Photoelectron spectroscopy of size-selected

transition metal clusters: Fe−n, n=3–24

Cite as: J. Chem. Phys. 102, 9480 (1995); https://doi.org/10.1063/1.468817
Submitted: 08 November 1994 • Accepted: 21 March 1995 • Published Online: 04 June 1998

Lai-Sheng Wang, Han-Song Cheng and Jiawen Fan

ARTICLES YOU MAY BE INTERESTED IN

The design and construction of a high-resolution velocity-map imaging apparatus for

photoelectron spectroscopy studies of size-selected clusters
Review of Scientific Instruments 85, 083106 (2014); https://doi.org/10.1063/1.4891701

Invited Review Article: Laser vaporization cluster sources

Review of Scientific Instruments 83, 041101 (2012); https://doi.org/10.1063/1.3697599

Time-of-Flight Mass Spectrometer with Improved Resolution

Review of Scientific Instruments 26, 1150 (1955); https://doi.org/10.1063/1.1715212

J. Chem. Phys. 102, 9480 (1995); https://doi.org/10.1063/1.468817 102, 9480

© 1995 American Institute of Physics.

Photoelectron spectroscopy of size-selected transition metal clusters:
Fe2
n , n 53–24
Lai-Sheng Wanga)
Department of Physics, Washington State University, Richland, Washington 99352 and Environmental
Molecular Sciences Laboratory, Pacific Northwest Laboratory, Richland, Washington 99352
Han-Song Cheng
Air Products and Chemicals, Inc., Allentown, Pennsylvania 18195
Jiawen Fan
Environmental Molecular Sciences Laboratory, Pacific Northwest Laboratory, Richland, Washington 99352
~Received 8 November 1994; accepted 21 March 1995!
A higher resolution magnetic bottle photoelectron spectrometer for the study of the electronic
structure of size-selected metal clusters is presented. The initial study on Fe2 n ~n53–24! is reported
at a photon energy of 3.49 eV. The photoelectron spectra of these clusters exhibit sharp features
throughout the size range. The spectra for Fe2 3– 8 show large size dependence with many resolved
features. The spectra for Fe2 9–15 exhibit some similarity with each other, all with a rather sharp
feature near the threshold. An abrupt spectral change occurs at Fe2 2 2
16 , then again at Fe19 and Fe23 .
These photoelectron spectral changes coincide remarkably with changes of the cluster reactivity
with H2 . Extended Hückel molecular orbital ~EHMO! calculations are performed for all the clusters
to aid the spectral interpretations. The calculations yield surprisingly good agreement with the
experiment for clusters beyond Fe9 when body-centered cubic ~bcc! structures are assumed for
Fe9–15 and a similarly close-packed structure with a bcc Fe15 core for the larger clusters. The EHMO
calculations allow a systematic interpretation of the sharp photoelectron spectral features in Fe2 9–15
and reproduced the abrupt spectral change taking place from Fe2 2
15 to Fe16 . Most importantly, the
reactivity changes of the clusters with H2 are successfully explained based on the detailed electronic
structures of the clusters, as revealed from the photoelectron spectroscopy ~PES! spectra and the
theoretical calculations. The calculations also correctly predict the existence of magnetism in these
clusters and yield reasonable values for the cluster magnetic moments. © 1995 American Institute
of Physics.

I. INTRODUCTION chemical reactivity of iron clusters reveals a strong size

effect.6 –14 In particular, the reactivity of small iron clusters
One of the fundamental issues in the study of metal clus- with H2 exhibits striking size dependency with orders of
ters is to understand the electronic structures of these inter- magnitude variations in reaction rate between sizes 15 and
mediate systems and their evolution toward bulk band struc- 16, 18 and 19, and, to a less extent, between 22 and 23.8,9
ture as the cluster size increases.1 Of all the metal clusters, Because of the unique ferromagnetic properties of bulk iron,
transition metal clusters are of particular interest due to their studies on the magnetic properties of the small iron clusters
diverse physical and chemical behavior. Unusual physical have drawn considerable interests. All iron clusters are found
and chemical properties are expected for these clusters. to be magnetic with higher moments than those found in the
Some of these clusters are found to be magnetic with signifi- bulk.2–5 In other experimental aspects, ionization potentials
cantly higher magnetic moments than in bulk materials.2–5 ~IPs! ~Refs. 24 –26! and dissociation energies of the positive
Their chemical reactivity is also found to be strongly clusters27 have been measured.
size-dependent.6 –14 Primarily due to the presence of partially Several possible interpretations have been proposed to
filled d orbitals, however, the transition metal clusters have account for the dramatic size dependency of the chemical
presented considerable challenges theoretically. Even the reactivity of the iron clusters with H2 . Whetten et al. first
dimer systems are found to have very complicated electronic made the interesting observation between the cluster reactiv-
structures and the nature of their bonding properties has not ity and the cluster IPs.9 Their proposed model for the reac-
been completely understood.15–21 Heretofore, very little is tivity involved the cluster-to-hydrogen charge transfer. Sub-
known about the detailed structures of transition metal clus- sequently, Siegbahn et al. focused on the detailed electronic
ters and their electronic properties. Two early review articles structure of the clusters on the reactivity and showed that the
have summarized the experimental and theoretical aspects of correlation between the IP and reactivity is only a secondary
transition metal clusters.22,23 consequence of the electronic structure of the clusters.28
One of the most extensively studied transition metal Moreover, Richtsmeier et al. stressed the geometrical struc-
cluster systems is the iron cluster systems. Research on the ture effect on the cluster reactivity.8 Recently, Conceicao
et al. found that there is a better correlation between the
a!
To whom correspondence should be addressed. cluster reactivity and the difference of the cluster IP and

9480 J. Chem. Phys. 102 (24), 22 June 1995 0021-9606/95/102(24)/9480/14/$6.00 © 1995 American Institute of Physics
 Wang, Cheng, and Fan: Size-selected metal clusters 9481

EA.29 They proposed that the reactivity is determined by an

entrance channel barrier, with a magnitude proportional to
the ~IP2EA! difference.
Theoretically, several studies have attempted to under-
stand the electronic structure of the iron clusters.30–36 How-
ever, the complicated nature of these systems only allows a
few selected high symmetry and very small systems to be
studied. It is worth mentioning the IP calculations by Pastor
et al. with a tight-binding method.33 They systematically
treated Fen from n53 to 25 and found good agreement with
measured IPs by assuming a body-centered cubic ~bcc!-type
structure for n>9. Ab initio calculations have only been per-
formed for the iron dimer.19,20,34 –36
Because of the high density of electronic states found in
transition metal clusters, conventional laser spectroscopic FIG. 1. Schematic view of the laser-vaporization/magnetic-bottle photoelec-
techniques cannot be applied to these clusters except to tron spectroscopy apparatus.
dimers and, perhaps, a few trimers. Photoelectron spectros-
copy ~PES!, involving single-photon, nonresonant transitions
from the ground state of a size-selected anion cluster to the cluster chemical reactivity with H2 and the cluster magne-
tism. It will be shown how the electronic structure informa-
various states of the neutral cluster, is ideal to probe the
tion obtained is used to give a microscopic understanding of
electronic structure of clusters.37– 43 We have completed a
the size dependence of the cluster reactivity. Finally, Sec. VI
modified magnetic-bottle time-of-flight ~MTOF! PES appa-
summarizes the conclusions derived from the present work.
ratus with high mass resolution and improved electron en-
ergy resolution.38,44 It is aimed to obtain high resolution PES
II. EXPERIMENT
spectra for a wide range of cluster systems.45– 48 The combi-
nation of a laser vaporization cluster source and the high Figure 1 shows a schematic view of our MTOF-PES
efficiency and higher resolution capabilities of this apparatus apparatus. It consists of a laser vaporization cluster source, a
will allow many novel clusters to be studied in considerable modified Wiley–McLaren time-of-flight ~TOF! mass spec-
detail. trometer, a mass gate, a momentum decelerator, and a MTOF
In this paper we report the details of the MTOF-PES electron analyzer. In the following paragraphs we describe
apparatus and the first PES study on Fe2 n clusters for n53– each part in detail.
24. Well-resolved PES spectra at 3.49 eV photon energy for A. Cluster source
these clusters are presented. In contrast to the expectation
that the high density of states for transition metal clusters A sufficiently cold and intense negative cluster beam is
would yield broad spectra, we have observed consistent essential for high resolution PES experiments. Due to the
sharp spectral features near the threshold through the whole electron attachment energy, negative cluster ions can carry
size range studied. More interestingly, the observed PES substantial internal energies and are more difficult to cool.
spectral variations correspond well with the cluster chemical We use a large ‘‘waiting room’’ nozzle and an intense super-
reactivity changes with H2 , suggesting the importance of the sonic expansion to maximize cluster cooling. A drawing of
electronic structure to the cluster reactivity. To aid the spec- the details of the nozzle is shown in Fig. 2. The vaporization
tral interpretations and gain insight into the details of the laser beam is directed collinearly with the cluster beam and
cluster electronic structure, we carry out extended Hückel
molecular orbital ~EHMO! calculations for all the clusters.
Surprisingly, the EHMO results yield qualitative agreement
with the experiment. The calculations also provide important
insights into the chemical reactivity of the clusters, as well as
the cluster magnetism.
The organization of this paper is as follows. In Sec. II,
we report the details of our MTOF-PES apparatus, which
currently can achieve a mass resolution (M /DM ) of more
than 500 and an electron energy resolution of better than 20
meV at 1 eV electron kinetic energy. We use EHMO theory
to calculate the electronic states for a variety of chosen clus-
ter structures. The details of the EHMO calculations are de-
scribed in Sec. III. The PES spectra of the Fe2 n cluster for
n53–24 are then presented in Sec. IV. The calculated results
are used to help interpret the experimental observations. In
FIG. 2. Schematic cross-sectional view of the large ‘‘waiting room’’ laser
Sec. V we give further in-depth discussions on the iron clus- vaporization nozzle, showing the double Jordan valves and the sample disk
ter electronic structure. We pay particular attention to the assembly.

J. Chem. Phys., Vol. 102, No. 24, 22 June 1995

9482 Wang, Cheng, and Fan: Size-selected metal clusters

transistor switch ~EUROTEK, HTS-30GM! is used to deliver

sharp and variable width pulses for the mass gate.
Following the mass gate, the selected cluster beam en-
ters a momentum decelerator. Once the cluster packet passes
the third grid of the mass gate, a positive square high voltage
pulse ~EUROTEK, HTS-30GM! is applied to this grid for
the momentum deceleration. The high voltage is pulsed to
ground before the ion packet leaves the deceleration stack,
which consists of ten guarded rings to ensure a uniform de-
celeration field. Both the pulse amplitude and the pulse width
can be varied to achieve the best deceleration effect. All ions
experience the same decelerating force within the same time
period, thus will be decelerated by the same amount of linear
momentum. Markovich et al. first used this deceleration
FIG. 3. The mass gate and the momentum decelerator. strategy in their MTOF-PES apparatus and have shown that
it works quite effectively.51
The large volume ion extraction imparts a substantial
hits the target disk after passing the 2 mm nozzle orifice. The amount of initial energy spread on the cluster beam, making
laser beam, typically 20 mJ from the second harmonic of a the usual retarding field deceleration very ineffective.38 The
Nd:YAG laser, is focused down to a 1 mm diam spot onto the momentum decelerator not only allows more deceleration,
target. Two pulsed molecular beam valves ~R. M. Jordan Co., but also yields less ion spread. In contrast to the retarding
CA!, symmetrically mounted, as shown in Fig. 2, are used to field deceleration where the ion energy distribution does not
deliver a short and intense helium carrier gas pulse ~at 10 change after the deceleration, the final ion energy distribu-
atm backing pressure!, which mixes with and cools the laser- tion after the momentum deceleration is reduced approxi-
induced plasma to produce clusters, including both neutral mately as the following for a mass selected cluster packet:
and charged species. The cluster/He mixture undergoes a su-
personic expansion and is skimmed twice to form a well- DE f 5 ~ E f /E i ! 1/2 DE i , ~1!
collimated cluster beam into the ion extraction chamber. A where E i and DE i are the initial ion mean kinetic energy and
liquid-nitrogen-cooled nozzle can be used to supplement the the initial ion energy distribution, and E f and DE f are the
supersonic cooling of the clusters. This is not used in the final ion mean energy and energy distribution after the mo-
current experiment. Different nozzle geometries can be used mentum deceleration, respectively. As will be shown below,
to alter the clustering conditions and cluster cooling. The this deceleration step is crucial to improve the MTOF elec-
large waiting room nozzle favors cluster formation and is tron energy resolution due to the minimization of the Dop-
known to produce colder negative clusters.49 pler broadening.
Figure 4 illustrates the working of the mass gate and the
B. Mass selection and momentum deceleration momentum deceleration. Figure 4~a! shows the mass spec-
trum of Cu1 from laser vaporization of a Cu target. The two
The negative clusters are extracted perpendicularly from
isotopes of Cu are well-resolved. Each peak has a width
the beam by a 1 kV high voltage pulse ~EUROTEK, HTS50-
~FWHM! of about 35 ns, yielding a mass resolution of about
21! and are subjected to a TOF mass analysis. The mass
480. Figure 4~b! shows the mass selected 65Cu1 isotope
spectrometer used is a modified Wiley–McLaren-type, which
peak, while Fig. 4~c! shows the decelerated 65Cu1 peak by a
was first described by de Heer and Milani for large volume
2 kV, 550 ns wide pulse. The ions after the deceleration have
ion extraction and simultaneous high mass resolution.50 The
a mean energy of about 30 eV in this case, compared to 1000
major modification is an addition of a short free-flight zone
eV initial beam energy with a roughly 10% spread. There is
in between the two acceleration stages of the original Wiley–
still moderate beam spread after the deceleration as shown in
McLaren design. As shown by de Heer and Milani, this
Fig. 4~c!. However, this represents the spread at the detector
modification can achieve high mass resolution with a large
that is much further down stream from the decelerator. At the
extraction zone appropriate for extracting metal clusters from
interaction zone, only 3 cm away from the decelerator, the
beams. We have obtained a mass resolution (M /DM ) of
beam spread should be much better. Furthermore, the beam
more than 500 at low masses. It deteriorates at higher
spread will be much less severe for heavier clusters due to
masses, mainly limited by the fringe field effect due to the
the smaller velocity distribution.
ion steering to compensate for the transverse velocity of the
clusters. Implementation of a dynamic ion steering system50
is expected to increase the mass resolution to over 1000. C. Magnetic-bottle time-of-flight photoelectron
spectrometer
A three-grid mass gate is used for mass selection as
shown in Fig. 3. The first and third grids are at ground, and The magnetic-bottle type of TOF photoelectron spec-
the middle grid is at negative high voltage ~21.1 kV! so that trometer first described by Kruit and Read44 is ideal for the
no clusters are able to pass. Once the desired cluster arrives study of clusters due to its high collecting efficiency ~2p
at the first grid, the high voltage is pulsed to ground for a solid angle!, which compensates for the low cluster number
short period allowing the cluster to pass unaffected. A fast density. This design has since been implemented for negative

J. Chem. Phys., Vol. 102, No. 24, 22 June 1995

 Wang, Cheng, and Fan: Size-selected metal clusters 9483

FIG. 4. ~a! Mass spectrum of Cu1 from laser vaporization of a Cu disk,

showing the two well resolved isotopes of Cu ~63Cu, 69%, and 65Cu, 31%!.
The mass resolution here is M /DM 5480. ~b! Mass-selected 65Cu1 isotope
peak. ~c! Decelerated 65Cu1 isotope peak with a high voltage pulse ~2 kV
pulse height, 550 ns pulse width!.

cluster studies and configured to collect all 4p solid

angles.38,39 Our apparatus is a modified 4p version of the
MTOF, using a pulsed electromagnet for the high field
~;500 G!. This externally mounted magnet allows ultrahigh FIG. 5. Photoelectron spectra of Cu2 at 3.49 eV photon energy showing the
vacuum conditions ~10210 Torr! to be obtained in the elec- effect of ion deceleration on the photoelectron energy resolution. Decelera-
tron detachment zone.38 tion pulse ~a! 2 kV, 200 ns; ~b! 2 kV, 400 ns; ~c! 2 kV, 600 ns.
As shown in Fig. 1, the MTOF is located at the end of
the TOF mass spectrometer. The detachment zone is 3 cm
downstream from the momentum decelerator. A set of in-line
microchannel plates, about 12 cm down from the detachment the transitions from the ground state of Cu2 ( 1 S 0 ,3d 10 4s 2 )
zone, is used as the cluster mass detector. The MTOF tube is to the ground state of Cu ( 2 S 1/2 ,3d 10 4s 1 ) and the first two
3.5 m long, at the end of which is located a fast Z-stack excited states of Cu ~2D 5/2 and 2D 3/2, 3d 9 4s 2 !.52 The bottom
electron detector. Not shown in the figure are the high field panel shows a deceleration of the Cu2 with a 2 kV and 600
external electromagnet, and the low field solenoid and the ns pulse. The Cu2 ions have a mean kinetic energy of less
double m-metal shielding along the 3.5 m MTOF tube. than 10 eV. The 2S peak has a FWHM of 50 meV, and the
As mentioned above, the Doppler broadening is a seri- two excited state peaks have a FWHM of about 30 meV.
ous resolution limiting factor for the MTOF due to the initial Further deceleration of the Cu2 does not improve the reso-
ion kinetic energy. Since photoelectrons are collected from lution any more, suggesting that this resolution is no longer
all 4p solid angles, the electrons acquire a maximum veloc- limited by the Doppler broadening. We have demonstrated
ity spread of 2 v in the laboratory frame as a result of the that by simply replacing the high field electromagnet with a
finite parent ion velocity, v . The resultant electron energy permanent magnet ~yielding ;700 G at the interaction re-
spread can be expressed as gion! about 20 meV resolution can be achieved at 1 eV elec-
tron kinetic energy.53 Even better resolution is expected at
DE e 54 @~ m e /m !~ E e E !# 1/2 , ~2!
lower electron kinetic energies ~Ek! since the resolution
where E e and E are the kinetic energies of the photoelectrons (DE) varies as ~Ek!3/2. Cha et al. also achieved impressive
and the parent negative cluster ions, respectively, while m e resolution in a similar MTOF PES apparatus by using essen-
and m are the corresponding masses. Figure 5 shows the tially a retarding field deceleration method, which works
effect of the cluster deceleration on the electron energy reso- well in their case primarily due to the very low extraction
lution of Cu2 at 3.49 eV photon energy. The spectra show voltage used.43

J. Chem. Phys., Vol. 102, No. 24, 22 June 1995

9484 Wang, Cheng, and Fan: Size-selected metal clusters

Oxide contamination is known to be a serious problem structure of the dimer, the only iron cluster that has been well
in laser vaporization generation of metal clusters. There are studied, is still not been fully understood. Leopold and
three major oxygen sources; ~1! oxide layer on the target Lineberger obtained a beautiful vibrationally resolved PES
surface; ~2! trace amount of oxygen impurity in the carrier spectrum of Fe2 17
2 at 2.540 eV photon energy. In contrast to
gas; and ~3! adsorbed water and oxygen on the interior walls previous calculations, which predicted Fe2 would have 112
of the gas lines and the pulsed molecular beam valves. To states within 0.5 eV above its ground state,16 they only ob-
minimize the problem, we use ultrahigh purity He served two states. This was interpreted to be due to a rather
~99.9999%! as the carrier gas with an inline liquid nitrogen strong d – d interaction.18 The ground state configuration of
trap. Before each experiment, the gas lines and the pulsed the anion was assigned to be 3d 13 s 24s s 4s
* 2 and the two PES
valves are fully baked with He flushing to desorb any water features were assigned to be from detachment of a s 4s * elec-
and oxygen. The target surface oxide layer can be substan- tron either high- or low-spin coupled with the unpaired 3d
tially reduced after the target is laser ablated several times. electrons—leading to a 3d 13 s 24s s 4s
* 1 ground state configura-
With these measures, the Fen O2 signals are reduced to less tion for Fe2 . Two ab initio calculations have followed this
than 5% of the pure clusters and they are diminished after work, and the most recent calculation yields a 3d 14 s 24s
continuous use of the target. For n up to 8 we could separate ground state configuration for Fe2 , in disagreement with the
the pure clusters from the oxide clusters completely. How- experimental assignment.19,20 Not surprisingly, there has
ever, due to the 5.8% natural abundance of 54Fe, the Fen O2 been no ab initio calculation on the iron clusters beyond the
peak would overlap with the pure clusters for n larger than 8. dimer.
We estimate the Fen O2 contamination is no more than 1% in Therefore, it is a formidable task to interpret the PES
the worst case. Using an O2/He mixed carrier gas, we have spectra of the iron clusters. As a first step to systematically
studied Fen O2 clusters for n up to 16, which are interesting understand the PES spectra and gain better insight into their
in their own right.48 electronic structure, we have carried out EHMO calculations
The PES spectrum of Cu2 is used for the spectrometer on all the clusters presented here. The geometrical structures
calibration. All the PES spectra of the Fe2n clusters are taken of the clusters are systematically selected except for Fe6 – 8.
at comparable resolution as that of the best Cu2 spectrum, as The EHMO theory has been used previously to study the
shown in Fig. 5~c!. The iron clusters are heavier than the Cu structures and bonding of several metal clusters and has been
atom and can be decelerated more easily than the Cu2 anion shown to yield reasonable results for alkali clusters.54 –56 We
since the heavier ions have narrower velocity distributions. attempt to use the EHMO results as a guide in the spectral
The third harmonic of a Nd:YAG laser ~3.49 eV! is used for interpretation and compare them with other results where
the detachment. Each spectrum is averaged with between available. Despite its simplicity, the EHMO results give
30 000 and 60 000 laser shots at 10 Hz repetition rate. The qualitative agreement with the current experiment, leading to
electron kinetic energy spectra are subtracted from the pho- important insight into the electronic structure of these clus-
ton energy to obtain the electron binding energy spectra pre- ters. Emphasizing the qualitative nature of the EHMO re-
sented. sults, we look for insight that can be obtained from these
calculations about the complicated electronic structure of the
Fe clusters.
III. INTERPRETATION OF PHOTOELECTRON
SPECTRUM AND EXTENDED HÜCKEL MOLECULAR
B. Extended Hückel molecular orbital calculations
ORBITAL (EHMO) CALCULATIONS
The extended Hückel theory is a semiempirical quantum
A. Interpretation of photoelectron spectrum
mechanical method57 that has considerable success in study-
Photoelectron spectroscopy probes the electronic struc- ing chemisorption phenomena on surfaces and a variety of
ture of matter. Photodetachment transitions occur between solid state chemistry and physics problems by Hoffmann and
the ground state of an anion to the ground and various ex- co-workers.58 – 60 It provides a simple and heuristic view on
cited states of the corresponding neutral. PES spectra are many physical and chemical processes in the condensed
usually interpreted based on the single electron approxima- phase, despite its admitted lack of numerical accuracy.
tion, where single electron removals from occupied MOs The EHMO computational procedure can be briefly de-
yield the spectral features. Although this approximation is scribed as follows. One starts with the Fock equation
quite crude and ignores relaxation effects, multiplet split-
FC5ESC, ~3!
tings, and electron correlations, it still provides a useful
means to interpret the main PES spectral features and can where C denotes the molecular orbital coefficient matrix and
yield much insight into the molecular bonding and electronic E is the energy matrix. The overlap matrix S can be conve-
structures. Due to their high numbers of unpaired d elec- niently calculated from the basis set. In the EHMO scheme,
trons, iron clusters have been very difficult to treat theoreti- the Fock matrix F is defined as the following: its diagonal
cally. Several theoretical studies have been performed on the elements are completely determined by the valence state ion-
electronic structure of a few selected Fe clusters.30–36 How- ization potentials of the atomic orbitals; the off-diagonal el-
ever, a detailed interpretation of the current PES spectra ements are defined by the diagonal matrix elements as well
based on these calculations is not yet possible. as the overlap matrix elements and interaction matrix ele-
To illustrate the difficulties in treating these systems, it ments that are purely empirical parameters. The Fock matrix
might be instructive to point out that even the electronic possesses correct symmetry properties arrived from the given

J. Chem. Phys., Vol. 102, No. 24, 22 June 1995

 Wang, Cheng, and Fan: Size-selected metal clusters 9485

molecular structures, and the built-in atomic ionization po- TABLE I. The numbers of unpaired electrons in the iron clusters obtained
tentials are directly obtained from experimental measure- from the EHMO calculations. The average magnetic moment ~mB ! is given
in the parentheses.
ments. Therefore, EHMO theory yields correct symmetry in-
formation on the molecular orbitals, and the calculated Fe3 8~2.67! Fe14 30~2.14!
energy levels usually correlate reasonably well with the mo- Fe4 10~2.50! Fe15 32~2.13!
lecular ionization potentials. The EHMO scheme normally Fe5 12~2.40! Fe16 36~2.25!
uses a noniterative procedure. One can solve the Fock equa- Fe6 14~2.33! Fe17 38~2.23!
Fe7 16~2.28! Fe18 42~2.33!
tion by direct matrix diagonalization to obtain the molecular Fe8 18~2.25! Fe19 46~2.42!
orbitals and the energy levels. Thus, computationally, it is Fe9 20~2.22! Fe20 48~2.40!
highly efficient. Fe10 22~2.20! Fe21 50~2.38!
Unfortunately, in the EHMO scheme, there is no repul- Fe11 24~2.18! Fe22 52~2.36!
sive term built into the Fock matrix. As a consequence, the Fe12 26~2.17! Fe23 54~2.35!
Fe13 28~2.15! Fe24 56~2.33!
original EHMO theory cannot be used to optimize molecular
geometry. Recently, several studies have shown that when
the repulsive forces are included, the EHMO theory can yield
quite reasonable optimized geometries for a large number of cluster with numerous unpaired electrons; thus, the magne-
molecules, as well as several cluster systems, while still tism in these clusters is ‘‘naturally’’ revealed. The same ef-
maintaining its high computational efficiency.61 fect ~i.e., the exchange interaction! produces magnetism in
The geometrical structures of the small iron clusters are bulk metals.65
not known, and they are difficult to obtain both experimen- The numbers of unpaired electrons from the EHMO cal-
tally and theoretically. Riley and co-workers have done ex- culations are listed in Table I. The structures of these clusters
tensive work using chemical means to probe the structures of that are used are shown in Figs. 6 –10. The occupied energy
the iron clusters.12,14,62,63 One noted conclusion is that a bcc levels, according to their degeneracy and electron occupancy,
closed shell structure is a very likely candidate for Fe15 .12 are plotted as stick spectra in Figs. 6 –10 by aligning the
Polyicosahedral types of structures have been proposed for HOMO with the threshold features in the experimental PES
Fe19 and Fe23 when saturated with NH3 .62 For Fe3–5 and spectra.66
Fe10–15, several calculations have been performed, and their
geometries optimized.35,36,64 Interestingly, calculations of the
IPs by assuming bcc-type structures for n>9 yield reason- IV. RESULTS AND SPECTRAL INTERPRETATION
able agreement with the measured IPs.33 The PES spectra of Fe2 n ~n53–24! are shown in Figs.
In the present calculations, we choose the cluster geo- 6 –10 with the EHMO occupied energy levels plotted as the
metrical structures based on this limited previous knowledge. stick spectra. The cluster structures used in the EHMO cal-
In particular, we choose the bcc structure for Fe9–15, encour- culations are also shown in the figures. Our results for Fe2 9–16
aged by the work of the IP calculations33 and the experimen- have been reported briefly before.67 The weak features at the
tal indication about the bcc structure for Fe15 .12 Beyond low binding energy side, common to all the spectra, are due
Fe15 , we use a close-packed structure based on the bcc Fe15 to hot band transitions or minor isomers with low electron
core. For Fe3–5, we use the most recent results from density affinities ~EAs! since their intensity varies with the cluster
functional calculations.36 In some cases, several structures source conditions while the main features remain the same.
are tested. Well-resolved spectral features are shown for the small clus-
In the EHMO scheme, the 3d, 4s, and 4p of Fe are ters, and they are quite different from each other, indicating
taken as valence orbitals, of which the 4s and 4p are repre- the molecular nature of the small clusters. The spectra of the
sented with a set of single zeta basis functions and the 3d larger clusters get increasingly diffuse and they seem to show
with double zeta basis functions. For each cluster, we obtain certain systematic changes. One of the most striking features
a set of electronic energy levels and the corresponding mo- in the spectra is the sharp peak near the threshold for Fe2 9–15,
lecular orbitals at a given geometry. A central issue is the in particular, Fe213 to Fe 2
15 , and the abrupt spectral changes
assignment of the electron occupancy in these levels. While occurring from Fe2 2 2
15 to Fe16 . The spectra from Fe16 to Fe18
2
2
the lower levels are doubly occupied, the occupancy in the are similar—the small feature appearing in the Fe16 spectrum
upper levels may be singlet, depending on the specific energy seems to increasingly gain intensity in Fe2 2
17 and Fe18 . In
2
level structures. For the iron clusters, the highest occupied Fe19 , this small feature becomes a well-resolved peak, and at
molecular orbital ~HOMO! and the lowest unoccupied mo- the same time the whole spectrum shifts to lower binding
lecular orbital ~LUMO! would be nearly degenerate if all energy. This pattern seems to repeat from Fe2 2
20 to Fe23 , where
electrons were paired in the occupied orbitals. Therefore, ac- a well-resolved threshold peak appears and the spectrum is
cording to the Hund’s rule,65 the electron pairs in the upper again shifted to a lower binding energy. Interestingly, these
levels are broken into singly occupied levels; the excess elec- spectral pattern changes coincide exactly with the chemical
trons are then transferred into the higher energy levels until reactivity changes of the iron clusters.8
an apparent energy gap ~0.5 eV or larger! is encountered. The first spectroscopic information obtained from the
This is natural to do since it lowers the electron-electron spectra is the EA of the clusters and its size evolution. There
repulsion due to the exchange energy.65 In doing so consis- are two definitions of the EA; the vertical EA is the transition
tently, we obtain a set of occupied energy levels for each energy from the ground state of the anion to the neutral

J. Chem. Phys., Vol. 102, No. 24, 22 June 1995

9486 Wang, Cheng, and Fan: Size-selected metal clusters

FIG. 7. Photoelectron spectra of Fe2 2

9 to Fe12 at 3.49 eV photon energy. See
also caption in Fig. 6.

ing procedure to estimate the adiabatic EA. Draw a straight

line along the slope of the threshold feature and find the
binding energy value where this line is crossed with the bind-
ing energy axis. This value plus a constant value of 0.07 eV
~this accounts for the finite thermal excitations and instru-
mental resolution! is taken to be the adiabatic EA. The EAs
obtained in this manner for the iron clusters are listed in
Table II and plotted in Fig. 11. We generously assign an error
bar of 0.08 eV although the EAs can be more accurate when
FIG. 6. Photoelectron spectra of Fe2 2
3 to Fe8 at 3.49 eV photon energy. The
sharp features appear at the threshold.
stick spectra are the occupied energy levels from the extended Hückel mo-
lecular orbital calculations by aligning the highest occupied level~s! with the
experimental threshold feature. The stick height is proportional to the orbital A. Fe3 –Fe8
degeneracy and occupancy. The assumed structures for the corresponding
calculations are also shown.
1. Fe3
The spectrum of Fe2 3 shown in Fig. 6 is more compli-
ground state at the same nuclear configuration; the adiabatic cated than that of Fe2 2 .
17
Beside the two well-separated
EA is the 0–0 transition energy from the anion to the neutral. bands at the lower binding energies, there are more unre-
In the absence of resolved vibrational structures, the adia- solved transitions at the high binding energy side. The most
batic EA cannot be determined accurately. The vertical EA recent local spin density approximation ~LSDA! calculation
can be obtained from the peak maxima of the first PES fea- by Castro and Salahub suggests an equilateral triangular
ture. However, even the latter cannot be determined accu- structure for Fe3 with a bond length of 2.10 Å.36 Using the
rately when there are broad and overlapping features near the same structural parameters, we performed EHMO calcula-
threshold such as seen in Fe2 2
6 and Fe7 . Therefore, to facili- tions that yielded eight unpaired electrons in agreement with
tate comparisons of the EA change as a function of cluster the LSDA calculations. The energy level diagram from the
size and to be systematic and consistent, we use the follow- current calculations is displayed in Fig. 12. Note the large

J. Chem. Phys., Vol. 102, No. 24, 22 June 1995

 Wang, Cheng, and Fan: Size-selected metal clusters 9487

FIG. 8. Photoelectron spectra of Fe2 2

13 to Fe16 at 3.49 eV photon energy. See
also caption in Fig. 6. FIG. 9. Photoelectron spectra of Fe2 2
17 to Fe20 at 3.49 eV photon energy. See
also caption in Fig. 6.

gap between the 1a 82 and 4e 8 orbitals. If all the electrons

were paired, the HOMO-LUMO gap would be between the
3e 8 and 2e 9 orbitals, which are nearly degenerate. energy used and the spectrometer transmission while the first
Those occupied levels within the experimental spectrum band exhibits a splitting as well as other unresolved fine
range are plotted in Fig. 6 as the stick spectrum by aligning features. The most recent calculation by Castro and Salahub
the HOMO level with the threshold peak. The HOMO, yields a tetrahedral structure for the cluster with a bond
1a 28 , is mostly composed of 3d orbitals. The 1a 19 orbital is a length of 2.22 Å and twelve unpaired electrons.36 We use the
strong bonding orbital from the 4s. We assign the first two structural parameters obtained by Castro and Salahub in our
bands in the spectrum to be from these two orbitals, respec- EHMO calculations, which are shown in Fig. 6 as the stick
tively. In particular, the broad width of the second band spectrum, displaying those occupied orbitals in the same en-
agrees with its being from a strong bonding orbital. There are ergy range as the experiment. The EHMO calculation gives
reproducible fine structures on both of these two bands. ten unpaired electrons, compared with twelve obtained by
However, the insufficient resolution plus the seemingly ir-
Castro and Salahub from their LSDA calculation.
regular spacing prevents a vibrational assignment. Since Fe3
Figure 13 shows the EHMO energy level diagram, which
has a degenerate ground state in the equilateral triangular
structure, it is likely to be Jahn–Teller active, which will displays two groups of orbitals near the HOMO; a triply
further complicate the spectrum. It can be seen from Fig. 6 degenerate 2t 1 orbital followed by the densely spaced second
that there is qualitative agreement between the one-electron group ~3t 2 , 2e, 1t 1 , and 2a 1!. We assign the first PES band
EHMO energy levels and the PES spectrum except that the to be from the removal of a 2t 1 electron and the second band
spacings of the calculated levels are too close. to be from the second group of orbitals. Thus, the EHMO
calculation can qualitatively account for the experimental
PES spectrum with a tetrahedral structure. The tetrahedral
2. Fe4 Fe4 cluster has a degenerate ground state, suggesting it is
The spectrum of Fe2 4 shows two broad and well- Jahn–Teller active. This is evident from the broad width and
separated bands. The second band is limited by the photon the splitting of the first band in the PES spectrum.

J. Chem. Phys., Vol. 102, No. 24, 22 June 1995

9488 Wang, Cheng, and Fan: Size-selected metal clusters

FIG. 11. Electron affinities of the iron clusters as a function of cluster size
(n) for n52–24 ~n52 from Ref. 17!.

two other structures; a symmetric bipyramid and a square

pyramid. We found that the symmetric bipyramid gives the
best result, which is shown in Fig. 6 as the stick spectrum.
The calculation does indicate that there are more levels be-
tween the first occupied level and the more densely spaced
levels at higher binding energy. The sharpness of the PES
feature for Fe25 suggests that the Jahn–Teller effect in Fe5
may be rather small if it takes place.

4. Fe6 – 8
For clusters with size larger than Fe5 , there is no calcu-
lation intending to optimize the cluster geometries. There-
fore, the structural parameters are all assumed in our EHMO
FIG. 10. Photoelectron spectra of Fe2 2
21 to Fe24 at 3.49 eV photon energy. See calculations. The nearest neighbor internuclear distance used
also caption in Fig. 6. for all the higher clusters is 2.43 Å, compared to the bulk
lattice constant of 2.87 Å. The increase of the number of
3. Fe5
The spectrum of Fe25 displays a sharp peak at the thresh-
old. The spectral pattern looks similar to that of Fe2
4 except
that there are weak features between the gap of the two
strong bands. The LSDA calculation suggests that a Jahn–
Teller distorted bipyramid is the most stable structure for
Fe5 .36 We performed EHMO calculations with the LSDA
optimized parameters and obtained rather unsatisfactory re-
sults compared to the experiment. Consequently, we tried

TABLE II. Electron affinities of the iron clusters.a

n EA~eV! n EA~eV!

3 1.47~8! 14 2.23~8!
4 1.72~8! 15 2.25~8!
5 1.81~8! 16 2.01~8!
6 1.51~8! 17 2.06~8!
7 1.39~8! 18 2.14~8!
8 1.66~8! 19 2.01~8!
9 1.75~8! 20 2.01~8!
10 1.85~8! 21 2.11~8!
11 2.00~8! 22 2.15~8!
12 2.12~8! 23 2.04~8!
13 2.20~8! 24 2.13~8!
FIG. 12. The extended Hückel molecular orbitals for Fe3 . The arrow indi-
a
The adiabatic EAs. See Sec. IV for the threshold determination. cates the highest occupied level.

J. Chem. Phys., Vol. 102, No. 24, 22 June 1995

 Wang, Cheng, and Fan: Size-selected metal clusters 9489

ment for the larger clusters. However, the structure III for
Fe8 , which is similar to those bcc structures, gives quite poor
agreement with the experiment.
The EHMO calculations do not yield satisfactory inter-
pretations of the PES spectra for these three clusters. This is
not surprising since their structural parameters are not known
at all. The unusually diffuse spectra of Fe2 2
6 and Fe7 suggest
that they may have strong Jahn–Teller effects. The sharp
feature in Fe28 implies that it should have a rather high sym-
metry.

B. Fe9 –Fe15
We briefly reported our results on Fe2 9–16 previously.
67

For completeness, we present them here and focus on those

aspects that could not have been included in the previous
discussion. One of the most striking features in the PES
spectra of this size range is the sharp features exhibited for
all these spectra and their close similarity, as shown in Figs.
7 and 8. Particularly, the sudden spectral change from Fe2 15 to
Fe216 coincides exactly with their chemical property changes.
On the basis of the various previous structural hints, we use
FIG. 13. The extended Hückel molecular orbitals for Fe4 . The arrows indi-
cate the highest occupied levels.
the bcc structure for Fe9–15 in our EHMO calculations.12,33
The EHMO results for these clusters are plotted in Figs.
7 and 8. For the simplicity of these calculations, the agree-
ment is impressive. The sharp threshold features are espe-
atoms in the clusters also means that there are more possible cially well accounted for. All these spectra show a gap fol-
structures. For each of these clusters we performed EHMO lowed by a broad spectral feature. This is well reproduced by
calculations in several geometries. the EHMO calculations. We attribute this good agreement to
The spectrum of Fe2 6 shows rather broad features near the correct structural assumptions.
the threshold, suggesting that there is significant structural Upon inspecting the EHMO wave functions, we find that
change from the anion to the neutral or that there is a strong all the MOs are fairly delocalized. In particular, the HOMOs
Jahn–Teller effect. We carried out EHMO calculations in of these clusters, which are responsible for the sharp features
three geometries; octahedron ~I!, pentagonal pyramid ~II!, near the threshold, are all predominantly composed of 3d
and capped trigonal bipyramid ~III, shown in Fig. 6!. None orbitals. Further more, these HOMOs all have the largest d
of the structures gave satisfactory agreement with the experi- contributions from the central atom in the bcc clusters. These
ment. The stick spectrum shown in Fig. 6 is the EHMO d orbitals from the central atom overlap fairly well with
result of structure III. The densely spaced levels illustrate the orbitals from the other atoms. The orbital delocalization in-
complexity of this cluster. The diffused features of Fe2 6 are dicates that the cluster electronic properties as well as their
also unusual compared to its neighbors. chemical/physical properties might be sensitive to local
The spectrum of Fe2 2
7 , like that of Fe6 , is also lack of structure changes, such as that taking place from Fe15 to
sharper features as seen in the other clusters. However, the Fe16 . It is also significant to observe that the major d contri-
first band in the Fe2 7 spectrum shows unresolved features, butions to the HOMOs are from the bcc central atom. This
indicating the energy levels are more congested. The results plus the structural similarity among these clusters explains
from EHMO calculations, by assuming a pentagonal bipyra- the similarity of the cluster PES spectra in this size range. It
mid structure, are plotted in Fig. 6 as the stick spectrum. It is worthwhile to point out that the striking similarity among
shows rather congested levels, consistent with the experi- the spectra of Fe2 2
13 to Fe15 strongly suggests their structural
ment. Both Fe6 and Fe7 have rather low electron affinities similarity. The bcc structures are the most consistent with
compared with their neighbors ~as is evident in Fig. 11!. this observation.
The spectrum of Fe2 8 displays a rather sharp feature at
the threshold. Previous IP calculations have assumed a cubic
C. Fe16 –Fe24
structure for Fe8 .33 We tested three structures in our EHMO
calculations; cubic ~I!, capped pentagonal bipyramid ~II, The Fe2 16 spectrum is shown in Fig. 8 along with Fe13–15
2

shown in Fig. 6!, and a bcc fragment ~III! by removing a to emphasize the abrupt spectral change from Fe15 to Fe22
16 .
corner atom of a bcc Fe9 ~see below!. None of the calcula- The spectra of Fe2 17–24 are displayed in Figs. 9 and 10. In the
tions yielded satisfactory results. The stick spectrum plotted Fe216 spectrum, a new feature appears at the threshold. Except
in Fig. 6 is obtained from structure II, as shown. From its the new feature in the PES spectrum of Fe2 16 , other spectral
appearance, the spectrum of Fe2 8 is similar to those of the features are quite similar to those of Fe2 15 , and the sharp
larger clusters. As will be shown below, EHMO calculations, threshold feature observed in Fe2 15 becomes the second band
by assuming simple bcc structures, yield rather good agree- in Fe2 16 . The new feature observed in Fe 2
16 persists in the

J. Chem. Phys., Vol. 102, No. 24, 22 June 1995

9490 Wang, Cheng, and Fan: Size-selected metal clusters

higher clusters and seems to gain intensity as the cluster size ter size increases. On the other hand, the HOMO, while
increases until Fe2 19 , where it becomes a distinct sharp fea- maintaining significantly s-like, continues to evolve with in-
ture and at the same time the electron affinity becomes creasing participation from the d orbitals.
smaller. Although less dramatic compared to the spectral
change from Fe2 2
15 to Fe16 , the PES spectral change from Fe18
2
2 2 2 V. DISCUSSION
to Fe19 is significant. From Fe19 to Fe22 , the first feature near
the threshold becomes less sharp and the EA increases A. Electronic and geometrical structural evolution
steadily until Fe2 23 . Then, as seen before, a rather abrupt The electronic structure and EAs of the iron clusters ex-
sharp feature appears and the EA becomes smaller. The PES hibit strong size effect. The small clusters display rather well
spectral changes, taking place from Fe2 2 2
18 to Fe19 , and Fe22 to resolved spectral features. Fe3 and Fe4 have equilateral trian-
2 2 2
Fe23 , as well as from Fe15 to Fe16 , correlate remarkably well gular and tetrahedral structures, respectively,35,36 and their
with the chemical reactivity changes of the clusters with H2 , main PES features are reasonably accounted for by the
on which we will give an in-depth discussion later. EHMO calculations. However, for Fe5– 8, higher level calcu-
It seems that there is a four-atom periodic variation in lations are required to interpret their PES spectra; the current
the PES spectral pattern, from Fe16 to Fe19 and from Fe19 to EHMO calculations are insufficient probably due to their
Fe23 . This leads us to suspect that the cluster growth after the structural complexity. Fe6 and Fe7 have rather low EAs and
bcc Fe15 proceeds with a closely packed route; each addi- unusually diffused PES spectra. The EHMO calculations ap-
tional atom attaches to a cubic edge around an apex atom of pear to agree better for the larger clusters ~n>9!. We believe
Fe15 . There are 12 such positions. At every fourth atom ~i.e., this is due to the judicious choices of the cluster struc-
Fe19 and Fe23!, a rather high symmetry cluster is recovered, tures12,33 and their insensitivity to precise structural param-
consistent with the sharp PES spectral patterns observed for eters.
Fe2 2
19 and Fe23 . It is important to point out that these struc- The density of states as a function of cluster size in-
tures are different from the exclusive bcc structures used in creases rather rapidly. It is expected that, beyond Fe24 , little
the previous IP calculations.33 sharp features will be resolved even at the present instrumen-
We perform the EHMO calculations for these clusters tal resolution. However, comparing with nickel clusters in
using the structures described above. The results are dis- the same size range the iron clusters exhibit much more in-
played in Figs. 8 –10 as stick spectra. The agreement be- teresting features. This can be attributed to the stronger in-
tween the calculated energy levels and the PES spectral pat- teractions of the d orbitals in the iron clusters, that is already
terns is rather good, particularly below n520. Quite evident in the dimers.17 While iron dimer yields an interest-
impressively, the extra feature in Fe2 16 is well reproduced. ing and simple PES spectrum with only two bands within 1
Two other structures for Fe16 , a true bcc fragment and one eV above the neutral ground state, the nickel dimer spectrum
with atom-16 attached to a cube corner of the bcc Fe15 , yield is much more complicated.68 The nickel trimer69 also has a
quite different energy levels. As the cluster size increases, the much more complicated spectrum than the iron trimer spec-
energy level structures become more complicated and are trum reported currently.
increasingly difficult to compare with the experimental PES The reasonable agreement between the EHMO results
pattern, especially beyond Fe20 . This is understandable since and the PES spectra beyond Fe9 suggests that the bcc-type
there are more structural possibilities for the larger clusters. structure from Fe9–15, and the close-packed structure beyond
For Fe19 alone, we performed calculations with two addi- Fe15 are highly probable structures. In fact, the PES spectra
tional structures; fcc and polyicosahedral, two alternative themselves are already strongly suggestive that a new shell is
structures which have been discussed for Fe19 , previously.62 started at Fe16 in the cluster growth and this shell continues
The resultant energy levels are definitely worse compared to to grow for the larger clusters. This is consistent with the
the PES pattern than the one shown in Fig. 9. conclusions of the IP calculations by Pastor et al. for n59–
Upon inspecting the calculated wave functions, we 15 although they assumed bcc-like structures even for
found that the HOMO of Fe16 is composed mainly of s or- n.15.33 If proven to be true, this would be highly unusual
bitals; the orbitals responsible for the second band in the PES and interesting since these clusters already have similar
spectrum of Fe2 16 are mostly of d orbitals with considerable structures as the bulk at such small sizes. Parks et al.,
contributions from the d orbitals of the central atom, similar through their extensive chemical studies, have proposed a
to the situation found in the first band of Fe2 15 . Further ex- bcc structure for Fe15 .12 However, they suggest polyicosahe-
amination of the wave functions indicates that as the cluster dral structures for Fe19 and Fe23 when saturated with NH3 .62
increases its size from Fe15 to Fe16 , the HOMO of the former They have observed adsorbate-induced structural changes.
evolves into the second HOMO ~HOMO-2! in Fe16 , whose Thus, it is highly possible that the bare clusters may have
HOMO becomes the s-like orbital corresponding to the different structures. Therefore, their conclusions are not in
threshold feature of the PES spectrum. As will be discussed conflict with our structural assumptions. A recent simulated
in the next section, this abrupt electronic structure change annealing calculation concludes that the small iron clusters
underlies the reactivity changes of the clusters. have high symmetry and rather hollow structures.64 Since
As the cluster size continues to increase, the second band this calculation predicts the wrong ground state structure for
remains to be overwhelmingly d-like with a significant con- bulk Fe ~it yields a fcc rather than the bcc structure for Fe!,
tribution from the central atom, whose contribution can be we conclude that higher level calculations are needed to con-
traced even to Fe24 with its amplitude decreasing as the clus- firm the structures of these clusters.

J. Chem. Phys., Vol. 102, No. 24, 22 June 1995

 Wang, Cheng, and Fan: Size-selected metal clusters 9491

The ‘‘central atom effect’’—the first sharp PES peaks for degenerate or closely spaced levels as possible. This is due to
Fe2 9–15 all have large contributions from the central atom 3d the same exchange interaction that gives bulk magnetism.65
orbitals—provides an interesting support for the bcc struc- The correct prediction of the magnetism from the EHMO
tural assumptions and explains nicely about the seemingly calculations, although not from the first principle, is interest-
similarities in the clusters with n>9. Beyond Fe15 , the cen- ing.
tral atom effect manifests itself in the second PES band, and The dramatic size-dependent reactivity of these clusters
it can be traced up to Fe24 . Interestingly, a unrestricted with H2 is well known.8,9 Several interpretations have been
Hartree–Fock calculation, by using a tight-binding Hubbard advanced,8,9,28,29 but it is still not well understood. The cen-
model on the bcc Fe15 , also predicts that the central atom has tral question in interpreting the size-dependent reactivity of
a large contribution to the density of states near the Fermi these clusters is why the change of only one atom can affect
level.34 The spectrum of Fe2 8 is similar to the larger clusters the cluster reactivity so dramatically.
and should be classified with the Fe9–15 series. However, the In the present study, we intend to show that the iron
central atom effect cannot be found in Fe8 . Calculations are cluster chemical reactivity with H2 can be best described by
performed by using a true bcc fragment for Fe8 , in which a considering the detailed electronic structures of the clusters,
corner atom is removed from the bcc Fe9 . The energy levels based on the experimental PES spectra and the symmetries
do not agree well with the experiment as they do for the of the cluster frontier orbitals. First, from the calculated
higher clusters. It is possible that we do not find the right wave functions, we find that all the cluster molecular orbitals
structure for Fe8 in the EHMO calculations since, with only are fairly delocalized. The orbital delocalization indicates the
eight atoms, it cannot form a true bcc fragment. Yet, the sensitivity of the electronic structure to minor geometric
sharp feature at the threshold and the overall spectral pattern structure changes in the iron cluster ~e.g., addition or sub-
of Fe2 8 look quite similar to the higher bcc-type clusters. traction of a single atom!. If such a change results in a
How big a cluster is a metal has been an interesting change of the nature of the frontier orbitals, one expects that
question in cluster research. However, the answer may de- a cluster chemical reactivity change will follow. As described
pend on what properties one is interested in. Recent mag- above, the behavior of the HOMOs in the iron clusters varies
netic studies2 indicate that the cluster magnetic moments ap- with cluster size as revealed from the EHMO wave func-
proach the bulk limit at a cluster size of approximately 700 tions, as well as from the PES spectral changes. This will
atoms for iron.2 Clearly, the density of states increases rap- naturally lead to different cluster reactivity. More specifi-
idly with cluster size. However, even for the largest cluster cally, when H2 approaches a cluster, the effective orbital
investigated in the present study ~Fe24!, there are still recog- overlap between the H2 s* and the HOMO of the cluster is
nizable features in its PES spectrum. Recently, Zhao et al. critical in determining the reactivity. In general, two factors
have investigated the metal–nonmetal transition in transition will influence this orbital interaction; the energy level and
metal clusters with close-packed structures.70 They con- the symmetry of the cluster HOMOs. It is thus not surprising
cluded that the critical size for the metal–nonmetal transition to find a correlation between the IP and the reactivity9 be-
is about 30–50 atoms, and their criterion was that at the cause the IP is a measure of the HOMO levels. However, the
critical point the density of states of the cluster exceeds HOMO energy level change in the cluster size range of in-
1/k B T. Our instrumental resolution of 30 meV is close to terest is relatively small.24 In fact, the IP change is less than
k B T at room temperature. As mentioned above, up to Fe24 we 0.1 eV from Fe15 to Fe16 and less than 0.25 eV from Fe18 to
can still resolve features in its PES spectrum. The EHMO Fe19 , while the reactivity changes by orders of magnitude in
energy levels are also less dense compared to 1/k B T. These the same size ranges. The HOMO energy level match, there-
seem to agree with the theoretical predictions. With careful fore, is probably not the major effect in these cases. Siegbahn
account for the Franck–Condon profiles in the photodetach- et al. arrived at the same conclusion.28 Thus, the symmetry
ment process, further PES studies of the larger clusters of the cluster HOMOs becomes the dominating factor in de-
should be able to quantitatively test the conclusions for the termining the orbital overlap—hence the reactivity. We be-
critical sizes of the anticipated metal–nonmetal transition. lieve this provides the key insight into the dramatically dif-
ferent reactivity with H2 observed for the iron clusters.
B. Iron cluster magnetism, chemical reactivity, and As mentioned in Sec. IV, the HOMO levels, including
their electronic structure
those levels responsible for the sharp features in the spectra
Two of the most interesting properties of iron clusters of Fe2 9–15 and those levels responsible for the second band in
are their magnetism and chemical reactivity, which are Fe216 are mainly of 3d characters with large contributions
,
closely related to their electronic structures. The magnetic from the central atom. The symmetry of these HOMOs
properties of the iron clusters have been well studied and matches with the H2 s* orbital, leading to high reactivity.
they are all shown to be magnetic.2– 4,30–36 The magnetism of However, the HOMO of Fe2 16 , corresponding to the threshold
these clusters is also well revealed from the EHMO calcula- feature in its PES spectrum, is distinctly different; it mainly
tions and the number of unpaired electrons obtained from the has 4s characters with minor 3d contributions. It is not
EHMO calculations, as shown in Table I, are quite reason- symmetry-matched with the H2 s* orbital, hence an ex-
able compared with other theoretical calculations.30–36 The tremely low reactivity. The s-type HOMO of Fe16 provides a
unpaired electrons of these clusters arise from the nature of ‘‘shielding effect’’ to protect the cluster from being attacked
the EHMO energy level structures and the tendency to re- by the incoming H2 molecule despite the presence of the
duce the electron repulsion by occupying as many nearly symmetry-matched d-type orbitals corresponding to the sec-

J. Chem. Phys., Vol. 102, No. 24, 22 June 1995

9492 Wang, Cheng, and Fan: Size-selected metal clusters

ond band in the PES spectrum. This will lead to a high acti- and Fe2 4 can be interpreted well by assuming equilateral tri-
vation barrier for the reaction. There is about a factor of 20 angular and tetrahedral structures, respectively, and both
difference in the reactivity with H2 between Fe15 and Fe16 . should be Jahn–Teller active. The spectra of Fe2 5– 8 cannot be
From Fe16 to Fe18 , the HOMOs of the clusters are all interpreted in detail from the EHMO calculations and high
nondegenerate and are mainly composed of s orbitals. Fur- level calculations are required. Both Fe2 2
6 and Fe7 show
thermore, the HOMO energy spacing to the next occupied rather diffuse spectra and both yield unusually low electron
level is fairly apart. Therefore, the reactivity of these clusters affinities. Sharp spectral features are observed for larger
remains low. At Fe19 , the HOMO becomes twofold degener- clusters. For Fe2 9–15, the spectra are quite similar all with a
ate, due to the high symmetry of the cluster. It has four atoms sharp feature at the threshold. The EHMO results agree quite
arranged around an apex atom of the bcc Fe15 core, as shown well with the PES spectra when bcc structures are assumed
in Fig. 9. Although these two orbitals are still composed for these clusters. The spectrum of Fe2 16 is quite different
mostly of s orbitals with substantial d contributions, they can from that of Fe2 15 and an extra feature appears at the thresh-
readily recombine to form symmetry-matched molecular or- old, suggesting the start of a new shell in the cluster growth.
bitals with the H2 s* orbital. Therefore, it should exhibit a Both Fe2 2 2
17 and Fe18 have similar spectra as Fe16 , with in-
2
very different reactivity, compared to Fe18 . This is indeed creasing intensity for the extra feature. At Fe19 , this extra
what is observed in the experiment, that shows a reactivity feature becomes a well resolved peak and the spectrum shifts
increase of nearly 60 times in Fe19 . to a lower binding energy. This pattern of change from Fe2 16
From Fe20 to Fe24 , the calculated results suggest that to Fe2 2 2
19 is repeated from Fe20 to Fe23 . All these PES spectral
there are more energy levels near the HOMOs, whose d changes coincide with the chemical reactivity changes ~with
components continue to increase as the cluster size increases. H2! occurring in the same cluster sizes. The EHMO calcula-
They become increasingly easy to form symmetry-matched tions, by assuming close-packed structures with a bcc Fe15
frontier orbitals to the s* orbital of the incoming H2 mol- core, are in reasonable agreement with the spectra, in par-
ecule. One thus observes sustained high reactivity in this size ticular, for Fe2
16 . From the symmetry of the HOMO levels of
range. In particular, at Fe23 , the HOMO once again becomes the clusters and the PES spectral changes, a detailed micro-
twofold degenerate due to a higher symmetry of this cluster scopic explanation of the dramatic cluster size-dependent re-
as shown in Fig. 10. This results in another reactivity jump in activity with H2 is proposed. The EHMO calculations also
Fe23 over Fe22 . correctly predict the existence of magnetism in these clus-
In the above, we give a detailed microscopic interpreta- ters. The electronic and structural properties of these iron
tion of the dramatic size dependence of the iron cluster re- clusters are interesting and it would be desirable to confirm
activity with H2 , based on the cluster electronic structure. these conclusions by using high level theories.
Finally, we offer the following comments on the seemingly
natural correlation between the high structural symmetry and ACKNOWLEDGMENTS
high reactivity. Initially, the structural symmetry seems to be
L.S.W. gratefully acknowledges Dr. L. Bloomfield, Dr.
a dominant effect in determining the reactivity. This is ob-
O. Cheshnovsky, Dr. D. Ray, and Dr. R. E. Smalley for vari-
served in the cases of Fe15 , Fe19 , and Fe23 , which all possess
ous helpful discussions about the instrument and Dr. S. D.
high geometric symmetry with abruptly high reactivity.
Colson and Dr. L. Lou for valuable discussions of this work.
However, all clusters beyond Fe19 have high reactivity with
H.S.C. would like to acknowledge Dr. P. M. Mathias and Dr.
H2 , while they do not necessarily possess high structural
G. P. Pez for their critical reading of the manuscript. He is
symmetries. The EHMO calculations show that the ultimate
also grateful to Professor M.-H. Whangbo for providing the
origin lies in the electronic structure of these clusters. Nev-
extended Hückel program and to Dr. J. Ren for valuable
ertheless, the high geometric symmetry often leads to high
discussions. We thank Pacific Northwest Laboratory and the
degree of degeneracy. The degenerate or near-degenerate
U.S. Department of Energy, Basic Energy Sciences for the
HOMOs can be recombined to form symmetry-matched or-
development of the cluster PES apparatus. Support for the
bitals to interact effectively with the H2 s*, which subse-
study of transition metal clusters from the National Science
quently results in high reactivity.
Foundation ~Grant No. CHE-9404428! is gratefully acknowl-
edged. This work was performed at Pacific Northwest Labo-
VI. SUMMARY
ratory, a multiprogram national laboratory operated for the
An improved magnetic bottle time-of-flight photoelec- U.S. Department of Energy by Battelle Memorial Institute
tron spectrometer is described for the study of the electronic under Contract No. DE-AC06-76RLO 1830.
structures for a wide variety of clusters. The 30 meV resolu-
tion achieved is no longer limited by the Doppler broadening 1
See, for example, Physics and Chemistry of Finite Systems: From Clusters
due to the implementation of a momentum deceleration for to Crystals, edited by P. Jena, S. N. Khanna, and B. K. Rao ~Kluwer
the cluster beam. Thus, further improvement on the spec- Academic, Boston, 1992!, Vols. I and II.
2
I. M. L. Billas, A. Chatelain, and W. A. de Heer, Science 265, 1682
trometer resolution is possible.53 ~1994!.
The PES spectra of small iron clusters, Fe2 3–24, are pre-
3
D. M. Cox, D. J. Trevor, R. L. Whetten, E. A. Rohlfing, and A. Kaldor,
sented, and the data are interpreted with the help of EHMO Phys. Rev. B 32, 7290 ~1985!.
4
W. A. de Heer, P. Milani, and A. Chatelain, Phys. Rev. Lett. 65, 488
calculations. This is the first attempt to experimentally probe ~1990!.
the iron cluster electronic structures. Small clusters from Fe3 5
J. P. Bucher, D. C. Douglass, and L. A. Bloomfield, Phys. Rev. Lett. 66,
to Fe8 show rather well resolved features. The spectra of Fe2 3 3052 ~1991!.

J. Chem. Phys., Vol. 102, No. 24, 22 June 1995

 Wang, Cheng, and Fan: Size-selected metal clusters 9493

6 38
M. E. Geusic, M. D. Morse, and R. E. Smalley, J. Chem. Phys. 82, 590 O. Cheshnovsky, S. H. Yang, C. L. Pettiette, M. J. Craycraft, and R. E.
~1985!. Smalley, Rev. Sci. Instrum. 58, 2131 ~1987!.
7 39
M. D. Morse, M. E. Geusic, J. R. Heath, and R. E. Smalley, J. Chem. G. Gantefor, K. H. Meiwes-Broer, and H. O. Lutz, Phys. Rev. A 37, 2716
Phys. 83, 2293 ~1985!. ~1988!.
8
S. C. Richtsmeier, E. K. Parks, K. Liu, L. G. Pobo, and S. J. Riley, J. 40
K. M. McHugh, J. G. Eaton, G. H. Lee, H. W. Sarkas, L. H. Kidder, J. T.
Chem. Phys. 82, 3659 ~1985!. Snodgrass, M. R. Manaa, and K. H. Bowen, J. Chem. Phys. 91, 3792
9
R. L. Whetten, D. M. Cox, D. J. Trevor, and A. Kaldor, Phys. Rev. Lett. ~1989!.
54, 1494 ~1985!. 41
K. Rademann, Ber. Bunsenges. Phys. Chem. 93, 653 ~1989!.
10
E. K. Parks, K. Liu, S. C. Richtsmeier, L. G. Pobo, and S. J. Riley, J. 42
T. N. Kitsopoulos, C. J. Chick, Y. Zhao, and D. M. Neumark, J. Chem.
Chem. Phys. 82, 5470 ~1985!. Phys. 95, 1441 ~1991!.
11
P. J. Brucat, C. L. Pettiette, L. S. Zheng, M. J. Craycraft, and R. E. 43
C.-Y. Cha, G. Gantefor, and W. Eberhardt, Rev. Sci. Instrum. 63, 5661
Smalley, J. Chem. Phys. 85, 4747 ~1986!.
12 ~1992!.
E. K. Parks, B. H. Weiller, P. S. Bechthold, W. F. Hoffman, G. C. Nieman, 44
P. Kruit and F. H. Read, J. Phys. E 16, 313 ~1983!.
L. G. Pobo, and S. J. Riley, J. Chem. Phys. 88, 1622 ~1988!.
13
45
J. Fan and L. S. Wang, J. Phys. Chem. 98, 11 814 ~1994!.
D. M. Cox, K. C. Reichmann, D. J. Trevor, and A. Kaldor, J. Chem. Phys.
88, 111 ~1988!.
46
J. Fan, L. Lou, and L. S. Wang, J. Chem. Phys. 102, 2701 ~1995!.
14
E. K. Parks, G. C. Nieman, L. G. Pobo, and S. J. Riley, J. Chem. Phys. 88,
47
L. S. Wang, Surf. Rev. Lett. ~to be published!.
6260 ~1988!.
48
L. S. Wang, J. Fan, and L. Lou, Surf. Rev. Lett. ~to be published!.
49
15
J. Harris and R. O. Jones, J. Chem. Phys. 70, 830 ~1979!. L. S. Wang, J. Conceicao, C. Jin, and R. E. Smalley, Chem. Phys. Lett.
16
I. Shim and K. A. Gingerich, J. Chem. Phys. 77, 2490 ~1982!. 182, 5 ~1991!.
17
D. G. Leopold and W. C. Lineberger, J. Chem. Phys. 85, 51 ~1986!.
50
W. A. de Heer and P. Milani, Rev. Sci. Instrum. 62, 670 ~1991!.
18
D. G. Leopold, J. Almlof, W. C. Lineberger, and P. R. Taylor, J. Chem.
51
~a! G. Markovich, R. Giniger, M. Levin, and O. Cheshnovsky, J. Chem.
Phys. 88, 3780 ~1988!. Phys. 95, 9416 ~1991!; ~b! G. Markovich, S. Pollack, R. Giniger, and O.
19
M. Tomonari and H. Tatewaki, J. Chem. Phys. 88, 1828 ~1988!. Cheshnovsky, ibid. 101, 9344 ~1994!.
20
T. Noro, C. Ballard, M. H. Palmer, and H. Tatewaki, J. Chem. Phys. 100, 52
C. E. Moore, Atomic Energy Levels, Natl. Bur. Stand. Circ. ~U.S. GPO,
452 ~1994!. Washington, D.C., 1971!, Vol. II.
21
D. R. Salahub, in Ab Initio Methods in Quantum Chemistry—II, edited by 53
L. S. Wang, J. Fan, and H. Wu ~unpublished!.
K. P. Lawless ~Wiley, New York, 1987!. 54
Y. Wang, T. F. George, D. M. Lindsay, and A. C. Beri, J. Chem. Phys. 86,
22
M. D. Morse, Chem. Rev. 86, 1049 ~1986!. 3493 ~1987!.
23
J. Koutecky and P. Fantucci, Chem. Rev. 86, 539 ~1986!. 55
D. M. Lindsay, Y. Wang, and T. F. George, J. Chem. Phys. 86, 3500
24
S. Yang and M. B. Knickelbein, J. Chem. Phys. 93, 1533 ~1990!. ~1987!.
25
E. K. Parks, T. D. Klots, and S. J. Riley, J. Chem. Phys. 92, 3813 ~1990!. 56
A. Yoshida, T. Dossing, and M. Manninen, J. Chem. Phys. 101, 3041
26
E. A. Rohlfing, D. M. Cox, A. Kaldor, and K. H. Johnson, J. Chem. Phys. ~1994!.
81, 3846 ~1984!. 57
R. Hoffmann, J. Chem. Phys. 39, 1397 ~1963!.
27
L. Lian, C. X. Su, and P. B. Armentrout, J. Chem. Phys. 97, 4072 ~1992!. 58
J. V. Saillard and R. Hoffmann, J. Am. Chem. Soc. 106, 2006 ~1984!.
28
I. Panas, P. Siegbahn, and U. Wahlgren, in The Challenge of d and f 59
R. Hoffmann, Rev. Mod. Phys. 60, 601 ~1988!.
Electrons ~American Chemical Society, Washington, D.C., 1989!, p. 125.
29
60
E. Canadell and M.-H. Whangbo, Chem. Rev. 91, 965 ~1991!.
J. Conceicao, R. T. Laaksonen, L. S. Wang, T. Guo, P. Nordlander, and R. 61
S. L. Dixon and P. C. Jurs, J. Comput. Chem. 15, 733 ~1994!.
E. Smalley, Phys. Rev. B 51, 4668 ~1995!. 62
30 E. K. Parks, B. J. Winter, T. D. Klots, and S. J. Riley, J. Chem. Phys. 96,
C. Y. Yang, K. H. Johnson, D. R. Salahub, J. Kaspar, and R. P. Messmer,
Phys. Rev. B 24, 5673 ~1981!. 8267 ~1992!.
31
K. Lee, J. Callaway, and S. Dhar, Phys. Rev. B 30, 1724 ~1984!.
63
E. K. Parks and S. J. Riley, J. Chem. Phys. 99, 5898 ~1993!.
32
G. F. Holland, D. E. Ellis, and W. C. Trogler, J. Chem. Phys. 83, 3507
64
O. B. Christensen and M. L. Cohen, Phys. Rev. B 47, 13 643 ~1993!.
65
~1985!. W. A. Harrison, Electronic Structure and the Properties of Solids—The
33
G. M. Pastor, J. Dorantes-Davila, and K. H. Bennemann, Chem. Phys. Physics of the Chemical Bond ~Freeman, San Francisco, 1980!.
66
Lett. 148, 459 ~1988!. In the extended Hückel theory, the extra electron in the anion does not
34
G. M. Pastor, J. Dorantes-Davila, and K. H. Bennemann, Phys. Rev. B 40, alter the energy level structure of the cluster; it enters the lowest energy,
7642 ~1989!. singly occupied level.
35
J. L. Chen, C. S. Wang, K. A. Jackson, and M. R. Pederson, Phys. Rev. B
67
L. S. Wang, H. S. Cheng, and J. Fan, Chem. Phys. Lett. 236, 57 ~1995!.
68
44, 6558 ~1991!. J. Ho, M. L. Polak, K. M. Ervin, and W. C. Lineberger, J. Chem. Phys. 99,
36
M. Castro and D. R. Salahub, Phys. Rev. B 47, 10 955 ~1993!. 8542 ~1993!.
37
D. G. Leopold, J. Ho, and W. C. Lineberger, J. Chem. Phys. 86, 1715 69
K. M. Ervin, J. Ho, and W. C. Lineberger, J. Chem. Phys. 89, 4514 ~1988!.
~1987!. 70
J. Zhao, X. Chen, and G. Wang, Phys. Rev. B 50, 15 424 ~1994!.

J. Chem. Phys., Vol. 102, No. 24, 22 June 1995