Published on

18-Dec-2016View

212Download

0

Embed Size (px)

Transcript

This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 134.151.40.2This content was downloaded on 29/01/2014 at 03:33

Please note that terms and conditions apply.

Equilibrium Stranski-Krastanow and Volmer-Weber models

View the table of contents for this issue, or go to the journal homepage for more

2009 EPL 86 16002

(http://iopscience.iop.org/0295-5075/86/1/16002)

Home Search Collections Journals About Contact us My IOPscience

April 2009

EPL, 86 (2009) 16002 www.epljournal.orgdoi: 10.1209/0295-5075/86/16002

Equilibrium Stranski-Krastanow and Volmer-Weber models

D. B. Abraham1 and C. M. Newman2

1 Rudolph Peierls Centre of Theoretical Physics, University of Oxford - Oxford OX13NP, UK, EU2 Courant Institute of Mathematical Sciences, New York University - New York, NY 10012, USA

received 11 December 2008; accepted 11 March 2009published online 16 April 2009

PACS 68.43.Hn Structure of assemblies of adsorbates (two- and three-dimensional clustering)PACS 68.35.Rh Phase transitions and critical phenomenaPACS 64.60.De Statistical mechanics of model systems (Ising model, Potts model, eld-theory

models, Monte Carlo techniques, etc.)

Abstract An equilibrium random surface model in 3d is dened which includes versions of boththe Stranski-Krastanow and Volmer-Weber models of crystal surface morphology. In a limitingcase, the model reduces to one studied previously in a dierent context for which exact results areavailable in part of the phase diagram, including the critical temperature, the associated specicheat singularity and the geometrical character of the transition. Through a connection to the 2dIsing model, there is a natural association with the Schramm-Loewner evolution that has alsobeen observed experimentally in a nonequilibrium deposition setting.

Copyright c EPLA, 2009

Introduction. In recent times, there has been arenascent interest in surface science promoted partlyby experimental methods like Scanning TunnelingMicroscopy (STM) and Atomic Force Microscopy(AFM) [1], which allow direct examination of the surfacestructure on an atomic length scale. Such structure is ofgreat relevance for understanding the technically signi-cant areas of heterogeneous catalysis and nanotechnologyat a fundamental level. At the same time, a system-atic theoretical treatment is emerging, in which exactstatistical-mechanics results have played a signicantrole [2]. In this letter, we address the general questionof whether equilibrium statistical mechanics can predictthe existence of mounds of molecules which are observedexperimentally on the substrate in certain systems. Theproblem is that on naive entropic grounds, the moundsmight be expected to dissociate into pieces of monolayerbecause of the greater available conguration space. Butthis neglects entirely excluded area eects.A notable conjecture about surface phase transitions is

due to Burton, Cabrera and Frank (BCF) [3]. Considera classical, low-temperature uni-axial ferromagnet (suchas the Ising system, or its lattice-gas analogues) atcoexistence. Let both equilibrium phases be present, butorganised in a symmetrical way into two dierent regionsof space separated by an interface or domain wall. BCFmade the bold conjecture that, because the interfacialregion is acted on by equal and opposite and thereforecancelling mean elds, the interface should undergoa transition like the bulk one in 2d, already found by

Onsager [4]. Later, using series expansions, Weeks, Gilmerand Leamy [5] gave compelling evidence that there is aphase transition, but not one of 2d Ising type. Rather,in the high-temperature phase the interface manifestslarge spatial uctuations which invalidate the customarythermodynamic picture of a sharp interface localised inlaboratory-xed axes. There has never been a proof of aroughening transition in the 3d Ising model, distinct fromthe usual transition between ferromagnetic and paramag-netic phases. It is known that the transition temperatureTR satises the inequality Tc(3) TR Tc(2), whereTc(d) is the usual critical temperature in dimension d.This result was established by an ingenious exact asso-ciation with the 2d Ising magnet [6]. On the other hand,exact results are available for associated Solid-On-Solid(SOS) systems showing both roughening and an essentialsingularity in the incremental free energy, quite unlike theBCF conjecture [7]. This raises the interesting question ofwhether the conclusions of the BCF scenario are realisedphysically in another type of surface phase transition. Itis the purpose of the present paper to answer this.To begin, we need to review three modes of crystal

surface growth: rstly, we have the Franck-van der Merwe(FM) mode in which adsorbate is added in completelayers [8]. The analogue at thermodynamic equilibrium iswetting [9]; this is normally characterised by sessile drops,contact angles and the Young rule. Although there areno exact results yet in 3d, the situation for the Isingmodel in 2d is well understood as is the equivalenceof divergent lm thickness and vanishing contact angle

16002-p1

D. B. Abraham and C. M. Newman

(properly dened) [10]. The second growth mode is theVolmer-Weber (VW) one [11]. In this, monolayer islandsof adsorbate of nite extent are formed, and then islandsgrow on top of islands and so on. The natural questionis this: are there equilibrium statistical-mechanical modelswhich show these turreted congurations, which haveindeed been observed experimentally [12]? The thirdgrowth mode is due to Stranski and Krastanov (SK) [13];it is a hybrid of the rst two, in which a nite number oflayers of FM type are laid down, followed by growth of VWtype. A signicant property of the turreted crystallites isthat they can manifest greatly enhanced chemical activitywhen compared with the bulk material, a matter ofconsiderable relevance in heterogeneous catalysis.We now detail the purpose of this paper more closely;

we will dene a three-dimensional surface model instatistical mechanics which contains cases of the VW andSK scenarios and for which exact results have becomeavailable, some of which are in agreement with the BCFpredictions. Before our detailed discussion, we pointout that our modelling aords another example of theSchramm-Loewner Evolution (SLE) [1418], a recentmathematical approach to interfaces in 2d systems whichis nding increasing relevance [19].

Equilibrium surface model. We now constructan equilibrium model analogous to the VW and SKgrowth models rst by prescribing allowed congurations,secondly assigning energies to such congurations andthirdly by giving them a Boltzmann weight. The modelrepresents the substrate as a at plane and inscribes onit a square lattice. Adatoms are physisorbed onto thesubstrate and, in the spirit of Kossel-Stranski [20,21],they are represented by cubes, the bottom faces of whicht exactly to unit cells of the underlying square lattice.From the outset, we exclude discommensuration and asso-ciated elastic phenomena because our theoretical methodsappear not to be useful in that case.Adatoms can occupy neighboring unit squares; such

congurations are energetically stabilized by assumingattactive interactions between neighboring adatoms. Onenergetic grounds, these adatoms tend to assemble intorafts. We assume that there are no holes in the raftssince such congurations would be energetically unstableagainst lling. Moreover, this restriction is essential forour theoretical analysis.The energy of a conguration of rafts conned to the

rst layer can be written as

E() = L()+ ( 0)A1(). (1)In the above equation, A1() is the total area of contactof the rafts with the substrate; by construction this is thesame as the area of the upper surface. The term A1() isthe surface energy of the upper surface while 0A1()is the binding energy of the rafts with the substrate.L() is the total length of the curve(s) formed by theintersection of the rafts with the substrate. itself is a

Fig. 1: Side view of four possible locations for adsorption of anadatom on a train of ledges.

collection of simple closed loops, = {1, . . . n}. Each jis the intersection with the substrate of plaquettes with thenormal parallel to the substrate plane. The term L() isthe total surface energy of these plaquettes.At this juncture, the substrate plane is in general

partially covered by molecular rafts. We now repeat thisprocess by placing additional rafts strictly on top of thosealready in place, once again with the exclusion of holesin the added rafts. By strictly on top, we mean thatvertical overhangs are forbidden. Thus the perimeter ofany raft in the second layer must be contained within theperimeter of some raft in the rst layer. We have departedby this point from the usual lattice gas models of wettingand lm thickening which allow internal holes.The remaining factor to be considered is the energetics

which results when elements of the perimeters of dierentrafts coincide. Ehrlich-Schwoebel [22,23] phenomenologystates that an adatom adsorbed on a terrace in a train ofledges is more likely to be adsorbed at the up-going ledgethan the down-going one. This is explained by activationenergy for the adsorption processes see g. 1.In our model, the placement of a molecule like that

numbered 4 in the gure is ruled out entirely, as wehave said above, because it is too energetic. Similarly, weregard molecule 3 in the gure as having higher energythan molecule 1, when not only neighboring, but alsonext-nearest-neighboring interactions are included. Thisis incorporated into the energetics described by (1) byinserting an additional term:

E() = L()+ ( 0)A1()+ 1N(1, 1), (2)where N(1, 1) counts the total number of coincident edgesand 1 is a positive energy. We note that molecule 1 inthe gure has higher energy than molecule 2 because ofthe L() term in (2). The specication of the modelthen allows further additions of molecular rafts on topof those already laid down. Unfortunately, as far as weknow, rather little of substance can be said about thisloop gas with congurational energy given by (2). Butif we take the limit 1, we recapture the Multi-Ziggurat (MZ) model [24], about which there is muchuseful information [25,26].At this juncture, it is worth making the following points:

rstly, what does our equilibrium statistical mechanicalmodel imply about whether a raft is added on the baseplane or on top of other rafts to make multilayeredstructures? Simple entropic arguments might favor theformer, but this matter can only be decided by detailed

16002-p2

Equilibrium Stranski-Krastanow and Volmer-Weber models

Fig. 2: A typical surface structure having three mounds abovethe substrate. The left one is a tower, the middle one is a dimerand the right one corresponds to a chessboard pattern of spinvalues assigned by the Peierls contours. The numbers indicateheights or the regions (or terraces).

investigation. Secondly, there is an entropic repulsionwithin nests of contours describing multilayered struc-tures, that is, within individual ziggurats as well as betweenthem; this is taken into account implicitly in what follows.To begin with, in the MZ model, the square lattice

underlying the conguration space no longer containsmore than singly occupied edges. The conguration spaceis dened uniquely in terms of the boundaries of the rafts,which are simple closed walks. According to the rulesof the model, dierent walks can meet at corners butthey cannot have common edges, because of the assumedlimit 1 of the ledge-ledge interaction, which has justbeen stipulated. Thus the congurations are the same asthose of the Peierls contours of the two-dimensional Isingferromagnet with xed parallel boundary spins, say allplus and their energy will be given by (1). An exampleis shown in g. 2.To make the matter quite denite, take the dual lattice

of the original square lattice ; this has vertices at thecenters of the unit cells of . At each such vertex i,place a spin (i) =1, and for i, take (i) =+1. Inthe interaction term ( 0)A1() for the conguration ,the quantity A1() satises

A1() =A()A(+()), (3)

where A denotes the total area and +() denotes, in

the language of Ising percolation theory, the plus clusterof the boundary of [26]. Evidently, we have arather curious object namely, a planar Ising model witha magnetic eld applied to a set of sites which is random,in that it is generated from the spin conguration.Note, though, the special case 0 = for which this eld

vanishes, giving the standard planar Ising model. Finally,an integer-valued height variable h(i) may be assigned toeach site i of . This height is the least number of Peierlscontours crossed in going from i to .

Mounds in the Multi-Ziggurat model. In thecase 0 = , it follows from Onsagers seminal work [4] thatthere is a phase transition with logarithmically divergentspecic heat on both sides of the transition. This transitionpersists for 0 , since then the substrate is covered with

a monolayer [10] thus aording a potential example of theSK scenario. For 0 = , there is no monolayer and one hasan example of the VW scenario.A diculty is to gauge the propensity to form mounds.

The probability that the surface at any point, say(0, 0), has at least height k is denoted by P (h(0, 0) k).It satises for 0 = , the bound [10],

P (h(0, 0) k) (1m)k, (4)where m is the spontaneous magnetization, given by

m =(1 (sinh())4)1/8 . (5)

For 0 > , because of the monolayer, the height h in (4)should be replaced by h 1. The droplet of the origin,denoted D, is dened as the connected cluster of sitesr with h(r) 1 which contains the origin. If h(0, 0) = 0,then D is the empty set for consistency. Thus, it followsfrom [24] that for 0 > , the mounds, or individualziggurats, are nite. It has also been shown that inthe thermodynamic limit, the height and therefore alsothe basal area of a typical mound, such as D, diverges asT Tc(2). Consequently, the phase transition impliesincorporating a macroscopically thick lm on the sub-strate in the high-temperature phase. Nevertheless, thelater work in [26] indicates that it is quite inappropriateto think of MZ as a satisfactory model of wetting. Firstly,above the critical temperature, the surface has a roofstructure, rather than the complete layer-by-layer oneassociated with equilibrium FM. Secondly, the constancyof the transition temperature is unnatural, as is thelogarithmic divergence of the specic heat on both sidesof criticality; neither of these results corresponds withaccepted ideas for wetting derived from RG and MCstudies [9].Our version of the SK and VW scenarios allows us to

study the thermal uctuations of two-dimensional domainwalls, both subcritically and at the transition point. In thelatter case, their statistics, in the scaling limit of vanish-ingly small lattice scale, should follow the SLE schemewith the parameter = 3 [14]. SLE() curves are randomcurves in the plane constructed out of a one-dimensionalBrownian motion via conformal m...