322
Nuclear
Symmetric versus asymmetric
from silicon
Instruments
collisions
and Methods
in Physics
in ion-induced
Research
858 (1991) 322-327
North-Holland
Auger emission
R.A. Baragiola
Dept.of Nuclear Engineering and Engineering Physics Uniuersity of Virginia, Charlottesville,
VA 22901, USA
L. Nair and T.E. Madey
Dept.of Physics and Astronomy and Lab. for Surface ~odi~c~tion,
Rutgers
Uttioersity, Piscatawqv,
NJ 08855, USA
We investigate the production
of Si 2p Auger electrons during bombardment
of Si surfaces by 1.0-5.1 keV Ar+ ions. Yields of
Auger electrons emitted at - 146 a with respect to the ion beam are measured as a function of ion energy. We observe a threshold of
- 3.8 keV for formation of a double L vacancy in Si, which is due to Ar-Si collisions. This value is higher than those of previous
studies, which were presumably
influenced
by contamination
of the Ar+ beam with multiply charged ions, as follows from our
measurements
of the variation of the Auger yields with the energy of the electrons in the ion source. We conclude that Auger
emission in our energy range is dominated
by symmetric collisions involving a fast Si recoil, with excitation in asymmetric,
Ar-Si
collisions growing in importance
above the - 3.8 keV threshold. The conclusion is also supported
by the appearance
of additional,
Doppler shifted, Auger peaks at high projectile energies. Our threshold value for Ar-Si collisions is in good agreement with previous
studies of gas-phase collisions, and should be used to reevaluate recent computer simulations
which fail to predict the observations.
1. Introduction
When keV ions bombard
solid targets of light elements, they may produce holes in shallow inner-shell
(core) levels of atoms (11. These holes can decay by
X-ray emission but more likely by ejecting Auger electrans, easily identified
in the high-energy
tail of the
energy distribution
of ejected electrons. Auger transitions normally involve two outer electrons, one filling
the core hole and the other being ejected, though a
small percentage
of the decays can be expected
to
involve three or more electrons.
Inner-shell
excitations
result from violent binary
atomic collisions between the projectile (p) and a target
(t) atom (asymmetric or pt collisions) or between a fast
recoil and another target atom (symmetric or tt collisions). Depending on the nature of the colliding pairs,
the core excitations can occur in the projectile and in
the target atom. Excitations
in the adiabatic region, i.e.
for impact velocities much lower than those of orbital
electrons, are explained by describing the slow collision
as a time evolution of electronic states of the quasimolecule formed transiently during the encounter [2]. Excitation occurs by compression
of the electron clouds during the collision and promotion
of some molecular
states due to the Pauli principle. Electron promotion
occurs in many cases over a very narrow range of
internuclear distances, R,, which causes excitation cross
0168-583X/91/$03.50
0 1991 - Elsevier Science
Publishers
sections to have a sharp kinetic energy threshold, which
is that required for the distance of closest approach to
equal R,.
The lifetimes of 2p holes in the light elements (Na to
Si) [3] are shorter than or comparable
to the lifetime of
collision cascades in solids. Therefore the Auger decay
subsequent
to the excitation collision can occur in the
bulk or in vacuum (for scattered projectiles or sputtered
atoms). Auger spectra will be different in these cases,
and reflect the different
distribution
of valence electrons at the location of the core hole, An important
consequence
of Auger decay of excited projectiles
outside the solid is a relatively high ion/neutral
ratio for
projectiles traversing thin foils [4], and backscattering
from solids [5,6]. On the other hand, Auger emission
from sputtered target atoms can produce high ion yields
in secondary ion mass spectrometry
[7-91.
In the case of second-row elements, the Auger spectrum from L-shell holes in target atoms consists of a
broad structure,
and sharp (1 eV wide or narrower)
peaks. The broad structure is readily associated with
Auger transitions
in the bulk, very similar to those
induced by electron bombardment.
In the case of LW
Auger emission from the decay of L-shell holes, the
spectrum is in a first approximation
the self-convolution of the density of states of the valence (V) band.
weighted by transition
probabilities
which depend on
the energy and symmetry of the state. Therefore
the
B.V. (North-Holland)
323
R.A. Baragiola et ul. / Colirsions in ran-induced Auger emission from Si
width of the LVV spectrum is large, close to twice the
width of the valence band, or lo-30 eV, depending on
the type of solid.
The understanding
of the origin of the sharp peaks
has not been so straightforward,
and has been surrounded by some controversy.
The clarification
came
from the analysis of Doppler shifts and broadening
of
the peaks [lO,ll], and from the observation
of energy
shifts accompanying
work function changes induced by
alkali adsorption [12]. It is now generally accepted that
these sharp peaks correspond
to atomic transitions
in
core-excited sputtered target atoms decaying in vacuum.
A topic still subject of controversy is the relative role
of pt versus tt collisions in core excitations
at keV
energies. Here the evidence is indirect, and comes from
comparison between experiments and computer simulations, which have reached a high degree of development.
The Monte Carlo simulations
of Be K emission [13]
showed an excellent agreement with the experimental
energy dependence
of the Auger yields. This provided
firm grounds to determine the relative importance of pt
and tt processes in X-ray emission yields, more accurately than possible using analytical
theory. Recent
simulations
of L-shell excitations
in bombardment
of
solid Al targets with Ar ions [14] conclude
that tt
collisions are largely responsible
for Auger yields at
very low energies. This is consistent with the observation that Auger yields from Al [15] and Si [16] compounds vary with the square of the concentration
c of
the light element which is excited. At higher energies, pt
collisions are expected to be dominant, and the growth
of the yields with c’ are seen to include a linear component.
However, and in contrast with these work, recent
theoretical
studies by Shapiro and Fine [17] have reopened the question proposing
that tt collisions are
unimportant
and that pt collisions are responsible
for
the Auger emission
from Al near the excitation
threshold.
In this paper we address the question of the threshold
behavior and relative importance
of Ar-Si and SiPSi
collisions in the emission of Si Auger electrons.
We
leave for a subsequent publication the detailed analysis
of the Si 2p Auger spectra.
2. Experimental
The measurements
were performed
in a UHV apparatus (fig. 1) with a base pressure of < IO-to Torr.
The sample was a mirror-polished
Si (111) crystal, but
the original crystalline structure is not important in this
experiment
since the sample was amorphized
during
cleaning by ion bombardment.
Ions were produced with
a differentially
pumped, electron bombardment
ion gun
not equipped with mass analysis. However, we took care
50mm
rad.
Electrostatic
Detector
50-1000
ev
Electron
Gun
Fig. 1. Schematic drawing of the UHV experimental system.
to obtain a pure ion beam by feeding the thoroughly
outgassed ion source with > 99.9995% pure Ar and by
controlling
the amount of doubly charged ions in the
beam by varying the energy of the bombarding
electrons.
Electron spectra were taken with a VSW, 50 mm
radius, hemispherical
electron
spectrometer
working
normally at 0.4% resolution,
with an acceptance
angle
of - 3O. The angle between the ion beam and the
entrance cone of the spectrometer
was fixed at 146O.
The sample could be rotated continuously
around an
axis perpendicular
to the plane determined
by the ion
beam and the spectrometer
axis.
To study the effect of oxidation on the Auger spectra
we needed to produce stable oxidized silicon surfaces.
This was achieved by replenishing
the oxygen removed
by sputtering
with adsorption
from an ambient
of
oxygen at 1.6 x lo-’
Torr maintained
during ion
bombardment.
3. Results
Typical spectra are shown in figs. 2-4. The main
features have been identified before: the broad structure
corresponds
to Si decaying in the bulk, whereas the
sharp peaks correspond
to transitions
in sputtered Si”
and Si + with one L-shell vacancy. A weak structure
around 105 eV (fig. 3) corresponds
to the decay of
atoms with two L-shell holes and is observed only for
large incidence angles. Excitation of Ar Auger lines is
not observed in our energy range.
The importance
of the geometry on the observed
spectra is shown in fig. 2, which compares data taken at
normal take-off and 2” grazing take-off angles. We
expect the emission from sputtered atoms to be independent of the observation
angle, if the Auger emission
is isotropic in the frame of the decaying Si [18]. On the
other hand, electrons
produced
in the bulk of the
target at a depth d will be attenuated
by a factor
I. ELECTRON EMISSION
R.A. Baragiola et al. / Collisions in ion-induced Auger emission from Si
324
5.1 keV
Ar+
- Si
5.0 keV Ar+ - Si
I
~
01
90
80
Energy (eV)
70
60
J
10
1
100
Fig. 2. Electron energy spectra of Si bombarded
by 5.1 keV
Ar+ in the region of LW and LMM Auger transitions,
for
emission (a) normal to the surface and (b) at 2O from the
surface. The angle between the directions of the ion beam and
the ejected electron is 146 O.
exp( -d/L
cos 8,), where L is the attenuation length
and 0, the angle of emission with respect to the surface
normal. For grazing observation conditions (6’, - 90 o )
most of the bulk emission is eliminated and the spectrum is due mainly to Auger processes in sputtered
atoms [19].
The Auger yields increase very sharply with energy,
but the shape of the spectra is not affected much until
the impact energy exceeds - 4 keV. Then, as illustrated
in fig. 3, a tail appears on the low energy side of the
sharp atomic lines, the peak at - 90 eV grows in
relative importance and a peak due to double 2p excitation becomes visible at - 104 eV.
Oxidation causes a dramatic change in the Auger
spectra and on in its dependence with angle, as shown
1
7
1
!/
0
Ar+ - Si
L
55
65
75
85
95
105
115
Electron energy (eV)
with
Fig. 3. Electron energy spectra of Si under bombardment
and 5 keV (-)
Arf ions incident at an
3 keV (0.0)
angle of 8.5” with respect to the surface. Auger decay from
double 2p excitations
is clearly seen at -103
eV, in the
magnified portion of the 5 keV spectrum.
65
75
85
95
105
115
Electron energv (eV)
Fig. 4. Electron energy spectra of Si bombarded
by 5 keV
Ar+ions.
(a) Clean surface, and (b) surface oxidized by exposure to 1.6 x lo-’ Torr 0, during ion bombard ment. Emission is at 2 o from the surface.
in fig. 4 for spectra taken at a 2O take-off angle to the
surface, for clean and oxidized surfaces. No atomic
peaks appear over the broad LW
feature when
bombarding oxidized samples near normal incidence.
We measured relative Auger yields for different ion
energies. To avoid problems of changes in the beam size
with energy, we scanned the beam over an area of
constant size and larger than the acceptance area of the
electron spectrometer. The spectra were normalized to
the ion beam current after suppressing secondary electrons. Since we suspected that previous yield measurements by other workers were influenced by the presence
of multiply charged projectile ions we measured spectra
with 50 eV bombarding electron energy in the ion
source, a condition known to produce a negligible fraction of A?+ [20].
The yields were defined as the area of the peaks,
after the subtraction of a background, divided by the
incident beam charge used to record the data. The
background was calculated by fitting a function of the
form A + B/E” to the high-energy side of the spectra,
where A, B and n are constants. In fig. 5 we present
yields for the single L-hole Auger transitions (between
80 and 95 eV to minimize uncertainties in background
subtraction). In fig. 6 we plot yields for the double
L-hole Auger decay. A very different kinetic energy
dependence is observed for the these two lines. We can
estimate thresholds by extrapolating the measured data.
These are - 1.0 keV for single L-shell excitations and
- 3.8 keV for double L-shell excitations.
We also measured L’MM yields using 235 eV electrons in the ion source, which is a typical operating
condition for this type of ion source. These yields are
also shown in fig. 6. They suggest lower apparent
thresholds as a result of contamination of the ion beam
with Ar2+ of twice the energy, consistent with results
R.A. Baragiola et al. / Collisions in ion-induced Auger emission from Si
g
4000
.
2000
.
325
2
E
is
3
s
E
Ar
energy
(keV)
Ar
energy
(keV)
Fig. 5. Auger yields from single L-shell vacancies in Si. (a) Linear scale, and (b) logarithmic scale. The data were measured with the
ion source operated in conditions to produce a negligible Ar”
by Valeri and Tonini [21] obtained with a conventional
ion source.
4. Discussion
4. I. Analysis of threshold data
Fig. 6 demonstrated that low thresholds as obtained
in previous experiments
can result from contamination
400
contamination.
normal.
Angle of incidence is 74O with respect to the surface
by doubly charged ions. With the values obtained in the
absence of contamination by Ar2+ we can now derive
the distance of closest approach required for excitation,
once we identify the collision partners: Ar-Si or Si-Si.
This identification can be done for the double core
excitation, which occurs when the distance of closest
approach R, in the collision is equal or smaller than
R,, the distance at which the promoted molecular orbital
(MO) crosses empty MOs. The threshold occurs when
R, = R,, and should be essentially the same for single
i
4
I
Si - :MM
G 300
I=
-2
E
z 200
2
u
Tii
*_
>-
P
100
P
0
0
2.0
0
3.0
Ar
*
.,*e.
4.0
energy
0
I
,
5.0
(keV)
Ar
energy
(keV)
Fig. 6. Auger yields from double L-shell vacancies in Si as a function of E, the energy of bombarding Ar+ ions. (a) Linear scale and
(b) logarithmic scale. The energy of the electrons in the ion source is: l 50 eV and o 235 eV. In the latter case, the ion beam is
contaminated by a significant fraction of Ar 2+ ions of energy 2E. Angle of incidence is 39O with respect to the surface normal.
I. ELECTRON
EMISSION
326
R.A. Baragmlu et (11./ Colhsions in ion-induced Auger emission from Si
and for double excitation. In both asymmetric (Ar--Si)
and symmetric
(Si-Si) collisions,
one or two Si 2p
vacancies can be produced by electron promotion in the
4fo MO [22]. However, as pointed out previously [23], if
two vacancies are produced
in a symmetric collision,
they will end up one in each collisions partner, rather
than two vacancies in one atom. This is easily understood by noticing that the formation of a Si(2p 2, + Si
final state is - 23 eV more inelastic than that of a
Si(2p-‘) + Si(2p-‘) final state [24].
Therefore, double L-shell excitation results from ArSi collisions, and the - 3.8 keV threshold energy is that
required for the 4fo MO to cross empty MOs in the
ArSi quasimolecule
and end up with one or two 2p
vacancies in the Si atom. Using the Moliere interatomic
potential, we obtain R, = 0.358 A for Ar-Si excitation
collisions, in excellent agreement with R, = 0.344 A, the
value calculated by Schneider et al. [22]. Based on the
similarity of cross sections for Si-Ar collisions as compared to Ar-SiH,
collisions [25], we do not expect large
solid state effects, though a slight increase in R, can be
expected due to enhanced screening in the solid.
These results then suggest that Auger emission below
- 3.8 keV is essentially determined by symmetric Si-Si
collisions initiated by fast Si recoils set in motion by the
Ar projectiles, This conclusion, which is consistent with
previous work in Al, disagrees with the simulations
of
Shapiro and Fine. The reason for this appears to be that
these workers based the simulation on the assumption
that the already artificially low threshold value obtained
by Vrakking and Meyer [26] was due in part to asymmetric collisions.
Our interpretation
also serves to explain the reason
for the nonobservation
of double excitations at normal
incidence. The angular distribution
of the doubly excited Si resulting from Ar-Si collisions lies in a cone
near the incident direction. The aperture of the cone is
determined
by the deflection in a collision where the
distance of closest approach equals R,. It increases with
increasing Ar energy but is limited to 42” for 5.1 keV
Ar (our maximum energy). Therefore the primary excited Si recoils will move into the solid for incidence far
from glancing. The small proportion
which will be
backscattered
into vacuum will be further attenuated by
Auger decay in the solid.
4.2. Double peaks
At our highest energies, additional atomic peaks or
shoulders appear at the low-energy side of the atomic
lines (e.g. 85 eV peak), as shown in fig. 3. We interpret
the double peaks in terms of the Doppler shift of two
distinct groups of excited Si. One, with low energies,
leading to sharp peaks, present at all impact energies.
This we assign to Si-Si collisions. The other group of
excited Si have higher energies and produce a noticeable
Doppler
shift (to lower energies for our geometry).
These faster Si recoils are assigned to Ar-Si collisions
since they start to be noticed above - 4 keV.
The question that remains is, at what energy does the
contribution
of Ar-Si collisions to the total Si yield
becomes comparable
(or larger) to that from Si-Si
from pt
collisions? Fig. 3 suggest a - 15% contribution
collisions at 5 keV and 81.5” incidence. Our previous
simulations for Al suggests a 50% contribution
at around
9 keV, for 45” incidence.
In fact, the very detailed
measurements
of Bonnano and co-workers
at 10 keV
[27] can be fitted with a model which includes only
Ar-Si collisions, and which gives a fitted critical interatomic distance
for excitation
equal to 0.355 A, in
remarkable
agreement with the findings of the present
work. Thus, a consistent picture taking into account all
previous experiments
is one in which Si-L emission in
Ar-Si collisions is dominated
by symmetric Si-Si collisions involving fast recoils, below 3.8 keV incident
energy. Above this threshold,
direct excitation by the
projectile
grows in importance,
possibly
becoming
dominant above 10 keV.
4.3. Emission from oxidized surfaces
By analogy with the work on silicides [16], it was
expected that oxygen will lower the probability of Si-Si
excitation collisions and that the Auger spectra would
be dominated
by Ar-Si collisions. We indeed observed
a very strong reduction in Auger intensity upon oxidation (fig. 4). Furthermore,
this also explains why we
observe atomic peaks from sputtered Si only for ions
incident obliquely. Although a recoil directed towards
the interior of the sample can be backscattered
and
ejected outside the solid, the probability
of doing so
while retaining the core vacancy (without decaying in
the bulk) will be low compared with a directly ejected
recoil. This interpretation
is also consistent
with the
absence of the sharper features attributed
to recoils
originating from SiCSi collisions.
5. Conclusions
Observations
of electron spectral features, energy
dependence
of the yields and the effect of oxidation
point to a consistent
picture of ion induced Auger
emission from Si. At low energies, from a threshold at
- 1.0 to - 3.8 keV, excitation occurs only to single hole
states, in collisions between a fast Si recoil and another
Si atom. Above - 3.8 keV, direct excitation is possible
in Ar-Si
collisions
which produces
more energetic
sputtered Si(Zp-‘) and also Si(2pP2) atoms. Oxidation
suppresses
most excitations
from Si-Si collisions and
produces drastic changes in the Auger spectra.
References
I51
161
[‘I
[Xl
I91
1101
1111
1121
iI31
[I41
1151
[I61
[I71
R.A. Baragiola. Radiat. Eff. 61 (1982) 47.
U. Fano and W. Lichtcn, Phys. Rev. Lett. 14 (1965) 627;
M. Barat and W. Lichten, Phys. Rev. A6 (1973) 211.
M.O. Krause. J. Phyx. Chem. Ref. Data 8 (1979) 307.
R. Baragiola.
P. Ziem and N. Stolterfoht.
J. Phys. B9
(1976) L447:
I’. Zirm. R. Baragiola and N. Stolterfoht.
in: Proc Int.
C‘onf. Phyh. X-Ray Spccrra. US GPO. Washington
(1976)
p. 278.
R.A. Baragiola.
in: Proc. 7th. Int. Conf. Atom. CotI.
Solids. Moscow St. Univ. Press (1977) p. 15 1.
0. Grwi. M. Shi. H. Bu. J. Rabalais and R.A. Baragiola,
Phys. Rev. B41 (1990) 4789.
K. Wittmaack.
Surf. Sci. 90 (1979) 557.
E.W. Thomas. Vacuum 34 1031 (1984):
J. Mischler and N. Benazeth, Scanning Electron. Microsc.
II (19X6) 351.
V.A. Pazdrersky
and B.A. Tsipinyuk.
Vacuum 35 (1985)
255.
R.A. Bsragola.
in: Inelastic Particle Surface Collisions,
Springer Ser. Chem. Phys. 17. ed>.. E. Taglaucr and W.
Heiland (Springer, Berlin. 1981) p. 3X.
R.A. Baragiola. F..V. Alonso and H. Rairi. Phys. Rev. A25
(19X2) 1969.
G.F.. Zampwri and R.A. Baragiola. Phys. Rev. B2Y (1984)
14x0.
0. Grwi and R.A. Baragiola. Phyh. Rev. A30 (1984)2297.
0. Grixi and R.A. Baragiola. Phys. Rev. A35 (1987) 135.
P. Viaris de Lesegno and J.-F. Hennequin,
Surf. Sci. 103
(lYX1) 257.
S. Valcrl. R. Tonini and G. Ottaviani.
Phys. Rev. B38
(1988) 13282. and references therein.
M.11. Shapiro and J. Fine. Nucl. Instr. and Meth. B44
(19X9) 43.
11x1K. Saiki and S. Tanaka,
Nucl. Instr. and Meth. B2 (1984)
512.
1191 L. de Ferrariss. 0. Gri7z.i. G.E. Zampieri. E.V. Alonso and
R.A. Baragiola. Surf. Sci. 167 (1986) 1.175.
[201 W. Bleakney, Phys. Rev. 36 (1930) 1303; Electron Impact
Ionization.
eds., T.D. Yaerk and G.H. Dunn, (Springer.
Wien. 1985).
S. Valcri and K. Tonini, Surf. Scl. 230 (1989) 407.
PI
Phys.
1221 I>. Schneider. G. Nolte. U. Wille and N. Stolterfoht.
Rev. A28 (1983) 161. Excitation
by 3d&3dn-3do
rotational coupling
is expcctcd
to be neghgible at keV enrrgics.
1231 M.F. Rudd, B. Fastrup. P. Dahl and F.D. Schowengerdt.
Phys. Rev. AX (1973) 220.
from D.A. Shirley. R.L. Martm. S.P.
1241 Binding energies
Kowalcqk.
F.R. McFeely and L. I.ev. Phys. Rev. B15
(1977) 544. The cncrgy of the S1(2p ‘) configuration
ix
assumed to be approximately
that of P + (2~ ’ ). by using
the Z + 1 equwalent
model [see. e.g. P. I)ahl. G. llcrmann, B. Fastrup
and M.t:. Rudd. J. Phys. B9 (1976)
15X1].
and I>. Schneider, in: Inner1251 R.D. Duhois. N. Stolterfoht
Shell and X-ray Physics of Atoms and Solids, eds., D.J.
Fabian, H. Kleinpoppen
and L.M. Watson (Plenum. NY.
19X1) p. 63.
J.J. Vrakking and A. Krocs. Surf. Sci. X4 (1979) 153. The
WI
authors have acknowledged
the presence of Ar”
contamination
in their beam and their threshold energies are
smaller
than
those for Al excitation
obtained
with
charge/mass
analysis of rhe primary
beam and higher
sensitivity (ref. [IO]).
[27] A. Bonanno.
F. Xu. M. Comarca.
R. Siciliano and A.
Oliva. Nucl. Inhrr. and Mcth. 84X (1990) 371 and privarc
commumcalions.
1. ELECTRON
EMISSION