NATO Science Series A Series presenting the results ofscientific
meetings supportedunder the NATO Science Programme.
The Series is published by lOS Press, Amsterdam, and Kluwer
Academic Publishers in conjunction with the NATO Scientific Affairs
Division
Sub-Series
I. Life and Behavioural Sciences II. Mathematics, Physics and
Chemistry III. Computer and Systems SCience IV. Earth and
Environmental SCiences V. Science andTechnology Policy
lOS Press Kluwer Academic Publishers lOS Press Kluwar Academic
Publishers lOS Press
The NATO Science Series continues the series of books published
formerly as the NATO ASI Series.
The NATO Science Programme offers support for collaboration in
civil science between scientists of countries of the Euro-Atlantic
Partnership Council.The types of scientific meeting generally
supported are "Advanced Study Institutes" and "Advanced Research
Workshops", although other types of meeting are supported from time
to time. The NATO Science Series collects together the results of
these meetings. The meetings are co-organized bij scientists from
NATO countries and scientists from NATO's Partner countries -
countries of the CIS and Central and Eastern Europe.
Advanced Study Institutes are high-level tutorial courses offering
in-depth study of latest advances in afield. Advanced Research
Workshops are expert meetings aimed at critical assessment of a
field, and identification of directions for future action.
As a consequence of the restructuring of the NATO Science Programme
in 1999, the NATO Science Series has been re-organised and there
are currently Five Sub-series as noted above. Please consult the
following web sites for information on previous volumes published
in the Series, as well as details of earlier Sub-series.
http://www.nato.jntlscience htt~r//www.wkap nl
http:/twww.iospress.nl http://www.wtv-boolss.delnato-oco btm
I
-$- I
Atomistic Aspects of Epitaxial Growth
edited by
Miroslav Kotrla Institute of Physics, Academy of Sciences of the
Czech Republic, Prague, Czech Republic
Nicolas I. Papanicolaou Physics Department, Solid State Division,
University of loannina, loannina, Greece
Dimitri D. Vvedensky The Blackett Laboratory, Imperial College,
London, U.K.
and
LucT. Wille Department of Physics, Florida Atlantic University,
Boca Raton, U.S.A.
Springer-5cience+Business Media, B.V.
Proceedings of the NATO Advanced Research Workshop on Atomistic
Aspects of Epitaxial Growth Dassia, Corfu, Greece 25-30 June,
2001
A C.I.P. Catalogue record for this book is available from the
Ubrary of Congress.
ISBN 978-1-4020-0675-3 ISBN 978-94-010-0391-9 (eBook) DOI
10.1007/978-94-010-0391-9
Printed on acid-free paper
All Rights Reserved @ 2002 Springer Science+Business Media
Dordrecht Originally published by Kluwer Academic Publishers in
2002 No part of this work may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means, electronic,
mechanical, photocopying, microfilming, recording or otherwise,
without written permission from the Publisher, with the exception
of any material supplied specifically for the purpose of being
entered and executed on a computer system, for exclusive use by the
purchaser of the work.
CONTENTS
Preface
Adatom, Vacancy, and Dimer Diffusion
Experimental Study of Surface Diffusion in Metal Overlayers on
Anisotropic Metal Surfaces
A. T. Loburets, N. B. Senenko, Yu. S. Vedula, and A. G. Naumovets
1
Ab Initio Modeling of Free Energy Profiles in Thermally-Activated
Processes
I. Stich, M. Hytha, J. D. Gale, K. Terakura, and M. C. Payne
19
Surface Diffusion with a Realistic Damping Coefficient O. M. Braun
31
Vibrational and Structural Properties of the Nb(OOI) Surface with
and without a Nb Adatom by Tight-Binding Molecular Dynamics
Ch. E. Lekka, G. A. Evangelakis, N. I. Papanicolaou, and D. A.
Papaconstantopoulos 43
Adatoms and Vacancies on AsB (001) Surfaces Ch. E. Lekka and G. A.
Evangelakis 51
Long-Time-Scale Simulations of A1(I00) Crystal Growth G. Henkelman
and H. J6nsson 63
Diffusion of Dimers on Silicon and Germanium (001) Surfaces H. J.
W. Zandvliet, E. Zoethout, and B. Poelsema 75
Island Nucleation and Multilayer Growth
Island Nucleation in Metal Thin-Film Growth K. A. Fichthom, M. L.
Merrick, R. Pentcheva, and M. Scheffler 87
Capture-Numbers and Island Size-Distributions in Irreversible
Homoepitaxial Growth: A Rate Equation Approach
M. N. Popescu, F. Family, and J. G. Amar 99
Island Statistics ReOecting Growth Processes P. A. Mulheran
111
vi
Growth of an Anisotropic Surface: The Case of AglAg(llO) C. Mottet,
R. Ferrando, F. Hontinfinde, and A. Videcoq 121
Vibrational Properties of 2D Copper Islands on the Cu(lll) Surface
by MD Simulations
E. Vamvakopoulos and G. A. Evangelakis 129
Irreversible Nucleation in Multilayer Growth P. Politi and C.
Castellano 135
Second Layer Nucleation and the Shape of Wedding Cakes J. Krug and
P. Kuhn 145
Steering Epitaxial Growth B. Poelsema and S. Van Dijken 165
Coarsening Mechanisms in Surface Morphological Evolution T.
Michely, M. Kalff, G. Comsa, M. Strobel, and K.-H. Heinig 185
Realistic Atomistic Modeling of Mound Formation during Multilayer
Growth: Metal(IOO) Homoepitaxy
K. J. Caspersen and J. W. Evans 197
Vicinal and Patterned Substrates
Patterning Surfaces by Self-Organized Growth K. Kern 207
Growth and Ion Erosion: Two Methods for Patterning Surfaces F.
Buatier de Mongeot, C. Boragno, and U. Valbusa 221
Oscillatory Driving of Crystal Surfaces: A Route to Controlled
Pattern Formation
O. Pierre-Louis and M. Haftel 243
Reconstruction-Determined Growth of Silver on Silicon(1l1)-(7x7) P.
Sobotik, I. Ost'cidal, J. Mysliveeek, T. Jarolfmek, F. Lavicky, and
P. Smilauer 255
Electromigration of Si Adatoms on Si Surfaces: A Key to Understan
ding Step Bunching Instabilities during Sublimation and MBE
Growth
S. Stoyanov, J. J. Metois, and V. Tonchev 267
vii
Atomic Steps on a Single-Crystal Surface Studied with in situ UHV
Reflection-Electron Microscopy
A. V. Latyshev, S. S. Kosolobov, D. A. Nasimov, V. N. Savenko, and
A. L. Aseev 281
Heteroepitaxy, Strain Relaxation, and Quantum Dots
Mechanisms and Anomalies in the Formation of InAs-GaAs(OOl) Quantum
Dot Structures
B. A. Joyce and D. D. Vvedensky 301
Ab initio Study of Stability of Surfaces and Nanostructures J.
Kollar, L. Vitos, and B. Johansson 327
Atomistic and Continuum Elastic Effects in Heteroepitaxial Systems
A. C. Schindler, D. D. Vvedensky,'M. F. Gyure,G. D. Simms, R. E.
Caflisch, and C. Connell 337
An Initio Thermodynamics and Statistical Mechanics of Diffusion,
Growth, and Self-Assembly of Quantum Dots
M. Scheffler and P. Kratzer 355
Atomistic Aspects of SiGe Nanostructure Formation by Molecular
Beam Epitaxy
O. P. Pchelyakov, Yu. B. Bolkhovityanov, A. 1. Nikiforov, B. Z.
Olshanetsky, L. V. Sokolov, S. A~ Teys, and B. VoigtHinder
371
Stress-Induced Surface Modulation C. MisbaJ:t, P. Berger, and K.
Kassner 383
Entropy Effects in the Self-Organized Formation of Nanostructures
V. A. Shchukin, N. N. Ledentsov, and D. Bimberg . 397
Dislocation-Free 3D Islands in Highly Mismatched Epitaxy: An
Equilibrium Study with Anharmonic Interactions
1. Markov and 1. E. Prieto 411
Self-Assembly of Few-Atom Clusters in a Model of a Strained
Submonolayer
V. 1. Tokar and H. Dreysse 429.
Ab initio Study of the Influence of Epitaxial Strain on
Magnetoelastic Properties
M. Komelj and M. Hihnle 439
viii
M. Siakavellas, A. G. Kontos, and Y. S. Raptis 449
Effect of Annealing at High Hydrostatic Pressure of Silicon
Implanted with Helium and Oxygen
A. Misiuk, J. Katcki, J. Ratajczak, V. Raineri, J. Bak-Misiuk, L.
Gawlik, L. Bryja, and J. Jon 457
Effect of High Temperature-Pressure on Strain Relaxation in Thin
Layers of Semiconductors Epitaxially Grown on GaAs and Si
Substrates
J. Bak-Misiuk, A. Misiuk, J. Adamczewska, M. Calamiotou, A.
Kozanecki, D. Kuristyn, K. Reginski, J. Kaniewski, and A.
Georgakilas 467
Multi-Component Systems
Atomic View of Surfactant Action in Epitaxial Growth: From STM to
Computer Simulation
J. Camarero, A. L. Vazquez de Parga, J. E. Prieto, J. J. de Miguel,
R. Miranda, C. Slutzky, J. Ferron, and L. G6mez 477
Effects of Atomic Interactions in Two-Component Submonolayer
Growth
M. Kotrla and J. Krug 489
Ultrathin Ionic Films Epitaxially Grown on m-v Semiconductors
Studied with Atomic Resolution
M. Szymonski, J. 1. Kolodziej, B. Such, P. Czuba, P. Paitkowski,
and F. Krok 499
Ultraviolet-Assisted Pulsed Laser Deposition of Thin Oxide Fllms V.
Craciun and R. K. Singh 511
Atomistic Theory of the Growth Mode for a Thin Metallic Film on an
Insulating Substrate
E. A. Kotomin, Yu. Zhukovskii, S. Dorfman, and D. Fuks 525
Structure and Formation Mechanism of Nanogranular CoCu Fllms V. M.
Fedosyuk 535
ix
Photoemission Studies of Bimetallic Ultrathin Films: Au-Ni on
Vttria Stabilised Zr02(lOO)
S. Kennou and S. zafeiratos 551
Ab-initio Calculations on the Structural and Electronic Properties
of BaOIBaTi03 and SrOlSrTi03 Interfaces
J. Junquera and P. Ordej6n 561
Atomic Ordering and its Influence on the Optical and Electrical
Properties of InGaP Grown by MOVPE
1. Novak 573
Nanoscopic Study of Zirconia Films Grown by Atomic Layer
Deposition
V. Sammelselg, 1. Karlis, A. Kikas, J. Aarik, H. Mandar, and T.
Uustare 683
Index 593
Preface
Epitaxial growth is at the heart ofa wide range of industrial and
technological appli cations, including magnetic storage media,
electronic or photoelectronic devices, and catalytic converters, to
name just a few. Relevant substances grown in such a way include
magnetic multilayers, semiconductor heterostructures, and metal
oxide systems. In these materials precisely controlled growth and
composition are essential for the proper functioning of the
devices. Recent breakthroughs, experi mental as well as
theoretical, allow atom-by-atom manipulation and understanding of
these processes, opening up a totally new era of unprecedented
nanostructuring. In view ofthese developments, it was felt
particularly timely to hold a workshop that brought together
experimentalists and theorists from various disciplines, to discuss
common problems and solutions, to stimulate interdisciplinary
cross-fertilization, and to review recent progress and remaining
challenges.
The Scientific Affairs Division of NATO kindly agreed to support
such a meeting under its Advanced Research Workshop program.
Entitled "Atomistic Aspects of Epitaxial Growth" the conference was
held at the Corfu Chandris Hotel in Dassia, Corfu· (Greece), June
25-30,2001. One of the key objectives was to provide a gathering
for several communities that do not in general interact much.
Scientists studying metallic systems do not usually meet with those
studying semiconductors, while the oxide-community is separated
from both ofthem. Although there are clear differences between
these various materials and the ways they are studied, there are
also many common aspects that deserve further and deeper
exploration. Moreover, there is great poteptial for transferofideas
and techniques between the various fields. In addition, the
workshop was roughly equally divided between experimental and
theoretical talks. Speakers were strongly encouraged to address
both aspects in their presentations.. Judging from participants'
subsequent comments, it was felt that the meeting had largely
succeeded in these ambitions.
The workshop was particularly topical because of very recent
breakthroughs both in theory and in experiment. On the theory side,
progress in computer hardware and software (including parallel
computation) has made it possible to study larger systems than
before at the first-principles level. This means that true 'ab
initio' calculations of diffusion and growth, without any
adjustable parameters, are now feasible. On the other hand kinetic
Monte Carlo simulations allow one to obtain results for both
realistic diffusion barriers and time scales. At the same time, new
methods aimed to bridge the gap between the time scale of molecular
simulation and the much longer typical time intervals of growth
processes are' being devel oped. On the experimental side, much of
the detailed microscopic understanding
xi
xii
of epitaxy comes from the wide availability of scanning tunneling
microscopes and their derivatives. Experts can now routinely make
'in situ movies' that show atomic motions, islanding, terracing,
etc. These can in tum be compared with atomistic simulations ofthe
same phenomena, leading to an unprecedented level of
testability.
The meeting was organized around three central themes:
1. experiments that permit atomic resolution of configurations and
kinetics (STM, AFM, AES, ISS, etc.);
2. first-principles calculations of binding energies and energy
barriers;
3. simulations of the dynamics of deposition, growth, migration,
and island forma- tion.
These themes are not isolated, however, and the interaction and
cross-fertilization between them was strongly emphasized. Some
specific issues that were discussed included: atomic manipulation
and probe microscopy; effects of strain and mis match; connecting
microscopic parameters to macroscopic models; role of adatom and
island diffusion in film growth; manipulation of growth by
additives, surfac tants, and ion bombardment; scaling relations
and growth instabilities; nature and effects of diffusion
barriers.
The present volume contains the proceedings of this stimulating
meeting. At the request of the organizers, speakers updated their
conference presentations to reflect discussions and subsequent
developments in the field. Thus this volume should be useful not
only to experts, but also to those newly entering this exciting
field or to anybody who wishes to have an overview of the state of
the art in epitaxy.
The organizers are deeply thankful to the Scientific Affairs
Committee of NATO and especially to the programme director, Dr.
Fausto Pedrazzini, for their spon sorship and enthousiastic
endorsement of this meeting. Further financial support was received
from the Physical Section of the Union of Czech Mathematicians and
Physicists, the Greek Ministry of Development, and the University
of Ioannina (Greece). These institutions are also gratefully
acknowledged. Finally, the organiz ers wish to thank all
participants for their cooperation, their sharing of knowledge and
friendship, and the good humor and promptness with which they
responded to various deadlines.
Miroslav Kotrla Nicolas I. Papanicolaou Dimitri D. Vvedensky Luc T.
Wille
EXPERIMENTAL STUDY OF SURFACE DIFFUSION IN METAL OVERLAYERS ON
ANISOTROPIC METAL SURFACES
A. T. LOBURETS AND N. B. SENENKO Yu. V. Kondratyuk State Technical
University, UA-63601, Poltava, Ukraine
AND
YU. S. VEDULA AND A. G. NAUMOVETS Institute of Physics, National
Academy of Sciences of Ukraine, 46 Prospect Nauki, UA-03028, Kiev
28, Ukraine
Abstract. We compare and discuss the diffusion kinetics of Li, Sr,
Dy and eu overlayers on the (112) Mo and W surfaces. The
experimental data are used to assess the role of such factors in
surface diffusion as the substrate and diffuser chemical nature,
substrate atomic structure, lateral interactions, phase transitions
and effect of coadsorbates.
1. Introduction
The mobility of surface atoms and molecules is one of the key
factors in epitaxial growth and many other surface phenomena
underlying a whole array of technologies. Recent investigations
repeatedly have demonstrated the many-body nature of surface
diffusion (SD), which essentially compli cates the comprehension
of SD processes and, at the same, time makes their studying really
enthralling (see, e.g., Ref. [1]).
A serious weakness of present-day experimental evidence on SD is
its rather fragmentary character. There are few systematic results
which could be used for extracting general regularities of SD
kinetics in relation to var ious relevant factors: the interaction
of diffusing species with the substrate and each other, the
substrate atomic structure and presence of surfaces of all kinds,
the phase transitions in the spreading overlayer, etc. [2-5]. An
almost unexplored area is the diffusion in coadsorbed (mixed)
overlayers.
1
M. Kotrla et al. (eds.), Atomistic Aspects ofEpitaxial Growth,
1-18. o 2002 Kluwer Academic PubUshers.
2 A. T. LOBURETS ET AL
Its regularities are of prime importance in epitaxial growth of
compound thin films, in catalysis, powder metallurgy of alloys and
other similar ap plications, where two or more components are
diffusing simultaneously.
A grave problem in the generalization of experimental SD results is
the wide scatter of the published data [6]. This stems from the
high sensitivity of SD parameters (diffusivity, activation energy
and pre-exponential factor) to experimental conditions as well as
to the method of their determination. This raises the worth of
comparative studies in which different systems are examined under
identical conditions, i.e., using the same substrates, measurement
techniques, etc. Such an approach ensures the most reliable and
conclusive revealing of common trends and peculiarities in
diffusion behaviour of different systems.
In this paper, we present and discuss experimental results obtained
in our study of surface diffusion of metals on metals. The
diffusers were Li, Sr, Dy and Cu, which represent alkali,
alkaline-earth, rare-earth and noble metals, respectively. The
(112) surfaces of Wand Mo served as substrates. This choice of
objects provided the possibility to compare the behavior of
different adsorbates on the same substrate in order to assess the
impact of the diffuser chemical nature on diffusion kinetics. It
should be remem bered that the polarity of Li, Sr and Dy
adsorption bonds on W and Mo is considerably higher than that of
Cu. On the other hand, the diffusion characteristics of the same
adsorbate can be contrasted on two substrates having different
chemical properties, but practically identical atomic struc ture.
Recall that both W and Mo have a bcc structure and their lattice
constants (aw = 3.165A, aMo = 3.147A) differ by a mere 0.57%.
The most salient feature of the (112) W and Mo surfaces is their
highly anisotropic ("channeled") atomic corrugation. They are built
of parallel close-packed rows of atoms, separated by atomically
deep "channels" (fur rows). Owing to this, the diffusion
activation barriers along the channels are much lower than across
the channels, at least for the adsorbates ex amined in this work.
As a result, the adatoms are found to diffuse almost exclusively
along the channels while their mobility across the channels is ac
tually blocked at not too high temperatures [7-9]. Thus, one has a
model of one-dimensional diffusion process, which in a way
facilitates the evaluation and interpretation of experimental
findings.
The structural information about diffusing overlayers is known to
be highly important to understanding the SD mechanisms and origins
of the variation of SD parameters with coverage and temperature.
Some LEED data are available for all the systems listed above
although their full phase diagrams are not always documented.
We have also investigated SD in some mixed overlayers composed of
Li and Dy, Li and Sr, Sr and Cu on W and Mo(112). The aim of
these
EXPERIMENTAL STUDY OF SURFACE DIFFUSION 3
experiments was to gain insight into. the mechanisms of changes in
SD kinetics caused by foreign additives, which can be regarded as a
kind of surface defect introduced deliberately.
2. Experimental
In this work, as in our previous investigations [5,10], the data on
SD ki netics were obtained from observations of the time evolution
of coverage distributions (proffies) shaped initially as coverage
steps. The coverage is defined here as the ratio of the
concentration of adsorbed atoms to the surface concentration of
substrate atoms. The latter is equal to 8.2x1014
cm-2 on W(1l2) and 8.3x1014 cm-2 on Mo(1l2). The initial step-like
cov erage profiles were formed by vacuum evaporation of the
adsorbate onto the substrate, half of which was screened with a
mask. The initial coverage in the step could be chosen within the
interval from 0.1 to 3 monolayers, depending on the task of
particular experiments. The mask had a straight edge, which was
always oriented. normally to the atomic channels on both W(1l2) and
Mo(1l2). Thus, the results obtained in this work relate to SD along
the channels, i.e. along the close-packed direction [III].
To recor~ a coverage proffie, we first measured the distribution of
the work function over the surface under study. This was done by
scanning the surface with a narrow elect):'on beam (from rv 1 to 15
/-Lm in diame ter) [10,11]. The beam is generated by an electron
gun, and a longitudinal magnetic field is used to improve the beam
focusing. The local current voltage (CV) characteristics of the
current between the gun cathode and the sample serving as a
collector (anode) are measured in different points of the surface.
The position of the falling (retarding field) section of the CV
curve is known to depend on the contact potential difference
between the gun cathode and the collector. This effect underlies
the Anderson method determining contact potential differences.
Using it one can record a con tact potential (work function) map
of the surface. The dependence of the work function on coverage is
carefully calibrated for each system in separate experiments. Since
the work function depends generally not only on cover age, but
also on the overlayer structure, it is necessary to calibrate the
work function versus coverage at different annealing temperatures
corresponding to temperatures of diffusion experiments. Such
measurements allow one to account for the work function changes
that are caused by structural rear rangements of overlayers due to
annealing. This is essential to guarantee a good accuracy of
coverage determination for adatoms which have a low mobility at the
temperatures of the overlayer deposition.
The advantages of the contact potential technique are its
non-destructive character, the high coverage sensitivity (ranging
from 10-2 to 10-1 of a
4 A. T. LOBURETS ET AL
«x'» 'fl(cm)
D=10"cm'/s
10' I /.:::::om
'" I i resolution
10 10' lis)
Figure 1. Measurement possibilities of scanning contact-potential
microscopy with re spect to paths and times of surface diffusion
(shaded area). See text for explanations.
monolayer) and the spatial resolution rv 1-;-15 /-Lm, which proved
to be sufficient for revealing rather fine features in the coverage
profiles. All of these features motivate calling this technique
"scanning contact potential microscopy" .
It is important to note that none of the adsorbates studied in this
work show significant solubility in Wand Mo. Thus, the adsorbate
drawing off into the substrate volume during annealing can be
neglected, and the pro cess of adsorbate spreading can be treated
as "pure" surface diffusion. It seems likely, however, that for the
same systems and under certain condi tions, SD may be accompanied
by surface reconstruction. This results in a mixing of adsorbate
and substrate atoms, Le. in the formation of a kind of surface
alloy (see Sec. 3.3).
The data on the evolution of the coverage profiles () = ()(x, t)
(() is coverage, x is the coordinate along the diffusion direction
and t is time) were evaluated using the Boltzmann-Matano method
[12]. A prerequisite for its applicability is the "normal".
character of SD, i.e. the requirement that the displacement of the
profile from its initial position (the Matano plane) be a linear
function of t 1/ 2 at each () = con$t. It was found that in most
cases this requirement is fulfilled with fair accuracy (see, e.g.,
Ref. [13]). However, we have also observed a number of
manifestations of anomalous diffusion when x is a sublinear
function of t 1/ 2 • These relate mostly to SD in coadsorbed layers
and are discussed in Sec. 3.4. All the measurements were made in a
vacuum of rv 10-11_10-12 Torr. Further details of our experimental
technique have been given in Refs. [8-11,13J.
EXPERlMENTAL STUDY OF SURFACE DIFFUSION 5
...... -2..,.-----------------, ~
0.0 0.5 1.0 1.5 2.0 2.5 3.0 coverage 9
Figure 2. Surface diffusivities vers1J.S coverage for Li, Sr, Dy
and eu on the Mo(1l2) surface.
Figure 1 shows the range of diffusion distances and times
measurable in our experiments. The lower and upper bounds of
distances are set by the spatial resolution (10-4-10-3 cm) and the
sample dimension (tv 5 x 10-1 cm), respectively. The lower bound of
the diffusion time (tv 10 s) is determined by the time required to
set a prescribed value of the diffusion temperature. The upper time
bound is imposed by vacuum conditions, i.e. by the necessity to
guarantee sufficient surface cleanliness of impurities coming from
residual gases. In any event, the amount of impurities should not
exceed tv 10-2 of a monolayer, but generally it is desirable to
have even a higher level of cleanliness because of the extreme
sensitivity of the diffusion kinetics to coadsorbates. In
consequence, we could determine the diffusivities within the range
of tv 10-10 to 10-3 cm2/s.
3. Results and Discussion
3.1. DIFFERENT ADSORBATES ON THE SAME SUBSTRATE
Figure 2 shows a summary of our data on the diffusivities (D) of
Li, Sr, Dy and eu in their relation to coverage (0) on the Mo(112)
surface [8,9,13,14]. A number of conclusions can be drawn from the
comparison of these findings.
3.1.1. Role of the Adsorbate Chemical Nature There is a very strong
distinction between the diffusion properties of the adsorbates
studied. For this reason, as well as due to very broad range of
variation of the diffusivities with coverage, it was impossible to
acquire experimentally the curves D versus 0 for. all the
adsorbates at the same temperature. However, one can obtain such
data by extrapolating the tem-
6 A. T. LOBURETS ET AL
perature dependencies of D and, in this way, compare the adatom
mobilities at the same temperature. For instance, according to data
in Ref. [9], the diffusivity of Li on Mo(112) in the interval 0
< () < 0.4 should be equal to", 10-4 cm2/s at T = 600 K.
Under the same conditions the diffusivity is '" 10-7 - 10-6 cm2Is
for Cu, '" 10-8 - 10-7 cm2/s for Sr and '" 10-10
cm 2 Is for Dy. Even at T = 300 K, DLi in the range 1.0 < ()
< 1:5 is sub stantially higher than Dcu determined at T = 600 K
in the same () range. Therefore we have a series of inequalities:
DLi > Dcu > DSr > DDy.
3.1.2. Root-Mean-Square DiJJusion Path until Desorption At
near-monolayer and supermonolayer coverages, there exists a
striking difference in the ability of the adsorbates being studied
to diffuse over detectable distances ('" 10-3 cm) within reasonable
times of the experiment ('" 104 s). Recall that the mean residence
time of an adparticle on the surface until desorption is T = TO
exp(EalkT) , where TO is approximately the period of adatom thermal
vibrations, Ea is the desorption activation energy (equal to the
adsorption energy in the case of metals on metals) and k is
Boltzmann's constant. Thus the root-mean-square diffusion distance
until desorption is (x2)~~ ()( (DT)I/2 ()( exp[(Ea-Ed)/2kT].lfthe
difference (Ea- Ed) is small, it may be necessary to perform the
experiment at rather
low temperatures to obtain (x2)~~ exceeding the spatial resolution
bound. However, the diffusivity in this case may become so low that
the time needed to wait until such an adparticle displacement will
exceed the allowable upper time bound (Fig. 1).
It is precisely this situation that is typical of Sr and Dy, as
well as of other alkaline-earth and rare-earth diffusers [15]. As
it is seen from Fig. 2, it was possible to evaluate the
diffusivities for Sr and Dy at () < 0.65 only. At higher
coverages, the difference (Ea - Ed) for these elements becomes too
small to allow the determination of their Ds, for the reasons
discussed above. In contrast, the diffusion of Li and Cu at () ~ 1
and () > 1 can easily be observed over macroscopic distances ( ~
1 mm in our experiments). This finding demonstrated that the value
!(Ea - Ed) is large enough for Li and Cu at all coverages
studied.
3.1.3. DiJJusivity versus Phase Transitions in Overlayers Let us
now compare more closely the variation of the diffusivity versus
coverage for different adsorbates. First we will consider the data
for elec tropositive adsorbates Li, Sr and Dy. For Sr and Dy as
well as for Cu, the region of low coverages (0 < () < 0.15)
is characteristic of somewhat enhanced diffusivity. Such regularity
was not reliably observed for Li on Mo(112), in which case D
remains nearly constant in the range 0 < () < 0.4. However, a
related system Li/W(112) exhibits a significant increase in D
at
EXPERIMENTAL STUDY OF SURFACE DIFFUSION 7
() < 0.15 [16]. Thus, the enhanced diffusivity at low coverage
is rather typ ical behaviour. Most probably, it can be attributed
to the fact that surface diffusion in this () range is effected by
jumps of individual adatoms and/or their small clusters
(oligomers). It is known from structural data obtained by LEED that
Li, Sr and Dy adatoms tend to form chains at low coverages on Mo
and W(112) [17-23]. These chains consist of adatoms located in the
channels and are oriented across the channels. This feature mirrors
the high anisotropy of the lateral interaction imposed by the
anisotropic structure of the substrate [20]. Therefore, it may be
anticipated that the electropositive adatom clusters (oligomers) on
Mo and W(112) represent short segments of chains. Such chain-like
clusters were observed on the channeled (110) surfaces of Ir and Pt
[24]. They were shown to diffuse mainly via successive
displacements of individual adatoms forming the clusters. This
mechanism resembles the reptational motion of large organic
molecules [25]. Actually, the formation of clusters imposes a
pronounced collective character on SD even at very low (average)
coverages.
The range of medium coverages (0.15 ::; () ::; 0.5) in
electropositive over layers on Mo and W(112) is typically the
region of extensive first-order phase transitions from the rarefied
chain structures to rather dense com mensurate phases [20]. Figure
2 shows that the diffusivities of all the adsor bates under study
are comparatively low at medium values of (). We ascribe this to
the domination of the attractive lateral interaction in the regions
'of the first-order phase transitions. The lateral attraction
reduces the driving force of the diffusion and, for this reason,
slows down the diffusion rate [26]. It should be recalled, however,
that the overlayer becomes heterogeneous in the first-order
transition region. Owing to this, the diffusivity determined by a
macroscopic method represents a value which is, in a way, averaged
over the D values characteristic of the coexisting phases [27].
.
An important stage limiting the net diffusion rate in the
heterogeneous overlayer is the detachment of adatoms from the
islands of the dense phase. In this step, the adatoms must overcome
the additional barrier caused by the attractive forces which
assemble them into the islands. This seems to be the most probable
reason for the low (averaged) diffusivity observed in the
first-order transition regions.
As the coverage approaches a close-packed monolayer,
electropositive adatoms usually arrange into incommensurate
structures. This is caused by the combination of three factors: the
difference in size of adatoms and substrate atoms (lattice misfit),
the strong attraction of the adatoms to the substrate and, at the
same time, their strong lateral repulsion at values of ()
approaching the dense monolayer. The transition from a commensurate
to an incommensurate structure (the C-I transition) is marked with
a strong rise in the diffusivity (Fig. 2). For Sr and Dy on
Mo(112), a sharp D max-
8 A. T. LOBURETS ET AL
imum is observed in the course of the C-I transition. For Li, there
occurs a rather steep, but practically monotonic, increase of the
diffusivity as () ---+ 1, which continues into the transition to
the second monolayer packing.
According to structural data [20], the C-I transitions in
electropositive overlayers on the (112) Mo and W surfaces result in
uniaxial compression of the commensurate adatom lattices along the
chamiels. This is the con sequence of the strong anisotropy of the
surface corrugation inherent to the (112) planes of bcc crystals.
At the early stage of the C-I transition, the overlayerjsubstrate
structural coherence is not lost at once over the whole surface,
but is localized within incommensurate domain walls. They separate
the areas (domains) of the commensurate structure and represent
misfit dislocations in the overlayer. The wall width depends on the
am plitude of the substrate potential relief and on the lateral
interaction [28]. The incommensurate domain walls repel each other
and can form their own ordered structures. In the case of uniaxial
overlayer compression, such structures represent a system of
parallel domain walls. They are oriented across the atomic channels
or, in the more general case, can be tilted at some angle with
respect to the channels [22]. The incommensurate domain walls
possess, from the mathematical viewpoint, the properties of
topologi cal solitons [28]. Their motion across the commensurate
phase can provide fast mass transport (the soliton mechanism of
surface diffusion) [28-30]. In the situation when the coverage is
higher than its stoichiometric value corresponding to the perfect
commensurate structure, the domain walls in corporate the
"excessive" adatoms with respect to the commensurate phase. Thus,
the motion of such a domain wall (soliton) across the C- phase
trans ports the adatoms over the surface. It should be noted that
this transport is effected by a relay mechanism, which is evidently
collective in nature.
So far we considered the fast moving domain walls (solitons) in the
C-phase which appear at superstoichiometric coverages and contain
extra adatoms. However, the experimental data obtained for Sr and
Dy reveal that their diffusivity substantially grows also while
approaching the stoi chiometric coverage from below. In this case,
the walls between the com mensurate domains contain vacancies with
respect to the C-phase. The adatoms situated near such a domain
wall should also be somewhat shifted from the sites that are
regular for the C-phase. Thus, the vacancy domain wall also
represents a locally incommensurate configuration (a vacancy soli
ton, or "antisoliton," with respect the soliton containing
superstoichiomet ric adatoms). The results depicted in Fig. 2
suggest that vacancy solitons have a lower mobility than the
superstoichiometric ("interstitial") solitons.
As noted above, the diffusivity of Li on the Mo(112) surface
increases nearly monotonically in the interval 0.5 < () < 1
where a uniaxial com pression of the overlayer occurs along the
atomic channels [19,22]. Only a
EXPERIMENTAL STUDY OF SURFACE DIFFUSION 9
minor D maximum is observed at () ~ 0.95, i.e. close to the
complete first monolayer. Li diffusivity in the second monolayer is
higher than at () ~ 1, which is the prerequisite for the operation
of the "unrolling carpet" diffusion mechanism [31]. This mechanism
is known to result in spreading of the first monolayer via
diffusion of the adatoms in the second (more mobile) mono layer.
It is interesting to note that Li adatoms adsorbed on the
close-packed (llO)W surface seem to show a higher mobility within
the first monolayer [32]. This distinction demonstrates the
dramatic sensitivity of the diffu sion kinetics and mechanisms to
the peculiarities of adatom/substrate and adatom/adatom
interactions on different surfaces. Obviously, the operating
diffusion mechanism should also depend on the observation
temperature.
We proceed now to discuss the diffusion kinetics of Cu on Mo(112).
In contrast to electro-positive adsorbates considered above, Cu on
transition metal surfaces forms an adsorption bond with a low
polarity. The work function dependence on () recorded for Cu on
Mo(112) represents a broken line with breaking points corresponding
to () = 1 and () = 2 (the mea surements were performed at () ~ 3
only). The work function grows linearly with () in the first and
third monolayers, and decreases linearly in the second monolayer.
Qualitatively, a very similar work function variation is observed
for Cu on W(112) [14,33]. Such a nonmonotonic behaviour of the work
function with the filling of successive monolayers can primarily be
ascribed to an alternate smoothening and roughening of the surface
by adsorbed Cu atoms (Smoluchowski's effect). The linear coverage
dependence of the work function is typical of overlayers undergoing
a first-order phase transition, which in turn, manifests the
existence of the attractive lateral interaction.
We are not aware of any structural data for the Cu/Mo(112) system.
However, such data are available for the related system CujW(1l2)
[33], which shows a close similarity to Cu/Mo(112) both in its work
function and surface diffusion characteristics [14]. According to
Ref. [33], successive Cu monolayers on W(112) are each filled by
first-order transitions from a rarefied 2D gas phase to a dense
structure which closely resembles (or is isomorphous to) the
substrate structure. Assuming that analogous tran sitions occur in
the Cu overlayer on Mo(1l2), one can attribute the high D peaks
found at () --t 0, () ~ 1 and () ~ 2 (Fig. 2) to the enhanced mo
bility of individual Cu adatoms and their oligomers in the 2D gas
phases formed at these coverages. An additional contribution to the
D maxima may come from the thermodynamic factor [5], which should
be a maximum near coverages () ~ 1, and () ~ 2, where the
adsorption energy undergoes fast drops in transitions to packing
the next monolayers. The soliton dif fusion mechanism may be
expected at the final stage of filling the second monolayer. Here,
the isomorphous Cu structure is additionally compacted via a
C-I-transition [33].
10
-5
-7
-8
T=600K
0.0 0.5 1.0 1.5 2.0 2.5 coverage 8
Figure 9. Surface diffusivities versus coverage for Cu on Mo(112)
and W(112) surfaces.
The heights of the D peaks grow with increasing () (Fig. 2) which
seems to reflect the changes in the chemical nature of the
underlying surface. The relatively low values of the Cu diffusivity
between the D peaks are coupled to the regions of the first-order
phase transitions. The interpretation of this effect was given
above.
3.2. THE SAME ADSORBATE ON DIFFERENT SUBSTRATES
An important issue is the role of the substrate chemical nature in
SD kinet ics. As noted in Sec. 1, Mo and W have almost identical
atomic structures, so the comparison of SD characteristics on them
is particularly informative. An example of such a comparison is
shown in Fig. :3 for Cu on the Mo(112) and W(112) surfaces. One can
see a close qualitative similarity of the dif fusivity curves
plotted against the coverage (both the curves correspond to T = 600
K). The D peaks are observed at nearly the same coverages, and
their heights at () ~ 1 and () ~ 2 are practically the same.
However, there are some quantitative distinctions, amounting to
about one order of magnitude, at submonolayer coverages and at
medium coverages within the second and third monolayers. These
coverages are just the regions where the first-order phase
transitions take place. Thus, the distinctions in the diffusivity
observed on Mo and W -reflect the specific interactions of the
diffusing adatoms both with the substrates and with each other. It
is inter esting to note that Cu diffusivity in some () intervals
is higher on Mo, while in others, it is higher on W. To explain
these dissimilarities, detailed cal culations of the corrugation
in the potential and of the lateral interactions are needed which
take into account the electronic structure of Mo and W
surfaces.
EXPERIMENTAL STUDY OF SURFACE DIFFUSION 11
4
1.6
12
1.0
0.8
0.6
Dy/Mo(112)
-2
-6
-8
-10
e
Figure 4. Activation energy of surface diffusion (Ed, left scale)
and pre-exponential factor (Do, right scale) for Dy on the Mo(112)
surface as functions of coverage.
An analogous comparison of the diffusivities on the Mo(112) and
W(112) surfaces has been made for Li and Sr. We shall not present
the results here and refer the readers to our recent wOJ:k [16] for
details. In brief, the diffusivities on (112) Mo and W are quite
close to each other both in their absolute values, relating to the
same temperature, and in the shape of their dependence on coverage.
.
On the other hand, the comparison of Li diffusion on the (112) and
(110) surfaces of W reveals their dramatic distinction (the
diffusion data for Li on W(110) were obtained in Ref. [32]). This
result evidences the critical importance of the substrate atomic
structure in SD kinetics. The effect stems both from the surface
potential corrugation, directly connected with the atomic
structure, and from the structure of the overlayer, also correlated
with the substrate structure. As a consequence, the difference in
the substrate chemical nature may prove less important for SD
kinetics than the similarity of the atomic structures of the
substrates compared.
3.3. DIFFUSIVITIES VERSUS TEMPERATURE
The diffusivities for all the systems investigated in this work
were obtained as functions of temperature and plotted on the
Arrhenius coordinates to determine the SD activation energy and the
pre-exponential factor Do. The results are shown in Fig. 4 for the
DyjMo(112) system [13]. Generally, Ed and Do are found to vary with
coverage in an intricate way, and a compen-
12 A. T. LOBURETS ET AL
sation effect is observed (Le. Ed and Do change "in phase"). The
variation of Do usually spans the range of a few orders of
magnitude. For some sys tems and at some coverages, it appears
close to the value a211 rv 10-2 cm2Is (a rv 3 X 10-8 cm is a
typical substrate lattice period and II rv 1013 s-1 is the thermal
vibration frequency). However, it is rather the exception than the
rule. The "anomalous" (i.e. deviating from Do rv 10-2 cm2 Is) Do
values are often ascribed to the presence of surface defects
(traps) or to possi ble long adatom jumps [6,31,34]. It seems,
however, that at least in some cases this deviation can result from
the formal application of the Arrhenius law to situations where its
validity can hardly be expected. We mean the situations when
variations in temperature lead to qualitative changes in the state
of the system which cannot be described merely in terms of the
Boltzmann occupancies of the vibration levels. The obvious examples
are surface phase transitions. However, even in the case of very
low coverages when the overlayer consists of individual adatoms and
small clusters, the temperature variation will change the size
distribution of the clusters. The diffusion parameters (and even
the diffusion mechanisms) were shown to depend on cluster size
[35-37]. It is apparent that the approximation of the temperature
dependence of the diffusivity by the Arrhenius law can give for
such a system consisting of different particles nothing more than
some effective values of Ed and Do. Obviously, these effects that
complicate the simple meaning of Ed and Do originate essentially
from the lateral interac tion of the diffusing particles, Le. from
the many-body character of surface diffusion (see also the
discussion of this problem and of role of surface het erogeneities
in Ref. [38]).
In a recent study of the structure of Dy overlayers on Mo(112),
car ried out in a broad temperature interval, we have also
revealed the pos sibility of an irreversible change in the
mobility of adatoms that occurs with increasing temperature [23].
Dy on Mo(112) was found to form an or dered overlayer immediately
in the course of evaporation onto the surface at T = 100 K. This
fact testifies that Dy adatoms posses a sufficient sur face
mobility under such conditions. Unexpectedly, in the coverage range
0.07 < () < 0.57, the annealing of such an ordered system at
T > 400 K leads to an irreversible loss of long-range order. It
should be recalled that diffusion experiments [13] showed that Dy
adatoms migrate quite quickly on Mo(112) at T ~ 750 K. However, the
cooling of the system from these temperatures does not restore the
long-range order in the overlayer (con trary to the situation·
observed at () > 0.58). These findings suggest that the
DyIMo(112) system may undergo a vitrifying transition which results
in formation of a two-dimensional glass at coverages 0.07 < ()
< 0.57. This effect has been hypothetically attributed to the
thermally activated recon struction of the DyIMo(112) system,
which may produce a surface Dy-Mo
EXPERlMENTAL STUDY OF SURFACE DIFFUSION 13
alloy [23]. The intermixing of Dy and Mo atoms in such an alloy can
sub stantially reduce the Dy mobility in comparison with the
situation when Dy migrates over the unreconstructed Mo(112)
surface. This can create the specific conditions necessary for the
occurrence of a glass-like state. The assumption about the
substantial structural changes which result from the annealing of
the Dy jMo(112) system is corroborated by the observation of
considerable work function variations occurring in the process of
annealing.
3.4. DIFFUSION IN BINARY OVERLAYERS
The results presented above illustrate the processes involved in SD
which are connected with lateral interactions and phase transitions
in the diffus ing overlayer, the possibility of substrate
reconstruction in the course of diffusion, etc. Obviously, a number
of additional degrees of complexity are added when we consider
diffusion in mixed overlayers. One should realize, however, that it
is this complicated situation that corresponds to the dif fusion
processes in the epitaxial growth of compound films, catalysis and,
actually, in all SD processes that occur under ambient
(non-ultra-high vac uum) conditions. Binary overlayers are clearly
the simplest model systems to investigate SD in coadsorbed layers.
To further simplify the situation, it is reasonable to examine
systems where the coadsorbates have substantially differing
mobilities. As is seen from Fig. 2, the possible candidates for
such investigations are, for example, Li and Dy, Li and Sr, Sr and
Cu.
The idea of our experiments was to investigate the impact of a less
mo bile ("slow") coadsorbate on the diffusion kinetics of a more
mobile ("fast") coadsorbate. To this end, we first covered the
whole substrate surface with a (macroscopically) uniform
preadsorbed layer ("base") of the slow adsor bate. Then a
step-like initial () profile of the fast adsorbate was prepared on
such a surface. The evolution of this profile was followed in a
usual way and compared with its evolution on the clean surface.
Using this approach, we have investigated the systems Li on
DyjMo(112), Li on SrjW(112), and Sr on CujW(112). Some results for
Li on DyjMo(112) , which will be discussed in detail below, are
shown in Fig. 5. Data for Li on SrjW(112) and Sr on CujW(112) were
presented in our recent work [16J. From these results, one can see
an extremely strong suppressing effect of the slow coadsorbates on
the diffusivity of the fast ones. The effect can amount to one or
two orders of magnitude even in the case when the coverage of the
slow coadsorbate is as low as "'. 10-2 of a monolayer [16]. There
are at least two reasons that can account for such a strong
inhibition. First, the structure of the (112) bcc surfaces results
in the transport of adatoms mainly along the atomic channels, so
the presence of a "stopper" in a channel can effectively block its
throughput. Second, due to the collective character of the
diffu-
14
Li-Dy-Mo(112) T=550 K
max=1
0.0 -i""'::,-.--r---'1r-r--,.---r---r--.--,-........--r---'1r-1 o 5
10 15 20 25 30 35
ell(S'~
Figure 5. Plots of x versus t 1 /
2 for evolution of Li coverage profiles on Dy/Mo(112). x
corresponds to OLi=O.l in the profiles. The initial profile is
positioned at x = O. See text for details.
sion mechanisms,the blocking (pinning) effect of a stopper
encompasses not just a single diffusing adatom but some' group of
the adatoms such as an oligomer, a chain or a domain wall (soliton)
[28, 30].
Let us now inspect more closely the findings for Li on Dy/Mo(112)
shown in Fig. 5. Linearizing the profile shifts (at 8Li = const)
versus t 1/ 2 in accordance with Fick's law as x <X (Dt)1/2, one
readily estimates that Li dif fusivity drops by a factor of", 102
on passing from 8Dy = 0.1 to 8Dy = 0.25. However, it can be seen
that the curves exhibit a slight but unquestionable deviation from
linearity (they are actually sublinear). In our previous work [39,
40], we presented examples for the same system where such effects
are even more pronounced and thus represent a manifestation of
anomalous surface diffusion ("subdiffusion"). The phenomenon of
anomalous diffusion is characteristic of complex systems
("structures with variations" [41], in particular glasses), and
attracts much interest (see, e.g. Refs. [?]). It will be recalled
that, according to recent LEED observations [23], Dy overlayers on
Mo(112) at 0.07 < () < 0.57 subjected to high-temperature
annealing form a two-dimensional glassy state at low temperatures
(see Sec. 3.3). It can be speculated that this peculiarity may be
related to anomalous Li diffusion observed on Dy/Mo(112) substrate.
In any event the preadsorp tion of a slow adsorbate leads to an
intentional introduction of ne,w defects (in addition to the
intrinsic substrate defects) into the diffusing overlayer. This can
create a surface with a complex structural hierarchy imposing an
anomalous diffusion scenario.
In contrast to Dy base overlayer, which slows down Li diffusion, an
Li base overlayer (deposited uniformly over the Dy base on (112)Mo)
substan-
EXPERIMENTAL STUDY OF SURFACE DIFFUSION 15
1.0 a)
0.8 Li-Dy-Mo(112)
0.2
b) X (mm) 1.0
0.6
0.2 1 - t=O 2 - t=120 s
0.0 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
X (mm)
Figure 6. Comparison of the spreading of Li step profile on two
substrates: (a) Mo(112) surface precovered with a Dy base overlayer
(8vl/=O.I); (b) Mo(112) surface precovered with Dy base 8v l/=O.1
and additionally with a Li base overlayer 8Li=O.4. In both cases,
total coverage in Li step is 8Li=1.
tially accelerates the spreading of Li coverage profile (Fig. 6).
Obviously, SD in multicomponent systems represents an interesting
and almost unex plored area.
4. Conclusions
Using the Mo(112) and W(112) surfaces as substrates and recording
the evolution of coverage profiles by local contact potential
measurements, we have studied and compared the surface diffusion
kinetics of four metals: Li, Sr, Dy and Cu. They represent,
respectively, alkali, alkaline-earth, rare earth and noble metals
of the periodic table. The nature of the adsorption bonds of these
diffusers is substantially different from the standpoint of their
adsorption energy, binding orbitals and polarity. It comes thus as
no sur prise that their diffusivities, under otherwise equal
conditions, show strong
16 A. T. LOBURETS ET AL
differences. On the other hand, a common feature is the dramatic
and in tricate variation of the diffusivity with coverage, which
spans 2+3 orders of magnitude for Sr, Dy and eu and almost 6 orders
of magnitude for Li. This variation stems primarily from the
lateral adatom interactions, which drive phase transitions and
many-particle diffusion mechanisms in the overlayers. The kinetics
of the phase transitions depends on adparticle diffusivity, but, in
turn, the diffusivity depends on the structure of the emerging
surface phases. This causes a dynamical self-organization of the
diffusion zone. In particular, the commensurate-incommensurate
transitions in metal over layers entail a substantial rise in the
diffusivity. The changes in the diffu sivity due to phase
transitions go in parallel with changes in other physical and
chemical properties such as chemical reactivity, wettability,
electron emission, and metallicity[45, 46]. In other words, one
deals with a deep rearrangement of all surface properties in the
course of surface diffusion, which is of great importance to
epitaxy, catalysis, etc.
Our recent investigations suggest the possibility of surface
reconstruc tions in some systems that drastically inhlbit
adsorbate diffusion and can, at sufficiently low temperatures, end
up with the formation of glass-like surfaces. Interesting effects
have been found in diffusing binary overlayers whose components
have substantially different mobilities. In addition to the very
strong blocking action of slow coadsorbates .(partially caused by
the channeled structure of the substrates studied) there are also
manifesta tions of anomalous diffusion regularities. This may
signal that the overlayer acquires a structure with some peculiar
inhomogeneity.
Thus, judging from the experimental results presented in this work,
we still face ma.n:y physically interesting and practically
important problems in surface diffusion.
Acknowledgements. The financial support of this work by the
Ministry of Ukraine for Education and Science, by the
Volkswagen-Stiftung and by the INTAS-Ukraine Program (project
95-0186) is gratefully acknowledged. We thank also O.L. Fedorovich
for help in the preparation of the typescript.
References
1. Tringides, M.C. (00.), Surface Diffusion: Atomistic and
Collective Processes, Plenum, New York.
2. Bonzel, H.P. (1977) in Vanselow, R. and Tong, S.Y. (eds),
Chemistry and Physics of SolirI Surfaces CRC Press, Cleveland,
Ohio. p. 87.
3. Bonzel, H.P. (1990) in Mehren, H. (00.), Diffusion in Solid
Metals and Alloys, Landolt-Bornstein, Group III: Condensed Matter,
Vol. 26. Springer-Verlag, Berlin. p.717.
4. Kellogg, G.L. (1996) Atomic view of cluster diffusion on metal
surfaces, Progr. Surf. Sci. 53, pp. 217-223.
EXPERIMENTAL STUDY OF SURFACE DIFFUSION 17
5. Loburets, A.T., Naumovets, A.G. and Vedula, Yu.S. (1997) Surface
diffusion and phase transitions in atomic overlayers, in Tringides,
M.C. (ed.), Surface Diffu sion:Atomistic and Collective Processes.
Plenum, New York. pp. 509-528.
6. Seebauer, E.G. and Allen, C.E. (1995) Estimating surface
diffusion coefficients, Progr. Surf. Sci. 49, pp. 265-330.
7. Bayat, B. and Wassmuth, H.W. (1983) Directional dependence of
the surface diffu sion of Potassium on Thngsten(112), Surf. Sci.
133, pp. 1-8.
8. Loburets, A.T., Senenko, N.B., Naumovets, A.G. and Vedula, Yu.S.
(1995) Sur face diffusion of strontium on the molybdenum(112)
plane, Phys. Low-Dim. Struct. 10/11", p. 49.
9. Naumovets, A.G., Paliy, M.V., Vedula, Yu.S., Loburets, A.T. and
Senenko, N.B. (1995) Diffusion of Lithium and Strontium on Mo(112),
Progr. Surf. Sci. 48, pp. 59 70.
10. Naumovets, A.G. and Vedula, Yu.S. (1985) Surface diffusion of
adsorbates, Surf. Sci. Rep. 4, p. 365.
11. Loburets, A.T. (1999) An experimental technique for recording
adsorbate surface concentration profiles based on local
measurements of the contact potential, Met allofiz. Nov. Tekh. 21,
pp. 43-46.
12. Philibert, J. (1991) Atom Movements. Diffusion and Mass
Transport in Solids. Les Editions de Physique, Les Ulis. p.
11.
13. Loburets, A.T., Naumovets, A.G. and Vedula, Yu.S. (1998)
Diffusion of dysprosium on the (112) surface of molybdenum, Surf.
Sci. 399, pp. 297-304.
14. Loburets, A.T. (1999) Surface diffusion and phase transitions
in copper overlayers on the (112) surfaces of molybdenum and
tungsten, Metallofiz. Nov. Tekh.21, pp. 47-51.
15.. Vedula, Yu.S., Loburets, A.T., Lyuksyutov, I.F., Naumovets,
A.G. and V. V. Poplavsky, V.V. (1990) Surface diffusion and
interaction of adsorbed particles of electropositive elements on
refractory metals, Kinet. Catal. 31, pp. 270-288.
16. Loburets, A.T., Senenko, N.B., Vedula, Yu.S. and Naumovets,
A.G. (to be pub lished) in Tringides, M.C. and Chvoj, Z. (eds.),
Collective Diffusion on Surfaces: Collective Behavior and the Role
of Adatom Interactions. Kluwer, Dordrecht.
17. Medvedev, V.K., Naumovets, A.G. and Smereka, T.P. (1973)
Lithium adsorption on the (112) face of tungsten, Surf. Sci. 34,
pp. 368-384.
18. Medvedev, V.K. and Yakivchuk, A.1. (1975) Ukrainskii Fiz.
Zhurnal20, p. 1900. 19. Gupalo, M.S., Medvedev, V.K., Palyukh, B.M.
and Smereka, T.P. (1979) Adsorption
of lithium on the (112) face of a molybdenum crystal, Fiz. Tverd.
Tela 21, 973; English translation: (1979) Sov. Phys. Solid State
21, 568 (1979).
20. Braun, a.M. and Medvedev, V.K. (1989), Usp. Fiz. Nauk 157, pp.
631--{)66; English translation: Interaction between particles
adsorbed on metal surfaces, Sov. Phys. Uspekhi 32328-348
(1989).
21. Gonchar, F.M., Medvedev, V.K., Smereka, T.P. and Babkin, G.V.
(1990) Dyspro sium adsorption on Mo(112), Fiz. Tverd. Tela 32, pp.
1872-1875; English transla tion: Sov. Phys. Solid State 32 (1990)
1092.
22. Fedorus, A., Kolthoff, D., Koval, V., I. Lyuksyutov, I.,
Naumovets, A.G. and Pfnuer, H. (2000) Phase transitions in the
adsorption system Li/Mo(112), Phys. Rev. B 62, pp. 2852-2861.
23. Fedorus, A., Koval, V., Naumovets, A.G. and Pfnuer, H. (2001)
Fizika Nizkikh Temperatur (Low Temperature Physics) 27 (to be
published).
24. Kellogg, G.L. (1994) Field-ion microscope studies of
single-atom surface diffusion and cluster nucleation on metal
surfaces, Surf. Sci. Rep. 21, pp. 1-88.
25. de Gennes, P.-G. (1979) Scaling Concepts in Polymer Physics.
Cornell University Press, Ithaca, New York. Ch. 8.
26. Bowker, M. and King, D.A. (1978) Adsorbate diffusion on single
crystal surfaces I. The influence of lateral interactions, Surf.
Sci. 71, pp. 583-598; Bowker, M. and King, D.A.· (1978) Adsorbate
diffusion on single crystal surfaces II. Extension to next nearest
neighbor interactions, Surf. Sci. 72, pp. 208-212.
18 A. T. LOBURETS ET AL
27. Ala-Nissila, T. and Ying, S.C. (1992) Theory of classical
surface diffusion, Progr. Surf. Sci. 39, pp. 227-323.
28. Lyuksyutov, I., Naumovets, A.G. and Pokrovsky, V. (1992)
Two-Dimensional Crys tals. Academic Press, Boston.
29. Lyuksyutov, I.F., Naumovets; A.G. and Vedula, Yu.S. (1986) in
Trullinger, S.E., Zakharov, V.E. and Pokrovsky, V.L. (eds),
Solitons, Elsevier, Amsterdam. p. 605.
30. Lyuksyutov, LF., Everts·H.-U. and Pfniir, H. (2001) Diffusion
in a strongly corre lated anisotropic overlayer, Surf. Sci. 481,
pp. 124-134.
31. Gomer, R. (l~O) Diffusion of adsorbates on metal surfaces, Rep.
Progr. Phys. 53, pp.917-1002.
32. Loburets, A.T., Naumovets, A.G., and Vedula, Yu.S. (1982)
Surface diffusion of Lithium on (001) face of Tungsten, Surf. Sci.
120, pp. 347-366.
33. Stolzenberg, M. (1992) Zweidimensionelle Phasenubergiinge in Cu
und Te Adsor batsystemen, Dissertation, Physikalisches Institut,
TU Clausthal, Kapitel 7.
34. Ferrando, R., Spadacini, R., Tommei, G.E. and Caratti, G.
(1994) Correlation func~
tions in surface diffusion - The multiple-jump regime, Surf. Sci.
311, pp. 411-421. 35. Wang, S.C., Kiirpick, U. and Ehrlich, G.
(1998) Surface diffusion of compact and
other clusters: Ir", on Ir(111), Phys. Rev. Lett. 81, pp.
4923-4926. 36. Tsong, T.T. (2001) Mechanisms of surface diffusion,
Progr. Surf. Sci. 67, pp. 235
248. 37. Naumovets, A.G. and Zhang, Zh. Fidgety particles on
surfaces: How do they jump,
walk, group, and settle in virgin areas? Surf. Sci. 500 (to be
published). 38. Zgrablich, G. (1997) in W. Rudzinski, W. A. Steele
and G. Zgrablich (eds) Equi
libria and Dynamics of Gas Adsorption on Heterogeneous Solid
Surfaces. Elsevier, Amsterdam. p. 373.
39. Loburets,' A.T., Naumovets, A.G. and Vedula, Yu.S. (1999) in R.
Kutner, A. Pekalski and K. Sznajd-Weron (eds\ Anomalous Diffusion.
.From Basics to Applications. Springer, Berlin. p. 340.
40. Loburets, A.T., PaIiy, M.V., Senenko, N.B., Vedula, Yu.S. and
.Naumovets, A.G. (1999) Processes of self-organization in the zone
of surface diffusion of metal adsor bates on metals, Functional
Materials 6, pp. 194-202.
41. Goldenfeld, N. and Kadanoff, L.P. (1999) Simple lessons from
complexity, Science 284, pp. 87-89.
42. Bouchaud, J.P. and Georges, A. (1990) Anomalous diffusion in
disordered media statistical mechanisms, models and physical
applications, Phys. Rep. 195, pp. 127 293.
43. Kutner, R., Pekalski, A. and Sznajd-Weron K. (eds.) (1999)
Anomalous Diffusion. Prom Basics to Applications. Springer,
Berlin.
44. Metzler, R. and Klafter, J. (2000) The random walk's guide to
anomalous diffu sion: A fractional dynamics approach, Phys. Rep.
339, pp. 1-77.
45. Katrich, G.A., Kliniov, V.V. and Yakovkin, LN. (1994)
Interrelation between atomic and electronic structures of
alkaline-earth adlayers on MO(112) and Re(1010), J. Electr. Spectr.
Rel. Phenom. 68, pp. 369-375.
46. Dowben, P.A. (2000) The metallicity of thin films and
overlayers, Surf. Sci. Rep. 40, pp. 151--:245.
AB INITIO MODELING OF FREE ENERGY PROFILES IN THERMALLY ACTIVATED
PROCESSES
1. STICH CCMS, Dept. of Physics, Slovak Technical University (FEI
STU), Ilkovicova 3, SK-812 19 Bratislava, Slovakia
M.HYTHA National Center for High Performance Computing, Hsinchu,
Taiwan
Institute of Physics· of the Czech Academy of Sciences,
Cukrovarnicka 10, 162 53 Prague 6, Czech Republic
J. D. GALE
Department of Chemistry, Imperial College, London SW7 2AY, United
Kingdom
K. TERAKURA JRCAT, Angstrom Technology Partnership, 1-1-4 Higashi,
Tsukuba,Ibaraki 305-0046, Japan
CREST, Japan Science and Technology Corporation, Kawaguchi, Saitama
332, Japan
AND
M. C. PAYNE Cavendish Laboratory, University of Cambridge Madingley
Road, Cambridge CB30HE, United Kingdom
Abstract. The quantitative modeling of many surface processes, such
as diffusion or chemical reactions, requires accurate knowledge of
free energy profiles. The need to go beyond the internal energy is
especially impor tant in entropy-controlled processes which may
happen at both high (the thermally-activated regime) and low (the
quantum tunneling regime) tem peratures. We present results for a
thermally-activated process, namely, the formation of the first
intermediate in the methanol-to-gasoline process, catalyzed by
acidic zeolites. At high temperatures of 700 K, the entropic
contribution cannot be correctly evaluated in the harmonic
approximation
19
M. Kotrla et al. (eds.), Atomistic Aspects ofEDitaxial Growth.
19-29. © 2002 Kluwer Academic Publishers.
20 1. STICH ET AL.
and we use ab initio thermodynamic integration within density
functional theory. We find that, at reaction temperatures, the
entropic contribution qualitatively alters the free energy profile.
Different transition states are found from the internal energy and
free energy profiles. The entropic contri bution varies
significantly along the reaction coordinate and is responsible for
stabilizing the products and for lowering the energy barrier. An
outlook is given for a proper treatment of entropically-controlled
processes in both the thermally-activated and quantum
regimes.
1. Introduction
The modern atomic-scale theoretical description of surface
processes is typ ically based on the concept of potential energy
surfaces (PES). These give valuable insights into the mechanisms
involved in processes such as sur face diffusion, crystal growth,
epitaxial growth, chemical reactions, surface catalysis, etc. When
this approach is taken several nagging problems have to be
addressed, such as:
• the high dimensionality of the underlying PES • the need for
free, rather than internal, energy profiles at finite temper
atures • the need for highly accurate PES
The problem of high dimensionality is typically solved by applying
suit ably chosen periodic boundary conditions and limiting the
study of the PES to just one ora few cuts through the PES. The path
connecting two min ima on the PES, the "reactant" and "product"
well, is chosen for a given process by postulating the "reaction"
coordinate. The saddle point of this curve determines then the
"transition" state [1]. For many processes such an approach, based
on internal energy, is entirely adequate and sufficient. On the
other hand, there may be situations where the reaction coordinate
may take a more complicated shape that may be difficult to
determine in an ad hoc manner, so more sophisticated methods, such
as transition path sampling [2], have to be used.
In the case that the PES is shallow around the minimum energy path,
a more complete and wider sampling of the configuration space is
required. A shallow PES means more configurational possibilities
and a significant en tropy contribution to the free energy. Two
very different sources for this be havior can occur. At high
temperatures, in the thermally activated regime, the system may
have enough kinetic energy to explore large parts of the
configuration space. Alternatively, at low temperatures the system
may visit parts of the configuration space far from the. minimum
energy due to
AB INITIO MODELING OF FREE ENERGY PROFILES 21
quantum tunneling. In both cases the system may be driven into a
regime where its behavior is no longer determined by internal
energy. In the oppo site limit, the behavior of the system may be
predominantly determined by entropy. The most complicated scenario
is represented by the case where both internal energy and entropy
are of comparable importance. Examples will be shown in the next
section. In such a case the customary assumption of the dominance
of internal energy in transition state theory is not valid and the
free energy profiles cannot be approximated by internal energy
profiles. As discussed below, the wider sampling of the
configuration space on the PES can be accomplished by methods of
computational statistical mechanics, such as molecular dynamics
(MD) or Monte Carlo methods.
In the discussion above we have tacitly assumed that the PES is
known with absolute accuracy. This may not be the case in practice,
as the shallow PES implies that we are dealing with a weakly
corrugated PES,around the minimum energy path. This issue is
important, firstly, because the internal energy contribution is
significant and, secondly, because the accuracy of the entropy
contribution is also computed by sampling the internal energy PES.
Even seemingly very simple processes, such as diffusion of an
adatom on a surface, involve the multiple breaking and forming of
chemical bonds, a process which is rooted in electronic structure.
Given the necessity of statistical sampling, the underlying PES is
almost invariably computed using mean-field methods for electronic
structure, such as Hartree-Fock or methods based on density
functional theory (DFT) [3]. The former has exact exchange but
lacks any correlation; the latter treats both exchange and
correlation at an approximate level. The DFT technique which we use
below in most cases correctly describes the minima on the PES but
the accuracy of the description of energy barriers is usually
lower.
This discussion points to the enormous complexity of the energy
profiles which are the basis for the atomic-scale understanding of
complicated sur face processes, such as epitaxial growth. In the
next section we give more details on computing the energy profiles.
An application of these techniques to a very complex catalytic
reaction, the methanol to gasoline conversion (MTG) catalyzed by
acidic zeolites [4], will highlight the differences be tween the
different approaches. The conclusions will give a broader view on
the subject and an outlook into the possible further
developments.
2. The Energy Profiles
Until recently, the ab initio computation of PES or energy profiles
for com plex systems was considered not feasible. Only in the last
decade has ex tensive and reliable ab initio modeling of PES and
energy profiles started to shed light on surface processes on an
atomic scale. The most straight-
22 1. STICH ET AL.
forward estimate of energy profiles is based on internal energy. In
a more sophisticated approach, the entropy correction along a
reaction coordinate ~ in the configuration space can also be
estimated.
Once the reaction coordinate ~ is chosen the thermodynamic
potential must be determined along ~. At any finite temperature the
proper ther modynamic potential is the free rather then the
internal energy. In the simplest approach transition state theory
[1] is used with the free energy profile approximated either by the
internal energy or with entropy esti mated from the internal
energy in the harmonic approximation. However, as shown below, such
an approach may not be sufficient if both internal energy and
entropy are of comparable importance, and where the shallow PES
precludes treatment of the entropy in the harmonic
approximation.
The free energy profile beyond the harmonic approximation can be
com puted by thermodynamic integration using the "Blue Moon"
ensemble [5]:
b.F(~, T) = J: (>'~)~o,Td{O , (1)
where R refers to reactants and P to products, >.~ is the
Lagrange multi plier fixing the system at a given point ~o of the
reaction coordinate ~, and the thermodynamic averaging O~o,T is
performed using methods of com putational statistical mechanics
such as MD or Monte Carlo methods. We note that formula (1) is
correct only for a constraint consisting only of a distance. The
considerations here will be limited to this case. The formulae for
more general cases can be found in Ref. [6].
This is a well known approach which, however, has not been applied
very often to very complex systems [7]. One way of computing >.~
is by using constrained MD by adding a holonomic constraint to the
Lagrangian generating the MD
(2)
where the first term on the right-hand side of Eq. (2) is the
kinetic energy of the N ions, U({ri}) is the many-body PES, and ~o
is the externally fixed value of the constraint.
There is one potential problem with the application of formula (1)
to complex systems such as chemical reactions with a simple control
by one distance constraint ~. In a chemical reaction, typically one
chemical bond is broken and a new one is formed while all degrees
of freedom, except for that constrained by ~, must be in
equilibrium along the reaction coordinate to give the correct free
energy. For example, up to the transition state, the reaction A +B
- C ~ A - B + C can be controlled by constraining the
AB INITIO MODELING OF FREE ENERGY PROFILES 23
distance between A and B with all other degrees of freedom in
equilibrium and with the process being reversible. However, past
the transition state it may not be possible to control the distance
between B and C by the constraint ~ corresponding to the distance
between A and B. Hence, in such a case, the distance between A and
B no longer corresponds to the reaction coordinate and the free
energy cannot be obtained from Eq. (1).
If the number of degrees of freedom is small and the process
studied sim ple, an educated guess can be made for the reaction
coordinate connecting the reactant and product wells. However,
there are known examples where this approach may not lead to
correct results and ~ so constructed may not correspond to· the
true reaction coordinate. An example will be shown in Sec. 3. Such
behavior is symptomatic of systems with complicated high
dimensional transition states [2] which may not be known a priori.
An elegant solution to this problem, the so called transition path
ensemble, was proposed recently by Chandler et al. [2]. The only
technical problem with this promising approach is its additional
computational cost. To our knowledge, the method has not yet been
applied to a complex process in an ab initio fashion. Hence, the
approach we apply here, the method of ther modynamic integration
with an educated guess for the reaction coordinate ~, goes mid-way
between the customary approximations based on transi tion state
theory and the one-dimensional internal energy surface, and a
transition-path-ensemble approach.
3. Application to a Chemical Reaction
In order to demonstrate the above points we nowapply the
above-mentioned techniques to a complex catalytic reaction, namely,
the catalytic conversion of methanol to hydrocarbons catalyzed by
acidic zeolites. This is one of the most studied industrial
applications of zeolites in current commercial pro duction [4].
The whole process involves a number of steps [8, 9, 10, 11],
namely: (i) the initial methanol adsorption; (ii) activation of the
adsorbed species; (iii) dehydration to dimethyl ether (DME); (iv)
formation of the C-C bond. We focus here on the formation of DME in
the zeolite under reaction conditions (high temperatures and high
methanol loadings) [11] to highlight the importance of a proper
treatment of the entropic contribution.
We consider here the "direct" pathway [8, 12]
CH3 - OHt + CH3 - OH+ ZO- -4 CH3 - 0 - CH3 +ZO- +H20 +H+ (3)
where both methanol molecules react with each other inside the
zeolite en vironment, which acts merely as a solvent. In reaction
(3) Z stands for the zeolite. We assume here that one of the
methanol molecules to be proto-
24 1. STICH ET AL.
Figure 1. The model for DME formation. Four methanol molecules are
loaded in the 8-ring. Molecule # 2 undergoes spontaneous
protonation and forms a methoxonium cation (CH3-0H;). The holonomic
constraint eis applied to oxygen #1 and carbon #2.
nated as the proton transfers from an active site occurs
spontaneously at higher methanol loadings [9, 10].
The commercial zeolite catalyst Z8M-5 has a unit cell with ~300
atoms, which is too large for the present simulations to be
practical [13]. For that reason the simulations were performed in
ferrierite [14], which has a much smaller unit cell of only 54
atoms, but a structure very similar to that of Z8M-5. The
ferrierite structure is the closest mimic to the Z8M-5 structure we
were able to find. Only one active site (the H-compensated Al
defect) was considered. The reaction conditions have been simulated
by loading four methanol molecules into the 8-ring channel and
associated intersection regions of ferrierite (Fig. 1). The system
was prepared so that two methanol moiecules (# 1 and # 2 in Fig. 1)
can react along the 8N2 pathway. The postulated reaction coordinate
~ is also shown in Fig. 1.
The temperature in the simulation was taken to be at a temperature
of 700K. This system was shown to form strongly activated
methoxonium species [10]. The activated species are expected to be
susceptible to a nu cleophilic attack by another methanol molecule
to follow the reaction (3). The ability of this system to exhibit
activation makes it a strong contender for the present
purpose.
AB INITIO MODELING OF FREE ENERGY PROFILES 25
>' 4.0 .!!.
'".!I! El 3.0 ElotCDc W .Q c. 2.0 Fec w Ii
1.0 TSl!! LL
1.0 3.0 4.0
Figure 2. The variation along the reaction coordinate { of the free
energy profile li.F; total energy profile li.Etot , and the entropy
contribution TS. The zero of the vertical scale is arbitrary.
(a)-(d) label the configurations shown in Fig. 3.
Ab initio MO simulations [3] have been performed for the formation
of OME. All technical details of our simulations are as described
in Ref. [10]. It suffices to say that simulations were run in the
(N, V, T) ensemble using OFT in its plane-wave pseudopotential
formulation. Gradient corrected functionals are required for an
accurate description of the OME formation and we use the PW'91 [15]
variant of the GGAapproximation to OFT. We use norm-conserving
pseudopotentials to represent the core electrons and the wave
functions of the valence electrons are expanded in plane waves at
the r point of the supercell with a cut-off of 40 Ry. The accuracy
of OFT in the present GGA approximation was extensively tested
previously [8, 10, 16]. It was found that it yields excellent
equilibrium methanol geometries and harmonic frequencies, proton
affinities, quartz formation energies, etc.
The Lagrange multipliers along the reaction coordinate, taken to be
the distance between the C atom on the methoxonium cation and the 0
atom on the other methanol molecule (Fig. 1), required to compute
the free energy profile from formula (1), were evaluated at 10
different values of ~o.
The computed free energy profile, the total energy
(4)
are shown in Fig. 2.
26 1. STICH ET AL.
Figure 9. Ball and stick model with superimposed valence electronic
charge densities for points (a)-(d) along reaction coordinate
defined in Fig. 2. The electronic charge density is shown on a
plane defined by the oxygen # 1, carbon # 2, and the Al defect
(Fig. 1).
To give a better insight into the reaction process, we show in Fig.
3 char acteristic configurations sampled from the MD trajectories
of the reacting molecules in (a) the reactant well, (b) and (c)
near the transition state, and (d) in the product well. Additional
understanding of the complexities of this reaction, including the
mobility of the zeolitic proton, reactants, and products can be
obtained from a computer graphics animation [17].
Given the fact that a very simple form of the reaction coordinate
~
was assumed with a single applied constraint to control the
reaction, it is important to assess the correctness of this choice.
Two processes take place as the system climbs the reaction barrier.
With the applied constraint a chemical bond is enforced between the
oxygen on the methanol # 1 and the carbon atom on the methoxonium
ion # 2 (Fig. 1). During the reaction (3) methoxonium is
dehydrated, which breaks the HaC - OHt bond. However, these two
processes do not take place simultaneously. We find that the water
molecule from the CHa - OHt complex dissociates near ~ ~ 2.38 A,
before the other (e-O) bond is formed around ~ ~ 2.0
AB INITIO MODELING OF FREE ENERGY PROFILES 27
A. The global maximum/saddle point corresponds to the transition
state from the Etot (~) profile. Hence, there is no cOmpetition
between breaking and forming chemical bonds. In particular, for ~
< 2.38 A the dissociated water is not taking any active part in
the DME formation and comes to equilibrium by optimizing the
alignment of its dipole moment. Otherwise, at least two constraints
would be required. In order to check the reversibility of the
process, tests have been made around the maximum/saddle point.
Hence, the thermodynamically stable reaction path is the one given
in Fig. 2 and our choice of the reaction coordinate ~ is
meaningful.
From Fig. 2 we see that the total energy and free energy profiles·
along the reaction coordinate differ appreciably even at a
qualitative level. In par ticular the entropic contribution to the
reaction barrier is of the same order of magnitude as the internal
energy contribution and, hence, any conclu sion reached without
explicitly including the entropy will be incorrect. This finding
may not appear surprising at T = 700K. However, to the best of our
knowledge, the huge nonuniform entropic corrections have never been
properly treated in the theoretical modeling of the MTG process and
is ignored in the study of most other processes.
The main findings from Fig. 2 can be summarized as follows:
1. The transition states deduced from the F(~) and Etot(~) profiles
do not coincide, hence different triggering processes for the
reaction are deduced from F(~) and Etot(~).
2. The total energy curve Etot(~) shows a local minimum close to
the transition state.
3. The entropy profile S(~) varies considerably along the reaction
coordi nate.
4. The minima of F(~) and Etot(~) curves do not coincide, hence
they yield different reactant and product equilibrium
geometries.
5. In contrast with the result for total energy, the minimum on the
prod uct side of the free energy curve is significantly (~ O.5eV)
lower than on the reactant side, and hence is entropy
stabilized.
6. The free energy barrier is entropically lowered compared to the
inter nal energy barrier.
We now discuss these features in more detail. The internal energy
curve Etot(~) exhibits two activated processes; dissociation of
water from the
28 1. STICH ET AL.
methoxonium cation around { ~2.38 A and reaction of the methyl
group with the other methanol around {=2 A, separated by a minimum.
The lat ter process corresponds to the transition state from Etot
({). On the other hand, the transition state from the F({) profile
corresponds roughly to the former process of dissociation of water
from the methoxonium cation. This clearly shows that the customary
assumption of the dominance of the in ternal energy is not valid
and that a more complex statistical sampling of the internal energy
surface is required.
The sampling of the flat anharmonic multi-minima internal energy
sur face leads to the huge and nonuniform variation of the
entropic profile S({). As the entropy associated with the zeolite
catalyst is approximately con stant, the complicated S({) profile
can be understood in terms of elementary molecular processes
[11].
We are currently applying similar techniques to study the final and
most complicated step of the MTG process, namely, the formation of
the first C-C bonds. There are more possible mechanisms [18], but
we ~onsider the so-called formaldehyde mechanism, where the
formation of the e-C bonds proceeds via internal reconfiguration of
DME in the zeolite:
(6)
Contrary to the above process of DME formation, in this case we
find that the assumption of a simple reaction coordinate based on
application of one simple constraint (the C-C distance) is not
sufficient and either application of more constraints or the method
of Ref. [2] is required. Simulations with more constraints are now
under way.
4. Conclusions
We have discussed the calculation of energy profiles for complex
processes. Special attention was paid to entropically controlled
processes with com plicated reaction coordinates. The conversion
of methanol to gasoline cat alyzed by acidic zeolites was shown as
a demonstration of possible subtleties which may occur.
The approach adopted here combines the well-known technique of
ther modynamic integration, required to extract the entropy
contribution be yond the harmonic approximation, with ab initio MD
needed to sufficiently accurately describe the breaking/forming of
chemical bonds in the chemical reaction. The main complication with
this approach is the high computa tional cost. However, the class
of systems and processes with comparable entropic and internal
energy contribution and/or with complicated multi dimensional,
difficult to locate, transition states, including the quantum
entropy, is large, and the techniques of thermodynamic integration
[6] and
AB INITIO MODELING OF FREE ENERGY PROFILES 29
transition path ensemble [2J will play an increasingly important
role in a realistic study of complex systems.
References
1. See, for instance, Benett, C.H. (1977) Algorithms for chemical
computations, ACS series 46, (ed. Christofferson, R.E.), p.
63.
2. Dellago, C., Bolhuis, P.G., and Chandler, D. (1998) Efficient
transition path sam pling: Application to Lennard-Jones cluster
rearrangements, J. Chern. Phys. 108, pp. 9236-9245. .
3. See, for instance, Payne, M.C., Teter, M.P., Alan, D.C., Arias,
T.A. and Joannopou los, J.D. (1992) Iterative minimization
techniques for ab initio total-energy calcula tions, Rev. Mod.
Phys., 64, pp. 1045-1097.
4. Meisel, S.L., McCullogh, J.P., Lechthaler, C.H. and Weisz, P.B.
(1976) Chern. Tech nol., 6, 86.
5. Carter, E.A., Ciccotti, G. and Hynes, J.T. (1989) Constrained
reaction coordinate dynamics for the simulation of rare events,
Chern. Phys. Lett., 156, pp. 472-477.
6. Sprik, M. and Ciceotti, G. (1998) Free energy from constrained
molecular dynamics, J. Chern. Phys., 109, pp. 7737-7744.
7. See, for instance, Boero, M., Parrinello, M., and Terakura, K.
(1998) First principles molecular dynamics study of Ziegler-Natta
heterogeneous catalysis, J. Am. Chern. Soc., 120, pp.
2746-2752.
8. Shah, R., Gale, J.D. and Payne, M.C. (1997) In situ study of
reactive intermediates of methanol in zeolites from first
principles calculations, J. Phys. Chern., BIOI, llP. 4787-4797.
.
9. Stich, I., Gale, J.D., Terakura, K. and Payne, M.C. (1998)
Dynamical observation of the catalytic activation of methanol in
zeolites, Chern. Phys. Lett., 283, pp. 402-408.
10. Stich, I., Gale, J.D., Terakura, K. and Payne, M.C. (1999) Role
of the zeolitic environment in catalytic activation of methanol, J.
Am. Chern. Soc, 121, pp. 3292 3302.
11. Hytha, M., Stich, I., Gale, J.D., Terakura, K. and Payne, M.C.
(2001) Thermody namics of catalytic formation of dimethyl ether.
from methanol in acidic zeolites, Chern. Eur. J., 7, pp.
2521-2527.
12. Blaszkowski, S.R and van Santen, R.A. (1996) The mechanism of
dimethyl ether formation from methanol catalyzed by zeolitic
protons, J. Am. Chern. Soc., 118, pp. 5152-5153.
13. Ab initio MD simulations for ZSM-5 on a time-scale of the order
of '"1 ps have already been performed [9, 10]. Simulations for the
time-scales relevant for the present simulation should be possible
in due course.
14. Vaughan, P.A. (1966) Crystal structure of zeolite ferrierite,
Acta Cryst., 21, 983. 15. See, for instance, Perdew, J.P., Chevary,
J.A., Vosko, S.H., Jackson, K.A., Pederson,
M.R, Singh, D.J. and Fiolhais, C. (1992) Atoms, molecules, solids,
and surfaces applications of the generalized gradient
approximation for exchange and correlation, Phys. Rev. B, 46, pp.
6671-6687.
16. Shah, R, Payne, M.C., Lee, M.-H. and Gale, J.D. (1996)
Understanding the cat alytic behavior of zeolites: A
first-principles study of the adsorption of methanol, Science, B
271, pp. 1395-1397. /
17. A computer graphics animation of the simulation can be
downloaded from: http://kf-lin.e1f.stuba.sk/ccms/index.htm1.
18. See, for instance, Tajima, N., Tsuneda, T., Toyama, F. and
Hirao, K. (1998) A new mechanism for the first carbon-carbon bond
formation in the MTG process: A theoretical study, J. Am. Chern.
Soc. 120, ,pp. 8222-8229.
(1)
O.M.BRAUN Institute of Physics, National Ukrainian Academy of
Sciences, 03650 Kiev, Ukraine
Abstract. We study the diffusion of a particle in a two-dimensional
exter nal potential. Simulation results show that, in the
underdamped limit, the average jump length (>') scales with the
damping coefficient r, as (>') ex r,-u). with 1/2 ~ u)., ~ 2/3,
so that the diffusion coefficient behaves as D ex r,-u with 0 ~ U ~
1/3. We then introduce a realistic friction coefficient for the
phonon damping mechanism. The study of diffusion in this model
shows that long jumps play an essential role for diffusing atoms of
small masses, especially in two limiting cases: a large substrate
Debye frequency, when the rate of phonon damping is low, and a
sniall Debye frequency, when the one-phonon damping mechanism is
ineffective. As an applicati
LOAD MORE