SURFACE AND INTERFACE ANALYSIS
Surf. Interface Anal. 2004; 36: 1564–1574
Published online 7 October 2004 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/sia.1984
Investigation of multiplet splitting of Fe 2p XPS
spectra and bonding in iron compounds
A. P. Grosvenor,
1,2∗
B. A. Kobe,
1
M. C. Biesinger
1
and N. S. McIntyre
1,2
1
Surface Science Western, Room G-1, Western Science Centre, University of Western Ontario, London, Ontario N6A 5B7, Canada
2
Department of Chemistry, University of Western Ontario, London, Ontario N6A 5B7, Canada
Received 4 May 2004; Revised 14 July 2004; Accepted 26 July 2004
Ferrous (Fe
2+
) and ferric (Fe
3+
) compounds were investigated by XPS to determine the usefulness of
calculated multiplet peaks to fit high-resolution iron 2p
3/2
spectra from high-spin compounds. The
multiplets were found to fit most spectra well, particularly when contributions attributed to surface peaks
and shake-up satellites were i ncluded. This information was useful for fitting of the complex Fe 2p
3/2
spectra for Fe
3
O
4
where both Fe
2+
and Fe
3+
species are present. It was found that as the ionic bond character
of the iron — ligand bond increased, t he binding energy associated with either the ferrous or ferric 2p
3/2
photoelectron peak also increased. This was determined to be due to the decrease in shielding of the
iron cation by the more increasingly electronegative ligands. It was also observed that the difference in
energy between a high-spin iron 2p
3/2
peak and its corresponding shake-up satellite peak increased as the
electronegativity of the ligand increased. The extrinsic loss spectra for ion oxides are also reported; these
are as characteristic of each species as are the photoelectron peaks. Copyright 2004 John Wiley & Sons,
Ltd.
KEYWORDS: XPS; iron; multiplet splitting; electronegativity
INTRODUCTION
X-ray photoelectron spectroscopy has become a highly
surface-sensitive technique that has found use in many
different areas of chemistry. The effects of chemical changes
on the photoelectron peak shape have been of particular
interest. For studies involving iron, the 2p
3/2
peak for high-
spin Fe
3C
and Fe
2C
compounds is broadened compared with
Fe
0
metal or low-spin Fe
2C
.
1
The broadness of these peaks
has been shown by Gupta and Sen to be due to the inclusion
of electrostatic interactions, spin-orbit coupling between the
2p core hole and unpaired 3d electrons of the photoionized
Fe cation and crystal field interactions.
2,3
These interactions
were calculated using the Hartree–Fock free-ion method.
2
Studies performed in the past have shown that the
2p
3/2
envelope from high-spin Fe
3C
compounds can be well
fit using peaks constrained to conform to the multiplets
calculated by Gupta and Sen.
4,5
Owing to resolution
limitations, it has been found that only those peaks predicted
from electrostatic and spin-orbit coupling interactions are
best used because the inclusion of crystal field effects
increases the spectral complexity.
5
McIntyre and Zetaruk
4
as well as Kowalczyk et al.
6
were among the first to show the
practical use of the Gupta and Sen (GS) multiplets as well
as the limitations of the calculation. For example, a pattern
of peaks consistent with the GS multiplet predictions was
Ł
Correspondence to: A. P. Grosvenor, Surface Science Western,
Room G-1, Western Science Centre, University of Western Ontario,
London, Ontario N6A 5B7, Canada. E-mail: agrosven@uwo.ca
Contract/grant sponsor: NSERC.
observed for high-spin Fe
3C
in ˛-Fe
2
O
3
, but the multiplet
components were spaced more closely by 25% compared
with the predicted spacing.
4
The difference in energy spacing
between the calculated peaks compared with the observed
peaks was ascribed to the use of a high-spin free-ion model
used during the calculation that was unable to properly
represent a real system.
3
The 2p
3/2
peaks from low-spin Fe
2C
compounds display
no multiplet interactions because all six 3d electrons are
spin paired.
7
However, the Fe
2C
low-spin 2p
3/2
peak can
exhibit an asymmetric tail to higher binding energy.
7
This
tail is ascribed to surface structures that exhibit different
binding energies compared with that of the bulk structure
because of surface termination.
7
Such termination frequently
could result in a decrease in symmetry.
8
Studies involving
ligand field theory have shown that if the symmetry of a
low-spin Fe
2C
cation is decreased due the loss of a ligand,
the binding energy of its 2p
3/2
peak would be increased.
8
A decreased crystal field energy for Fe ions located at the
surface (top two atomic layers) compared with those located
within the bulk has also been suggested to be the cause of
this high-binding-energy tail.
9
In addition to multiplet structures, the presence of
satellite peak structures has been used to determine the
presence of high-spin ferrous species. Such satellites have
been ascribed to shake-up
10
or charge transfer processes.
11
Yinandcolleagueswereabletoshow that shake-up processes
were attributed to the movement of an electron from a
3d orbital to the empty 4s orbital during ejection of the
core 2p photoelectron.
10
These authors have suggested that
Copyright 2004 John Wiley & Sons, Ltd.
Multiplet splitting of Fe 2p XPS spectra 1565
charge transfer processes (either ligand to metal or metal to
ligand) cannot be the cause of high-spin Fe
2C
and Fe
3C
satellites because comparable shake-up satellites do not
appear on photoelectron lines associated with the ligand.
The calculations performed by Gupta and Sen also predict
high-binding-energy shake-up peaks to be present in the Fe
2p spectrum in addition to the multiplet peaks.
3
In this work, various Fe compounds (both high-spin and
low-spin) have been analysed by XPS to show the usefulness
of GS multiplets and surface peaks for fitting the Fe 2p
3/2
envelope. The specific compounds that will be discussed
include ˛-Fe
2
O
3
, -Fe
2
O
3
, ˛-FeOOH, -FeOOH, FeBr
3
,
FeCl
3
,FeF
3
,Fe
3
O
4
,FeBr
2
,FeCl
2
,FeF
2
,FeSO
4
,Fe
1x
Oand
K
4
Fe(CN)
6
. The application of XPS to a qualitative determina-
tion of the degree of ionic character of the Fe —ligand bond,
as well as the strength of crystal field splitting by the ligands,
will also be discussed. Extrinsic loss structures for the major
oxide structures studied are included in Appendix A.
EXPERIMENTAL
Measurements by XPS
All spectra were measured using a Kratos Axis Ultra XPS
instrument. A monochromatic A K˛ x-ray source was used
for all samples, along with pressures in the analysis chamber
of 10
6
–10
7
Pa. The resolution function of the instrument
has been found to be 0.35 eV using the silver Fermi edge
(unpublished results). To control charging of the samples, the
charge neutralizer filament was used during all experiments.
The conditions used for all of the survey scans were as
follows: energy range D 1100–0 eV, pass energy D 160 eV,
step size D 0.7 eV, sweep time D 180 s and x-ray spot size D
700 ð 400
µm. For the high-resolution spectra an energy
range of 40–20 eV was used, depending on the peak being
examined, with a pass energy of 10 eV and a step size of
0.05 eV.
Sample preparation
All powder samples examined were prepared and intro-
duced into the spectrometer via a glove box. The glove box
was filled with N
2g
or Ar
g
so as to limit the chance that the
samples would react with the air or airborne contaminates.
The powders were placed on Cu sample holders and crushed
to reveal clean surfaces, which were then heated in a vacuum
if it was felt to be necessary for removal of any adsorbed
H
2
O. Mineral samples were fractured in a vacuum so that
fresh, clean faces were present during analysis.
All powder samples (except for Fe
1x
O) were obtained
from AlfaAesar, Ward Hill, MA. The sample of ˛-Fe
2
O
3
(haematite; Boot Hill, NSW, Australia) was obtained from
the Dana Mineral Collection found at the University of
Western Ontario, Department of Earth Sciences. The Fe
3
O
4
(magnetite; Mesabi Range, MN, USA) sample was obtained
from a mineral collection owned by Mr Ross Davidson.
Owing to the instability of the non-stoichiometric oxide
Fe
1x
O,
12
it was synthesized immediately prior to examina-
tion using a synthetic route based on the method discussed
by Moukassi et al.
13
whereby ˛-Fe
2
O
3
was reduced by H
2g
while being heated in a Lindberg Mini-Mite tube furnace
at 600
°
C. During the synthesis, the powder changed colour
from dark red to dark grey/black and underwent sintering
to form a hard pellet. After 3 h had passed the sample was
quenched with liquid N
2
and transported in liquid N
2
to
the spectrometer for analysis. In the spectrometer the sam-
ple was reheated to 600
°
C for 1 h to make sure that only
Fe
1x
O was present. Analysis of the XPS spectra indicated
that the stoichiometry of the compound was actually Fe
1.1
O,
its presence was confirmed using powder x-ray diffraction.
Spectral analysis
Collected XPS spectra were analysed using CasaXPS
software.
14
All spectra were calibrated using the adventitious
C 1s peak with a fixed value of 284.8 eV. After calibration,
the background from each spectrum was subtracted using a
Shirley-type background to remove most of the extrinsic loss
structure.
11
All survey scans were analysed to determine the
stoichiometry of the compound by using the appropriate sen-
sitivity factors and to determine the amount of adventitious
carbon and contaminates present.
For analysis of the high-resolution Fe 2p spectra, two
Shirley backgrounds were used, one each for the 2p
1/2
and
2p
3/2
envelopes. The Fe 2p
3/2
envelope from compounds
containing a high-spin Fe cation was fit using peaks
Figure 1. Most intense GS multiplet peaks calculated for
high-spin Fe
3C
(a) and Fe
2C
(b) compounds. Component
information for each of the multiplets came from a digitized
reproduction of the graphs found in the original paper.
3
The
linewidth used to fit the Fe
3C
data was 1.1 eV, whereas a
linewidth of 1.3 eV was required to fit the Fe
2C
data. The
requirement of different linewidths depending on the oxidation
state is unexplained.
Copyright 2004 John Wiley & Sons, Ltd. Surf. Interface Anal. 2004; 36: 1564–1574
1566 A. P. Grosvenor et al.
corresponding to the GS multiplets, surface structures and
shake-up-related satellites. The main peak GS-predicted
multiplets for the Fe 2p
3/2
envelope are shown in Fig. 1. The
experimental spectra were fitted using peaks that had a full
width at half-maximum (FWHM) ranging from 1.0 to 1.6 eV
and intensities similar to those found in Fig. 1. In this work,
the intensity has been defined as the peak area rather than
the peak height. The FWHM range used was chosen based
on the use of a similar range by previous authors.
4,5
To fill the
rest of the envelope, a peak reporting to surface structures
was added with a higher binding energy than the multiplets.
This peak was allowed to have a larger FWHM than the
individual multiplets. A single large peak representing the
satellites found due to shake-up, which may include peaks
attributed to both t
2g
and e
g
3d transitions, was also added.
A single low-intensity peak on the low-binding-energy (BE)
side of the envelope was added to account for the formation
of Fe ions with a lower than normal oxidation state by the
production of defects in neighbouring sites.
15
The defect sites
for many of the compounds studied were more than likely
formed during sample preparation when the surface was
either cleaved in a vacuum or crushed under the cover of
inert gas. In the following sections, this peak is referred to as
the ‘pre-peak’. After the 2p
3/2
envelope was fitted adequately
using the above method, the main peak centre of gravity (CG)
was determined using the GS multiplets only.
RESULTS
Table 1 lists the multiplet peak parameters used when
theFe2p
3/2
envelopes from the high-spin Fe
3C
and Fe
2C
compounds were fitted. Each of the compounds followed the
GS predictions well, allowing for relatively small deviations
to be observed.
High-spin Fe
3+
compounds
Figure 2 shows the Fe
3C
2p
3/2
spectra from four different
oxide species (˛-Fe
2
O
3
, -Fe
2
O
3
, ˛-FeOOH and -FeOOH).
Themaindifferencebetweenthetwosetsofsamplesis
coordination of the Fe
3C
cations. In the ˛-compounds, the
crystal structure is oriented in such a way that all of the
cations are octahedrally coordinated.
13
In the -compounds,
on the other hand, three-quarters of the Fe
3C
cations are
octahedrally coordinated whereas the other quarter of the
cations are tetrahedrally coordinated.
12
Cation vacancies are
also present in the crystal structure of the -compounds
to balance the overall charge. The shape of the Fe
2
O
3
spectra and the spectrum from ˛-FeOOH (Figs. 2(a), 2(b)
and 2(c)) resemble those reported previously by McIntyre
and Zetaruk.
4
All four spectra were fitted using the GS multiplets, high-
BE surface structures, low-BE ‘pre-peaks’ and satellite peaks
(see labels in Fig. 2(a)). The 2p
3/2
envelopes from each of
the four compounds were fit well with the GS multiplets,
with intensities close to those seen in Fig. 1(a). The difference
in energy between the peaks was lower than predicted by
Gupta and Sen
5
(see Table 1) and lower than those found by
McIntyre and Zetaruk;
4
in the case of the latter this is due to
the increased resolution of the instrument used during this
study. It was also found when fitting the four compounds
that the multiplets from the two ˛-compounds tended to have
a smaller FWHM than the -compounds.Thisobservation
Figure 2. Shirley background-subtracted Fe 2p
3/2
spectra of ˛-Fe
2
O
3
(a), -Fe
2
O
3
(b), ˛-FeOOH (c) and -FeOOH (d).
Copyright 2004 John Wiley & Sons, Ltd. Surf. Interface Anal. 2004; 36: 1564–1574
Multiplet splitting of Fe 2p XPS spectra 1567
Table 1. Gupta and Sen (GS) multiplet peak parameters used to fit the high-spin Fe
2C
and Fe
3C
compounds
Compound
Peak 1 (eV)
[FWHM] %
Peak 2 (eV)
[FWHM] %
E
peak2peak1
(eV)
Peak 3 (eV)
[FWHM] %
E
peak3peak2
(eV)
Peak 4 (eV)
[FWHM] %
E
peak4peak3
(eV)
˛-Fe
2
O
3
709.8 [1.0] 34.0 710.7 [1.2] 28.8 0.9 711.4 [1.2] 22.8 0.7 712.3 [1.4] 14.4 0.9
-Fe
2
O
3
709.8 [1.2] 32.6 710.8 [1.3] 32.6 1.0 711.8 [1.4] 24.1 1.0 713.0 [1.4] 10.8 1.2
˛-FeOOH 710.2 [1.3] 31.4 711.2 [1.2] 29.7 1.0 712.1 [1.4] 24.7 0.9 713.2 [1.4] 14.2 1.1
-FeOOH 710.3 [1.4] 31.9 711.3 [1.4] 32.2 1.0 712.3 [1.4] 23.5 1.0 713.3 [1.4] 12.2 1.0
FeF
3
714.1 [1.4] 45.0 715.3 [1.4] 33.5 1.2 716.5 [1.3] 14.2 1.2 717.7 [1.4] 7.3 1.2
FeCl
3
711.3 [1.4] 41.7 712.4 [1.4] 33.2 1.1 713.3 [1.4] 16.2 0.9 714.2 [1.4] 8.9 0.9
FeBr
3
709.2 [1.2] 37.7 710.3 [1.4] 32.0 1.1 711.3 [1.4] 21.9 1.0 712.4 [1.4] 8.7 1.1
Fe
3
O
4
(Fe
3C
) 710.2 [1.4] 34.9 711.3 [1.4] 31.8 1.1 712.4 [1.4] 22.6 1.1 713.6 [1.4] 10.8 1.2
Fe
3C
GS multiplets
b
39.9 30.4 1.6 19.6 1.3 10.1 0.6
FeF
2
710.6 [1.4] 32.7 711.5 [1.6] 41.2 0.9 712.7 [1.6] 26.1 1.2
FeCl
2
709.8 [1.2] 30.8 710.5 [1.5] 41.3 0.7 711.5 [1.4] 27.9 1.0
FeBr
2
707.1 [1.2] 31.3 708 [1.6] 42.4 0.9 708.9 [1.3] 26.3 0.9
FeSO
4
(H
2
O) 710.0 [1.3] 35.3 711.2 [1.5] 39.5 1.2 712.1 [1.6] 25.2 0.9
FeO 708.4 [1.4] 35.2 709.7 [1.6] 43.7 1.3 710.9 [1.6] 21.1 1.2
Fe
3
O
4
(Fe
2C
) 708.3 [1.2] 41.6 709.3 [1.2] 43.2 1.0 710.4 [1.4] 15.2 1.1
Fe
2C
GS multiplets
b
36.1 46.4 1.4 17.5 1.6
a
Peak 1 represents the multiplet with the lowest binding energy for all compounds whereas peak 4 represents the highest binding energy multiplet for theFe
3C
compounds; peak 3 represents
the highest binding energy multiplet for the Fe
2C
compounds.
b
Component information for each of the theoretically derived GS multiplets came from a digitized reproduction of the graphs found in the original paper.
3
Copyright 2004 John Wiley & Sons, Ltd. Surf. Interface Anal. 2004; 36: 1564–1574
1568 A. P. Grosvenor et al.
has been attributed to the differences in orientation of the
Fe
3C
cations in the two sets of compounds.
The high-BE surface peak used to fit the spectra indi-
cated that the relative intensity of the surface structure was
larger for the cleaved mineral sample (˛-Fe
2
O
3
)compared
with the three crushed powder samples (-Fe
2
O
3
, ˛-FeOOH
and -FeOOH). As was noted in the introduction, multi-
ple reasons for the presence of the the higher BE surface
peak have been put forth. One possible cause of the surface
structures is the reduction in coordination of the molecules
located at the surface
8
after the surface has been cleaved in
a vacuum or crushed under the cover of N
2
or Ar. With a
decrease in coordination, the Fe
3C
ion would be surrounded
by a lower electron density requiring more energy to be
used to produce a photoelectron. A decrease in the crystal
field energy of the Fe
3C
ions located at the surface compared
with those found within the bulk cannot be ruled out as the
cause of the high-BE surface peak.
9
Droubay and Chambers
have reported for ˛-Fe
2
O
3
that this is more likely the cause
of the surface peak and that a decrease in coordination of
the Fe
3C
cations located at the surface provides little effect
on the BE of the various multiplets.
9
It should be noted
also that these authors have indicated that a peak with a
BE of ¾715 eV (identified as the surface peak in this report)
has been found in ˛-Fe
2
O
3
XPS spectra comparable to that
shown in this report, regardless of electron take-off angle.
9
This indicates that the high-BE surface peak includes not
only peak(s) pertaining to differences between the bulk and
surface structures but also low-intensity peak(s) from the
bulk structure not well described by the GS-calculated spec-
tra. Although only a single high-BE surface peak is shown
in Fig. 2, this state could still undergo multiplet splitting.
The other multiplet peaks attributed to this surface structure
are overshadowed by those from the bulk material, making
them impossible to be represented in the fitting.
TheFe2p
3/2
lineshapes of FeBr
3
,FeCl
3
and FeF
3
were
also analysed with a view to improving the fitting of high-
spin Fe
3C
compounds (see Fig. 3). For FeF
3
,theF1speakis
very close in binding energy to the Fe 2p
3/2
envelope, with
some of its plasmon loss peaks overlapping the Fe
3C
peak
(Fig. 3(c)). To correct for this, a spectrum of CaF
2
was taken
to determine the binding energies of the plasmon loss peaks
and their intensity ratios compared with the F 1s main peak,
so that a fitting procedure could be determined for both
FeF
3
and FeF
2
. The peaks used to fit the FeBr
3
,FeCl
3
and
FeF
3
Fe 2p
3/2
spectra were similar to those used to fit the
iron oxides discussed above. The relative intensities of the
GS multiplets used for all three compounds were found to
resemble the calculated values more closely than for the Fe
oxides (see Table 1). This difference is due to the fact that Br
,
Cl
and F
are much weaker crystal field splitting ligands
than the OH
and O
2
ligands,
16
therefore the Fe
3C
cations
in these compounds would better resemble the free ion
approximation used to calculate the GS multiplets compared
with the Fe
3C
cations found in the oxides.
High-spin Fe
2+
compounds
Spectra showing the high-spin Fe
2C
2p
3/2
envelope from
FeBr
2
,FeCl
2
,FeF
2
,FeSO
4
and Fe
1.1
OareshowninFig.4.
Figure 3. Background-subtracted Fe 2p
3/2
spectra from FeBr
3
(a), FeCl
3
(b) and FeF
3
(c).
The FeSO
4
sample was first placed in the spectrometer as
FeSO
4Ł
H
2
O
7
; this was heated to a temperature of ¾300
°
C
to form the monohydrate of FeSO
4
before being analysed.
Each spectrum was fit using the GS multiplets as well as
a high-BE surface peak, a low-BE ‘pre-peak’ and a satellite
peak attributed to shake-up. A large peak located within the
Fe 2p
3/2
envelope just above the third multiplet was found
in all of the Fe
2C
spectra. This peak was ascribed to high-BE
surface structures, as well as multiplets resulting from the
presence of small amounts of Fe
3C
, which may have formed
either during sample preparation or while the sample was
being analysed. The high-BE surface peak found in the FeBr
2
spectrum shown in Fig. 4 is believed to not include Fe
3C
or
other contaminates because its FWHM and intensity were
reminiscent of the surface peaks found for the various Fe
3C
Copyright 2004 John Wiley & Sons, Ltd. Surf. Interface Anal. 2004; 36: 1564–1574
