Gmchimica d Cosmochimicn Acfo Vol. 52. pp. 1641-1648 Copyright 8 1988 Pergamon Press pk. Printed in U.S.A.

OOl6-7037/88/$3.00 + .Hl

Aspects of silicate surface and bulk structure analysis using X-ray photoelectron spectroscopy (XPS)

MICHAEL F. HOCHELLA, JR. and GORDON E. BROWN, JR.

Department of Geology* and Center for Materials Research, Stanford University, Stanford, CA 94305, U.S.A.

(Received October 2 1, 1987; accepted in revised form March 1 I, 1988)

Abstract-X-ray photoelectron spectroscopy (XPS) has been used to study the average local environment and polar- inability of oxygen in the near-surface region of crystalline and amorphous SiOz, NaAlS&O*, NaA1Si206, NaAlSiO,, and CaMgSi206, as well as crystalline MgzSiO, and rhyolitic and basaltic composition glasses. For these materials, the 0 1s chemical shift due to different oxygen environments is not as large as in previously studied alkali silicate and alkali aluminosilicate systems. However, it was found that the 0 1s peak width (measured as FWHM) for the minerals is proportional to the number of chemically distinct oxygen environments in each structure. Further, 0 Is FWHM values for the mineral and rock composition glasses suggest certain structural details that correlate well with previously proposed amorphous structure models. Taken together, these results suggest that the shape and width of the 0 1s line may be important in monitoring atomic-level structural changes that may occur on silicate mineral and glass surfaces during, for example, reactions with aqueous solutions. In addition, for the 0 1s line to be useful for bulk structural information, approximately the upper 60A of the sample must be representative of the bulk.

The 0 2s photopeak and X-ray induced 0 &,3Lz,, Auger line were also carefully studied, although they did not yield further structural information beyond that provided by the 0 1s line. However, the oxygen Auger parameters derived using the energy difference between the 0 1s and 0 KL~JL~,~ lines for the crystals and glasses in this study within the Na20-A&O,-Si02 system show a systematic trend with bulk composition. This trend reelects a change in the polarization energy (extra-atomic relaxation energy) for oxygen in these materials.

INTRODUCTION

X-RAYPHOTOELECTRONSPECTROSCOPY (XPS),sinceitwas introduced for general laboratory use around 1970, has be- come the most practical and useful spectroscopic tool that can yield both elemental and structural/oxidation state in- formation from the near-surface regions (the upper several tens of angstroms) of materials. In geochemical research, XPS- derived surface chemical information has been used exten- sively to gain insight into, for example, mineral dissolution (see e.g., PETROVIC et al., 1976; HOLDREN and BERNER, 1979;

SCHOTT et al., 198 1; S~HOTT and BERNER, 1983 and HOCH-

ELLA et al., 1988) and sorption reactions at mineral/water interfaces (see, e.g., COUNTS et al., 1973; BANCROFT et al., 1977; MURRAY and DILLARD, 1979 and DILLARD et al., 198 1). In addition, there has been some interest in developing XPS as a structural probe for silicates (e.g., YIN et a/., 197 1; ADAMS et al., 1972; HUNTRESS and WILSON, 1972; ANDER-

SON and SWARTZ, 1974 and WAGNER et al., 1981, 1982). Although most of the early studies met with only limited success, WAGNER et al. (1982) showed that the Auger param- eter, obtainable from XPS experiments, is a dependable vari- able in determining the structural state of 0, Al, and Si in certain aluminosilicates. Further, there has been considerable success in using XPS to help determine the structure of silicate glasses in the NaZO-SiOZ and Na20-A120,-Si02 systems (FENN and BROWN, 1974; BRUCKNER et al., 1976, 1978a,b, 1980; KANEKO and SUGINOHARA, 1977, 1978; JEN and KALI-

NOWSKI, 1979; SMETS and LOMMEN, 1981; LAM et al., 1980; KANEKO et al., 1983; ONORATO et al., 1985; TASKER et al., 1985 and GOLDMAN, 1986; for similar work on other silicate

compositions, see NAGEL et al., 1976; VEAL et al., 1982; PUGLISI et al., 1983, 1984; SMETS and KROL, 1984). It has

* Address correspondence to the Department of Geology.

1641

been shown in these studies that clearly resolvable features in the 0 1 s photopeak can be attributed to bridging and non- bridging oxygens. ONORATO et al. (1985), TAXER et al. ( 1985) and GOLDMAN ( 1986) have recently refined this anal- ysis by showing that, for glasses in the NazO-Al203-Si02 sys- tem, the bridging oxygen portion of the 0 1s photopeak should be fit with two peaks representing both the Si-0-Si and the Si-O-Al’” linkages. Using this more sophisticated ap- proach, the bridging to non-bridging oxygen ratios measured directly from the 0 1s photopeaks of these glasses closely agree with the same ratios calculated from the theoretical structural model for these glass compositions.

This study was undertaken to extend the usefulness of XPS as a structural probe for geologically important crystalline and amorphous silicates by evaluating the structural infor- mation obtainable from the 0 1s and 0 2s photopeaks, as well as the oxygen Auger parameter. To accomplish this task, we have collected XPS spectra from homogeneous mineral specimens with known compositions, including selected samples of quartz, albite, nepheline, jade&e, diopside, and forsterite. Clean and ordered surfaces are achieved by frac- turing these mineral standards under at least mid-range ultra- high vacuum conditions (lo-” torr) just prior to XPS analysis. In addition, six glass samples, including four with mineral compositions (albite, nepheline, jadeite, and diopside) and two with average rock compositions (rhyolite and basalt) were analyzed in the same way as the mineral standards. The results from the crystalline standards were used to interpret the XPS spectra obtained from the glasses.

It should be noted that the results from this study can be applied to the study of the bulk structure of silicates only when the upper 40-8OA of the material is representative of the bulk. This depth estimate comes from a recent reassess- ment of electron escape depths in silica which indicates that 95% of 1 keV electrons (the approximate energy of 0 1s

1642 M. F. Hochella, Jr. andG. E. Brown, Jr.

electrons excited with Al Ka X-rays) come from approxi- mately the top 60A of the sample (HOCHELLA and CARIM. 1988). The escape depth in other silicates should not vary by more than lo-2076 from this value. Otherwise, the results of this study can be applied to the study of structural changes that occur in the near-surface regions of minerals during, for example, reactions with aqueous solutions.

MATERIALS AND METHODS

The sources and compositions of the 12 crystalline and amorphous silicate samples used in this study are shown in Tables I and 2. The minerals and mineral com~sition glasses are very close to ideal stoi- chiometry unless otherwise noted.

Each sample was prepared by cutting the glass or crystal into a prism approximately IO mm long with a cross-section of 2.5 X 2.5 mm. The prism was notched one-third of the way along its length to facilitate fracture in ultra-high vacuum (see below). Finally, the sample was degreased by boiling in acetone and ethanol before mounting onto a specially adapted fracture stage and insertion into the XPS instrument.

The XPS instrument used in this study is a VG ESCALAB Mk II which is equipped with both Al monochromatic and non-mono- chromatic X-ray sources (Al Kol = 1486.6 eV). The non-monochro- matic X-ray source was used in this study because of dramatically higher X-ray intensities compared with the monochromatic source. This becomes important when observing weak lines such as the 0 2s photopeak, which has a photoion~zation cross-section 21 times less than that of 0 Is for Al Ka radiation (SCOFIELD, 1976, YEH and LINDAU, 1985). In addition, despite the width of the undiffiacted Al Kcu excitation line from the non-monochromatic source, the spectra of insulating materials have only slightly degraded energy resolution compared with the same spectra produced with monochromatic sources. Normally, monochromatic sources produce substantially improved spectral resolution due to a narrower incident X-ray line width, but these sources can also create charging instability on in- sulating surfaces which substantially degrades the resolution. Nev- ertheless, in order to obtain the very best energy resolution possible with the non-monochromatic source, the XPS scans in this study were collected with a step size of 0. I eV while reducing instrumental broadening by using a spectrometer pass energy of 20 eV and a narrow entrance slit to the hemispherical electron energy analyzer.

Each sample was fractured in the analytic chamber of the ES- CALAB (mid-lO-‘” torr range measured near the sample position) just before data collection commenced. The notch in the prism assured a relatively controlled fracture and a reasonably flat surface for anal-

Table 1. Minerale and mineral composition glasses used in this study and their sources.

NsA1Si,08

NaAISi,O,

albite from Taylor and Amelia, VA *rwn f 19798)

jadeite from Clear Taylor and Creek, San Benito Brovn (1979b1

co., CA

NaALSiO, nephellne from

Bancrafr. Ontario Tayior and

Brown (1979b33 Canada’

diopside from storringcon Tovnsllip, Hachella (1981)

Ontario, Canada

Fo,, from San car1o.s. AZ

*Approxidw~te composition of this mineral estimated from XPS survey npecrrum: (Na,.,K,.,=ao.1)A1,.,Sis,,012.

SiOp 71.63 54.00 Al&‘, 14.32 15.06

W,O 3.63 2.77 KzO 5.19 1.0x cao 1.70 10.63 w 0.38 17.03

Fe,03 3.60 100.45

o.00 100.5

#Classes analyzed by electron microprobe. SC: Wochella and Brown (1984) for details.

*Composition based on Nociwld’s (1954) biotite- hornblende granite. Synthesized by Hochella and Brown (19843.

**Composition based on an Fe-free version of Yoder and Tllley’s (1962) high-silica basalt. Synthe-

sized by Hochella (19Sl).

ysis. Survey scans (from O-1000 eV binding energy) were collected first, followed by narrow scans over the energy regions of interest (0 Is and 0 2s photopeaks, 0 KLL Auger line). The line positions and shapes remained constant during repeated analyses. Deconvolution routines for removing the X-ray line shape from the observed line shape were not used. All curve fitting described below was performed with a least squares Ga~sian/~~ntz~an computer curve fitting rou- tine.

Peak positions were charge corrected by assummg that the very weak Cls photopeak present for each sample was at 284.6 eV (WAG- NER ei al, 1982). The weak carbon signal is from adventitious carbon on the prism sides (a result of air exposure and residuals from organic solvents), as there is a small probability that photoelectrons coming from the prism sides near the fractured surface will make it into the electron energy analyzer. The corrected peak positions reported here using this simple charge referencing method are thought to be inter- nally consistent because the adventitious carbon on each sample should be the same. However, the absolute position of this adven- titious line is uncertain (SWIFT, 1982), and we estimate that the ab- solute energies of the lines reported here could be in error by up to 0.3 eV.

Other methods of charge correction were not used for various rea- sons. The gold dot method (e.g., see STEPHENSON and BINKOWSKI, 1976) of charge correction was not used due to potential problems which have been pointed out recently by KOHIKI and OKI (1985). They have shown that there can be significant variations in the binding energy of the Au 4f7,* line as a function of the thickness of the gold Iilm and the sample upon which it is deposited due to changes in extra-atomic relaxation between the gold and the substrate. We also attempted to perform charge correction by implanting small quantities of Ar at low energies and assuming a material independent energy for the Ar 2p,,, line as suggested by KOHIKI et al. (1983a,b). However, we found that our samples can suffer significant surface damage from even very light implantation due to differential sputtering, electric field stimulated migration, and surface mixing. We will address this latter subject in a separate publication.

The modified oxygen Auger parameters were carefully measured for each sample and the definition of this parameter and the method of measurement follows. The Auger parameter, iy, first described by WAGNER (1972, 1975), can be measured for any element as long as one photoelectron peak and one Auger peak for that element can be observed. It is defined as

CK = &!?(A) -. KE(P) (11

where K&A) is the kinetic energy of an X-ray induced Auger line and KE(P) is the kinetic energy of a photoelectron line for the same element observed in the same spectrum. Rewriting Eqn. (1) in terms of binding energy, and adding the photon energy, hv, in order to keep the Auger parameter positive and independent of the photon

Silicate surface structure 1643

energy used to collect the spectrum, we arrive at the modified Auger parameter, (Y’, where

a’ = ty + hv = BE(P) - u&!?(A) + hv (2)

and where &E(A) is the apparent binding energy of the measured Auger line and BE(p) is the binding energy of the measured photo- peak.

Because the Auger and modified Auger -meters are based simply on the difference in energy between an Auger and a photoelectron line for the same element in the same XPS spectrum, absolute peak positions, and therefore static charge corrections, are not needed. The accuracy of an Auger parameter depends onfy on the accuracy of the peak position measurements. Such measurements can easily be accomplished to within kO.05 eV; therefore, Auger parameters are dependable to +O. 1 eV or less. For this study, the relative positions of the 0 Is and 0 KI+3L2,J lines have been used to generate the oxygen modified Auger paranteters (Eqn. 2). The meaning and use- fulness of the Auger parameter is discussed below.

RESULTS

The 0 1s and 0 2s peak positions (charge corrected) and widths (measured as full width at half maximum, or FWHM) for the six crystalline and six amorphous materials used in this study are shown in Table 3. The modified oxygen Auger parameters are also listed for each sample.

The 0 1s peak positions for all samples are between 53 1 .O and 532.0 eV except for quartz (532.9 eV). These peak po- sitions are consistent with reported charge corrected positions for the 0 1s lines of most silicates (e.g., see ADAMS el al., 1972; ANDERSON and SWARTZ, 1974; TASKER et al., 1985; ONORATO, 1985). The 0 2s peak positions are slightly more variable and range from 23.3 to 24.8 eV except again for quartz (25.6 eV).

The 0 1 s photopeaks for all samples except jadeite, diopside, and diopside glass are symmetric or sl&htly asymmetric. The 0 1 s photopeaks for jadeite, diopside, and diopside glass show definite asym~et~, but do not have distinct shoulders or other clearly resolvable features. The peak widths vary from 1.7 eV for quartz to 2.9 eV for diopside and diopside glass (Fig. 1).

Although the signal-to-noise level of the 0 2s photopeaks is relatively poor (e.g., see Fig. 2), these peaks have a consistent shape from sample to sample and vary only slightly in width (3.2 eV for forsterite and rhyolite glass to 3.5 eV for diopside

Table 3. Positiona and widths for oxyi~~” le and 2s photoemisaion peaks md oxygen Auger parameterel.*

01s 028 0 All~er BE(@) PUiiM Be<&‘) WIW para%eer(eV)

qu.artz 532.9 1.7 25.6 1040.1 albire 531.0 2.1 24.7

:*: 1039.9

alblre glass 531.7 2.2 24.4 3:4 1040.0 jadefre 531.1 2.5 24.1 3.9 1039.8

jedere glass 531.6 2.1 24.2 nepheline 531.0 2.0 23.8

::: 1039.9 1039.7

neph. &Se 531.0 2.0 23.3 1039.6 di0pside 531.6 2.9 23.7

::a 1040.0

diop. 81USS 531.3 2.9 23.3 3.5 1040.0 forsterite 531.2 1.8 23.7 3.2 1039.6

rhyolite +SS# 532.0 2.3 24.8 3.2 1040.0 basalt 81aSS# 531.6 2.6 23.8 3.4 1039.8

*Peak positions have been adjusted uirh reference to the Cls line (sea terr) and peak widths are messured aa full width at half maximum (S’WHM).

#See Table 2 for idemification of them glasses.

f I

Bindlrtg energy (eV?

FIG. 1. Charge corrected 0 1 s photoemission lines (intensity scale in relative counts per second) for quartz and diopside showing the range of the 0 1s peak widths for all samples tested in this study. The peak widths are given as full width at half maximum (FWHM) and are in units of eV. Note that although the diopside line consists of a bridging and non-bridging component in a I:2 ratio, the Iine is only very slightly asymmet~c.

glass). Only jadeite (3.9 eV) falls outside this range. It should be noted that alkali and alkaline earth shallow core level pho- topeaks appear near the 0 2s line (Fig. 2). In fact, when Ca is present, the Ca 3p line partially overlaps the 0 2s line on its high binding energy side. Therefore, in order to obtain a width for the 0 2s for diopside (crystal and glass) and the basalt composition glass, the 0 2s-Ca 3p overlapped peaks were curve fit. Although the best fits obtained here are not unique solutions, the FWHMs of the 0 2s peaks for the Ca- bearing materials obtained in this way are consistent with those measured directly for the other samples (Table 3).

The X-ray induced oxygen KLz,~L~,~ Auger lines (apparent binding energy near 980 eV in these spectra) show very slight shape changes among the 12 samples. The modified Auger parameters listed in Table 3 vary slightly and range from 1039.6 eV for nepheline glass and forsterite to 1040.1 eV for quartz.

DISCUSSION

0 Is photopeak

The 0 1s photopeaks observed for jadeite and diopside do not have the easily resolvable features attributable to bridging and non-bridging oxygens as commonly recorded for glasses in the Na20-A&O-Si02 system (ONORATO et al., 1985; TAS- KER et al., 1985; and references therein). It is not possible to attribute this observation to a lack of energy resolution in our spectra; judging from the FWHM of the observed 0 1s peaks from compounds along the SiOz-NaAlO2 join measured in this and other studies, the energy resolution obtained here is equivalent to or better than previous work (ADAMS et al., 1972; BRUCKNER et al., 1978b; LAM et al., 1980; SMETS and LOMMEN, 198 1; TASKER et al., 1985; ONORATO et al., 1985). On the other hand, it is likely that the components of the 0 1s photopeaks for jadeite and diopside are much closer in energy than the bridging and non-boding com~nen~ of the 0 1s line for materials in, for example, the Na@-SiO2

1644 M. F. Hochella, Jr. and G. E. Brown, Jr.

.

f E

Na2p

/

J 4vi

K~P

FIG. 2. Unsmoothed XI’!3 spectrum of the low binding energy region for nepheline (intensity scale in relative counts per second). The 0 2s line is relatively weak and the shape is difficult to characterize irrespective of smoothing. Shallow core level lines from Na and K do not overlap the 0 2s position, but the Ca 3p line does. making it even more difficult to characterize.

binary system. ONORATO et al. (1985) showed that the energy separation between the bridging and non-bridging compo- nents of the 0 1s photopeak for a glass of composition 0.25Na20-0.75Si02 is approximately 2.0 eV. The energy separations of two Gaussian/Lorenztian peaks fit to the 0 1 s lines for jadeite and diopside are 1.2 and 1.4 eV, respectively. In addition, if the higher binding energy component for these two materials is assumed to represent the bridging oxygen as has been determined for the Na20-Al203-Si02 system, the bridging to non-bridging peak area ratios are close to 1:2 in accordance with these structures. As an example, Fig. 3 shows the best fit obtained for the 0 1s photopeak of jadeite. For jadeite, the bridging and non-bridging components have FWHMs of 1.7 and 1.9 eV, respeciively, and the area ratio is 1:2.4. The only constraints in the fitting were the FWHM of the bridging component which was fixed at 1.7 eV (the width of the quartz peak) and the relative position of the bridging component which was constrained to remain at higher binding energies than the non-bridging component.

Although ONORATO et al. (1985), TASKER et al. ( 1985), and GOLDMAN (1986) have found good reason to fit 3 com- ponents to 0 1s photopeaks in the sodium aluminosilicate system (one non-bridging and two bridging peaks), appro- priate positions and widths of all the possible components for 0 1s lines for silicates are generally not experimentally or theoretically known. Especially when the 0 1 s line is sym- metric or nearly so, fitting multiple components to a pho- topeak becomes rather arbitrary without constraints on both peak positions and widths. This is the case for the curve fitting shown in Fig. 3 for jadeite. There are two types of non-bridg- ing oxygens in jadeite (Table 4), and therefore this 0 1s en- velope should really be fit with three peaks. However, there is little or no independent information on how the two non- bridging components should be fit.

On the other hand, it is apparent for the mineral and min- eral composition glasses tested in this study that the FWHM of the entire 0 1s photopeak (i.e., without curve fitting) is

directly proportional to the number of chemically distinct oxygen coordination environments in each structure. A cor- relation of this type was first alluded to by YIN et al. (197 1) and ADAMS et al. (1972), but it was never developed. Table 4 shows, for the six mineral standards used in this study, the FWHM for each 0 1s line and a schematic representation of the oxygen coordination complexes from each structure. The six minerals are divided into three groups (designated I, II, and III) on this basis, i.e., with 1, 2 or 3 oxygen environ- ments within the structure. It should be emphasized that this number of oxygen coordination complexes is not based on the number of crystallographic oxygen sites, but simply on the number of chemically distinct coordination groups around oxygen. Therefore, quartz and forsterite are grouped together despite the fact that their crystal structures are un- related, and both have the narrowest 0 1s FWHMs (below 1.9 eV) because there is only one unique oxygen coordination complex in each structure. Albite and nepheline, with two unique oxygen complexes per structure, have 0 1 s FWHMs of 2.0 and 2.1 eV, respectively. Finally, jadeite and diopside, with three unique oxygen complexes, have 0 1s FWHMs above 2.4 eV.

The four mineral composition glasses measured in this study are arranged in Table 5 according to the same scheme used in Table 4, i.e., where the width of the 0 Is lines indicates the number of distinct oxygen coordination complexes which should exist. This grouping is reasonable according to what is known about each glass structure. Albite and nepheline composition glasses have framework structures which are to- pologically distinct from their crystalline equivalents (e.g., see SHARMA et al., 1978; TAYLOR and BROWN, 1979a), but they should nevertheless have the same number of distinct oxygen coordination complexes. Our results support this no- tion, with the 0 1s FWHM of the mineral/glass pairs differing by no more than 0.1 eV. On the other hand, jadeite glass is known to have a framework structure similar to that for albite and nepheline composition glasses with two distinct oxygen

527 529 531 533 5

Binding energy (eV)

FIG. 3. Observed 0 1s line from jadeite with computer fit bridging (br) and non-bridging (ribr) Gaussian components (&en&y scale in relative counts per second). The separation of the centers of the two components is 1.2 eV. The sum of the two components very closely overlies the observed line and is not shown. See text for further details.

Silicate surface structure 1645

Table 4. Oxygen coordination complexes, 01s FWHM, and group designation for mineral standsrds used in this study.

OXUP Mineral 01s designation standard FWHM

Oxygen coordination complexes

quartz 1.7 Si-0-Si

I MS I

forsterite 1.8 Y+"'

I albite 2.1 0

Si' 'Si b

Si' 'AIIV

II

nepheline 2.0

Na Na

Si-‘t)-_A1Iv I K

I I I jadeite 2.5 Si-0-Si Si-0-Na 0

I Si' 'ApI Na

III

Ca 4 Mg I

diopside 2.9 Si-0-Si Si-0-Ca A I Si’ ‘Ca

Ca MS

environments (TAYLOR and BROWN, 1979b; SHARMA et al., 1979; HOCHELLA and BROWN, 1985), while crystalline jadeite is a chain silicate with three distinct oxygen environments. In this case, the glass should have a significantly smaller 0 1s FWHM than the crystal, and in fact we observe that the glass 0 1s FWHM is smaller than that of the crystal by 0.4 eV. Therefore, as expected, jadeite composition glass falls within group II, while crystalline jadeite is in group III. For the eopside crystal/glass pair, evidence from X-ray radial distribution analysis (HOCHELLA, 1981) show no major dif- ferences between the short-range bonding configurations of the two materials. The XPS results from this study sup- port this observations, as both have the same 0 1s FWHM (2.9 eV).

The classification scheme presented above would seem to be more difficult to apply to the rhyolitic and basaltic com- position glasses because of their relatively complex compo- sitions and presumably larger but unknown number of ox- ygen coordination environments. Apparently, however, these coordination complexes do not produce components of the 0 1 s photopeak that are significantly different from the ones present in the minerals and mineral composition glasses al- ready ‘discussed. Therefore, the widths of the rhyolitic and basaltic 0 1s photopeaks can be rationalized in the context of the same scheme used for the simpler compounds. Radial distribution analysis of the rhyolitic composition glass indi- cates that its structure has primarily a framework character similar to that for albite composition glass, although a few non-bridging oxygens are expected to be present as suggested from its composition (HOCHELLA and BROWN, 1984). Thus, it is not surprising that the rhyolite composition glass, with an 0 1s FWHM of 2.3 eV, falls in group II, but it is closer to group III than any other material examined in this study.

For the basaltic composition glass, the observed radial dis- tribution function (RDF) shows features that are intermediate between RDFs of feldspar and diopside composition glasses, suggesting that the structure of basalt composition glass has components of both a framework and a chain type structure (HOCHELLA, 1981). In fact, the 0 1s FWHM for the basalt glass (2.6 eV) is intermediate between those for the diopside and feldspar composition glasses.

It should be noted that the 0 1s peak width can be very sensitive to sample history. For example, the quartz 0 1 s line width will increase from 1.7 to around 2.0 eV by exposing the quartz surface to ethanol or acetone. This effect is pre- sumably due to oxygen-containing solvent residues on the surface and must be taken into account or avoided when performing a study of this type.

0 2s photopeak and 0 KL2.3L2.3 Auger line

The 0 2s photopeak and 0 IU2,3L2,3 Auger line are of interest because they are produced by electronic transitions that are either close to or involve valence levels; therefore,

Table 5. Classification of glasses in this study based 0" the groups defined in Table 4.

Group Glass Range of designation composition 01s n?iH

I

II 2.0-2.3

III caMgsi,o, basalt glass 2.6-2.9

i 646 M. F. Hochella, Jr. and G. E. Brown, Jr.

their shape and width variations from sample to sample may be more sensitive to structural variations than the 0 Is pho- topeak. The 0 2s line is within approximately 20 eV of the 02p valence level and the 0 KL2.,L2,X Auger line produces two final state holes in valence levels. However, as mentioned earlier. the variations in the shapes and widths of the 0 2s photopeaks and the 0 KL2,3L2,3 Auger lines are small for the samples examined in this study. The width of the 0 2s line follows only a weakly defined trend with atomic structure (group I, 3.2 eV; group II, 3.3-3.4 eV; and group III, 3.4- 3.9 eV; the only exception is quartz which has an 0 2s FWHM of 3.4 eV), and the very slight KLL line shape variations do not define any apparent trends. Part of the problem with the 0 2s lines is that the widths of these very weak peaks are difficult to measure accurately and also that they are subject to interference from other shallow level core lines as discussed earlier. Concerning the Auger line, WAGNER et (11. (1982)

showed that certain silicates do have distinctive oxygen KL2,3L2,1 line shape variations, but it is apparent from the present study that this line cannot be used consistently for silicate structural information.

It has been well establish~ (K~WALCZYK et al., 1974; WAGNER, 1975; THOMAS, 1980) that a change in the Auger parameter of any element is directly praportional to the change in the extra-atomic relaxation (also referred to as po- larization energy) for that element in the compound being measured. This relaxation involves the screening of the final state ion in the Auger electron emission process by electrons of neighboring atoms (not necessarily next-nearest neighbors) and/or by electrons in the conduction band. Within this study, the variation in the oxygen Auger parameter is most easily rationalized in the series of compositions along the SiOz- NaAIOz join which have the same framework structure type. Therefore, from this study, we have included quartz (Qt), albite and albite glass (Ab), jade&e glass (Jd), and nepheIine and nepheline glass (Ne). Along the Qt-Ab-Jd-Ne series, the oxygen Auger parameter systematically decreases (Fig. 4). Therefore, as more NazO and A&O3 are added to the silica framework, the screening of Auger-induced doubly charged oxygen ions becomes less efficient. However, despite this ap- parent sensitivity to compositional change, the oxygen Auger parameter does not seem to reflect structural variations within the samples studied. For example, the quartz and diopside oxygen Auger parameters are different by only 0. i eV, as are the jadeite and jadeite glass oxygen Auger parameters. Therefore, despite its ease of measurement and its apparent usefulness as a structurai parameter in some systems (WAG- NER et ai., 1982), the oxygen Auger parameter does not appear to be a useful structural indicator for several geologically im- portant silicates.

SUMMARY AND CONCLUSIONS

1) The 0 Is photopeaks of the 10 silicate mineral and mineral composition glass samples tested in this study (Tables 1 and 2) do not show clearly resolvable features which can be easily assigned to known oxygen species as has been done previously for glasses on the perakaline side of the Na20-

\. \ \

0 1039.6

I I

0 IO 20 30 40 50

Na,O + Ai203 (at. %I

FIG. 4. Modified oxygen Auger Parameter vs. composition for quartz (Qt), albite and albite composition glass (Ab), jadeite (Yd), and nepheline and nepheline composition glass (Ne). See text for discussion.

A1203-Si02 system. However, it was found for these samples that the full width at half maximum (FWHM) of the 0 1s photopeak is pro~~ion~ to the number of chemically dis- tinct oxygen bonding environments in each structure. The widths range from 1.7 eV for quartz which has only one oxygen coordination environment to 2.9 eV for diopside which has three (see Fig. 1, Tables 3 and 4). For the more complex rhyolitic and basaltic composition glasses, the 0 1 s widths can be rationalized in terms of the 0 1s widths of the mineral glasses to which they are most closely related struc- turally. Therefore, the width of the rhyolite glass 0 1s peak is only 0.1 eV from that of the albite glass, and the width of the basalt glass 0 1s peak is between those of the albite and diopside glasses.

2) Considering the kinetic energy of the 0 IS electrons which are being detected (approx. 1 keV) and the silicate materials from which they emanate, we can estimate that 95% of the signal is coming from within 60A of the surface (see HOCHELLA and CARIM, 1988, for details). Therefore, if the near-surface is representative of the bulk, the results of this study can be used for bulk structure analysis. On the other hand, these results can also be directly applied to, for example, the study of silicate surface interactions with so- lutions where a surface structural change could be verified by a change in the width or shape of the 0 Is peak. However. surface contamination must be controlled or carefully mon- itored before the shapes and widths of the 0 is line can be used for structural info~ation.

3) The correlation between the width of the 0 1 s tine and the range of oxygen bonding environments should not be extended to other silicates without further study. The energy differences among 0 1s lines for various oxygen bonding en- vironments are quite variable and can be very small. For example, the energy difference between the bridging and the non-bridging 0 Is components for glasses along the Na@- Si02 join is approximately 2 eV @though it is slightly variable depending on composition), while the difference between these 0 Is components in jadeite is closer to 1 eV. In the latter case, the non-bridging component (FWHM = 1.9 eV.

Silicate surface structure 1647

Fig. 3) represents two distinct nonbridging environments (Table 4). Judging from the width of 0 Is peaks which rep resent only one oxygen ending ~n~gurat~on (quartz and forsterite, Table 4), the two nonbridging 0 1s peaks for jadeite are probably within a few tenths ofan electron volt in energy.

4) The 0 2s and 0 KL2,3L2,3 lines were of limited use in this study despite the fact that they have the potential to be more sensitive to structural change because of their proximity to or direct involvement in valence levels. The 0 2s photopeak was observed in this study to be a very weak line due to its small cross-section for Al Ka radiation, and its shape is dif- ficult to define accurately due to relatively noisy spectra. The 0 KL2,3L2,3 line is inherently broad (typically between 5 and 6 eV FWHM), and this may help mask subtle line shape changes from sample to sample. The modified oxygen Auger parameter also did not prove to be a useful structural indicator for these samples, although this parameter does show the average change in the oxygen extra-atomic relaxation energy (polarization energy) along the series quartz-albite-jadeite- nepheline.

5) Further advancements in the use of XPS as a structural probe for silicates will come with improved spectral resolu- tion. As line widths become narrower, even closely spaced 0 Is components will become resolvable. Photopeak widths from conducting and semi-conducting materials can be most easily reduced with a reduction in the energy spread of the excitation source, but unstable charging phenomena on in- sulating surfaces generally keep photopeaks for these materials relatively broad even when using monochromatic X-ray sources, Therefore, higher resolution photoemission spectra will have to await the development of better charge stabili- zation and neutralization schemes.

Acknowledgements-This study was generously supported by the National Science Foundation through Stanford’s Center for Materials Research and NSF Grant EAR-85 13488 (Experimental and Theo- retical Geochemistry Program). S. Didziulis and K. Butcher expertly maintained the VG ESCALAB during the course of this project. In addition, the comments of an anonymous reviewer resulted in the clarification of several points in the text.

Editorial handling: P. C. Hess

REFERENCES

ADAMS I., THOMAS J. M. and BANCROFT G. M. (1972) An ESCA study of silicate minerals. Earth Planet. Sci. Lett. M&429-432.

ANDERSON P. R. and SWARTZ W. E., JR. (1974) X-ray photoelectron spectroscopy of some aluminosilicates. Inorg. Chem. 13, 2293- 2294,

BANCROF? G. M., BROWN J. R. and FYFE W. S. (1977) Quantitative X-ray photoelectron spectroscopy (ESCA): Studies of Ba*’ sorption on c&ite. Chem, Gebl. 19, 13il i44.

BRUCKNER R.. CHUN H.-U. and GORE~ZKI H. ( 1976) ~~~mination between bridging and non-bridging oxygen in’sodium silicate glasses by means of X-ray induced photoelectron spectroscopy (ESCA). Glastechn. Ber. 49, 2 1 I-2 13.

BRUCKNER R., CHUN H.-U. and GORETZK! H. (1978a) XPS mea- surements on alkali silicate and soda aluminosilicate glasses. Japan. J. Appl. Phys. 17,291-294.

BRUCKNER R., C&UN H.-U. and GORETZKI H. (1978b) Photoelectron spectroscopy (ESCA) on alkali silicate- and soda ~uminosili~te glasses. Glustechn. Ber. 51, 1-7.

BRUCKNER R., CHUN H.-U., GORETZKI H. and SAMMET M. (1980) XPS measurements and structural aspects of silicate and phosphate glasses. J. Non-Cryst. Solids 42,49-60.

COUNTS M. E., JEN J. S. C. and WIGHTMA~ J. P. (1973) An electron spectroscopy for chemical analysis study of lead adsorbed on montmorillonite. J. Phys. Chem. 77, 1924-1926.

DILLARD J. G., KOPPELMAN M. H., CROWTHER D. L., SCHENCK C. V., MURRAY J. W. and BALISTRIERI L. (1981) X-ray photo- electron spectroscopy (XPS) studies on the chemical nature of metal ions adsorbed on clays and minerals. In Adsorption~om Aqueous Solutions (ed. P. H. TEWARI), pp. 227-240. Plenum Press.

FENN P. M. and BROWN G. E., JR. (1974) X-ray induced electron emission spectroscopy of crystalline and amorphous silicates (abstr.). Eos 55, 1202.

GOLDMAN D. S. (1986) Evaluation of the ratios of bridging to non- bridging oxygens in simple silicate glasses by electron spectroscopy for chemical analysis. Phys. Chem. Glasses 27, 128- 133.

HOCHELLA M. F., JR. (198 1) Aspects of structure and rheology of ~uminosilicate melts. Ph.D. dissertation, Standord Univ.

HOCHELLA M. F., JR. and BROWN G. E., JR. (1984) Structure and viscosity of rhyolitic composition melts. Geochim. Cosmochim. Acta 48,263 I-2640.

HOCHELLA M. F., JR. and BROWN G. E., JR. (198.5) The structures of albite and jadeite composition glasses quenched from high pres- sure. Geockim. Cosmochim. Acta 49, 1 I37- I 142.

HOCHELLA M. F., JR. and CARIM A. H. (1988) A reassessment of electron escape depths in silicon and thermally grown silicon diox- ide thin films. Surface Sci. Lett. 197, L260-L268.

HOCHELLA M. F., JR., PONADER H. B., TURNER A. M. and HARRIS D. W. (1988) The complexity of mineral dissolution as viewed by high resolution scanning Auger microscopy: Labomdite under hy- drothermal conditions. Geochim. Cosmochim. Acta S&385-394.

HOLDREN G. R., JR. and BERNER R. A. (1979) M~h~ism of feldspar weathering-I. Ex~~mental results. Ge~him. Cosm~h~m. Acfa 43, 1161-1171.

HUNTRESS W. T., JR. and WILSON L. (1972) An ESCA study of lunar and terrestrial materials. Earth Planet. Sci. Lett. 15, 59-64.

JEN J. S. and KALINOWSKI M. R. (1979) An ESCA study of the bridging to non-bridging oxygen ratio in sodium silicate glass and the correlations to glass density and refractive index. J. Non-Cryst. Solids 38-39,2 l-26.

KANEKO Y. and SUGINOHARA Y. (1977) Fun~men~l studies on quantitative analysis of d, O-, and @- ions in silicate by X-ray photoelectron spectroscopy. Nippon Kinzoku Gakkai J. 41, 375- 380.

KANEKO Y. and SUGINOHARA Y. (1978) Observation of Si 2p binding energy by ESCA and determination of O”, O-, O*-, ions in silicate. Nippon Kinzoku Gakkai J. 42,285-289.

KANEKO Y., NA~MURA H., YAMANE M., MIZOGUCHI K. and SUGINOHARA Y. (1983) Photoelectron spectra of silicate glasses containing trivalent cations. Yonvo-Kyokaishi. 91. 321-324.

KOHIKI S., GHMURA T. and KUS~O K:( 1983a) A new charge-cor- rection method in X-ray photoelectron spectroscopy. J. Electron Spectros. Relat. Phenom. 28, 229-237.

KOHIKI S., OHMURA T. and Kusno K. (1983b) Appraisal of a new charge correction method in X-ray photoelectron spectroscopy. J. Elefiron Speetros. Relat. Phenom. 31, 85-90.

KOWALCZYK S. P.. LEY L.. MCFEELY F. R.. POLLAK R. A. and SHIRLEY D. A. (1974) Relative effect of extra-atomic relaxation on Auger and binding-energy shifts in transition metals and salts. Phys. Rev. B9,38 i-39 I.

LAM D. J., PAULIKAS A. P. and VEAL B. W. (1980) X-ray photo- electron spectroscopy studies of soda aluminosilicate glasses. J. Non-Cryst. Solids 42,4 i-48.

MURRAY J. W. and DILLARD J. G. (1979) The oxidation of~b~t(II) adsorbed on manganese dioxide. Geochim. Cosmochim. Acta 43, 781-787.

NAGEL S. R., TAUC J. and BAGLEY B. G. (1976) X-my photoelectron studv of soda-lime-silica alass. Solid State Camm. 20,245-249.

N~CKOLDS S. R. (1954) Average chemical compositions of some igneous rocks. Geol. Sot. Amer. Bull. 65, 1007-1032.

ONORATO P. 1. K., ALEXANDER M. N., STRUCK C. W., TASKER G. W. and UHLMANN D. R. (1985) Bridging and nonbridging

1648 M. F. Hochella, Jr. and G. E. Brown, Jr.

oxygen atoms in alkali aluminosilicate glasses. J. Amer. Cerum. Sot. 68, C148-C150.

PETROVIC R., BERNER R. A. and GOLDHABER M. B. (1976) Rate control in dissolution of alkali feidspar-I. Study of residual feldspar grains by X-ray photoelectron spectroscopy. Geochim. Cosmochim. Acta 40,537-548.

PIJGLISI O., MARLE~A G. and TORRISI A. (1983) Oxygen depletion in electron beam bombarded glass surfaces studied by XPS. J. Non-Cryst. Solids 55, 433-442.

PUGLISI O., TORRISI A. and MARLETTA G. (1984) XPS investigation of the effects induced by the silanization on real glass surfaces. J. Non-Cryst. Solidr 68, 2 19-230.

SCHOTT J. and BERNER R. A. (1983) X-ray photoelectron studies of the mechanism of iron silicate dissolution during weathering. Geo- chim. Cosmochim. Acta 47,2233-2240.

SCHOTT J., BERNER R. A. and SJOBERG E. L. (1981) Mechanism of pyroxene and amphibole weathering-I. Experimental studies of iron-free minerals. Geochim. Cosmochim. Acta 45, 2 123-2 135.

SCOFIELD J. H. (1976) Hartree-Slater subshell photoionization cross- sections at 1254 and 1487 eV. J. Electron Spectros. Relat. Phenom. 8, 129-137.

SHARMA S. K., VIRGO D. and MYSEN B. 0. (1978) Structure of melts along the join SiOz-NaAlSiO, by Raman spectroscopy. Carnegie Inst. Wash. Yearb. 17,652-658.

SHARMA S. K., VIRGO D. and MYSEN B. 0. (1979) Raman study of the coordination of aluminum in jadeite melts as a function of pressure. Amer. Mineral. 64, 779-787.

SMETS B. M. J. and KROL D. M. (1984) Group III ions in sodium silicate glass. Part 1. X-ray photoelectron spectroscopy study. Phys. Chem. Glasses 25, 113-l 18.

SMETS B. M. J. and L~MMEN T. P. A. (1981) The incorporation of aluminum oxide and boron oxide in sodium silicate glasses, studied by X-ray photoelectron spectroscopy. Phys. Chem. Glasses 22, 158-162.

STEPHENSON D. A. and BINKOWSKJ N. J. (1976) X-ray photoelectron spectroscopy of silica in theory and experiment. J. Non-Crysf. Solids 22,399-42 I.

SWIFT P. (1982) Adventitious carbon-the panacea for energy ref- erencing? Surface Interface Anal. 4, 47-5 1.

TASKER G. W.. UHLMANN. D. R.. ONORATO P. I. K. and ALEX- ANDER M. N. (1985) Str&ure oisodium aluminosilicate glasses: X-ray photoelectron spectroscopy. J. Physique C8, 273-280.

TAYLOR M. and BROWN G. E., JR. (1979a) Structure of mineral glasses: I: The feldspar glasses NaAISiJOS, KAISiTOs, and Ca- A12Si20s. Geochim. Cosmochim. Acta 43,6 i-75.

TAYLOR M. and BROWN G. E., JR. (1979b) Structure of mineral glasses: The Si02-NaAISi04 join. Geochim. Cosmochim. Acta 43, 1467-1473.

THOMAS T. D. ( 1980) Extra-atomic relaxation energies and the Auger parameter. J. Electron Spectrosc. Relat. Phenom. 20, 117-127.

VEAL B. W., LAM D. J., PAULIKAS A. P. and CHINC W. Y. (1982) XPS study of CaO in sodium silicate glass. J. Non-Crwf. Solids 49,309-320.

WAGNER C. D. (1972) Auger lines in X-ray photoelectron spectros- copy. Anal. Chem. 44,967-973.

WAGNER C. D. (1975) Auger parameter in electron spectroscopy for the identification of chet&al species. Anal. Chem. 47, I20 I-1203.

WAGNER C. D.. SIX H. A. and JANSEN W. T. (1981) Imorovina the accuracy of hetermination of line energies‘by E&Z& Chemical state plots for silicon-aluminum compounds. AppL Surl: Sci. 9. 203-213.

WAGNER C. D., PASSOJA D. E., HILLERY H. F., KINISKY T. G., SIX H. A. JANSEN W. T. and TAYLOR J. A. (1982) Auger and pho- toelectron line energy relationships in aluminum-oxygen silicon- oxygen compounds. J. Vat. Sci. Tech. 21,933-944.

YEH J. J. and LINDAU I. (1985) Atomic subshell photoionization cross sections and asymmetry parameters: I < % <: 103. Atomir. Date and Nuclear D&a Tables 32, 1 - 155.

YIN L. I.. GHOSE S. and ADLER I. (197 1) Core binding enerav dif- feren& between bridging and nonbhdgiig oxygen ato& in a silicate chain. Science 173, 633-635.

YODER H. S. and TILLEY C. E. (1962) Origin of basalt magma: An experimental study of natural and synthetic rock systems. J. Pefrol. 3, 342-532.