American Mineralogist, Volume 73, pages 161-167, 1988
X-ray photoelectronspectroscopyfor the direct identification of Ti valencein
[Ba,Cs"l[(Ti,ADl,++Jif1r,-rlOr. hollandites
Svennn Mvnn-l
School of Science,Grifrth University, Nathan, Queensland,Australia 411I
TrnnorHv J. Wrurn
Materials Division, Australian Nuclear Scienceand Technology Organization, Lock Mail Bag No. I,
Menai, N.S.W., Australia 2234
Sur E. KnssoN
ResearchSchool of Earth Sciences,Australian National University, GPO Box 4, Canberra,A.C.T., Australia 2600
JonN C. Rrvrnnn
SurfaceAnalysis Section, Materials Development Division, Harwell Laboratory, Oxon OXl l 0ItA, United Kingdom
AssrRAcr
A priori knowledge of the valence adopted by transition-metal elements in minerals
frequently provides useful insight into their petrogenesis.Generally, one or more of the
several well-establishedspectroscopicmethods (e.g., Mdssbauer, UV-visible, X-ray absorption) are used to determine the chemical state of cationic species.A more recent
technique, which is now gaining increasingacceptancefor such studies, is X-ray photoelectron spectroscopy(xrs). In this paper, we report the way in which xPSmay be used to
extract, in a semiquantitative manner, the ratio of trivalent to tetravalent Ti speciesin
synthetic [Ba,Cs,][(A1,Ti3*)2,+yTifi2"-,]O,uhollandites. The Ti3+/Ti4+ratios found by xrs
are in qualitative agreementwith those derived indirectly from classicalelectron-microprobe analysis.The quantitative discrepanciescan be ascribedto the presenceof rutile in
minor amounts. Generalized guidelines for the use of xps to obtain compositional and
chemical information about mineral specimensare outlined.
INrnonucrroN
Although the atomic structuresof many minerals have
been accuratelydetermined by X-ray or neutron diffraction, the details oftheir electronic structuresare usually
less well established.This is particularly so for minerals
containing transition-metal cations that may exist in two
or more oxidation states.Often the valenceof the elements
that constitutea mineral areinferred from its petrogenesis,
but this approach may prove ambiguous if extensive alteration has occurred. Therefore, it is inherently more
useful to determine directly the elementalvalence,as this
information will reflect the ambient oxygenfugacity during mineral formation.
In mineralogicalstudies,oxidation statesare most commonly derived from electron-microprobeanalyses(Eurn)
in which oxides are summed to the desired oxygen stoichiometry. However, this approach assumesthat the vaIence of all elements is known, and errors may occur if
the chemistry is complex or the oxygen stoichiometry
large(> l0). To overcometheselimitations, attemptshave
been made to develop more sophisticated evrpe techniquesin which the relative intensitiesofthe characteristic
l 024 l 6 l $02.00
0003-004x/88/0
X-ray emissions for a particular element are compared.
For example,O'Nions and Smith (1971)tried to use the
Lrr:Lr, ratio of Fe to establishthe content of ferrous and
ferric speciesin wiistite, hematite, and pseudobrookite.
However, they concluded that only qualitative information could be gained in this way.
Severalother spectroscopictechniquesare available to
determine chemical speciation. M0ssbauer spectroscopy
has been applied to over 40 elements (Herber, 1982),
though for mineralogical studies it is generally the 57Fe
nuclide that is exploited(e.g.,Aldridge et al., 1986;I)yar
and Burns, 1986). Optical absorption spectroscopy,
X-ray absorption spectroscopy,and electron-paramagnetic-resonancespectroscopyare also available, but these
methods are sometimeslimited by the large sample volume required (e.g.,Hofmeister and Rossman, 1984;Bonnin et al., 1985).Again, thesemethods have usually been
applied to the geologicallyinteresting fenous-ferric species.
spectroscopyhas been
Most recently,electron-energy-loss
used to identify the valence of elements and simultaneously (in combination with electron microscopy) provide structural information from very small crystallites.
l6l
r62
MYHRA ET AL.: Ti VALENCE IN HOLLANDITES
Chemicalinformation can be extractedfrom the spectrum
by detailed considerationsof the position (with a typical
resolution of +0. I eV) and shapeofpeak envelopes.This
valence
information
can be interpreted readily by comparison with
eledrons
spectraof standardcompounds,recordedin the sameexperiment, or by consulting the voluminous and mature
literature on xps investigations of various compounds
(Briggsand Seah,1983).
core
core
electrons
xes is generallythought ofas a broad-beam technique
electrons
becausethe normal lateral resolution is - I mm. However,
Fig. l. (a)Schematic
illustrationofthe process
ofphotoelec- some recent instruments are capableof "mapping" with
tron ejectionby X-ray photons.The photoelectrons
can come a lateralresolutionof - 100 pm. (Riviere, I 986).
fromeithercoreor valenceshells.(AdaptedfromBriggsandSeah,
1983.)(b) The ejectionofAugerelectrons
followsthe nonradia- CoprplnrsoN wrrH orHER suRFAcE ANALyTTCAL
TECHNIQUES
tive transitionofan electronto an inner-shellvacancy.(Adapted
from BriggsandSeah,1983.)
It is common to have xps availablein combination with
at least one other surface analytical technique-usually
It is, however, unable to separateor quantify mixed va- Auger electron spectroscopywith scanning Auger milencesof the sameelement(Otten et al., 1985).
croscopy (aEs-snu), and with static secondary-ionmass
As summarized in Table l, these
xps studies ofmineral surfaceshave not been reported spectrometry (ssrrvrs).
extensively (Nefedov et al., 1972; Adams et al., 1977; techniquesare surfacespecificand yield data that are esBancroft et al., 1979; Thames et al., l98l; Evans and sentially complementary.
Raftery, 1982) although application of the technique to
The application of AES-sAMto geologic problems has
such problems is increasing.In particular, xps has been been reviewedrecently(Hochella et al., 1986a, 1986b).
shown to be especiallyvaluable for determining the mech- Auger electronsare generatedby irradiation ofthe surface
anisms that are relevant to surfaceor interface reactions with monoenergetic electrons (typically 3-10 keV). As
suchasthosethat governthe dissolutionof minerals(Schott with xps, the primary interaction is the ejection of a core
and Berner, 1983;Myhra et al., 1984).
electron, but in this case the characteristic event is the
The present article describeshow xps may be used in radiationlessdecayofan upperlevel electroninto the core
mineralogical studies to ascertainsemiquantitatively the vacancy and the simultaneousejection of an Auger elecproportions of mixed-valencetransition-metalspecies.We tron with a characteristickinetic energy(Fig. lb). The low
have chosen to illustrate the utility of this technique by (30-1000 eV) kinetic energiesanalyzedin ees ensurethat
determining the relative proportions of trivalent and tet- the technique is surface specific for the same reasonsas
ravalent cations in a number of well-characterizedsyn- mentioned above for xrs. Again, a spectrum of characthetic hollandite sampleswith the generalformula [Ba"- teristic peaks indicative of atomic speciesin the nearCs,][(Al,Ti3*)2,*"Tif1r,_,JO,u(Kessonand White, I 986). surfacelayers can be generatedby electrostaticanalysis,
and peak heights can be related to atomic abundances.
Pnrxcrplps oF X-RAv pHoroELEcrRoN
However, chemical (i.e., bonding) information cannot be
SPECTROSCOPY
as readily extracted,as in the caseofxps, since electronBriefly, xps is basedon the principle that, when a surface excited Auger peaks are generally wider and weaker. In
is irradiated with X-rays, photoelectronswill be ejected particular, diferential Auger peak positions, which is the
(Fig. 1a). If X-ray lines of sufficiently narrow widths are most common operational mode, cannot be measured
used, the photoelectronshave characteristic energiesre- with the precision ofxrs peaks.The significantadvantage
lated directly to the atomic levels from which they came; ofens-senr over xps lies in its excellentlateral resolution
xps commonly uses either the AJKa (1486.6 eV) or the and the ability to scan a finely focused electron beam
MgKa(1253.6 eV) lines. With suchlow-energyexcitation, (-20-nm diameter for the most advanced instrument)
the photoelectronsmust originate in the outer few mono- over the sampleand produce elementalmaps showingthe
layers only ifthey are to escapewithout energyloss. The lateral distribution of surfacespecies.Such maps can be
resultant energeticelectronsare then collectedand count- relatedto surfacefeaturesobservedby secondary-electron
ed, after dispersion, by an electrostaticanalyzer.A pho- microscopy (seeCooper et al., 1986) for an example of
toelectronenergyspectrumthen consistsof a plot of counts Cs mapping on hollandite fracture faces).
Unlike xps and AEFSAM,static secondary-ion mass
as a function ofkinetic energy.Such spectramay be used
is a destructive technique,dynamic
for analytical purposesor to gain insight into the chemical spectrometry(ssrr'as)
bonding of the elements present. The technique is firlly srrrasis even more so, and at times this may limit its
quantitative insofar as the areaunder a characteristicpeak applicability. (There are also caseswhere the high current
can be related directly to the concentration ofthe corre- densitiesof ars-sau will be destructive.)During analysis
sponding atomic speciesin the surfacelayer. The sensi- by ssrus,an incident ion beam (usually 0.5-5 kev) is used
tivity of xes is of the order of 0. I aI.o/ofor most elements. to induce sputteringof atomic and molecular speciesfrom
r63
MYHRA ET AL.: Ti VALENCE IN HOLLANDITES
Treue1. Comparisonof surfaceanalyticaltechniques
2 (static),40
(dynamic)
Samplingdepth (monolayers)
1-10
1-5
Lateral resolution(nm)
Sensitivity(at %)
Nondestructive
Chargingproblems
Elemental-mapping
capability
Elementalrange
Accuracyof analysis
106
0.1
yes
yes
no
z>3
fullyquantitative
200-1000
01
yes-no
yes
yes
z>3
quantitafully(AES)
andsemi-(sAM)
trve
mapping
of surface
elemental
analysis
speciesandsmall-area
determrnation
of chemicalspeciationandcomposition
Principalapplication
a solid surface. The incident ions are commonly Ar or
oxygenbut liquid metal, particularly Ga, sourcesare also
available. The ejectedspeciesusually come from the first
one or two atom layers(0.5 nm), although interlayer mixing of speciesmay sometimes give rise to contributions
from greaterdepths (Smart, 1985).The sputteredions are
focusedonto a mass analyzer (quadrupole, electrostatic,
or magneticsector).The techniqueis extremely sensitive,
being able to detect many speciesat the ppm level, and
it is as a method of identifying trace elementsthat ssrMs
is far superior to xps and aes-seu (Table l). Although
ssrus is able to detect all elementalisotopes,such spectra
are extremely difficult to quantify owing to the largerange
of ionic sensitivities and the variable sputtering rates of
the same speciesfrom different matrices. However, reliable comparisonscan be drawn betweenmaterials of similar composition. Becausesrus gradually erodesthe surface under examination (especiallyin the so-called
"dynamic" mode), it is often usedin conjunction with the
other techniquesto yield depth profiled xps or AES-sAM
data.
Thus, for the purposesof the presentexperiment,which
is the determination of Ti valencestatesin hollandite, xps
is the most appropriate tool. If on the other hand, the aim
had beento study the spatial distribution ofCs or perhaps
the occurrenceof unexpectedtrace impurity atoms, then
AEFSAMor ssrMswould have been more suitable.
ExpnnrvlnurAl
Specimenpreparation
10n
0.001
no
yes
yes
all elements
semi-quantitative
trace-elementanalysis
hydrated titania. This powder was mixed with aqueous
solutions of aluminum, barium, and cesium nitrates in
quantities appropriate to the particular hollandite being
prepared.After slurrying at pH 9 with NH.OH, the filtered
powder was calcinedat 750'C for I h. R.eactivehot pressing was then carried out using the conditions listed in
Table 2. The specimenswere characterizedby X-ray diffraction and chemically analyzed using an electron microprobe;the derivation of structural formulae was based
on theselatter data.In addition to hollandite, lesserquantities of slightly reducedrutile (Ti,Or" ,) werealso present
and estimatedto account for I 5-30% of the bulk by volume. As well as the hollandite-rutile assemblages,a section of single-crystalrutile (TiOr) was used as a reference
surfacein which the majority of the Ti is tetravalent.
The hollandites were fractured immediately before
transfer to the spectrometerUHV chamber in order to
minimize oxidation and contamination and to present a
surface most likely to be representative of bulk abundancesand chemical environments for all atomic species'
The rutile specimenwas cut perpendicular to its crystallographic c axis (+2";, and the (001) surfacewas polished
(0.2-pm alumina powder) and etched in a bath of molten
NaOH at 800'C. The specimenwas then equilibrated in
high-purity oxygenat 700"C,quenchedto laboratory ambient, and transferred to the spectrometerairJock in an
oxygen atmosphere.
METHoDS
Analgical procedure
xps data were collected using a vc EscALABMark II at
A detailed description of the methods used to prepare AERE Harwell. A survey scan was carried out for every
the hollandite sampleshas been given elsewhere(Kesson specimenin order to confirm the presenceofall expected
and White, 1986).In brief, titanium-2-propoxide was hy- species,to determine the extent of adventitious C condrolized and dried to produce a highly reactive form of tamination (i.e., weakly bound polymeric C), and to as-
Tner 2. [Ba,Cs"][(Al,Ti3')",*Jit!", y]Or6hollandites
Specimen
no.
Hp 462
Hp46s
Hp463
Hp 410
Hp 402
Structuralformula
Preoarationconditions
hot-pressedin graphile,1250'C, 30 MPa,
[Bao4!Cso?o][Tiit6TiSr]O16
hot-pressedin platinum,1250 "C, 0.5 GPa,
[cs,4,][(AlouTi?;)Ti33j]o16
hot-pressedin graphite,1250"C. 30 MPa,
[Cs13,][Ti?IJi3;s]O16
[Baoscso5,][(AloToTqi,)Ti3is]O,6 hot-pressedin graphite,1250 "C, 30 MPa,
[Bao,,Csoe,][(Alo66Ti8;6)Tit:6]O,6hot-pressedin graphite,1250'C, 30 MPa,
h
n
h
n
n
164
MYHRA ET AL.: Ti VALENCE IN HOLLANDITES
HP4lO
TiO" nmdord
Q Auger
2pr/2
F.
3
Ols
Bo3d
c!tuilug.r
c,
F
=
5
I es
q
F
c
F
Tt 2p
E
S
Ed
S
p
2pvz
ah
F
so
cb
750
Al 29
=
f
o
o
250
500
BINDING
ENERGY(eV)
(EV)
8'ND'NGEI\IERGY
Fig. 2. Surveyxrs scanof the hollandite(HP 410) surface. Fig. 3. Detailedxps scanover the Ti 2pr,r.r,regionfor a
The major peaksarelabeledin accordwith standardnotation. single-crystal
TiO, surface.Background
hasbeensubtracted
and
DoubletssuchasTi 2pr,r.r,r,
arenot resolved.
the envelopesmoothedwith a polynomialbestfit in order to
arriveat the standardtetravalentspectrum.
certain the severityof charging.AlK" X-rays with a source
power of 300 W were usedas the incident radiation. Base widths. In practice, however, this approach may be tevacuum was l0-e torr or better; previous studies have dious and ambiguousas it is usually necessaryto include
shown that extreme UHV conditions are not required for "skewness"as well as adding Lorentzian contributions to
relatively inert titanate surfaces(Myhra et al., 1983).The Gaussian line shapes (Sherwood, 1983). Alternatively,
parametersfor thesesurvey scanswere as follows: 50-eV when suitable standardsare available, it is more straightpassenergy, l-eV increments, and l0-ms per increment. forward to compare two experimental envelopes,one of
A typical survey scan (HP 402) is shown in Figure 2; the which is characteristic of a known compound with the
major characteristicpeaksareconsistentwith the expected relevant speciesin a well-definedvalencestate.This latter
atomic species.
method was usedin the presentstudy, in which the stanDetailed scanswere made acrossall major peaks-oxydard TiO, surfacewas analyzedin order to generatea Ti
gen lr, Ti 2p, Ba 3d, Cs 3d, and Al 2p and 2s. (Ad- 2p envelopefor tetravalent Ti-O bonding.
ventitious C contamination was also monitored, although
the fractured surfacescontained only trace amounts of Depth'profiled xes
this element.) The detailed scanswere obtained with a
In order to justify the assumption that hollandite fracpassenergyof 20 eV, 0.2-eV increments,and count times ture faceswererepresentativeof the bulk, someadditional
of 20-100 ms per increment depending on the relative information was required. Elemental abundancesin the
sensitivity and the abundanceof atomic species.The sig- first few monolayersof the fracture faceswere determined
nal-to-noise ratios of the detailed spectrawere improved by the following standardprocedure.Areasunder selected
by carrying out 20-100 repetitions over the region of in- peaks were measured for all species.These areas were
terest.
normalized, then divided by empirical sensitivity factors
As the aim of theseexperimentswas to determine the so as to yield self-consistentestimates for atomic abunvalenceof Ti, particular attention was paid to the Ti 2p dancesthat could be compared to the bulk composition
envelopes.In the caseofthe pure tetravalent state, these of the starting materials. Ion-beam etching (5-keV Ar*
consistof a doublet 2pr,r.,,,with a splitting of 5.8 eV; the and current density of 0. l-l pA.min/cm'?) was then carbinding energycorrespondingto the tetravalent 2pr,rslate ried out to a depth of several nanometers(i.e., removal
is 458.7 eV in TiO, (Sayersand Armstrong, 1978).Fur- of the first few monolayers), and the xps measurements
thermore, it is known that the binding energies of Ti were repeated.In this manner, it was possibleto ascertain
2pr,relectrons in trivalent and divalent statesare shifted whether the surfacepresentedby fracture was comparable
to lower binding energiesby 2.0 and 3.6 eV, respectively to the bulk.
(Sayersand Armstrong, 1978).
DnrnnrrrN,l,TroN oF Tr var-nNcn RATros
There are two ways in which to proceedwith the fitting
ofpeak envelopesthat have contributions from more than
one valencestate.First, one can in principle generatetotal
envelopescharacteristicofany mixture ofTi valencestates
by suitable scalingand superpositionofshifted envelopes
and by using known binding energies,shifts, and peak
A detailed scanover theTi 2p envelopeofthe standard
rutile specimen is shown in Figure 3; background and
satelliteshave been subtractedand the data smoothed by
simple polynomial fitting. Detailed scansover the same
envelopewere obtained for all hollandite specimensand
165
MYHRA ET AL.: Ti VALENCE IN HOLLANDITES
(r,
F
2
f
F
E
(t
F
3
463
a
459
455
E|ND|wG ENERGY(cV,
459
463
SlNAt'rG ENERGYbV)
Fig. 4. Detailed scan of the Ti 2p envelope for specimensHP 462 (a) and HP 402 (b). Subtraction of the TiO, "standard"
tetravalent envelopeand the resultant contribution from trivalent speciesin the surfaceare shown, in stippled pattern.
subjectedto similar data manipulation. Finally, allTi 2p
envelopeswere scaledto the samearbitrary height for the
2pr,, center maximum. The tetravalent 2p envelope of
TiO, was then subtractedfrom the processedTi envelopes
of the hollandite samplesin order to identify any trivalent
Ti component. By way of example, the results of these
manipulations are shown in Figure 4 for the hollandites
HP 462 and HP 402 (Table 2). On the basisof microprobe
data, these two hollandites were expectedto exhibit the
widest extremes in Ti3*/Ti4* ratios; the presenceof Ti3*
inHP 462 is obvious (Fig. 4a), while for HP 402 the Ti3*
contribution is barely visible above background. Similar
spectrahaving Ti3* to Ti4. admixtures intermediate between thosepresentedwere generatedfor other hollandite
specimens,but for the sake of brevity are not shown. In
Table 3 are summarizedthe Ti3*/Tia* ratios asdetermined
by rvre and xps for the hollandites studied. The two sets
of data are in qualitative agreement; smaller Ti3+/Ti4+
ratios are detected by both techniquesfor HP 402 1T1t'
deficient) than for HP 462 (TF. rich). However, the differencesin absoluteterms betweenthe resultsare in places
subslantial. Significantly, the Tir+/Ti4+ ratios as determined by xps are consistentlysmaller than the eupe values.It is likely that this is attributable to the fact that xps
averagesover a relatively large sample surface,whereas
the microprobe data representa mean of many analyses
collected from individual hollandite crystallites. Therefore, xrs analyseswill be afected by an additional Ti4*
content,presentin the minor rutile phase,leadingto lower
than expectedTi3*/Tia* ratios.
The interpretation ofthe resultsaboveis clearly affected
by the extentto which the fracture surfaceis representative
ofthe bulk composition and chemicalstates.For instance,
it is known that fracture of hollandite-bearingpolyphase
titanate ceramics occurs predominantly along glassy intergranular films (usually - I nm wide), and that Cs may
be concentratedin thesefilms (Cooper et al., 1986).In
order to determine the significanceof this, hollandite fracture surfaceswere ion-beam etchedto an integrated dose
of about 40 pA'min/cm'?, which is equivalent to thg removal of some l-5 nm of surfacematerial. [The precise
etch rate is uncertain sincethe fracture surfaceis irregular,
and the relationship betweenion-sputtering rate and the
speedof monolayer removal is not well known for titanates (Myhra et al., 1985).1The areasunder the characteristic xrs peakswere monitored during the ion etching
so that compositional profiles could be obtained as a function of depth. Typical data are given for HP 462 and HP
402 in Figure 5. It can be seenthat Cs is enhancedin the
near-surfacelayers (i.e., along grain-boundary films) at
the expenseofTi but that the concentrationsofall species
approachthe bulk composition of the starting material at
a depth of approximately 0.5 nm. Theseresults are summarized in Table 4. Ti valenceratios were not calculated
for the etchedsurfaces,sinceion-beam bombardment introduces electronic damageand both divalent and trivalent speciesare known to becomemore prevalent (Myhra
et al., 1983). Therefore, it is generally impracticable to
determine by ion-beam etching the chemical changesin
speciationwith depth. However, it is possibleto useangleresolved xps measurements for nondestructive depth
analysisof chemical environments in the first few monolayers.
From consideration of the data presentedin Figure 5
and Table 4, it is apparentthat Cs segregationis confined
to a few monolayers and that in these regions, Ti abundanceis no lower than 800/oof the bulk value. Small changes
ratios for hollandite[Ba,Cs"]TaeLe3. Ti3+/Ti4+
[(Al,Ti3.)r,+Jigi,,_"]O,u
Tir*/Ti.. (%)
Specimenno.
HP 462
HP465
HP 463
H P4 1 0
HP 402
26
18
20
18
10
18
15
15
10
3
r66
MYHRA ET AL.: Ti VALENCE IN HOLLANDITES
--"------
of Ti within
TnaLe
4. Relative
enhancement
of Cs anddepletion
films(lF)compared
to the bulkcompointergranular
sition
Intergranular
film
thickness'
s
_/
u
z
z
o
I
too
L
Specimenno
Cs,./
Cs*,.
Tt,rl
Tiou,*
HP 462
HP465
HP463
H P4 1 0
HP 402
3
3
4
2.5
3
0.8
0.8
0.8
0.9
0.9
lon dose
(pA min/cm'?) Width (nm)
0.5
5 + 0.5
0.5
0.5
3+ 0.5
JA
t.3 a
0.3
0.5
0.5
0.2
03
* Intergranularfilm thicknessesare estimatedfrom the ion dose required
to reachbulk composition.An averageetchingrate of 0.1 nm per pA min/
cri, has been assumed.The systematicuncertaintywithin this conversion
factor may be as large as a factor of two, but it will not atfect the relative
accuracy
DD.o
,oN oosE
or cleanedsurfaceshould be prepared immediately prior
Fig.5. Elemental
vs.ion etchingdose(D) for HP
abundances
to insertion into the UHV chamber.If the surfaceis highly
462 and HP 402. The nominal bulk compositionis shown.
> D : 0 dataarefor a freshfractureface.O : oxygen,O : Ti; reactive, cleavagemust be carried out in situ under extreme UHV conditions. The experimentalist should also
tr:Ba,I:Cs.
be aware that some cleaved or fractured surfaces may
in the characteristicsof interfacial regions (Cs enhance- undergosubstantialreconstructionor be highly defective.
3. Curve synthesisof"standard" spectrais feasibleonly
ment and layer thickness) are primarily due to different
eutectic reactions between rutile and hollandite for the in rare instances(Sherwood,1983).In general,it is more
various compositions.Those hollandites with the highest straightforward to generate standard spectra from the
Cs concentrationhave the lowest eutectic, so that during analysis of compounds that are related chemically and
hot pressing,partial melting occurs,and the formation of structurally (i.e., possessthe same anionic coordination
wider intergranular films are favored. However, beyond number) to the "unknown" specimen. Standards must
this modified intergranular region, the fracture faces are have the relevant atomic speciesin known valencestates.
4. The xps technique is not applicable to trace-element
representativeof the bulk composition.
concentrations.Routine identification of valencestatesis
CoNcr-usroNs
feasible only when the atomic speciesis present in con>l al".o/o.
Although most speciesare detectable
centrations
It is useful to summarize the conditions for which the
ratio may not be suflevel,
signal-to-noise
below
this
the
proceduresused in the present limited context may be
ficiently high to deterrnine peak position and the shape
exploited generally.
l. Routine xes techniquesare capable,in principle, of ofthe envelopewith the required accuracy.
5. Most minerals are good electrical insulators. Thereestablishingthe valenceof multivalent elementswhen these
fore
the application ofsurface analytical techniquessuch
speciesexhibit chemical shifts of I eV or more. Smaller
xps,
AES-SAM,
and dynamic sIr,tsgives rise to charging.
as
greater
require
more
shifts
detailedanalyses,and
attention
of the signal-to-noise ratio. For the majority of ionic This is a seriousproblem when accurateinformation on
species,the characteristicxps binding eneryiesare known binding energiesis required.For xps there are severalways
and are documented in the review and/or applications of taking account of charging shifts (Swift et al., 1983).
literature [for collated data, seeBriggs and Seah(1983)]. For instance, X-ray-induced Auger peaks can be meaAlthough we have emphasizedthe application of xps to suredin order to determinethe so-calledAuger parameter,
transition-metal species,it has been successfullyapplied which is independent of charging and most other experto determining the valence of rare-earth elementsand U imental artifacts.Or, one may be able to deducethe charging shift and correct for it by scanningover a peak with
in titanateminerals(Szajmanet al., 1987).
2. Formedy, the major constraint of xrs analysis was known binding energy,i.e., by using a "marker." Another
the necessityfor a large (> I mm,) representativesurface popular method is to use a flood of low-energyelectrons
to be accessible.However, recent instrument advances from an auxiliary source,either to dischargethe surface,
make it possible to do selecled-areaxps under certain or at least to stabilize the degreeofcharging.
favorablecircumstances(Riviere, I 986). Furthennore, the
surfacemust be clean, and the chemical environment of
the relevant atomic speciesin the first two or three monolayers must be substantially similar to that in the bulk.
For a nonreactive material (i.e., one resistant to change
by exposureto laboratory ambient conditions) a fractured
AcxNowr-nocMENTS
The present study was undertaken with support from the Australian
National Energy Research,Development and Demonstration Program.
One of the authors (S.M.) is especiallyindebted to the SurfaceAnalysis
sectionat the Harwell Laboratory for the hospitality and support extended
during a period ofattachment
MYHRA ET AL.:Ti VALENCE IN HOLLANDITES
RnrnnnNcns cItpo
Adams,J.M.,Evans,S , Reid,P.I.,Thomas,J.M , andWalters,M. J.(1977)
Quantitative analysis of aluminosilicates and other solids by X-ray
photoelectronspectroscopy.Analytical Chemistry, 49, 2001-2008.
Aldridge, L.P., Bill, E., Bliis, R., Lauer, S., Marathe, V R., Sawaryn,A.,
Trautwein, A,X., and Winkler, H. (1986) Electronic structure of Fe in
some minerals, derived from iterative extendedHiickel theory (IEHr),
multiple scatteringXd (MS-Xc) calculations,and Mdssbauermeasurements AmericanMineralogist,71, l0l5-1021.
Bancroft,GM., Brown, J.R., and Fyfe, W.S. (1979) Advancesin, and
applicationsoi, X-ray photoelectronspectroscopy(esce)in mineralogy
ChemicalGeology,25, 227-243.
and geochemistry.
Bonnin,D., Calas,G., Suquet,H., and Pezerat,H. (1985)Sitesoccupancy
ofFest in Garfield nantronite:A spectroscopicstudy.Physicsand Chemistry of Minerals, 12, 55-64
Briggs,D., and Seah,M.P., Eds ( I 983) Practicalsurfaceanalysisby Auger
and X-ray photoelectronspectroscopy.Wiley, New York.
Cooper,J A., Cousens,D.R , Hanna,J.A., Lewis,R.A., Myhra, S.,Segall,
R L., Smart,R.St.C.,Tumer, P.S.,and White, T.J (1986)Intergranular
films and pore surfacesin Synroc C: Structure, composition and dissolution characteristics.American Ceramic Society Joumal, 69,3475)Z
Dyar, M.D., and Burns,R.G. (1986)Mdssbauerspectralstudy of femrginous one-layertrioctahedralmicas. American Mineralogist, 7 l, 955965
Evans,S., and Raftery, E. (1982) Determination ofthe oxidation state of
manganese
in lepidolite by X-ray photoelectronspectroscopyClay Mine r a l s I. / . 4 t t - + 6 1 Herber, R.H, (1982) Mdssbauer etrect. In Encyclopediaof scienceand
technology,vol. 8, p. 754-751. McGraw-Hill, New York.
Hochella,M F, Harris,D.W., and Turner,A.M. (1986a)ScanningAuger
microscopy as a high-resoluton microprobe for geologic materials
1, 124'7-1257.
AmericanMineralogist,'1
Hochella,M.F., Turner,A.M., and Harris,D W (1986b)High resolution
scanningAuger microscopyof mineral surfaces.ScanningElectron MicroscopyI1,337-349, SEM Inc., AMF O'Hare, Chicago.
Hofmeister, A.M., and Rossman,G R. (1984) Determination of Fe3"and
Fe2*concentrationsin feldsparby optical absorption and epn spectros'
copy.Physicsand ChemistryofMinerals, ll,213-224
synKesson,S.E, and White, T.J (1986)[Ba,Cs"][(Ti,Al)#*"Ti!1,,-"]O'o
roc-type hollandites:I Phasechemistry. Proceedingsofthe Royal Society (London),A-405, 73-101.
Myhra, S., Bishop,H.E., and Riviere,J.C. (1983)Investigationby xrs of
surfacefeaturesofsome titanate minerals.SurfaceTechnology,I 9, I 6 lt'72.
t67
Myhra, S., Savage,D., Atkinson, A., and Riviere, J.C. (1984) Surface
modification of some titanate minerals subjected to hydrothermal
chemical attack. American Mineralogist, 69, 902-909.
Myhra, S.,Bishop,H.E.,and Riviere,J.C.(1985)Featuresoffracturefaces
in SynrocC. SurfaceTechnology,25,259-272.
Nefedov,V I, Urosov, V.S., and Kakhana,M M' (1972)X-ray photoelectron spectroscopyof bonds in Na, Mg, Al, and Si minerals. GeochemistryIntemational,9, 7-13
O'Nions, R.K., and Smith, D.G.W. (1971) Investigationsof the lFr
X-ray emission spectraof Fe by electron microprobe. Part 2' The Fe
-Ln-nr
spechaofFe and Fe-Ti oxides.American Mineralogist, 56, 14521463.
Otten, M.T., Miner, B., Rask,J H., and Buseck,P.R (1985)The determination of Ti, Mn and Fe oxidation statesin minerals by electron
energyloss spectroscopy.Ultramicroscopy, 18, 285-289.
Riviere, J.C (1986) Surfaceanalysis-Current capabilities Zeitschrift fiir
AnalytischeChemie,324, 728-744.
Sayers,C.N., and Armstrong,N.R. (1978)X-ray photoelectronspectroscopy ofTiO, and other titanateelectrodesand various standardtitanium
oxide materials: Surfacecompositional changesof the TiO, electrode
SurfaceScience,77, 301-309.
during photoelectrolysis.
Schott,J, and Berner,R.A. (1983) X-ray photoelectronstudiesof the
mechanismof iron silicate dissolution during weathering,Geochimica
et cosmochimica Acta, 47, 2233-2240.
Sherwood,P.M.A. (1983) Data analysisin X-ray photoelectronspectroscopy. In D. Brrggsand M P Seah,Eds., Practical surfaceanalysisby
App 3, p.445475.Wi.'
Augerand X-ray photoelectronspectroscopy,
ley, New York
Smart, R St.C. (1985) The validity of snrrsobservationsof alkali metal
segregationinto intergranularfilms in ceramics.Applications ofSurface
Science,22/23, 90-99
Swift, P, Shuttleworth,D., and Seah,M.P. (1983) Appendix 2. In D.
Briggs and M.P. Seah, Eds., Practical sulface analysis by Auger and
X'ray photoelectronspectroscopy.Wiley, New York
Szajman,J., Smart,R.St C , and Myhra, S. (1987)xps studiesof valenca
statesofcerium and uranium in Synroc C. Surfaceand CoatingsTechnology, 30, 333-342
Thames,M.T., Peterson,D.A., Hartley,J.N, and Freeman,H.D. (1981)
Solving mineral processingproblems with advanced surface science
techniques.In D.M Hausen and W.C. Park, Eds., Proceedingsof the
I lOth AIME annualmeeting;Processmineralogy,extractivemetallurgy,
mineral exploration, energyresources,p, 63-88. American Institute of
Mineralogy and Metallurgy, Chicago,Illinois.
Apnu-21, 1987
MaNuscnrsrRECEIvED
AccEprEDSeFrsl{sen22, 1987
MANUSCRTpT