Examination of the bonding in binary transition-metal monophosphides MP (M = Cr, Mn, Fe, Co) by X-ray photoelectron spectroscopy

Article abstract of DOI:10.1021/ic051004d

The binary transition-metal monophosphides CrP, MnP, FeP, and CoP have been studied with X-ray photoelectron spectroscopy. The shifts in phosphorus 2p _3/2 core line binding energies relative to that of elemental phosphorus indicated that the de

Full text of DOI:10.1021/ic051004d

Inorg. Chem. 2005, 44, 8988−8998
Examination of the Bonding in Binary Transition-Metal Monophosphides
MP (M Cr, Mn, Fe, Co) by X-Ray Photoelectron Spectroscopy
)
Andrew P. Grosvenor,* Stephen D. Wik, Ronald G. Cavell, and Arthur Mar
Department of Chemistry, UniVersity of Alberta, Edmonton, Alberta, Canada T6G 2G2
Received June 20, 2005
The binary transition-metal monophosphides CrP, MnP, FeP, and CoP have been studied with X-ray photoelectron
spectroscopy. The shifts in phosphorus 2p core line binding energies relative to that of elemental phosphorus
indicated that the degree of ionicity of the metal phosphorus bond decreases on progressing from CrP to CoP.
The metal 2p core line binding energies differ only slightly and show similar line shapes to those of the elemental
metals, reaffirming the notion that these transition-metal phosphides have considerable metallic character. The
satellite structure observed in the Co 2p X-ray photoelectron spectra of Co metal and CoP was examined by
3/2
−
3/2
3/2
reflection electron energy loss spectroscopy and has been attributed to plasmon loss, not final state effects as has
been previously suggested. Valence-band spectra of the transition-metal phosphides agree well with the density of
states profiles determined from band structure calculations. The electron populations of the different electronic
states were extracted from the fitted valence-band spectra, and these confirm the presence of strong M
weak P P bonding interactions. Atomic charges determined from the P 2p core line spectra and the fitted valence-
band spectra support the approximate formulation M1+P1 for these phosphides.
−P and
−
-
1
. Introduction
metal phosphides may have applications as hydrodesulfur-
ization and hydrodenitrogenation catalysts.
7
The first-row transition metals form a series of mono-
To analyze the bonding in these monophosphides in more
detail, it is desirable to focus on a series of isostructural
members. Here we select the four MnP-type members: CrP,
phosphides, MP (M ) Sc-Ni), whose structures and
properties have been extensively investigated. There is a
progression of structure types, from NaCl-type (ScP) to TiAs-
type (TiP), NiAs-type (VP), MnP-type (CrP, MnP, FeP,
CoP), and NiP-type (NiP), in which successive distortions
arise from the interplay of metal-metal and phosphorus-
phosphorus bonding. These compounds exhibit a wealth of
interesting physical properties; they are all metallic, and many
display unusual magnetic ordering transitions. For example,
a large body of work is available on MnP alone, which
undergoes transitions to ferromagnetic and helimagnetic
8
,9
MnP, FeP, and CoP. This structure, shown in Figure 1, is
based on a distortion of the NiAs-type, one of the most
celebrated structure types in intermetallic chemistry.10 The
metal atoms occupy all the octahedral sites between hcp
layers of phosphorus atoms, so that columns of face-sharing
metal-centered octahedra develop along the stacking axis.
To a first approximation, the electronic structure of these
compounds consists of bands that arise from the interaction
of metal 3d, 4s, and 4p orbitals with the phosphorus 3s and
3p orbitals. The result is a filled, lower-energy, set of M-P
bonding levels, a partially filled set of mostly metal-based d
1
2
3,4
phases. Study of the size-dependent properties of MnP and
FeP nanoparticles has now begun. CoP and other transition-
5,6
*
To whom correspondence should be addressed. E-mail:
levels (split into t2g and e
higher-energy, set of M-P antibonding levels.
g
components), and an empty,
andrew.grosvenor@ualberta.ca.
1) Aronsson, B.; Lundstr o¨ m, T.; Rundqvist, S. Borides, Silicides and
11-13
In the
(
Phosphides; Methuen: London, 1965.
(
(
2) Hulliger, F. Struct. Bonding 1968, 4, 83-229.
(6) Perera, S. C.; Fodor, P. S.; Tsoi, G. M.; Wenger, L. E.; Brock, S. L.
Chem. Mater. 2003, 15, 4034-4038.
(7) Oyama, S. T. J. Catal. 2003, 216, 343-352.
(8) Rundqvist, S. Acta Chem. Scand. 1962, 16, 287-292.
(9) Rundqvist, S.; Nawapong, P. C. Acta Chem. Scand. 1965, 19, 1006-
1008.
3) Huber, E. E., Jr.; Ridgley, D. H. Phys. ReV. Sect. A 1964, 135, 1033-
1040.
(
4) Okabayashi, J.; Tanaka, K.; Hashimoto, M.; Fujimori, A.; Ono, K.;
Okusawa, M.; Komatsubara, T. Phys. ReV. B: Condens. Matter Mater.
Phys. 2004, 69, 132411-1-132411-4.
5) Perera, S. C.; Tsoi, G.; Wenger, L. E.; Brock, S. L. J. Am. Chem.
Soc. 2003, 125, 13960-13961.
(
(10) Pearson, W. B. The Crystal Chemistry and Physics of Metals and
Alloys; Wiley-Interscience: New York, 1972.
8988 Inorganic Chemistry, Vol. 44, No. 24, 2005
10.1021/ic051004d CCC: $30.25
© 2005 American Chemical Society
Published on Web 10/22/2005

Bonding in Binary Transition-Metal Monophosphides
phosphorus bonds. For the first time, the metal 2p3/2 peak
shapes in these monophosphides have been analyzed in detail
and compared to those in clean samples of the parent
transition metal. The origin of the satellite structure found
in the Co 2p3/2 spectra in Co metal and in CoP will also be
discussed. Valence orbital electron populations in these
compounds were extracted, for the purpose of bonding
analyses, from peak fitting deconvolution of the valence-
band spectra.
2
. Experimental Section
.1. Synthesis. The monophosphides were prepared from sto-
2
ichiometric reaction of Cr (99.95%, Alfa-Aesar), Mn (99.95%,
Cerac), Fe (99.9%, Cerac), or Co (99.999%, Spex) metal powder
with red P (99.995%, Cerac), placed in evacuated fused-silica tubes.
The tubes were heated to 1323 K over a 36-h period and maintained
at this temperature for 4 days before being cooled over 12 h to
room temperature. The products were isolated and stored in a
glovebox under argon to limit exposure to air. The powder X-ray
diffraction patterns, obtained on an Inel powder diffractometer
equipped with a CPS 120 detector, were in good agreement with
those calculated from the literature crystallographic data and
revealed that the samples represented pure phases.
Figure 1. MnP-type crystal structure of MP (M ) Cr, Mn, Fe, Co) viewed
down the b axis. The small solid spheres are M atoms and the large lightly
shaded spheres are P atoms. The dashed lines indicate the metal-metal
bonding network (2.6-2.8 Å).
orthorhombic MnP-type structure, there is metal-metal
bonding not only along the columns of face-sharing octahedra
along the a direction but also between the columns in the
form of zigzag chains running along the b direction. The
metal-metal bonding network originates from the interaction
between the metal-based t2g orbitals, whose lobes are directed
away from the metal-phosphorus bonds. These metal-metal
interactions are quite strong, as indicated by bond lengths
2.2. XPS Analysis. All measurements were performed on a
Kratos AXIS Ultra spectrometer equipped with a monochromatic
Al KR X-ray source. The pressures throughout the analysis chamber
-
6
-7
were 10 -10 Pa. The resolution function for this instrument
has been determined to be 0.4 eV by analysis of the cobalt Fermi
edge.
Samples were finely ground in a glovebox and pressed into 0.5-
mm thick In foil (Alfa-Aesar) before being placed on a Cu sample
holder and transferred to the XPS instrument in a sealed container
to reduce exposure to air during transport. After being loaded into
9
of 2.6-2.8 Å. The t2g-based bands are expected to broaden
g
and overlap significantly with the e -based bands with no
gap between them. The structural distortion also generates
zigzag chains of phosphorus atoms running along the b
+
the instrument, the samples were sputter-cleaned with an Ar ion
9
direction but at distances (2.6-2.7 Å) that are somewhat
beam (4 kV, 10 mA) to remove any surface oxide or phosphate
which had formed. The time required for sputter-cleaning depended
on the degree of oxidation present (as ascertained by initial survey
spectra of the as-received samples) and ranged from 10 (CoP) to
45 min (MnP). After the samples were sufficiently clean, both
survey (broad range) and high-resolution spectra were obtained.
Survey spectra were collected with a binding energy (BE) range
of 0-1100 eV, a pass energy of 160 eV, a step size of 0.7 eV, a
longer than the typical single P-P covalent bond length of
14
2
.2 Å. Several questions arise. To what extent does electron
transfer occur from metal to phosphorus atoms? How
3+ 3-
reasonable are formulations such as M P (which assumes
isolated P anions) or M P (which assumes 2 center-2e-
1+ 1-
bonds in the zigzag chains of P atoms)? How closely does
the metal-metal bonding character in these compounds
resemble that in the elemental metals?
2
sweep time of 180 s, and a spot size of 700 × 400 µm . High-
resolution spectra were collected with an energy envelope of 20-
The electronic structure of some first-row transition-metal
monophosphides has been previously investigated with both
experimental X-ray photoelectron spectroscopy (XPS)
45 eV (depending on the peak being examined (phosphorus 2p,
metal 2p3/2, or valence band)), a pass energy of 20 eV, and a step
size of 0.05 eV. All results were analyzed with use of the CasaXPS
1
5-18
11-13,19,20
21
studies
and theoretical calculations.
Here we
software package. During this study, it was found that charge
describe new high-resolution XPS measurements of CrP,
MnP, FeP, and CoP, with the goal of clarifying the electronic
structure and bonding in these important compounds. In
particular, the metal 2p and phosphorus 2p binding energies
have been measured, and the trends in these energies have
been related to the degree of covalency in the metal-
correction was not required, most likely because of the metallic
(
15) Domashevskaya, E. P.; Terekhov, V. A.; Ugai, Ya. A.; Nefedov, V.
I.; Sergushin, N. P.; Firsov, M. N. J. Electron Spectrosc. Relat.
Phenom. 1979, 16, 441-453.
(16) Myers, C. E.; Franzen, H. F.; Anderegg, J. W. Inorg. Chem. 1985,
2
4, 1822-1824.
(
17) Okuda, H.; Senba, S.; Sato, H.; Shimada, K.; Namatame, H.; Taniguchi,
M. J. Electron Spectrosc. Relat. Phenom. 1999, 101-103, 657-660.
(
(
(
(
11) Yanase, A.; Hasegawa, A. J. Phys. C: Solid State Phys. 1980, 13,
989-1993.
12) Perkins, P. G.; Marwaha, A. K.; Stewart, J. J. P. Theor. Chim. Acta
981, 59, 569-583.
13) Tremel, W.; Hoffmann, R.; Silvestre, J. J. Am. Chem. Soc. 1986, 108,
(18) Shabanova, I. N.; Mitrochin, Y. S.; Terebova, N. S.; Nebogatikov, N.
M. Surf. Interface Anal. 2002, 34, 606-609.
1
(19) Nol a¨ ng, B.; Eriksson, O.; Johansson, B. J. Phys. Chem. Solids 1990,
51, 1033-1046.
1
(20) Scott, B. A.; Eulenberger, G. R.; Bernheim, R. A. J. Chem. Phys.
1968, 48, 263-272.
5174-5187.
14) Pauling, L. The Nature of the Chemical Bond, 3rd ed.; Cornell
(21) Fairley, N. CasaXPS, version 2.2.19; Casa Software Ltd.: Teighn-
mouth, Devon, UK, 2003 (www.casaxps.com).
University Press: Ithaca, NY, 1960.
Inorganic Chemistry, Vol. 44, No. 24, 2005 8989

Grosvenor et al.
nature of these compounds. To fit the high-resolution spectra, a
Shirley-type function was first used to remove the background
arising from energy loss. The extracted spectra were then fitted
with a combined Gaussian (70%) and Lorentzian (30%) line profile
to account for spectrometer and lifetime broadening, respectively.
The chemical compositions of the samples were determined from
both these X-ray photoelectron (XP) survey spectra and independent
EDX analysis (Hitachi S-2700 SEM; 20 kV beam voltage). Within
experimental accuracy, all samples were essentially stoichiometric
(50 ( 5% M; 50 ( 5% P).
XP spectra of metal standards were also collected. Samples (in
the form of foils or pieces) of Cr (Fisher Scientific), Mn (99.98%,
Alfa-Aesar), Fe (99.995%, Alfa-Aesar), and Co (99.0%, British
Drug House) metal were sputter-cleaned to remove any surface
oxide present and analyzed using similar parameters as above. Both
survey and high-resolution spectra (metal 2p3/2 and valence band)
were obtained.
2.3. REELS Analysis. To examine the nature of the satellite
peaks observed in the XP spectrum of Co metal (99.995%, Alfa-
Aesar), reflection electron energy loss spectroscopy (REELS) was
performed on a PHI 660 scanning Auger microprobe with electron
beam energies of 3-0.2 keV. Before analysis, the sample was
+
cleaned in situ by Ar ion-beam sputtering at 3 kV. The sample
was considered suitably clean when no O KLL line was observed
in the Auger spectra. This analysis was performed at Surface
Science Western (SSW) at the University of Western Ontario
(UWO).
2.4. Band Structure Calculations. Although theoretical calcula-
tions have been previously performed on some MnP-type com-
pounds, it was desirable to extract partial density of states profiles
for the metal and phosphorus valence states to illustrate the fitting
of the experimental valence-band spectra here. A tight-binding
extended H u¨ ckel band structure calculation on MnP was performed
with 96 k-points in the irreducible portion of the Brillouin zone,
22,23
with use of the EHMACC suite of programs,
parameters (valence shell ionization potentials Hii (eV) and orbital
exponents ú ) were as follows. For Mn 4s, Hii ) -9.8, ú ) 1.80;
for Mn 4p, Hii ) -5.9, ú ) 1.80; for Mn 3d, Hii ) -11.7, úi1
) 0.514, úi2 ) 1.70, c ) 0.693; for P 3s, Hii ) -18.6, ú
1.84; for P 3p, Hii ) -14.0, ú ) 1.45.
The atomic
i
i
i
)
i
5
)
.15, c
1
2
Figure 2. High-resolution P 2p XP spectra of (a) CrP, (b) MnP, (c) FeP,
and (d) CoP. Here and in other spectra, the intensity is in counts per second
i
(cps).
3
. Results and Discussion
.1. Phosphorus 2p High-Resolution X-Ray Photoelec-
tron Spectra. Figure 2 shows the high-resolution phosphorus
p XP spectra collected for CrP, MnP, FeP, and CoP. Each
Table 1. Binding Energies (eV) of Core Photoelectron Peaks in
Transition-Metal Monophosphides MP (M ) Cr, Mn, Fe, Co) and Metal
Standards
3
CrP
MnP
FeP
CoP
130.4
129.5
778.3
2
P 2p1/2
P 2p3/2
M 2p3/2
130.0
129.1
573.9
130.2
129.2
638.7
130.2
129.3
706.9
spectrum shows two peaks corresponding to the P 2p3/2 (low
binding energy) and P 2p1/2 (high binding energy) states.
Sharp line shapes were observed which could be fitted with
component peaks having a FWHM as small as 0.7 eV. In
all cases, the intensity ratio of the P 2p3/2 and 2p1/2 peaks
a
a
(781.4, 783.0)
Cr
Mn
Fe
Co
M 2p3/2
574.1
638.7
707.0
778.1
was 2:1, which is consistent with the expected ratio of (2j
1)/(2j + 1), where j and j represent the coupled orbital
l) and spin (s) angular momentum quantum numbers from
respective spin-up and spin-down states of the unpaired core
1
(781.1, 783.1)
+
(
2
1
2
a
Satellite peaks arising from plasmon loss processes.
24
electron which remains after photoionization. On progress-
ing through the series CrP to CoP, the P 2p3/2 and 2p1/2
binding energies gradually increase (Table 1). On the high-
binding-energy side in all spectra, there is a small peak which
is attributed to the P 2p1/2 signal from some remaining
unreacted phosphorus; the corresponding P 2p3/2 peak is not
indicated, as it would be buried under the main P 2p1/2 peak
from the transition-metal monophosphide.
(
(
(
22) Whangbo, M.-H.; Hoffmann, R. J. Am. Chem. Soc. 1978, 100, 6093-
6098.
23) Hoffmann, R. Solids and Surfaces: A Chemist’s View of Bonding in
Extended Structures; VCH Publishers: New York, 1988.
24) Briggs, D., Seah, M. P., Eds. Practical Surface Analysis, Vol. 1: Auger
and X-ray Photoelectron Spectroscopy, 2nd ed.; Wiley: Chichester,
1990.
8990 Inorganic Chemistry, Vol. 44, No. 24, 2005

Bonding in Binary Transition-Metal Monophosphides
Figure 3. Plot of P 2p3/2 binding energies, BE, versus the difference in
electronegativity (Allred-Rochow) between the phosphorus and metal
Figure 4. Plot of the charge on P atoms (estimated from binding energy
values) versus the difference in electronegativity, ∆ø, for CrP, MnP, FeP,
and CoP.
25
atoms, ∆ø, for several first-row transition-metal monophosphides. Values
indicated by solid circles (b) are from this work, and those by open circles
(
O) are from ref 16.
Indeed, the trend in P-P bond lengths through the series
9
CrP (2.644 Å) to CoP (2.701 Å), as well as charges
Figure 3 shows a plot of the P 2p3/2 binding energies versus
1
3
calculated in theoretical studies, is opposite to that sug-
gested here, implying that decreased metal-to-phosphorus
electron transfer (from CrP to CoP) is the more important
effect in determining the binding energies observed.
2
5
the difference in electronegativity (Allred-Rochow) be-
tween the phosphorus and metal atoms for these and other
first-row transition-metal monophosphides. The binding
energies are lower in the monophosphides than in elemental
phosphorus, indicating the presence of anionic phosphorus
in these compounds. The trend of increasing phosphorus
binding energy through the series from CrP to CoP clearly
confirms our expectation of decreasing ionic character in the
metal-phosphorus bonding as the electronegativity differ-
ence between the elements decreases. Interestingly, theoreti-
cal studies on the diatomic gas-phase molecules MP (M )
Sc-Cu) also show the same general trend of decreasing ionic
character.26
Since core-level binding energies are affected by the
chemical environment, the charge on the phosphorus atoms
in these compounds may be estimated from the P 2p3/2
binding energies. Binding energies for a large number of
compounds containing the same transition metals and
phosphorus in a variety of oxidation states (as well as that
of elemental phosphorus) are available from the NIST XPS
3
.2. Metal 2p3/2 High-Resolution X-Ray Photoelectron
Spectra. Figure 5 shows the high-resolution metal 2p3/2
spectra for CrP, MnP, FeP, and CoP. The binding energies
(Table 1) are similar to those found for the elemental metal
(
within 0.1-0.2 eV) and to values previously reported by
16
Myers et al. However, the high-resolution spectra presented
here show a distinct asymmetric line shape that, to our
knowledge, has not been previously interpreted for these
compounds. This feature is characteristic of significant
metal-metal bonding such as has been proposed for MnP-
1
3
type structures.
To substantiate the inference of metal-metal bonding, we
have collected high-resolution 2p3/2 spectra for the elemental
metals Cr, Mn, Fe, and Co, which are shown in Figure 6.
Each spectrum shows an asymmetric line shape skewed to
higher binding energy. According to Doniach and Sˇ unji c´ ,28
this type of line shape arises when valence electrons,
interacting with the core hole (produced after photoioniza-
tion), are excited and scattered from filled states below the
Fermi edge to empty conduction states above. Since there is
a continuum of states above and below the Fermi edge in a
metal, an asymmetric tail comprising many closely spaced
states is observed instead of a few distinct satellite peaks.29
The spectrum for Co metal, however, reveals not only an
asymmetric line shape but also a distinct satellite structure
2
7
database. If a linear relationship is assumed between the P
3/2 binding energy and the oxidation state, an interpolation
2
p
of the binding energies for the transition-metal monophos-
phides suggests P charges of -1.4 in CrP, -1.1 in MnP,
-
1.0 in FeP, and -0.7 in CoP. A plot of these values versus
the electronegativity difference is shown in Figure 4,
consistent with the decreasing ionicity of the metal-
phosphorus bond through the sequence from CrP to CoP.
3
+ 3-
These values also suggest that the formulation M P is
untenable, and in fact, they are remarkably consistent with
30,31
(Figure 6d) which has been observed before.
To properly
fit the spectrum, two satellite peaks centered at 5.0 and 3.0
eV away from the main core line were required. Note that
the spectrum for CoP also shows two satellite peaks at similar
energies as in Co metal, but with much lower intensities
1+ 1-
the formulation M P mentioned earlier in which some
P-P bonding is implicated. However, caution must be
applied because it is difficult to separate the effect of electron
count, which modifies the degree of P-P bonding, and the
effect of decreased ionicity in the metal-phosphorus bond
on the ultimate charge adopted by the phosphorus atoms.
(Figure 5d). A detailed interpretation of these satellite peaks
is important but is deferred to a later section.
The similarity in the binding energies and line shapes of
the metal 2p3/2 spectra for both the transition-metal mono-
(
(
(
25) Allred, A. L.; Rochow, E. G. J. Inorg. Nucl. Chem. 1958, 5, 264-
68.
26) Tong, G. S. M.; Jeung, G. H.; Cheung, A. S.-C. J. Chem. Phys. 2003,
18, 9224-9232.
27) Wagner, C. D.; Naumkin, A. V.; Kraut-Vass, A.; Allison, J. W.;
Powell, C. J.; Rumble, J. R., Jr. NIST X-ray Photoelectron Spectros-
copy Database, version 3.4 (web version); National Institute of
Standards and Technology: Gaithersburg, MD, 2003 (srdata.nist.gov/
xps).
2
(28) Doniach, S.; Sˇ unji c´ , M. J. Phys. C: Solid State Phys. 1970, 3, 285-
291.
(29) H u¨ fner, S. In Photoemission in Solids II; Ley, L., Cardona, M., Eds.;
Springer-Verlag: Berlin, 1979.
(30) Raaen, S. Solid State Commun. 1986, 60, 991-993.
(31) Richardson, D.; Hisscott, L. A. J. Phys. F: Metal Phys. 1976, 6,
L127-L134.
1
Inorganic Chemistry, Vol. 44, No. 24, 2005 8991

Grosvenor et al.
Figure 6. High-resolution 2p3/2 XP spectra of clean samples of (a) Cr,
Figure 5. High-resolution metal 2p3/2 XP spectra for (a) CrP, (b) MnP,
c) FeP, and (d) CoP. The inset in Figure 5d shows the observed low-
(b) Mn, (c) Fe, and (d) Co metal.
(
intensity satellite structure for CoP.
ture is also localized. At first glance, these binding energy
results seem to imply a problem with charge neutrality in
these transition-metal monophosphides, as the metal atoms
are assigned charges of nearly 0 whereas the phosphorus
atoms are assigned charges of nearly -1. It is interesting to
note that a similar problem was encountered in an earlier
band structure calculation, which revealed that the clear
presence of some P-P bonding (as evaluated by overlap
populations) was not reflected correctly in the charges borne
phosphides and the elemental metals argues for the presence
of metal atoms with nearly zero charge and substantial
metal-metal bonding in the monophosphides. Relative to a
metal atom in a cationic state with a localized electronic
structure, photoelectrons from a metal atom participating in
extended metal-metal bonding may have a lower binding
energy because of the delocalization of electrons in conduc-
tion states allowing the atom to experience, on average,
greater nuclear screening. As well, the strong hybridization
of metal- and phosphorus-based states arising from the
primarily covalent nature of the bonding interactions reduces
the difference in binding energy of the metal atoms in the
monophosphides versus in the elemental metals. In contrast,
the difference in binding energy observed for the P atoms
in the monophosphides compared to elemental phosphorus
suggests that the electrons in the monophosphide P atoms
are localized and any bonding within the anionic P substruc-
13
by the phosphorus atoms. Resolving this problem involves
a more sophisticated interpretation of the XP spectra, given
below.
3
.3. Satellite Structure in Co Metal and CoP Co 2p3/2
Spectra. The satellite structure in the Co 2p3/2 spectra of
Co metal and CoP noted above is also observed in the
spectrum of Ni metal, which has received considerable
32
3
3-35
attention in the past.
Two explanations have been
(
(
33) H u¨ fner, S. Photoelectron Spectroscopy; Springer-Verlag: Berlin, 1995.
34) Kemeny, P. C.; Shevchik, N. J. Solid State Commun. 1975, 17, 255-
258.
(
32) Anno, H.; Matsubara, K.; Caillat, T.; Fleurial, J.-P. Phys. ReV. B:
Condens. Matter Mater. Phys. 2000, 62, 10737-10743.
(35) H u¨ fner, S.; Wertheim, G. K. Phys. Lett. A 1975, 51, 299-300.
8992 Inorganic Chemistry, Vol. 44, No. 24, 2005

Bonding in Binary Transition-Metal Monophosphides
developed to account for this satellite structure: plasmon
loss (loss of kinetic energy of the photoelectron when it
interacts with bound valence electrons, causing them to
3
6
33
oscillate) and the “two core hole” theory. The former was
initially abandoned when no evidence of plasmon loss was
33
found in optical or electron energy loss spectra. Instead, it
was suggested that the satellite structure found at 6 eV above
the core line in the XP spectrum of Ni metal could arise
33
from final state effects, similar to those proposed by Kotani
and Toyozawa.37 After photoionization, the resulting core
hole increases the effective charge operating on the remaining
electrons. This increased charge pulls conduction states below
the Fermi energy. If the conduction state is empty, then the
satellite peak is observed because of a final state containing
two holes (one in the core level present after photoionization
and one in the valence band). If a filled conduction state is
pulled below the Fermi energy, then this final state represents
the observed core line.33 The same description has been
applied to describe the satellite peak observed in the spectrum
3
0
of Co metal.
Although the “two core hole” theory has been widely
accepted, there is evidence that the plasmon loss explanation
deserves further consideration. Previous electron energy loss
studies of the first-row transition metals have revealed the
presence of two peaks at <10 eV below the main elastic
line for Co and Ni metal.3 These peaks were attributed to
plasmon loss, as well as to M4,5 ionization.38 It should be
noted that these results were not specifically used to
rationalize the nature of the satellite peak found in the XP
spectrum of Ni metal during the period in which the “two
core hole” theory was developed. Recently, a careful
re-examination of REELS data for Ni metal has revealed
peaks at a similar energy separation from the main elastic
line, as is found for the satellite peak from the core line
8,39
Figure 7. (a) REEL spectrum of Co metal taken at 0.3 keV. (b) Expanded
spectrum showing the plasmon loss region located just below the main
elastic line. (c) First derivative of the expanded spectrum.
Further, REELS analysis of NiO showed no such similar
peaks, as expected since no plasmon loss satellite structures
are observed in its corresponding Ni 2p3/2 XP spectrum.40
This confirms that the peaks observed for Ni metal are real
and not due to noise or other associated spectrometer effects.
Plasmon loss peaks normally occur at much higher energy
than was observed during this study of Ni metal (typically
40
observed in the Ni 2p3/2 XP spectrum. These features were
attributed to both bulk and surface plasmon structures.
Plasmon loss events occurring by interaction of the exiting
photoelectrons with valence electrons of atoms located closest
to the surface (surface plasmon loss) are lower in energy
than those occurring by interaction with valence electrons
3
8,39
∼20 eV as determined by EELS).
However, similar low-
energy plasmon loss has been observed by EELS of Mo (110)
after deposition of two monolayers of Ba onto the surface.41
The surface plasmon peak found at 2.6 eV was proposed to
arise from interaction of the exiting electrons with the layer
of Ba atoms located closest to the surface, whereas the bulk
plasmon peak found at 6.7 eV was attributed to interactions
of the exiting electrons with the layer of Ba atoms beneath
4
1
of atoms located below the surface (bulk plasmon loss). If
the presence of the satellite peak in the XP spectrum were
really due to the process involved in the “two core hole”
theory, then no corresponding transitions should be observed
in the REEL spectrum because photoionization does not
40
41
occur in this experiment. The fact that these features were
indeed observed in the REELS data militates against the “two
core hole” explanation and instead favors the plasmon loss
explanation.
it. These observations imply that the interaction of electrons
with atoms found within the uppermost region of a surface
results in a much lower energy plasmon loss than is the case
deeper within the surface. Moreover, in the initial reports
that suggested the presence of lower energy surface plasmon
peaks, it was observed that, although these losses were found
during experiments using EELS in reflection mode, they were
not specifically observed when transmission EELS was
(
36) Egerton, R. F.; Malac, M. J. Electron Spectrosc. Relat. Phenom. 2005,
143, 43-50.
(
37) Kotani, A.; Toyozawa, T. J. Phys. Soc. Jpn. 1974, 37, 912-919.
(
38) Robins, J. L.; Swan, J. B. Proc. Phys. Soc., London 1960, 76, 857-
869.
39
used.
(
(
(
39) Misell, D. L.; Atkins, A. J. J. Phys. C: Solid State Phys. 1972, 5,
To understand the nature of the satellite peaks observed
in the XP spectra of Co metal and CoP, a REELS analysis
was performed on Co metal (Figure 7). The overall spectrum,
taken at 0.3 keV (Figure 7a), compares well with that
95-106.
40) Grosvenor, A. P.; Biesinger, M. C.; Zou, F.; Smart, R. St.-C.; McIntyre,
N. S. Surf. Sci. 2005, submitted.
41) Gorodetsky, D. A.; Melnik, Yu. P.; Proskurin, D. P.; Sklyar, V. K.;
Usenko, V. A.; Yas’ko, A. A. Surf. Sci. 1998, 416, 255-263.
Inorganic Chemistry, Vol. 44, No. 24, 2005 8993

Grosvenor et al.
Table 2. Bulk Plasmon Loss Peak Energies (eV) in Co Metala
energy separation
line
peak energy
from elastic peak
elastic peak
295.5
291.5
290.0
289.0
287.5
286.0
A
B
C
D
E
4.0
5.5
6.5
8.0
9.5
a
Energies were determined from the first derivative of the REEL
spectrum (Figure 7c) taken using a beam energy of 0.3 keV. Surface plasmon
loss peaks were also observed in spectra taken at 0.2 keV with energies of
2
and 3 eV from the main elastic line.
obtained by Robins and Swan.38 In the energy-loss region
at ∼10 eV below the elastic line, five low-intensity peaks
were observed (Figure 7b). The positions of these peaks,
which are attributed to plasmon loss, were located more
precisely by analyzing a plot of the first derivative of the
intensity (Figure 7c) and are listed in Table 2. The peaks
labeled A and B are the most intense (after subtraction of
the background due to inelastic electron scattering); they are
believed to represent bulk plasmon loss and also appear in
the corresponding XP spectrum as a broad peak centered at
5
6
eV above the main 2p3/2 core line (satellite 2 in Figure
d). The bulk plasmon event might be represented by two
peaks rather than one because of interactions with different
valence states, i.e., Co 3d and Co 4s states. Peaks C, D, and
E may represent higher-energy plasmon loss or multiple
plasmon loss events. The broad satellite peak centered at
3.0 eV above the core line in the XP spectrum of Co (labeled
as satellite 1 in Figure 6d) was not observed in the REELS
data taken at 0.3 keV, likely because it is overlapped by the
intense and broad elastic peak. A REEL spectrum was also
taken at a lower beam energy of 0.2 keV (not shown). Intense
peaks separated only slightly from the elastic line, at 2 and
3
eV away, were observed and attributed to surface plasmon
Figure 8. Fitted valence-band spectra of sputter-cleaned samples of (a)
Cr, (b) Mn, (c) Fe, and (d) Co metal. The spectra have been fitted using
two peaks representing 3d5/2 and 3d3/2 states with a fixed intensity ratio of
loss since such peaks were not observed in spectra taken
using higher-energy and, thus, less-surface-sensitive electron
beams. Thus, it appears that satellite 1 arises from surface
plasmon loss and satellite 2 arises from bulk plasmon loss
in the XP spectra of Co metal (Figure 6d) and CoP (Figure
3:2.
Co metal. That is, analysis of the satellite structure points
strongly to the conclusion that, despite their similar binding
energies, the Co atoms in CoP are definitely cationic relative
to the elemental metal. By analogy, we assert that this
argument can be extended to the other monophosphides
studied as well.
5d). As pointed out earlier, these losses most likely arise
from interactions of the exiting photoelectrons and valence
electrons of atoms located within the uppermost region of
the surface. It should also be noted that as the electron beam
energy was increased, the higher energy loss peaks were
consistently observed, indicating that these are real transitions
and not noise.
The Co 2p3/2 XP spectra of Co metal and CoP could be
fitted with individual peaks representing the different plas-
mon loss structures observed in the REELS data, but it was
found to be much simpler to model these loss events using
two broad peaks (FWHM > 2 eV). The observation of the
plasmon loss peaks in CoP at similar energy separations as
those found in Co metal (Table 1) indicates that similar
valence electrons are involved in the plasmon loss process
3.4. Valence-Band Spectra. Comparing the valence-band
spectra of the elemental metals and the metal phosphides
provides further insight into the bonding interactions present
in these compounds. The valence-band spectra of the metals
(Figure 8) show a broad peak located at the Fermi edge which
can be fitted by two peaks representing 3d5/2 and 3d3/2 states
with a fixed intensity ratio of 3:2. As was earlier observed
in the metal 2p3/2 spectra (Figure 6), it was important to
model these peaks as asymmetric to higher binding energy
to properly fit the spectra, indicating that Doniach- Sˇ unji c´
processes also occur during excitation of the 3d valence
2
9
(most likely Co 3d and 4s electrons). However, the lower
electrons. No attempt was made to include a peak repre-
senting 4s states since it would be broad and have an energy
similar to the 3d states with a lower intensity due to a low
intensity of the satellite peaks in CoP suggests that there are
fewer valence electrons around the Co atoms in CoP than in
8994 Inorganic Chemistry, Vol. 44, No. 24, 2005

Bonding in Binary Transition-Metal Monophosphides
1
5
to P 3s states whereas the broad region above it (8-3 eV)
1
8
has been attributed to P 3p bonding states. The transition-
metal 3d states (containing both t2g and e components) have
been suggested to lie close to E along with P 3p nonbonding
g
F
states having a similar to slightly lower binding energy.18
The position of the P 3p nonbonding states lying above the
transition-metal 3d states has been implied previously in band
1
8
structure calculations of transition-metal monophosphides,
44
2
as well as in the valence-band spectrum of Ni P. As an
example, the unfitted valence-band spectrum for MnP can
be compared to the P 3s, P 3p, and Mn 3d projections of the
density of states, as determined from an extended H u¨ ckel
calculation (Figure 10). The highest BE region can be
assigned to P 3s states, the region above it to P 3p states,
followed by Mn 3d states; there is also a small contribution
of P 3p nonbonding states closest to the Fermi edge. Guided
in this manner, we have developed a fitting methodology
for the valence spectra of all four compounds, described
below.
The superior resolution of the spectra obtained here allows
deconvolution into components, labeled from 1 to 7 (Figure
9
), in an attempt to separate individual valence states.
Although there is some degree of overlap, each of the peaks
shown is required to optimize the fit to the spectrum,
especially the shoulder features, which appear reproducibly.
Consistent with calculated band structures and previously
reported XPS results, the following assignments are pro-
posed. Peak 1 represents P 3s states, and peaks 2 and 3
represent P 3p1/2 and 3p3/2 states, respectively, with a fixed
intensity ratio of 1:2. Peaks 4 and 5 represent the t2g set of
the metal 3d states and were fitted with an asymmetric line
shape profile (as was observed in the valence-band spectra
of the elemental metals, discussed above) with a fixed
intensity ratio of 3:2 (3d5/2/3d3/2). Peak 6 represents the e
g
Figure 9. Fitted valence-band spectra of sputter-cleaned samples of (a)
set (M-P antibonding) of the metal 3d states and was fitted
with a single profile constrained to represent spin-up
electrons. Peak 7 likely represents essentially nonbonding P
3p states. To properly fit this state, only a Gaussian line shape
was required, suggesting that spectrometer effects are
dominant; lifetime effects are probably small, given the low
energy of this feature, which appears just below the Fermi
edge. The identity of peaks 2a and 3a found in the spectrum
of CrP is discussed below. The binding energies and FWHM
values of the component peaks used to fit the valence-band
spectra are listed in Table 3 and were optimized to provide
the best fit.
Consistent with the expectation that the major orbital
interactions should be between the metal 3d and phosphorus
3p states, these peaks show considerable overlap in energy.
In this analysis, we have made a distinction for the
phosphorus 3p states, which are dispersed over a very wide
energy range according to band structure calculations (Figure
10c), into those that are clearly bonding in character (near
5-7 eV) and those that are essentially nonbonding in
character (close to the Fermi edge). Justifying this analysis,
we note that the P 3p bonding peaks are wide and their
CrP, (b) MnP, (c) FeP, and (d) CoP.
electron population.13,33 In principle, the d orbitals are not
quite degenerate and the fit could be improved by employing
multiple 3d peak sets, but the two peaks used here were
deemed sufficient to give a reasonable fit.
The fitted valence-band spectra of the transition-metal
monophosphides are shown in Figure 9. In general terms,
the measured valence-band spectra are in excellent qualitative
agreement with calculated band structures, which show a
sharp Fermi edge and a narrow, metal-based, d band with a
high density of states superimposed partially on a broader,
largely phosphorus-based p band.1
1-13,17,18
They also resemble
4
,16,17
16,18
spectra previously reported for MnP,
FeP,
and
4
2
CoSb, but the use of different excitation energies in
different studies affects the photoionization cross-sections
and thus modifies the peak intensities.43
To fit the spectra presented in Figure 9 with both
transition-metal and phosphorus states, they were compared
to previously reported band structure calculations and XPS
F
results. The region 15-10 eV below E has been attributed
(
(
42) Liang, K. S.; Chen, T. Solid State Commun. 1977, 23, 975-978.
43) Scofield, J. H. J. Electron Spectrosc. Relat. Phenom. 1976, 8, 129-
37.
(44) Kanama, D.; Oyama, S. T.; Otani, S.; Cox, D. F. Surf. Sci. 2004, 532,
8-16.
1
Inorganic Chemistry, Vol. 44, No. 24, 2005 8995

Grosvenor et al.
Table 3. Binding Energies (eV) of Component Peaks in Valence-Band
Spectra of MP (M ) Cr, Mn, Fe, Co)a
peak assignment
CrP
MnP
FeP
CoP
1
2
P 3s
P 3p1/2
11.8 (3.4) 11.7 (3.4) 11.9 (3.1) 12.3 (2.9)
8.6 (1.3),b
.8 (1.8)
6.7 (1.9)
4.8 (2.4)
7.3 (2.4)
5.6 (2.2)
8.2 (2.0)
6.1 (2.5)
5
b
3
P 3p3/2
6.8 (1.7),
.3 (2.0)
4
4
5
6
7
M 3d3/2 t2g
M 3d5/2 t2g
M 3d5/2 eg
P 3p
3.2 (2.2)
2.0 (1.0)
1.4 (0.7)
0.6 (1.0)
3.3 (1.8)
2.0 (1.3)
1.2 (0.8)
0.5 (0.8)
3.9 (2.0)
2.3 (1.5)
1.2 (1.1)
0.5 (0.8)
4.3 (2.3)
2.2 (1.3)
1.1 (1.0)
0.5 (0.7)
a
FWHM values (eV) are indicated in parentheses. b A second set of
higher-binding-energy P 3p1/2 and P 3p3/2 peaks, labeled, respectively, as
peaks 2a and 3a in Figure 9a, is present in the spectrum of CrP.
exiting photoelectrons interacted with the valence electrons
of the metal atoms in the pure metal and in the monophos-
phide, the photoelectrons lost similar amounts of energy. The
general increase in the binding energies of the metal 3d3/2
t
2g peak with a concomitant increase in the energy separation
between this peak and the corresponding 3d5/2 2g peak as
the atomic number of the metal increases is normally
t
2
4
observed in XP spectra.
The electron populations of the different orbital states in
the valence-band spectra were extracted from normalized
4
5
peak intensities, C
i
, according to eq 1.
I /(σ λ )
i
i i
C )
(1)
i
n
I /(σ λ )
∑
j
j j
j)1
In this equation, I
of the peak under consideration and ∑
i
i i
/(σ λ ) represents the corrected intensity
n
I
j
j j
/(σ λ ) represents
j)1
the sum of the corrected intensities from all of the peaks
present in the valence-band spectrum. The values of the
photoionization cross-section (σ) and inelastic mean free path
(
IMFP, λ) corrections are listed in Table 4. Much of the
variation in the shapes of the valence-band spectra shown
in Figure 9 arises from the different cross-sections for the
metal 3d states. The number of electrons in each state was
then calculated assuming a total of 11, 12, 13, and 14
electrons in the valence band of CrP, MnP, FeP, and CoP,
respectively. The results of these calculations are summarized
in Table 5.
Figure 10. (a) Valence-band spectrum of MnP compared to the calculated
projections of the density of states for (b) P 3s, (c) P 3p, and (d) Mn 3d
orbitals. The shaded regions represent the projections, whereas the solid
line represents the total density of states. The dashed line marks the Fermi
level.
binding energies show a significant shift through the series
CrP to CoP (Table 3), which is to be expected if they are
truly involved in bonding interactions. In contrast, the P 3p
nonbonding peaks are narrow and their binding energies
show little change. Interestingly, the binding energies of the
P 3s states (near 12 eV) show only a slight shift on going
from CrP to CoP, similar to that observed for the obviously
nonbonding P 2p states, implying that they are less important
than the 3p states in bonding interactions as has been
From the total electron populations, a charge of about +0.7
for the metal and about -0.7 for the phosphorus atoms in
each of these compounds was deduced, providing further
1
+ 1-
support for the approximate formulation M P described
earlier. The electron populations of individual states also
correlate with some of the general trends seen in the crystal
structures. The population of the metal t2g states, which are
involved in metal-metal bonding, increases through the
-
-
15
series CrP (4.2 e ) to CoP (5.2 e ), consistent with the
suggested previously. The metal 3d states, split into t2g and
components, have similar energies as those in the valence-
e
g
(
45) Watts, J. F.; Wolstenholme, J. An Introduction to Surface Analysis by
band spectra of the pure metals. This verifies our earlier
conclusion derived from analysis of the satellite structure in
the metal 2p3/2 core spectra of Co metal and CoP which
showed that during the plasmon loss process in which the
XPS and AES; Wiley: Rexdale, 2003.
(46) Brillson, L. J.; Ceasar, G. P. Surf. Sci. 1976, 58, 457-468.
(
47) Tougaard, S. QUASES-IMFP-TPP2M: Database for Calculation of
IMFPs by TPP2M Formula, version 2.1; QUASES-Tougaard, Inc.:
Odense, Denmark, 2000 (www.quases.com).
8996 Inorganic Chemistry, Vol. 44, No. 24, 2005

Bonding in Binary Transition-Metal Monophosphides
Table 4. Corrections Applied to Intensities of Peaks in Valence-Band
Spectra of MP (M ) Cr, Mn, Fe, Co)a
photoionization
IMFP,
atom
Cr
state
3d3/2
cross-section, σ
λ (Å)
0.0264
0.0387
0.0424
0.0622
0.0694
0.1017
0.1082
0.1582
0.1116
0.0354
0.0708
26.2
25.9
25.9
24.0
33.9
3
d5/2
3d3/2
d5/2
3d3/2
d5/2
3d3/2
d5/2
Mn
Fe
Co
P
3
3
3
3s
3
3
p1/2
p3/2
a
The P 3p1/2 and 3p3/2 cross-sections were corrected on the basis of
comparison of experimental cross-sections determined by Brillson and
4
6
43
Ceasar to those calculated by Scofield. All other cross-sections listed
4
3
are those calculated by Scofield. These cross-sections were determined
for an excitation energy equal to that of Al KR X-rays. The IMFP values
were determined using the QUASES IMFP calculator.
Table 5. Electron Populations in MP (M ) Cr, Mn, Fe, Co)
4
3
4
7
state
CrP
1.3
3.3
4.2
1.1
1.1
5.7
5.3
11.0
MnP
FeP
CoP
P 3s nonbonding
P 3p bonding
M 3d t2g
M 3d eg antibonding
P 3p nonbonding
total P
1.4
3.1
5.0
1.4
1.1
5.7
6.3
12.0
1.6
1.9
5.2
2.3
2.1
5.6
7.4
13.0
1.4
2.4
5.2
3.0
2.0
5.8
8.2
14.0
a
total M
-
total number of e
a
.1 e- in P-P bonding states (peaks 2a and 3a in Figure 9a) and 2.2
1
e- in Cr-P bonding states (peaks 2 and 3 in Figure 9a).
shortening of the average metal-metal bond lengths (2.782-
.678 Å),9 if contacts less than 3.0 Å are considered
2
significant. The population of the phosphorus 3p bonding
states, which are involved in both metal-P and possibly P-P
-
bonding, decreases on going from CrP (3.3 e ) to CoP (2.4
-
e ), consistent with the lengthening of P-P bond lengths
9
(
2.644-2.701 Å). Given that these are weak interactions at
Figure 11. Energy-level diagrams for (a) CrP, (b) MnP, (c) FeP, and (d)
CoP.
best, any manifestation of P-P bonding states should be most
prominent in CrP. Indeed, the valence-band spectrum of CrP
some of these compounds also complicates the interpretation
of the electronic structure.
(Figure 9a) reveals extra shoulders that can be fit by invoking
a second set of higher-binding-energy P 3p states (peaks 2a
and 3a in Figure 9a) which result from P-P bonding, in
addition to the first set of P 3p states (peaks 2 and 3 in Figure
From the above analysis of the electron populations
deduced from the valence-band spectra, experimental energy-
level diagrams for these four monophosphides can be
constructed. Figure 11 presents plots of the electron popula-
tion versus energy. These diagrams are consistent with
theoretical band structure calculations. Band dispersion and
electron populations of different metal- and phosphorus-based
states agree well with partial density of states curves and
the bonding, nonbonding, or antibonding character of states
agrees well with crystal orbital overlap population curves.13
9a) at lower binding energy which are due to Cr-P bonding.
Table 5 shows that the electron population in the P-P
bonding states is only half as large as in the Cr-P bonding
states. On progressing to the other transition-metal mono-
phosphides of the series, a trend toward populating the
nonbonding P 3p states, at the expense of the bonding P 3p
states, is observed which is consistent with the gradual
disappearance of the already weak P-P bonds. In the case
of FeP and CoP, the values of the electron populations of
the phosphorus-based states suggest that there is negligible,
if any, P-P bonding. Size effects cannot be ruled out as
being responsible for these trends in both metal-metal and
phosphorus-phosphorus bonding behavior. Progressing along
the first-row transition-metal series, we see a contraction of
the structure, which may serve as the true driving force for
the observed trends. The occurrence of ferromagnetism in
4
. Conclusion
Core and valence-band X-ray photoelectron spectra have
been obtained for the series of isostructural MnP-type
transition-metal monophosphides CrP, MnP, FeP, and CoP.
The high-resolution phosphorus 2p spectra indicated that the
charge of the P atoms is close to -1, as determined from
their binding energies. The increase in the phosphorus 2p
Inorganic Chemistry, Vol. 44, No. 24, 2005 8997

Grosvenor et al.
binding energies through the sequence CrP to CoP confirms
the decrease in ionicity of the metal-phosphorus bond as the
difference in electronegativity between the atoms diminishes.
The high-resolution metal 2p3/2 spectra of the monophos-
phides closely resemble those of the parent transition metals
in terms of both binding energy and line shape. The metal
atoms in the monophosphides possess a delocalized electronic
structure arising from the presence of extended metal-metal
bonding, which provides greater nuclear shielding than for
isolated atoms. The satellite structure observed in the Co
involves highly delocalized states, similar in character to
those of the pure metal, whereas the phosphorus-phosphorus
bonding network, which is already weak for CrP and MnP
and probably negligible for FeP and CoP, involves localized
covalent bonding. Energy-level diagrams were constructed
from the valence-band spectra and extracted electron popula-
tions, which compare well to calculated band structures. This
type of bonding analysis derived from analysis of experi-
mental photoelectron spectra is complementary to theoretical
calculations and may be generalized to study the bonding
interactions involved in other inorganic solids.
2p3/2 spectra of both Co metal and CoP was confirmed by
REELS analysis of Co metal. This feature arises from
plasmon loss and not as the result of a “two core hole” final
state effect. The common satellite structure for both Co metal
and CoP implies that the valence electrons of the metal atoms
have similar energies, as was also confirmed by analysis of
the valence-band spectra. However, there are fewer valence
electrons around the metal atoms in CoP than in Co metal,
implying a cationic state which is consistent with the anionic
state of the phosphorus atoms.
The valence-band spectra agree well with calculated band
structures. We demonstrate a useful approach for analyzing
the valence-band spectra by fitting these bands with com-
ponent peaks representing electronic states and their popula-
tions. The major orbital interactions arise between the metal
Acknowledgment. The Natural Sciences and Engineering
Research Council (NSERC) of Canada supported this work
through Discovery Grants to R.G.C. and A.M.. We thank
Dr. Murray Gray (Director) for providing access to the Kratos
XPS and Dr. Dimitre Karpuzov (Facility Manager) for
assistance with instrumentation at the Alberta Centre for
Surface Engineering and Science (ACSES) at the University
of Alberta. ACSES is operated through capital funding from
the Canada Foundation for Innovation (CFI), which provides
interim operating support, and Alberta Innovation and
Science. We thank Dr. S. Ramamurthy (Surface Science
Western, University of Western Ontario) for performing the
REELS analysis of Co metal, as well as Ms. Christina Barker
(Department of Chemical and Materials Engineering, Uni-
versity of Alberta) for assistance with the SEM-EDX
analysis. A.P.G. thanks the University of Alberta for support.
3d and phosphorus 3p states. The electron count for the metal
and phosphorus atoms, extracted from the corrected intensi-
1+ 1-
ties of these component peaks, favors the formulation M P
in which both metal-metal and phosphorus-phosphorus
bonding are present. The metal-metal bonding network
IC051004D
8998 Inorganic Chemistry, Vol. 44, No. 24, 2005