Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

.

Surfa<:.: S<:ience 247 (1991) 125-132 125


Nonh-Holland

Image potential states at surfaces *

P.M. Echenique
Departamento de Física de '''fateriales. Facultad de Química. Universidad del País Vasco. Apdo /072. 20080 San Sebastián. Spain

and

M.E. Uranga
Departamento de Matemática Aplicada. Escuela Técnica Superior de Ingenieros Industriales. Universidad del País Vasco. Bilbao. Spain

Received 21 may 1990; acc.:pted for publication 30 May 1990

Tbe main physics of image potential induced surface states is presented in a simple manner. Recent experimental data on .Iifetime

~ is compared with theoretical predictions. The role of image states in STM. as a two-dimensional
noise at disorded surfaces is discussed.
electron gas ando as a source of a 1//

1. Introduction energy and momentum of an incident electron and


the energy of the emitted photon. In the last years
That electrons can be localized at surfaces has two photon experiments have provided good reso-
been recognized for a long time. The electrons in lution data bn binding energies and effective
such state cannot penetrate into the bulk when the masses [6]. and even lifetime data [7.9] has been
material (solid or liquid) shows a negative electron extracted. Energy loss experiments have produced
affinity and do not have enough energy to escape [10] some data on the binding energy of image
into the vaccum because'its own image potential. states on simple surfaces.
i.e.. the potential created by the polarization charge
it induces at the surface. In fact the Coulombic
tail of the image potential of l/z aIlows an in- 2. Phase shift model
finite Rydberg-type series to exist.
Hydrogenic states of this kind were studied in A simple way of understanding the physics
liquid helium [1]. in the case of a surface polaron involved in image states is provided by the phase

- [2]. and in metals were such states can arise if a


gap in the direction normal to the surface contains
the vacuum level.
Johnson and Smith [3] pointed out tha't these
states could be observed by angle-resolved inverse
shift model [11,\2]. This model has been widely
employed to give a transparent picture of surface
states wbich can be used either for precise numeri-
cal calculations. or for simple analytic approxima-
tions. In this picture the surface state is considered
photoemission. Inverse photoernission (or brems- as a wave trapped between the bulk crystal and
strahlung spectroscopy) [4,5] is well suited for . the surface barrier. This assumes a region of con-

probing bound stat~s directly by measuring the stant potential between the crystal and the surface
barrier but since tbis need only be infinitesimally
wide. this places no restriction on the arguments
* 1nvited lecture presented by P.M. Echenique. we are about to make.

0039-6028/91/$03.501) 1991 - Els.:vier Science Publishers B.V. (North-Holland)

:1
¡

¡
126 P.M. Echenique. M.E. Uranga / Image pOlenlialslales al surfaces

Thc
Summing the repeated scattering, taking into spending time on the crystal or the barrier side
sta tic ti
account all paths via the scattering formalism of [11]:
magnitu
Pendry and Gurman [13] the condition for a dlf>o
-o:: liquid h
surface state becomes dE .. s.de 1'1'12dz,
1barner close to
thus the
-ddEIf>co::
det{l- ReRo} =0, (1)
¡cryslal side
1
'1' 1
2 dz
' (4) ten ths (
away ( -
where I is the identity matrix and Re and R o are '1' being the wavefunction describing the elec- an elect
the crystal and barrier reflection matrices. tronic state. It is worthwhile to stress the point surface
A simple. approximation containing the essen- that variation of either If>cor If>ois an important and Eb
tial physics consists in reducing the problem to a element in producing a surface state: the greater = -0.8.
one-dimensional one; then the sum of the re- the variation the greater ihe chance of sweeping is reasol
peated scattering gives for the total amplitude of through the 27Tcondition. Thus Echenique and The (
the wavefunction at the surface Pendry referrred to a surface state induced prim- the crys
arily by a rapid variation of If>c as a crystal penmen
1 induced sta te, and that induced by a rapid varia- so-calleó
(2)
1 - rore ~i(4)B+4>c)' tion of If>oas a barrier induced state. We stress g?'-"':'~p.
the point that the two sorts of states are described tht gap,
where rco; If>coare the modulus and the phase of by condition (3). v~ fn
the crystal and the surface barrier reflectivity re- ga'fF.'íf t
spectively. The conditions for a surface state are gap, as
them 3. Binding energy metals,
eV/n2;
(3) A useful expression for If>Bis the WKB solu- proximat
rero=l, If>o + If>c= 27Tn.
- 21/z,
tion [14] for the classical image potential situation

221 metals q\
Because of flux conservation re and ro must be -0.85 e'
(5)
equal to unity. This means that the energy (E) lf>o(E)= ( V-2E ~1 ) 7T, for such
and momentum parallel to the surface (kp) of the where E is the energy measured from the vacuum more tigt
waves must be such that there are no propagating faces. Th
level, as E approaches the vacuum leve!, the phase
states within the crystal and no possibility of flux
varies very rapidly with energy. Many
escaping from the crystal into vacuum, i.e., a gap cated mI
In a real crystal, when image states exist, one
in the projected band structure and energy below the barri,
the vacuum leve!. has usually a band gap several electron volts wide
while the Rydberg series lies within an eV below results C0
This still leaves condition (3) for the phases to
be satisfied and this will only happen for certain
the vacuum level (21 = ~ in this case) therefore discussed
because of the rapid variation of the barrier phase o ¡th
values of energy and momentum. In particular the treII..~ th
reflectivity we can take 1>e as a constant and
crystal phase If>ewill vary at most by 7Tfor the
obtain for the binding energy ener:A
energy values we are concerned with here. Clearly surfa~1
if this would also hold for the barrier phase reflec- _ 2
tivity, condition (3) on the phases will have one or Eb = _ 2\ (6) applied ti
2(n+a)'; 11=1.2..... gaps asso
no solutions (this is actually the case for a step
where not paraIl
model of the surface barrier). Ir If>o,however, We sh(
varies rapidly with energy, as is the case for the 1 Es- 1 review 01
image potential, there is a good chance that many 2\ = "4 Es + 1'
(7)
solutions can be found. importan,
model re;
The variation of the phase with energy is di- (8)
a = i (1- ~e). systematil
rectly related to the probability of the electron


.,

I
'-- - - -
- ---

P.M. Echenique, M.E. Vranga / Image potential states at surfaces m


side The coupling constant Z\ and therefore the both crystal induced and image potential induced
static dielectric constant Es fixes the order of surface states. For accurate quantitative calcula-
magnitude of the binding energy. In the case of tions it is well established [31] that surface-state
liquid helium the static dielectric constant is very energies (and the work-function) are correctly re-
close to unity (1.0572 for 4He and 1.0428 for 3He), produced only in calculations which have been
thus the binding energy is of the order of a few performed self-consistently. The scattering for-
(4) tenths of meV (- 3.73/n2 meV); the state is far malism of Pendry and co-workers [13), or equiv-
away (- 100 Á) from the surface. For the case of alent ones, can still be used for accurate calcula-
elec- an electron attracted by its image potential to the tions via eq. (1). We insist, however, that the
Joint surface of an ionic crystal, such as LiF Es= 8.65 importance of the one-dimensional phase-shift
rtan t and Eb:::: -0.5/n2 eVo For metals Es= - 00; Eb model is to provide a simple explanation of sys-
eater = -0.85/n2 eVoIn the last two cases the electro n tematic trends. Even within this simple model
:ping is reasonably close to the surface - 2.3Á. some further assumptions have to be made to
and The details of the binding are fixed by (])c,i.e., justify the use of a two-band nearly-free-electron
>nm- the crystal scattering. In metals and in most ex- model, outside the gap region (Au(111» [33), or
)'stal perimental cases the image states appear in the for cases in which the states at the gap are pre-
ana- so-called Shockley inverted gap case, where the dorninantly d-like. Perhaps it is not surprising that
.tress gap is p-like at the bottom and s-like at the top of the biggests discrepancies reported between the
ribe~ the gap, then a is greater than zero since (])c simple phase shift model (with a nearly free elec-

e variés from O at the bottom to 17at the top of the


gap. Ir the vacuum level occurs at the top of the
gap, as it occurs in the [111] faces of some fcc
tron (])c) and the experimental values have been,
in the n = 1 state of Au [27,31) where the state is
nearly 2 eV above the top of the gap; and in
metal s, then (])c::::17, a = O and Eb = -0.85 Ni(lll) [34-36] where the states at the bottom of
eV/n2; while if the vacuum level occurs ap- the gap are predominantly d-like. As pointed out
solu- proximately at the rniddle of the gap, such a by Steinmann [15] even if a better calculation of
?:l/Z, situation arises in the [100] faces of the same fcc (])c could resol ve the disagreement about the n =1
metals quoted above, (])c= 117,a = 0.25 and Eb = state, it is questionable whether it could account
(5) -0.85 eV/(n + 0.25)2. Therefore one expects that for the binding energy of the n = 2 state, which is
for such fcc "metals the image states would be higher than the hydrogenic value. A careful analy-
:uum more tigthly bound in the [111] than in the [100] sis of the positioos of the image plane rnight have
Jhase faces. This has been found experimentally [6,7,15). some bearing on these cases since one would ex-
Many more calculations using more sophisti- pect that the d-baod character will tend to dis-
.one cated models both for the crystal scattering and place the ceotroid of the induced charge density
wide the barrier one have been presented [16,30). Their towards the material compared with the results of
>elow results corrobora te the basics of the physical model the jellium model. A similar conclusion has been
-efore discussed above. reached by Pitarke et al. [37) io their analysis of
:>hase Srnith [31) has presented an elegant study of the the influence of the shape and image potential 00
_ ar'-..- trends that should be expected for the binding the oscillations of current versus voltage in STM.
~ energies and effective masses for a variety of
~ surface states on different surfaces. He has also
(6) applied the phase shift formalism to the case of 4. Effective mass
gaps associated with a reciprocal lattice vector g
not parallel to the surface normal. lo the preceding section we have coocerned
We should hke to emphasise as stressed in the mainly with surface slates at the bottom of the
(7) review of Echenique and Pendry [32) that the surface band, and \Ve have nol takeo into account
importance of the one-dimensional phase shift aoy possible variation (other Ihan the implicit tk;
model relies on the correct description of the free-electron-like one) with the direction parallel
(8) systematics and trends of the binding energies of to the surface. Within the one-cleclron picture of a

A
.

I
128 P.M. Echenique. M.E. Uran/{a / Ima/{e potential.<tates at surfaces

solid the dispersion of Eh(kp) can only be due to effective mass through the shift in potential energy these tI
a scattering with the ion cores of the surface, i.e., 8V( z). allows
variation of the crystal reflectivity with k p or a momen
8V(z)k;
variation of the dynamic image potential with the V(z. kp) = V(z, O) + '). (9) leading
electron parallel velocity. such el'
A rough estimate of the effective mass correc- for a rH
tion (m * = 1 + 8m) can then be obtained by using of Ih.: (
4.1. Crysta/ effects the undispersed surface plasmon to represent the value o
surface response and taking the value of 8 V( z) at for Ihe
Echenique and Pendry suggested that for small the position of maximum probability density possiblt
kp the effective mass should be very close to 8V( zmax)' This gives, for a hydrogenic wavefunc- centre e
unity. The qualitative reason is that even for the tion to represent the first image state at 2.5%
crystal
most tightly bound Rydberg states the wavefunc- correction for rs = 2.67 (Cu) and 3.5% correction
rapidly
tion lies mainly in the vacuum, and the electron for r, = 3 [40,41]. certain
moves in a regíon with no potential variation The use of a more adequate surface response broader
parallel to the surface. In variance with this idea function including particle-hole excitation does governe
Garcia. et al. [38] suggested that image-potential not change the above conclusions [22,42]. Analo-
absorpt
surface states are strongly affected by surface cor- gous result is obtained when explicit inclusion of
rugations. This implies a relation between E h and the wavefunction penetration into the crystal is
8E a:
effective mass m* which does not agree with the taken into account [22].
accurate two-photon photoemission measurements ~J
of Giesen et al. [7,34] showing that the binding Therefo
energy is independent of m* within error bars of 5. Lifetimes pace WI
- 0.03 eVoMoreover the [100] surfaces orientation n. AII n
exhibits a more pronounced surface corrugation So far we have concentrated on the binding are in ¡:
but as mentioned above the binding energy of energy and effective mass of the sta te. and we member
image states in such surface is less than in the have assumed that the modulus of the crystal find ag
[111]. reflectivity re is unity. In fact flux can escape wavefur
A multiple scattering calculation for the energy from the state. The electron in such a state is not crystal ,
infinitely long-lived but can in principIe, make a fected \
dispersion of the first image state ~inding energy
in the [100] silver surface performed by Pendry et Iransition to other states, i.e.. to an empty crystal r
al. [39] gíves an effective electron mass equal to current-carrying state of the crystal. For finite l/S. A t:
unity. When effective masses are defined outside momentum parallel to the surface additional in-
the region of small kp' significant deviations from elastic channels are present !eading to an extra 8E a:-
unity appear: Giesen et al. have explained [7] broadening. In principie one could think that in
satisfactorily the experimentally observed effective metals electron-electron interaction w0uld result
masses using a one-dimensional phase-shift model in decay processes that broaden surface states to .-.....
with explicit inclusion of the shrinking width of resonances. This interaction may be so se\"ere dS to
the gap and the behaviour approaching band destroy all but the lowest member of Ihe Rydherg
edges. A comparison between the results of the senes. r~.
by b:fie
simple phase shift model and the experimental As the energy normal to the surface. ap- separale
ones has recently presented by Steinmann [15]. proaches the vacuum leve!. Ihe stales get closer in the ~ta:,
energy6.EJ. a: l/n} but at the same time the centre the ener
of gravity of the resonant sta te charge distrihution Ihe initi
4.2. Dynamica/ image potentia/ effects moves further away from the surface, having less the em~
probability of coupling to surface excilations. When ti
From the definition of local effective potential hence the energy broadening of the levels becomes lO Ihe ':
one could for operational purposes define a local smaller. Echenique and Pendry [11] proved th:1I wiJth .!

...
...

I
--- --

P.M. Echenique, M.E. (jranga (/muge PQlenlial Slales al surfa('es


129

:rgy these two effects compensa te in such a way that state decays into a state localized in the surface
allows resonant levels to be resolved. For the having the same z-dependent wavefunction but
moment we shall not discuss in detail the processes with a different k-parallel component (intraband
(9) leading to an energy broadening, and represent transition). The results of two calculations of the
such effects by an absorptive potential 1:;, which energy broadening r(1); r(2) for the two processes
rec- for a bulk free electron of the same energy can be described above are given in ref. [43.44] and in the
,;Ing of the order of 0.5-1 eV (depending on the actual review of Echenique and Pendry [32].
the value of the energy). ClearIy if these values held Typical values of model calculations of Pl)
.) at for the image states, no Rydberg series will be and r(2), with no wavefunction penetration are of
1sity possible. What saves the day is the location of the the order of 5 meV for the case of a silver surface
unc- [43]. When penetration of the wavefunction is
centre of gravity of the surface state outside the
2.5% taken into account the halfwidth increases, while
crystal because' these inelastic' processes decay
;tion the one associated to the momentum parallel re-
rapidly in strength outside the surface. Beyond a
certain range. zC' they can be neglected, and the mains of the same order. For a 30% penetration of
onse broadening, SE.J. of an image state wiII be the wavefunction de Andres et al. [44] quote 33
does meV for !r(1) in Ag, which is about a factor 10
governed by its overIap with this region of finite
1alo- absorption: greater than the value obtained with no wavefunc-
n of tion penetration. We should like to emphasize that
al iSI ZC 2 a4lB 1 in spit(: of their complexity the calculations of the
j
SE .J.'a:. -00 I'1'I dz a:. aE a:.3"
n
(10)
lifetime of image states use very simplified models
~
~ for the surface response that might not be very
Therefore the broadening scales as 1/n3 and keeps adequate tó describe appropriately the response
pace with the leve! separation in the limit of large function of materials such as Cu or Ag. The
n. All resonances converging on a given threshold calculations [43,44] use simplified models for the
Iding are in principIe resolvable. If we can see the first surface response, based either on the specular
j we members of a series the rest should follow. We model of Ritchie and Marusak [45] for the surface
)'stal find again the same idea. the Rydberg state's response, together with the small w expansion fo
;cape wavefunctions are effectively decoupled from the the bulk RPA dielectric response function of Pers-
s not crystal wavefunctions and therefore would be af- son and Andersson [47] for the case of an electron
Ike a fected very little by scattering processes at the totally outside the crystal [43]. At the energies
mpty crystal region. This, of course, is exact only for high involved preliminary analysis [48] indicates that
finite n's. A typical value of SE.J. can be obtained: the use of the experimentally derived undispersed
11 in- response functlon may lead to energy widths.
-1: a4lclaE
I
extra greater by a factor of 2 ar more than the ones
8E.J. a:. a4lB/aE
at in obtained using the respon se functions quoted
result -1:¡ 50 meV above. Recent experimental data by Schoenlein et
¡es to (11 ) al. [8] and Nielsen et al. [49] on Ag(lOO) .have
32(gap width)n3 n3
as l, measured a lifetime of about 25::!: 10 fs [8] or
~ .g More detailed calculations have been presented 20::!:5 fs [49]. These correspond to a line width of
~.

by Echenique. Flores and co-workers [43,44]. They about 35::!:la meV which is more than twice the
. ap- separate two contributions. First they calcmate for calculated' value with no wave-penetration [43].
ser ID the state at the bottom of the surface state band Nielsen et al. [49] and Steinmann [15] have sug-
;en tre the energy broadening related to transitions from gested that the discrepancy is presumably due to
)ution the initial image state to a final state defined by an additional decay channel, e.g., surface imper-
g less the empty. current-carrying states of the metal. fections. This is .an interesting suggestion that
ltions, When the Rydberg states have momentum parallel should be explored. but also one should take into
::omes to the surface, an extra contribution to the energy account the abbve comments on the wavefunction
j that width arises from a process in which the initial penetration and on the innuence of the different

~
,

I
I
130 P.M. Echenique. M.E. Uranga / /mage po/ential sta/es a/ surfaces

models of the response function. A calculation simplified model of Louis et al. [52], but because cient (e
along these lines is in progress [48]. the exponential dependence of the tunneling life- functiol
Very recently Schoenlein et al. [50] have re- time this should be a good value for any real another
ported the first time-resolved studies of higher situation.
phase i
order image potential states on Ag(lOO).
expericI
They measure the lifetime of the n = 2 image
state on Ag(100) to be - 180 fs, substantially 7. Two-dimensional electron gas
befort: .
high de
longer (- a factor of 7) than that of the n = 1
delay ir
state. This result is in agreement with the original Image states represent a very simple and funda- tor is c(
prediction of Echenique and Pendry [11] (see eqs. mental class of surface states which is useful in
surface
(14) and (15». The n3 scaling prediction is exact studying the two-dimensional (20) electron gas disordel
only for high n 's. More recent calculations by de [54]. To obtain a 20 Ferrni electro n gas a high of enco;
Andres et al. [51] using a many-body self-energy electron density is needed. In liquid helium the
whose p
formalism predict n3 lifetime scaling for states binding energy is weak (- meV) and an instability the elecl
n ~ 5. For lower order states, the model predicts a of the 20 electron gas develops when the density disorder
lifetime scaling less than n3, with is increased [55]. In metals the binding energy is of tum to
the order of 1 eV but a problem arises due to the
('Tn_2/'Tn= 1) =:::4. produce
processes that contribute to the energy width of w- res
The hydrogenic model of de Andres et al. as the image state. The interband transition (decay wOI1cers
stressed by Schoenlein et al. [50] overestimates into allowed bulk states) contribution to the en-
'Tn_1 since it neglects wavefunction penetration ergy width is high and comparable to the in- h~
m~~re
into the bulk. Thus one expects the ratio of life- traband (decay into image sÜltes having smaller
times 'Tn_2/'T"=1to be between four and eight, parallel momentum within the band) contribution. d4>B('
which is in good agreement with the experimental
measurements of Schoenlein et al. [50]. In the case
This means that electrons may not be able to ( dE
thermalize to the bottom of the 20 band because The con
of silver (and copper) (111) surfaces the n = 2 they decay into bulk states and the high density lifetimes
image state lies in the continuum and therefore needed to obtain a Ferrni electron gas would not tions th.
the surface state presents an elastic linewidth, be achieved. In insulators, such as diamond or For quit
originating from the lack of total reflectivity, Le., LiF, relatively strong binding appears (a large cut-off f
re < 1. This elastic half width for the n = 2 state of enough dielectric constant) and the escape outside measure
Cu and Ag has been estimated to be around 25 the band into buIk states is prevented by the insulatOl
meV [44]. electronic structure. Amau and Echenique [56]
have evaluated variationaIfy that the intraband
energy halfwidth is 150 meV for parallel momen- Aclrnowlt
6.STM tum of 0.15 (Á -1). They conclude that if one
The :
could populate the first image-state electron band
Louis et al. have presented an analysis of the Flores, J
up to parallel momentum of k p = 0.13 Á -1 then a . ~
conduction process through image surface states density of about 1013 cm - 2 could be a.ttained, Sl( an.
in scanning tunneling microscopy [52,53]. For low enough to think in terms of a20 Ferrni electron Rese3:il
electric fields the Auger inverse lifetime discussed gas [57]. ibert~'J.
above, becomes larger than the tunneling lifetime Guipuzk,
CAICYl
ando therefore, the tunneling process controls the
current while for large fields a saturation occurs, 8. Disordered surfaces
since the Auger process is not fast enough as to Referenc
allow new electrons to come from the tip into the Pendry [58] has suggested toat image states at
crystal, thus further increases of the field do not the surfaces of insulators may playa central role [1]W.T.~
lead to an increase in the tunneling current. The in the generation of 1/1 noise at the surfaces of M.("..
transition occurs at fields around 0.40 V/ Á in the these materials. The phase of the reflection coeffi- V.B.S

..

~
!,
¡.¡
~j

~
.

P.M. Echenique, M.E. Uranga I lmage po/en/ia} sta/es a/ surfaces Dt

cause [2] J. Sak, Phys. Rey. B 6 (1972) 3981.


cient (eq. (4» is related to the amount of wave-
~ life- function Iying inside the material. It has also [3] P.D. Jonhson and N.V. Srnith, Phys. Rey. B 27 (1983)
! real 2527.
another significance [58]: the derivative of the [4] V. Dose, W. Altmann, A. Goldmann, U. Kolac and J.
phase is also directly proportional to the delay Rogozik, Phys. Rey. Lett. 52 (1984) 1919.
experienced by a pulse of electrons, energy E, [5] D. Straub and F.J. Himpsel, Phys. Rey. Lett. 52 (1984)
before the reflection process is complete, thus, a 1922.
high density of states in the material will trap and [6] K. Giesen, F. Hage, F.J. Himpsel, H.J. Riess and W.
Steinmann, Phys. Rey. B 35 (1987) 971.
delay incident electrons. The surface of an insula- [7] K. Giesen, F. Hage, F.J. Himpsel, H.J. Riess, W. Stein.
unda-
tor is covered by a potentially conducting sheet of mann and N.V. Srnith Phys. Rey. B 35 (1987) 975.
fuI in surface states. As an electron runs around the [8] R.W. Shoenlein, J.G. Fujimoto, G.L. Eesley and T.W.
n gas . disordered surface, it will have a finite probability Copehart, Phys. Rey. Lett. 61 (1988) 2596.
lhigh of encountering a crystal reflection coefficient re [9] H.B. Nielsen, G. Brostrom and E. Matthias, Z. Phys. B 77
;n the (1989) 91. .
whose phase, <Perises sharply in energy, and then [10] H. Conrad, M.E. Cordesch, R. Scala and W. Stenzel, J.
lbility the electron will be trapped inside the solid in the Electron Spectrosc. Relat. Phenom. 38 (1986) 289.
ensity
y is of
to the
. disorder induced localized state to eventually re-
tum to the surface state bando This process will
[11] P.M. Echenique and J.B. Pendry, J. Phys. C 11 (1978)
2065;
P.M. Echenique, PhD Thesis Cambridge Uniyersity, 1976.
produce fluctuations in the electrostatic field and
lth of [12] E.G. McRae, Rey. Mod. Phys. 51 (1979) 541.
will result in noise. Pendry, Kirkman and co- [13] J.B. Pendry and S.J. Gurman, Surf. Sci 49 (1975) 87.
dec"
workers [59] have shown that the trapping states [14] E.G. McRae and M.L. Kane, Surf. Sci. 108 (1981) 435.
le have a characteristic distribution of lifetimes as
;i$ [15] W. Steinmann, Appl. Phys. A 49 (1989) 365.
le ._-:-
measured by [16] N.V. Srnith, Appl. Surf. Sci. 22 (1985) 349.
maller [17] M. Ortuño and P.M. Echenique, Phys. Rey. B 34 (1986)
jd<PB(E)d<PB(E+W)
)_.!.
ution. 5199.
( )
,le to \ dE dE w. 12 [18] P. Gies, J. Phys. C 19 (1986) L209.
[19] G. Borstel and G. Thomer, Surf. Sci. Rep. 8, (1988) 1.
cause The consequence of this distribution of trapping [20] N. Srnith, Rep. Prog. Phys. 51 (1988) 1227.
ensity lifetimes is that the trapped charge shows fluctua- [21] D. Straub and F.J. Himpsel, Phys. Rey. B 33 (1986) 2256.
Id not tions that have a "1/1 noise" power spectrum. [22] M. Weinert, S.L. Hulbert and P.D. Johnson, Phys. Rey.
nd or For quite modest thickness of insulator the inverse Let!. 55 (1985) 2055.
[23] M. Stéslicka, M. Zagórski and L. Jurczyszyn, Surf. Sci.
large cut-off frequency (wmin::=exp(-52L/2); 52 is a 200 (1988) 209.
.utside measure of disorder and L is the thickness of the [24] R.F. Garret and N.V. Srnith, Phys. Rey. B 33 (1986) 3740.
IY the insulator) will be of the' order of days or longer. [25] A. Goldmann, M. Donath, W. Altmann and V. Dose,
e [56] Phys. Rey. B 32 (1985) 837.
aband [26] G.D. Kubiac, Surf. Sci. 201 (1988) L475.
omen- Acknowledgements [27] N.V. Smith. c.T. Chen and M. Weinert. Phys. Rey. B 40
(1989) 7565.
if one
[28] C.T. Chen and N.V. Smith, Phys. Rey. B 40 (1989) 7487.
1 band . The authors gratefully thank Professors F.
[29] A. Goldmann, V. Dose and B. Borste!, Phys. Rey. B 32
then a Flores, J.B. Pendry, and R.H. Ritchie for discus- (1985) 1971.
aim sions and Iberduero S.A. for help and support. [30] Z. Lenac, M. Sunjic, M. Conrad and M.E. Kordesch,
n Research sponsored jointly by Euskal Herriko Un- Phys. Rey. B 36 (1987) 9500.
ibertsitatea, Hezkuntza Saila, Eusko Jaurlaritza, [31] N.V. Smith, Phys. Rey. B 32 (~8S).3549.
[32] P.M. Echenique and J.B. Pendry, Prog. Surf. Sci. 32 (1989)
Guipuzkoako Foru Aldundia and the ~Spanish 111.
CAICYT.
[33] D.P. Woodruff. W.A. Royer and N.V. Smith. Phys. Rey. B
34 (1986) 764.
[34] K. Giesen. F. Hage, F.J. Himpsel, H.J. Riess and W.
References Steinmann, Phys. Rey. B 33 (1986) 5241.
ltes at
[35] K. Giesen, DoctOral Thesis, München. 1986.
al role [1] W.T. Sommer, PhD Thesis, Stanford Uniyersity, 1964; [36] F. Hage, Doctoral Thesis. München, 1986.
lces of M. Cole, Phys. Rey. B 2 (1970) 4239; [37] J.M. Pitarke. F. Flores and P.M. Echenique, Surf. Sci. 234
coeffi- V.B. Shikin, SOy. Phys. JETP 31 (1970) 936. (1990) 1.

lO
~

I
132 P.M. Echenique. M.E. Uranga / ¡mage pOlen/ial s/ales al surfaces

[381 N. Garcia. B. Reihl. K.H Frank and A.R. Williams. Phys. [51) P.L. de Andres, P.M. Echenique and F. Flores. Phys. Rev. ",,'~"'"
Rev. Lett. 54 (1985) 59\. B 39 (1989) 10356.
[39] J.B. Pendry. CG. Larsson and P.M. Echenique. Surf. Sci. [52] E. Louis, F. Flores and P.M. Eehenique. Solid State
166 (1986) 68. Commun. 59 (1986) 453.
[40) P.M. Echenique and J.B. Pendry. J. Phys. C 19 (1986) [53] E. Louis, F. Flores and P.M. Eehenique, Phys. Ser. 37
5437. (1988) 359.
[41] G. Borstel and V. Dose. unpublished work. [54] T. Ando, A.B. Fowler and F. Stern. Rev. Mod. Phys. 54
[42] P.M. Echenique and J.B. Pendry, Surf. Sci. 166 (1986) 69. (1982) 437.
[43] P.M. Eehenique, F. Flores and F. Sois, Phys. Rev. Lett. 55 [55] L.P. Gor'kov and D.M. Chernikova, Pis'ma Zh. Eksp.
(1985) 2348. Teor. Fiz. 18 (1973) 119 [JEPT 18 (1973) 68].
[44] P.L. de Andres, P.M. Eehenique and F. Flores. Phys. Rev. [56] A. Arnau and P.M. Eehenique. Phys. Rev. B 38 (1988)
B 35 (1987) 4529. 10897.
(45) R.H. Ritehie and A.L. Marusak. Surf. Sei. 4 (1966) 234. [57] K. Giesen, F. Hage, H.J. Riess, W. Steinmann. R. Haight,
[46] F. Gareia Moliner and F. Flores, 1ntroduetion to the R. Beigang, R. Dreyfus, Ph. Avouris and F.J. Himpsel,
Theory of Solid Surfaces (Cambridge University Press. Phys. Ser. 35 (1987) 578.
Cambridge, 1979). [58] J.B. Pendry, in: Proe. 10th Werner Brandt Workshop on
J
[47] B.N.J. Persson and S. Andersson, Phys. Rev. B 29 (1984) Penetration Phenomena. ORNL-CONF 870155 (1988) p. '1
4382. 107.
(48) P.M. Echenique and A. Rivaeoba. to be published. [59] J.B. Pendry. P.D. Kirkman and E. Casta no, Phys. Rev.
[49] H.B. Nielsen, G. Bostrom and E. Mathias, Z. Phys. B 77 Lett. 57 (1986) 2983.
(1989) 91.
[50] R.W. Sehoenlein, J.G.. Fujimoto, G.L. Eesley and T.W.
Capehart. Phys.'Rev. B 41 (1990) 5436.

----..

~,::~

,
J

..
"!

----

You might also like