PHYSICAL REVIE% 8
15,
VOLUME
15 FEBRUARY 1977
NUMBER 4
Metal surface properties in the linear potential approximation
V. Sahni*
The New School
of Liberal Arts, Brooklyn College of the City University of
J. B. Krieger*
Department
of Physics, Brooklyn
College
of the
and
¹w
York, Brooklyn,
¹wYork 11210
J. Gruenebaum
City University
of
¹wYork, Brooklyn, ¹wYork 11210
(Received 29 June 1976)
Jellium metal surface properties including the dipole barrier, work function, and surface energy are obtained
in the linear-potential approximation to the effective potential at the surface. The metal surface position and
field strength are determined, respectively, by the requirement of overall charge neutrality and the constraint
theorem. The surface energies are obtained both
set on the electrostatic potential by the Budd-Vannimenus
within the local density approximation and by application of a sum rule due to Vannimenus and Budd, and
the two methods compared. The calculations are primarily analytic and all properties, with the exception of
the exchange-correlation energy, are given in terms of universal functions of the field strength, The effects of
for the
correlation on the various properties are studied by employing three different approximations
correlation energy per particle. The results obtained employing the Wigner expression for the correlation
energy closely approximate those of Lang and Kohn. The use of different correlation functions, however, leads
to only small differences in the results for the dipole barrier, work function, and the exchange-correlation
contribution to the energy, but the results for the total surface energy are significantly different.
I. INTRODUCTION
In this paper we present the results of a model
potential calculation of jellium metal' surface
properties. The principal advantage of such model
calculations" is the elimination of the requirement of a numerical solution of the Schrodinger
equation for particles moving in a self-consistently obtained effective potential which is intrinsic
Together
to other more complex formalisms.
with the application of certain theoretical constraints, which help to fully define the model.
effective potential, it is then possible to determine various properties of interest, such as
work functions and surface energies. Examples
of typical constraints applicable to model metal
surface calculations are the requirement of selfconsistency of the surface dipole barrier, ' the
condition on the electrostatic potential at the
metal surface set by the Budd-Vannimenus theorem' (BVT), and the Bayleigh-Ritz' energy minimum criteria, in addition to the charge neutrality
condition. ' The choice and number of these constraints to be satisfied mould, of course, depend
upon the complexity of the model potential employed.
%e consider here the linear potential model.
approximation to the effective potential at a metal
surface (see Fig. l). For a given value of the
field strength, the metal surface position is fixed
The field
by requiring overall charge neutrality.
strength may then be determined by application
of the BVT. The choice of the BVT criteria as
the constraint is governed by the fact that its
"
'"
application in our previous mork ' consistently
l.ed to good results for both the work function and
surface energy rather than, for example, just
the latter, as is the case' on application of the
variational principle for the energy. In Sec. II
we give a discussion of the calculations, and
definitions of properties and theorems employed.
The results for the surface dipole barrier, work
function, and surface energies are presented in
Sec. III. The surface energies are determined
both within the local-density approximation (LDA),
which is meaningful provided both exchange and
correlation are treated locally ii. u and by ap
plication of a sum rule due to Vannimenus and
Budd. The use of the Vannimenus-Budd theorem"
(VBT) appears particularly attractive, as it completely eliminates the requirement of determining
the individual components of the energy and any
questions regarding the accuracy of the I DA "'i4'5
and the effects of inclusion of the gradient terms.
These sets of results thus afford a comparison'
'
Y
eft
FERMI
P+
I
= Fx
efx)
EVEL
k3
=—
F
—k
3TT2
2 F
t
o
a
X
FIG. 1. Schematic representation of the linear potential model of a metal surface indicating all relevant energies. The hatched region represents the uniform
positive background beginning at the metal surface.
J. B.
V. SAHNI,
1942
KRIEGER,
of the two methods for obtaining metal surface
energies. In order to study the effects of correlation on the various surface properties, we
have also employed three different correlation
functions due to signer, Pines and Nozieres,
and Vashista and Singwi, '" in our calculations.
The correlation energy per particle as given by
these expressions can differ by as much as twenty
percent for low-density metals. Finally, we
compare our results obtained by empl. oying the
Wigner approximation for correlation with those
of the self-consistent calculations of Lang and
Kohn' (LK).
"
"
II. CALCULATION OF METAL SURFACE PROPERTIES
In this model calculation,
tial
V, «(x)
the effective poten-
"
strength defined in terms of
'k~z is
x» as E=kz/2xz, —,
and e(x) is the step function.
function g» (x) for x- 0 is
given as
(t, (x) = —(2/L)'~' sin[ kx+ 6(k)].
"
In the region x~ 0, the Schrodinger
the Airy differential equation'
(2)
equation is
xpr(x) dx.
&Q =4m
The electrostatic potential V„(x) required for
application of the BVT, VBT, and the electrostatic contribution to the surface energy is obtained by solution of Poisson's equation
"
d2V„ = —
4vpr(x),
,
(10)
x'
t'x
V„(x) = A(t) —4v
dx'
~co
dx"pr(x").
00
The surface energy of a metal E, which is the
energy required to split the crystal in two along a
plane may be obtained from the sum of the kinetic, electrostatic, exhange and correlation contributions. The kinetic energy E~ is given
(3)
(12)
where
whose solution is
ll»(g) = C»Ai(&)
&»i" = (k~»/160m)e,
(4)
where Ai(f) is the Airy function, C» is a normalization factor, g = (x-E/E)(2E)'(', and E is the
energy. The factor C, and the phase shift 6(k)
are determined by the requirement of the continuity of the wave function and its logarithmic
derivative at the origin. Thus,
C» = —(2/L)'(' sin6(k,
xz)[Ai(-
0=1k,
'k'
-,
Jl
)~,
Ai'(- g, )
.
'),
(6)
(
where go = (k»/k~+)(kzxz}'~',
and where
and
A
ff
P8i
ff
p
(x)=
(
0
(k»
—k')[g )»dk
pe
(14)
gy
E„ is
to the surface ener-
and the sum of the exchange and correlation contributions E„, which must be taken together within
the LDA is given as'
wc=
ky
Pe i
p&xV xdx
Ai'(1) is
the derivative of the Airy function.
The fundamental quantities from which all other
surface properties may be obtained are the electronic and total charge densities defined as
t
k'k(k)kk)
0
The electrostatic contribution
1
kg
kll(k)dk —
(6)
f»}]
and
(
(-
()) = 0.
0) = V,',
with the boundary conditions V,
Applying the charge neutrality condition, the
electrostatic potential may thus be written
as"
—fg» —
—0,
cot6(k, x~) =
(8)
(-
where
is the field
the slope parameter
the Fermi energy,
The electronic wave
d g»
15
respectively. Implicit in the definition of the
total charge density is the assumption that the
positive ions of the metal are smeared out and
replaced by a uniform charge background of den-kz/3)(' ending abruptly at the metal sursity p+ —
face position at & =a.
The surface dipole barrier contribution to the
work function is given by the expression
sumed to be
I
GRUENEBAUM
pr(x) = p, (x) —(k~z/3v')e(-x+a),
at a metal surface (see Fig. 1} is as-
V„~x) =Exe(x),
J.
AND
~ac pe &
—&gp
pe pg &
&&p
where &„, = &„+&~, &„, and &, are the average
exchange and correlation energies per particle
(16)
METAL SURFACE PROPERTIES IN THE LINEAR POTENTIAL. . .
for a uniform electron gas, and where p, =k'„/
3m' is the mean interior electronic density.
The expression for the derivative of the surface energy with respect to the signer-Seitz
radius r, which according to the VBT is given
as
dE
9
4~r4
dr,
.„
[ V., (- ~) —V„(x)]dx,
1948
1.5
1.0
0.5
(17)
to obtain the surface enerdirectly by integration over r, , together with
a suitable choice for the constant of integration.
In this model the two variable parameters are
the metal surface position a and the field strength
E. For a given E the metal surface position may
be determined either by the charge neutrality
may also be employed
—0.5
gy
~e3'
I'IG. 2. Plot of the universal function for the metal
surface position y, vs the slope parameter yz.
condition
pr(x)
dx=0,
(18)
till. it satisfies the BVT.
The work function of the metal
or by application of the phase-shift rule of Sugiyama,
'
according to which
—,
8k„k
g =—
3~
I
k6(k) dk.
(19)
The field strength, or equivalently, the slope
parameter x~ is adjusted so as to satisfy the requirement set on the electrostatic potential by
the BVT such that'
~V= V„(~) —V„(-")=p. dc~
d
(20)
&~ is the sum of the kinetic, exchange, and
correlation energies per particle for a uniform
electron gas.
—q,
With the transformation y =kzx, and k/kz —
such that the metal. surface position is now defined to be at y, =k~a, it can be shown that the
quantities y, , p, /k3z, AP/kz, V„/kz, E~/kz~, E„/
kz', and r~dE, /dr, are all universal functions of
the slope parameter y~ = k~x~. Furthermore, all
the spatial integrals in the expressions for the
metal surface position, the surface dipole barrier, the electrostatic potential, and the kinetic
Thus, together
energy can be done analytically.
with the electronic density, the determination of
these properties reduces to a simple numerical
calculation of k-space integrals ranging from 0
in the
to 1. The explicit expressions'employed
present calculations are given in the Appendix.
Plots of the universal functions y, , aP/k~,
160vE~/kz', E„/k~z, and (9v)r4dE, /dh, are given
in Figs. 2-6. The metal surface position, surface dipole barrier, and surface energies are
then easily determined for a specific metal [defined by its Fermi momentum kz --1/ar, ; o.
= (-, w)' '] by adjusting the slope parameter x~
where
C
is then given
as
4
=ay ——,'k2~ —p, „,,
(21)
where p. „, is the exchange and correlation part
of the chemical potential. of a uniform electron
gas defined as
Wxc
=
d(p, e„,)
(22)
The exchange energy" per particle for the uniform electron gas is e„= —0 458/r, . . For the
correlation energy per particle valid at metallic
densities we employ the correlation functions"
due to Wigner" (W), Pines and Nozihres" (PN),
and Vashista and Singwi" (VS),
e, = — .044/(r, +7.8),
c~" = 0.0155 in/, —0.0575,
= 0.016 t5 lnx,
—0.056.
0, 3
0.2
0, 1
I
I
I
I
1.0
2.0
3.0
4.0
FIG. 3. Plot of the universal function Aft)/kz, where
is the surface dipole barrier, vs the slope parameter yp.
4Q
V. SAHNI,
1944
J. B.
KRIEGER,
J.
AND
Sx1Q
GRUENEBAUM
-2
4TT
9
4x10
3x1Q
15
4
dEx
dr
( a. u.
-2
-2
-2
2x10
-2
1x10
-1 x10
FIG. 4. Plot of the universal function (160m/kz)E»
where E~ is the kinetic energy, vs the slope parameter
pp
~
FIG. 6. Plot of the universal function (T4 x)r ~ dE +dr,
where E~ is the surface energy, vs the slope parameter
pp
the VS and PN expressions are similar,
their respective values for &„. can differ by as
much as 15/p for r, =6, for which value of r, the
VS expression differs by 20% from that due to
Wigner. The results for the various properties
within the linear potential model employing these
different correlation functions are presented in
-2
~
Although
Sec. III.
III. RESULTS
A. Surface dipole barrier and work functions
In Table I we present the results for the surface dipole barrier and work function for the three
different correlation functions. For r, = 2-4, both
the BVT and charge neutrality are exactly satisfied, whereas for r, ~ 4. 5, no choice of parameters
in the linear potential model satisfy both these
Thus for r, ~ 4. 5, the present work
requirements.
corresponds to satisfying the BVT as closely as
11
x10
-4
9 x10
7x10
5x10
3x 10
1x10
YF
FIG. 5. Plot of the universal function E,~/&z, where
energy, vs the slope parameter yz.
E,~ is the electrostatic
possible in the l. imit of this model, i. e. , by a
potential of infinite slope, and by the charge neutr ality condition.
A comparison of the Wigner (W) results with
those of LK indicate that the model reproduces
the large dipole moments required for high-density metals, differing by 0.23, 0. 04, and 0. 1S eV
for r, =2. 0, 2. 5, and 3.0, respectively. The reason for the accuracy of these results is that by
adjusting ~ U so as to satisfy the BVT, one has
already obtained approximately 40% of the dipole
barrier, exactly. The remaining contribution
to ~Q from charge outside the metal is determined
accurately since the model permits a large electronic spillover. In addition, satisfaction of the
BVT also leads to very accurate electronic densities in this range. For r, =3.5 and 4. 0, the
model. becomes progressively more reflecting
as the potential rises more steeply leading to
underestimates for the dipole barrier. Also,
~U in this range comprises only about 25% of
the LK value. For r, ~ 4. 5, for which the contribution of the dipole barrier to the work function. is small, the results for AP are within 0.28
eV of those obtained by LK. In this range, the
LK results' also do not satisfy the BVT, although
their results more closely satisfy this condition
than do the results of the infinite barrier model. '
Thus over the entire metallic range the results
of this model. calculation for the dipole barrier
and work function are within 0.32 eV of those
due to LK.
The use of the PN and VS correlation functions
lead to results for &Q and 4 which are consistently I.ower than those obtained employing the
W function. The PN and VS results for ~Q are
within 0.03 eV of each other, the VS result dif-
'
METAL SURFACE PROPERTIES IN THE LINEAR POTENTIAL. . .
15
1945
TABLE I. Results for the surface dipole barrier b, p and work function C for the correlation functions due to Wigner
(W), Pines and Nozieres (PN) and Vashista and Singwi (VS). The values for the metal surface position y, and slope parameter y& quoted are those obtained using the Wigner expression for the correlation energy.
2.0
2.5
3.0
3.5
4, 0
4.5
5.0
5.5
6.0
Surface dipole barrier AP (eV)
Lang-Kohn
Present work
g =kgb
pp =kpxp
3.760
2.801
1.967
1.214
Present work
PN
VS
3.68
3.31
3.01
4.09
3.63
3.26
2.95
2.74
2.59
2.63
2.61
2. 58
2.68
2.57
2.60
2.58
2.54
4.02
3, 55
3.17
2.85
2.57
2.49
W
PN
VS
1.488
1.108
7.03
6.90
3.69
6.88
3.66
6.80
4.12
3.83
0.763
0.410
2.05
2.02
2. 32
1.18
0.53
1.09
0.50
0.28
0.25
0.23
0.21
1.43
0.91
0.56
0.35
0.16
3.79
2.13
-0.0370
0.507
0.000
0.000
0.000
0.000
Work function 4(eV)
Lang-Kohn
"
-1.178
—1.178
-1.178
—1.178
0.59
0.28
0.25
0.23
0.21
1.12
0, 28
0.25
0.23
0.21
W
W
0.04
2.51
2.49
2.45
'
(eV)
W
W
3.89
3.72
3.50
3.26
3.06
0.23
0.04
0.19
0.25
0.32
0.28
0.10
0.07
0.17
2.87
2.73
2.54
2.41
~See Ref. 5.
In Table II we present the results for the surface energy for the three different correlation
functions determined in the LDA. The parameters
employed are the same as those of Table I ob-
conditions, and the results given only for those
values of r, for which both these conditions are
exactly satisfied, viz, r, =2-4. The use of the
infinite barrier model for r, ~ 4. 5 leads to unrealistic results within the LDA, due primarily
to the sensitivity of this method of computing
energies to the vanishing of the el. ectronic density at the artificial barrier. ' Other methods
however, such as the VBT employing the same
model do lead to meaningful results for these
low densities, as discussed in Ref. 3. We have
also included in Table II the results for the infinite barrier model. for r, =2-4.
A study of Table II indicates that the results
for the total energy of the W column closely approximate those due to LK (see also Fig. 7). For
r, =2-3, the results for the individual componthe accuracy
ents are also close approximations,
of the electronic densities being particularly well
reflected in the results for the kinetic and ex-
tained by satisfying the BVT and charge neutrality
change-correlation
fering by at most 0. 15 eV from the W result. On
the other hand, it is the PN and W values for 4
which are more similar, being within 0.06 eV of
each other, the VS and W results differing by
a maximum of 0.17 eV. Thus these different correlation functions give rise to only small differences in the results for the dipole barrier and
work function. Since our results are close to
those of LK using the W correlation function, we
expect that a completely self-consistent calculation empl. oying the other correlation functions
would differ from LK by similarly small amounts,
i. e. , of the order of 0. 1 or 0.2 eV.
B. Surface energies-local density approximation
energies.
Since for
TABLE II. Surface energies in ergs/cm2 as obtained in the local density approximation for
the different correlation functions due to Wigner (W), Pines and Nozieres (PN), and Vashista
and Singwi (US). The individual component values correspond to the Wigner expression for
the correlation energy.
$
Surface energy components
(Wigner correlation function)
Present work
Surface energies
Infinite
&xc
2.0
2.5
3.0
3.5
4.0
Present work
Barrier
potential
—5897
-1861
-666
—236
-51
See Ref. 3.
See Ref. 5.
3386
1458
718
383
213
PN
1567
4077
-944
447
1832
964
565
359
44
205
206
59
23
Lang-Kohn"
~
&es
153
E,
185
—1008
-863
-843
80
223
218
201
92
36
231
199
194
158
224
208
r, =3.5
y.
~R1EgER, ,
SAHN&
GRUENEBA™
15
I.DA is q uite
' 'insensitive to cha n g es in the
ffective potenti ial in sharp conntrast
ra to the kinIn fact it can be s h ownn by expanding
EgC about the exacct density th a tth
th
calculated E„, is prroportional to thee error in the
calculated diip ol. e moment whic, in
'tt
the infinite barb
ermitting
me a ramp, the resu itss for the total
b obrastically altere,d a
energy are dra
o the infinite
served by comp
m aring the resu itss of
r otentials given in a e
The surface ene r g ies obtaine
bta e
S correlation functions m
' '
si nificantl.
rent from those
y di ff eren
the signer expressi
ress on These
rise primarily due too the numerica
differences in the values for th e k'i t'
lesser degree, from thee differences
energy. The exchange-corgV for both the PN an
in the
200
LD
E
at g/cm
-200
-400—
-600—
+ Lang
-1000 —+
—Kohn
.
'
on of the surface energie s obtained
nsity approximation (-A),
imenus-Budd dth
orem (VBT), emp oyt core
P lC
i ner expression for corre a lO .
ing the Wig
resent the results o an
repre
Wl
AN& J
l
-db. "-
and 4. 0, the efffective poten
e
t t iaaal rises too steeply,
or E„are larger an
the results for
n the corresp
an
sponding values
. , as was
obtained by I.K. However,
ow
e c
'nfinite barrier mo e s,
p
change-corre lation
a i energies for a v
s are considera, bl y more accura t e l.y obtained
ine i
t tic contribua
t
xam le, for r, =2. , „, i
s E and
nd are within
q o
lumn. Thus, althoug
e
'on functions affect
cor relation
ct thee total energy im&V and thus E~, their
ei explicit contribution
ri
taken togeether with exc an
mately the same.
- udd theorem
-Use oof thee Vannimenus-Budd
C. Surface energies-Use
tiou we present results (see
see Tab
Tablee
'th
tive of the sur face
) o
m"
r
as obtainedd bby the sum rule of VanThe parame
meters are again
t the BVT is ei'th er
tl
the case for r, &
d as p ossible, as y h
""Budd.
&„di er
4% respective p, from those o
I K. These result s to g ether wi th those of the
other models s tu died
i indicate that E as obtained
„a
"
ppo
sion for the derivative
of
e en
of th
fa
th respect to the Wi gn er-Seitz radius dE /dy
gy
and surface eneer gi es in ergs cm,
m as obtained
o aine by a licat'on o
employing the Wiigneer approxima t'ionn for correlation.
dE,
S
Present work
2.0
2.5
3.0
3.5
4.0
45
5.0
5.5
6.0
~See Hef.
13.
S
S
Lang-
Kohn ~
3365
669
4447
755
119
—9
—36
—34
—23
—15
—ll
79
—54
—71
-67
—34
—37
-33
Present work
—868
—41
123
143
130
111
97.5
Lang-Kohn
-980
49
197
191
157
122
97.5
88
80
82
62
METAL SURFACE PROPERTIES IN THE LINEAR POTENTIAL.
finite barrier model obtained from Eq. 17 is given
in Ref. 3, and a plot of dE, /dr, over the entire
metallic range is given in Fig. 8. The crosses
in the figure represent the values obtained by
Vannimenus and Budd employing the electrostatic potential. of LK. The surface energy is
determined by integration over r, with the constant of integration" chosen such that the energy
matches the LK value at r, =5. Although we could
determine this constant by a physical criterion
such as the t', ~ limit for which case the surface energy vanishes, we restrict ourselves to
the above choice in order to enable comparisons
with the work of Vannimenus and Budd. A graph
of the surface energies thus obtained is al. so
plotted in Fig. 7.
We note that for high densities, in contrasts
to both the step and infinite potential models, the
present model does permit dE, /dr, to become
positive and large. However, although the graph
for the derivative appears very similar to the
LK results, a comparison of the explicit values
given in Table III indicate them to be quite different. This is interesting because use of the
LDA within this model. with the same choice of
parameters leads to very accurate surface energies. Since the LDA method depends primarily
on the electronic density, and the VBT method
on the total charge density inside the metal, the
reason for the differences in the results of the
two methods becomes evident on a comparison
1000—
+
Lang-Kohn
600—
dES
drs
400—
200—
2.5
I
-)00-
30 3.5 4.0
I
I
4, 5
5.0
5.5
I
I
I
I
+
+
+
6.0
+
FIG. 8. Plot of the derivative of the surface energy
E~ with respect to the signer-Seitx radius &~ vs ~~
for the Wigner expression for correlation. The crosses
represent the results obtained by Vannimenus and Budd
employing the electrostatic potentials of Lallg and Kohn.
..
1947
of the electronic and total charge densities and
electrostatic potentials obtained with those of
LK. Inside the metal, for example, the electronic
density normalized with respect to the bulk value
for r, =2 is within 4% of the LK density, differing
1.2 to
by less than a percent in the range from
—0.275 Fermi wavelengths. On the other hand,
the total charge density near and at the metal
surface can differ by as much as 30%. The major
contribution to the integral of Eq. 1V for dE, /dr,
arises from a region extending from the surface
to approximately a third of a Fermi wavelength
inside the metal. . In this range, the results for
the electrostatic potential differs by (6-9)% from
those of LK. Furthermore, inside the metal,
however, the differences in V„are comparable
to those of the electronic density. Thus it is these
differences in total density, and hence V„ inside
the metal which lead to the differences ln the
derivative of the surface energy and to its integral, i.e. , the surface energy. The adjustment
of the electrostatic potential such that it corresponds to the exact value at one point, viz. ,
at the metal. surface, cannot be expected to lead
to precise values elsewhere. However, the results for the surface energy, though not as accurate as those of the LDA calculation, are fair
approximations to the LK values, and meaningful
over the entire metallic range (see Fig. 7). Thus
a study of this more accurate model reaffirms
the conclusion' that the VBT method for obtaining
surface energies is particularly sensitive to the
choice of effective potential.
We observe, in conclusion, that the linear potential model together with the constraint of the
sum rule due to Budd and Vannimenus leads to
results for all surface properties comparable
to those obtained by Lang and Kohn for jellium
metal. The fact that the effective potential does
not become constant but increases indefinitely
is unimportant, since the effective potential is
in substantial error only in the region far from
the metal. surface where the electron density has
exponentially decayed to a small fraction of its
value at the surface. The majority of the calculations, as shown in the Appendix, are primarily analytic„and the universal curves permit
a direct determination of surface properties on
application of any theoretical, constraint. Furthermore, due to the accuracy of the results and semianalytic nature of the density, it is reasonable
to employ this model in order to study the effects
of the gradient and higher-order terms on the
exchange and correlation energies, and such an
investigation is in progress. We also observe
that although the LDA and VBT methods for obtaining energies lead to good results in compari-
-
V. SAHXI,
J. 8.
J.
KRIKGER, AND
son with those of Lang and Kohn, the former method proves to be superior, and it would thus be
of interest to determine how the results of the
two methods compare on inclusion of the gradient
term contribution to the existing LDA value.
Finally, we note that use of different correlation
functions does not affect in any significant manner
either the work function or the exchange-correlation contribution to the surface energy, as
obtained via the LDA, but do affect considerably
the results for the total energy. This as explained
earlier is because the component of the energy
most sensitive to small variations in the effective
potential is the kinetic energy, and its contribution to the total energy for high and medium densities is always significant.
GRUENEBAUM
erties were obtained by use of the foll. owing integral expressions.
Ai2
df
g
&Ai'
=fbi' f -Ai"
—
g d&= ', g'Ai' g
(A7)
i"
—&A.
&
+xi (g) A~'(g)],
A.
i"
dg=-,'
g
-g'Ai'g +gAz"
+ 2xf
g
(g)Af''(g)].
(AS)
Metal surface position
Application of the charge neutrality
of Eq. j.8 leads to the expression
With the transformation y =kzx, 0/k~ =q, the
slope parameter is y„=k~x~, the metal surface
is at y, = k„a, and the variable g is
~=(y
f,
2)
(1
sin25(q,
condition
yp )
(A10)
(Al)
q »1 )yZ
present below expressions for the phase shift,
electronic density, metal surface position, surface dipole barrier, electrostatic potential, and
the kinetic energy in terms of universal functions
of the slope parameter y~.
'tI)%t'e
alternate expression is obtained from the
Sugiyama phase shift rule of Eq. 19:
An
—
Sm
y, = — —3
dq q5(q,
yz).
D
Surface dipole barrier
&0/&p = (4/3&)[d(yz)+&(yz)],
%lith the definition
where
1yS»(-q'4")
«(Syz)=ye
,,
«(0) =yg'
Aji(
2
ags)
~
»(0)— =-1.37172»g',
&.
—,
&(q, y~) =cot '[1/q~(
(A2)
2
3
1 ——y',2
J(y„) = —
4
3
y
(A4)
o
dq(1 —q') cos2~[qy+5(q, yz)].
y
~V, (y) = 4
3,
3s
d(y~)+ 8
+
dq(l —q') sin'5(q, y~)
~
&f'{{y-q'y
gg2(
+
)y
q2»2/3)
The universal. functions for the remaining
dq(1-q')qsin20(q, y&).
«0,
2
p~
1
3, y~- 2'
+ 2'
v. (y)
Q'
(A14)
For
j.
o
Electrostatic potential
«0
=1 ——,'
2
(1 —q')«'
dq—
1+K
{A13)
@y~) =
yq~)].
g
+ —«(0)+
(A3)
Electronic density
For
(A12)
prop-
3
(,
e{y.-y)+-'»{2». -y)e(y-y.
i
3
8
y
)
2, cos26{q, y~) sin qy
dq(l —q )
. -g , s'
dg(l
~
25(q, y )Hi 2qy)
Q'
(A15)
METAL SURFACE PROPERTIES IN THE LINEAR POTENTIAL. . .
For y~ 0
1949
Kinetic energy
V„(X)
hQ
k~
k~
+
g(i)
2
(7 —Vo) 8(X —S)
3w
k~4
= (1/160s)8,
(A1'7)
where
80
8 = 1+ —
dq q6(q, y~)
—,
1
x [ 2g'Ai'(t;) —21. Ai "(g )
0
A-i (g) Ai
'(g)l
.
of
York Faculty Research Program.
N. D. Lang, in Solid State Physics, Advances in Research and Applications, edited by H. Ehrenreich,
F. Seitz, and D. Turnbull (Academic, New York, 1973),
Vol. 28, p. 243.
V. Sahni, J. B. Krieger, and J. Gruenebaum, Phys.
Bev. B 12, 3503 (1975).
V. Sahni and J. Gruenebaum, Phys. Rev. B 15, 1929
New
(1977).
J.
W. Kohn and L.
Sham, Phys. Rev. 140, A1133 (1965).
5N. D. Lang and W. Kohn, Phys. Rev. B 1, 4555 (1970);
3, 1215 (1971). The self-consistency procedure has
been numerically refined by Lang, and the improved
results for the surface energy and dipole moments
have been quoted in the text.
H. F. Budd and J. Van». menus, Phys. Rev. Lett. 31,
1218 (1973); 31, 1430(E) (1973).
Variational Principles (lnterscience, New York, 1966), p. 153.
J. Bardeen, Phys. Bev. 49, 653 (1936).
W. J. Swiatecki, Proc. Phys. Soc. Lond. A 64, 236
(1951).
A. Sugj, yama, J. Phys. Soc. Jpn. 15, 965 (1960).
B. L. Moiseiwitsch,
E~"'l&z =(1/«')[&(Xz)h~l
(A16)
Supported in part by a grant from the City University
des &(ax )),
(A18)
(A19)
D. Lang and L. J. Sham, Solid State Commun. 17,
581 (1975).
D. C. Langreth and
P. Perdew, Solid State Commun.
17, 1425 {1975).
~3J. Vannimenus and H. F. Budd, Solid State Commun.
~~N.
J.
'
15, 1739 (1974).
J. Harris
and R. O. Jones, J. Phys. F 4, 1170 (1974).
'5E. Wikborg and J. E. Inglesfield, Solid State Commun.
16, 335 (1975).
D. Pines, Elementary Excitations in Solids (Benjamin,
New York, 1963), p. 94.
D. Pines and P. Nozieres, The Theory of Quantum
Liquids {Benjamin, New York, 1966), p. 330.
P. Vashista and K. S. Singwi, Phys. Rev. B 6, 875
(1972).
~~Atomic units are used: ~e = K= m =1. The unit of energy is 27.21 eV.
~
L. D. Landau
and E. M. Lifshitz, Quantum Mechanics
(Pergamon, New York, 1965), p. 71.
M. Abramowitz and I. A. Stegun, Handbook' of Mathematical Functions (Dover, New York, 1965), p. 446.
S. Raimes, 8'ave Mechanics of Electrons in Metals
(North-Holland,
Amsterdam, 1961), p. 177.