption in silicates, then effect as opposed to some type of direct electronic exci- one might expect to seean increasingly down-shifted or tation. asymmetric ls to 3d peak as the tetrahedral Ti4+ content The correlation ofthe ls to 3dand sixth-featureinten- increases.Since the tetrahedral absorption intensity should sity demonstratesthe effect of distortion on the two dif- overwhelm the octahedral intensity for any appreciable fering types ofedge features,but does not yield an expla- amount of tetrahedral Ti4*, and the shift is presumably nation for its source. This could conceivably be small smaller than the peak width (about 2 eV), a shift rather amounts of tetrahedral Ti4+, but another explanation de- than a peak shoulder should be seen. Such is not ob- rives from the chemical analyses.The San Benito mela- served, and this is also evidence against significant Ti3+ nite has the smallest tetrahedral Si deficiency,but it has in the garnets,as will be consideredbelow. the largest ls to 3d intensity and largest sixth feature. In Eventually, the clearestdistinction betweentetrahedral contrast, the Texas schorlomite has the largest Si defi- Ti4* and any octahedralTi4* in a distorted environment ciency and the smallest ls to 3d intensity and sixth fea- must be made with the use of specially characterized ture. Sinceone expectsSi deficiencyto correlatewith tet- standards. For example, Ti xaNes and exers spectra rahedral Tia+ occupancyif the latter occurs at all, these should be obtained for Ti-bearing garnetsthat have had observationssuggest that the ls to 3d intensity is not due high-quality site-occupancyrefinements made with neu- to tetrahedral Tio*. tron-diffraction methods. A possible factor influencing site distortion may be chargebalancing between the octahedral and tetrahedral Effect of site symmetry on XANESstructure sites. Ti4+ on the octahedral site may force the structure The spectalsequence depicted in Figure 2 was arranged to position a 3+ ion on one of the nearest tetrahedral to suggesta progressionin Tio* edgestructure as the site sites. In the San Benito melanite there are not enough symmetry varies. In benitoite, the Ti site has 32 sym- trivalent ions on thesetetrahedral sites to locally balance WAYCHUNAS: xeNes SPECTROSCOPYOF Ti 97

Table 3. Ti site symmetry, bond length, and bond-anglevariance information

Sample Ti sitqs) Symmetry Metal-oxygen bond length (A) d (deg') References Benitoite One, octahe- 6-1.942 4.46 Fischer(1969) dral Schorlomite-melanite One, octahe- 6-2.024' 1ea Novak and Gibbs garnets dral (1e71 ) Neptunite Two, octahe- 2.045 2.067 40.0,50.6 Cannilloet al. dral 1.918 2.067 (1e66) 2.096 2.195 2.065 1.986 2.218 2.010 1.990 2.057 2555.. 2J64.. One, octahe- 2.014 Speer and Gibbs dral 1.991 (1976) 1.984 1.973 2.025 1.766 rg5g.. Acmitic clinopyroxene M1, octahe- 2 2-2.1091 41.9t clark et al. (1969) dral 2-2.029t 14.8ft 2-1.9361 [email protected]. Kaersutite,barkevi- M1,M2, M3 2,2,21m 2-2.005 2-2.131 4-2.089 47.9,31.2,71 .9 Hawthorne (1978) kite octahedral 2-2.188 2-2 045 2-2.O01 56.4,34.0,79 4 Kitamuraet al. 2-1.950 2-1.977 2T.59* (1975) 2M-7.. 2.0-5r- Biotite Ml, M2 octa- 21n,2 2-2.142+ 2.137$ 31.7,26.4 Hazen and Burn- hedral 2-2.088+ 2.1429 ham (1973) 2-1.947+ 2.091$ Ohta et al. (1982) 2.059.. 2.091$ 1.950$ 1.947$ rf60- One, octahe- 3-2.0886 Wechsler and dral 3-1.8744 Prewitt (1984) 1.-96-1S. One, octahe- 2-1 984 29.5 Cromer and Her- dral 4-1.946 rington(1955) 1-.959'. 't07.2 One, octahe- 42m 2-1.964 Cromer and Her- dral 4-1.937 rington(1955) i-.9;f6,- 'Andradite. .' Mean metal-oxygen distance. t Acmite. tt Hedenbergite. t M2 site, 1M structure. $ 2Ml structure.

the octahedral Ti4+, whereas in the Texas schorlomite sity changes,most of the Tia+ in the structure must be there are plenty. The two other schloromite garnetshave afected by a site distortion. This is most consistentwith intermediate situations. The poorer the chargebalancing, distortion of the octahedral Ti4* sites to similar degrees, the more distortion presumably occurs in the octahedral rather than tetrahedral occupation by a small fraction of Ti4+ site, and thus the larger the ls to 3d intensity. the available Tia+ ions. This explanation also involves most of the Tia+ in the The Ti site in ilmenite has site symmetry 3 (Wechsler garnet structure, whereasthe assumption of a small por- and Prewitt, 1984), and its absorption spectrum is very tion of the Ti4+ content on the tetrahedral site presum- similar to the spectraof the garnets,particularly the Tex- ably affectsonly that fraction. On the basis of chargebal- as schorlomite. However, the ls to 3d intensity is larger ance, Ti4+ should not prefer tetrahedral sites sharing than in the garnet samples.The bond-angle variance of corners with octahedral sites also containing Ti4+. Thus, the ilmenite Ti site is 86 deg'z,some 63 times larger than any tetrahedral Tia+ would tend not to affect other octa- the andradite garnet site. This is due to the loss of the hedral Tia+ in the garnet structure. However, owing to center of symmetry and two considerablydiffering bond the edge-featureenergy shifts observedas ls to 3d inten- lengths. Nevertheless,it is clear that the 3 and 3 sym- 98 WAYCHUNAS: xeNEsSPECTROSCOPY OF Ti

having a center of symmetry, and a bond-anglevariance of 29.5 deg'?.The ls to 3d intensities in these oxides are thus inconsistent with the results for the silicates, and with symmetry considerations in general. Extension of this study to Ti oxides with lower-symmetry sites and / SHIFT solid solutions may help to explain this anomaly. Ulti- I cP( I ,r' mately, MO calculations must be performed on model K.: .?!c!' - SILICATES NaTiSi206 ;AH M6bB BEN clusters that take into account possible metal-metal N 4s67 bonding and the silicons bonded to coordinated oxygens of silicates. r"Ttor=oru#uJr=rn","l,tto in the case Fig.7 . Theenergy shift of the ls ' 3dfeature due to valence Identification of Ti3* in silicates changebetween Ti3* and Tia+and systematicstrucrural and bulk- The determination of the Ti3*/Ti", ratio in silicates chemistrydifferences in oxidesand silicates. from the Ti xruvesrequires the separation ofvalence effects from those due to mean Ti-O bondJength, site-symme- metries produce similar multiple-scattering featuresand try, and bulk-chemistry variations. From an examination that these featuresmay evolve from a simpler set gener- of Figure 3, it is clear that the main-edgefeatures (num- ated by a 32 (benitoite) case. bers 4, 5, and 6) are largely unaffected by bond length The other silicateshave presumably much-lower sym- and hence cannot be used to discriminate between long metry Ti sites, but also fit into the general trends. For Tia+-O bonds and short Ti3"-O bonds.However, the pre- example, in the kaersutite edge,the sixth feature has the edgefeatures, in particular the strong I s to 3d absorption, greatestintensity, and the fifth feature has been reduced are more affected by valence changesand relatively in- to a shoulder. The fourth and fifth features are down- sensitive to bond length. shifted in energy relative to the presumed higher-sym- In Figure 7 the energiesof the Ti3" and nominally Tia* metry garnet features. samplesare compared. The mean ls Io 3d energyfor the The biotite, barkevikite, kaersutite, and clinopyroxene two Ti3+ samplesis 4967.5(l) eV, and for the 13 Ti4+ edgesare all roughly similar with one large main edge samplesis 4969.1(l) eV, a shift of 1.6 eV. The oxide ls absorption featureand two lower energyshoulders on this to 3d energiesfor all samples are larger than the corre- edge. However, the geometry of the various Ti sites in sponding silicates.This possiblechemical effect is 0.4 eV these speciesdiffer considerably, though they are all of for the Ti3* caseand about 0.6 eV for the Tio* case.It is similar low symmetry (kaersutite Ml, M2, and M3 sites difficult to select the appropriate set of samples for va- have symmetry 2, 2 and 2/m, respectively; biotite Ml lence calibration becauseof this difference between the and M2 siteshave symmetry 2/mand 2, respectively;Ml silicate and oxide samples. There may also be a more site in clinopyroxene has symmetry 2). It thus appears subtle shift within the silicate samplesowing to the vary- that the edgestructure observedon these speciesmay be ing nature of the silicate linkages in these samples.For dependent on the 2 and 2/m site symmetry rather than instance,the garnet and titanite energiesare similar, and any specifictype ofsite distortion. these structures have isolated SiO. tetrahedra in their The scatter of feature energiesis expectable,because structures. Benitoite has isolated three-member silicate multiple-scattering resonancesshould only change rings and has a ls to 3denergy closeto the orthosilicates. smoothly in energy for smooth variations in atom-cage The energy of the ls to 3d feature in the clinopyroxene geometry. Thus, changesas a function of site distortion is somewhat lower than that in the orthosilicates,and it should be dependent on the initial symmetry as well as has a chain structure consisting of single-corner-sharing the specifictype ofdistortion. The Ti garnet shifts alone silicate groups. The biotite, kaersutite, and barkevikite are relatively smooth, suggestingthat only one type of features are lower still, and these structures have sheets distortion is occurring as the ls to 3d intensity increases. and double chains ofsilicate units. The anataseedge is interesting becauseof its unique Becauseof thesefactors, estimation of the Ti3* content structure. The Ti site symmetry is 42m, a highly sym- in the samplesthat are most likely to have finite Ti3* is metric site, but the bond-angle variance with respect to semiquantitative at best. None of the garnet samplesap- an ideal octahedronis 107 deg'z,and the site has no center pear to have major Ti3*, although the uncertaintieswould of symmetry. Despite this, the ls to 3d transition is rel- permit about l0o/oTi3*/Ti.,. However, the garnet ls to atively weak, as it is in ilmenite. Given such site distor- 3d energiesdo not differ significantly from one another, tions and lack of centrosymmetry, it is clear that these as might be expectedif there were variations in Ti3* con- oxides must have reduced d-p orbital mixing in the 3d tent among them. The Mdssbauerstudies of Schwartzet MO state relative to the silicates. al. (1980) and Huggins et al. (1977a, 1977b) and the The Ti site in rutile differs significantly from anatase chemical study of Howie and Wooley (1968) suggestthat and ilmenite, though the ls to 3d transition has similar the San Benito garnetspecimen may have significantTi3*, intensity. The Ti site in rutile has mmm symmetry, thus on the order of 20o/oTi3+/Ti. However, the particular San WAYCHUNAS: x,llrEs SPECTROSCOPYOF Ti 99

Benito specimenin this study clearly does not have a ls from oxygen 2p depending on whether the oxygen is to 3d feature position consistentwith more Ti3+ than the bonded to other metal ions or Si. other garnets. 7. The shape ofthe Ti near-edge spectra and the ls to If there is no structural effect on the energy of the ls 3d fealure intensity are consistent with the results of past to 3d feature,then the energiesof the ls to 3d featuresof refinements for Ti site occupation for the kaersutite and the other silicateswould suggestdefinite Ti3+ when con- biotite samples. However, even given optimal conditions, trasted to the garnets.However, the presenceof any Ti3* changes in the site partitioning between octahedral sites in the kaersutite sample, which is strongly oxidized, is of similar symmetry may not affect edge structure in a quite unlikely. Furthermore, it is most unlikely that the dramatic enough manner to allow better than qualitative two amphibole samplesand the biotite sample selected determinations of the partitioning. for this study would all have considerableTi3+ contents. Thus the variations observed are more probably due to AcxNowr,nocMENTS structural effectsor experimental uncertainties. The author is grateful for assistancewith a portion ofthe data exars analysis should aid in the analysis of Ti3+ in collection at SSRL to Michael Apted, Gordon Brown, Jr., David The mi- silicates since the metal-oxygen bond distances should McKeown, Carl Ponader,and Jason Pressesky. silicate croprobeanalyses were performed by Anne Vaughan.Donations differ by 0.05 to 0.10 A between the two valence states. of Ti sampleswere made by Gordon Brown, Jr., Wayne Dollase, However, preliminary analysis of the ex,lFs spectra for and GeorgeRossman. Reviews by Jack Tossell,Wayne Dollase, the Ti garnetsin this study does not reveal the presence and Max Otten improved the quality of the manuscript. Typing of significant concentrationsof Ti3*. and text editing of the manuscript and assembly of the tables were ably performed by Jean King. This work was supported by the NSF-MRL program through the Center for Materials Re- searchat Stanford University. The work at SSRL was supported Cor.ccr,usroNs by the Department of Energy, Ofrce of Basic Energy Sciences; Analysis ofthe near-edgefeatures (xaNes) in a suite of and the National Institute of Health, Biotechnology Resource Ti silicate and oxide minerals indicates the following: Program, Division of ResearchResources. l The near-edgeTi spectrum is sensitive to site sym- metry, and the edge structure differs significantly for RrrnnnNcns markedly differing symmetry. It may be possible to de- Anderson,A.T., Crewe,A.V., Goldsmith,J.R., Moore, P.B., duce Ti site symmetry solely from near-edge Jr., structure, Newton,R.C., Olsen, E.J., Smith, J.V., and Wyllie, P.J.(1970) even in the absenceof a quantitative theoretidal model Petrologic history of Moon suggestedby petrography,miner- for the near-edgeabsorption processes. alogy, and crystallography.Science, 167, 587-590. 2. The multiple-scattering features on the Ti edge are Apted, M.J., Waychunas,G.A., and Brown, G.E., Jr. (1981)Bxers sensitive to Ti-O distances,decreasing in energy as the study of iron and zinc complexesin hydrothermal solutions. 8th Annual SSRL Users Group Meeting. SSRL report 81-03, bond length increases.This verifies trends already ob- 11 1a servedin other transition-metal compounds. -(1985) Structure and specification of iron complexesin 3. There is no compelling evidence for significant aqueoussolutions determined by X-ray absorption spectros- amounts of tetrahedral Tio* in any of the samplesdespite copy. Geochimica et Cosmochimica Acta, 49,208l-2089. Bailey, S.W. (1984) Crystal chemistry of the true micas. Min- relatively high Ti4+ concentrationin general and substan- eralogicalSociety of America Reviews in Mineralogy, 13, 13- tial Si deficienciesin the garnet samples. The ls to 3d 60. intensification that is observed can be explained by charge Bair, R.A., and Goddard, W.A. (1980) Ab initio studies of the imbalance-induced octahedral-sitedistortions. X-ray absorption edgein copper complexes. l. Atomic Cu'?* -27 7 4. The presenceof Ti3+ in the silicates examined is and Cu(II)Cl'. Physical Review B, 22, 276'7 6. Balzarotti, A., Comin, F., Incoccia, I., Piacentini, M., Mobilio, possible but difficult to determine unambiguously with- S., and Savoia,A. (1980) K-edgeabsorption oftitanium in the out standards of similar structure and composition. There perovskites SrTiOr, BaTiO., and TiOr. Solid State Commu- is no good evidence for significant Ti3+ in the melanite- nications,35, 145-149. schorlomite garnets,which are the best candidatesof all Bianconi, A., Fritsch, E., Calas,G., and Petiau, J. (1985) X-ray near-edgestructure of 3d transition elements in of the samplesfor Ti3+ absorption substantial content. tetrahedral coordination: The etrectof bondJength variation. 5. Relative Ti4* site distortion in the silicates can be Physical Review B,32, 4292-4295. judged from the intensity ofthe ls to 3d feature,because Bland, J.A. (1961) The crystal structure of barium orthotitanite, the intensity is not usually due to the presenceof tetra- BarTiOo.Acta Crystallographica,14, 875-881. (1978) hedral Tia+, and from the energy shift of edge features. Brown, G.E., Jr., Keefer, K.D., and Fenn, P.M. Extended X-ray absorption fine structure (rxers) study ofiron-bearing The energyshift is due to the generaltrend of larger sites silicate glasses:Iron coordination environment and oxidation. having greaterdistortion. Geological Society of America Abstracts with Programs, 10, 6. Variations in the intensity of the ls - 3d transition J/J. in responseto Ti-site symmetry appear to differ between Calas, G., Levitz, P., Petiau, J., Bondot, P., and Loupias, G. (1980) Etude I'ordre local autour de fer dans des verres oxidesand silicates.This is presumablydue to differences de silicates naturels et synthetiquesd I'aide de la spectrom6trie in the degree of p character orbital mixing into the 3d d'absorptionX. Revuede PhysiqueAppliqu6, 15, I l6l-l 167. final state, and hence may reflect differing contributions Cameron, M., and Papike, J.J. (1980) Crystal chemistry of sili- 100 WAYCHUNAS: x-cNEsSPECTROSCOPY OF Ti

cate pyroxenes.Mineralogical Society of America Reviews in Howie, R.A., and Wooley,A.R. (1968)The role of titanium and Mineralogy, 7, 5-92. the effectofTiOr on the cell-size,refractive index, and specific Campbell, I., and Schenk, E.T. (1950) Camptonite dikes near gravity in the andradite-melanite-schorlomiteseries. Miner- Boulder Dam, Arizona. American Mineralogist, 35, 67 l-692. alogical Magazine, 36, 77 5-7 90. Canillo, E.,Mazzi, F., and Rossi, G. (1966) The crystal structure Huggins,F.8., Virgo, D., and Huckenholz,H.G. (1977a)Tita- of neptunite. Acta Crystallographica,21, 200-208. nium-containing silicate garnets.I. The distribution of Al, Fe3+, Clark, J.R., Appleman, D.E., and Papike,J.J. (1969) Crystal- and Tia* between octahedral and tetrahedral sites. American chemical characterization of clinopyroxenes based on eight Mineralogist, 62, 47 5-490. new structure refinements.Mineralogical Society of America - (1977b) Titanium-containing silicate garnets. II. The SpecialPaper 2, 3l-50. crystal chemistry of melanites and schorlomites. American Cotton, F.A. (1971) Chemical applications of group theory. Sec- Mineralogist, 62, 646-665. ond edition. Wiley, New York. Kitamura, M., and Tokonami, M. (1971) The crystal structure Cromer, D.T., and Herrington, K. (1955) The structuresof an- ofkaersutite. Tohoku University ScienceReports, ser. 3, 11, ataseand rutile. American ChemicalSociety Joumal, 77, 4708- 125-141. 4709. Kitamura, M., Tokonami, M., and Morimoto, N. (1975)Distri- Dollase, W.A., and Newman, W.I. (1984) Statistically most bution of titanium atoms in oxy-kaersutite.Contributions to probable stoichiometric formulae. American Mineralogist, 69, Mineralogy and Petrology, 5 l, I 67-17 2. 55 3-5 56. Kutzler, F.W., Natoli, C.R., Misemer, D.K., Doniach, S., and Dowty, E. (1971) Crystal chemistry of titanium and zirconium Hodgson, K.O. (1980) Use of one electron theory for the in- garnet:I. Review and spectralstudies. American Mineralogist, terpretation ofnear edgestructure in K-shell X-ray absorption 56, 1983-2009. spectra of transition metal complexes. Journal of Chemical 'l Dowty, 8., and Clark, J.R. (1973) Crystal structure refinement Physics, 3, 3274-3288. and optical propertiesof a Tir* fassaitefrom the Allende me- Lager, G.A., Armbruster, T., Ross, F.K., Rotella, F.J., and Jor- teorite. American Mineralogist, 58, 230-242. genson,J.D. (198 l) Neutron powder diffraction study ofdefect Erickson, R.L., and Blade, I.V. (1963) Geochemistry and pe- spinel structures: Tetrahedrally coordinated Ti4* in trology of the alkalic igneous complex at Magnet Cove, Ar- Ni2 62Tio6eO4 and Ni, orTio,osio ojoo. Journal of Applied Crys- kansas.U.S. Geological Survey ProfessionalPaper 425. tallography,| 4, 261-264. Faye, G.H. (1968) Optical absorption spectraof certain transi- Laird, J., and Albee, A.L. (1972) Chemical composition and tion metal ions in muscovite, lepidolite, and fuchsite. Cana- physical, optical and structural properties of benitoite, nep- dian JournalofEarth Sciences.5, 3l-38. tunite and joaquinite. American Mineralogist, 57, 85-102. Fischer, D.W. (1972) X-ray band spectraand molecular-orbital Lehijarvi, M. (1960) The alkaline district of Iivaara, Kuusamo, structureof TiO,. PhysicalReview 8,5,42194226. Finland. GeologinenTutkimuslaitos (Bulletin de la Commis- Fischer,K. ( 1969)Verfeinerung der Kristallstruktur von Benitoit sion G6ologiquede Finlande) no. 185. BaTilSirOJ. Zeitschrift fiir Kristallographie, I 29, 222-243. Mansker,W.L., Ewing, R.C., and Keil, K. (1979) Barian-titanian Geller, S., Sherwood,R.C., Espinosa,G.P., and Williams, H.J. biotites in nephelinites from Oahu, Hawaii. American Min- (1965) SubstitutionofTi4*, Cr3*, and Ruo* ions in yttrium eralogist,64, 156-159, iron garnet. Journal of Applied Physics,36, 321. Newnham,R.E., and DeHaan,Y.M. (1962)Refinement of the Greegor,R.B., Lytle, F.W., Ewing, R.C., and Haaker, R.F. (1982) aAl2O3,TirOr, VrO, and CrrO, structures.Zeitschrift fiir Kris- Investigation of titanium in metamict Nb-Ta-Ti oxides using tallographie,| 17, 235J37. the extendedX-ray absorption fine structure techniQue.In W. Novak, G.A., and Gibbs, G.V. (1971)The crystalchemistry of Lutze, Ed. Scientific basis for radioactive waste management, the silicate garnets.Arnerican Mineralogist, 56, 79 l-825. V. Elsevier, Amsterdam. Ohta, T., Takeda,H., and Takeuchi, Y. (1982) Mica polytypism: Greegor, R.B., Lytle, F.W., Sandstrom, D.R., Wong, J., and Similarities in the crystal structuresof coexisting lM and 2M, Schultz,P. (1983) Investigation of TiOr-SiO, glassesby X-ray oxybiotite. American Mineralogist, 67, 298-3 lO. absorption spectroscopy.Journal of Noncrystalline Solids, 55, Peacor,D.R. (1967) Refinement of the crystal structure of a py- 27-43. roxene of formula MrMrr (SirsAlor)Ou. American Mineralo- Grunes, L.A. (1982) A study of core excitation spectra in 3d gist, 52, 3 l-4 1. transition metals and oxides by electron energyloss spectros- Prewitt,C.T., Shannon,R.D., and White, W.B. (1972)Synthesis copy and X-ray absorption spectroscopy.Ph.D. thesis,Cornell of a pyroxenecontaining trivalent titanium. Contributions to University, Ithaca, New York. Mineralogy and Petrology, 35, 77 -82. Grunes,L.A., Leapman,R.D., Wilker, C.M., Hoffman, R., and Robert, J .-L, (1976) Titanium solubility in synthetic phlogopite Kunz, A.D. (1982) OxygenKnear-edge fine structure:An elec- solid solutions.Chemical Geology, 17, 213-227. tron energyloss investigation with comparisonsto new theory Robinson,K., Gibbs, G.V., and Ribbe, P.H. (1971)Quadratic for selected3d transition metal oxides.Physical Review B, 25, elongation: A quantitative measure of distortion in coordi- 7157-7173. nation polyhedra. Science,17 2, 567-570. Hahn, J.E., Scott,R.A., Hodgson,K.O., Doniach, S., Desjardins, Robinson,K., Gibbs, G.V., Ribbe,P.H., and Hall, M.R. (1973) S.R., and Solomon, E.I. (1982) Observation of an electric Cation distribution in three hornblendes. American Journal quadrupole transition in the X-ray absorption spectrum of a of Science.273A'. 522-535. Cu(II) complex. Chemical PhysicsLetters, 88, 595-598. Roe, A.L., Schneider,D.J., Mayer, R.J., Pyrz, J.W., Widom, J., Hartman, P. (1969) Can Ti4* replaceSia+ in silicates?Mineral- and Que, L., Jr. (198a) X-ray absorption spectroscopyofiron- ogical Magazine,3'l , 366-369. tyrosinate proteins. American Chemical SocietyJournal, 106, Hawthorne, F.C. (1978) The crystal chemistry of the amphi- 1676-168r. boles. VL The stereochemistryofthe octahedral strip. Cana- Rossman, G.R. (1980) Pyroxene spectroscopy.Mineralogical dian Mineralogist, 16, 37-52. Society of America Reviews in Mineralogy, 7,93-115. -(1981) Crystal chemistry of the amphiboles. Mineralog- -(1984) Spectroscopyof micas. Mineralogical Society of ical Society of America Reviews in Mineralogy, 9A, l-I02. America Reviewsin Mineralogy,13, 145-181. Hawthorne, F.C., and Grundy, H.D. (1973) The crystal chem- Schwartz,K.B., Nolet, D.A., and Burns, R.G. (1980) Mijssbauer istry of the amphiboles.II. Refinement of the crystal structute spectroscopyand crystal chemistry of natural Fe-Ti garnets. of oxy-kaersutite.Mineralogical Magazine, 39, 390-400. American Mineralogist, 65, 142-1 53. Hazen, R.M., and Burnham, C.W. (1973) The crystal structure Shannon,R.D., and Prewitt, C.T. (1969) Effective ionic radii in of one layer phlogopite and annite. American Mineralogist, oxides and fluorides. Acta Crystallographica,825, 925-9 46. 58,889-900. Speer,J.A., and Gibbs, G.V. (1976) The crystal structure of syn- WAYCHUNAS:x.mrss SPECTROSCOPY OF Ti t0l

thetic titanite, CaTiOSiO", and the domain texture of natural pounds: II. Exers. Physicsand Chemistry of Minerals, 12, 3l- titanites. American Mineralogist, 6 l, 238-247. 47. Tossell,J.A., Vaughan,D.J., and Johnson,K.H. (1974)The elec- Weber, H.P., Virgo, D., and Huggins, F.E. (1975) A neutron tronic structure of rutile, wiistite, and hematite from molec- difraction and 57FeMtjssbauer study of a synthetic Ti-rich ular orbital calculations.American Mineralogist, 59, 319-334. gamel Camegi€ Institution of Washington Year Book 74,57, Verhoogen,J. (1962) Distribution oftitanium between silicates 579. and oxides in igneous rocks. American Journal of Science, Wechsler, B.A., and Prewitt, C.T. (1984) Crystal structure of 260,211-220. ilmenite (FeTiOr) at high temperature and high pressure. Waychunas,G.A., Apted, M.J., and Brown, G.E",Jr. (1981)Co- Amedcan Mineralogist, 69, 17G185. ordination and valenceof Ti in kaersutite,schorlomite, titan- Whipple, E.R. (1979) Enors in chemical analysesof two titanian ite and titanite glass from X-ray absorption spectroscopy. micas. Amerian Mineralogist, 64, l3l l. Geological Society of America Abstracts with Programs, 13, Wong, J., Lytle, F.W, Messmer,R.P., and Maylotte, D.H. (1984) 577. A study ofthe K-edge absorption spectra ofselected vanadium -(1983) X-ray K-edge absorption spectra of Fe mineral compounds. Physical Review B, 30, 5596-5610. and model compounds: Near-edge structure. Physics and Chemistry of Minerals, 10, l-9. Waychunas,G.A., Brown, G.8., Jr., and Apte4 M.J. (1986)X-ray Moruscnrrr REcETvEDMnncx 23, 1986 K-edge absorption spectra of Fe minerals and model com- Mem;scnrrr AccEprEDSrPTEMBER 2, 1986