HAL Id: jpa-00208528
https://hal.archives-ouvertes.fr/jpa-00208528
Submitted on 1 Jan 1976
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, estdestinée 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.
A simple stochastic description of desorption rates
G. Iche, Ph. Nozières
To cite this version:
G. Iche, Ph. Nozières. A simple stochastic description of desorption rates. Journal de Physique, 1976,
37 (11), pp.1313-1323. �10.1051/jphys:0197600370110131300�. �jpa-00208528�
A SIMPLE STOCHASTIC DESCRIPTION OF DESORPTION RATES
G.
ICHE,
Ph.NOZIÈRES
Institut
Laue-Langevin,156 X, Centre
de Tri38042 Grenoble
Cedex,
France(Reçu
le 11 mars1976, accepti
le3 juin 1976)
Résumé. 2014 La théorie
cinétique
des vitesses dedésorption
d’un adatome en surface est reprisesous un angle stochastique, permettant de
généraliser
les résultats de Kramers (fondés sur uneéquation
de Fokker Planck). Pour un puits depotentiel profond,
1a vitesse dedésorption
est reliéeà la
probability
de piégeage moyenne d’un atome thermique incident. Celle-cidépend
à son tour de l’énergie moyenne cédée au substrat par la particulependant
une oscillation. En général, le transfertd’énergie
aux phonons est si efficacequ’il
justifie a posteriori le modèle absolu de Eyring. On peutgénéraliser
le modèle à des situations pluscompliquées (adparticule
avec degré de liberté interne,barrières activées, etc...). Enfin, un formalisme
dépendant
du temps, valable pour despuits
peuprofonds,
estesquissé
2014 pour despuits profonds,
ce formalismegénéral
se réduit au modèlesimple
du
depart.
Abstract. 2014 The kinetic
theory of desorption
rates for an adatom on a solid surface is reconsidered from a stochastic point of view. The results of Kramers, obtained within the framework of a Fokker- Planck equation, are therebygeneralized.
For a deep potential well, thedesorption
rate is relatedto the average
sticking
probability of an incoming thermalparticle.
The latter in turndepends
onlyon the average energy loss 03B4 of the
particle
to the substrateduring
asingle
oscillation.It is shown that in most cases the loss to lattice vibrations is so strong as to
justify
a posteriorithe absolute rate theory of
Eyring.
The theory is extended to morecomplicated
situations(adparticle
with internal
degrees
of freedom, activated barrier, ...). Finally, afully time-dependent
formulation,valid for shallow wells, is sketched; it is shown how, for
deep
wells, it reduces to the abovesimple
formulation, based on detailed balanceonly.
Classification
Physics Abstracts
7.854
1. Introduction. - The
theory
of chemical reaction rates is a well-established field - and as such it has reached a level of considerablesophistication
overthe past
forty
years. As aresult, simple physical
ideasoften tend to be hidden.
In the present paper, we consider the
problem
froma somewhat unusual
point
ofview,
stochastic in nature, whichclearly points
to some of the essentialphysical
facts. Our purpose is tolay
down a formula-tion,
and to stressqualitative
results : we thus deli-berately
chooseoversimplified
models.We consider
only
situations in which thereacting species
are in close contact with a heatbath,
whosemany
degrees
of freedom caneasily
absorb orprovide
energy. The
simplest example
of such a reaction is thedesorption
of asingle
atom Atrapped
on a solidsurface :
there,
theunderlying
substrateprovides
thethermal
agitation
thatultimately
will lead the atom A into free space. The chemical reaction may be viewedas a brownian motion in a
potential
well.(In
section4,
we consider
briefly
thegeneralization
to morecompli-
cated cases,
using
the standard concept of a reactionpath.)
The
application
of stochastic theories to chemical kinetics is not new. In1958,
forinstance,
Montrolland Shuler
[1]
treated in that way thedesorption of a particle trapped
inside a truncated harmonic well.They
relied on a masterequation
in order to describetransitions between the successive
quantized
levels.The present
approach
is similar inspirit
to thispioneering work,
but rather different in its formula- tion. We consideronly
apurely
classicalregime.
Also,
weemphasize
the role of inelastic processesnear the
top
of thepotential
barrier - a featurewhich was
apparently
notincluded
in the work ofMontroll and
Shuler.
As usual in stochastic
theories,
all that we say is standardpractice
forprofessional
statisticians : ourproblem
ofdesorption
isisomorphous
to thegeneral
Markov chain with
absorbing
barriers(as treated,
for
instance,
in the standard text of Feller[2]).
Grantedthat the stochastic method is in no way
original,
we found it easier to construct from scratch our own
probabilistic formulation,
rather thantranslating
the usual mathematical
language
into ourproblem.
In section
2,
we set up thegeneral problem
ofArticle published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:0197600370110131300
desorption,
and we discuss thecorresponding
absoluterate uo as obtained from the well-known
Eyring theory. Using
a detailed balance argument, we show that the actualdesorption
rate u differs from uo, the ratiouluo being
related to the averagesticking probability
that an incident thermal atom remainstrapped
whenhitting
the surface. That u shoulddepart
from uo was noted back in 1940
by
Kramers[3]
whotreated brownian motion in the framework of a
Fokker-Planck
equation.
Theproblem
was revivedrecently by
Suhl and coworkers[4].
The variousregimes
of Kramers appearnaturally
in ourlanguage.
Such a stochastic
description
is based on theassumption
that the process is markovian on anappropriate
time scale t(here
theperiod
of oscillation of the adatom in the surfacepotential well).
One may take theopposite view,
and consider a non-stochasticsubstrate,
whichresponds
to the adatomaccording
to the usual deterministic laws of mechanics. The calculation of the
sticking probability
is then apurely
classicalproblem
- as described for instance in recent articlesof Beeby et
al.[5].
Several
possible
extensions of oursimple
model aresketched in section 3 : influence of the three-dimen- sional structure
(motion
of the adatomparallel
to thesurface),
of the internaldegrees
of freedom of the adsorbedparticle.
We also consider reactions between twopotential wells,
instead of one well vs outer space.In section
4,
we return to the case ofdesorption,
and we show that our
simple
detailed balance argu- ment isonly approximate : actually,
it assumes thatone has two very different time scales : a fast scale
(a
few oscillationperiods)
for theadparticle
to fallin the
well,
and a slow scale(the desorption time)
forthe
trapped particle
tofinally
escape.Trapping
anddesorption
can be identifiedonly
if their two scalesare well
separated,
a condition which is met if thepotential
welldepth
V is muchlarger
than the tempe-rature T.
In the most
general
case, we show how the calcula- tion of the reaction rate may be reduced to asimple
mathematical
form, namely
the search foreigen-
values of the
appropriate
stochastic operator. The results of section 2 are reached in thelimit V >
T.2.
Desorption
rate vssticking probability.
- Weconsider
point particles,
with mass m,trapped
in apotential
wellV(x)
near a solid surface. In thissection,
we consider
only
motionalong
thex-direction
normalto the surface : the
problem
is one-dimensional. Thepotential profile
is sketched infigure
1. We assumethat there is no activation barrier to be overcome
before
entering
the well(the
influence of such a barrier is discussed in section3).
Theorigin
of energy is chosen at the vacuumlevel;
thedepth
of the well is V.Were it not for the friction on the
substrate,
aparticle
with energy 8 in the range
(- V, 0)
would oscillate with afrequency w(E), easily
found from m andV(x).
Implicit
in thisdescription
is theassumption
that weFIG. 1. - The profile of a one-dimensional, unactivated surface
potential barrier.
can separate a static interaction with the substrate
(yielding
thepotential
wellV(x))
from adissipative
friction mechanism. To the extent that it is the same
interaction that
provides
thebinding
and thefriction,
such a
separation
is far from obvious. It can be achie- ved forheavy adparticles
that moveslowly
as com-pared
totypical
substratefrequencies (R6sibois
andLebowitz
[6]).
We limit ourselves to a
purely
classicalregime,
validif hw « T. We can then build a wave
packet
the size of which is smallcompared
to the width of the well.The temperature is
imposed by
the substrate. In thermalequilibrium,
each state has anoccupation.
The net number of
trapped particles
iswhere p
=1/hw(e)
is thedensity
of states.The
integral
in(1)
is controlledby
aregion
ofwidth - T near the bottom of the well. For
deep
wells( V > T),
we canreplace W(B) by
its value Wm near thebottom : we thus find
In
principle, (2)
holdsonly
if the adsorbed atoms arein thermal
equilibrium
with a gasphase having
thesame
(u, T).
Assume now that there is no gas
phase.
The adsorbed atoms are in thermalequilibrium
with the substrate except for a thinlayer
near the top of the well. Fora
deep well, (2)
remains valid. Let J be the net flow ofparticles
thatquit
the well per unit time. Thedesorption
rate u is defined asIn the standard
approach
ofEyring [7],
one assumesthat the thermal
equilibrium
distributionsfo(e)
extendsto all
energies, including
those atoms above vacuumlevel
(E
>0)
which move away from the surface. The net flow is found fromelementary
kinetictheory :
where dx = v dt is the distance travelled
during
time dt. On
comparing (2)
and(3),
we find the so-called absolute rate
of desorption
(4)
has the usual Arrhenius form. Note that(4) only
involves
equilibrium quantities (1).
As shown
by
Kramers[3]
the result(4)
cannot holdin all cases. Let us describe the
dissipation
on thesurface as a friction force
proportional
to thevelocity
vIf n >>
m, the mean freepath
of theadparticle
is shortas
compared
to the wellsize,
and thedesorption
may be viewed as aplain
diffusion process in the presence of anapplied
force(- grad V).
The net currentdensity J(x)
is a result of conduction and diffusionFor a
deep
well understationary conditions,
Kramersshows that
(5)
can be solvedexactly, yielding
a reactionrate
(xm
is the wellminimum).
In the
opposite limit I
ill, the motion of theparticle
is adamped
oscillation. In thelimit q - 0,
Kramers shows that thedesorption
may be viewed asdiffusion in energy space; the
corresponding
rate u isproportional to q :
it vanisheswhen q - 0,
for lack ofa mechanism which could
replenish
the upper states(c
>0)
once the firstparticles
are gone.More
generally,
let us consider an adsorbedlayer
inthermal
equilibriwn
with a vaporphase.
Consider a gasparticle
with energy s that hits the surface : letP(s)
be the
probability
that it remains stuck.By
stuck wemean that the
particle
fallsdeep enough
into the well(losing
an energybigger
thanT),
in such a way as to remaintrapped
near the bottom for along
time(see
section 5 for moredetails).
The net flow ofparticles
that hit the surface is
given by (3).
Ofthese,
a fraction P(1) In (4), the factor wm comes from no, and it has nothing to do
with kinetics. The argument according to which it is the frequency
at which the particle tries to escape is meaningless.
is
trapped :
hence in thermalequilibrium
a netingoing flux
Jin
must be balancedby
thedesorption flow, Jout
= no u.We thus obtain an exact
expression
for thedesorption
rate u
Eyring’s
rate uo is correctedby
the averagesticking probability
P(2).
Note that the result(8) depends only
on detailed balancearguments.
It iscompletely general, equally
valid in theoverdamped
case(where,
combined with
(6),
ityields
easy access toP ),
and inthe
underdamped limit q
m.We now limit ourselves to the low friction case,
" n to. The
trajectory
is then anoscillatory
roundtrip
with random energy. We focus our attention on the
particle
when it is reflected on the external side. Let si 0 be the energy at such aturning point :
thebrownian motion may be viewed as
hopping
fromsi to ei+1
in
the course of one roundtrip. Eventually,
the
particles
will reach apositive
energy en at theedge
of the barrier
(beyond
which friction isnegligible) :
the
particle
is then desorbed.Conversely,
anincoming particle
with energy e > 0 willhop
intoturning energies
81, E2, ...,falling
down into the bottom unless isagain
hits apositive
En.Let
W(s, 8’)
de’ be theprobability
that theparticle hops
from energy e to the range(e’,
s’ +de)
in thecourse of one round
trip
between twoturning points.
W(s, s) provides
theonly physical ingredient
of ourformalism. We assume that the process is markovian from one round
trip
to the next : there are no statistical correlations between two successive oscillations.(Such
a markovian
approximation
is less restrictive than the Fokker-Planckdescription
usedby Kramers,
as itapplies only
on the finite scale of one roundtrip).
We do not make any
assumption
on theshape
ofw(Bg g’).
Consider the
sticking probability P(8)
of an incidentparticle
with energy e > 0. After one roundtrip,
ithas energy
8’,
with aprobability
distributionW(e, c’).
If s’ >
0,
theparticle
escapes and it is nottrapped ;
if e’
0,
it starts overagain
with asticking probability
now
equal
toP(e’).
The functionP(s)
thusobeys
thesimple integral equation
(9)
must be solved with theboundary
condition P - 1when s -+ - oo. The reaction rate u is thus
entirely
(1) The relation between desorption rate and sticking probability
is well known-for instance, in the case of solid evaporation.
determined
by
the energy lossprobability
for asingle
round
trip, W(s, s’).
Such apoint
of view has been used in the past in otherproblems,
forinstance,
the capture ofelectrons by
traps in semi-conductors - the so-called Giant traptheory of
cascade processes.The
pioneering
work of M. Lax[8]
in this respect is based on anequation
similar to(9).
In order to
proceed further,
we examine in somedetail the behaviour of W.
Obviously,
it is normalised to 1 :For
simplicity,
we assume that Wdepends only
onthe difference
(e’
-E)
in theimportant
range(of
width
T)
near thetop
of the barrier(3).
It thendisplays
the behaviour sketched in
figure
2 : an average loss per oscillation6,
and fluctuations around this value with anamplitude
4. W issubject
to a further cons-traint : it must
respect
detailed balance in thermalequilibrium.
The condition ofequal
flow betweenthe energy ranges
(s, e
+de)
and(c’,
e’ +de) implies
FIG. 2. - A sketch of the energy transfer probability W(8,8) for
a single oscillation in the potential well.
We may
distinguish
tworegimes : (i) 6 «
T : we canexpand
the factorand
consequently (11) implies
(12)
isnothing
but Einstein’s relation in energy space,relating
energy diffusion( N A 2’)
to average energy loss 6 .(ii) 6 »
T : then(11) implies
W = 0 if e’ - e > 0 :we can
only
lose energy(except
for a small tail ofwidth
T).
(3) Such a simplification may be questioned in the real case of a
barrier which is rounded off near the top : then the oscillation period becomes longer and longer when 8 approaches 0. If however we assume that dissipation occurs only in a region of finite width a near
the surface, the energy transfers to the heat bath do not vary appre-
ciably when 8 goes through zero : in that case, W depends only
on (E’ - e).
We now return to the
integral equation (9) governing
the
sticking probability P(e),
and we consider firstcase
(ii), i.e., 6 >>
T. If W werestrictly
zero whene’ > s, the
question
oftrapping
would be answered after thefirst
roundtrip (since
theparticle
cannotregain energy).
The solution of
(9)
is thenIn
practice,
the smallpositive
tail inW(8’ - s)
willround off P near
-s = 0,
over a range - T : this is a small effect which leavesP(s)
very close to 1 in thatregion.
HenceP ~
1. In the limit6 » T,
the absoluterate
uo is therefore
correct,irrespective
of all the details of individual energy losses.We consider next the
opposite
limit6 «
T. Thebehaviour of
P(s)
is then sketched infigure
3.P(8)
vanishes
if 8 »
L1(in
which case theparticle
leavesafter the first
oscillation).
Fornegative
s, it goes to 1 when s - - T : the maximum energy theparticle
can
regain
from the thermostat is T(after n
oscilla-tions the
particle
loses an energynb,
while it can atbest
regain In
L1 fromdiffusion ;
the maximum energy it can accumulate isd 2/b
=T).
What we wantis the average
P
defined in(7) (shaded
area ofFig. 3a).
FIG. 3. - The behaviour of the sticking coefficient P(c) as a func-
tion of E, in the limiting cases 3 « T.
Actually, (9)
can be solvedexactly
in thatlimit,
asshown in
appendix
A. One finds thatwhere 6 is the average energy
loss,
defined as the firstmoment of W
The actual
desorption
rate u is thus reduced ascompared
to uo.As an
example,
let us consider Kramers’ model wheredissipation
occursthrough
a friction force- mnv. Then
(the integral
runs over oneperiod
ofoscillation).
Thevalue of 6
depends
on the exactpotential profile ;
it is of order
where V is the well
depth
andWõ 1
the time it takes to cross adissipation layer.
We recover the low friction result ofKramers,
whichapplies
ifAs a
conclusion,
let us stress a fewimportant points.
First of
all,
any informationregarding dissipative
mechanisms appears
only
inP.
The other factors in u areequilibriwn quantities,
that havenothing
to dowith friction.
Second,
and moreimportant, P
isentirely
controlledby
whathappens
in a narrowlayer
of width T near
the top of
the barrier : whentrying
to escape from the
potential well,
it is the last T that counts, thehighest
hurdle to overcome. Below thatlayer,
the distribution is in thermalequilibrium, independent
of friction.In
principle,
these resultsdepend
on our markovianapproximation,
which as we shall see in section 3 is often doubtful.Actually,
such a restriction does not hold if 6 »T,
in which case after asingle
oscillation anincoming
thermalparticle
isalready
well below thetop layer
of width T : the chance thatsubsequent
oscilla-tions will make for this lost energy are
negligible (the
heat bath can at best
provide
an energyT).
Thus wemay
safely
concludethat, irrespective
of statistical correlations between successiveoscillations,
the trap-ping probability P
is 1 when d > T : the absolute ratetheory
holds.3. Generalization to more
complicated problems.
-Up
to now, we were concernedonly
with a verysimplified problem : desorption
of apoint particle, trapped
in a one-dimensionalwell,
in the absence of any activation barrier. We now consider how such acrude model could be
extended,
at least inprinciple.
Let us first assume that the
adparticle
has additionaldegrees
offreedom, beyond
the motion in the x-direction normal to the surface. These
degrees
offreedom may be for instance motion
along the y
and z axis
parallel
to the surface(in
the three-dimen- sionalproblem). They
may also be internal motion of a molecularadparticle,
whether rotational orvibrational. Let n denote the state of these extra
degrees
of freedom. The net energy of theparticle
isIt is clear that
only Ex
is relevant asregards desorption
and
trapping :
theparticle
will escape if it hits theedge
of the barrier with a
positive
Ex. The extension of ourapproach
is thenstraightforward.
LetPn(ex,)
be theprobability
that aparticle
with normal energy Ex,internal state n, be
ultimately trapped
in thepotential
well.
Similarly,
letWnn,(Ex, E;x
be theprobability
thatin one round
trip
theparticle
goes from(n, E,x
to(n’, Ex)
xFrom W,
we infer theequation obeyed by
thesticking probability
(18)
is to be solved(for
instanceby iteration),
with theboundary
conditionsPn(Ex)
- 1 if Bx --> - oo. Oncewe know
Pn(Ex),
weproceed
to the averagesticking
coefficient of
incoming
thermalparticles
We are now in a
position
toapply
oursimple
detailedbalance
argument.
The number ofparticles trapped
in the well is
where
Zm
is the internalpartition
function forparticles
near the bottom
(note
the difference with(2)
due tothe additional
degrees
offreedom). Similarly,
the fluxof
particles arriving
from the gasphase
onto thesurface is no
longer (3)
but ratherwhere
Zo
is the internalpartition
function near the top of the well(not necessarily equal
toZm).
Onequating
the in and out
fluxes,
we find thedesorption
rateThe absolute rate is corrected
by
theinternal degrees
of freedom - but the relation
u/uo
=P
is unaffected.The main effect of the internal
degrees
of freedomis to blur the
probability
W : the normal energyloss, (ex - 8§),
may now arise from a transfer between internal energy uand ex itself,
apurely
elastic process(which
adds to the former inelasticexchange
with theheat
bath).
In any case, whatever the average loss 6 in normal energy was, additionaldegrees of freedom
will make it
bigger.
The two
limiting
cases areagain
trivial. Letbn
be theenergy loss
averaged
over all final channels n’ :If
dn > T,
thetrapping probability
P is1, irrespective
of the
details,
andEyring’s
absolute rate is correct,a situation even more
likely
than wasanticipated
insection 3. If
by
chance elastic processes alloweddn T,
asimple
extension ofAppendix
A wouldshow that
We now return to a one-dimensional
point particle,
and we consider another
complication, arising
whenthe
potential profile V(x)
is not monotonic. In the activateddesorption
barrier offigure 4,
theparticle
must climb a hill of
height Va (the
activationenergy)
before
falling
into the well ofdepth Vm. Clearly,
thereverse
desorption
process shoulddepend
on the netdesorption
barrierheight
FIG. 4. - An activated desorption barrier.
More
generally,
we can consider reactions betweentwo
equilibrium positions
1 and 2(one
of them meta-stable),
as shown infigure
5.FIG. 5. - A reaction path between two equilibrium positions 1
and 2.
Such a
picture
is often used to represent reactions inside a molecule(for
instance in an adsorbedlayer).
The one-dimensional model
(4)
is then the reaction (4) In such molecular reactions, the one-dimensional reactionpath is only a first approximation. Eventually, one must also worry about motion in the directions orthogonal to that path (up the
flanks of the hill at the saddle point in V). Such motion can be viewed as additional internal degrees of freedom superimposed on
the motion in the x-direction along the reaction path.
path,
the steepest dscentpath
that overcomes thesaddle
point separating
twovalleys
in the multidi- mensional energy surface.We take the
origin
of energy at the top of thebarrier,
and we denote
by Vi
=Va - Vmi,
thedepth
of thewell on either side. As
before,
we assumedeep wells,
i.e.
Vi >>
T. In thermalequilibrium,
the number ofparticles
on each side is thuswhere Wi is the oscillation
frequency
at the bottom of each well(see (2)).
We now empty well 2. LetJ1-+2
be the flow of
particles coming
from 1 that fall to thebottom of 2 : the reaction rate is defined as
A similar definition holds for the reverse rate U2-+ 1.
In the absolute rate
theory,
one assumes thermalequilibrium everywhere, including
near the top of the barrier. Then the kinetic argument ofEyring yields
at once
Of course, detailed balance is verified
Such a result is
subject
to the same limitations asin section 2 : it assumes
complete trapping
ofparticles
that pass
through
the barrier at x = xa. If infigure
5they
cross the barrier from left toright, they
aretrapped
in2 ;
ifthey
go fromright
toleft, they
fall in 1.Assuming
this to be true,Eyring’s
result(22)
is exact.In order to calculate
departures
from the absolute rates, we must describe moreprecisely
thisproblem
of
trapping.
In thelarge
friction case(il
>w),
themotion is
diffusive,
and it is easier to calculate udirectly,
as doneby
Kramers. One recovers result(6),
except that the
integral
in the denominator extends from xl to X2(anyhow,
it is controlledby
thevicinity
of
xa). Here,
we consideronly
the weakdamping
case
(n w),
where the motion isoscillatory.
Let
Pi(e)
be theprobability
that aparticle entering
well i
(1 or 2)
will betrapped
in that well without evercrossing again
the barrier at xa(it
remains inwell
i until it falls to thebottom). Conversely,
letQ(e)
debe the
probability
per unit time that aparticle starting
from the bottom of well i crosses the barrier at xa with an energy in the range
(s,
B +ds)
andfor
thefirst
time.
Clearly, Pi
andØi
are relatedby
detailedbalance;
it is
easily
shown thatNow,
when aparticle
with energy 8 enters well1,
it must end upsomewhere,
withprobability Ql(g)
to be in well
1, [1 - Ql(e)]
to be in well 2. Thistime,
there is no restriction on
Q, :
theparticle
may returnas much as it wants to the 2 side before
eventually falling
into 1.Similarly,
we denote asQ2(e)
the pro-bability
that aparticle crossing
the barrier toward 2 willeventually
end up in 2. With all thesedefinitions,
the net rates from one well to the other are
easily found,
e.g.(we identify
thefirst
passagethrough
x =xa). Using (22)
and(23),
we see that(24)
extends our former result(7).
It is shown inappendix
B that a12 = a21, so that detailed balance ispreserved.
In order to
proceed
further,
we introduce the condi- tionalprobability W1 (8, E)
thatduring
one roundtrip
in well
1,
theparticle
goes from energy 8 to 8’. Simi-larly, W2(E, 8’)
refers to one roundtrip
in well 2.Upon isolating
thefirst
roundtrip
in1,
we obtaineasily
theequations obeyed by Pl
andQ 1
(equations
forP2
andQ2
are obtainedby
the inter-change
1 ->2).
Onceagain,
theimportant
parametersare
61
and62,
the average energy lossesduring
oneoscillation in well 1 or well 2. The same two
limiting
cases are
simple.
(i) If ð1, ð2 > T, trapping
iscomplete
in the energy range ofinterest,
8 - T :According
to(24),
the absolute ratetheory
is correct,a conclusion which is insensitive to our various appro- ximations
(markovian evolution, etc...) :
aparticle
that crosses the barrier cannot
regain
the energy lost in one oscillation.(ii) If ð1, ð2 T,
we know how to calculatePi(e).
Moreover, we show in
appendix
B that in therange
of
interest,
0 8 %T,
theprobabilities Ql
andQ2
are
essentially
constant((26)
was indeed to beexpected :
sincestatistically trapping
occursonly
after manyoscillations,
it doesnot matter either where or in which direction one crosses the barrier : hence
Q,
= 1 -Q2, prorated by
the energy losses
61
and62)- Using (13)
and(26),
we find the correction to
Eyring’s
rateAs an
application,
let us considerdesorption
withan activation
barrier,
as shown infigure
4. «1» isthe
well,
0 2 » the open space. The absolute rate,given by (22)
involves the totalheight V
=(Ya - V oJ
of the barrier to be climbed. In order to find the
correcting
factor(24),
we note that aparticle entering region
2 never returns. HenceThe correction to the absolute rate thus retains the
same form
(7)
as in the case of no activation.4. Time
dependent
stochastic formulation. - All theforegoing analysis
is based on animplicit
assump- tion that we canunambiguously
separate two pro-cesses :
(i) trapping
in thepotential well, (ii)
subse- quentdesorption.
The two processes are treated asstatistically independent.
We now examine to whatextent such an
assumption
is valid.Let us consider an
incoming
gasparticle. Strictly speaking,
it is nevertrapped !
Even if it falls down to the bottom of thewell,
it willeventually evaporate again.
Thus absolutetrapping
ismeaningless.
Whatwe must consider instead is the
probability P(8, t)
that the
incoming particle
be stilltrapped after
atime t. As a function
of t,
we then expect two stages :(i)
Atfirst,
theparticle
remains close to the top of thebarrier,
within alayer
of width T. Then there is still a fair chance that it canpick
upenough
energy from the heat bath to escaperight
away. The corres-ponding
time scale t, istypically
a fewperiods
ofoscillation. If one waits
too long,
thesystematic
lossafter n round
trips, n3,
overcomes any diffusioneffect ~ ilL1] :
with anoverwhelming probability,
the
particle,
if it has notescaped before,
falls towardthe bottom of the well.
(ii)
Thenbegins
thedesorption period :
the fallenparticle
triesagain
andagain
to climb up thepotential
well - and it will
eventually succeed,
after atime t2
of order
1 /u.
The
corresponding
behaviour ofP(8, t)
is sketchedin
figure
6. At t =0,
we start from some energy 8 close to the top : theparticle
is in andP(8, 0)
= 1.At each round
trip,
theparticle
has someprobability
to escape, and