HAL Id: jpa-00247094
https://hal.archives-ouvertes.fr/jpa-00247094
Submitted on 1 Jan 1995
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Elastic Interaction Between Modulated Steps on a Vicinal Surface
B. Houchmandzadeh, C. Misbah
To cite this version:
B. Houchmandzadeh, C. Misbah. Elastic Interaction Between Modulated Steps on a Vicinal Surface.
Journal de Physique I, EDP Sciences, 1995, 5 (6), pp.685-698. �10.1051/jp1:1995160�. �jpa-00247094�
Classification
Physics
Abstracts61.50Cj 05.40+1
68.55-aElastic Interaction Between Modulated Steps
on aVicinal Surface
B. Houchmandzadeh and C. Misbah
Laboratoire de
Spectrométrie Physique,
UniversitéJoseph
Fourier(Grenoble I)
lcCNRS,
B-P.87, Saint-Martin
d'Hères,
38402Cedex,
France(Received
28 November 1994, revised 14February1995, accepted
6 March1995)
Abstract. We
study
in this paper trie elastic interaction between modulated steps, in bothhomoepitaxy
andheteroepitaxy.
The influence of this interaction on the step fluctuation spec- trum is discussed. There are twoimportant
consequences that emanate from ouranalysis,
1)in the
homoepitaxy
case the elastic interaction favors a modulated stepprofile. However,
thegain
in the elastic interaction does notgenerally
overcome the fine tension effect andii)
moreimportantly,
for theheteroepitaXial
case the elastic contribution to the energy modulation is of the same order as that of the fine tension part.Elasticity
results in an effective fine tension(but singular),
which may be eitherpositive
ornegative depending
on thephase
shift between twoadjacent
steps. Thisanalysis
shows that the elastic contribution for non-zero wavevectorsshould be
incorporated
in any statisticaltheory
offluctuating
steps, if one wants ta accountproperly
forexperimental
observations. Wepoint
eut similarities between thisproblem,
and othersarising
inquite
different systems. We suggestthat,
mgeneral,
tensorial(dipolar
andmultipolar)
interactions betweenfines, surfaces,
etc. ofelectrostatic, magnetic, hydrodynamic,
etc. nature, should lead ta the same
generic
behaviour.1. Introduction
Interaction between
steps
on a vicinal surface ofcrystals
is along-standing subject.
It isa crucial
question
in manyimportant physical problems,
such as trieequilibrium shape
ofcrystals,
trie formation of domains oncrystal surfaces,
trie fluctuationspectrum
of steps, trie terrace width distribution atequilibrium,
etc.Steps
on vicinal surfaces are interface defectsleading
to elastic stress fields in trie bulk. Trielong
range elastic field ofsteps
interferes so that an effectivestep-step
interaction follows. Trieproblem
of interaction betweenstraight steps
in trie frarnework of continuumelasticity
bas been triesubject
of vanousinvestigations il, 2].
Trie mostimportant
result which follows from these studies is that trie interaction energy of twostraight
steps,separated by
adistance1,
is+w
1/é~
in triehomoepitaxy
case(1.e.,
when trie lastlayer
of triecrystal
bas trie saineequilibrium
structure as trie
bulk),
and+w lui in trie
heteroepitaxy
case(1.e.,
when trie lastlayers
bave a different structure than triebulk,
due to surface reconstruction ordeposition
of otherspecies).
Several authors used trie result for
straight steps,
in a mean-fieldapproach,
tostudy
both trie©
Les Editions dePhysique
1995fluctuation of
steps
atequilibrium
and terrace width distributions[3],
orstep dynarnics
mstep
flow[4, Si. Incorporation
ofelasticity
in trie total energy ofstep
modulations(1.e.,
for situationswith nonzero
wavevectors)
remains to beformulated,
and itsfar-reaching
consequences to beelucidated.
Dealing
with thesequestions
is trie main purpose of this paper. Thisrequires extending
trie work of Marchenko-Parshiniii
to modulatedstep profiles.
Whendealing
with thesequestions
we bave alsorecognized
that ourinvestigation
may be relevant to a wide list ofphysical systems,
asexplained
below.Let us outline trie main results to emerge from our
analysis. Firstly, contrary
to what could bave beenexpected
àpriori,
m triehomoepitaxial
case(where,
as we shall see, triestep
is alocation of force
doublets)
a modulation of triestep profile
results in a decrease of trie elastic energy. Aninspection
of trie order ofmagnitude
of trie elastic interactionshows, however,
that litre tension should inhibit such a(potential) instability. Secondly,
forheteroepitaxy (where
trie
step
is a location of elastic forces in contrast to forcedoublets)
trie elastic contribution to modes with nonzero wavevectors results in an effective linetension,
which can be eitherpositive
ornegative, depending
on triephase
shift between twoadjacent
steps.Moreover,
aninspection
of trie order ofmagnitude
shows that trie elastic contribution bas trie same order ofmagnitude
as trie litre tension. As a consequence, its contribution must not bedisregarded
when
dealing,
forexample,
withstep
fluctuations. We shall also see that trie elasticrepulsion
betweenstraight steps (for
q=
0,
where q is trie triewavevector)
may overcome, m somesituations,
trie dilfusive attraction whichusually
leads tostep-bunching dunng
sublimation. In otherwords, elasticity
may inhibit such aninstability.
Another
important
feature of this work is trierecognition
that trietype
of(potential)
insta- bilities causedby
elastic interactions mayarise,
m a somewhat similar way, in othersystems.
Trie
hydrated phospholipidic systems,
which exhibit a lamellarphase,
aretypical examples.
Indeed,
triedipolar
interaction between twolayers
is reminiscent of force-force interaction be- tweensteps.
It comes out that a modulation of triebilayers
red~ces trie energy, and may thus account for trie appearance ofrippled phases
in thesesystems.
We shall alsosuggest
some ex-perimental protocols
in othersystems (e.g., magnetic systems)
where trie interaction betweendipole
lines isexpected
to lead to a spontaneous modulation of trie hnes.Here is a brief survey of trie
organization
of trie paper. In Section2,
we formulate trieproblem
ofstep-step
interaction for anarbitrary step
modulation. We consider triehomoepitaxial
andheteroepitaxial
casesseparately.
In Section 3 we discuss trie order ofmagnitude
of trie elastic interaction and itsimplication
on trie fluctuationspectrum.
Section 4 is devoted to a discussion of trie results. Section 5 sums up our results andpresents
some outlooks. Details of trie basicformulation
together
with some technicalcomputations
arerelegated
toappendices.
2. Elastic Interaction of Modulated
Steps
To
study
triestep interaction,
we bave to establish a close distinction between hetero- andhomo-epitaxy.
In trie first case, trie atomsbelonging
to trie lastlayer
are not of trie same species as those in trieunderlying
substrate.Since these
questions
are notusually
substantiatedenough,
we bave felt it worthwhile to devote a detailed discussion to thisproblem.
Starting
fromelementary elasticity
we shall useconcepts
of force or force doublet distributionalong steps
to derivegeneral expressions
for trie interaction energy between two stepshaving arbitrary profiles.
We shall thenspecialize
our calculation to trie case of smallperturbations
of trie stepprofile.
This will be sullicient if one isonly
interested in trie linearstability analysis.
Let us discuss trie
heteroepitaxial
andhomoepitaxial
casesseparately.
_~~
t~
~Y
xf
Fig.
l. Flux of forcedipoles
across the step give rise ta force localizedalong
it.2.1. THE HETEROEPITAXY CASE. First we consider an adatom on trie substrate. This
adatom
locally
creates a force on trie substrate. Let f denote this force. Trie elasticdisplace-
ment created
by
trie force on trieplane
isgiven
in standard textbooks ofelasticity [6],
and conbe written as
U~°~~~(r) =
A(r)f, (1)
where r is a two-dimensional vector. A is a second-tank tensor
where1 is trie unit tensor,
rr~
trie second-rank tensor(r~rj),
E trieYoung
modulus and « trie Poisson coefficient.Equilibrium implies
that the total force feltby
the substrate atoms is zero.However,
ingeneral,
thedipole
is not zero.Thus,
the presence of an adatom is modelledby
a forcedipole.
More
precisely
in order tocompute
triedisplacement
inducedby
theadatom,
we bave to takea force distribution fz at distance az
(small compared
tor)
around trie adatom located at apoint
M(Fig. 1),
with trie total forceequal
to zero. fz is trie force exertedby
trie adatom atpoint
M on trie substrate atom 1in itsneighborhood.
Triedisplacement
at apoint
P(Fig. 1)
is
easily
obtained fromequation il)
to beu~~P°~~(r)
=(VA.az)fz
=VA.(azff)
= VA.D
(3)
where VA
=
(ôAzj /ôx~)
is a third-rank tensor and(D~~)
=(az,~ fi)
is a second-rank tensor ofdipolar
moments. It is understood thatrepeated
indices are to be summed over. Since D issymmetric iii
and does notdepend
on r, we can transformequation (3)
VA.D =
~~~JD~j)
=
(~~~J~J") =div(A.D) (4)
ôXv ôXv
To calculate trie total
displacement
atP,
we bave tointegrate
over aildipoles
in trie halfmonolayer
hmitedby
trie step.Applying
triedivergence theorem,
we findu~~P°~~(r)
=/ /dTdiv(A.D)
=/ dsA.D.n, (5)
s
where n is trie
in-plane
unit vectorperpendicular
to triestep. So,
triedisplacement
causedby
trie half
monolayer
isequivalent
to that of a force distribution f= Dn localized
along
triestep.
In trie
isotropic
case we considerhere, Djj
=Dôzj
and trie vector f = Dn isperpendicular
to triestep.
Inequation (5)
dT stands for trie surface element on trie halfmonolayer,
while dsdesignates
triearclength
elementalong
triestep.
Thus it appears that trie total elfect amounts to
considering
a force distributionalong
triestep.
This is not surpnzing.Indeed,
because trie Laméequations
are conservativeexpressing
trie fact that a volume interact
elastically
with itssurrounding only through
itssurface,
and thus trie total force must be conservative trieintegrated displacement
over trie halfmonolayer gives
a contribution
only through
trie contour, i-e-, the step, m terms of a forcedipole
distributionflux across the
step.
This amounts to forces localized on thestep.
In conclusion of thisdiscussion,
we reiterate that trie elasticproblem
reduces toconsidering steps
as locations of force distribution.A localized force
fi
on agiven step
creates an elastic field ui, whose expression isgiven by equation (1).
The interaction energy between twopoints belongmg
each to twoneighboring steps separated by
r reads(where
trie forces located at the twopoints
whose interaction isconsidered are
fi,
andf2)
E~°~~~
=
fiui (r)
=
f2u2(-r) (6)
where uz is the
displacement
induced at theapplication point
of fzby fj. Using expression (1),
we obtain
E~°~~~
=
~ ° °
fi
f2 +(rfi (rf2 (7)
XE r r
Note
that, fixing
r, trie minimum energy is reached when ail the three vectorsfi,f2,r
areparallel
withfi
"
-f2.
Forstraight steps
on a vicinalsurface,
all forces point to the samedirection and this
configuration
bas trie maximum energy.Modulating steps
modifies this resultand,
as we shall see, lowers trie energy underprecise
circumstances.Let us now
compute
trie totalstep-step
interaction energy. This isobviously
givenby mtegrating
E~°~~~ over the twostep profiles
~~
~~~~ ~~~~j~~~ / / ~81d82
~
~
(~l~2)
~~(~~l)(~~2)
,
(~)
~
where p is the
density
of localized forces atsteps (p
+w1la
where a is trie atomicspacing)
andf
triestrength
of localizedforces,
r trie distance between twopoints
at trie twoadjacent
stepsduring integration
and nz trie local normal to step 1.Expression (8)
is triegeneral
form of trie interactionirrespective
of the stepprofile.
Since our atm is toinvestigate
triestability
ofstraight steps against
infinitesimalperturbations,we
shall confine ourselves tocomputing
triestep-step
energy up to second order m trie deformation. Since Fourier modes do notcouple
to this
order,
it suffices to consider one Fouriercomponent. Moreover,
we shall allow for anarbitrary phase
shift betweensteps.
Moreprecisely,
twoneighbonng step
modulations aretaken to bave trie
following
form(see Fig. 2)
vi " e
cos(qxi)
v2" 1+
ecos(qx2
+#) (9)
The
quantities
r, n,etc.,
which appear mequation (8),
areeasily expressed
in terms of vi and y2. Trie calculation isstraightforward,
triealgebraic
details arepresented
inAppendix
A.We
find,
per unitlength
of triestep Ef~~~~P
=
~~
°~~~~~~~ (Ci In(ilL)
+(eli)~ il iqKi (iq) cos11 (10)
XE
where
Ci
"2a/(1- a)
andKi
is trie modified Bessel function of first order. L is trie steplength.
Figure
3represents
trie contributionproportional
to e~(that is,
we take trie energy origm thatcorresponds
tostraight steps)
inEf~~~~P
uersw
iq
for vanous values of#.
A few remarks2~
q Q
Fig.
2.Modelhng
of the elastic interaction of two modulated steps.2.5
2.0 4"K
1 5
k4 -Q
1 0
____,
__---~_--
_-- _'
_-' -'
"
Ô 5 "
,' é"~'
,
0.0 '
0 1 2 3 4
iq
Fig.
3. Elastic interaction energy of two modulated steps vernis the wave number q, for different values of thephase shift,
in theheteroepitaxy
case.are in order. From
Figure
3 one sees that a modulationautomatically
leads to an increase of trie elastic energy, since ail curves bavepositive
values. Forin-phase
fluctuations(#
=
0)
trie energy is an
increasing
function of q, while it is adecreasing
function for out-ofphase
fluctuations(#
=
x).
Moregenerally,
for#
>x/2
trie energydecreases,
and it increases otherwise.It is instructive to
expand
trie energy for small values of q. To second order in q, trie energy reads~~~
~~~~ ~~Î~~~~~ ~ Î~ ~
~~Î
~ ~~ ~ ~~~~~~~
~~~°~
Î
~~ ~~~~
where AE
designates
trie energyby taking
triestraight step
as a reference. Trieimportant
issue of trie
present
calculation is that trie energy behaves as q~lnq (to
becompared
with trie line tensionterm,
which behaves asq~).
Sinceqi
< trie factormultiplying
q~ bas a sign fixedby
that ofcos#.
For#
<x/2, elasticity
acts as apositive
linetension,
whereas for#
>x/2,
trie elastic elfect may bethought
of as an effectivenegatiue
line tension! If trie energymagnitude
werelarge enough
videinfra),
one would expect to be in a situation of spontaneoweiasticity-ind~tced moTphological instability
whennon-equilibrium phenomena
are involved. Trie basic reason is that trie(non-variational)
dilfusive mechanismduring
sublimation enforces trie#
= «-mode. As we shall seelater, according
to trieexploration
of variousregions
ofparameter
space made sofar,
it seems that diffusion of adatoms(which
isstabilizing during sublimation)
inhibits such a
potential instability.
A remark should be made at this
point. Equation (5)
shows that astep
can be modelledby
a distribution of localized forces. This is ageneral
result: D represents trieforce-dipole
inducedby
adatoms on trie lastlayer
and trie localized forces falong
triesteps
aresimply given by
D.n.Assuming
that D isisotropic,
we conclude that these forces areperpendicular
to triestep.
It ispossible
to extend this calculation to trieanisotropic
case. If trie exact form of D isknown,
we caneasily
deduce trie direction andstrength
of forces at trie steps.An
important example
isprovided by Si(001)
where dimerisation occurs on trie lastlayer.
There are two kinds of
domain,
say A andB, separated by
steps, and trie direction of dimeri- sation rotateby
90° bewteen consecutive domains.Taking
trie x-directionparallel
to one of trie dimensationdirections,
we can model Dby
[7]D =
dll °
(12)
0
dl
'This amounts to localized forces f
=
~(djj dl
atsteps
whose direction isopposite
from one step to trie next one. This situation contrasts with that considered above where D was taken to beisotropic.
Of course, ingeneral
ailcrystals
areanisotropic,
andstrictly speaking,
D isanisotropic accordingly.
The case ofSi(001)
isspec1al
in as much as theanisotropy
results inan alternation of the force direction between two
adjacent
steps due to the 1 X 2 and 2 X 1 reconstruction. The totalstep-step
energy isgiven by
the sameexpression
asequation (10) multiplied by
-1. This entails thatin-phase
step fluctuations will be enhancedby
elastic effects. This is trie case studiedby
Tersoff and Pehlke [8] who found indeed thatin-phase step
fluctuations were
unstable,
inagreement
with ouranalysis.
It must beemphasized, however,
that these authors confined theirstudy
to trie case of#
= 0. Dur calculation treats
arbitrary
fluctuations withregard
to triephase.
As isclearly
seen fromFigure
4 trie most unstableconfiguration
is obtained for#
= x and q =
0,
and not#
= 0 and q= q*
(where
q* is a finite value obtained from acompromise
betweenelasticity
and linetension;
seeFig. 4(b) ).
Otherwisestated, elasticity
leads to(deterministically straight) step-pairing. Therefore,
their conclusionregarding elasticity-induced morphological instability
in a realistic situation(1.e.,
where ailphases
arepermissible)
is called intoquestion.
We must however be somewhat cautious: theircalculation
(including
trie bulkresponse)
isfully anisotropic,
and our present remark would be conclusiveonly
after trie authors bave included anarbitrary phase
in their treatment.Before
proceeding farther,
let us now bave adigression.
Since trie interaction energy between two forces behaves as m1/r,
itsintegration
over triestraight steps diverges.
This is trieanalog
of trie interaction between twocharged
lines m electrostatics. This term is noteztensiue,
i.e., the interaction energy of Nsteps
grows as NlnN. Thislongstanding puzzle
is still a matter of debate.2.2. THE HOMOEPITAXY CASE. In this subsection we deal with
homoepitaxy.
The mostimportant point
is that the adatoms of the lastlayer
are of trie same nature as those of trie substrate. Besides triefact,
which holds for allsituations,
that trie total force exertedby
trie lastlayer
atoms on theirneighbors
is zero, here trie totaldipole
vanishes too. It should beemphasized, however,
that an isolatedadatom,
be it of trie same nature as trieunderlying
iq°
-05 ', ,, , .
, , , ,'
' ,
"»,__1_,,*"
' .~.5
.15
.15
0 1 2 3 4 0 1 2 3 4
1q lq
Fig.
4.Step
energy in trie case ofSi(001) (Solid
fine:#
= x; dashed fine:
#
=
0) (a)
elasticinteraction; (b)
elastic interactionplus
fine tension contribution.substrate or
not,
creates a non-zero forcedipole (see
lastsubsection).
Trie difference arisesonly
when one considers a halfmonolayer, Indeed,
in triepresent
case triemonolayer
is not constrained(because
there is nomisfit).
It behavesexactly
as trie bulk. We canpresent
oufreasoning by imagining
triecrystal
structure to be a result of energy minimization withrespect
to trie lattice spacing. Since trie half
monolayer
atoms are m their natural bulkposition,
this amounts to
requiring
trie forcedipole
to vanish. One can, if needbe,
convince oneselfby resorting
to a broken bond model.Thus,
an adatombelonging
to trie halfmonolayer
basa zero force
dipole.
Trie firstnon-vanishing
contribution comes then fromforce q~tadrupoles only.
A remark is in order. When we refer to trie naturalposition
of trie atombelonging
to trie lastlayer,
we bave in mind trieposition
in trieplane
of trie surfacecrystal.
Of course, trie verticalposition
is affectedby
surfacerelaxation,
which decreasesexponentially
with triepenetration distance,
ascompared
to triealgebraic decay
ofdisplacement
field due toin-plane
forces. The contributionstemming
from surface relaxation can thus beignored
for ailpractical
purposes [9].
Having
shown that an adatom on trie surface can be modelledby
a forcequadrupole,
weare in a
position
to determine triestep-step
elastic energy. Trieprocedure
followsexactly
thatpresented
above forheteroepitaxy.
Trie total elastic energy associated with triemonolayer
canbe written as
being
due toforce
do~tblets localizedalong
triestep,
since trie flux ofquadrupoles implies dipoles along
trieintegration
contour. This distibution of force doubletsperpendicular
to trie
step
is what wasinitially proposed by
Marchenko and Parshiniii
and rederived in amore
rigourous
mannerby
Andreev and Kosevitch [2].An
important
remark should be made. All we discussed here was about surface defects vithzero total moment.
However,
aspointed
outby
Marchenko and Parshiniii, steps,
as surfacedefects,
must bavenon-vanishing
total moments tocompensate capillary
forces. Consider alarge
radiusregion
around agiven step. Capillary
forcesacting
on trieedge
of thisregion
create a moment ~fa per unit
length,
where~f is trie surface tension of trie substrate and a trie
step height.
Intemalstress,
modelled as forcedipole
with non-zero total moment and peTpen-dic~tlar ta the
s~trface,
shouldcompensate
trie moment inducedby capillary
forces.However,
in triefollowmg,
weneglect
this effect.Actually,
we are interested in trie interaction energy of modulatedsteps,
as a correction to trie zeroth orderapproximation
ofstraight steps.
Triederivation of this non-zero total moment
dipole
was based oncapillary
forcesacting
on alarge
region around trie
steps,
and does notdepend
on themorphology
of triestep. Therefore,
weneglect
it in triefollowing.
Trie interaction energy between two force doublets
(each
on a differentstep) separated by
adistance r is
given by
~~~~~~~~ ~~ ~
~~~'~~
~~ ~Î ~~~~l)~
+(~Ll2)~j
+~)~(rLll)~(rLl2)~~
~~~°~
~~(rni)(rn2)(nin2)
+~°
~
(nin2)~ (13)
r r
For
complementary
information about detailsleading
to trie above expression seeAppendix
B(compare
withEq. (7)).
Note thathere, f'
is the forcedoublet, 1.e.,
bas trie dimension of a forcemultiplied by
alength.
For a fixed r, trie interaction energy is minimum whenfi'
andf2'
areparallel
and make anangle
with r. Thisangle
isgiven by:
cos~
=
~(
~
~
(14)
which is m
x/3
fortypical
values of a, a tz1/3.
This effect may favormodulating steps (see below).
To obtain trie total step-step interaction energy,
expression (13)
ismtegrated
over trie twostep profiles, exactly
as in trieheteroepitaxy
case. When wespec1alize
our expression to small deviations(with
one Fourier mode asbefore)
about triestraight configuration,
we obtainE~~~~~~
=
~~~
)j~Î~~'~~
+ ~~ ~
)j(Î'~~~~ 1(1 a)(6 (iq)~)+
l((1- 7a)(iq)~ a(iq)~) Ki(iq)+
(i a)(-3(iq)2
+(iq)4)K~(<q)j cosij. (is)
As
before,
p is trie number of doublets per unitlength
of trie step. The first term m the aboveexpression
is trie interaction energy of twostraight steps,
asgiven by
Marchenko and Parshiniii.
This is trie classicallli~ repulsion
betweensteps.
We are interested here m trie e~ term in thisexpression
whichcorresponds
to trie contribution due to trie modulation.Figure
5displays
trie behaviour ofAE~~~~~P (where
trie contribution fromstraight
steps lias beensubtracted)
as a function ofqi.
Astriking
feature of trie outcome of this calculation is thatelasticity
destabilizesstraight steps,
above a certain value ofqi
m 1- 3;(AE~~~~~P
becomes
negative).
This is a newqualitative result,
discovered here for trie firsttime,
to trie bestknowledge
of trie authors. This is asurprising
feature masmuch as twostraight steps having
trie samesign repel
each other. That is to say, contrary to what could bave apriori
beenexpected
on trie basis of trie well-knownrepulsion
of twostraight steps,
their modulationresults in a decrease of energy.
This result can
easily
berecognized
fromequation (15) by noting
that modified Bessel functions decrease asexp(-iq)
for q »1Ii,
so thatAE~~~~~P
+~
-q~
forlarge
values of q, 1-e-,AE~~~~~P
is adecreasmg
function of qElasticity
ateq~tilibr~~tm
should lead to aspontaneous
step modulation if it were
acting
alone.However,
anystep
modulation isaccompanied
witha line tension energy. Thus, a modulation can take
place only
if trie elastic energyamplitude
overcomes trie hne tension effect. In
fact,
as we shall see m trie nextsubsection,
trie hne tensionis
sufficiently large (typically
two or three orders ofmagnitude larger
than trie elasticenergy)
so that it
penalizes energetically
any modulation.We would like here to
anticipate by making
triefollowing important
remark. Trie expression of the interaction energy between forces(as
m the lastsubsection,
seeEq. (7))
or force doublets8 o.o
-0 2 4
~-0.4
~4~ ~ 0
0 6
-4 -0 8
-1 o -8
0 0 0 5 10 15 2 0 0 1 2 3 4
1f/ 1f/
Fig.
5. Elastic interaction energy of two steps m thehomoepitaxy
case versus the wavenumber,
for dilferent values ofphase shift, a) #
= 0;b)
dashed fine: #=
r/2,
solid fine:#
= r.(Eq. 13)
are very reminiscent ofdipole-dipole
andquadrupole-quadrupole
interactions in electrostatics.Therefore,
theimplication
of our results may begeneric
in the sense thatthey
con
apply
toelectric, magnetic, hydrodynamic,
etc.systems
as it will be documented in trie conclusion of triepresent
paper. Inparticular,
aninteresting system
is trie purephosphohpidic
membranes which may
organize
themselves for a certain range oftemperature
and watercontent m lamellar
phases
of theamphiphilic bilayers
where trie molecules have apolar
head
[11].
Triedipolar
interaction energy between two consecutivelayers
leads to a similarinstability [12],
which can manifest itself since such asystem
is devoid of a surface tension. We shall corne back to thispoint
later.Let us come back to expression
(15)
For small q, trie e~ term readsTaking
atypical
value of a m1/3,
weeasily
find that forcos#
<2/3,
1.e.approximately
#
>x/4, AE~~~~~P
is adecreasing
function of q for small q. Hereagain
one sees thatelasticity plays
the role of an effective line tension.2.3. ORDER OF MAGNITUDE OF ELASTIC INTERACTIONS. A
potential instability
causedby elasticity
may manifest itselfonly
if it iscapable
ofovercoming
the line tension elfect. It is therefore ofgreat importance
toprovide
orders ofmagnitude
of the elastic contribution. Let usconcentrate on the
small-q regime,
where the elastic contribution(due
to thestep modulation)
is +w
q~.
The q~prefactor
is our definition of theamplitude (which
weeventually
compare to the linetension).
Let us considerheteroepiatxy
andhomoepitaxy separately.
. In
heteroepitaxy,
it is clear fromequation (Il)
that thisamplitude,
denotedby Ai
is givenby
Ai
+w
~(~~ (17)
and
f designates
a force. We cangive
a crudeapproximation
of thisamplitude
basedon a dimensional
analysis. f
is aforce,
anddepends
on trie mismatch between atomicspacing
in the surface and in triebulk, ôala.
Trieonly
way to construct a force is(with
the additional
assumption
of linearelasticity
whichimplies
thatf
+w
ôala) f
~-
(ôala)Ea~. (18)
Taking
p11la,
ôala
m3.10~~
and for E atypical
value forsilicon,
E1 10~~Pa,
a m 3À,
we obtainAi
m10~~~ J/m
The
question
of order ofmagnitude
is not documentedenough
in the literature. Alerhand et ai. [7] and Webb et ai. [13] used triespontaneous
formation of stress domams onSi(001)
to evaluate the force
strength
at steps. Theiranalysis yields pf
= 0.07
eV/À~. Using
this
value,
~ve obtainAi
m10~~~ J/m,
which is coherent with our dimensionalanalysis.
This
amplitude
is to becompared
to trie line tension. A dimensionalanalysis (a typical
cohesion energy, a fraction of an eV per an atomic
length)
leads to ~f+w
10~~° -10~~~ J/m.
This is consistent with the value measured
by
Alfonso et ai.[14]
from trieeqmhbrium
fluctuation spectrum onSi[111].
In trie case ofSi[001], however,
where reconstruction occurs, this value is much smaller. Trie value measuredby
Bartelt et ai.[15],
lv~hich is based on triestudy
of trie step fluctuation spectrum, is m trie range of1-10mev/Àfor
trie two
types
ofsteps, SA
andSB,
i-e-,10~~~ -10~~~ J/m,
which is very close to trieamplitude
of elastic interaction.. In
homoepitaxy,
trieamplitude
of elastic interaction is obtained fromequation 16) (we
recall that we are interested in the q~
term)
Ad
+~~jj/~ (19)
~vhere
f' designates
a force doublet. Note here that theinterstep
distance 1enters theamplitude,
~vhich is a consequence of triealgebraic
interaction(as opposed
to thelogarithmic one).
From a dimensional
analysis,
we obtainf'
+w
Ea~. (20)
Taking
atypical
stepspacing
1m 100 atornic spacing we obtainAd
~310~~~ J/m.
Alfonso et ai.
[14],
studied trie terrace width distribution on cleanSi(111)
surface.They give i~Ad
m10~~°
J m. For astep
spacingof100a,
we findAd
"10~~~ J/rn,
~vhich isnot too far from the value derived from dimensional considerations.
3. Discussion
The main result obtained here is
that,
at least in triehetero-epitaxy
case, values of trie hne tension and elastic interaction ofsteps
are close. This should bave an important influence on trieanalysis
of trie fluctuationspectrum
ofsteps
atequilibrium.
For
large wavelengths [10],
trie fluctuationspectrum
of a train ofsteps
readsw +~
~~~~
~~°~
~~
+
(A ie~~~
cos( +'i)q~. (21)
1
where we have
disregarded
theIn(iq)
whichorigmates
from trie elasticpart.
The two
important
results can be stated as follows:At low values
of1, however,
and for ont-ofphase fluctuations,
the energy cost islarger
than forin-phase
fluctuations. As aresult,
one istempted
toexpect
that for small 1thein-phase
fluctuation modesprevail.
In a such caseelasticity
results in an(effective) positive
line tension.2- The elastic term is
important only
for q <é~~.
In thisregime, elasticity
and line tensionare additive. For
#
>x/2, elasticity
acts as aneffective negatiue
fine tension. For such atype
of modeselasticity
may cause aspontaneo~ts modulation,
if it islarge enough.
Whether thistype
of mode may be relevant or not is atpresent
an openquestion
whichrequires
adetailed statistical
study
of afluctuating
train. For q Ȏ~~,
trie elastic term isunimportant,
as
represented by
trieexponential
decrease of trie Bessel function forlarge arguments.
Trieinterstep
distance éplays
trie role of a cut-offlength
below whichelasticity
is irrelevant withregard
tostep
modulations.Usually,
in theexperimental analyses
of thestep
fluctuationspectrum,
the elastic renormal- ization of the linetension,
which can attainlarge values,
is not taken into account. While thetheoretical
study
of a train fluctuation is aproblem
which isrelatively
hard to tackle even atequilibrium,
it is animportant
task for futureinvestigations
toproperly incorporate elasticity
m trie model
equations.
4. Conclusion and Outlook
Ta
conclude,
we extended trieproblem
of elastic interactions to trie case of modulatedsteps
witharbitrary profiles.
Forhomoepitaxy,
we found thatelasticity
favors modulatedsteps!
This is a
striking
feature inasmuch as twosteps having
trie samesign repel
each other. Thisphenomenon
occursirrespective
of triephase
shift. While line tension shouldpenalize step modulations,
this result isinteresting by
itself on trieconceptual
level.Realizing
that thegeneral
form of trie force-force anddipole-dipole
interactionenergies
beara
strong
resemblance with electrostaticdipole-dipole
andquadrupole-quadrupole interactions,
we were
naturally
led to ask whether trie ideasput
forward here couldapply
to a wide list of systems. Ofparticular
interest seem to be triehydrated amphiphilic
systems which form for a certain range of water contents andtemperatures
lamellarphases,
where thebilayer
entities form a
parallel
array.Upon
variation of the centralparameter (e.g.,
thetemperature)
trie lamellae become
rippled [11].
TOdate,
trieorigm
of trierippling
transition bas remainedobscure,
albeit somekey ingredients
arebegmning
to emerge. Since trieamphiphilic
molecules bave apolar head,
it was natural to see whether triedipole-dipole
interaction betweenadjacent bilayers
may lead to an energy reduction due to a modulation. This is indeed what cames Dut frein Dur calculation[12].
Because trie membrane bas(virtually)
a zero surface tension(but
arigidity, giving
rise to a q~ contribution instead ofq~),
weexpect
triedipolar
energy to becomecompetitive
with therigidity
energy. While trie order ofmagnitude
of trieripple wavelength
which we calculated is smaller
by
one order ofmagnitude
than thatobserved,
this idea may open aninteresting
line of futureinquiry.
Other situations where our ideas may stimulate new progress concern a
variety
ofsystems, going
from electrostatics tohydrodynarnics.
Beside the above cited case, one can, forexample,
conceive of ways of
testing
our ideas onmacroscopic magnetic
systemsby setting
up expen-ments with lines
supporting magnetic
campasses. On the otherhand,
inhydrodynamics,
it isknown that two von Karmann vortex lines
[16]
aresusceptible
to modulations. Since thehy- drodynarnic equations
are,
for a two-dimensional
flow,
reminiscent of electrostaticequations,
it ispossible
to formulate trie line-fine interaction in triepresent
terms. Other variousexamples
may
be, probably,
found in trie case of vortex lines insuperconductors,
etc.Thus,
it appears thatanisotropic
interactions(of vectonal,
and tensonal think ofmultipoles origins)
mayembrace a
disparate variety
ofphysical systems.
Trie other line of
investigation reported
here was triediscovery
of trie relevance of trie elastic contribution to trie total energy modulation in trieheteroepitaxy
case. Future works on step fluctuations shouldincorporate
such an effect with trie aim tointerpret expenmental
resultsproperly.
Forphase
shiftslarger
thanx/2,
trie effect ofelasticity
isopposite
to that of trie linetension;
itplays
trie role of anegative
fine tension. It is notyet
clear whether inequilibrium
the modes with
#
>x/2
are relevant or not. Thisquestion
iscurrently
underinvestigation.
What
might
trie consequences in trie out-ofequilibrium regime
be? It is known thatduring step
flowgrowth
[17] thestraight step
may becomemorphologically unstable,
the mode with#
= 0 [18]being
the mostdangerous
one.During sublimation,
thestraight step profile
isstable,
while the vicinal surface sufiers a
step-bunching instability (mode
q= 0 and
#
=x).
Becauseduring
sublimation the difiusivephenomenon
enforces the mode#
= x, we may expect
elasticity (recall
that for#
= x itplays
the role of anegative
linetension)
to lead to amorphological mstability.
Theincorporation
of elastic efiects in the Burton-Cabrera-Frank(BCF)
[19] model results in a modified condition at the steps[12].
Sofar,
we have found thatduring sublimation,
diffusion(which
isstabilizing) precludes
an elastic inducedmorphological instabihty.
This isattributed to the fact
that, generally,
the diffusion "force" is muchlarger
than the elastic one.In
conclusion,
trie workpresented
here should serve as a basic framework for future inves-tigations dealing
with theproblem
offluctuating steps.
On the otherhand,
it has led to theidentification of the relevance of the
(tensorial)
interaction m other systems that lookextremely
diverse. What we have learnt so far is that these
types
of interactions(e.g., dipolar interaction).
familiar as
they
may appear, lead tounexpected
results.Appendix
AInteraction
Energy
of Two ModulatedSteps:
TheHeteroepitaxy
CaseWe derive here the interaction energy of two lines of forces
(Fig. 2). Step
modulations aredefined
by
vi " e
cos(qxi ),
v2" 1+
e
cos(qx2
+çi). (A.1)
Their interaction energy reads
(Eq.(7))
E~~~P =
~
°~~~~~
dsi ds2
°(nin2)
+(rni)(rn2)
,
(A.2)
XE
~
r r
where fz =
fnz,
and p is the forcedensity
atsteps.
For twopoints Pi (xi, vi), P2(x2,v2)
onsteps,
wedevelop
terms mintegral
at the second order m e. To shorten thenotation,
we use:(
"(Xi X2)/t,
1"(Ii, #
"
tq,
A
=
cas(qxi) cas(qx2
+ çS)=
cas(qx2
+#() cas(qx2
+ çS), II=
sin(qxi) sin(qx2
+ çS)=
sin(qx2
+#() sin(qx2
+ çS),Z
=
sin(qx2
+#()
+sin(qx2
+ çS),gi(i)
" 1(1 +
2)(1
+i~)~i~~~~/~ i(i
+i~)~i~~~~/~ (A'3)
Integration
of the first term inequation (A.2)
readsIl dsids2 ~~~~~~
r
=
/ / d(dx2 ((1+ (~)~~/~
+ ê~
(H(~(1+ (~)~~/~
+(1/2)A~gi(()j
= -L
In(t IL)
+ Lil fKi (#)
cas#] ê~. (AA)
where L is trie
step length.
Integration
of the second term inequation (A.2) gives
Surprinsingly,
ê~ terms inequation (A.5)
vanishexactly.
For thisterm,
the energy decrease due to modulationcompensates exactly the
energy increase due toaugmentation
of forces perprojected length.
Adding
these two contributions(A.4-A.5),
weget
theexpression given
inequation (10).
Appendix
BInteraction
Energy
of Two ModulatedSteps:
Tl~eHomoepitaxy
CaseHere we
compute expression là), using
the energy of two force doubletsgiven by equation (13).
We use notations of
Appendix
A. Trie interaction energy of two modulatedsteps
readsE~~~P
=
/ / dsi
d82E~°~~~~~
(fi, il, r). (B. Ii
Integration
of different terms ofequation (13) gives
(<2
IL) j j dsi ds~
r
= 2 +
(6
+ #~ +(f~K2(#) #~K3 (#)j
cas#)
ê~= 2 +
(6
+ #~(#~Ki (#)
+3#~K2 (f)j
casçi) ê~. (B.2)
(t~ IL) / / d81ds2 ((rni
)~ +
(rn2)~j
r
= 8
/3
+(24
+ 2@~ +(-4@~Ki (#) 4#~K2 (4)
+4@~K2(4)
+3
10#~K3 (#) 2#~K4(#)j
casçi)
ê~=
8/3
+(24
+2#~
+(-6#~Ki (#) 12#~K2(4)+
3
2#~K2 là))
casçi) i~. (B.3)
(i~ IL) / / dsi ds2 (rni
)~
(rn2)~
r
=