Nuclear Instruments and Methods in Physics Research B 182 (2001) 73±83
www.elsevier.com/locate/nimb
Plasmon excitation in ion±solid interactions
Ra
ul A. Baragiola
a
a,*
, Catherine A. Dukes a, Pierfrancesco Riccardi
a,b
Laboratory for Atomic and Surface Physics, Engineering Physics, University of Virginia, Charlottesville, VA 29904, USA
b
Laboratorio IIS, Dipartimento de Fisica, Universit
a della Calabria, and INFM Unit
a di Cosenza,
87036 Arcavacata di Rende, Cosenza, Italy
Abstract
We discuss recent advances in the area of plasmon excitations in ion±solid interactions. The basic aspects of plasmons are described, including surface modes. Kinetic plasmon excitation, a major contribution to the energy loss of fast
charges in matter, should occur above a threshold ion velocity of 1.3 times the Fermi velocity. The unexpected observation of sub-threshold excitation is analyzed with the conclusion that the usually dominant mechanism is indirect
excitation by fast secondary electrons. Slow ions with suciently high potential energy can excite multipole surface
plasmons, as evidenced by the plasmon decay structure in the electron emission spectra. Ó 2001 Published by Elsevier
Science B.V.
1. Introduction
Plasmons are quantized collective charge-density oscillations in solids [1,2]. They were studied
theoretically in the 1950s [3], although the ®rst
experimental evidence appeared in 1942 [4], in the
form of multiple energy losses of electrons in thin
solid ®lms. Plasmons arise due to the overshoot of
the screening response of the electronic system to a
rapidly varying electric ®eld like that produced by
fast charges moving in solids. These excitations are
evidenced not only in the discrete energy losses of
electrons in electron energy loss spectroscopy
(EELS), photoelectrons or Auger spectroscopy.
Plasmons also occur as a ``shake-up'' excitation
*
Corresponding author. Tel.: +1-804-982-2907; fax: +1-804924-1353.
E-mail address: [email protected] (R.A. Baragiola).
produced by sudden changes in the occupation of
localized electron states, e.g., in the photoionization and decay of inner shells. In photoionization,
the electrons rush to screen the suddenly created
hole but overshoot the average static screening
position, leading to an oscillation of the screening
charge. The switching of the perturbation must be
fast, so that the transient electric ®elds have Fourier components with frequencies overlapping with
the plasmon frequencies at 1±5 1015 Hz (in the
range of ultraviolet light). The study of these
shake-up energy losses in X-ray photoelectron
spectroscopy is useful in surface analysis of solids
[5].
Plasmons occur in the volume of the solid and
are also localized in the surface region. The energy
of the plasmon is quantized at Ep ˆ hx, where x is
its frequency, and varies with the wavelength of
the oscillation, k ˆ 2p=k (called the dispersion of
the plasmon). Here, k is the plasmon wave-vector
0168-583X/01/$ - see front matter Ó 2001 Published by Elsevier Science B.V.
PII: S 0 1 6 8 - 5 8 3 X ( 0 1 ) 0 0 7 2 3 - 6
74
R.A. Baragiola et al. / Nucl. Instr. and Meth. in Phys. Res. B 182 (2001) 73±83
and q ˆ hk, its momentum. The plasma frequency
for a gas of free electrons of charge e and mass m is
related to the electron density n by
s
ne2
xp ˆ
;
…1†
me0
where e0 is the permittivity of vacuum. In reality,
electrons do not move freely but are subject to
forces due to the ion cores and to the polarization
cloud of correlated electrons. Hence, the term
``quasi-particle'' is preferred to electron, and an
e€ective mass m is used in Eq. (1). Plasmon energies deviate strongly from the free-electron value
in the case of metals with localized d-bands, like
silver.
A boundary introduces di€erent surface or interface oscillation modes, which have been treated
in various reviews, e.g. [2,6,7]. The normal or
monopole
surface plasmon (nSP) has an energy
p
Ep = 2 at the solid/vacuum interface, for q ˆ 0,
where Ep is the energy of the volume plasmon
(VP). Multipole surface plasmons (mSP) have energies closer to that of the VP at q ˆ 0, (0.80±
0.85)Ep , and a positive dispersion [2,8]. The mSP,
which can be viewed as a VP of the low electron
density surface region [9], is responsible for the
large enhancement of photoelectron emission
[10,11] at the mSP energy.
Plasmons are short-lived (K1 fs), enduring at
most a few oscillations before they decay into
single-particle interband transitions. The short
lifetime s introduces an uncertainty broadening
Cˆ
h=s of the plasmon loss peak seen in EELS.
Electrons excited in plasmon decay may escape
into vacuum and be detected [12,13]; their energy
distribution, Np …E†, is broad, as wide as the valence band if Ep is larger than the Fermi energy
[14±16]. The discontinuity at the Fermi level causes
a shoulder in Np …E†, broadened by the plasmon
width. Since the energy spectrum of all emitted
electrons, N …E†, also contains electrons excited by
other mechanisms, visualization of the plasmon
decay shoulder is enhanced in the derivative
dN …E†=dE [17], a standard method in Auger
spectroscopy. An example is the secondary electron spectra shown in Fig. 1, taken in our laboratory.
Fig. 1. Electron energy spectrum from clean Al(1 1 1) excited by
226.1 eV electrons at normal incidence, together with its derivative. The energy loss structure near the elastic peak is assigned to excitation of single and multiple SPs and VP. The
derivative indicates the structure due to the decay of SP and VP.
The central region of the derivative spectrum is magni®ed to
show the structures due to multiple plasmon and 2p Auger
decay.
2. Kinetic excitation of plasmons
Fast charges traversing solids create a wake of
electron density ¯uctuations behind them. If the
velocity is high enough, the valence electrons
cannot react adiabatically; e.g., electrons that rush
to screen a fast positive ion ®nd that it is gone by
the time they reach its path, creating an excess of
screening charge behind the ion that starts the
plasmon oscillation. The energy loss to plasmon
excitations is a major mechanism for stopping of
fast electrons and ions in solids, especially for light
elements, where the number of core electrons is
small. In spite of its importance, most studies
of plasmon excitation by fast ions have been
theoretical, due to experimental diculties. For
R.A. Baragiola et al. / Nucl. Instr. and Meth. in Phys. Res. B 182 (2001) 73±83
instance, one cannot use the energy loss of the
projectile, as in EELS [1], since it is obscured by
multiple continuum energy losses of the ion to
target nuclei. The alternative to study plasmons is
to observe their decay in characteristic electrons
or, far less likely, photons [18].
Allowed electronic excitations in solids are
typically described with a diagram of energy
transfer vs. momentum transfers, like that in Fig. 2
for the case of a free-electron gas. The region of
single-particle excitation is bound by E=EF ˆ
…q=qF †‰…q=qF † ‡ 2Š and E=EF ˆ …q=qF †‰…q=qF † 2Š
resulting from energy and momentum conservations. The plasmon line in the ®gure represents its
dispersion relation, and the width is omitted for
clarity. There is a maximum or cut-o€ momentum
qc for the plasmon, which is where the plasmon
line enters the single-particle region. At such high
momenta, the wavelength becomes comparable to
the average distance between electrons, the idea of
a density oscillation loses its meaning, and the
width becomes large by coupling to single-particle
excitations. The maximum energy transfer is given
by DEmax ˆ q…2q0 q†=2M, where q0 and M are
the momentum and mass of the projectile charge.
Thus, for electrons, the maximum energy loss
forms a parabola in E q space, where for ions,
q0  q and the maximum energy transfer is the
straight line DEmax ˆ qv, where v is the velocity of
the ion.
The rate of energy loss of fast charges in an
electron gas is given by
W …q; x† ˆ
8pe2
Im‰ 1=e…q; x†Š;
q2
…2†
where Im… 1=e† ˆ e2 =…e21 ‡ e22 † is called the energy
loss function and e ˆ e1 ‡ ie2 is the complex longitudinal dielectric function of the material. The
Coulomb divergence for small q (large distances)
in Eq. (2) is suppressed by screening, described by
Im‰ 1=e…q; x†Š and initially proportional to q2 .
The condition for plasmon excitation is that
e1 1 and e2 1; other criteria have been proposed [19,20].
Kinetic plasmon excitation by ions has attracted several theoretical studies [15,21±26]. Early
experimental work using ions of tens and hundreds
of keV [27±31] was limited to identifying the
plasmon decay structure and comparing plasmon
energies with theoretical predictions based on
plasmon dispersion. Indirect evidence of plasmon
excitations is provided by alignment and altered
energy losses of molecular ions in solids [32±34].
Recently, Ritzau et al. [35] discovered that, for
proton impact on Al and Mg, plasmons are excited
below the threshold predicted by theory from the
condition qc v ˆ Ep …qc †. For a free-electron gas
with Fermi velocity vF , the cut-o€ momentum
qc Ep =vF , the dispersion relation for the VP is
Ep2 …q† ˆ Ep2 …0† ‡ 0:6v2F q2F . Therefore
vth 1:3vF :
Fig. 2. Diagram showing the region of allowed energy transfer
hx and momentum transfer q (in units of the Fermi momentum
qF ) in an electron gas. Excitations can be to single particles
(shaded) or along the VP line (VP, shown for Al). Higher-order
processes are not included. The straight lines indicate maximum
energy transfers for a given q, for protons at the indicated energy. qc is the cut-o€ q (maximum plasmon momentum).
75
…3†
In the case of incident protons, Eth 40 keV [21]
for Al and 25 keV for Mg; these values are indicated in Fig. 3. The VP energy, determined by
the interception of qc v with the plasmon line, increases with increasing v due to the positive plasmon dispersion, as can be inferred from Fig. 2.
76
R.A. Baragiola et al. / Nucl. Instr. and Meth. in Phys. Res. B 182 (2001) 73±83
Fig. 3. Plasmon yields for proton impact on Al and Mg (data
from [35,36] …H †). Line R indicates R
osler's theory for direct
excitations [15]; lines SE and A denote the contribution from
secondary electrons and Auger capture, respectively [35]. The
data labeled H …† are from the measurements of Hasselkamp
and Scharmann [36]. The vertical lines indicate the threshold for
direct excitation, according to Eq. (3).
The dependence of plasmon decay energies with v
is found to agree with theory except near or below
vth [35,36].
3. Potential excitation mechanisms
Plasmon excitation can occur at even lower
velocities …v < vth † as a result of electron capture
by ions of suciently high potential energy. This
was discovered a few years ago in experiments
using 50±4500 eV He‡ and Ne‡ ions on Al and Mg
and explained by shake-up excitation during ion
neutralization (Fig. 4) [37]. Several theories of
potential plasmon excitation during neutralization
have been published [38±44]. Electron emission
from this process competes with the more studied
Auger neutralization (AN) (also called Auger
capture, AC), which can occur if the potential
energy of the ion exceeds twice /, the work function of the surface (Fig. 5). For low / surfaces, a
competing channel is resonance neutralization
followed by Auger de-excitation (AD) [45]. The
potential plasmon excitation mechanism is allowed
if the potential energy released when the ion neutralizes near the surface, En ˆ I 0 / eh equals
Ep . Here I 0 is the ionization potential of the ion I
shifted by the image interaction (2 eV) [37] and eh
is the energy of the ®nal hole in the solid, measured
from the Fermi level. With a work function of
4.3 eV for Al, slow He‡ (I ˆ 24:6 eV) and Ne‡
(I ˆ 21:6 eV) have enough energy to excite the VP
of Al but Ar‡ (I ˆ 15:8 eV) has not (Epv …q† >
15 eV) [2,7]. On the other hand, monopole SPs can
be excited by the three ions (Eps …0† ˆ 10:6 eV).
Potential plasmon excitation competes with
Auger electron emission due to AN and with AD
of excited atoms [45]. The characteristic spectral
signature of the plasmon decay electrons allows
their separation from electrons originating from
AN, AD, direct excitation in ion±atom and ion±
electron collisions, and Auger decay of core excitations [45,46]. For slow He‡ ions on Al and Mg,
plasmon decay is more important than AN; this is
more clearly seen for Mg in Fig. 4, where the different groups of electrons are well separated in
energy. The di€erence between the high-energy
edge of the two distributions is I 0 Ep /.
Therefore, a more general way to separate the
plasmon decay structure from that of AN is to use
incident ions with very di€erent I, like He‡ , Ne‡ ,
and Ar‡ . In addition, the width of the high-energy
edge in AN depends on ion velocity normal to the
surface [45], while that of plasmon decay is nearly
®xed, determined by the plasmon lifetime plus a
possible distribution of Ep …q†. From these arguments, it is clear that the prominent plasmon
shoulder that appears for impact with He‡ and
Ne‡ ions (and fast electrons), but not with Ar‡
ions (Fig. 4), cannot be attributed to AN involving
structure in the density of valence states [47], since
its energy is not correlated with I and is not
broadened by increasing the ion velocity. For Be,
only AN is seen (the structure shifts with the
R.A. Baragiola et al. / Nucl. Instr. and Meth. in Phys. Res. B 182 (2001) 73±83
Al
Mg
77
Be
Fig. 4. Electron energy distributions and their derivatives for polycrystalline Al, Mg and Be induced by 106 eV He‡ , Ne‡ and Ar‡ ions,
and 1 keV electrons (adapted from [37,66]). The vertical dotted and dashed curves represent the position of the monopole SPs and VPs,
which are only evident for Al and Mg due to their small widths. The short solid lines are the positions of the high-energy edge of AN
assuming an image shift of 2 eV of the ionic levels.
Fig. 5. One-electron energy diagrams appropriate for Auger capture, electron capture with plasmon excitation, and electronic excitation transfer from the projectile to a plasmon.
78
R.A. Baragiola et al. / Nucl. Instr. and Meth. in Phys. Res. B 182 (2001) 73±83
potential energy of the ion), probably due to the
relatively large plasmon width.
The plasmon dip in dN =dE for slow He‡ and
‡
Ne impact is similar to the VP dip excited by
incident electrons. The problems with assigning
the structure to VPs in the case of slow ions are
twofold. The energy is lower by 1 eV than that of
the lowest energy …q ˆ 0† VP and slow ions are
neutralized outside the surface. On the other hand,
when ions penetrate the solid, like the 4.5 keV
multiply charged Ne ions in the experiments of
Niemann et al. [48], the plasmon decay energy
corresponds to VP. The discrepancy is signi®cant
because of the high precision of both experiments
(1 eV) achieved by comparing results for incident ions and electrons under the same experimental conditions.
Recently, it was argued theoretically that signi®cant VP excitation could occur by external
charges not too distant from the surface [49].
However, in core-level photoemission of adsorbates, where both the hole and the emitted electron
remain outside the surface, only the surface plasmon (SP) is excited. These SP have low q and thus
energies that are too low compared to those seen in
slow ion neutralization. Surface plasmons with
high momentum appear in the theory of Monreal
[50] and the SP assignment is supported by the
sensitivity of the measured plasmon structure to
slight cesiation or oxidation of the surface [66].
However, the observed plasmon energy and width
are not consistent with the available data for
monopole SPs [7]. A better assignment, consistent
with the surface sensitivity, larger energies, and
moderate widths is the multipole SP (mSP). The
density ¯uctuation of the mSP has dipole character
[2] and could couple more easily with the surface
dipole formed by the incoming ion and its image
charge, that rapidly disappears upon neutralization (the image charge transfers to the ion and the
incoming hole transfers to the solid). Multipole
plasmons with energies of 13 eV have been inferred in photoyield experiments in Al [11], seen in
our laboratory in EELS spectra of Al(1 1 1) excited
by 140 eV electrons (Fig. 6) and shown in a contemporary EELS study using 50 eV electrons [51].
The plasmon losses cannot be understood by only
a monopole SP and a VP, but require a third ex-
Fig. 6. Electron energy loss spectrum of Al(1 1 1) for 140 eV
electrons, and plasmon loss structure after subtracting a smooth
background of single-particle excitations (dashed line). Incidence is normal to the surface and emission at 43° from the
normal. The plasmon loss structure is ®tted by three Gaussians
assigned to the nSP (monopole), the mSP, and the VP, with
respective peak energies of 10.36, 12.60 and 15.10 eV. Higher
energy losses are due to multiple excitations.
citation at an intermediate energy (12.60 eV),
which we assign to the mSP.
We now examine why mSP are excited for He‡
and Ne‡ below 1 keV [37] but VP for Ne‡ at
4.5 keV [48]. The solution of the apparent discrepancy came from Riccardi et al. [52], who
found that for 1 keV Ne‡ on Al, the plasmon
intensity is independent of incidence angle, while
at 5 keV, it increases rapidly with angle. These two
contrasting behaviors are characteristic of potential electron emission (occurring outside the solid)
and kinetic electron emission (occurring mainly
inside), respectively [45]. On average, excitations
occur deeper inside for normal than for oblique
incidence projectiles and thus the decay electrons
are more attenuated on their way to the surface.
The transition from potential to kinetic excitation
is apparent in Figs. 7 and 8 [37]. The plasmon
yields in Fig. 8 are obtained by subtracting a
background in the region 6±16 eV and integrating
the plasmon dip, a procedure discussed elsewhere
in these proceedings [53]. In potential excitation,
the energy spectra and the yields are nearly constant with incident energy. Above 1 keV, kinetic
emission sets in, increasing with incident energy,
R.A. Baragiola et al. / Nucl. Instr. and Meth. in Phys. Res. B 182 (2001) 73±83
79
(a)
(b)
Fig. 7. (a) Electron energy spectra dN =dE for 50±4500 eV Ne‡
ions on aluminum at 12° grazing incidence. (b) dN =dE. The
structure above 20 eV is due to auto-ionization from backscattered, doubly excited Ne . From [37].
and the energy of the plasmon dip increases from
that of an mSP to that of a VP, in agreement
with the 4.5 keV results of Niemann et al. [48].
Recently, Barone et al. [54] were able to separate
the contribution of mSP and VP by careful
analysis of the electron emission spectra. For
increasing projectile energy, the excitation of
mSP was found to be nearly constant or slightly
decreased, while that of VP increased from a
threshold at 1 keV. Similar conclusions were
reached from experiments in Berlin using 4.5 keV
Nen‡ …n ˆ 1; 6† on Al [16,48], which produce
plasmons with the same energy as VP excited by
incident electrons, indicating that the de-excitation of the multiply charged ions occurs after they
penetrate the surface [48]. This conclusion is
supported by the cosine angular distribution of
Fig. 8. Intensity of plasmon decay electrons for Ne‡ on polycrystalline Al at 12° grazing incidence. The intensity is derived
from the strength in dN =dE in Fig. 8 [37], after background
subtraction, as detailed in [53,55].
plasmon decay electrons, characteristic of a
source of electrons below the surface seen with
incident Ne [16] and Ar ions [53,55].
We note here that an additional mechanism for
potential plasmon excitation is possible during the
fast redistribution of charge that occurs during
the de-excitation of an atom or ion at a surface.
The resonant condition is achieved when the energy di€erence between the (broadened) atomic
levels equals the plasmon energy [56]. This excitation transfer (Fig. 5) may contribute to the decay
of the excited cloud of the ``hollow atom'' that
forms when a multiply charged ion captures electrons from a surface.
4. ``Sub-threshold'' kinetic plasmon excitation
Mechanisms that can excite VPs at velocities
lower than vth (Eq. (3)) were ®rst discussed by
Ritzau et al. [35]. These are: (a) the e€ect of ®nite
80
R.A. Baragiola et al. / Nucl. Instr. and Meth. in Phys. Res. B 182 (2001) 73±83
width of plasmons that allows lower-energy
excitations with energies lower than Ep …q ˆ 0†,
(b) second-order electronic processes, (c) absorption of momentum by lattice atoms by an Umklapp process, (d) the increase in neutralization
energy in potential excitation a€orded by the shift
in target electron energy in the projectile frame,
and (e) excitation by electrons faster than the
projectile.
The plasmon width can only introduce a small
correction to vth , for Al and Mg, since it is small
compared with Ep (the e€ect may be important in
other metals). Constraints of energy and momentum conservation are more easily met if plasmon
excitation is accompanied by an additional singleparticle excitation; such a second-order process
appears intuitively weak but has yet to be analyzed
theoretically. Alternatively, the second excitation
can go to a target atom that can easily absorb the
required momentum [57]. In a solid, atomic displacements are coordinated, and the small momentum transfer will be absorbed collectively by
the lattice (with or without phonons). Ritzau et al.
[35] considered lattice-assisted plasmon excitation
as a mechanism for sub-threshold excitation and
concluded that it is unlikely, since plasmon excitation in photoionization of valence electrons (also
requiring the lattice) is very weak and since lattice
e€ects are unimportant in the energy loss of H‡ in
Al for v < vF . We add that achieving energy and
momentum conservation is necessary but not suf®cient, since it is also important that the electron
gas be perturbed at high frequencies, which is not
obvious to occur in the lattice-assisted mechanism.
Nevertheless, the idea of absorption of momentum
by lattice atoms was retaken by Van Someren et al.
[58] at Utrecht, to explain prominent peaks in the
energy distribution of electrons they measured for
grazing collisions of 2±6 keV protons with Al(1 1 1)
surfaces. The structure was attributed to plasmons, but this assignment is unlikely since the
electron energy spectrum extends to more than
20 eV for 6 keV protons. S
anchez et al. [61] in
Bariloche did not ®nd these peaks in measurements by using glancing protons on very ¯at
Al(1 1 1), but rather the VP shoulder similar to that
reported by Ritzau et al. [35] for polycrystalline
Al. We have since done experiments using 4.4 keV
H‡
2 at 12° incidence on a clean Al(1 1 1) surface;
the spectra (Fig. 9) do not show the structure
measured in Utrecht; instead, there is a clear
indication of normal surface plasmons. Our observation conditions ± electrons collected at all
azimuths, at 43° with respect to the surface normal ± wash out di€raction e€ects that might have
a€ected the Utrecht results. Recently, Eder et al.
[59] in Vienna obtained evidence supporting the
Utrecht results using protons on Al(1 1 1), but the
peak energies and widths were found to depend
strongly on emission angle, and to be absent in
polycrystalline Al. Thus, Eder et al. concluded that
the structure is not due to plasmon decay but rather results from electron di€raction, which is well
known to modulate normal secondary electron
energy and angular distributions.
Fig. 9. Electron energy spectra from Al(1 1 1) excited by 4.4 keV
H‡
2 at 12° grazing incidence, compared to that excited by 1 keV
electrons at normal incidence [66].
R.A. Baragiola et al. / Nucl. Instr. and Meth. in Phys. Res. B 182 (2001) 73±83
4.1. E€ect of the Doppler shift of electron velocities
The plasmon excitations seen in the neutralization of slow (v vF ) rare gas ions at surfaces
can also occur for fast ions. For Al and Mg, VP
cannot be excited by stationary protons (I 0 11:6
eV) because the potential energy available for excitation En ˆ I 0 / is insucient. For moving
protons, the velocity distribution of valence electrons is Doppler shifted in the frame of the ion,
increasing the maximum energy release accompanying neutralization by DEs ˆ mvvF ‡ 12mv2 [35].
The shift may then enable neutralization, e€ectively increasing the potential energy available for
electron capture. Also contributing is the broadening DE ˆ hvn =a caused by the ®nite time the ion
spends near the surface, where a is the width of the
tail of electron density spilling outside the surface,
and vn is the ion velocity normal to the surface.
Although the argument is being used speci®cally
for ion±surface collisions, it is actually the same
occurring in plasmon-assisted electron capture
inside the solid [60]. In the atomic picture, the
broadening occurs due to the ®nite collision time,
and the transition probability peaks at the velocity
v
ha=DE (the Massey criterion), where DE is the
adiabatic energy defect in an electron capture
collision.
4.2. Excitation by fast secondary electrons
Ritzau et al. [35] concluded that the most likely
sub-threshold mechanism for protons is the indirect plasmon excitation by fast secondary electrons
produced in binary proton±electron interactions.
The minimum energy that an electron must have
to excite a VP is 23 (17) eV for Al (Mg), measured
from the Fermi level, when calculated in the random-phase approximation of a free-electron valence gas. To test the importance of excitation by
fast secondary electrons, one may look for a correlation between the number of plasmon decay
electrons and the number of electrons ejected with
sucient energy to excite a plasmon, in a given
electron energy spectrum. However, this comparison will depend on the path of the fast electrons
near the surface, which can be a€ected by the angle
of ion incidence [61]. The comparison may be
81
misleading since the fast emitted electrons represent only a tiny fraction of the total number of fast
electrons moving inside the solid, due to the small
escape depths (1±2 nm). In particular, most of
the electrons that have excited plasmons (and thus
lost energy) will appear in the low-energy part of
the energy spectrum. A better approach is to calculate the energy distribution of fast electrons
inside the solid from that of ejected electrons plus
a model for attenuation and escape [35]. An alternative is to do a full theoretical estimate that
includes the calculation of the initial distribution
of electrons ejected in binary collisions and a
transport calculation [62]. The results of both
methods di€er, not surprisingly given the approximations needed in both cases.
Plasmon excitation due to fast electrons from
binary collisions requires a threshold velocity, because the maximum energy transfer from a proton
to an electron at the Fermi level, 2mv…v ‡ vF † [63]
must lead to a ®nal electron energy exceeding the
minimum required for plasmon excitation. The
value of this threshold v0th is 0:74…0:64† 108 cm/s
for Al (Mg), quite lower than vth for direct excitation, and corresponds to 2.9 (2.1) keV/amu.
The situation di€ers somewhat for heavy ions.
Ar‡ ions excite plasmons at energies of tens of keV
[27,28] but not below 1 keV [37,52], as previously
mentioned. The question is then what is the
threshold energy for this type of projectiles. Riccardi et al. [64] found in Cosenza that plasmons
are readily excited by Ar‡ incident on Al at 60°
above 4 keV, and that the energy dependence extrapolates to a threshold at 1±2 keV, similarly to
the case of Ne‡ impact (Fig. 8). Since this
threshold is similar to that for Al-2p shell ionization in Ar±Al collisions [65] but above the value
expected from kinematic broadening (8 keV), it
was concluded that plasmon excitation is produced indirectly by the Al-LVV Auger electrons
(72.7 eV above EF ). This conclusion is supported
by the separate analysis of the SP and VP contributions in the spectra induced by Ne‡ impact [54].
In experiments with multiply charged Nen‡ ions
[16,48], the plasmon decay intensity increases when
going from n ˆ 1 to n ˆ 2 but then remains constant up to n ˆ 5, within errors. Modeling the
excitations resulting in the electronic cascade that
82
R.A. Baragiola et al. / Nucl. Instr. and Meth. in Phys. Res. B 182 (2001) 73±83
accompanies the neutralization of the multiply
charged ion shows a discrepancy between model
and experiment that increases with n [16]. This
di€erence can be explained by an increased probability of excitation by fast secondary electrons,
which become more abundant when n (and thus
the potential energy) increases.
5. Conclusions
In the last few years, substantial new results
were produced in di€erent laboratories on mechanisms for plasmon excitation in ion±surface collisions. Potential excitation can occur during
apparently adiabatic conditions (slow ion motion)
because the energy and high frequencies required
for excitations are provided by the fast capture of
a surface electron. This type of excitation can be
tuned by choosing projectile ions of di€erent potential energy. Moreover, by choosing suciently
low perpendicular velocity to prevent penetration,
slow ions become the only known way of exciting
plasmons purely outside solids.
Plasmon neutralization and AN are separable
by their distinct energy distributions and also, in
principle, by their di€erent time dependences
(Auger emission occurs promptly during neutralization, whereas plasmon decay is delayed by the
plasmon lifetime). The energy separation is very
clear in Mg due to the relatively small di€erence
between its plasmon energy and work function,
and serves to show that neutralization leads preferentially to plasmon shake-up rather than to an
Auger electron. A further di€erence between Auger and plasmon processes is that the electron
energy distribution in AN broadens strongly when
changing the velocity of the ion perpendicular to
the surface while that from plasmon decay is determined by the solid and depends only slightly on
excitation conditions.
Plasmons excited outside the surface by the
potential mechanism are most likely multipole
surface modes, as judged by their sensitivity to
surface conditions and by their energy that lies
between that of low-q VPs and that of high-q
monopole SPs. Plasmon excitation is predicted to
dominate neutralization when it is energetically
allowed, and thus may be relevant in secondary
ion mass spectrometry, and electron-stimulated
desorption of ions from surfaces. A related way to
excite SPs, distinct form electron capture, is the deexcitation of an excited atom or ion in front of the
surface, which may be a pathway for the relaxation of hollow atoms formed when slow highly
charged ions interact with surfaces.
In the impact regime where kinetic electron
emission is important, plasmons can be produced
additionally by secondary electrons inside the solid
(e.g., Auger and fast binary electrons) that have
energies above the threshold for plasmon excitation. Fast secondary electrons are responsible for
sub-threshold plasmon excitation by protons, for
kinetic plasmon excitation by keV Ne and Ar ions,
and contributes likely to plasmon observed in the
interaction of multiply charged ions with solids.
Acknowledgements
This research has been supported by an NSF±
CONICET cooperative research grant and by the
NASA Cassini Program.
References
[1] H. Raether, Excitation of Plasmons and Interband Transitions by Electrons, Springer, Berlin, 1980.
[2] A. Liebsch, Electronic Excitations at Metal Surfaces,
Plenum, New York, 1997.
[3] D. Pines, D. Bohm, Phys. Rev. 85 (1952) 338.
[4] G. Ruthemann, Naturwissenschaften 29b (1941) 648.
[5] T.L. Barr, Modern ESCA, CRC Press, Boca Raton, FL,
1994, Chapter 6.
[6] H. Raether, Surface Plasmons on Smooth and Rough
Surfaces and on Gratings, Springer, Berlin, 1988.
[7] M. Rocca, Surf. Sci. Rep. 22 (1995) 1.
[8] K.-D. Tsuei, E.W. Plummer, A. Liebsch, K. Kempa,
P. Bakshi, Phys. Rev. Lett. 64 (1990) 44.
[9] J.J. Quinn, Solid State Commun. 84 (1992) 139.
[10] H.J. Levinson, E.W. Plummer, P.J. Feibelman, Phys. Rev.
Lett. 43 (1979) 952.
[11] S.R. Barman, P. Haberle, K. Horn, Phys. Rev. B 58 (9)
(1998) R4285.
[12] N.B. Gornyi, L.M. Rakhovich, S.T. Skirko, Sov. Phys.
J. 20 (1967) 15.
[13] C. von Koch, Phys. Rev. Lett. 25 (1970) 792.
[14] M.S. Chung, T.E. Everhart, Phys. Rev. B 15 (1977) 4699.
R.A. Baragiola et al. / Nucl. Instr. and Meth. in Phys. Res. B 182 (2001) 73±83
[15] M. R
osler, W. Brauer, Springer Tracts Mod. Phys. 122
(1991) 1.
[16] N. Stolterfoht, D. Niemann, V. Ho€mann, M. R
osler,
R.A. Baragiola, Phys. Rev. A 61 (2000) 52902.
[17] L.H. Jenkins, M.F. Chung, Surf. Sci. 28 (1971) 409.
[18] Yu.A. Bandurin, L.S. Belykh, I.W. Mitropolsky, S.S. Pop,
Nucl. Instr. and Meth. B 58 (1991) 448.
[19] U. Fano, Phys. Rev. 118 (1960) 451.
[20] H. Ehrenreich, H.R. Philipp, in: A.C. Strickland (Ed.),
Proceedings of the International Conference on Phys.
Semiconductors, Bartholomew Press, Dorking, UK, 1962,
p. 367.
[21] M. R
osler, Nucl. Instr. and Meth. B 69 (1992) 150.
[22] A.A. Lucas, M. Sunjic,
Surf. Sci. 32 (1972) 439.
[23] M. R
osler, Appl. Phys. A 61 (1995) 595.
[24] R. Zimmy, Surf. Sci. 260 (1992) 347.
[25] F.J. Garcõa de Abajo, P.M. Echenique, Nucl. Instr. and
Meth. B 79 (1993) 15.
[26] J.A. Gasspar, A.G. Eguiluz, D.L. Mills, Phys. Rev. B 51
(1995) 14604.
[27] N.M. Omel'yanovskaya, S.Ya. Levedev, Sov. Phys. Solid
State 15 (1974) 1848.
[28] C. Benazeth, N. Benazeth, L. Viel, Surf. Sci. 78 (1978) 625.
[29] D. Hasselkamp, A. Scharmann, Surf. Sci. 119 (1982) L388.
[30] M.F. Burkhard, H. Rothard, K.-O.E. Groeneveld, Phys.
Stat. Sol. (b) 147 (1988) 589.
[31] N.J. Zheng, C. Rau, J. Vac. Sci. Technol. A 11 (1993)
2095.
[32] R.H. Ritchie, W. Brandt, P.M. Echenique, Phys. Rev. 14
(1976) 4808.
[33] D.S. Gemmell, J. Remillieux, J.C. Poizat, M.J. Guillard,
R.E. Holland, Z. Vager, Phys. Rev. Lett. 34 (1975) 1420.
[34] N.R. Arista, Nucl. Instr. and Meth. B 164±165 (2000) 108.
[35] S.M. Ritzau, R.A. Baragiola, R.C. Monreal, Phys. Rev. B
59 (1999) 15506.
[36] D. Hasselkamp, A. Scharmann, Surf. Sci. 119 (1982) L388.
[37] R.A. Baragiola, C.A. Dukes, Phys. Rev. Lett. 76 (1996)
2547.
[38] P. Apell, J. Phys. B 21 (1988) 2665.
[39] A.A. Almulhen, M.D. Girardeau, Surf. Sci. 210 (1989)
138.
[40] R. Monreal, N. Lorente, Phys. Rev. B 52 (1995) 4760.
[41] F.A. Gutierrez, Surf. Sci. 370 (1997) 77.
[42] M.A. Vicente Alvarez, V.H. Ponce, E.C. Goldberg, Phys.
Rev. B 27 (1998) 14919.
83
[43] F.A. Gutierrez, J. Jouin, S. Jequier, M. Riquelme, Surf.
Sci. 431 (1999) 269.
[44] H. Jouin, F.A. Gutierrez, C. Harel, S. Jequier, J. Rangama,
Nucl. Instr. and Meth. B 164 (2000) 595.
[45] R.A. Baragiola, in: J.W. Rabalais (Ed.), Low Energy IonSurface Interactions, Wiley, New York, 1994, Chapter 4.
[46] S. Valeri, Surf. Sci. Rep. 17 (1993) 85.
[47] A. Hitzke, J. G
unster, J. Kolaczkiewicz, V. Kempter, Surf.
Sci. 318 (1994) 139.
[48] D. Niemann, M. Grether, M. R
osler, N. Stolterfoht, Phys.
Rev. Lett. 80 (1998) 3328.
[49] A. Bergara, J.M. Pitarke, R.H. Ritchie, Phys. Lett. 256
(1999) 405.
[50] R.C. Monreal, Surf. Sci. 388 (1997) 231.
[51] G. Chiarello, V. Formoso, A. Santaniello, E. Colavita, L.
Papagno, Phys. Rev. B 62 (2000) 12676.
[52] P. Riccardi, P. Barone, A. Bonanno, A. Oliva, R.A.
Baragiola, Phys. Rev. Lett. 84 (2000) 378.
[53] P. Riccardi, P. Barone, M. Camarca, N. Mandarino, F.
Xu, A. Oliva, R.A. Baragiola, Nucl. Instr. and Meth. B
182 (2001) 84.
[54] P. Barone, A. Bonanno, M. Camarca, A. Oliva, F. Xu, P.
Riccardi, R.A. Baragiola, Surf. Sci. Lett. (in press).
[55] N. Stolterfoht, J.H. Bremer, V. Ho€mann, M. R
osler,
R. Baragiola, Nucl. Instr. and Meth. B 182 (2001) 89.
[56] J.I. Gersten, N. Tzoar, Phys. Rev. B 9 (1973) 4038.
[57] K.C. Pendey, P.M. Platzman, P. Eisenberger, E. Foo,
Phys. Rev. B 9 (1974) 5046.
[58] B. Van Someren, P.A. Zeijlmans van Emmichoven, I.F.
Urazgil'din, A. Niehaus, Phys. Rev. A 61 (2000) 32902.
[59] H. Eder, F. Auymar, P. Berlinger, H. Stori, H. Winter,
Surf. Sci. 472 (2001) 195.
[60] P.M. Echenique, F. Flores, R.H. Ritchie, Solid State Phys.
43 (1990) 230.
[61] E.A. Sanchez, J.E. Gayone, M.L. Martiarena, O. Grizzi,
R.A. Baragiola, Phys. Rev. B 61 (2000) 14209.
[62] M. R
osler, Nucl. Instr. and Meth. B 164±165 (2000) 873.
[63] R.A. Baragiola, E.V. Alonso, A. Oliva-Florio, Phys. Rev.
B 19 (1979) 121.
[64] P. Riccardi, P. Barone, M. Camarca, A. Oliva, R.A.
Baragiola, Nucl. Instr. and Meth. B 164±165 (2000) 886.
[65] R.A. Baragiola, E.V. Alonso, H. Raiti, Phys. Rev. A 25
(1982) 1969.
[66] R.A. Baragiola, S.M. Ritzau, R.C. Monreal, C.A. Dukes,
P. Riccardi, Nucl. Instr. and Meth. B 157 (1999) 110.