parallel electron energy-loss spectroscopy (peels) study...

American Mineralogist, Volume 80, pages 1132-1 144, 1995

Parallel electron energy-loss spectroscopy (PEELS) study of B in minerals:The electron energy-loss near-edge structure (ELNES) of the B K edge

LlunnNcn A. J. G,l,nvrE,* ALAN J. Cnl,vnNDepartments of Physics and Astronomy, University of Glasgow, Glasgow G12 8QQ, U.K.

Rrr BnyosoNDepartment of Materials Science and Engineering, University of Surrey, Guildford, Surrey GU2 5XH, U.K.

ABSTRAcT

The B K-edge spectra of a wide variety of minerals have been measured using thetechnique of parallel electron energyJoss spectroscopy (PEELS) conducted in a scanningtransmission electron microscope (STEM) from sample areas of nanometer dimensions.The B K edges of the minerals exhibit electron energy-loss near-edge structure (ELNES)characteristic of B coordination. For threefold-coordinated B ft3;, the spectra are domi-nated by a sharp peak at ca. 194 eV because of transitions to unoccupied states of zr*character, followed by a broader peak at ca. 203 eV attributed to states of a* character.The ELNES on the B K edge (B K ELNES) of fourfold-coordinated B (tat3; consists of asharp rise in intensity with a maximum at ca. 199 eV followed by several weaker structures.For tolB, the ELNES is interpreted as transitions to states of antibonding o* character.Minerals that possess both tstB and talB exhibit an edge shape that is composed of B Kedges from the respective BO. and BOo units, and we demonstrate the feasibility of quan-tifrcation of relative site occupancies in minerals containing a mixture of B coordinations.The origins of the B K ELNES are discussed in terms of both molecular orbital (MO) andmultiple scattering (MS) theory. We also present the B K-edge spectra of selected non-minerals and show how differences in edge shapes and energy onsets allow these nonmi-nerals to be readily distinguished from borates and borosilicates.

Ir.mnooucrroN

B is impossible or difficult to analyze by the conven-tional techniques used by geoscientists, namely electronmicroprobe and X-ray fluorescence, but can be readilyexamined by several spectroscopic techniques. One of themost widely used bulk techniques for studying B is "Bmagic-angle-spinning nuclear magnetic resonance (MASNMR), from which threefold-coordinated planar (t3iB)triangles and fourfold-coordinated (lolB) sites are distin-guished by their differing chemical shifts (Turner et al.,1986; Marchetti et al., l99l; Mii l ler et a1.,1993; Sil ls etal.,1994; Sen et al., 1994). Additional spectroscopic tech-niques that are used to distinguish between t31B and I4lBare X-ray emission spectroscopy (XES) (Luck and Urch,1990), Raman spectroscopy (Meera et al., 1990), and in-frared spectroscopy (Moustafa et al., 1994). A few elec-tron energy-loss spectroscopy (EELS) and X-ray absorp-tion spectroscopy (XAS) studies have been performed onB-bearing minerals (Brydson et al., 1988b, 1988c; Saueret al., 1993) but have mainly concentrated on inorganicmaterials such as BF, and KBF. (Hallmeier et al., l98l;Schwarz et al.. 1983). XES and EELS studies on the un-

* Present address: Department of Geology, Arizona State Uni-versity, Tempe, Arizona 85287-1404, U.S.A.

0003-004x/95 / | | 12- | r32s02.00

occupied states of B were reviewed by Garvie (1995).These early studies illustrated the large differences in theB K-edge shapes from t3lB and talB.

Spectroscopic techniques, such as XES and X-ray pho-toemission spectroscopy (XPS), have become establishedmethods for probing the occupied electronic states (main-ly bonding states) of minerals. To develop a more com-plete understanding of the properties of minerals, such astheir electronic, thermal, and optical behavior, it is nec-essary to investigate the unoccupied, mainly antibondingstates. Until recently, the study of occupied states in min-erals has commanded most attention. Unoccupied statesmay be probed by a variety ofspectroscopic techniques,including XAS, EELS, bremsstrahlung isochromat spec-troscopy @IS), inverse photoelectron spectroscopy (IPES),and X-ray Raman spectroscopy (Fuggle and Inglesfield,1992). Of these spectroscopic techniques, EELS and XASare most suited to mineral studies.

Recent improvements in EELS instrumentation haveresulted in renewed interest in the study ofthe electronenergyJoss near-edge structure (ELNES) of core-loss edg-es from solid-state materials. EELS performed in a con-ventional or scanning transmission electron microscope(CTEM or STEM) is a valuable technique for determin-ing solid-state information, such as elemental composi-tions, coordinations, valences, and spin states of atoms.

rt32

One of the benefits of EELS as an analytical technique isits ability to probe materials at high-spatial resolution,which facilitates the study of multiphase systems and evenallows one to probe bonding at interfaces (Bruley et al.,1994) or directly identify single atomic columns (Brown-ing et al., 1993). Parallel detector systems have recentlybecome available for EELS (PEELS), and this has greatlyincreased the ease of ELNES collection and vastly ex-tended the range of materials that may be studied. Underthe correct experimental conditions, the ELNES is theresult of dipole-allowed transitions from atomic core-lev-el states to unoccupied states above the Fermi level. TheELNES often exhibits a shape that is characteristic of thearrangement and types of atoms surrounding the excitedatom. This "fingerprint" of the local coordination arisesin cases where there is a localized electronic interactionof the central excited atom with the atoms in the nearest-neighbor coordination sphere. The experimental basis forthe ELNES coordination fingerprint is that the basic edgeshape for a particular atom in a specific coordinationshould be similar regardless of the material in which theatom is bonded.

Several EELS and XAS studies have demonstrated thecoordination fingerprint technique: Si Zr' edge in the SiOounit (McComb et al., l99l; Li eI al., 1993; Garvie et al.,1994); the C K edge in CO3- (Garvie et al., 1994, 1995);the P, S, and Cl Ir., edges in POI-, SO? , CIO;, respec-tively (Sekiyama et al., 19841, Hofer and Golob, 1987;Brydson et al., 1992); the Al Ir,, and K edges of ratAl andt6rAl (Hansen et al., 1992; Ildefonse et al., 1992, 1994);and the B K edge sf trtg t.6 IorB (Sauer et al., 1993). Theconcept of the coordination fingerprint must be used withcare because the exact symmetry of the ions surroundingthe excited atom can have a large efect on the shape ofthe ELNES, as has been demonstrated for the Si Ir., edgein both zircon (McComb et al., 1992) and forsterite (Mc-Comb et al., l99l). In zircon and forsterite the large an-gular variance of the SiOo tetrahedra causes the Si I',,ELNES to exhibit edge shapes different from those inminerals such as a-qtnrtz or talc, in which the angularvariance of the SiO4 tetrahedron is small.

B is present in many minerals, bonded almost exclu-sively to O or OH anions and only rarely to other anionssuch as F. In minerals, B is present in the two anions, thetetrahedral BOi- and the planar triangular BOI- groups.The O may be partially or totally replaced by OH anions.B triangles and tetrahedra join in a multitude of ways toform many polynuclear anions, and these polyanions inturn can polymerize to form infinite sheets, chains, andthree-dimensional networks (Christ ar'd Clark, 1977;Farmer, 1982; Heller, 1986). In the borosilicates, B sub-stitutes for Si. This is exemplified by datolite, in whichSiOo tetrahedra are replaced by BOr(OH) tetrahedra.

In this study, we investigate the B K edges of mineralscontaining planar t3lB and IalB. We demonstrate that thecoordination of the B in the minerals is unequivocallyidentified from the ELNES observed at the B K edges.Where mixed B-O coordinations are present in the same

I 1 3 3

mineral, we demonstrate that the I3rB/l4lB ratio can bedetermined from the B K ELNES. In several of the ma-terials the B K edge exhibits structure related to the dis-tortion of the B-O polyhedron.

Mltnnurs AND ExPERTMENTAL DETATLS

Minerals containing either t3tB or talB were studied (Ta-ble l), as were minerals with both ptB and IarB. For com-parison, crystalline B and several B-containing materials(Table l) were studied to illustrate some of the other pos-

sible B K-edge shapes. The minerals used in this studycome from a variety ofsources and have been identifiedby powder X-ray difraction. All minerals, except for du-mortierite and sussexite, were single crystals. Of the non-minerals, cubic BN (c-BN) was studied as single crystals,whereas the other compounds were either chips (crystal-line B) or high-purity powders.

Samples for PEELS analysis were prepared by crushingselected grains in acetone. A drop of the finely dividedmaterial, suspended in acetone, was then dried on a lacycarbon film supported on a Cu TEM grid' Data were thenacquired from thin areas overhanging the holes in the Iacycarbon film.

PEELS spectra were recorded on a VG HB5 STEM.This microscope is equipped with a cold field-emissiongun (FEG) and postspecimen lenses and operated at anaccelerating voltage of 100 kV with a probe semiangle ofI I mrad and a collection angle of 12.5 mrad. The exactacquisition parameters for the spectra depended on thestability of the sample under the electron beam. In gen-

eral, data were recorded from areas ranging from 130 x

100 nm'z to 32 x 25 nm2, using a probe current of ca. 0.5nA with recording times of ca. 2-4 s.

Great care must be taken when recording the B K-edgespectra of lalB-containing minerals, especially those thatare hydrated, because these minerals are susceptible toelectron-beam damage. This damage is a result of the easewith which IalB converts to t3tB by electron-beam irradi-ation. Sauer et al. (1993) studied the radiation damage ofthe mixed BO;- and BO;-containing minerals colemaniteand howlite. The initial B K ELNES of these mineralswere consistent with their expected t31B/t4rB ratio. Pro-longed electron-beam irradiation of these borates pro-

duced B K ELNES consistent with t3tB only.The susceptibility to electron-beam damage of lalB-con-

taining minerals studied in this work varied enormously'Anhydrous minerals such as sinhalite and rhodizite werebeam stable even after prolonged irradiation. Preobra-zhenskite, and especially tunellite, rapidly suffered elec-tron-beam-induced effects accompanied by the growth ofthe I31B r* peak. To reduce the effects of beam damage,the data were recorded with a defocused probe. The beam-sensitive minerals, such as tunellite, required special pro-

cedures for the acquisition ofPEELS spectra. These pro-

cedures involved the use of short integration times (<2

s), a defocused probe, and continuous movement of thescanned area over the sample during acquisition. Themajority of the B-containing nonminerals, with the ex-

GARVIE ET AL.: PEELS STUDY OF THE B K EDGE OF MINERALS

rt34 GARVIE ET AL.: PEELS STUDY OF THE B K EDGE OF MINERALS

Trsle 1, Experimental materials

Mineral name Structural formula Reference'

DumortieriteHambergiteLudwigiteSussexiteGaudefroyite

DatoliteDanburiteAxiniteSinhaliteRhodizite

ChambersiteBoraciteTunellitePreobrazhenskite

B-rhombohedral BChromium borideGlassy boric oxideHexagonal boron nitrideCubic boron nitride

IsrB-bearing minerals(Al, Fe),O.[(BO"[SiO).]Be,(OHXB03)(Mg,Fe,+ )rFe3+O"(BO.)Mn,(OH)[8,O3(OH)]Ca.Mn"O.(COr)(BO")"

r1tB-bearing mineralsCa4[Bi(SiOl)4(OH)4]CalBSiO4l,(Ca,Mn, Fe,+ )sAl,ISi,06][Si,O6BOs(OH)]MgAl(BO4)(Cs,K)Be.AloB,, O*(OH),

PtB- and l.lB-bearing mineralsMnsClIB?O13]Mg"Cl[8,O,"]Srl86O,(OH),1.3H,OMg3[B11O1.(OH)6] HrO

B-containing nonminerals

Moore and Araki, 1978zachariasen et al., 1963Bonazzi and Menchetti. 1989Hoffmann and Armbruster, 1995Garvie et al., in preparation

Foit et al., 1973Phillips et al., 1974Takeuchi et al., 1974Hayward et al., 1994Pring et al., 1986

Honea and Beck, 1962Dowty and Clark, 1973Clark, 1964Burns and Hawthorne, 1994

Hoard et al., 1970Guy and Waz, 1976Prabakar et al.. 1990Pease,1950Solozhenko et al., 1990

0-Bt*CrB2B"o.h-BNC-BN

- Structural reference where appropriate. No structure refinement was conducted for chambersite.

ception of glassy BrO., which was beam sensitive, weresufficiently beam stable to allow the use of integrationtimes of up to l0 s.

The following data-processing steps were performed onthe spectra, as recorded: the dark current was subtracted;a background of the form AE-. was subtracted from be-neath the core-loss edges; and the effect ofthe tailing ofthe zeroJoss peak was deconvoluted from the core-lossedge using the Fourier-ratio technique. The effects ofmultiple inelastic scattering were not deconvoluted fromthe coreJoss edges because such scattering typically af-fects the structures approximately 20 eV above the edgeonset, i.e., above the region presented for qualitativecomparison of ELNES. For a more detailed discussion ofEELS data processing, see Egerton (1986).

Rrsur,rsMinerals

The B K edges of the minerals that contain single-co-ordinated B are illustrated in Figures I and2,which showminerals containing t3lB and talB, respectively. Large dif-ferences between the ELNES of the two B coordinationsare apparent. Minerals with t3rB have ELNES dominatedby an intense sharp peak at ca. 194 eV (peak A), calledthe zr* peak, followed by a broader peak with a maximumat ca.203 eV (peak C), called the o* peak. The reasonsfor the use of these names will be given in the next sec-tion. Minerals that contain only t4lB have an ELNESdominated by an initial sharp rise in intensity with amaximum at ca. 199 eV (peak A). Table 2 gives the en-ergies ofthe principal features for selected minerals.

None of the minerals in Figure I has been studied pre-

viously by PEELS, although our spectra are similar inshape to the spectra from other materials with t3lB (Ishi-guro et al., 1982; Schwarz et al., 1983; Hallmeier et al.,l98l; Esposto et al., l99l1. Sauer et al., 1993). There areno major differences between the shapes of the r* peakfrom the minerals in Figure l, although the width at halfheight of this peak varies from 0.6 eV (dumortierite) to1.5 eV (gaudefroyite). The structure of the o* region var-ies considerably among the minerals studied. Dumorti-erite and sussexite exhibit a broad structureless peak witha maximum centered at ca. 204 eV. For dumortierite, thelack of structure in this region may be a result of the poorcounting statistics caused by its low B content, which istypically in the range l-6 wto/o BrOr. The o* regions ofludwigite, hambergite, and gaudefroyite exhibit maximaat several energy losses (Table 2). The shape of the o*region of hambergite is different from that of previouslydescribed t3tB-containing minerals because the intensityrises greatly to a sharp peak at 200.6 eV before graduallydropping. The height of the zr* peak is significantly moreintense than the o* peak in dumortierite, ludwigite, sus-sexite, and hambergite. The ratio of the peak heights r*/o* varies with 1.2 in gaudefroyite, 2.8 in dumortierite,3.4 in sussexite, 4.1 in hambergite, and, 4.9 in ludwigite.Although the intensity of the zr* peak relative to the o*region is often sensitive to crystal-orientation effects, es-pecially in the case of hexagonal BN (h-BN), such effectsalone cannot account for the large differences exhibitedbetween the spectra of hambergite and gaudefroyite.

Of the spectra illustrated in Figure 2, only rhodizite hasbeen previously studied by PEELS @rydson et al., 1988c).Our B K-edge spectrum from rhodizite exhibits the samestructure as the spectrum from the earlier PEELS study.

190 195 200 205 210 215

Energy Loss (eV)

Fig. l. Experimental PEELS B K-edge spectra for t3rB-con-taining minerals. The spectra presented here and in the followingfigures are shown on an arbitrary intensity scale. The labels inthis figure and subsequent figures refer to major or importantfeatures described in the text.

All the minerals exhibit peaks A and B. After peak B thespectra differ in the number of peaks visible, but, exceptfor the sinhalite spectrum, these additional peaks are broadand weak. Rhodizite, danburite, and axinite exhibit a weakpeak at ca. 194 eV, which, as we saw above, is indicativeof r3iB. This weak peak is caused by electron-beam dam-age during data acquisition. Sinhalite differs from the restofthe spectra illustrated in Figure 2 by exhibiting an ad-ditional distinct peak, A', below peak A. As we shall seelater, the intensity ofpeak A' varies greatly depending onthe orientation of the electron beam with respect to theprincipal axes of the crystal. The poor quality ofthe spec-trum from axinite is a result of the low B content of thismineral.

Nonminerals

The B K-edge spectra of a variety of nonminerals areshown in Figure 3, and the energies of the major featuresare tabulated in Table 2. Elemental B occurs as severalallotropes, the most stable of which is the B-rhombo-hedral form (0-B,or) containing 105 atoms in the unit cell(Hoard et al., 1970). The B Kedge of B-B,0, is similar in

I 135

190 195 200 205 210 215

EnergY Loss (eV)

Fig. 2. Experimental PEELS B K-edge spectra for talB-con-taining minerals. The small prepeak observed at ca. 194 eV inrhodizite, danburite, and axinite is due to electron-beam dam-age.

shape to published data (Esposto et al., l99l) and to theKedges of elemental Li and Be (Garvie, unpublished data).The initial sharp peak at 190.0 eV is followed by a broad-er maximum at ca. 196 eV. The initial sharp peak isbroadened on the low-energy side and can be modeled as

TreLE 2. Energies (eV) of the principal features of selected ma-terials illustrated in Figures 1-3

I3lB-containingmaterials

Peaks


D'

Gaudefroyite 193.7 197.0Hambergite 194.4 -Ludwigite 194.0 198.4Glassy-B,O3 194.1 197.5h-BN 192.1 194.7

203.4 - 207201.1 -

200.0 203.1 - 205.6 -202.7 -198.3 199.5 204.4 207

199.5 203 209200.3 203200.5 -200199.8 203.9 209.2

215.3

t41B-containingmaterials

SinhaliteRhodiziteDatoliteDanburitec-BN

197.1 197.9198.0197.7198.1

194.2 197.0

214213.3212

214.O

Note: Efiot in EELS values is +0.2 eV.

l 1 3 6

190 200 210 220 230

Energy Loss (eV)

Fig. 3. Experimental PEELS B K-edge spectra for nonmrner-als; B-B,0, : B-rhombohedral B, c-BN : cubic-BN. and h-BN:hexagonal-BN.

two peaks separated by 1.44 eV. These two peaks areprobably the result ofcore-hole effects. As expected, theB K-edge onset in B-B,0, is lower in energy than that foundin plB- and tarB-containing materials, in which B possessesan effective positive charge.

Chromium diboride has a B KELNES similar in shapeto B, although the first peak in CrB, is more pronouncedand at a lower energy, 188.6 eV. The lower energy of thefirst peak in the diboride, relative to that in elemental B,may be rationalized by consideration of the charge trans-fer from the Cr to the B. The initial peak at 188.6 eV inthe diboride represents transitions to a r* antibondingstate resulting from the presence of sp, bonding withinthe structure, whereas the following structure, centered atca. 200 eV, may be associated with transitions to o* an-tibonding orbitals. In the diboride, B is present in layersof I3rB, each B atom being sp, bonded to three other Batoms. The B planes are sandwiched between layers ofCr atoms.

Boron nitride occurs in two principal polymorphs: hex-agonal (h-BI.[) and cubic (c-BN). Hexagonal BN is a lay-ered material with B atoms bonded to three N atomswithin the plane. Cubic BN has a diamondlike structurewith each B atom bonded to four N atoms. Hexagonal


BN (containing tnB) and cubic BN (containing t4rB) ex-hibit B K-edge spectra that are qualitatively similar tothose from the minerals. This resemblance is rationalizedby the similar electronegativities of the N and O atomsas well as the similar local B coordination. Previous mea-surements on both the B K and N K edges of severalpolymorphs of BN (Hosoi et al., 1982; Terminello et al.,1994) have already identified the zr* and o* features inthe spectra of h-BN and the o* feature in the spectra ofc-BN that arise from their differing local coordinations.

Boron oxide (BrOr) occurs in two principal poly-morphs: glassy and crystalline. The glassy form is mostcommon. The glassy structure is still debated, althoughthe consensus is that the B is predominantly threefold-coordinated to O, comprising a random network of bo-roxol rings (Mozzi and Warren, 1970; Prabakar et al.,1990). This coordination is confirmed by its B K edge,which is similar in shape to the B K edges of mineralscontaining BO, units. Crystalline varieties of B oxide aredifficult to prepare (Prewitt and Shannon, 1968; Gurr etal., 1970) and have not been studied in this work.

Quantification of the I3lB/l4tB ratio

Borates containing both BO, and BOo groups are com-mon. The four mixed t3lB- and lalB-containing mineralsstudied have the following t3lB/t4lB ratios: tunellite : I(3A + 3T); preobrazhenskire:0.64 (4A + 7T); and bor-acite and chambersite :0.167 (lA + 6T). The terms inparentheses refer to the number of triangles, A, and tet-rahedra, T, in the fundamental borate building block. InFigure 4, the spectra from the mixed-coordination B min-erals are arranged in order ofdecreasing height ofthe firstzr* peak, at 194 eY (characteristic oft,rB), relative to theintensity ofthe second peak at ca. 199 eV. This orderingfrom tunellite to boracite directly follows the decrease inthe t3lB/la1B ratio determined from crystallographic data.As the t3lB/t41B ratio decreases, the B K edge more closelyresembles that of tolB, and in boracite and chambersite,the B K ELNES is similar in shape to that of rhodizite.The present data suggest that the B KELNES of the mixed-coordination B minerals is just the superposition of theB K edges of B from the respective BO, and BO, units,which would follow if the characteristic final states forexcitation at both ptB and t41B sites were sufficiently lo-calized as to prevent significant interaction.

Analysis of the B K edges in minerals containing botht3tB and ratB should therefore allow the t31B/t41B ratio to bedetermined. This is in agreement with the findings of Saueret al. (1993), who quantified the fraction of B atoms int3lB sites, /o, in two hydroxyborates by assuming that thisfraction was proportional to the intensity in the r* peak,,I(zr*), relative to the total integrated intensity in a suffi-ciently large energy window above the edge onset, J(AE).A similar analysis was employed to determine the frac-tion of sp'z bonding relative to sp3 bonding in carbona-ceous materials (Katrinak et al., 1992). We used a pro-cedure similar to that used by Sauer et al. (1993), namely

GARVIE ET AL.: PEELS STUDY OF THE B K EDGE OF MINERALS tt37

TABLE 3. Fraction of l3rB-coordinated sites, fo, derived from thecrystal structure (theoretical) and from analysis of theB K ELI'IES using the procedure described in the text

Mineral fo(theoretical) f"(experimental)

0.420.350.120 . 1 2

190 195 200 205 210 215Energy Loss (eV)

Fig. 4. Experimental PEELS B K-edge spectra from mineralscontaining both t31B and talB. The spectra have been arrangedfrom top to bottom with respect to decrease in the zr*/o* heightratio, thus reflecting a decrease in the t3rB/t41B ratio.

. _ lJ(tr*\/J(AE)lr a

lJQr*)/J(a,E)1,6'

Here we take,f(zr*) to be the intensity in a 2.5 eV windowover the zr* peak, and "I(AE) is the intensity in an energywindow of width 14.6 eV commencing 1.4 eV above thezr* peak. The chosen energy window for ,/(r*) measuredthe intensity of the whole r* peak but excluded the in-tensity in the o* region. The energy window AE for,I(AE)was chosen following two guidelines: first, it must be wideenough so as not to be governed by changes in the ELNES;and second, the window must not extend too far abovethe edge onset where multiple scattering effects are mostdominant. The denominator in the above equation refersto the ratio of spectral intensities in the B K edge of areference material assumed to contain 1000/0 t3lB. We takelJGr*)/J(LE)l*, to be 0.354, which represents an averagevalue obtained from the t3rB minerals ludwigite (0.353),gaudefroyite (0.25 6), hambergite (0.40 I ), sussexite (0. 3 9 5),and glassy BrO3 (0.363).

The results of the above procedure for determining /lin the mixed-coordination B minerals are presented inTable 3 together with the theoretical values based on thecrystal structures. There is a good match between the the-oretical and experimental values for preobrazhenskite,chambersite, and boracite, whereas the fraction of talB fortunellite was overestimated. The result from tunellite willbe discussed in more detail in the next section.

TunellitePreobrazhenskiteChambersiteBoracite

It is difficult to estimate the accuracy of the I3lB/rarBratios determined from the B Kedge using EELS (or XAS).A more systematic study is needed to examine a largerset of materials with a wide range of I31B/l4lB values. Wesee that the overall shapes of the B K edges are charac-teristic of B coordination, implying that the ELNES aredominated by the effects of the nearest-neighbor atoms,with minor contributions from atoms surrounding theBO3- or BOi anions. This insensitivity of the B K edgeto the structure outside the immediate coordination sphereis likely to make the determination of the polyanion dif-ficult or impossible, e.g., it is difficult to differentiate be-tween the isolated BOI- unit and the triborate polyanion

[B3O3(OH)6]3-, both of which contain only t3rB. Likewise,the distinction between B coordinated to O anions and Bcoordinated to OH is likely to be difrcult because of thesimilar scattering powers of both anions.

Use of EELS (or XAS) to determine the coordination-specific site occupancy is likely to find wide applicationin the study ofborate glasses, where the coordination andcoordination ratio vary with B content (Prabakar et al.,1990; MiiLller et al., 1993). Determination of coordinationand coordination ratios of B in glasses is typically re-served for "B MAS NMR, although PEELS conductedin a TEM offers the unequaled advantage of being ableto analyze nanometer-sized regions.

Orientation effects

The orientation dependence of core-loss edges wasstudied in anisotropic materials by EELS and XAS. Ori-entation dependence arises because of the directionalityofthe transitions to the unoccupied states into which thebound electron is excited. In XAS, the polarization vectorselects the direction probed within the crystal, whereas inEELS, the scattering vector, q, selects the direction. InEELS, the direction of q is determined by the differencebetween the wave vectors of the incident and the scat-tered high-energy electrons. Ifthe incident electrons havean angular divergence that is much less than the charac-teristic angle of the inelastic scattering process, 0., andthe angular range over which the scattered electrons arecollected is also much less than dr, then the direction ofq is well defined. Ifthe angle ofscattering is zero, then qis parallel to the incident beam, but if the scattering angleis much greater than dr, q is approximately perpendicularto the incident beam. This allows the ELNES to be stud-ied as a function of scattering angle and crystal orienta-tion (Leapman et al., 1983). Many polarized studies have

0.500.360.140.14

I 138

180 190 200 210 220 230Energy Loss (eV)

Fig. 5. Spectra from h-BN, sinhalite, and danburite illustrat-ing orientation effects. The B K edges for h-BN are for the elec-tron beam parallel (e- ll c) and perpendicular (e- I c) to the caxis. The spectra for h-BN have been normalized at the r* peak.For sinhalite and danburite, orientation of the electron beamwith respect to the axes of the BO4 tetrahedra is not known. Anintermediate spectrum for sinhalite is illustrated in Fig. 2. Thesmall prepeak atca.194 eV in sinhalite and danburite is causedby electron-beam damage.

been conducted on graphite because of its anisotropicstmcture (Leapman et al., 1983; Cheung, 1985; Rosen-berg et al., 1986). The basic structure of graphite consistsof o bonding between the C atoms perpendicular to the caxis and the zr bonds parallel to the c axis. When thepolarization vector in XAS, or q in EELS, is parallel tothe c axis, states ofr character are greatly enhanced overthose with o character, whereas if the vector is perpen-dicular to the c axis, the reverse is true. The EELS ex-periments reported here differ from the above situationbecause the sum ofthe angular divergence ofthe incidentelectrons and the angular range over which the scatteredelectrons are collected is much less than dr. The electronbeam is focused into a probe with the halfangle ofcon-vergence of I I mrad, and the scattered electrons are col-lected over a cone with the half angle of 12.5 mrad havinga mean scattering angle of zero but a range of scatteringangles that exceeds 0.by an order of magnitude. Hence,scattering events over a wide range ofq vectors are col-


lected, and those with q perpendicular to the electronbeam are dominant. In this situation, an electron probeincident parallel to the c axis (e ll c axis) of graphiteresults in the detection of more excitations into states ofd character than into states ofzr character, whereas a probeincident perpendicular to the c axis (e- I c axis) resultsin detection of similar numbers of excitations into statesof each character. Although orientation effects were notexplicitly studied in this work, they were visible in thespectra recorded from several materials and can be usedto explain the incorrect I31B/I41B ratio determined for tu-nellite.

Figure 5 illustrates orientation-dependent spectra fromh-BN, sinhalite, and danburite. Hexagonal BN has astructure similar to graphite, with o bonds between the Band N atoms (i.e., perpendicular to the c axis) and r or-bitals parallel to the c axis. In the case of h-BN, Figure 5shows that the zr* peak is significantly enhanced relativeto the o* peaks when e- I c axis in comparison with thecase when e- ll c axis. Sinhalite and danburite exhibitedarange of ELNES shapes (the main differences are labeledA-C), although the changes were subtle in danburite. TheB K ELNES of sinhalite can be explained in light of thedistorted BOo tetrahedra, which have three similar andone long B-O bond length (1.442 L, 2 , 1.483 A, and1.586 A; Hayward et al., 1994). Thus, the orientation-dependent near-edge structure depends on whether theelectron beam is parallel or perpendicular to the long B-Obond. The differences in the B K edge of danburite aremore difrcult to explain. In danburite, B is present inseparate BrO, groups with the B-Ob,-B bond (Oo.: bridg-ing O) roughly parallel to the c axis, and the B-Ob. dis-tance is shorter than the remaining B-O distances (Phil-lips et al., 1974).It is probable that the bonds between Band the bridging O have slightly different characteristicsthan the bonds between B and the nonbridging O atoms.Thus, we would expect differencos in the B K ELNESdepending on whether the incident electron beam is par-allel or perpendicular to the c axis.

Turning now to the compounds containing the planarBOI- group, if the BO3- triangles are arranged in severalorientations within a crystal, as in glassy BrOr, gaude-froyite, and hambergite, then we would not expect to seesignificant orientation effects. On the other hand, if theplanes of the triangles are all aligned, then such orienta-tion effects may be observed. For example, if the incidentelectron beam is perpendicular to the BOI- triangle, thenthe o* region is enhanced in relation to the zr* region(assuming that the sum of the convergence and scatteringangles is >> dr, as in the experimental conditions used torecord the spectra here). However, determination of thet31B/r4rB ratio depends on the value of "I(z*)/J(A,E) fromthe I3rB component being approximately equal to V(tr*)/J(AE)l*r. If the electron beam is perpendicular to theBOI- triangles, then J(zr*)/J(AE) decreases as a result ofthe increase of the o* region from the t3lB component,with a corresponding decrease in f, and overestimationof r4tB.

We are now in a position to explain the anomalously

TABLE 4. Unoccupied orbitals calculated by the SCF-LCAO MOmethod on a BO8- anion (D3h symmetry) using theGaussian 90 code of Frisch (1990) and STO-3G basissets

I 139

TABLE 5. Unoccupied orbitals calculated by the SCF-LCAO MOmethod on a BO!- anion (Td symmetry) using theGaussian 90 code of Frisch (1990) and STO-3G basissets

Relative RelativeOrbital Atomic character energyt energy-.


Orbital Atomic characterRelative Relativeenergy- energy'.

B 2p, + O 2p 211.0B 2s + O 2s + O 2p 221.8B2p, + B 2pr+ O 2s + O2p 227.7

Nofej The B-O bond length is 1.31 A.t Energy relative to the B 1s level at -169.4 eV.

-. Energy ot the first unoccupied MO.

low value of t determined for tunellite from its B Kedge.In tunellite, the BO3- triangles are condensed into an in-finite sheet parallel to the (100) plane. Another importantfeature of tunellite is its perfect {100} cleavage. Dryingthe finely divided mineral in suspension onto the lacycarbon film resulted in alignment of the thin mineral flakesparallel to the carbon film and thus perpendicular to theelectron beam. Therefore the B K edges of tunellite wereinvariably recorded with the electron beam perpendicularto (100), and so J(tr*)/J(AE), from ther3lB component oftunellite, would be less than lJQr*)/J(AE)L"r, giving riseto the anomalously low value of fo and overestimationof the I41B component.

INrnnpnnurroN oF THE B I( ELNES

ELNES may be interpreted and modeled by a varietyof techniques (Brydson, l99l; Rez et al., 1991, 1995;Rez, 1992). In terms of single-electron theory, the tran-sitions that result in the observed ELNES simply reflectthe unoccupied electronic structure ofthe solid projectedonto the particular atomic site in question. Interpretationof the edges is most easily achieved by comparison withthe results of band structure, multiple scattering (MS), ormolecular orbital (MO) calculations. These one-electroncalculations neglect the interactions between the core hole,created during the excitation process, and the resultantfinal state. Such interactions may have appreciable effectsupon the shape of the ELNES and generally result in adistortion of the electronic structure close to the onset ofthe conduction band. Core-hole effects are probably mostdominant in low atomic number materials, which havefew valence electrons with which to shield the hole leftby the excitation process (Brydson et al., 1988a; Weijs etal., 1990). For example, the large prepeak to the B Kedgeof B-B,0, is thought to be caused by the interaction of thecore hole with the conduction-band final states, resultingin the formation of a separate unoccupied energy levelbelow the conduction-band onset (P. Rez, personal com-munication).

The spectra of the minerals presented here are com-pared with our own SCF-LCAO (self-consistent field lin-ear combination of atomic orbitals) calculations onBOI- and BOI- anions using the Gaussian 90 code ofFrisch (1990) and STO-3G basis sets, the correspondingMO calculations of Vaughan and Tossell (1973), and the

221.2223.1

Notei The B-O bond length is 1.46 A.- Energy relative to the B 1s level at -153.7 eV.

-- Energy of the first unoccupied MO.

results of cluster-based MS calculations using theICXANES program (Vvedensky et al., 1986). MO theorywas used to interpret qualitatively ELNES by comparingthe results of the calculations with the features observedwithin the first few electron volts of the ionization edgeonset.

Molecular orbital theory

The symmetry of the O or OH ligands surrounding theB atoms determines the energies and degeneracy of theunoccupied molecular orbitals. In icosahedral (In), octa-hedral (O), and tetrahedral (Tj symmetries the B p or-bitals are triply degenerate and belong to a MO of t2 char-acter. Reduction of the degeneracy occurs in hexagonal(Du), tetragonal (Do), and triangular (DrJ symmetries,with the p" orbital being of all character and the p- andp, orbitals being of e' character. Distortions in the sym-metry of the ligands from the ideal Tu or Dro symmetryare expected to cause degenerate levels to lose their de-generacy, and this would be manifested in the ELNESeither as a broadening or splitting offeatures. The differ-ences between the B K ELNES of ptB and talB may beexplained by the different unoccupied MOs of the twoanions. The relative eneryies of these unoccupied MOsare given in Tables 4 and 5. In BOI , the lowest unoc-cupied MO (LUMO) is the ai (zr*) antibonding MO, whichis derived from B 2p electrons, followed by the a', (o*)and the doubly degenerate e' (o*) MOs, which are pre-dominantly B 2s and B 2p in character, respectively(Vaughan and Tossell, 1973). Our own MO calculations,as well as those of Vaughan and Tossell (1973), indicatethat there is considerable mixing in the ali state betweenthe B and O 2p electrons. Addition of another O atomand rearrangement to form the BOI- anion (with B in Tosite symmetry) significantly change the MO structure. TheLUMO becomes the triply degenerate t, (ox) MO derivedfrom B 2p states, closely followed by the a' (o*) MO,which is predominantly B 2s in character.

The B Kedges of minerals that contain only BO, groupsall have two main structures: an initial intense peak (peakA, Fig. l) followed by a broader peak that varies in in-tensity and shape. It is now fairly well agreed that the firstsharp peak ofthe B K edge ofh-BN (Barth et al., 1980)and BFr- (Ishiguro etal.,19821' Schwarz et al., 1983) andBO.-containing minerals (Sauer et al., 1993) is the resultof transitions to an unoccupied r MO. For B(OH,)

O2

aie'

0.010.816.7

0.01 . 9

B2p + O2pB 2 s + O 2 p

I 140

TABLE 6. Unoccupied orbitals calculated by the SCF-LCAO MOmethod on a BOI anion (C" symmetry) using theGaussian 90 code of Frisch (1990) and STO-3G basisSEIS

Orbital Atomic characterRelative Relativeenergy' energy-'


a'

B 2p, + B 2W + O 2p 217.7 0.0B2s + B 2p ,+ B2py+ O2s + O2p 220.9 2 .6B 2p. + O 2p 220.7 3.0B 2 s + 8 2 p , + 8 2 p " + O 2 s + O 2 p 2 2 4 . 1 6 . 4

Nofe.' The B-O bond lengths and angles for sinhalite are from Fang andNewnham, 1965.

- Energy relative to the calculated energy of the B 1s level at -154.1eV.

'. Energy of the first unoccupied MO.

and BFr, this first peak is described as aB 2ptr nonbond-ing or weakly antibonding orbital of a, symmetry (Tos-sell, 1986). The broad peak C, centered 7-10 eV higherin energy, then corresponds to transitions from a ls to ane' (o*) MO, which is dominantly B 2p in character. Theenergy separation between these features in the experi-mental spectra is in reasonable agreement with the the-oretical separation derived from our MO calculations onBOI- (Table 4). Several of the spectra show a weak peak,B, between the zr* and a* peaks, and this could be inter-preted as a dipole-forbidden ls - a\ (o*) transition. Atransition to this B 2s-like state would explain the weak-ness ofthis peak. An alternative explanation for the pres-ence of this feature, and some of the other features suchas peaks B' and D from ludwigite, involves the influenceof more distant atoms (i.e., next-nearest neighbor andbeyond). Such effects are more easily handled using amultiple scattering approach and are discussed later.

The zr* peak present in the I3lB K edge corresponds toa transition to a bound state lying below the conduction-band onsot al 195.7 + 0.5 eV (Brydson et a1., 1992). Thelatter figure was derived from XPS measurements of theenergy separation between the B ls peak and the top ofthe valence band together with the magnitude of the bandgap obtained directly from EELS low-loss measurements.This suggests that the position, and presumably intensity,of the r* feature is strongly afected by interaction withthe ls core hole left by the excitation process.

The high-energy resolution of the HB5 STEM facilityallows the true width of the z* peak from r3rB to be de-termined, provided that the width of the zrx peak is great-er than the width of the zeroloss peak. The width of thezr* peak varies from ca. 0.6 eV in dumortierite, ludwigite,and hambergite, to ca. 0.9 eV in sussexite, to 1.5 eV ingaudefroyite. The lower value of 0.6 eV is limited by thewidth of the zero-loss peak, which always exhibited a full-width half-maximum (FWHM) of 0.45-0.54 eV duringthe acquisition of the B Kedges. Barth et al. (1980) mea-sured a FWHM of 0.4 eV for the a-* peak from h-BN.Our measurement for the width of the r* peak from h-BNof 0.59 eV is consistent with the width of the zeroloss

5 1 0 15 20Relative Energy Loss (eV)

Fig.6. Comparison of the B Kedge (thin line) with the cor-responding anion K edge (thick line) for h-BN, hambergite, andglassy BrOr. The z* peaks from the B Kedges are aligned withthe first peak from the anion K edges, and the spectra are pre-sented on a relative energy scale.

peak. As mentioned above, distortions of the environ-ment around B in the BO, group are expected to causebroadening of the ELNES features. In dumortierite, lud-wigite, hambergite, and gaudefroyite, B sits in regular t3lBsites with three similar B-O bond lengths (Zachariasen etal.. 1963: Moore and Araki. 1978:. Bonazzi and Men-chetti, 1989; Garvie et al., in preparation). The largerwidth of the r* peak from sussexite is probably causedby the distorted BOI- triangle (Hoffrnann and Armbrus-ter, 1995) with B-O bond lengths of 1.370-1.408 A, whichdeviate significantly from the average value of 1.365 Afound for the BO, anion. The cause of the larger zr* peakfrom gaudefroyite is uncertain because the BO, triangleis regular, although a contributing factor may be the twoinequivalent B-O bonds. In gaudefroyite, two of the Oatoms are twofold coordinated, whereas the third O atomis threefold coordinated-

Although the spectra for t4rB all show considerable dif-ferences in their near-edge structure, they are character-ized by a sudden increase in the intensily at ca. 196-198eV. The MO picture for BOi (Vaughan and Tossell, 1973)shows one dipole-accessible MO, t, (o*) of B 2p character.Thus, the first peak (peak A, Fig. 2) may be assigned to

-5

tt41

experiment

7 shells

1 shel l

0 1 0 2 0 3 0 4 0Relative Energy Loss (eV)

Fig. 8. Comparison of the theoretical multiple scattering cal-culations for h-BN with an experimental B Kedge. The multiplescattering calculations for one- and seven-shell clusters are illus-trated.

zr* MO (Sauer et al., 1993). In Figure 6 we compare theB and N K edges of h-BN and the B and O K edges ofglassy BrO, and hambergite. Most of the fine structurevisible at the anion K edge is present at the cation K edge,thus attesting to the idea of hybridization between anionand cation electrons ofpJike character.

Multiple-scattering theory

Multiple-scattering (MS) theory is based on interfer-ence between the outgoing electron wave of the excitedelectron and the electron wave backscattered from thesurrounding atoms. MS calculations are performed in realspace using a cluster approach and, in the limit of largecluster size, are formally equivalent to electronic bandstructures calculated using Korringa-Kohn-Rostocker(KKR) theory. The scattering properties of the variousatoms in the cluster are described by phase shifts ob-tained by integration of the Schriidinger equation, assum-ing a muffin tin form for the crystalline potential. It isthis assumption that often limits the accuracy of the over-all MS calculation, especially in open structures, such asminerals, which possess substantial directionality (i.e.,covalence) in the bonding.

The cluster approach readily permits the investigationof the influence of different shells of neighboring atomson the overall ELNES and may be used to identifu the


30201 0Relative Energy Loss (eV)

Fig. 7. Theoretical multiple scattering calculations of the BK edge for a trigonal BO, cluster and a tetrahedral BOn cluster.The calculations have been conducted for a one-shell cluster.

transitions to the triply degenerate t, (o*) MO. The MOpicture for BOi- shows another unoccupied MO, the a,(o*), but because this is of predominantly B 2s character(Vaughan and Tossell, I 973), transitions from the I s corelevel to the a, (o*) MO are dipole forbidden and henceweak. In rhodizite, B is sited in regular tetrahedral sites(Pring et al., 1986) with only small angular and bondvariations from an ideal tetrahedron (T). Thus, the B KELNES of rhodizite can be described as an archetypal t41BK edge. Its edge is dominated by the intense o* peak, A,followed by several weaker, broader structures (Fig. 2,Table 2). All these peaks are present in sinhalite, but theyare more intense relative to the first o* peak. The extrastructures in sinhalite relative to the spectrum from rhod-izite result from the highly distorted BOo tetrahedron (an-gular variance 56.87',), with B-site symmetry C, (Hay-ward et al., 1994). This distortion causes the triplydegenerate t, (ox) MO to lose its degeneracy and splitinto two MOs, a" (p") and a doubly degenerate a' (p. andp"). The MO structure of such a distorted BOo tetrahe-dron is tabulated in Table 6. Qualitatively, the observedexperimental peak separations agree well with the sepa-ration between the calculated MO levels.

Finally, because the states probed by the excited elec-tron are essentially hybridized states consisting ofboth B2p and O 2p character, one would expect to see analogousfeatures at the O K edge, which essentially probe O 2p-derived unoccupied states. Indeed, the O K ELNES ofborates and borosilicates containing PIB exh'b'ts a sharpfeature at ca. 532 eV corresponding to transitions to the

l t 4 2 GARVIE ET AL.: PEELS STUDY OF THE B K EDGE OF MINERALS

Similar results may be obtained for the polymorphs ofBN. Figures 8 and 9 show the results of MS calculationson h- and c-BN, respectively. Core-hole effects were againincluded. The clusters employed consisted of seven andeight shells, respectively, and in each case were >0.5 nmin radius in an attempt to reproduce all the observedfeatures. Although all details are not reproduced, theoverall agreement is satisfactory. Also in Figures 8 and9, we show calculations on smaller clusters consisting ofone shell of N atoms threefold- and fourfold-coordinatedto a central B atom. These calculations essentially con-firm the existence of a B K ELNES coordination finger-print. In the case of h-BN, they reproduce the sharp r*and o* features expected on the basis of MO theory,whereas for c-BN we observe only the broader o* feature.

Acxxowr,nocMENTs

We would like to thank the following for donating specimens: J. Faith-full (Huntarian Museum, Glasgow, U.K.) for the specimen of axinite(M830a); L. Greenbank (Cumbria, U.K.) for the boracite; G. Cressey(British Museum of Natural History, London) for the hambergite (BMl9l l,l); and A. Hodgkinson (Glasgow, U.K.) for the sinhalite gemstone.We would also like to thank the University of I-ondon Computer Centerfor help in the MO calculations, and the Science and Engineering ResearchCouncil for the Gatan 666 PEELS and for a postdoctoral research assis-tantship for L.A.J.G.

RrrnnnNcns cITEDBarth, J., Kunz, C., and Zmkina, T.M. (1980) Photoemission investiga-

tion of hexagonal BN: Band structure and atomic effect. Solid StateCommunications. 36. 453-456.

Bonazzi, P., and Menchetti, S. (1989) Contribution to the crystal chem-istry of the minerals of the ludwigite-vonsenite series. Neues JahrbuchfiiLr Mineralogie Monatshefte, 2, 69-83.

Browning, N.D., Chisholm, M F, and Pennycook, S.J. (1993) Atomic-resolution chemical analysis using a scanning transmission electron mi-croscope. Nature, 366, 143-146.

Bruley, J., Hoche, T., Kleebe, H.-J., and Riihle, M. (1994) Recent at-tempts to detect magnesium in a heavily doped sapphire bicrystal byspatially resolved electron energy-loss spectroscopy. Journal of theAmerican Ceramic Society, 77, 2273-2276.

Brydson, R. (1991) Interpretation ofnear-edge structure in the electronenergy-loss spectrum. Electron Microscopy Society of America Bulle-tin,21, 57-67.

Brydson, R., Bruley, J., and Thomas, J.M. (1988a) Further evidence forcore-hole effects in the near-edge structures of light-element K-edges.Chemical Physics L€tters, 149, 343-347.

Brydson, R., Vvedensky, D.D., Engel, W., Sauer, H., Williams, B.G.,Z,e|tler,E., and Thomas, J.M. (1988b) Chemical information from elec-tron-energy-loss near-edge structure: Core hole effects in the berylliumand boron K-edges in rhodizite. Journal of Physical Chemistry, 92,962-966.

Brydson, R., Williams,8.G., Engel, W., Lindner, Th., Muhler, M., Schlogl,R., Zeitler, E., and Thomas, J.M. (1988c) Electron energyloss spec-troscopy and the crystal chemistry ofrhodizite: Part 2. Near-edge struc-ture. Joumal ofthe Chemical Society, Faraday Transactions, 84, 631-646.

Brydson, R., Sauer, H., and Engel, W. (1992) Electron energy loss near-edge structure as an anal)'tical tool: The study of minerals. In M.M.Disko, C.C. Ahn, and B. Fultz, Eds., Transmission electron energy lossspectrometry in materials science, p. 131-154. The Minerals, Metalsand Materials Society, Warrendale, Illinois.

Burns, P.C., and Hawthorne, F.C. (l 994) Structure and hydrogen bondingin preobrazhenskite, a complex heteropolyhedral borate. CanadianMineralogist, 32, 387 -396.

Cheung, T.T.P. (1985) Orientation dependence of the carbon K edge in

0 1 0 2 0 3 0 4 0Relative Energy Loss (eV)

Fig. 9. Comparison of the theoretical MS calculations forc-BN with an experimental B K edge. The MS calculations forone- and eight-shell clusters are shown.

existence of coordination fingerprints (i.e., ELNES struc-tures predominantly determined by the arrangement andtype of atoms in the nearest-neighbor coordination shell).Furthermore, excited atom potentials can be included soas to simulate the effect of the core hole on near-edgefeatures (Brydson et al., 1988a, 1988b). The energy scaleof MS calculations is entirely relative, and the usual pro-cedure is to align the initial features present in both ex-periment and theory. Further details of the XANES cal-culation may be found in Vvedensky et al. (1986) andBrydson (1991).

To demonstrate the existence of specific coordinationflngerprints for I3lB and t4lB in borate and borosilicateminerals, we present the results of MS calculations onplanar trigonal BO, and tetrahedral BOo clusters in Figure7. Core-hole effects were included by use of the (Z + l)*approximation for the central B atom in the cluster. Oatoms and ions are particularly strong backscatterers, andtheir spatial arrangement in the nearest-neighbor coor-dination shell dominates the overall near-edge structureobtained. In the case of t3lB, we are able to model theobserved zr* and a* features common to the experimentalspectra of Figure l, whereas for talB we obtain only anisolated o* feature. It is now our belief that some of theadditional structures observed after the initial peak in theexperimental spectra shown in Figures I and2 arise frommore distant shells of atoms than the nearest-neighborshell, thus explaining some of the differences observedbetween talB-containing minerals.

graphite measured by reflection electron-energy-loss spectroscopy.Physical Review B, 31,4792-4797.

Christ, C.L., and Clark, J.R. (1977) A crystal-chemical classification ofborate structures with emphasis on hydrated borates. Physics andChemistry of Minerals, 2, 59-87 .

Clark, J.R. (1964) The crystal structure of tunellite, SrB.O,(OH), 3H,O.American Mineralogist, 49, I 549- 1 568.

Dowty, E., and Clark, J.R. (1973) Crystal-structure refinements for or-thorhombic boracite, MgrClBrO,r, and a trigonal, iron-rich analogue.Zeitschrift fiir Kristallographie, I 38, 64-99.

Egerton, R.F. (1986) Electron energy-loss spectroscopy in the electronmicroscope, 410 p. Plenum, New York.

Esposto, F.J., Aebi, P., Tyliszczak, T., Hitchcock, A.P., Kasrai, M., Boz-ek, J.D., Jackman, T.E., and Rolfe, SR. (1991) Boron K-shell spec-troscopy ofboron-doped silicon. Journal ofVacuum Science and Tech-nology, A9, 1663-1669.

Fang, J.H., and Newnham, R.E (1965) The crystal structure of sinhalite.Minerafogical Magezins, 35, 196-199.

Farmer, J.B. (1982) Metal borates. Advances in Inorganic Chemistry andRadiochemistry, 25, 187 -237

Foit, F.F., Jr., Phillips, M.W., and Gibbs, G Y. (1973) A refinement ofthe crystal structure of datolite, CaBSiO.(OH). American Mineralogisr,58 ,909 -914 .

Frisch, M.J. (1990) GAUSSIAN 90 revision I. Gaussian, Pittsburgh,Pennsylvania

Fuggle, J.C., and Inglesfield, J.8., Eds. (1992) Unoccupied electronic states:Fundamentals for XANES, EELS, IPS and BIS, 359 p. Springer-Verlag,New York.

Garvie, L.A.J. (1995) Parallel electron energyJoss spectroscopy ofboronin minerals. In Mineralogical Society of America Reviews in Miner-alogy, 33, in press.

Garvie, L.A.J., Craven, A.J., and Brydson, R. (1994) Use of electron-energy loss near-edge fine structure in the study of minerals. AmericanMineralogist, 79, 4ll-425.

Garvie, L.A.J., Buseck, P.R., and Craven, A.J. (1995) Electron-loss near-edge structure (ELNES) as a valence and site-specific probe. CanadianMineralogist, in press.

Gun, G.8., Montgomery, P.W., Knutson, C.D., and Gones, B.T. (1970)The crystal structure of trigonal diboron trioxide. Acta Crystalloga-phica, B26,906-915.

Guy, C.N., andUraz, A.A. (1976) The chromium-boron system. Journalof the Irss-Common Metals. 48. 199-203.

Hallmeier, K.-H., Szargan, R., Meisel, A., Hartmann, E., and Gluskin,E.S. (198 1) Investigation ofcore-excited quantum yield spectra ofhigh-symmetric boron compounds. Spectrochimica Acta, A37, 1049-1053.

Hansen, P.L., McComb, D.W., and Brydson, R. (1992) ELNES finger-prints of Al coordination in nesosilicates. Micron and Microscopia Acta,23,169-t70.

Hayvard, C.L., Angel, R.J., and Ross, N.L. (1994) The srructural rede-termination and crystal chemistry of sinhalite, MgAlBO". EuropeanJournal ofMineralogy, 6, 313-321.

Heller, G. (1986) A survey ofstructural types ofborates and polyborates.Topics in Current Chemistry, l3l, 39-98.

Hoard, J.L., Sullenger, D.B., Kennard, C.H.L., and Hughes, R.E. (1970)The structure analysis ofB-rhombohedral boron. Journal ofSolid StateChemistry, 1, 268-277.

Hofer, F., and Golob, P. (1987) New examples for near-edge fine struc-tures in electron energy loss spectroscopy. Ultramicroscopy, 21, 37 9 -384.

Hoffmann, C., and Armbruster, T. (1995) Crystal structue of a (001)twinned sussexite MnrBrO4(OH), from the Kalahari Manganese FieldSouth Africa. Schweizerische Mineralogische und Petrographische Mit-teilungen, 7 5, 123-133.

Honea, R.M., and Beck, F.R. (1962) Chambersite, a new mineral. Amer-ican Mineralogist, 47, 665-67 l.

Hosoi, J., Oikawa, T., and Inoue, M. (1982) Study of boron nitride byelectron energy loss spectroscopy. Journal ofElectron Spectroscopy andRelated Phenomena.2T . 243-254.

Ildefonse, P., Calas, G., Kirkpatrick, R.J., Montez, B., Flank, A.M., andl:garde, P. (1992) Local environment of aluminium in arnorphousalumino-silicates by XANES and MAS NMR. In Y.K. Kharaka andA.S. Maest, Eds., Proceedings ofthe 7th International Symposium on

tt43

Water-Rock Interaction, p. 153-158. Balkema, Rotterdam, the Neth-erlands.

Ildefonse, P., Kirkpatrick, R.J., Montez, B., Calas, G., Flank, A.M., andLagarde, P. (1994) 'Al MAS NMR and aluminium X-ray absorptionnear edge structure study of imogolite and allophanes Clay and ClayMinerals, 42,276-287.

Ishiguro, E., Iwata, S., Suzuki, Y., Mikuni, A., and Sasaki, T. (1982) Theboron K photoabsorption spectra of BFr, BCl3 and BBr3. Journal ofPhysics B: Atomic and Molecular Physics, 15, l84l-1854.

Katrinak, K.A., Rez, P., and Buseck, P.R. (1992) Structural variations inindividual carbonaceous particles from an urban aerosol. Environmen-tal Science and Technology, 26,1967-1976.

I-eapman, R.D., Fejes, P.L., and Silcox, J. (1983) Orientation dependenceofcore edges ftom anisotropic materials determined by inelastic scat-tering of fast electrons. Physical Review B, 28, 2361-2373.

Li, D., Bancroft, G.M., Kasrai, M., Fleet, M.E., Feng, X.H., Tan, ILH.,and Yang, B.X. (1993) High-resolution Si r(- and L,.'-edge XANES ofa-quartz and stishovite. Solid State Communications, 87, 613-617.

Luck, S., and Urch, D.S. (1990) Boron Ka XRE and the electronic struc-ture ofboron compounds: Coordination and resonance emission. Phys-ica scripta, 41, 97 0-97 2.

Marchetti, P.S., Kwon, D., Schmidt, w.R., Interrante, L.V., and Maciel,G.E. (1991) High-field I'B magic-angle spinning NMR characterizationof boron nitrides. Chemistry of Materials, 3,482-486.

McComb, D.W., Hansen, P.L., and Brydson, R. (1991) A study of siliconELNES in nesosilicates. Microscopy, Microanalysis, Microstructures,2, 56 t -568.

McComb, D.W., Brydson, R., Hansen, P.L., andPayne, R.S. (1992) Qual-itative interpretation of electron energyJoss near-edge structure in nat-ural zircon. Joumal ofPhysics: Condensed Matter, 4, 8363-8374.

Meera, B.N., Sood, A.K., Chandrabhas, N., and Ramakrishna, J. (1990)Raman study of lead borate glasses. Journal of Non-Crystalline Solids,126,224-230.

Moore, P.8., and Araki, T. (1978) Dumortierite, Si3B[A16,50,5O','dOtI)o-]a detailed structure analysis. Neues Jahrbuch fiir Mineralogie Abhan-dlungen, 132, 231-241.

Moustafa, Y.M., Doweidar, H., and El-Dawrawi, G. (1994) Utilizationof infrared spectroscopy to determine the fraction of the four coordi-nated borons in borate glasses. Physics and Chemistry of Glasses, 35,104-106.

Mozzi, R.L., and Warren, B.E. (1970) The structure of witreous boronoxide. Journal of Applied Crystallogaphy, 3, 251-257.

Miiller, D., Grimmer, A.-R., Timper, U., Heller, G., and Shakibaie-Mog-hadam, M. (19931 'tg MAS NMR studies on the structure of borateanions. Zeitschrift ffir anorganische und allgemeine Chemie,619, 1262-1268.

Pease, R.S. (1950) Crystal structure ofboron nitride. Nature, 165,722-723 .

Phillips, M.W., Gibbs, G.V., and Ribbe, P.H. (1974) The crystal structureof danburite: A comparison with anorthite, albite, and reedmergnerite.American Mineralogist, 59, 7 9-85.

Prabakar, S., Rao, K.J., and Rao, C.N.R. (1990) 'rB NMR spectra andstructure of boric oxide and alkali borate glasses. Proceedings of theRoyaf Society of London, 4294,, l-15.

Prewitt, C.T., and Shannon, R.D. (1968) Crystal structue of a high-pres-sure form of BrOr. Acta Crystallographica,824, 869-87 4.

Pring, A., Din, V.K., Jefferson, D.A., and Thomas, J.M. (1986) The crys-tal chemistry of rhodizite: A re-examination. Mineralogical Magzinq50,163-172.

Rez, P. (1992) Energy loss fine structure. In M.M. Disko, C.C. Ahn, andB. Fultz, Eds., Transrnission electron energy loss spectrometry in ma-teriafs science, p. 107-129. The Minerals, Metals and Materials Soci-ety, Warrendale, Illinois.

Rez, P., Weng, X., and Ma, H. (1991) The interpretation of near edgestructure. Microscopy, Microanalysis, Microstructures, 2, 1 43-l 5 l.

Rez, P., Bruley, J., Brohan, P., Payne, M., and Garvie, L.A.J. (1995)Calculation ofnear edge structure. Ultramicroscopy, in press.

Rosenberg, R.A., Love, P.J., and Rehn, V. (1986) Polarization-dependentC(K) near-edge-absorption fine structure ofgraphite. Physical ReviewB,33, 4034-4037.

Sauer, H., Brydson, R., Rowley, P.N., Engel, W., and Thomas, J.M. (1993)Determination of coordinations and coordination-specific site occu-


tt44

pancies by electron energy-loss spectroscopy: An investigation of bo-ron-oxygen compounds. Ultramicroscopy, 49, 198-209

Schwarz, W.H.E., Mensching, L., Hallmeier, K.H., and Szargan, R. (1983)K-shell excitations of BF3, CFo and MBF. compounds. Chemical Phys-ics. 82. 57-65.

Sekiyama, H., Kitajima, Y., Kosugi, N., Kuroda, H., and Ohta, T. (1984)Photon Factory Activity Report 1983/84, VI-l 19. Tsukuba, Japan.

Sen, S., Stebbins, J.F., Hemming, N.G., and Ghosh, B. (1994) Coordi-nation environments of B impurities in calcite and aragonite poly-morphs: A I'B MAS NMR study. American Mineralogist, 79, 819-825.

Sills, J.A., Martin, S W., and Torgeson, D R. (1994) "B NMR studies ofthe short range order in wide composition range xNarS+(l-x)BrS,glasses. Journal ofNon-Crystalline Solids, 168, 86-96.

Solozhenko, V.L, Chernyshev, V.V., Fetisov, G.V., Rybakov, V.8., andPetrusha, I.A. (1990) Structure analysis ofthe cubic boron nitride crys-tals. Journal ofPhysics and Chernistry ofSolids, 51, 1011-1012.

Takeuchi, Y., Ozawa, T., Ito, I., Araki, T., Zoltai,T., and Finney, J.J.(1974) The B,SisOro groups oftetrahedra in axinite and comments onthe deformation of Si tetrahedra in silicates. Zeitschrift fiin Kristallo-graphie, 140,289-312.

Terminello, L.J., Chaiken, A., I-apiano-Smith, D.A., Doll, G L., and Sato,T. (1994) Morphology and bonding measured from boron-nitride pow-ders and films using near-edge X-ray absorption fine structure. Journalof Vacuum Science Technology, A12, 2462-2466.

Tossell, J.A. (1986) Studies ofunoccupied molecular orbitals ofthe B-O

CARVIE ET AL.: PEELS STUDY OF THE B K EDGE OF MINERALS

bond by molecular orbital calculations, X-ray absorption near edge,electron transmission, and NMR spectroscopy Arnerican Mineralogrst,71, t t70-1177.

Turner, G.L., Smith, K.A., Kirkpatrick, R.J., and Oldfield, E. (1986) Bo-ron-l I nuclear magnetic resonance spectroscopic study ofborates andborosilicate minerals and borosilicate glass. Journal of Magnetic Res-onance. 67. 544-550.

Vaughan, D.J., and Tossell, J.A. (1973) Molecular orbital calculations onberyllium and boron oxyanions: Interpretations of X-ray emission,ESCA, and NQR spectra and of the geochemistry of beryllium andboron. American Mineralogist, 58, 7 65-7 7 O.

Yvedensky, DD., Saldin, D.K, and Pendry, J.B. (1986) An update ofDLXANES, the calculation of X-ray absorption near-edge structure.Computer Physics Communications, 40, 421-440.

Weijs, P.J.W., Czyzyk, M.T., van Acker, J-F., Speier, W., Goedkoop, J.B.,van Leuken, H., Hendrix, H.J.M., de Groot, R.A., van der Laan, G.,Buschow, K.H.J., Wiech, G., and Fuggle, J.C. (1990) Core-hole effectsin the X-ray-absorption spectra of transition-metal silicides. PhysicalReview B. 41. I 1899- l 1910.

Zachariasen, W H., Plettinger, H A, and Marezio, M. (1963) The struc-ture and birefringence of hambergite, BerBOr.OH. Acta Crystallogra-phica, 16, 1144-1146.

MeNuscnrp,r RECETVED Decsllnrn 12, 1994MeNuscnrsr AcCEPTED Jrnre 27, I995

parallel electron energy-loss spectroscopy (peels) study...

Documents