American Mineralogist, Volume 79, pages 411-425, 1994

Use of electron-energy loss near-edge fine structure in the study of minerals

LlunBNcn A. J. G.lnvrnr* AuN J. Cnc'YEI'{Department of Physics and Astronomy, University of Glasgow, Glasgow Gl2 8QQ, U.K.

Rrx BnvosoNDepartment of Materials Science and Engineering, University of Surrey, Guildford, Surrey GU2 5XH, U.K.

Arsrucr

High-resolution electron-energy loss near-edge fine structure (ELNES) recorded in ascanning transmission electron microscope (STEM) is shown to provide information onthe local structure and bonding of specific types of atoms in minerals. The Lr,, ELNESfrom Fe (Fe'z+ and Fe3*), Mn (Mn'*, Mn3+, and Mno*), and Cr (Cr3+ and Cr6*) showvalence-specific multiplet structures that can be used as valence fingerprints. In general,

the Z, edge for a specific 3d transition metal exhibits a chemical shift toward higher energylosses with an increase in oxidation state. Examples of mixed valence Fe- and Mn-bearingminerals are presented where the presence of multiple valence states is distinguished by asplitting of the L, edge. The high spatial resolution that can be obtained using the STEMallows variations in the relative proportions of the oxidation states to be detected on ascale down to I nm2. This resolution is illustrated from a sample of hausmannite thatshows different lr-edge shapes consistent with variations in the Mn2+-Mn3+ ratio overdistances of ca. 100 nm. Furthermore, spectra of many elements exhibit ELNES shapescharacteristic of the nearest neighbor coordination, as is demonstrated for C in the car-bonate anion and Si in the SiO" tetrahedral unit. The C K edge from CO3- is comparedwith that for elemental forms of C that exhibit very different ELNES. Similarly the Si lr,,ELNES from a range of SiOf -containing minerals all show the same near-edge shape thatis very different from Si and SiC. ELNES allows for its semiquantitative analysis, whichis illustrated by two theoretical techniques. Finally, the effects of electron beam damageare discussed in relation to the experimental changes observed in the ELNES.

IxrnonucrroN

In a transmission electron microscope (TEM), thehigh-energy incident electrons that have been elasticallyscattered may be used to obtain images and diffractionpatterns from specific regions of the sample. This allowsimportant structural information to be deduced about thesample. The high spatial resolution offered by TEM hasbeen useful for studying submicroscopic twins and exso-lution lamellae in minerals. The TEM has also been veryuseful in mineralogy for high-resolution structure imag-ing. The incident electrons also undergo inelastic inter-actions with the sample (Joy, 1993) and can provide fur-ther information on the specimen. That has given rise toa number of analytical techniques, such as electron-en-ergy loss spectroscopy (EELS), energy-dispersive X-rayspectroscopy (EDX), and Auger emission spectroscopy(AES). In EDX and AES it is the secondary processes ofelectron and photon emission that are studied, whereasin EELS the signal is due to the primary process of elec-tron excitation. EELS can provide valuable informationabout the material under study, such as a quantitative

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

0003-o04x/94 / oso644 | l $02.00

chemical analysis on the nanometer scale. More impor-tant, the electron-energy loss near-edge structure (ELNES)of a core-loss edge can also provide information aboutthe valency, coordination, and site symmetry of the atomundergoing excitation (e.g., Brydson et al., 1992a).

Recent advances in detector design, specifically theavailability of parallel EELS (PEELS) spectrometers (Kri-vanek et al., 1987; Krivanek, 1989), allows data to becollected that are comparable with those obtained byX-ray absorption spectroscopy (XAS). Until recently, mostEELS systems employed serial detectors with which thespectrum was acquired by scanning the signal across asingle detector, one channel at a time. The developmentof parallel detectors, specifically multichannel arrays, hasallowed a large gain in the collection efrciency, typicallyby a factor of 500, compared with earlier serial spectrom-eters. Parallel detection allows high-resolution spectra tobe acquired from materials that were too beam sensitiveto be collected by serial EELS at comparable spatial res-olutions.

This paper demonstrates the great potential of PEELS,used in conjunction with a TEM, for providing infor-mation about the crystal chemical environment of atomsin minerals. Clay minerals and clay-sized minerals werechosen for this study, as they can be difficult to charac-

4tl

412 GARVIE ET AL.: ENERGY LOSS NEAR-EDGE FINE STRUCTURE

350 400 450Energy Loss (eV)

Fig. l. A representative EELS spectrum showing (a) the zero-loss peak and low-loss region and (b) a coreJoss edge from themineral gaudefroyite. To highlight the fact that most electronsdo not participate in an inelastic scattering event, the zeroJosspeak is shown to scale in a, together with an expanded viewillustrating the lowJoss region. Following the low-loss region isthe monitonically decreasing background. (b) The edge onsetexhibits two intense white-line features due to the Ca Lr.. edge.The EXELFS modulations are very weak and only just visibleon the tail ofthe edge.

terize by conventional analytical techniques because oftheir fine-grained and often heterogeneous nature. Thepaper concentrates primarily on two types of informationthat can be gained from a qualitative analysis of theELNES spectra: (l) determination of the valency of 3dtransition elements, and (2) identification of local anionand cation coordinations. The ability to model ELNES isillustrated by two theoretical methods that provide semi-quantitative information on the core-loss edges and henceabout the atom undergoing excitation. Consideration isalso given to effects induced by the electron beam on thesample as revealed by the ELNES. Only a brief introduc-tion to EELS is given; more detailed accounts can befound in Egerton (1986) and Buseck and Self(I992). Thereare many papers that deal with ELNES; introductory pa-pers include Coll iex et al. (1985), Brydson (1991), andBrydson et al. (1992a).

Tgn EELS SPEcTRUM

An EELS spectrum (Fig. l) displays the electron inten-sity as a function ofenergy loss and can be divided into

several regions. The large zero-loss peak (ZLP) resultsfrom transmitted electrons that have undergone elasticand quasi-elastic (mainly phonon) interactions, i.e., elec-trons that have lost no, or very little, energy. Its widthdepends primarily on the energy spread and stability ofthe electron gun and determines the best resolution thatcan be obtained at the core-loss edges. The region im-mediately following the ZLP and extending to energylosses of ca. 50 eV, called the low-loss region, is an areaof the spectrum dominated by plasmons. Plasmons canbe described as resulting from the collective excitationsof valence electrons. They can provide information aboutthe dielectric function (Buechner, 1975), valence electrondensities, and, in suitable cases, the phases present inalloys. Within the low-loss region are also found edgescaused by transitions from outer shell electrons. Extend-ing beyond the low-loss region in a spectrum from a thinsample is a monotonically decreasing background, arisingpredominantly from plasmon and single electron excita-tions, on which are superimposed the coreJoss edges.These features result from the transition ofcore electronsto unoccupied states in the conduction band; this requiresthe energy transfer between the incident and a core elec-tron to be greater than its binding energy. The coreJossedges usually take the form of a step, characterized by asudden increase in intensity (Fig. lb) that decreases inintensity with increasing energy loss. This sudden rise inintensity represents the ionization threshold, the energyof which approximately corresponds to the inner-shellbinding energy and hence is characteristic ofthe element.Close to the thresholds and for small scattering vectors,these core-loss edge transitions are governed by the atom-ic dipole selection rules for the electronic transitions A/: + l, and Aj : 0, +1, where I and j are the orbital andtotal angular momentum quantum numbers of the sub-shell from which the electron has been excited. The edgesare classified according to the standard spectroscopic no-tation, e.9., the Ca Lr' edges in Figure lb arise from tran-sitions from the 2p core level.

In the first instance EELS can be used for elementalanalysis, since the edges occur at energy losses equivalentto the binding energies of the core electrons in the sample,and the integrated intensity under the edge is related tothe number of atoms under the electron beam. UnlikeEDX, EELS can provide elemental data on all the lightelements except H and He, although the latter has beendetected under certain circumstances. e.g., in nanometer-sized He bubbles in irradiated Ni alloys (McGibbon,r 99 l ) .

Core-loss edges can be divided into two regions (Fig.lb), the ELNES, extending 30-50 eV above the edge on-set, and the weaker extended electron-energy loss finestructure (EXELFS). The equivalent XAS terms areXANES (X-ray absorption near-edge structure) and EX-AFS (extended X-ray absorption fine structure). In a sol-id, the core wave functions show very little change ascompared with an isolated atom. This is not the case forthe outer bonding and antibonding orbitals that can in-

0 50 100ELNES D

GARVIE ET AL.: ENERGY LOSS NEAR.EDGE FINE STRUCTURE

TABLE 1. Name, simplified formula, valence, coordination, and source of the 3d transition metal containing minerals

Mineral name Simolified formula*Valence andcoordination Source

413

Fe-bearing mineralsChromiteSideriteHematitelron analogue of leucite*.Chlorite (var. daphnite)CronstedtiteNontronite

RamsdelliteAsbolanNorrishiteBixbyiteGanophylliteManganositeHausmannite

CrocoiteVolkonskoiteChromite

(Fe,MgXCr,Al,Fe),04(Fe,Mn)COgFe,O"KFeSi,O6(Fe,*,AD,,(Si,AD6O,o(OH),6' nH,O(Fe'z*,Fe3* )6(Si,Al,Fe)1Olo(OH)6' nH,OCtu e5Fe4(Si,Al)sOro(OHL nH2O

r4tFe2+

r6tFe2+

16lFe3+

r41 Fe3+t61Fg2 +

l6iFe2+, {5lFe3+, I4lFe3+

rsr FeF*

Gebe lsland, Indonesialvigtut, GreenlandLangban, Sweden

Redruth, Cornwall, EnglandWheal Jane, Cornwall, EnglandLajitas, Texas

Pirika Mine, Hokkaido, JapanGebe lsland, IndonesiaHoskins Mine, N.S.W., AustraliaThomas Mt., UtahLangban, SwedenNordmark, SwedenKalahari Mine Field, S. Africa

Redlead Mine, Dundas, TasmaniaGlazov Region, UdmurtianGebe lsland. Indonesia

Mn-bealing mineralsMnO, t61Mn4+

Mn(O,OH).(Co,N|),(OH),, nH,O I6rMn1+

KLiMn4SisOr4 t6lMn3+

(Mn,Fe),O3 2 x I61Mn3+

(K,Ca,Na),(Fe,Mn)"(Si,Al),,O,e(OH),.nH,O 16rMn2+

Mno I6lMn2+

c-Mn"Oo l4lMn2+, I6lMn3+

Cr-bearing mineralsPbCrO. I61Cr6+

Ca,,(Cr,Fe,Mg)o(Si,Al).O,0(OH)o'nHrO 16ro13+

(Fe,MgXCr,Al,Fe),O4 t6rcr3+

' Maior elements based on qualitative EDX analysis't Synthetic Fe analogue of leucite

teract to form bands oforbitals that are governed by thesolid-state character of the specimen and can be appre-ciably modified by chemical bonding. The ELNES di-rectly probes the unoccupied orbitals and therefore re-flects the environment surounding the excited atom andcan provide information on bonding, valency, coordina-tion, and site symmetry. Note that when discussing mi-croanalysis the edge is taken to mean all transitions froma given initial state, whereas in discussions of ELNES orXANES, the edge is taken to mean the structure in theregion close to the threshold.

The EXELFS oscillations are very weak modulationsofthe intensity extending several hundred electron voltsabove the ELNES region. These oscillations are causedby an ejected electron with a low kinetic energy behavingsimilarly to a free electron. The ejected electron is back-scattered by the surrounding atom shells, which can pro-vide information on the coordination number and near-est neighbor distances. EXAFS studies most commonlyinvolve analysis of K-edge modulations, such as thosefrom the transition elements. These K-edges occur at en-ergy losses of several thousand electron volts, where in-dividual edges are well separated in energy. Unlike itssister technique EXAFS, EXELFS is unlikely to be widelyused, since the fine structure modulations often overlapwith other edges in the energy range used to collect EELSedges, typically <1000 eV.

ExpnnrprnNrAl. sETUP

The instrument used was a VG HB5 STEM operatingat 100 kV and equipped with a cold field emission gun(FEG). With this microscope, a probe size of ca. l-nmdiameter was scanned over the sample. The microscopewas operated with a probe semiangle of I I mrad, a col-lection angle of 12.5 mrad, and a probe current of ca.0.5nA. The small collection semiangle, B, ensures that the

experiment is in the dipole regime for electronic transi-tions. A Garan 666 parallel electron-energy loss spec-trometer (PEELS) is attached to the top of the microscopecolumn (Krivanek et al., 1987; Krivanek, 1989). To ex-emplify the advantage of parallel detection, an 8-s acqui-sition on the PEELS would require just over I h on a

serial detection system. As well as requiring patience from

the operator, long counting times have other associatedproblems, including C contamination, lack of microscopestability, and beam damage to the sample. The spectrom-eter was calibrated against the Ni I, edge in NiO that hasa peak maximum at 853.2 eV, as determined using syn-chrotron radiation studies (van der Laan et al.' 1986).

The samples used in this study have come from a va-

riety of sources or were collected by one of us (L.G.) andidentified by powder X-ray diffraction and EDX. Theminerals containing the 3d transition metal elements arelisted in Table l, together with their formulae, selectedcrystallographic data, and sources. Table 2lists mineralsused in the coordination fingerprint study. Materialsstudied were either checked prior to study in the HB5STEM by EDX in an SEM, or checked qualitatively byEDX in the HB5 STEM. Crystalline samples were pre-pared by crushing selected mineral grains in acetone andallowing a drop of the fine-grained material in suspensionto dry on a lacy C-coated Cu TEM edd. Alternatively,clay samples were prepared by dispersing a small amountof material in distilled HrO, and a small drop of the <2-

rrm clay fraction in suspension was dried on the abovegrids. Lacy C films were used so that data could be re-corded from thin electron-beam transparent grains pro-jecting over the holes. Core-loss edges were recorded with

a dispersion of 0.1 or 0.05 eV per channel in order torecord the detailed fine structure ofthe edges. The actualenergy resolution is limited by the energy spread of theFEG, which is close to 0.27 eV at low probe currents. At

4t4 GARVIE ET AL.: ENERGY LOSS NEAR-EDGE FINE STRUCTURE

TlsLe 2. Name, simplified formula, and source of C- and Si-containing minerals

Mineral name Simplified formula'

CalciteSideriteDesautelsite'.HydrotalciteAmorphous C**GraphiteDiamond

QuartzTalcNontroniteB silicon carbide-'si--

C-containing mineralsCaCO"(Fe,Mn)CO3MguMn.(CO.)(OH)16 4H,OMg6Al,(C03XOH)16. 4H,O

c

c-SiO,Si-containing minerals

(Mg,Fe)6(Si,Al)sO,o(OH)4 nH,OCao e5Fe4(Si,Al)so,o(OH)4 nHrOsic5 l

Llanelwedd Quarry, Waleslvigtut, Greenland

Snarum, Norway

South Western Graphite Mine, TexasSouth Africa

Llanelwedd Quarry, WalesLlanelwedd Quarry, WalesLalitas, Texas

' Major elements based on qualitative EDX analysis.- Synthetic material.

normal operating currents, the width increases to a min-imum of ca. 0.3 eV, which is further increased to ca. 0.45eV by residual interference. The EELS data were typicallyobtained with a scanned raster from areas in the rangeI 30 x 100 to I 3 x 10 nm, with acquisition times rangingfrom 0.5 to 8 s. With a sample thickness of 20 nm andan irradiated area of 13 x l0 nm2, this corresponds to avolume of roughly 2600 nmr and an approximate massof 8 x l0 " g (with p :3 g/cm3). Since the analyzedvolume is so small, the area probed is almost invariablypart of a single crystal. In a typical EELS experiment, thedose given at the sample varies from 1.25 x 1gs e/At fora 0.5-s Io 2 x l}e e/A' for an 8-s integration time. Thechoice of magnification and acquisition time is governedby the stability of the material under the electron beam.For each core-loss edge recorded, three additional spectrawere collected: the dark current for the core-loss edge, thelow-loss region, and its corresponding dark current. Thedark current arises from thermally excited current in thedetector. The dark current is then subtracted from thecorresponding spectrum. Before presenting the core-lossedges, a background of the form AE-' was subtractedfrom beneath the edge, and the effect of multiple scatter-ing and asymmetry of the ZLP were deconvoluted fromthe edge. The adjustable parameters A and r can be de-duced by least-squares fitting ofthe intensity ofa specificregion directly preceding the core-loss edges. The topicsof background subtraction and ZLp deconvolution aredescribed in detail in Egerton (1986).

Sample thickness is an important factor when collect-ing EELS spectra. In order to reduce the effect of multiplescattering the thickness should be <50-100 nm for 100-kV incident electrons (Egerton, 1986). Ifthis condition isnot met, then there is an increased probability that thetransmitted electron will be involved in more than oneinelastic scattering event. Although the effect of a smallamount of multiple scattering can be removed using de-convolution techniques, it becomes rapidly more difficultto detect edges as the specimen thickness increases. In thesamples studied, abundant thin areas could always be

produced by simply crushing the material in a mortar andpestle. The phyllosilicates are ideal materials for EELSstudy because oftheir platy nature, which provides abun-dant electron-beam transparent areas.

ELNES

Valency determination of 3d transition metals

All of the 3d transition metals are found in naturalphyllosilicates, and they can be a major chemical com-ponent. Examples include roscoelite (V3* ), volkonskoite(Cr.* ), glauconite (Fe.* ), pennantite (Mn'* ), and pime-lite (Nir* ). Fe is by far the most common 3d transitionmetal in minerals. Many spectroscopic techniques havebeen used to determine both the oxidation state and co-ordination of 3d transition metals in minerals: 57Fe Mdss-bauer spectroscopy (Cardile and Brown, 1988; Haw-thorne, 1988), optical spectroscopy (Rossman, 1988),XANES (van der Laan et al., 1992; Manceau et al., 1992:Cressey et al., 1993), electron paramagnetic resonance(Calas, 1988), photoelectron spectroscopy (Rao et al.,1979), and high-resolution X-ray emission spectroscopy(Koster and Mendel, 1970; Wood and Urch, 1976). Anal-ysis of K-edge XANES has proved useful in determiningthe coordination and oxidation state ofselected transitionmetals in minerals (Calas and Petiau, 1983; Manceau etal., 1992). The feature that most reflects the crystal chem-ical environment of the transition elements is a preedgefeature before the main K absorption edge. This has beeninterpreted as arising from predominantly dipole-forbid-den 1s - 3d transitions and so is fairly weak in intensity.In contrast, the dipole-allowed transitions that give riseto the Ir., edges exhibit numerous strong sharp featuresor multiplet structure. Fine structure is not readily ob-servable on the K-edge prepeak because of the greatercore-hole lifetime broadening associated with the K edgethan with the Ir,. edges. The broadening on the Ir., edgesof the 3d transition elements is typically 0. 1-0.3 eV,whereas it is three to four times larger on the correspond-ing K edge. The use of PEELS to determine the oxidation

GARVIE ET AL.: ENERGY LOSS NEAR-EDGE FINE STRUCTURE 4t5

state and, in certain circumstances, the coordination oftransition elements is illustrated using the lr., edges fromFe. Mn. and Cr.

Several ELNES and XANES studies have shown thatlhe Lr.. edges from the 3d transition elements exhibitstructures that are sensitive to the valence state of themetal (Otten et al., 1985; Otten and Buseck, 1987; Rasket al., 1987: Krivanek and Paterson, 1990; Paterson andKrivanek, 1990; van der Laan et al., 1992; de Groot etal., 1993; Garvie and Craven, 1993, 1994). The ELNESof the 3d transition melal Lr. edges are characterized bytwo whiteline features (so called because they were firstidentified as such on photographic plates) that are nar-row, intense peaks. These white lines arise from quasi-atomic dipole-allowed transitions from an initial state,2p63d', to final states of the form 2psJl'+t. The highintensity of the 3d transition metal Lr,, edges is due tothe high density of unoccupied 3d states. The two sepa-rate edges, l,, and lr, arise from the two ways that thespin quantum number, s, can couple to the orbital an-gular momentum, I giving the total angular momentum,j : I + s. That gives two separate peaks due to transitionsfrom 2p,, (j -- lr), forming the L' edge, and 2pn U :'/r), forming the I, edge. Final states of s character arealso dipole allowed, but they are less intense because thedipole-allowed 2p - s transitions are approximately anorder of magnitude less intense than 2p - d transitions'This, coupled with the fact that s and p densities of stateare rather broad compared with d states, implies that theZr., edges essentially represent transitions to the vacant3d orbitals. Hence the ELNES reflects the local environ-ment and charge state of the 3d transition metal ion inquestion.

An isolated 3d transition metal ion such as Mn2* hasfive degenerate 3d orbitals. In a solid, the effect of thefield from the surrounding ligands is to split the 3d or-bitals into two or more separate orbitals. In minerals, the3d transition metals are predominantly octahedrally co-ordinated and exhibit a wide range of symmetries. Forhigh-spin compounds the effect of the ligands is to splitthe degenerate 3d orbitals into the lower energy tr, andhigher energy e, orbitals. The energy difference betweenthese two sets of orbitals is defined as the crystal fieldparameter, designated as lODq (or A in some texts). Aswill be seen later, the crystal field can have large effectson the shape of tllre Lr.t edges. In the following sectionson the I- edges, the discussions will be primanly con-cerned with the analysis of the Z. edges, which are moreintense and exhibit more multiplet structure than thebroader I, edges. The increased broadening of the Itedge, in comparison with the L, edge, is due to the avail-ability of the extra Coster-Kronig Auger decay channel,which reduces the lifetime of the final state and henceincreases the width of the L, multiplet lines.

Fe. Fe is found as both Fe2* and Fe3* in minerals, andit is often the case that more than one valency and co-ordination is present in the same mineral. Minerals con-taining Fe2* and Fe3* are most commonly octahedrally

705 710 715 720 725

EnergY Loss (eV)

Fig.2. Fe I- edges from reference materials. Synthetic ironanalogue of leucite (torFet* ), hematite (r6rFe3+ ), siderite (t6lFe'?+ )'and chromite (lolFet* ). The spectra illustrated here and in sub-sequent figures are shown on an arbitrary scale.

coordinated, although Fe in tetrahedral coordination is

sometimes found. Fe in minerals is expected to be in thehigh-spin state, as only strong field ligands such as CN-can cause spin pairing. Before the discussion of the Fel,., edges for several phyllosilicates, the Fe Ir.. edges from

a number of reference minerals containing Fe2+ and Fe3+in both octahedral and tetrahedral coordination are dis-cussed. They are chromite (t4lFe2+ ), siderite (l61Fe2+ )' he-matite (r6rFe3+), and the Fe analogue of leucite (t41Fe3+).

The Lr.. edges from the reference compounds (Fig. 2)all show different edge shapes, as well as a distinct chem-ical shift between the divalent and trivalent Fe Z, edges(Table 3). These basic edge shapes are consistent withpublished Fe I, XANES (Cressey et al., 1993) and ELNES(I(rishnan, 1990; Kurata et al., 1990) spectra. The l,, edg-es for the divalent Fe minerals arc at ca. 707.8 eV, andthe trivalent Fe Z, edge is 1.7 eV higher, at ca. 709.5 eV.The separation of the L, and Lt peak maxima, due tospin orbit splitting, is ca. l l eV. The Fe I, edges fromthe reference compounds illustrate that different site sym-metries are recognizable by changes in the multipletstructures on the Z3 edges. Hematite has a distinctive Ir-

edge shape, with a prepeak leading the main L, peak,

whereas the Fe analogue of leucite, which contains talFe3+'

Fe2+ Fe L2,3

I

416 GARVIE ET AL.: ENERGY LOSS NEAR-EDGE FINE STRUCTURE

TABLE 3. Energies (eV) of the ln and In edge peak maximumfor selected 3d transition metal-containing minerals

Element valenceMineral name and coordination lo peak max. I? peak max.

SideriteHematiteManganositeNorrishiteBixbyiteAsbolanChromiteCrocoite

t6 lFg2+

16lFe3+

t6lMn2+

16lMn3+

l6lMn3+

161Mn4 +

I61Cf3+

r4tc16+

707.8709.5640.2642.5641.8643.8578.6580.9

720.4722.6650.3653.6o c z . o

6s4.3s86.6589.5

Note: efior in energy values is +0.2 eV

shows an Z, edge without any pronounced fine structure.fhs telpsz+ L, edge from siderite shows several partiallyresolved multiplets on both the low- and high-energy sidesof the l,, peak maximum, whereas these are not presenton the Z,3 edge of chromite. However, the peak at ca. 7l0eV is much stronger in chromite than in siderite. The I,edges from hematite, the Fe analogue of leucite, and sid-erite correspond closely with the theoretical atomic mul-tiplet spectra calculated by van der Laan and Kirkman(1992), which confirms the single valence and site sym-metry of the Fe in these samples. However, for chromitethe experimental edge has a much larger subsidiary peakaI ca.7I0 eV than the calculated edge for lalFer+ with theappropriate 10Dq. This subsidiary peak is at approxi-mately the correct energy for the L, edge from Fe3+. Ittherefore appears that the chromite contains several per-cent of Fe3* substituting for I6tCr3+, as is often the casefor chromite (e.g., Yang and Seccombe, 1993). The pres-ence of Fe3+ in octahedral sites was confirmed by EDX,which showed that the majority of octahedral sites arefilled with Cr and Al, with approximately one out of ev-ery 16 octahedral sites occupied by Fe.

The two peaks on the I, edge in hematite are separatedby 1.5 eV, as they are in many other minerals containingt6lFer* (unpublished results). These two peaks have beeninterpreted in terms of a simple crystal field model(Krishnan, 1990), where they have been assigned to thetr, and e, orbitals. Although such a simplistic ligand fieldapproach appears to explain the features on the Fe I.edge, as well as on the Mn I, edges from minerals con-taining Mn3+ and Mn4+ (Garvie and Craven, 1993) andcompounds with the valence of the 3d transition metalin the d0 state (Brydson et al., 1987, 1989a), that is notgenerally the case (de Groot et al., 1990a, 1990b). Thevalue of the crystal field strength for hematite can be es-timated by comparing its I, edge with the calculatedatomic multiplet spectra of van der Laan and Kirkman(1992). Their best fit to the experimental hematite I. edgeis to the spectrum calculated for Oo symmetry with a1 ODq value of 2 eY, which is close to the value of 1.9 eVdetermined by optical spectroscopy for hematite (Maru-sak et al., 1980). Similarly, the values of the 10Dq canbe estimated to be I and 0.5-l eV for the Fe analogue ofleucite and siderite, respectively.

705 7't0 715 720 725Energy Loss (eV)

Fig. 3. Fe Ir., edges from three phyllosilicates. Nontronite(dominantly t61Fe3+ ), cronstedtite (three distinct Fe sites, 16lFe2+,I6lFe3*, and totFer* ) and the chlorite daphnite (dominantly r6lFez+with a few percent lolFer+ ).

The Fe Zr.. edges from a number of phyllosilicates con-taining various Fer+-Fe2+ ratios were collected, and rep-resentative spectra are illustrated in Figure 3. The datawere collected from individual well-defined clay particles,typically from areas measuring 130 x 100 nm2. Daphnitehas an Fe Lr., edge shape similar to that from siderite,confirming both that the Fe is octahedrally coordinatedand divalent. The higher intensity of the partially re-solved peak on the high-energy side of the I, edge, incomparison with that on the siderite I. edge, may indi-cate the presence of a small amount of Fe3*. That hasbeen confirmed for this sample by 57Fe M6ssbauer spec-troscopy (Goodman and Bain, 1979). Cronstedtite con-tains a number of Fe Sites: r4tFe3+, r6tFe3+, and r6lFe2+(Bailey, 1988). The presence of both Fe3+ and Fe2+ isvisible from the double-peaked Z. edge, the two edgesbeing separated by 1.7 eY . The partially resolved peakon the low-energy side of the Fe2* L, edge indicates thatit is dominantly octahedrally coordinated, as expectedfrom the crystal structure. Although the presence of Fe3+is confirmed by the peak 1.7 eV higher in energy than theFe2* edge, it is not obvious from the L, edge that bothI6lFe3+ and [alFe3+ are present. Nontronite gives an Fe lr.r-edge shape identical to that from hematite, indicating thatthe majority of the Fe is teJ['sr+, as expected for nontron-ite (Giiven, 1988). The splitting between the two peakson the L, edge in nontronite is slightly less than in he-matite, indicating a slightly smaller crystal field value.

GARVIE ET AL.: ENERGY LOSS NEAR-EDGE FINE STRUCTURE 4t7

This is consistent with the l0Dq of 1.77 eY for nontron-ite determined by optical spectroscopy (Bonnin et al.,1985). As illustrated by comparing the Z, edge from he-matite with that from nontronite, it is possible to obtaina relative, and possibly quantitative, measure of the crys-tal field strength acting on the metal ion from the mul-tiplet splittings.

Since the Fe2* and Fe3* Z. edges are well separated inenergy, it is possible to determine Fe3+-Fe2+ ratios byfitting reference spectra to the experimental Fe Ir,, edges.This has been demonstrated by Cressey et al. (1993) usingXANES from a number of Fe-containing minerals havingmixed valency and coordination. The same would alsobe possible using EELS data, as long as a small collectionsemiangle, B, was used in the microscope, which wouldensure that only dipole-allowed transitions were collect-ed. Under this condition, EELS and XAS are equivalent.The success of determining Fe3+-Fe2+ ratios by fittingreference spectra in part depends on the constancy oftheFe l,rr-edge shapes in different materials and environ-ments. The spectra illustrated in Figures 2 and 3 andpublished Fe I,., EELS and XAS spectra (Krishnan, 1990;Krivanek and Paterson, 1990; Kurata et al., 1990; Pat-erson and Krivanek, 1990; Cressey et al., 1993) indicatethat the Fe2+ and Fe3* Zr., edges do not show large de-viations from the calculated Fe Lr., edge spectra of vander Laan and Kirkman (1992), and so Fe3*-Fe2* ratiodetermination should be possible for compounds con-taining Fe2* and Fe3* in roughly tetrahedral and octa-hedral coordination where the Fe to ligand bonds aredominantly ionic. The crystallinity of the mineral shouldalso have little effect on the shape of the Fe l., edges, ifone assumes that the metal to ligand bonds are domi-nantly ionic, since the spectra for Fe2* and Fe3+ are dom-inated by quasi-atomic transitions that give rise to the FeLr., edge, with only some modification due to the solidstate environment. Fe3*-Fe2* ratio determination is ex-pected to be more difficult when a multivalent or multi-site compound contains sites that are highly distorted fromTo or Oo symmetry, as in iron phthalocyanine (Koch etal., 1985), which contains Fe in a square planar site, orwhen the character ofthe ligands bonded to the Fe causesthe Fe to assume the low-spin state, as in FeS, (Thole andvan der Laan, 1988) and the iron isocyanates (Kurata etal., 1990). Although Cressey et al. (1993) have shown thatit is possible to determine the Fe3+-Fe2+ ratio in a num-ber of materials, such as spinels, amphiboles, Fe-contain-ing glass, and amethyst, this implies that one can deter-mine the Fe3+-Fe2+ ratio at the nanometer scale usingEELS, assuming the energy resolution of the EELS/TEMsetup is -0.5 eV or better. Such an energy resolution isusually only obtained with an ultra-high vacuum TEMfitted with a cold field emission gun.

Mn. Mn is another element common in the geologicalenvironment, and chemically it can exhibit a large num-ber of oxidation states and coordinations. In contrast toFe, Mn is found in the three oxidation states II, III, andIV in nature, although Mn in sedimentary and surficial

635 640 645 650 655Energy Loss (eV)

Fig. 4. Mn Zr., edges from representative Mn compoundsshowing the effect of different oxidation states on the shape andenergy of the L,, edges. Manganosite 1tet14nz*;, ganophyllite(r6lMnr+), bixbyite (two distinct [6rMnr* sites), norrishite (singlet6lMn3+ site), asbolan (t6lMna+ ), and ramsdellite (tullt4no*;.

Mn minerals is more likely to be Mn3* and Mna+. Thecoordination of Mn is almost always octahedral, althought41Mn2+ is found in some minerals such as jacobsite, ga-laxite, and hausmannite. Mn minerals are often fine-grained heterogeneous mixtures containing more than oneoxidation state. That makes the determination of theiroxidation state by conventional spectroscopic and chem-ical techniques very difficult and imprecise, as it onlygives a summed average of the oxidation states from thebulk sample. The high spatial and energy resolution of-fered by PEELS in a FEG-STEM is ideal for obtainingsuch information.

Figure 4 compares the Mn Zr., edges from a number ofMn-containing minerals. There are three obvious changesin the spectra with an increasing oxidation state of theMn: a move to higher energy losses of the I, peak max-imum (Table 3), a decrease in the L.-L, white line inten-sity ratio, and a characteristic shape change of the L,,edges. Manganosite illustrates a typical Mn2* Ir., edgewith an I. peak maximum at ca. 640 eV; many otherhigh-spin Mn2* compounds have been studied previous-

Mn2*

| "i'.Mn L2,3

Hausmannite

4 1 8 GARVIE ET AL.: ENERGY LOSS NEAR-EDGE FINE STRUCTURE

645 650635 640Energy-Loss (eV)

Fig. 5. Mn L, edges from a well-crystallized specimen ofhausmannite. The Z, edge is split into two peaks because of thetwo Mn sites in hausmannite (lalMn2+ and t6lMnr+ ).

ly, and all exhibit the same basic edge shape with littlevariation in coordination number or site symmetry (Rasketal., 19871' Garvie and Craven, 1993). The Mn l,, edgesfrom ganophyllite have the same shape and I, energy asin manganosite, indicating that the Mn is dominantlyMn2+, as is commonly assumed for this mineral (Dunnet al., 1983). The Mn Ir.. edges shown here are typical ofhigh-spin t6lMn2+. Low-spin t6lMn2+ is not thought to oc-cur in natural minerals, although it has a characteristicedge shape (Cramer et al., l99l) that can be used to dis-tinguish it from high-spin t6lMn2+. Two minerals are il-lustrated that contain Mn3*, bixbyite, and the oxybiotitenorrishite (Tyrna and Guggenheim, 1991). The Mn Zr.,edges in both minerals are very similar in shape, with abroad I, edge and a maximum atca.642 eV. In bixbyitethe I, peak maximum is split into two peaks separatedby ca. 0.5 eV, and there is further evidence ofunresolvedmultiplet splitting on the low-energy side of the I, peak.In norrishite only partially resolved multiplets are ob-served; this was observed for several Mn3* Z. edges (Gar-vie and Craven, 1993) where the Mn sits in a Jahn-Tellerdistorted environment due to the da electron configura-tion. The last two examples in Figure 4 are from the Mna *minerals asbolan and ramsdellite (Miura et al., 1990). Aswith the Mn2* and Mn3* Ir., edges, the Mna Ir.. edgeshave distinct edge shapes, with an l, peak maximum atca. 644 eV. All Mna+ Lr., edges have similar shapes, witha broad L, edge and a double peaked L, edge. The twopeaks on the Z, edge have been interpreted in terms ofcrystal field theory and assigned to the tr, and e, orbitals,respectively (Garvie and Craven, 1993). Similar assign-ments were made on the Mn2* I. edge. Interpretation ofthe Mn3+ I. edge is more complicated because of thetetragonal distortion ofthe octahedron caused by the Jahn-Teller effect, which further splits the tr, and e, orbitals.The Mn Z,. edge from manganite was assigned orbitalsbased on Do, symmetry in Garvie and Craven (1993).

636 638 640 642 644 646 648Energy Loss (eV)

Fig. 6. Representative Z, edges from another sample ofhaus-mannite. The .L, edges are arranged from top to bottom with anincrease in intensity of the Mn'* peak corresponding to an in-crease in the relative amount of Mnr+.

The ability of PEELS to observe variations in oxida-tion state at high spatial resolution was studied in twosamples of hausmannite. Both samples occurred as small,octahedral crystals of - I mm in size but were from dif-ferent localities. The first sample gave distinct and similarMn lr.. edge shapes from all the areas studied under themicroscope. Figure 5 illustrates a typical lr., edge fromthe first sample. A similar edge shape has been recordedby XANES (Cressey et al., 1993) and ELNES (Patersonand Krivanek, 1990). The Mn I, edge from hausmannitecan be interpreted by consideration of its crystal struc-ture. Hausmannite can be represented by the formular4tMn2+r6tMn3*Oo with an ideal Mn3+-Mn2+ ratiO Of 2:1,although experimentally synthesized hausmannite wasfound to have Mn3+-Mn2+ ratios varying from 2 to 3.5(Kaczmarek and Wolska, 1993). The L. edge shows twodistinct peaks separated by 1.6 eV. The first one is at 640eV, consistent with Mn'?+, and the second one is ca. 1.5eV higher and is from Mn3*. The second sample of haus-mannite exhibited a range of Mn L.-edge shapes. Rep-resentative spectra are illustrated in Figure 6. It is evidentfrom these spectra that there are large variations in theMn3+-Mn2+ ratio, ranging from almost pure Mn2* (at thetop of Fig. 6) to edges that resemble ideal hausmannite(middle of Fig. 6) to edges that are dominated by Mn3t

Mn2. Mn L3

GARVIE ET AL.: ENERGY LOSS NEAR-EDGE FINE STRUCTURE 4t9

(bottom of Fig. 6). These variations can occur over dis-tances as short as 100 nm.

Cr. In nature, Cr occurs in two oxidation states, III andVI. The Cr6+ ion occurs exclusively in the chromate anddichromate anions, as in crocoite, PbCrOo, and has notbeen found in clay minerals. The best-known Cr-bearingclay is the smectite volkonskoite, which has a bright greencolor due to the high percentage of Cr residing in theoctahedral layer (Giiven, 1988). the Cr lr., edges for threeminerals are illustrated in Figure 7. The Cr Ir., edgesfrom chromite show the same multiplet structure and ba-sic edge shapes as the theoretical atomic multiplet spectraof van der Laan and Kirkman (1992) for Cr3* in O, sym-metry and 10Dq values of 2-2.5 eV. This range of l0Dqvalues is consistent \Mith the crystal field strength of I6lCr3+compounds bonded to O that have optically determinedlODq values of 2.1-2.5 eV (Lever, I 984). The Cr l, edgefor chromite is at 578.6 eV. When one compares the CrIr., edges from volkonskoite and chromite, it is evidentthat the various features have the same shape and energypositions. This directly confirms the trivalent nature ofCr in the volkonskoite. The features on the I, edge in thevolkonskoite are slightly broader than those from thechromite, which can possibly be attributed to the moredistorted octahedral environment in the former mineral.Similar edge shapes have been recorded previously forthe Cr Ir., edges from CrrO, (Kurata et al., 1988; Kd-vanek and Paterson, 1990). For comparison, the Cr Lr'edge from crocoite is illustrated in Figure 7. This edge isconsistent with lower energy resolution PEELS work onKrCrOo (Kurata et al., 1988). The Z, ar.d L2 edges areresolved into two separate peaks, and there is a shift ofthe l,, peak maximum to a higher energy loss (Table 3),consistent with the higher effective charge on the Cr atom.The I4rCr6+ Lr:-edge shape is characteristic of tetrahe-drally coordinated XOX anions (X: Ti, V, Cr, and Mn)in general (Brydson et al., 1993) and provides a finger-print for these species.

Anion and cation coordination fingerprints

Many EELS spectra exhibit a near-edge structure char-acteristic of the arrangement and number of atoms in thefirst coordination shell surrounding an atom. This givesrise to the concept of a coordination fingerprint (Brydsonet a1., I 988; Sauer et al., I 993). A coordination fingerprintarises when the excited atom is surrounded by atoms thatare strong electron backscatterers, such as O'? and FIn that case the near-edge structure is dominated by scat-tering events within the first coordination shell surround-ing the atom. The ELNES structure may then be relatedto the molecular orbital (MO) structure of the excitedatom and its nearest neighbors (Sauer et al., 1993). Someexamples of ELNES coordination fingerprints include theB K edge from BO, and BOo (Sauer et al., 1993), the Sand P K edges from SO?- and POI- anions (Hofer andGolob, 1987), the Lr., and, K edges from talAl and t6lAl(Brydson et al., 1989b; Ildefonse et al., 1992), and the Kedges of t4tMg and I6rMg (Tafto and Zhu, 1982).

575 580 585 590 595Energy Loss (eV)

Fig. 7. Cr Zr.. edges for chromite (r6rcr3+ ), volkonskoite(l6lcrr+ ), and crocoite (tolCru* ).

Several allotropes of C exist, the most well known be-ing diamond, graphite, and the fullerenes such as Cuo.Between the extremes of sp3-bonded diamond and sp2-bonded graphite are a whole range of carbonaceous ma-terials that have intermediate bond types, such as glassyC and evaporated amorphous C (Robertson, 1986; Rob-ertson and O'Rielly, 1987). In minerals, inorganicallybonded C is found almost exclusively in the carbonateanion, COI-. Its structure consists of a central C atomthat is bonded to three planar trigonal O atoms. TheCOI- anion is isostructural and isoelectronic with BOI-and NO; anions, which all exhibit similar K ELNESshapes. The C K ELNES from the minerals containingthe carbonate anion exhibit an edge shape consisting ofa sharp initial peak a| 290.2 eV and a broader, less in-tense, feature with a maximum at 301.3 eV (Fig. 8). Thesefeatures may be attributed to transitions to the unoccu-pied zr* and o* antibonding molecular orbitals of aCOI- cluster, respectively. The C K edge from the COI-anion is illustrated for calcite, siderite, and two membersof the hydrotalcite group, hydrotalcite and desautelsite(Fig. 8). The extra structure on the o* peak from calciteis probably due to scattering from shells surrounding thecarbonate cluster. For comparison, Figure 8 illustrates theC K edge for evaporated amorphous C, graphite, and di-amond. Their edges are consistent with previously pub-

o* cK

Amorphouscarbon

420 GARVIE ET AL.: ENERGY LOSS NEAR-EDGE FINE STRUCTURE

100 1 1 0 140130120

280 290 300 310Energy Loss (eV)

Fig. 8. C Kedge from minerals containing the carbonate an-ion (calcite, siderite, desautelsite, and hydrotalcite) and threeallotropes of C (amorphous C, graphite, and diamond).

lished XANES and ELNES for these materials (Kincaidet a1., 1978; Rosenberg et al., 1986; Berger et a1., 1988;Weng et al., 1989a; Bernatowicz et al., 1990; Batson andBruley, 199 1; Katrinak et al., 1992). The fullerenes Cuoand Cro also have C K edge shapes similar to that fromgraphite but with differences due to the molecular natureof the Cuo and Cro molecules (Terminello et al., l99l;Shinohara et al., l99l). Amorphous C and graphite havebasically the same edge shapes, with an initial peak at285 eV and a second more intense feature at ca. 290 eY',the similarities can be explained in terms of the similar-ities in their bonding. Diamond has a different edge shapefrom graphite, with a peak maximum at 292.6 eV. Thedifferences between the C K edges in graphite and dia-mond may be explained by the presence of sp' bondingin graphite, which results in a peak at 285 eV, identifiedas transitions to the zr* molecular orbital, and a secondpeak at 290 eY, due to transitions to o* orbitals (Wenget al., 1989a). In diamond the bonding between the Catoms can be explained in terms of tetrahedrally directedsp3 hybrid orbitals, and the first peak is identified as aris-ing from transitions to molecular orbitals of o* character(Weng et al., 1989a).

Energy Loss (eV)

Fig. 9. Si Ir., edges from three silicates containing SiO" tet-rahedra (qtartz, talc, and nontronite). For comparison, the SiZ, edges from p-SiC and Si are also shown.

The presence of Si in minerals is almost exclusivelyassociated with the SiOo tetrahedron. These tetrahedramay join together to form a large number of crystallinesilica polymorphs (Keskar and Chelikowsky, 1992), aswell as amorphous and glassy phases. The silicates arealso characterized by the SiOo tetrahedron: these tetra-hedra are present in simple isolated tetrahedra, as in theneosilicates, but also join together to form a huge numberof structures (Liebau, 1985). A common cation finger-print observed in minerals is the Si Ir., edge from totSi.

The Si lr., edges from quartz, talc, and nontronite areillustrated in Figure 9, together with the Si Ir., edges fromSi and SiC. The Si Ir.. edges from quartz and the twoclay minerals exhibit very similar edge shapes, character-ized by two initial sharp peaks, peaks A (106.3) and B(108.4 eV), followed by a sharp peak C (115.0) and abroad peak D (ca. 133 eV). Peak A, from q:uarlz, is splitinto two peaks separated by 0.45 eV; a similar splittinghas been recorded by high-resolution XANES (Li et al.,1993) and EELS (Garvie, unpublished data). The ob-served structures have been assigned to the unoccupiedmolecular orbitals of a SiOX- tetrahedron (McComb etal., l99l;, Li et al., 1993). Thus based on the molecularorbital assignments in McComb et al. (1991) and Li etal. (1993), the four peaks from the quartz Si Zr., edgehave been assigned to the following unoccupied molec-

GARVIE ET AL.: ENERGY LOSS NEAR-EDGE FINE STRUCTURE 421

ular orbitals: peak A : a, (sJike), peak B : t, (pJike),peak C : primarily e (d-like), and peak D : tz (dJike).This Si L,r-edge shape is seen in the majority of silicates,although a recent study of the neosilicates (McComb etal., 1991, 1992) illustrates that the fingerprint techniquemust be used with care. Relative to quartz, the Si I*ELNES from zircon (McComb et al., 1992) shows a dis-tinctly different edge shape, which is related to the highlydistorted SiOi tetrahedra. Octahedrally coordinated Si,which rarely occurs in minerals, is present in stishovite.The Si l,,., absorption edge from stishovite (Li et al., 1993)is considerably different in shape from the Si Ir,3 ELNESin minerals containing regular or distorted SiOo tetrahe-dra, which implies that the edge shape may be used as aguide to the coordination of the Si. The different Si [.,ELNES from minerals containing regular or distorted SiOotetrahedra could possibly be used as a guide to the sitesymmetry of the Si, although further experimental workwill be needed to confirm this. For comparison, the Lr.edges from SiC and Si are illustrated in Figure 9. Thedifferences in the edge shapes between the Si Ir,. edgesfrom the silicates and the Si and SiC can be understoodin terms of the differing unoccupied densities of states(Weng et al., 1989b, 1990), whereas the different posi-tions ofthe edge onsets are related to the differing effec-tive charges on the Si atoms in the different materials.

Mounr-rNc coRE-r-ross EDGES

A host of theoretical techniques have been used tomodel EELS core-loss edges, which can be divided rough-ly into two types (de Groot et al., 1993): the local densityapproximation (LDA) and the ligand field multiplet (LFM)approaches. The former includes methods such as bandstructure, multiple scattering, and molecular orbital cal-culations (see Brydson, 199 I , for an introduction to thesemethods as applied to ELNES data). These are suitablefor weakly correlated systems where the interaetion of theexcited electron with the core hole created by the exci-tation process is small. The LFM method explicitly takesaccount of electron-electron and electron-core hole inter-actions in strongly correlated systems and has been mostsuccessfully used to model the 3d transition metal lr.,edges (de Groot et al., 1990a, 1990b; van der Laan andKirkman, 1992; Cressey et al., 1993) and M,., edges (vander Laan, l99l). Two contrasting methods for modelingthe different core-loss edges in MnO are illustrated: themultiple scattering method to model the O K edge andthe atomic multiplet method for the Mn Zr., edge.

Multiple scattering calculations

The multiple scattering calculations were performedusing the ICXANES computer code of Vvedensky et al.(1986) using full multiple scattering. Multiple scatteringtheory is based on the interference between the outgoingejected electron wave and the returning backscatteredelectron wave. Briefly, the transition rate ofexcitation iscalculated within the dipole approximation, using essen-tially an atomic contribution modified by multiple elastic

530 540 550 560 570Energy Loss (eV)

Fig. 10. Experimental O K edge from MnO compared withthe results of rhe multiple scattering calculations for differenrcluster sizes. The results for shells 3 and 5 are not illustrated, asthey had little effect on the final plot.

scattering of the excited electron by the surrounding at-oms. This approach, while formally equivalent to a Kor-ringa-Kohn-Rostocker band structure calculation, is con-siderably more flexible, allowing the inclusion of excitedatom potentials to account for the effects ofthe excitationprocess. The scattering properties of the various atomsare described by phase shifts that may be calculated bynumerical integration of the Schrddinger equation, as-suming a muffin-tin form for the atomic potentials in thecrystal, obtained by the simple superposition of neutralatomic charge densities. The multiple scattering calcula-tion is performed in real space by dividing the clusterinto shells of atoms surrounding the central excited atom.The multiple scattering is solved within each shell in turn,and the results are then combined to reassemble the finalcluster, use being made of any structural symmetry.

Many studies have illustrated the applicability of themultiple scattering approach in modeling ELNES andXANES spectra (Oizumi et al., 1985; Lindner et al., 1986;Brydson et al., 1988, 1992b; McComb et al., 1992; Sipr,1992; Kurata et al., 1993). The calculations for the O Kedge in MnO (Fig. l0) were performed for seven shells,extending to a cluster radius of I 1.86 au (where I au :

0.52918 A) around an O atom. The core hole was ac-counted for by the use of the (Z + l)* approximation for

422

636 638 640 642 644 646Energy Loss (eV)

Fig. I l. Comparison of the Mn L, edge from manganosite(r61Mn'?+ ) with the calculated Mn d5 atomic multiplet spectra forMn2+ in O, symmetry with crystal field strengths of 0.5-2.0 eV.

the central O atom in the cluster. In agreement with thefindings ofLindner et al. (1986) and Rez et al. (1991) onMgO, the dominant scatterers in oxide structures are theO atoms, and it is possible to reproduce many of themajor features using a cluster that ignores the Mn atoms.The following cluster surrounding a central O atom wasused: shell I : six Mn atoms at 4.19 au, shell 2 : 12 Oatoms at 5.93 au, shell 3 : eight Mn atoms at 7.26 au,shell 4 : six O atoms at 8.38 au, shell 5 :24 Mn atomsaI 9.37 au, shell 6:24 O atoms at 10.21 au, and shell 7: 12 O atoms at 11.86 au. By varying the cluster size,the features in the experimental O Kedge can be assignedto specific scattering events. The six main features of theexperimental O K edge, labeled a-f, are successfully re-produced by the multiple scattering calculation (Fig. 10).The first peak, a, can be adequately reproduced usingsolely the first coordination shell (shell l) of Mn. Thedominant scattering events are d-like scattering from Mnand pJike scattering from O. Hence this feature reflectstransitions to unoccupied O 2p states hybridized with thenarrow metal 3d band that is highly localized around themetal atom sites. This region therefore reflects the co-valence in the transition metal oxides. This is in agree-ment with the previous work on the O K edges of 3dtransition metal compounds (de Groot et al., 1989; Kura-Ia eI al.,1993). Peak b appears to arise only after inclu-sion ofa third O shell (sixth coordination shell) indicatingthat this region is also sensitive to longer range effects.

GARVIE ET AL.: ENERGY LOSS NEAR-EDGE FINE STRUCTURE

The sharp intense peak c arises because of intrashell mul-tiple scattering within the first O coordination shell andreflects transitions to unoccupied O 2p states hybridizedwith the more delocalized Mn 4s and 4p states. Peaks dand f are adequately reproduced after the addition of athird O coordination shell (shell 6) and arise from mainlyintershell multiple scattering. Finally, the large broad peake is modeled using the first O coordination shell and maybe adequately reproduced using solely single scattering.Such a basic resonant feature is known as a multiple scat-tering resonance, and its energy position follows the re-lationship E e t/nz (where ,E is the energy of the featureabove the edge onset and R is the O-O distance). Thussuch multiple scattering resonances provide a means ofdetermining bond lengths and bond length variations(Kurata et al., 1993). Multiple scattering resonances arisesince certain ions (notably the oxide anion) are relativelystrong backscatterers and give rise to a potential barrierfor electron scattering.

Atomic multiplet calculations

The 3d transition metal L23 edges cannot be modeledin terms of a one particle density of states because of thelarge electron correlation effects in the partially filled 3dorbitals and the strong interaction with the 2p core holecreated during the excitation process. The LFM approachhas been used to model the Ir., edges in a number of 3dtransition metals and compounds (de Groot et al., 1990a,1990b; van der Laan and Kirkman, 1992). The theoret-ical Mn2* Lr., data presented here were calculated by vander Laan and Kirkman (1992\ and take account of thefollowing effects in order to calculate the 2p63d' -2psJfl'+t dipole transitions: 2p-3d and 3d-3d Coulomband exchange interactions, 2p and 3d spin-orbit interac-tions, and the crystal field acting on the 3d states. Al-though transitions 2p - 4s are dipole allowed, they havebeen neglected in the calculations because the 2p - 36transitions dominate over the transitions to 4s states. The3d-3d and 2p-3d two particle interactions define theground state of the transition metal ion and split the EELS-XAS final state into a large number of configurations. Thecalculated atomic multiplet spectrum for the 3d" com-pounds (r > 0) typically consists of hundreds of unre-solvable final states spread over typically 15 eV for theMn Ir., edges. In order to compare the multiplet spectrawith experimental ones, the multiplet lines are broadenedto take into account effects such as lifetime and vibra-tional broadening and hybridization.

The validity of the atomic multiplet method to modelthe experimental Lr., edge is illustrated in Figure I l. Inthis figure the theoretical multiplet spectra for Mn2+ inO, symmetry with crystal field values of 0.5, l, 1.5, and2 eV are compared with the Mn I, edge from MnO. Theagreement between the experimental L, edge and the the-oretical L, edge with a crystal field value of I eV is ex-cellent. All the features on the experimental edge are re-produced in the theoretical spectrum. This crystal fieldvalue of ca. I eV, determined by comparing the experi-

ExoerimentalMn L3 edge

GARVIE ET AL.: ENERGY LOSS NEAR-EDGE FINE STRUCTURE 423

mental L, edge with the theoretical ones, is also in goodagreement with the l0Dq of 1.2 eY (Pratt and Coehlo,1959) determined for MnO by optical spectroscopy. Thegood agreement between the theoretical and experimentalMn I. edge for MnO is in part due to the dominantlyionic bonding between Mn and O. For more covalentcompounds such as Mn-O bonding in pyrolusite, the ex-perimental and theoretical edges show quite differentstructures (Garvie and Craven, 1993, 1994). Most theo-retical atomic multiplet spectra calculations assume sole-ly ionic-type bonding, although introducing terms in thecalculations for covalence can be used successfully tomodel more covalent compounds (van der Laan et al.,1986), but this is more difficult.

Errncrs TNDUCED By rHE ELEcTRoN BEAM

In minerals, the main effects of electron beam irradi-ation are element mobility and amorphization (Champ-ness and Devenish, 1992). These effects are most notableduring EDX in a TEM, where element mobility and lossis often a major problem, particularly with the longcounting times required for EDX. These problems can beespecially acute for the clay minerals (van der Pluijm eta1., 1988), which often contain mobile elements such asK, as well as HrO and OH. In many instances the effectsinduced by the electron beam change the nature of thesolid and therefore also the environment of specific at-oms. These changes can be followed using an atom-spe-cific probe such as PEELS. PEELS has been used to fol-low the changes occurring during the electron irradiationof solids, such as the transformation of calcite to CaO(Walls and Tenc6, 1989;Murooka and Walls, 1991) andthe transformation of BOo to BO. in hydroxyborates (Saueret al., 1993). Egerton et al. (1987) provided a good over-view of PEELS-related studies of chemical changes in-duced by the electron beam. In materials containing 3dtransition elements, the changes in the edge shape areconsistent with changes in the valency of the transitionmetal atom. This has been studied in detail in the mineralasbolan (Garvie and Craven, 1994), where it was possibleto follow the reduction of the Mn4* to Mn2+ with intenseelectron beam irradiation. During irradiation the Mn .L..,edge changes its shape and position, consistent with in-creasing proportions of Mnr+ and finally Mn'?+.

Although the majority of materials containing Mn inthe oxidation states ofIII or IV were found to be reducedto lower oxidation states, the opposite has occasionallybeen observed. In sussexite, Mn2+BO,(OH), and gano-phyllite, the Z, edge initially changed in shape and energyfrom an Mn2+ to an Mn3* I, edge. Further irradiationcaused the Mn3* edge to revert back to an Mn2* edge.Similar behavior was observed in all the phyllosilicatesthat contained Fe2*. After a short period of irradiation,the Fe2+ component was completely transformed to anL, edge having the same energy and shape as I61Fe3+. Theoxidation of the Fe'7* in the phyllosilicates and the Mn'?+in sussexite and ganophyllite is presumably linked to theloss of H in the OH anions, which is expected to be very

mobile under the electron beam, and subsequent oxida-tion of the metal.

Acxxowr-nncMENTsWe are all g.rateful to the following for providing samples: G.C. Allen

(Interface Analysis Centre, University of Bristol, U.K.) for the syntheticjacobsite; D.C. Bain (Macaulay Land Use Research Institute, U.K.) forthe samples of volkonskoite and daphnite; J. Faithfull (Hunterian Mu-seum, U.K ) for samples of ganophyllite (M9609), hausmannite (M8097),

and silicon carbide; H.C.B. Hansen (Royal Veterinary and AgriculturalUniversity, Denmark) for the synthetic desautelsite; C.M.B. Henderson(Department of Geology, University of Manchester, U.K.) for the syn-thetic iron analogue of leucite; S. Guggenheim (University of Illinois atChicago) for the sample of norrishite; and H. Miura (Department of Ge-ology and Mineralogy, Hokkaido University, Japan) for the ramsdellite.We are grateful to G. van der laan (SERC Daresbury Laboratory, U.K.)for the Mn lr 3 multiplet data The authors would like to thank the SERCfor providing the Gatan 666 PEELS and for a postdoctoral research as-

sistantship (L.A.J G.) under grant GWGI2832.

RrnnnBNcns crrEDBailey, S.W. (1988) Structures and compositions of other trioctahedral

l:l phyllosilicates. Mineralogical Society of America Reviews in Min-

eralogy, 1 9, 1 69- 1 88.Batson, P.8., and Bruley, J. (1991) Dynamic screening ofthe core exciton

by swift electrons in electron-energyJoss scattering. Physical Review

l€tters, 67, 350-353Berger, S.D., McKenzie, D.R., and Martin, P J. (1988) EELS analysis of

vacuum arc-deposited diamond-like films. Philosophical MagazineIrtters. 57. 285-290.

Bernatowicz, T.J., Cibbons, P.C., and kwis, R.S. (1990) Electron energyloss spectrometry of interstellar diamonds The Astrophysical Journal,3s9,246-255

Bonnin, D., Calas, G., Suquet, H , and Pezerat, H. (1985) Sites occupancyof Fer* in Garfield nontronite: A spectroscopic study. Physics and

Chemistry of Minerals, 12,55-64.Brydson, R. (1991) Interpretation of near-edge structure in the electron

energyJoss spectrum Bulletin of the Electron Microscopy Society ofAmerica. 21. 57-67 .

Brydson, R , Williams, B.G., Engel, W., Sauer, H,Zf-itler, E., and Thom-as, J.M. (1987) Electron energyJoss spectroscopy (EELS) and the elec-

tronic structure of titanium dioxide. Solid State Communications, 64,

609 -6 r2 .Brydson, R., Williams, B.G., Engel, W., Lindner, Th, Muhler, M, Schliigl,

R, Zeitler, E., and Thomas, J.M. (1988) Electron energy-loss sp€ctros-copy and the crystal chemistry of rhodizite. II. Near-edge structure.

Journal ofthe Chemical Society, Faraday Transactions l, 84, 631-646.Brydson, R., Sauer, H., Engel, W., Thomas, J M., Znitler, E., Kosugi, N.,

and Kuroda, H. ( I 989a) Electron energy loss and x-ray absorption spec-troscopy of rutile and anatase: A test of structural sensitivity. JoumalofPhysics: Condensed Matter, l, 797-812.

Brydson, R., Sauer, H., Engel, W., Thomas, J.M., and Zeitler, E. (1989b)

Co-ordination fingerprints in electron loss near-edge structures: Deter-mination of the local site symmetry of aluminium and beryllium in

ultrafine minerals. Journal of the Chemical Society, Chemical Com-municat ions. I 5. l0 l0- l0 l l .

Brydson, R., Sauer, H , and Engel, W . (1992a) Electron energy loss near-

edge structure as an analytical tool: The study of minerals. In M.M.

Disko. C.C. Ahn. and B. Fultz, Eds., Transmission electron energy loss

spectrometry in materials science, p. l3l-154 The Minerals, Metals

and Materials Society, Warrendale, Illinois.Brydson, R., Sauer, H., Engel, W., and Hofer, F. (1992b) Electron energy-

loss near-edge structures at the oxygen K-edges oftitanium (IV) oxygencompounds. Journal of Physics: Condensed Matter, 4, 3429-3437 .

Brydson, R., Garvie, L.A.J , and Craven, A.J. (1993) Zr,-edges of tetra-hedrally co-ordinated do transition metal oxoanions XOI . Journal ofPhysics: Condensed Matter, 5, 937 9-9392.

Buechner, U. (1975) The dielectric function of mica and quartz deter-

424

mined by electron energy losses. Journal ofPhysics C: Solid State Phys-ics, 8, 2781-2787 .

Buseck, P.R., and Sel{ P (1992) Electron energy loss spectroscopy (EELS)and electron channelling (ALCHEMI). Mineralogical Society of Amer-ica Reviews in Mineralogy, 27,l4l-180.

Calas, G. (1988) Electron paramagnetic resonance. Mineralogical Societyof America Reviews in Mineralogy, 18, 513-571.

Calas, G., and Petiau, J. (1983) Coordination of iron in oxide glassesthrough high resolution K-edge spectra: Information from the pre-edge.Solid State Communications- 48. 625-629.

Cardile, C.M., and Brown, IW.M. (1988) An 5?Fe Mdssbauer spectro-scopic and X-ray diffraction study of New Zealand glauconite. ClayMinerals. 23. 13-25.

Champness, P.E., and Devenish, R.W (1992) Radiation damage in sili-cate minerals: Implications for AEM. In A. I-6pez-Galindo and M.I.Rodriguez-Garcia, Eds., Electron microscopy 92, vol. II: Materials sci-ences, p. 541-545. Secretariado de Publicaciones de la Universidad deGranada, Granada, Spain.

Colliex, C., Manoubi, T., Gasgnier, M., and Brown, L.M. (1985) Nearedge fine structures on electron energy loss spectroscopy coreJoss edg-es Scanning Electron Microscopy, Il,489-512

Cramer, S.P., de Groot, F.M F., Ma, Y, Chen, C.T., Sette, F., Kipke,C.A., Eichhorn, D M., Chan, M.K., Amstrong, W.H., Libby, E., Chris-tou, G., Brooker, S., McKee, V , Mullins, O.C , and Fuggle, J.C. (1991)Ligand field strengths and oxidation states from manganese Z-edgespectroscopy. Journal of the American Chemical Society, I 13, 7 937 -7940.

Cressey, G., Henderson, C.M.B., and van der l-aan, G (1993) Use ofZ-edge X-ray absorption spectroscopy to characterise multiple valencestates of 3d transition metals: A new probe for mineralogical and geo-chemical research. Physics and Chemistry of Mineral, 20, I I I - I 19.

de Groot, F.M.F, Grioni, M., Fuggle, J.C., Ghijsen, J., Sawatzky, G.A.,and Petersen, H. (1989) Oxygen ls x-ray-absorption edges of transi-tion-metal oxides. Physical Review, B40, 5715-5723.

de Groot, F.M F, Fuggle, J.C., Thole, B.T., and Sawatzky, G.A. (1990a)L,. x-ray-absorption edges ofd0 compounds: K*, Ca'z*, Sc3*, and Ti4*in O, (octahedral) symmetry. Physical Review, B4l, 928-937.

-(1990b) 2p x-rcy absorption of 3d transition-metal compounds:An atomic multiplet description including the crystal field. Physics Re-view, B42, 5459-5468.

de Groot, F.M F, Abbate, M., van Elp, J., Sawatzky, G.A., Ma, Y.J.,Chen, C.T., and Sette, F. (1993) Oxygen ls and cobalt 2p x-ruy ab-sorption of cobalt oxides. Journal of Physics: Condensed Matter, 5,2277-2288.

Dunn, P.J., Peacor, D.R., Nelen, J.E., and Ramik, R.A. (1983) Gano-phyllite from Franklin, New Jersey; Pajsberg, Sweden; and Wales: Newchemical data. Mineralogical Magazine, 47, 563-566.

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

Egerton, R.F, Crozier, P.A., and Rice, P. (1987) Electron energyJossspectroscopy and chemical change. Ultramicroscopy, 23, 305-312.

Garvie, L.A J., and Craven, A.J (1993) Determination of the crystal fieldstrength in manganese compounds by parallel electron energy loss spec-troscopy. In A J. Craven, Ed., Proceedings of the Institute of PhysicsElectron Microscopy and Analysis Group, vol. 138, p. 3l-34. InstituteofPhysics Conference Series, Bristol, U.K

-(1994) Electron beam induced reduction of Mna* in manganeseoxides as revealed by parallel EELS. Ultramicroscopy, in press.

Goodman, B.A., and Bain, D.C (1979) Miissbauer spectra of chloritesand their decomposition products. Proceedings ofthe International ClayConference 1978, p. 65-74.

Gtiven, N. (1988) Smectites. Mineralogical Society of America Reviewsin Mineralogy. 19. 49'7 -559.

Hawthorne, F C (1988) Miissbauer spectroscopy. Mineralogical Societyof America Reviews in Mineralogy, 18,255-340.

Hofer, F., and Golob, P. (1987) New examples from near-edge fine struc-tures in electron energy loss spectroscopy Ultramicroscopy,21,379-384.

Ildefonse, P., Calas, G., Kirkpatrick, R.J., Montez, B., Flank, A.M, andLagarde, P. (1992) Local environment of aluminium in amorphous

GARVIE ET AL.: ENERGY LOSS NEAR-EDGE FINE STRUCTURE

alumino-silicates by XANES and MAS NMR. In Y.K. Klaraka andA.S. Maest, Eds., Proceedings ofthe 7th International Symposium onWater-Rock Interaction, p 153-158. Balkema, Rotterdam, Germany.

Joy, D.C. (1993) Electron specimen interactions. In A G Fitzgerald, B.E.Storey, and D. Fabian, Eds., Quantitative microbeam analysis, p. 213-245 SUSSP and IOP, Edinburgh and I-ondon.

Kaczmarek, J., and Wolska, E. (1993) Cation and vacancy distribution innonstoichiometric hausmanite. Journal of Solid State Chemistry, 103,387-393.

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.

Keskar, N.R., and Chelikowsky, J.R. (1992) Structural propenies ofninesilica polymorphs. Physical Review B, 46, l-13.

Kincaid, B.M., Meixner, A.E., and Platzman, P.M. (1978) Carbon Kedgein graphite measured using electron-energyJoss spectroscopy. PhysicalReview lrtters, 40, 1296- 1299.

Koch, E.E., Juenet, Y., and Himpsel, FJ. (1985) High-resolution softx-ray excitation spectra of 3d-metal phthalocyanines. Chemical PhysicsI€ t t e$ . 116 .7 -11 .

Koster, A.S., and Mendel, H. (1970) X-ray KB emission spectra and en-ergy levels ofcompounds of3d-transition metals: I Oxides Journal ofPhysics and Chemistry ofSolids, 3l,25ll-2522.

Knshnan, K.M. (1990) Iron 2,, near-edge fine structure studies Ultrami-croscopy, 32,309-311.

Krivanek, O.L. ( I 989) Improved parallel-detection electron-energy-lossspectrometer Ultramicroscopy, 28, I 18-121.

Krivanek, O.J., and Paterson, J H (1990) ELNES of3d transition-metaloxides. I. Variations across the periodic table. Ultramicroscopy, 32,313 -318 .

Krivanek, O.L., Ahn, C.C., and Keeney, R B (1987) Parallel detectionelectron spectrometer using quadrupole lenses. Ultramicroscopy, 22,103 - l 16 .

Kurata, H., Ishizuka, K., and Kobayashi, T. (1988) Near edge structurein electron energy loss spectra of chromium trioxide intercalated intographite and some chromium oxides. Bulletin of the Institute of Chem-ical Research, Kyoto University, 66, 57 2-57 9.

Kurata, H., Nagai, K., Isoda, S., and Kobayashi, T. (1990) ELNES of ironcompounds. In L.D. Peachey and D.B. Williams, Eds., Proceedings ofthe XIIth Intemational Congress for Electron Microscopy, vol. 2: An-alytical Sciences, p- 28-29. San Francisco Press, San Francisco, Cali-fornia.

Kurata, H., LefBvre, E., Colliex, C., and Brydson, R. (1993) ELNES struc-tures in the oxygen iK-edge spectra oftransition metal oxides. PhysicalReview. 847. 13763-13768

I-ever, A.B P (1984) Inorganic electronic spectroscopy: Studies in phys-ical and theoretical chemistry, vol. 33 (2nd edition), 863 p. Elsevier,Amsterdam

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

Liebau, F. (1985) Structural chemistry ofthe silicates: Structure, bonding,and classification, 347 p. Springer-Verlag, Berlin.

Lindner, Th., Sauer, H., Engel, W., and Kambe, K. (1986) Near-edgestructure in electron-energy loss spectra of MgO. Physical Review B,33,22-24.

Manceau, A , Gorshkov, A.I., and Drits, V.A. (1992) Structural chemistryof Mn, Fe, Co, and Ni in manganese hydrous oxides. L lnformationfrom XANES spectroscopy. American Mineralogist, 77, 1133-1143.

Marusak, L.A., Messier, R., and White, W.B. (1980) Optical absorptionspectrum of hematite, a-Fe,O, near Ir to UV. Journal of Physics andChemistry of Solids, 41, 981-984.

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

McComb, D.W., Brydson, R., Hansen, P.L., and Payne, R.S. (1992) Qual-itative interpretation ofelectron energy-loss near-edge structure in nat-ural zircon. Journal of Physics: Condensed Nfalter, 4, 8363-837 4.

McGibbon, A.J. (1991) The application of PEELS on a STEM to theproblem ofinert gas bubbles in solids. In F.J Humphreys, Ed., Pro-

ceedings of the Institute of Physics Electron Microscopy and AnalysisGroup conference, p. 109-1 12. Institute of Physics Conference Seriesno 1 19, Institute of Physics, Bristol.

Miura, H., Kudou, H., Choi, J H , and Hariya, Y. (1990) The crystalstructure of ramsdellite from Pirika mine. Joumal of the Faculty ofScience, Hokkaido University, Series IV, 22, 6ll-617.

Murooka, Y., and Walls, M.G. (1991) Beam damage on anisotropic ma-terials (CaCO,) in STEM. In F.J Humphreys, Ed., Proceedings of theInstitute of Physics Electron Microscopy and Analysis Group confer-ence, p. 337-340. Institute ofPhysics Conference Series no. ll9, In-stitute of Physics, Bristol.

Oizumi, H., Fujikawa, T , Ohashi, M., Maezawa, H., and Nakai, S. (1985)F-K XANES studies ofalkaline-earth fluorides. Journal ofthe PhysicalSociety of Japan, 54, 4027 -4033.

Otten. M.T.. and Buseck. P R. (1987) The oxidation state of Ti in horn-blende and biotite determined by electron energy loss spectroscopy,with inferences regarding the Ti substitution Physics and Chemistryof Minerals. 14. 45-51

Otten, M.T., Miner, B., Rask, J.H , and Buseck, P R (1985) The deter-mination of Ti, Mn and Fe oxidation states in minerals by electronenergy-loss spectroscopy. Ultramicroscopy, 18, 285-290.

Paterson, J.H.. and Krivanek, O J. (1990) ELNES of 3d transition-metaloxides. II. Variations with oxidation state and crystal structure. Ul-tramicroscopy, 32, 3 19-325

Pratt, G.W.J., and Coehlo, R. (1959) Optical absorption of CoO and MnOabove and below the Neel temperature. Physics Review, I I 6, 28 I -286.

Rao, C N R., Sarma, D.D., Vasudevan, S , and Hegde, M.S. (1979) Studyof transition metal oxides by photoelectron spectroscopy Proceedingsofthe Royal Society London, 4367,239-252.

Rask, J H , Miner, B A., and Buseck, P.R. (1987) Delermination of man-ganese oxidation states in solids by electron energyJoss spectroscopy.Ultramicroscopy , 21, 321-326.

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

Robertson, J. (l 986) Amorphous carbon Advances in Physics, 35, 3 17 -

374.Robertson, J., and O'Rielly, E.P. (1987) Electronic and atomic structure

ofamorphous carbon. Physical Review, B35, 2946-2957.Rosenberg, R.A., Love, P.J., and Rehn, V. (1986) Polarization-dependent

C(K) near-edge x-ray-absorption fine structure ofgraphite. Physical Re-view B. 33. 4034-4037

Rossman, G.R. (1988) Optical spectroscopy. Mineralogical Society ofAmerica Reviews in Mineralogy, 18,2O7-254.

Sauer, H., Brydson, R., Rowley, P N , Engel, W., and Thomas, J.M. (1993)Determination of coordinations and coordination-specific site occu-pancies by electron energyJoss spectroscopy: An investigation of bo-ron-oxygen compounds. Ultramicroscopy, 49, 198-209.

Shinohara, H., Sato, H., Saito, Y., Tohji, K., Matsuoka, I., and Udagawa,Y ( I 99 I ) Carbon r(-edge X-ray absorption near-edge structures of solidC,n. Chemical Physics I-etters, 183, 145-148.

Sipr, O. (1992) Real-space multiple-scatteringanalysis of x-ray absorptionnear-edge K spectra of Cu,O and CuO. Journal of Physics: CondensedMatter. 4. 9389-9400

'fafts, J., and Zhu, J. (l 982) ELNES, a potential technique in the studies

of focal atomic arrangements. Ultramicroscopy , 9, 349-354.

425

Terminello, L.J., Shuh, D.K., Himpsel, F.J., Lapiano-Smith, D.A , Stdhr,

J , Bethune, D.S., and Meijer, G. (1991) Unfilled orbitals of Cuo andC," from carbon K-shell X-ray absorption fine structure. ChemicalPhysics Irtters, 182, 49 l-496.

Thole. B.T., and van der l-aan, G. (1988) Branching ratio in x-ray ab-sorption spectroscopy. Physical Review, 838, 3158-3171.

Tyrna, P.L., and Guggenheim, S. (1991) The crystal structure of nomsh-ite, KIiMnl*Si,O,.: An oxygen-rich mica American Mineralogist, 76,266-27 | .

van der laan, G. (1991) M23 absorption spectroscopy of 3d transirion-metal compounds. Journal of Physics: Condensed Matter, 3, 1443-7454.

van der Laan. G., and I(irkman, I.W. (1992) The 2p absorption spectra

of 3d transition metal compounds in tetrahedral and octahedral sym-metry. Journal ofPhysics: Condensed Matter,4, 4189-4204.

van der Laan, G.,Zaanen, J., Sawatzky, G.A., Karnatak, R., and Esteva,

J.-M. (1986) Comparison of x-ray absorption with x-ray photoemission

ofnickel dihalides and NiO. Physical Review, B33, 4253-4263.van der l-aan, G., Pattrick, R.A.D, Henderson, C.M B., and Vaughan,

D.J.,1992]' Oxidation state variations in copper minerals studied with

Cu 2P X-ray absorption spectroscopy. Journal of Physics and Chem-istry ofSolids, 53, I 185-l 190.

van der Pluijm, B.A., I-ee, J.H., and Peacor, D.R. (1988) Anall'tical elec-tron microscopy and the problem of potassium diffusion. Clays andClay Minerals, 16. 498-504

Vvedensky, D.D., 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.

Walls, M G., and Tenc6, M.G. (1989) EELS study of beam-induced de-

composition of calcite in the STEM. In P J. Goodhew and H Y. Elder,Eds., Proceedings of the Institute of Physics Electron Microscopy andAnalysis Group and Royal Microscopical Society Conference, p. 255-258. Institute ofPhysics Conference Series no 98, Institute ofPhysics,Bristol.

Weng, X., Rez, P, and Ma, H (1989a) Carbon K-shell near-edge struc-ture: Multiple scattering and band-theory calculations. Physical Re-view.840. 4175-4178.

Weng, X , Rez, P., and Sankey, O.F (1989b) Pseudo-atomic-orbital bandtheory applied to electron-energy-loss near-edge structures. PhysicalReview, B40, 5694-5704.

Weng, X., Rez, P., and Batson, P E. (1990) Single electron calculationsfor the Si L.. near edge structure. Solid State Communications, 74,

l 0 l 3 - 1 0 1 5 .Wood, P R., and Urch, D.S. (1976) Valence-band X-ray emission spectra

of manganese in various oxidation states. Journal of the Chemical So-

ciety, Dalton Transactions, P art 3, 247 2-247 6.Yang, K., and Seccombe, P K. (1993) Chemical variation of chromite in

the ultramafic cumulates of the Great Serpentine Belt, Upper Bingarato Doonba, New South Wales, Australia. Canadian Mineralogist, 31,7 5-87 .

MaNuscnrm REcErwD Jut-v 2, 1993Meroscnrvr AccEr'rED Dpcevssr 29, 1993

GARVIE ET AL.: ENERGY LOSS NEAR.EDGE FINE STRUCTURE