...:.:.:.

~~~~~~,~,~,,
::::::,
I.,;;
,....:.:,)
,:,:.
~ _
:.:>;::::~~,;~,:~y+:
e ..... ........
‘.1’
I.... -. ‘,~:::~
..::
..~.‘.~:~‘:...~.:.:.:....:...:.:.::.:.:.:~.:........~
““‘.“““....~‘.,.~ .... . . .I . . ...~.. .. .. ...‘.............~.~....~~.
. . .. . .. . . . . . .. .

g;::::...

surface scienc6
x.:..>...
_ ‘::.1.
..i........a
:+? .(/ .*.....:.:‘:j$$.+::
:,:.:~::::::~:::::~~.~:::::?
,.,...
,,,,..
““‘.f,‘.‘.:..,.,.:..,...
.,.. “,..
‘“‘%. _.i..
i......:
‘.““.%W
...A ........:,:,:,:,..,.,
...n. :,:,: ,/,,
;,,,:. ,,,,
.,_:,,,,,
.........._...~ :,:, ),,,,,,
““:.~.:~;~~:::~::::::~::::::~:~:y,.;..:‘
/

ElSEVIER Surface Science 306 (1994) 427-433

Diffusion on stepped surfaces

II. Step attraction and adsorbate-adsorbatc repulsion
C. Uebing *
Institut fir Festk%perforschung, Forschungszentrum Jiilich, GrnbH, 52425 Jiilich, Germany

R. Gomer
Department of Chemtitry and James Franck Institute, The University of Chicago, Chicago, IL 60637 USA

(Received 2 August 1993; accepted for publication 15 November 1993)

Abstract

Diffusion of adparticles on stepped surfaces has been investigated via Monte Carlo simulations, with step sites
assumed to exert extra binding energy, Jste,. In addition repulsive nearest-neighbor interactions were added, either
JNN = -Jstep, or J,, = - i J,,,,. These interactions were also assumed to be present between nearest-neighbor step
sites and step and nearest-neighbor terrace sites. For both cases J,, largely nullifies trapping by steps and greatly
reduces diffusion anisotropy at all coverages and temperatures. However fairly subtle effects balancing NN repulsion
and step attraction remain and affect the temperature and coverage behavior of diffusion.

1. Introduction diffusion anisotropy, except at low temperature

where steps were nearly saturated, thereby con-
In paper I El] we presented results of Monte stituting a barrier for diffusion across them but
Carlo simulations for the diffusion of adparticles not parallel to them. At high coverage this
on square lattices with evenly spaced steps. In anisotropy became pronounced at higher temper-
most of the simulations step sites were assumed atures. In the present paper we consider the
to exert extra binding energy, so that diffusion effect of nearest-neighbor adsorbate-adsorbate
from a step site required added activation energy. repulsion, assumed to be effective also within
A few simulations were also carried out for repul- step rows. The results show some initially unex-
sive steps, i.e. those requiring extra energy to be pected features, which however can be readily
crossed. In both cases absence of ad-ad interac- understood.
tions was assumed. The results for attractive steps
indicated that at low coverage steps intercepted
atoms in such a way as to lead to very little 2. Method

The procedures used were identical to those

* Corresponding author. described briefly in Ref. Ill and in more detail in

0039-6028/94/$07.00 0 1994 Elsevier Science B.V. All rights reserved

SSDI 0039-6028(93)E0998-A
428 C. Uebing, R. Gomer /Surface Science 306 (1994) 427-433

a previous paper [2]. As in paper I 64 x 64 square

lattices with periodic boundary conditions and
steps every fourth row were used. Step coverages
were determined both in the canonical and grand
canonical ensembles. In the latter case a large
number of runs giving overall 8 as functions of p
and T were found from which values of ester at
fixed 0 could be obtained as function of reduced
temperature.
Simulations were carried out for two cases, (I)
J NN = -Jstep and (II) J,, = - +J,,,,. Here J,,
stands for nearest-neighbor repulsion and Jst,
for step attraction, the positive sign signifying
attraction #l. Temperatures were expressed in Surface Coverage
terms of k,T/I J,, 1 in both cases, even though Fig. 1. Phase diagram for adsorption on a simple square
this changes the scale relative to Jstep. The reason lattice with nearest-neighbor repulsions; T, is the critical
for this convention is that for purely repulsive NN temperature (after Ref. [3]).
interactions the phase diagram, shown in Fig. 1
has a critical temperature given by [3]
scale is identical to that used in paper I, while in
k,T,/I J,, 1 = 0.567 (1) case (II) T is reduced by a factor of 2 relative to
and it seemed worthwhile to relate T to T, in a this.
simple way. Thus for case (I) the temperature

3. Step coverage
#’ The use of the symbol J for interactions is not universal
but has been used by us in previous work [2] and we retain it It will be useful to start with step and terrace
here. populations as function of I J,, I/k,T, shown in

1 B=0.8 1
0.9
0.8
0.7

IJNNI~~ IJNNII~BT
Fig. 2. Row coverage l9, versus 1JNNI/k,T for various Orotalindicated on the figures. (a) J,, = -Jstep. (b) J,, = - ; Jstep. (0)
denotes the step coverage, (0) represents the rows adjacent to steps, and A gives the coverage in the middle of terraces.
 C. Uebing, R. Gomer / Surface Science 306 (1994) 427-433 429

Figs. 2a and 2b. We recall also that in the ab-

sence of steps the phase diagram, shown in Fig. 1
consists of disordered lattice gas, a region of
c(2 x 2) (with vacancies) extending from 0.4 G 8
Q 0.6 below T, and “lattice fluid” for 8 > 0.6 at
all T. The line separating lattice gas or fluid from
the c(2 x 2) region corresponds to second-order
transitions. Fig. 2 shows that at low and high 8,
8step increases with decreasing T, reaching limit-
ing values at 1JNN I/k,T a 2. OS,,, is in all cases
higher than overall 8, but considerably less than
in the absence of NN repulsions [l], for the
obvious reason that adjacent filled step sites re-
duce step binding energy from J,_ to J,,,r 0 0.2 0.4 0.6 0.8 1
- 1J,, 1 per adatom. This effect is obviously big-
ger for case (I) so that OS,,, in this case is consid- Coverage t9
erably less than in case (II). Concomitantly the
Fig. 3. Mean square fluctuation ((Shr)2)/(N> versus total
population of terrace atoms decreases because of coverage for .I,, = - Jstep (upper panel) and J,, = - $ Jstep
conservation of adpopulation. However, because (lower panel): (0) (J,,j/kBT= 3.44 and (0) IJ,, I/k,T =
of NN repulsion this decrease is more pro- 1.60. The dotted line represents the high temperature limit
nounced for atoms in rows adjacent to steps. (Langmuir gas) ((SAV>/(N> = 1- 0.
For 8 = 0.5 the situation is more complicated.
As T decreases from high values ester, increases at steps. This can be seen also from equilibrium
first but as T approaches T, 8,,,, decreases, even- “snapshots” (not shown) and from plots of
tually approaching overall 8 again. The reason is ((Slv>*)/(ZV) versus t9, shown in Fig. 3. For
that the ~$2 x 2) structure dominates at low tem- both cases (I) and (II) there is a minimum at
peratures and thus imposes itself also on the 8 = 0.5, which becomes very sharp at low T. This

Fig. 4. Normalized tracer diffusion coefficients versus I J,, I/k,T for 0 indicated on the figures. (a) JNN = --Jo_. (b) J,, = - $
Jstep’ (0) D */Do;(+I 0:/D’; (X ) Of/D”. Do is the chemical diffusion coefficient for the non-interacting system without steps.
430 C. Uebing, R. Gomer / Surface Science 306 (1994) 427-433

behavior is typical of repulsive interactions, as negative activation energies result. For case (II)
shown by Fig. 6 of Ref. [2]. The reason is that where repulsion is weaker J,, dominates only at
((Shrj2)/(N> is proportional to the compress- lower T. A small positive activation energy occurs
ibility of the adlayer [2], which becomes very at high T where step population is quite small.
small in the c(2 x 2) region of the phase diagram, For high coverage, typified by the 8 = 0.8 results
because of the very high stability of this structure. steps are so highly populated as to constitute very
As already mentioned, this extends from 0 = 0.4 little energy barrier. Since the overall coverage is
to 0.6, precisely where the dip in ((Shrl*)/(N) already high they also do not affect diffusion
begins and ends. Since the structure is most nearly markedly via jump availability. For cases (I) and
perfect at 0 = 0.5 this is also the coverage where (II) the numerical D*/D” values are virtually
the minimum is seen. It is interesting that step identical to those for JsteP= 0, shown in Fig. 7 of
attraction in excess of NN repulsion attenuates Ref. [2]. For case (I) there is virtually no
this minimum somewhat. anisotropy; for case (II) diffusion across steps is
slightly more difficult than diffusion parallel to
them, for the obvious reason that the latter can
4. Tracer and jump diffusion coefficients occur with no step crossing at all. For 8 = 0.5 the
situation is more complex. At high T the step
We present next results for these quantities, population is in both cases 8,,,, > 0.5 so that
D* and Dj respectively, defined in Refs. [1,2]. steps are largely neutralized, both because of
Figs. 4a and 4b show D*/D” versus 1J,, I/k,T vacant sites in them and because diffusion from
for cases (I) and (II) for 8 = 0.2, 0.5, and 0.8. Dj step sites is made easier by the fact that an atom
(not shown) is virtually identical to D* and both entering a step row can leave it easily because of
will be discussed together. Two general facts stand NN repulsion. Thus at high TJ,, dominates and
out: there is very little diffusion anisotropy, and E * is negative, i.e. D*/D’ increases as T de-
except at 8 = 0.8 relatively little temperature de- creases. However below T, the c(2 X 2) structure
pendence. In order to understand these results dominates. It was found previously [2] that even
we note that the presence of NN repulsion serves for repulsive NN interactions only, a positive
to make steps less of a trap for diffusing atoms, activation energy, i.e. decreasing D*/D” with
first because two adjacent filled step sites reduce decreasing T occurred below T, because diffu-
the extra diffusion activation energy to - 1.I,, 1 sion requires breaking of the very stable c(2 x 2)
for case (I) and to 0 in case (II); NN neighbors in structure. Since we have seen that the latter even
rows adjacent to steps have the same effect. Thus controls step population it is not surprising that it
steps act as traps mainly under conditions where dominates diffusion as well.
they are largely empty, and then reduce
anisotropy, just as in the case of J,, = 0 by
affecting equally the perpendicular and parallel 5. Activation energies of D*
random walk components of atoms. At high step
coverage there will again be very little anisotropy Fig. 5 shows activation energies, E * relative to
but for the opposite reason: steps now do not the non-interacting case as function of coverage
constitute very much of a barrier. We should also for case (I) in units of J,,. Values are shown for
note that for repulsive NN interactions and J,,,, the high temperature regimes and also for
= 0 negative activation energies and values of J,,/kBT = 2, i.e. below T,. The results for case
D*/D’ > 1 were found in Ref. [2], because J,, (II) (not shown) are very similar, indicating that
< 0 facilitates jumps relative to the non-inter- E * is dominated by J,, rather than by J_,. The
acting case represented by Do. nearly linear relationship at high T is similar to
We are now able to discuss the results in that found for NN repulsion only without steps at
detail. Starting with 8 = 0.2 we note that for case high T [2]. The positive peaks at 0 = 0.5 corre-
(I) the effect of J,, dominates at all T so that spond, as already pointed out to the need of
 C. Uebing, R. Gorner / Surface Science 306 (1994) 427-433 431

4 0 0.2 0.4 0.6 0.8 1

Covemge 0
Fig. 5. Normalized activation energies for tracer diffusion
tE, = E * - I?‘)/[ J,, 1versus total coverage etotal for J,, =
- Jaw (0) En,,= EAI for IJNNI/kBT-+O; (+) EAI for
I~,,I/ka~=2;and(x) EA,, forJ,,,,/k$‘=S.
1 2

disrupting the c(2 x 2) structure. It shouId ako be lJN~llk~T

noted that E * shows very little anisotropy, for Fig. 7. timparison of D, /Do (thin symbols, dotted line) and
the reasons already mentioned. D,, /Do (thick symbols, solid line) Btotal= 0.2. (0) l&/D”,
D,/D’; t+) D,, /Do- DKGl/D’; (~1 D,,,,‘D”,
D~q /Do. JNN= - Jstetep.
6. Chemical diffiwion coefficient
El1 and labeled D,,, and DFI respectively as
Figs. 6a and 6b show the chemical diffusion function of 1J,, j/k,T for 8 = 0.2, 0.5, and 0.8.
coefficients obtained by the fluctuation method Comparison of these-results with D Ko, the diffu-

b 1000
100
10
1

10

0 1 2

1JNNtlrcBT iJNNjlkBT
Fig. 6. Normalized chemical diffusion coefficients obtained with the fluctuation method, D,, versus )JNN l/k&f for8 indicated
on
the figures. (a) JNN = -J_; (b) JNN = - 1 Jstep: (of D,/D’; (+> D, I/Do; (X) DpII/Do. D, obtained from a square probe,
D FL and D,,, from the slit probe.
432 C. Uebing, R. Gomer / Surface Science 306 (1994) 427-433

D ko is good in all cases for JNN/kGBT d 1.2 but

below these temperatures all components of D,
fall signifi~a~~y below D,. At B = OS (not
shown) there is good agreement between I), and
D xo for cases (I) and (II) over the entire temper-
ature range probed. At 0 = 0.8 (Fig. 8) there is
good agreement for case (I) for all components
over the entire range. This is also true for case
(II) except for DrL at J,,/k,T a 1.2 where
D r I is smaller than D, I . The reason for this
deviation may have to do with the breakdown of
the fluctuation method under conditions where
steps are trapping, for small probes as discussed
in Ref. 111.
One other point is worth noting. We have
E &(j already seen that the temperature dependences
a, 1 2
of D* and Dj are relatively small. Fig. 9 indi-
e lJNNI fkBy
cates that the temperature
((&V)*>/(N)
dependence
is considerably larger and corre-
of

Fig. 8. Comparison of &/Do (thin symbols, dotted line) and sponds to positive activation energy, consistent
DK,/DO (thick symbols, solid line) for 61total = 0.8. (0) also with the results of Fig. 3 of Ref. 121. The
D,/D’, D, /Do; (+) DFI /Do. D,, ,. /Do; (XI reason for this is that
DF,,& DKG,,/D'. JNN = - Jste,.
C(Shr)2)/(N) = k,T((%‘~)K (2)
where K is the compressibili~ of the adlayer.
sion coefficient obtained by combining r>ij and When repulsive NN interactions dominate K will
((~~)2)/(~), as discussed in Refs. [1,2f is shown increase with increasing T. Since
in Fig. 7 for case (I) and 0 = 0.2. Results for case
D = [tN)/(sN)2)]Dj (3)
(II) are very similar. Agreement between D, and

0 1 2 3 4

IJNNV~BT IJNNI PBT

Fig. 9. Mean square fluctuations ((SN)2>/( N > versus fJ,, Ik,T for different total coverages 8,,! as indicated: (a) JNN = -Jstep.
(b)J,N=-iJ * step.
 C. Web&, R. Gotner /Surface Science 306 (1994) 427-433 433

the temperature dependence of [((Slv>2)/ two sub-cases the situation would lie between
(N)]-’ dominates that D, accounting for nega- that reported here and encountered in Ref. [l].
tive activation energies of D (relative to Do> over We have also ignored the possibility of attrac-
the whole temperature range studied. tion by the bottom edges of steps (ledges) and
weaker binding at the terrace outer edges, as well
as the various possibilities of different activation
7. Conclusion energies along these directions (i.e. parallel to
steps) as discussed in section 1 of Ref. [l]. We
The simulations presented here indicate that hope to address some of these situations in future
NN repulsions affect diffusion on stepped sur- simulations.
faces drastically if such repulsions also exist within
step rows, since they then tend to decrease trap-
ping by steps. The compensation is of course 8. Acknowledgments
greatest for the somewhat artificial example of
equal step attraction and NN repulsion, but is This work was supported in part by NSF Grant
still quite appreciable when step attraction ex- CHE9222051. We have also benefited from the
ceeds NN repulsion by a factor of 2. We have not Materials Research Laboratory of the NSF at the
treated the case of NN repulsions on terraces but University of Chicago.
no NN repulsion within step rows. For this latter
case one may also envision either repulsion be-
tween step and adjacent terrace adsorbed atoms, 9. References
or no such interaction. In the latter case the
effect of steps would be essentially that found in 111C. Uebing and R. Gomer, Surf. Sci. 306 (1994) 419.
Ref. [l], namely strong anisotropy at high cover- (21 C. Uebing and R. Gamer, J. Chem. Phys. 95 (1991) 7626.
age and at low T at all coverages. For the first [3f K. Binder and D.P. Landau, Phys. Rev. B 21 (1980) 1941.