• Aucun résultat trouvé

The Response to an External AC Drive of the Interface Between Solid and Liquid Helium Near the Roughening Transition

N/A
N/A
Protected

Academic year: 2021

Partager "The Response to an External AC Drive of the Interface Between Solid and Liquid Helium Near the Roughening Transition"

Copied!
32
0
0

Texte intégral

(1)

HAL Id: jpa-00247106

https://hal.archives-ouvertes.fr/jpa-00247106

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.

The Response to an External AC Drive of the Interface Between Solid and Liquid Helium Near the Roughening

Transition

S. Giorgini, R. Bowley

To cite this version:

S. Giorgini, R. Bowley. The Response to an External AC Drive of the Interface Between Solid and

Liquid Helium Near the Roughening Transition. Journal de Physique I, EDP Sciences, 1995, 5 (7),

pp.815-845. �10.1051/jp1:1995170�. �jpa-00247106�

(2)

Classification Physics Abstracts

61.50Cj 64.00 67.80-s

The Response to

an

External AC Drive of the Interface Between

Solid and Liquid Helium Near the Roughening llYansition

S.

Giorgini

and R-M-

Bowley (*)

Department of Physics, University of Nottingham, Nottingham NG7 2RD, UK

(Received

19 January1995, received in final form and accepted 17 March

1995)

Résumé. La théorie de la transition rugueuse développée par Noziéres utilise un programme de renormalisation tronquée pour traiter de l'effet des fluctuations de l'interface liquide-cristal.

Pour une direction facettée, la renormalisation s'arrête à la longueur de corrélation

en l'absence d'une force extérieure appliquée à l'interface. Cependant, le processus de renormalisation doit être arrêté plus tôt s'il y a une longueur plus courte dan8 le problème. Nozières a examiné de tels effets de taille finie soit pour des surfaces vicinales, lorsqu'une force extérieure continue est

appliquée pour provoquer une croissance

(ou

une fusion) de l'interface à vitesse finie. Ici on considère l'effet d'une force extérieure alternative qui introduit une nouvelle échelle de longueur;

c'est la distance de diffusion d'une perturbation pendant le temps

w~~,

w est la pulsation de la force appliquée. Nous avons étendu les relations de récurrence pour traiter le cas de courant

alternatif. Pourvu que la longueur de correlation soit plus grande que les autres échelles de

longueur, nous pouvons résoudre les relations de récurrence et obtenir la réponse de l'interface,

une approche qui est exacte au deuxième ordre

en potentiel d'ancrage. Quand la longueur de corrélation est plus petite que les autres longueurs, nous résolvons les relations de récurrence pour des échelles de longueur plus courtes que la longueur de corrélation; pour des échelles de longueur plus grandes nous résolvons l'equation de mouvement renormalisée sans tenir compte du bruit aléatoire. Cette approximation grossière nous permet de décrire semi-quantitativement

le passage du comportement mobile au comportement ancré de l'interface.

Abstract. The theory of the roughening transition developed by Nozières uses the notion

of a stopped renormalisation scheme when dealing with the effect of fluctuations of the crystal- hquid interface. For a faceting direction, the renormalisation is stopped at the correlation length

in the absence of any externat driving force on the interface. However the renormahsation process needs to be stopped earher if there is a shorter length scale in the problem. Nozières considered such finite size effects arising from either vicinal surfaces or from the effect of an external DC driving force which

causes the interface to grow

(or

melt) at a finite velocity. Here we consider

the effect of an applied AC driving force which introduces a new length scale; this length is related to the distance a disturbance m the interface diffuses m a time w~~, where w is the

angular frequency of the applied drive. We have extended the recursion relations to deal with

the AC case. Provided the correlation length is forger thon the other length scales we con

solve the recursion relations and obtain the response of the interface, an approach which is

exact to second order

m the pmnmg potential. When the correlation length is smaller than

(*) Author to whom correspondence should be addressed

Q Les Editions de Physique 1995

(3)

one of trie other lengths, we solve trie recursion relations for length scales shorter than the correlation length; for longer length scales

we solve the equation of motion renormalised up to trie correlation

length

by ignonng trie random noise as being small. This crude approximation

allows us to describe

semi-quantitatively

the crossover from trie mobile to trie pinned behaviour of the interface.

l. Introduction

The dassic work of Nozières and Gallet

iii

on trie

dynamical broadening

of trie

roughening

transition was

developed

in

parallel

with trie beautiful

expenmental

work of

Gallet,

Balibar and

Rolley

[2] on hehum

crystals.

In trie present paper we extend trie

theory

to describe trie

response of trie interface between solid and

liquid

helium to

applied

AC drives.

First let us review trie

theory.

Consider trie interface between a

crystal

and its melt for a direction in which trie interface can be facetted. The

Langevin equation

for the motion of the interface

is, according

to Nozières and Gallet

(lj

J~~~='fV~z+4l+R+~

Here J~ is a friction coefficient which is related to trie

mobility,

K, of trie interface

by

trie relation

J~ = pc

/K

with pc trie

crystal density;

'f is trie surface

stiffness;

R is a random force with a white noise spectrum;

Fla

is trie

driving force,

related to trie difference in chemical

potential

across trie interface

Fla

=

pc(~IL ~lc)

and 4l is trie force that comes from trie

pinning potential (throughout

we follow trie notation and

techniques

used

by

Nozières

iii ).

Trie

pinning potential

bas to be

periodic

with

period

a, trie

spacing

of trie

crystal planes,

for if trie

crystal

grows

uniformly by

one lattice

spacing

trie

potential

is unaltered. Since trie

potential

is

periodic along

trie z-direction we can

expand

it as a Fourier series.

Only

trie lowest Fourier component is relevant in trie

region

of trie

roughening transition;

trie other components vanish on

long length

scales and are irrelevant.

Therefore,

Nozières

iii

took trie

pmning potential

to be

V(z)

=

-Vcos(27rz la)

so that trie

pinning

force is

4l(z)

=

-(27rla)Vsin(27rzla)

Nozières and Gallet

iii

derived renormalisation

equations

for a flat interface

parallel

to trie

crystal planes. They

then extended trie calculation to vicinal surfaces whose direction lies close to trie

faceting

direction and to interfaces

subjected

to a constant

driving

force. Trie introduction of vicinal surfaces, or of a finite

driving

force

produces

new

length

scales whose effect is to terminate trie renormalisation group

equations

and broaden trie transition. Trie

driving

force introduces a

length

TF "

(a~i/F)~~~ (l)

This

length

is denoted as

Aj~ by

Nozières. We con also write this

length

as TF =

l'ft/J~)~/~

where t

= a~J~

IF

is trie time for trie mobile

crystal

to grow one atomic

layer,

a.

For temperatures above trie

roughening

transition trie interface is

rougir;

for temperatures

below,

it is faceted. Trie

roughemng

transition

belongs

to trie Kosterlitz-Thouless

universality

class of

phase

transitions. Nozières and Gallet obtained a set of recursion relations which

(4)

describe how trie surface

stiffness,

trie friction coefficient and trie

pinning potential

are altered under trie renormalisation scheme.

Correlations of surface

displacements

are described

by

trie correlation function

GIP)

=<

zir)zir

+

P)

>

For temperatures below trie

roughening

transition trie correlation function

decays exponentially

as

G(p)

=

Ae~P/~

where

(

is trie correlation

length.

For

length

scales shorter than

(

trie

pinning potential

is small and trie interface fluctuates

freely.

For

length

scales

larger

than

(

fluctuations

are blocked because trie cost in

pinning

energy is too

large

to allow trie interface to move from

one

crystal plane

to trie next.

Therefore, f

marks trie onset of

faceting

on

large Iength

scales.

Nozières and Gallet

iii give

trie

following picture

of trie

length

scale TF- Under an

applied

force a free surface bas a curvature R

=

ia/F.

An element of surface of

length

L bas a vertical

displacement,

a

bulge,

of order L~

/R

which is

superimposed

on trie thermal fluctuations. There

are two cases to be considered:

1) if trie

bulge

is very much less than a, it is irrelevant and trie surface element fluctuates

as if F were zero.

ii)

if trie

bulge

is mach

larger

thon a, then it averages trie

periodic potent1alto

zero, even in the absence of thermal fluctuations. Trie surface behaves as if it were free. Trie

length

scale where trie

displacement equals

trie lattice spacing is

just

TF- We are led to trie idea of

stopping

trie renormalisation program at trie

Iength

TF

beyond

which trie

pinning potential

is irrelevant.

For temperatures

just

below trie

roughening

transition trie correlation

length, (,

is much greater than TF so that it is TF which terminates the recursion relations. The interface behaves

as if it were

rougir,

even

though

it should appear to be faceted for these temperatures. Trie transition is broadened.

However,

when trie temperature decreases

further,

trie system reaches trie

region

where

(

is smaller thon TF and trie

crystal

behaves as if it were faceted. Trie interface does not move

freely.

In DC

experiments

it grows

slowly through homogenous

nucleation of

terraces on trie facets.

An

interesting question

concems trie response of trie interface to an

oscillating driving

force of trie form

(Fla)cos (wt).

Trie introduction of a finite

frequency

Ieads to a new

Iength

scale

=

ii/~~J)1/2

12)

This

length

scale

corresponds

to trie distance a localised disturbance of trie interface moves

along

trie interface in a time of order w~~

The

Iength

rd must be

compared

to both TF and trie correlation

Iength

to see which one dominates trie behaviour of trie interface. In other words there is a

competition

between these

Iength

scales. When w"~ is much

Iarger

than t

(trie

time to grow one atomic

Iayer)

then rd is much

larger

than TF

Trie situation is

simplest

in trie linear response

region

where TF is very much

larger

than trie other two

length

scales. There are two cases to be considered:

i)

If rd is less than

f

the recursion relations are

effectively stopped

at rd and trie interface

behaves as if it were

rougir.

For

example

when a

high frequency

pressure wave is incident

normally

on trie interface from trie

Iiquid

side there will be very little transnlission of pressure into trie

crystal

if trie interface bas a

large mobility;

this is a charactenstic property of a

rougir

interface.

(5)

ii)

If rd is

larger

than

(

trie recursion relations are

effectively

stopped at

f

and trie interface behaves as if it were faceted. When a pressure wave is incident

normally

on trie interface from the

liquid

side there is a

large

transmission of pressure into the

crystal

smce the solid and its melt are

acoustically

similar. The transmission of low

frequency

sound waves at the interface

depends

on the ratio

rd/(. Altematively

an

oscillating

pressure could be

exerted on the interface

by

a resonance

technique

[3]. The response at resonance

depends

on the relative sizes of the three

length scales,

TF, rd and

(.

Dur aim is to

develop

a

theory

of the AC response which can describe both the linear and the non-linear response of the interface near the

phase

transition. The non-linear case occurs when the

Iength

TF is

comparable

or smaller than other

length

scales. The

theory

of the non-linear

response is

simplest

when the correlation

length

is

longer

than other

length

scales for then the

theory just depends

on the ratio of rd to

TF In this case the interface is

effectively rough

and moves

approximately

in

phase

with the

driving

force. The response is

nearly

linear. We

develop

the renormalisation

equations

to deal

explicitly

with this case. The situation is more

complex

when

(

is

comparable

to the other

length

scales.

Effectively

the equation of motion for

long length

scales is dominated

by

the

pinning potential

which cannot be treated as if it

were small. We propose a way of

dealing

with this

complication by abruptly changing

from the renormalisation program to a method based on the

equation

of motion of the interface

neglecting

fluctuations.

The

plan

of the paper is the

following.

In Section 2 we set up the renormalisation scheme

followmg

the

techniques pioneered by

Nozières. In Section 3 we derive the recursion relations in the presence of an AC drive and introduce the dielectric function. Section 4

gives

our numencal

results for both the hnear and the non-linear

regimes.

In Section 5 we propose an

experiment

to test the

predictions. Finally,

in Section 6 we discuss the

strengths

and weaknesses of the

theory

and directions for future theoretical work.

2. Renormalisation in trie Presence of an AC Drive

The calculation m this section is an extension of that made

by

Nozières and Gallet

iii

to the

case of an AC

driving

force and uses the

technique

that

they

used. Consider a flat interface which lies

parallel

to the

crystal planes.

An AC

driving

force

(Fla)cos(wt)

causes the interface

to move. If there is no

pinmng potential

the interface moves at an average

speed

dzo

F

cos(wt)

$

aJ~

so at time t the

displacement

is

Fsin(wt)

zo =

aJ~w

The total

displacement

when there is a

pinning potential

is z

= zo + zi where zi

obeys

the

Langevin

equation

J~

~~~

=

iv~zi

~~~ sm ~~~~~ ~ ~°~ + R

(3)

dt a a

The

driving

force appears to bave vamshed but it is hidden m the term which descnbes trie

pinning

force. Because we have removed the

displacement

due to the

driving force,

all that

remains is the

displacement

due to fluctuations driven

by

the random force. Our atm is to renormalise equation

(3),

the

equation

of motion for the coordinate zi driven

by

the random fluctuations.

(6)

The random force

R(x, y)

can be Fourier transformed

R(x, y)

=

~j Rke'~.~

k

into its components Rh- The random force has a white noise spectrum <

Rk(t)Rk(t')

>=

Gkô(t t').

The spectrum must have the form Gk " 2J~kBTÙ(A

k)

in order to

satisfy

the fluctuation

dissipation theorem, something

which is

proved

in the

Appendix.

For convenience we consider a

sharp

cutoff in Rh at k

= A

although

the

precise shape

of the cutoff is irrelevant

ÎIÎ.

The

length

scale A~~ is of the order of the lattice spacing. In the renormalisation program

we

change

the cutoff from A to À and average over the Fourier components of the random force

with wavevectors in the range A to A. We do this in incremental steps

by gradually reducing

the cutoff wavevector, so that apart from the

spacing

of the

planes,

a, all the parameters in

equation (3)

are

changed

to renormalised

values;

these renormalised values are denoted

by

a

bar over the top of the

symbol.

Dur aim is to get the

equation

~Î~ ~'f~~zi

~~~ ~ 27r(21 + j

~ in

o

~

~ ~

When we have recovered the same

equation,

we can repeat the process

indefinitely

and carry out a

large

renormalisation of the

equation

of motion, at each stage

removing

the effect of

fluctuations in a narrow band of wavevectors. If we can carry out the renormalisation program for all wavevectors then we can account

completely

for the effect of fluctuations.

When

making

the renormalisation calculation we assume that the

pinmng potential

is small in comparison to the other terms so that we can use a

perturbation

expansion in the

pmning potential.

To obtain the

equations

of motion to describe the

roughening

transition it is

only

necessary to calculate terms which are second order in V. For temperatures below the transition the renormahsation

procedure

works well until we reach a

length

scale

(

=

1/A~,

where the pinning

potential

becomes

large

and the renormahsation

procedure

breaks down. In this region

we can treat the

pinning potential

as

large

and the fluctuations as

being

small m which case

we can solve the non-linear equation of motion for ai

by neglecting

fluctuations.

We separate the random force into two parts

R=É+ôR

with É the part

containing

wavevectors in the range 0 to À and ôR

contaimng

wavevectors in the range À to A. We write

é =

In(A Il)

We are

only

interested in an infinitesimal renormalisation at each stage, so the range

À

to A is small

compared

to A. Thus we consider a

tiny change

dé.

For any

history

of the random

force, R(t),

the

quantity zi(t)

is a well-defined function of

R(t')

at earlier times t'. The average

quantity

ai is defined as

ai =<

ziÎÉ+ ôR]

>ôR

with the average taken over the fluctuations m ôR. We can write zi

= ai +

ôzi

with

equations

of motion for both

ôzi

and ai

given by

(7)

and

dôzi

~ 27rV

27r(a

+ ôzi 27rV

27r(a

+

ôzi

~- =

iv ôzi

+ ôR sin + sin

dt a a a a

where a = ai + zo. V~e solve these

equations

for ai and

ôzi

as power serres in trie

pinning potential, V, by wnting

ôzi

= ôz)°1 +

ôzl~l

+

where ôzl°~ is of order

(V)°,

ôzl~~ is of order

(V)~

and so on. Trie contribution to ôzi of zeroth order is

ôzl°~

(r, t)

=

/ d~r' / dt'xc(r r',

t

t')ôR(r', t')

where trie response function

xo(p,

T) is

given by e-~P~/4~r

~°~~'~~

47riT

~~~~

The random force ôR is distributed as a Gaussian variable centred on zero. It follows that ôzl°~ is also distributed as a Gaussian with an average value of zero.

The correction to

ôzi

which is first order in the

pinning potential

is obtained

by

iteration of the equation of motion. Thus

ôz)~~ =

~~~

/ d~r' / dt'Ko(r r',

t

t') (sin

~~') sin

(~~~')

a a a

Here we have used the dash to indicate that we are concemed with the space time

point (r', t').

Thus

z'

=

2(r', t')

+ ôz)°~

(r', t')

Corrections of

higher

order in the

pinning potential

can be obtained

by iteration; fortunately

we

only

need to determine contributions to ôzi to first order in the

pinmng potential.

In what

follows,

we put ôzi " ôzl°~ + ôzl~~ and

ignore higher-order

ternis.

The

equation

of motion of ai to second order in the

pinning potential

is

~

ÎÎ '~~~~~

~ ~

~~

~~~

Î~Î~ ~°~ ~Î~~~

~ ~~~~

~~~~~

~~~

27rV

27r(a

+

ôzl°~)

2~rôZÎ~~

à = cos

a a a

the average

being

taken over the random force ôR.

Since the

quantity

ôzl°~ is distributed as a

Gaussian,

we have the exact relation

~

~~

~~~~Î~~

~-2w~<(ôz)°~)~>la~

~-nôg(0,0)

~

where the function

ôg(p,

T) is defined

by (see

Nozières's article,

Eq. (4.89)) ôg(p, r)

=

j jôzl°tir, t)ôzl°~ir

p, t

r)j

ÎÎÎ~~~~~~

~~~~~~

~~~~~~~~

~~~~~~

(8)

and where the

quantity

n is

7rkBT

n =

la~

The term in the

equation

of motion for ii which is first order in V becomes

~

(ai

+ zo

=

~~~

in

~~~~~~

27r -nôg(o,o) ~.

~~

~e

in ~

a ~

with V

=

Ve~"~~(°?°l.

The effect of the renormalisation at this stage is to leave the pin-

ning potential

term with the same form as the

corresponding

term in

equation (3)

but with

renormalised values of zi and V.

The term which is second order m the

pinmng potential, à(~),

can be found

by substituting

for ôz)~~ which

gives

à(2)

=

~~~)~ / d~r' / dt'xc(r r',

t

t')

a

27r(z

+ ôz)°~

7rz'j 27r(z

+ ôz)°~

7rz'j

cos sm cos sm

~ ~ ~ ~

Remember that the

primed

term z' refers to the space time

point (r', t')

whereas the

unprimed

terms describe the space time point

(r, t).

First consider the last term in braces on the

right

hand side of this equation.

By taking

Gaussian averages of terms

involving

ôz)°~ we get

l127r(z

cos +

ôz)°1)

sm

7rz'j

=cos

7rzj

sm

7rz'j

e_~~~~~~ ~j?

a a a a

Now consider the other term m braces on the

right

hand side. Before

taking

averages we rewrite this as

27r(z

+

ôz)°1) 7rz'j

1

27r(z'

+ z + ôz)°~ i

27r(z'

z

ôz)°1)

cos sm = sm + sm

~ ~ ~ ~ ~ ~

When we take averages over ôR this becomes

C°S ~~

~ ~

~

Zi°~ ~~~,

~~~

~

= ~i~

27r(2'

+

a)

2 ~

~_~~~~~~~~~ ~

+ ~~

27r(2' a)

~ ~~~~~~~~

~ ~ ~

e2nôg(p,~)~_~~~~ ~~~

The sum of the two terms m braces is therefore

given by

~~"Î~°'°~ SIR ~~~~l+

~~

ie~"~~~~~~ i)

+ SIR

~~~~l

~~

ie~"~~~~~~

i)1

But the

quantity ôg(p,

T) is small since it is

proportional

to

dé;

it follows that we con ignore terms in

ôg(p, T)2.

The final

expression

for

à(~)

is

~~~

47r3V~

/

~

/ 27r(z' z) 27r(z'

+

z)

à

= ~

d p

dTxo(p, T)2nôg(p, T)

sin sin

a a a

(9)

where we have written p

= r r' and T

= t t'. At this stage no approximation has been made.

The term which

depends

on

sin(27r(z'+ z) la) corresponds

to a

higher spatial

harmonic of the sort that has been

ignored

when

describing

the

pinmng potential.

It is irrelevant. It is the term in

sin(2~(z'- z) la)

which we want, for this is relevant close to the transition. We write z'- z as

ai

(r', t')

ai

(r, t)

~

(sin(~at) sin(~a(t r))

= ai

(r', t')

ai

(r, t)

a

a~~J with a

=

(2F/~~Ja) sin(~JT/2) cos(~J(t T/2)).

Nozières and Gallet

(ij

considered

only

the

w = 0 limit of this

equation

in which case a

= Fr

la~.

The new element of our

approach

is to extend their

theory

to finite

frequencies.

We

pick

ont of

sin(27r(z' z) la)

the

slowly varying

mode which we want and

thermally

average over the rest of the modes which are assumed to be m

equilibrium.

In this way we get

27r(z' z) 127r (ii (r', t')

ii

(r, t)

27ra 27ra _~~j~~ ~~

sm = cos -sm e ~

a a a a

where we have written

The

slowly varying

mode which we want can be

expanded iii

as a power series in T and p.

After

integration

over

d2p,

the lowest-order terms are

ai

(r', t')

ii

(r, t)

=

-T~~~

+ ~

V~zi

dt

so that we get in lowest order

Sm

~~~~l ~l

SQS

ll~ 1-T Î~

+

1v~zil

COS

l~i~l

SIR

l~i~l1e-~"~~~~~

Of course there are

higher-order

terms such as

)T2d2zi /dt2.

These terms are all

neglected

for

they

are irrelevant in the

long wavelength

limit. For

example,

the term

)T2d2zi/dt2

alters the effective mass of the

interface, something

which bas

already

been

neglected

since trie interface is

overdamped

in the

long wavelength

limit.

Dur aim is to recover the

original equation

of motion for zi but with renormalised parameters.

By collecting

terms we get the

equation

~~~~

=

'fv~zi

~~~ sm

~~~~~

~

°~)

dt a a

+

~~~~~~~ /d~p /dTxo(p, T)Jo(Ap)e~~~~~&e~~"~(P'~)

a

x

ll~ 1-T Î~ +1/4P~v~zil

COS

l~i~l

SIR

l~i~ Il

The first three terms on the

right

hand side of this

equation

have the correct form. In the last term on the

right

hand side there is an expression

proportional

to dal

/dt

which renormalises

(10)

the friction coefficient there is also a terni

proportional

to

V~zi

which renormalises trie surface stiffness coefficient. Both of these terms are affected

by

trie

quantity cos(27ra la);

the presence of this term

changes

trie recursion relation

for'f

and ~ from those obtained

by

Nozières.

Moreover,

we get an extra term on trie

right

hand side which

corresponds,

as we will see, to a renormalisation of trie

driving force,

or

altematively

to a renormalised value of

dzo /dt.

We use mathematical identities for

cos(fl cos(Ù))

and

sin(flcos(Ù)) given by

Abramowitz and

Stegun (Sections

9.1.44 and 9.1.45 of Ref.

(4]).

It is convenient to introduce trie

quantity

y defined as

47rF sin

(wT /2 47rr( sin(wT /2)

~ ~J~a~

r(

It follows that

ce

sin(27ra la)

= 2

~j(-1)~J2k+1IV)

cos

((2k

+

1)~J(t T/2))

k=o

and

ce

cos(27ro la)

= JoIV + 2

~j(-1)~

J2k

Iv)

cos

(2kw(t

T

/2)

k=i

We

only

need trie first term in each of these

expansions.

Trie next term

produces

a correction to the

equation

of motion which are second order in trie

pinning potential,

a correction that

oscillates at a

higher frequency.

This correction can feed back and

give

rise to a contribution at the fundamental

frequency,

but the

leading

contribution is fourth order in the

pinning potential.

We are

only working

to second order so

only

the first term in each of these

expansions

is needed.

Thus we take

sm(27ra la)StS2Ji

IV cos

(w(t

T

/2)

and

cos(27rala)StsJo(V)

The

neglected

terms oscillate at

multiples

of w. Remember we are interested in the response of the system at

frequency

w and not at

higher frequencies.

Dur immediate

objective

is to recover

the

original equation

of niotion for zi

By collecting

terms we get the renormalised

equation

'~~~~~ ~~

~~~

~~~~~/

~~~ ~ ~

~~~~~~~ /d~p /dTxo(p>

T)

Jo(Ap)e~~~~~/~e~~"~(P~~)

~~~

a

x2JiiY) cosjcojt r/2))

Here we have renormalised i

to1='f+di

where

dl

is given

by

dl

"

~~~Î~

~~~

/ d~pp~ / dTxo(p> T)e~~"~(P'~) Jo(Ap)e~~~~~/~ Jo Iv)

a

~

and have renormalised ~ to q =

~+d~

where

d~

is given

by 16~4p~~dé

d~

~

= ~

d~p dTTxo(p> T)e~~"~(P'~)Jo(Ap)e~~~ ~/~Jo(1l)

a

Î Î

(11)

Both

expressions

contain the terni

Jo(Y)

which

elfectively

cuts off trie renormalisation when y becomes

larger

than umty.

Equation (5)

still is not of trie form we

require,

for there is an extra terni which oscillates at

frequency

w. This extra term is

effectively

an additiona1force which renornialises zo to do.

To realise

this,

it is

simplest

to transform

equation (5) by writing

it in ternis of trie variable

z = ai + zo. Thus

dz

~~

2~V

j27rzj

flfcos(wt)

q- = y z sm + R +

dt a a J~a

~~~~~~~

/d~p/dTxo(p>T)Jo(Ap)e~~~~~/~e~~"~(P~~)

a

x

2Ji(Y) cos(w(t T/2))

The last terni can be

thought

of as a contribution to trie renormalised force; but it is

simpler

to define a renormalised

velocity

of trie interface as

dz/dt

= io+dào where trie incremental

change

in

velocity

is

dào "

~~~Î~~~~~ /

d~

p

/ dTXo(p, T)Jo(Ap)e~~~~~/~e~~"~(P~~)

JiIV)

cos(w(t T/2)

a il

When we rewrite trie

equation

of motion in terms of ii we recover equation

(4)

with do trie renormalised

displacement

of trie interface due to trie externat drive.

Trie renormalised

velocity

of trie interface now contains a part which oscillates as

sin(wt),

that is there is a part which is out of

phase

with trie extemal drive. It is convenient to express trie force as

Fcos(wt

+ ç§). Trie part of trie interface

velocity

which oscillates as e~~~~ is

~°~"~ ~°~"~ ~aÎ~

We con then write

d

Jnluolta))1

=

~~~(ÎÎ~~~ / d~p / dTxolP,

T)JOIÀP)e~~~~~/~e~~"~~P~~~Jilv)e~~~/~

3. Recursion Relations and trie Dielectric Function

3.1. REcuRsioN RELATIONS. Ii is convenient to

change

from variables p and r to dimen-

sionless variables

,~

~ 'lP~ ~

~~

We also define trie dimensionless

quantities

r = rd

/rF

fd

=

Àrd

The

quantity

r is

just

trie ratio of trie

length

scale rd to TF In terms of these new

quantities

trie parameter y becomes

y = 47rr~

sin(xp~ /2fd~) (6)

The

pinning

energy per unit area is converted into an energy

given by

U

=

V/À~.

Trie recursion relations express incremental

changes

in

U,

'f>

~1 and

uo(w)

as differential equations. In ternis of trie new variables these are

$

= 12

n)U Ii)

(12)

dl

~

2~~

~~ A(n,

td

r)

~~~

(

=

~~[~

B(n,

td

r)

~~~

d i a

and

d

lin juo1°'))1

=

~~~~~

Gin,

fd

r)

~~°~

72a4

~~~ ~~~~~~~~~ ~~ ~y~ ~~ c are

given by A(n,

td>

r)

= n

/ ~ dpp~ / ~

dz

e~~R~e~~"~(P'~) Jo(fl)e~P~~Jo(Y) Ill)

X

B(n,

fd>

r)

= n

/~ dpp~ /~ dxe~~R~e~~"~(P~~)

Jo

(P)e~P~~

Jo

Iv) (12)

o o

C(n,fd> r)

=

"~) /~ dpp /~ ~~e~~R~e~~"~(P?~) Jo (fl)e~P~~Ji(1l)e"P~/~~d~ (13)

7rr ~ ~ x

This is a

complete

set of recursion relations which is valid for trie non-linear

regime provided

the

pinmng potential

is small so that the second order calculation is sensible. For teniperatures

above the

roughening

transition the

quantity

n is

always larger

than 2 for all é and the second- order calculation is valid for ail

wavelengths.

The pinmng

potential

becomes

progressively

smaller as é increases.

By making

a succession of

tiny changes

of scale we renornialise over ail

length scales,

a

procedure

we can carry out without any

problems

for temperatures above the

roughening

transition. The result is a renormalisation of

U,

i and J~. The

siniplest

treatment is to

keep

the unrenormalised values of these pararneters on trie

right

hand side of

equations (7), (8)

and

(9).

In a more

sophisticated

calculation one uses trie renormalised values on trie

right

hard side of these

equations,

a

procedure

which goes

beyond

trie second-order calculation. When this

approach

is

adopted,

trie noise spectrum must be renormalised at each stage in order to

ensure that trie fluctuation

dissipation

theorem is

obeyed.

This point is discussed more

fully

in trie

Appendix.

Trie

velocity

of trie interface

given by

~~

uo(~J,é

=

0)

~~

i~a4 ~~"'~~'~~

where

uo(w,

é

=

co)

is trie

velocity

after renornialisation to

infinitely long wavelengths.

We define a "dielectric

constant",

e,

by

trie ratio

~j~~

~ ~01~°>é "

o) uo(w,é

=

co)

This constant tells us how the unrenornialised

velocity

is altered

by

fluctuations in trie random force on all

length

scales. Thus

i~

l~l~°))

"

/~

dé~)))ÎÎÎ~ Cl", fd>T) 1")

All we bave to do is to solve trie recursion relations so as to get trie renornialised

pinning potent1al, friction,

surface stiffness and

driving

force as a function of é and

integrate

trie

expression above to get trie dielectric constant.

There are

only

two

Iength

scales for temperatures above trie

roughening

transition: rd and TF If trie ratio r =

rd/rF

is very small then trie response is

Iinear;

as trie ratio increases and

becomes

comparable

with trie response becomes non-linear.

(13)

3.2. RESPONSE FOR TEMPERATURES BELOW THE ROUGHENING TRANSITION. For tem- peratures below trie

roughening

transition trie recursion relations break down for a value of é = Ç

(a precise

definition of Ç is

given

later

on),

where trie

pinning potential

becomes com-

parable

with

kBT.

As é increases one moves from a

regime

where trie

pinning potential

is small and trie fluctuations are dominant to a

reginie

where trie

pinning potential

is

large

and

fluctuations are small. This crossover occurs at trie correlation

length, (

=

Ap~exp(Ç).

There

are three

length

scales m this

regime:

rd> TF and

(.

We can use trie renormalisation niethod to

analyse

the response for

wavelengths

shorter than the correlation

Iength;

for

longer wavelengths

it is necessary to use a dilferent

technique.

We define trie dielectric constant for renormalisation up to Ç

by

trie ratio

éi~lta)

=

))))')f ()

,

where we renormalise trie

velocity

over trie range 0 to Ç. Thus Ç

~~4uj~)2 inléi~lta))

=

/

,y~~~4 CIn,fd,r)

o

At Ç trie equation of motion bas the

following

form

fl(Ç) ~~j)~~

=

Y(Ç)V~zi IL) ÎV(Ç)

sin

~~~~~~~~~

~ ~°~~~~~

+

É(Ç)

where ail trie terms are taken at the

point

é =

Ç.

This

equation

can

easily

be rewritten m terms of trie

quantity

z

= ii + do which

gives dé(Ç

~

27r

27rz(Ç) É cos(wt

+ ç§)

q(Ç)

=

1(Ç)V 2(Ç) VIL

sin +

R(Ç)

+

dt a a a

~~~~~

É cosiLdt

+

')

=

iié~) ~j(C~

a

Here

Éla

is the renormalised

driving

force and

ç§ is a

phase angle.

If we

neglect

the randoni force we can solve the equation of motion

analytically provided

the

pinning potential

is strong

enough

to allow

only

small

displacements

of the interface around the

equilibrium position.

Such an

approach

is not correct when the renormalised

driving

force is too

strong, causing large displacements;

when that is trie case we bave to solve trie

equation

of motion

nuniencally.

3.3. SMALL DISPLACEMENTS. Trie equation of motion for small

displacements

can be

written as

~ÎÎ '~~~~ Î

~

~~ ~ ~ ~

~

~~~ÎÎ~

~ ~~

which is a

simple

hnear

equation

which can

easily

be solved [7] The Fourier coniponent of trie

velocity

due to trie renormalised

driving

force which oscillates as e~~~~ is

~oj~a,

t

=

«)

= ~-

~~ '~

=

~

~-

uol~a,

t =

t~)

(i

+

1~~ ( ~~~ (i +1~~ (

taira taira

(14)

The dielectric constant is

In(e(w))

=

~~ dt~~(~~C(n,fd,r)

+ In

(1+

~_

~~~)

o 'f a wJ~ a

The

procedure

now is to evaluate trie recursion relations up to Ç and then to use trie renor- nialised values of

V(Ç)

and

fl(Ç)

to calculate the correction to trie dielectric constant for

large

é. This is a rather crude

approxiniation,

for we are

ignoring

the thermal fluctuations of

long wavelength generated by

the renormalised random noise which oct to reduce

slightly

the

pin-

ning

potential.

Such an elfect is

certainly

present, but it is

ignored

in our

approach

or rather it is subsumed in the choice of how to define

Ç.

If the choice of

Ç

is sensible we expect trie

predictions

of trie

theory

to be

relatively

insensitive to small

changes

in

Ç.

The calculation

only

makes sense when trie

displacement

of trie interface is small in compar- ison with

a/27r.

If this condition is not satisfied we need to use a numericalmethod to solve the

equation

of motion. The condition that the

displacement

is

sulliciently

small

depends

on

trie three dilferent

length

scales. There are two separate cases. When

(

is

bigger

than rd, trie condition for small

displacements

is that rd is less than TF- When

(

is smaller thon rd trie condition for small

displacements

is that

(

is less than TF More

succinctly,

the

length

TF must be

larger

than the smaller of the other two

lengths.

When this condition is not satisfied the

displacement

is

large

and we need to use a numerical method to obtain the dielectric constant.

3.4. LARGE DISPLACEMENTS. In our

simplified

treatment we

neglect

trie random force

because its elfect is small in comparison to trie

pinning potentiaJ.

Therefore we propose to solve trie equation of motion

~

ÎÎ '~~~~ Î

~ ~~~

~Î~

~

~

~~~ÎÎ~

~ ~~

numerically

when trie

displacements

are

large.

Trie surface stilfness term can be

neglected

because we are

considering only

a

spatially

uniform

driving

force.

In order to

perform

trie numencal work it is convenient to recast this

equation

into dimen- sionless form

by using

the

length

scales rd, TF and

(.

Let us call

w =

27rzla

x = uJt

and use the reduced variables [1,2]

21a~

2

~ 7rkBT n

47rU

~ kBT

Then the

equation

of motion becomes

Î ÎÎÎÎ

~~

~~

~~~~~~°~ ~ ~~~~~~~~

~°~~~~ ÎÎÎÎ~ ÎÎÎÎ ~~~~~~

where the realand

imaginary

part of the inverse dielectric function bave been introduced. In trie above

equation

ail trie parameters are evaluated at

Ç.

Références

Documents relatifs

instead of using a classic LIFO memory, proved to be sufficient to capture locality effects (Fried- mann et al. 2009) when “similar” NPs are pro- cessed (Warren &amp;

00 PRIX DU SYNDICAT DES MARCHANDS DE (Groupe A) Paris Simple Gagnant, Simple Placé, Couplé Gagnant, Couplé Placé, TrioD.

Placer ces points sur la figure

Our final implementation uses a red-black tree to store fil- tered transactions, item order is ascending according to their support, simultaneous traversal is used as a routing

Largement répandu dans les établissements d’enseignement, le photocopillage menace l’avenir du livre, car il met en danger son équilibre économique et prive les auteurs

• Pour que la visibilité du bouton dépende de l'état de la case pointM, ouvrir le panneau des propriétés du bouton, et, dans l'onglet Avancé , inscrire dans le champ Condition

• Pour que la visibilité du bouton dépende de l'état de la case pointM, ouvrir le panneau des propriétés du bouton, et, dans l'onglet Avancé , inscrire dans le champ Condition

➋ L'image initiale pour insérer une image nommée image1 dans la zone de travail : deux points A et B.. sont automatiquement créés aux coins inférieurs