HAL Id: jpa-00246450
https://hal.archives-ouvertes.fr/jpa-00246450
Submitted on 1 Jan 1991
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.
Repton model of gel electrophoresis and diffusion
B. Widom, J. Viovy, A. Defontaines
To cite this version:
B. Widom, J. Viovy, A. Defontaines. Repton model of gel electrophoresis and diffusion. Journal de
Physique I, EDP Sciences, 1991, 1 (12), pp.1759-1784. �10.1051/jp1:1991239�. �jpa-00246450�
Classification
Physics
Abstracts05.40 66.10 82.45
Repton model of gel electrophoresis and diffusion
B. Widom
(*),
J. L.Viovy
and A. D. DefontainesLaboratoire de
Physicochimie Thdodque,
ESPCI, 10 rueVauquelin,
F-75231Pads Cedex 05, France(Received
3 June 1991, revised13August, accepted14 August)
Abstract. We
analyze
the repton model of Rubinstein asadapted by
Duke as a model for thegel electrophoresis
of DNA. Parameters in the model are the number Nor reptons in the chain, alength
a, amicroscopic
transitionfrequency
w, and theproduct
eE of the electric field E(assumed constant)
and thecharge
e per repton.Fornlally
exact formulas are derived for the dimensionlessdiffusion coefficient
Dla~w
and driftvelocity Vlaw,
the latter as a function of the field.Calculation of Vlaw
requires
theeigenvector
associated with the leading eigenvalue of a 3~ x 3~ ' matrix. For short chains exact results are obtainedanalytically
:Vlaw
for all eE for I « N « 4, andDla~
w for I « N « 5. For large N we deduce thatDla~
w vanishes
proportionally
to
I/N~,
the standard de Gennes reptation result, but we have not evaluated the coefficientanalytically.
We have determinedDla~w
for N up to 150by
simulation and verified the I/N~
law.1. Indoducfion and outline.
Rubinstein
[I]
has introduced a model that he calls the repton model for the diffusion of apolymer
chain in a dense medium. It is a lattice model thatincorporates
and expresses de Gennes'reptation
mechanism[2]
in its purestform, viz.,
as the diffusion of storedlength along
the chain's own contour. The chain isrepresented by
a sequence of reptons(beads)
connected
by
links. A number 0 or I is associated with eachlink, according
to whether the associated beadsbelong
to the same site or toadjacent
ones. This model wasgeneralized by
Duke[3]
to a chain of three-state links(0, 1, 1),
as a model for thegel electrophoresis
of DNA. To avoid anyconfusion,
we refer in thefollowing
to theoriginal
modelIi
as the « two-state repton model » and to Duke's
generalization [3]
as the « three-staterepton
model ».This paper is concemed ~vith the latter.
For a
description
of thephysical
model in the real two- or three-dimensional space of thediffusing
chain we refer to theoriginal
papers[1, 3J;
here we shall describeonly
theequivalent
one-dimensional model into which that one maps. Themapping
was demonstratedby
Rubinstein[1].
(*)
Permanent address: Department ofChemistry,
BakerLaboratory,
ComellUniversity,
Ithaca, New York 14853, U-S-A-Figure
I shows a connected chain of N « reptons », eachrepresenting
some(large)
number of basepairs
of acharged
DNA chain. The discrete variable x~= .,
2,
1,0,
1,2,
(I
=I,.., N) gives
the location of the I-threpton,
hence of the I-th element of the DNAchain,
in some direction of space that isconveniently
chosen to be that of an extemal electric field E that acts on thecharges
of the chain. When E=
0,
so that the chain diffuses but has no net driftvelocity,
this direction isarbitrary.
The chainconnectivity requires
that Ji~ t x, =0 or ± I for all I
=
I,
,
N -1.
6 5
~
-i
2 3 4 5 7 8 9 II 13
repton number
Fig.
I.-Repton
chain with N =15 in arepresentation
with asingle spatial
coordinate x in the direction of anextemally applied
electric field E(after
Duke, Ref. [3]). The arrows represent allowedmoves.
Neighboring
reptons that have a common value of their x coordinate represent, via themapping,
a cluster of successivereptons occupying
a common cell in theoriginal
two- orthree-dimensional space. In the
example
offigure I,
taken from Duke[3], reptons
2 and 3 area cluster of two,
reptons 8, 9,
10 are a cluster ofthree,
and14,
15 are another cluster of two.The
reptation
mechanism in theoriginal
space then maps intodynamical
move rules in the space offigure
I. An interior repton I(I
=2,
...,
N
I)
is allowed to moveonly
if it is the endrepton
of acluster,
and then it is allowed to moveonly
in the direction of that one of its twoneighbors
that does notbelong
to the cluster. This is shown infigure
Iby
the arrows,I
or(,
on the interior reptons
2, 3, 8, 10,
and 14. An endrepton (I
= I or
N~
isalways
allowed tomove. If its coordinate x is not the same as that of its
neighbor,
so that it is not part of acluster,
it is allowed to moveonly
in the direction of itsneighbor,
asexemplified by repton
I infigure
I. If the end repton is at the same x as itsneighbor,
so that it is part of acluster,
it isallowed to move either
I
or
(,
asexemplified by repton15
infigure
I. A moveI
or( by
arepton changes
its x coordinateby
+ I orI, respectively.
Define
B
=
e~~W~~~
(I.I)
where E is the constant extemal electric
field,
e thecharge
on each repton, and a asingle
length parameter
that may be taken to be both thelength
per repton and the linear dimension of a cell(gel pore)
in theoriginal
space ; and let w be a constant transitionprobability
per unittime. We then
assign
wB and wB as the transitionprobabilities
per unit time for an allowedmove
I (direction
ofincreasing x)
or( (direction
ofdecreasing x), respectively.
This is a little different from theassignments
madeby
Rubinstein[I]
and Duke[3],
who inassigning
transition rates to arrows
distinguish
between arrows that are on interior and on endreptons,
and between the two arrows on an endrepton (when
it carriestwo).
Such differences in theassignment
of transition rates affect thequantitative
but not thequalitative
behavior of the model. We choose to haveonly
the transition rates wB or wB for everyI
or(
arrow without further distinction so as not to have any moreparameters
thanN,
a, w, and eE.The
configuration
of therepton
chain in the one-dimensionalrepresentation
offigure
I(one spatial dimension, x) changes
in timeaccording
to the rulesjust specified. Any
one suchchange
ofconfiguration, I-e-,
any allowed moveby
asingle
repton,changes
the meanposition
x of the
reptons,
and so, as time progresses, we see a drift of the chain as a whole in the x direction. This may be viewed as theprojection
onto one dimension of the drift of theoriginal
chain in its
original, lligher-dimensional
space. The driftvelocity
and diffusion coefficient of the chain in the x direction in therepresentation
offigure
I are the same as that of the realchain in the
original
space. Because the moves arestochastic,
the transition ratesbeing
transitionprobabilities
per unittime,
we should think of the chain as one member of an ensemble of many suchchains,
all of the samelength
N.Since x;~t -x; for each I
=
I,...,N-
I may have any of the three values 0 or±
I,
therepton
chain has3'~~~ possible
distinct internalconfigurations irrespective
of its location as a whole in the x direction. We index those internalconfigurations by
y. We choose any one of the x; to define the location of the chain in the x direction. We call that Jisimply
x. Then x and ytogether
define both the location and intemalconfiguration
of thechain.
Let
p~,~(t)
be theprobability
that at time t the chain will be at x inconfiguration
y. In section 2 we write andformally
solve theequation
that determinesp~,~(t)
for anygiven p~,~(0).
The solution involves theeigenvalues A~(q) (n =0,
1,2,
..,
3'~~~ -1)
of a3'~~~ x3'~-~
matrix of transitionprobabilities
the elements of whichdepend
on a(dimensionless)
wave number q as well as on B.We remark in section 2 that one of the
eigenvalues,
which we label n=
0,
vanishes at q =0,
and that thefield-dependent
driftvelocity
V and diffusion coefficient 5~ of the chain areobtained from the coefficients in the
expansion
ofAo(q)
in powers ofqla,
A
o(q)
= I
Vqla s~q ia~
+(1.2)
V and 5~ are
respectively
odd and even in eE,V(eE)
= V'
(0 )
eE + V"'(0 ) (eE)~
+5~(eE)
=
5~(°)
+5~"(°)(eE)~
+(l.3)
Here
5~(0)
=
D, (1.4)
the
ordinary phenomenological
diffusion coefficient in the absence of an external field.Since each repton carries the
charge
e the total force exertedby
the field E on the chain is NeE.Hence, by
the Nemst-Einstein relation[4],
the coefficientV'(0)
in the first ofequations (1.3)
and the diffusion coefficient D in(1.4)
must be relatedby
V'(0)/N
=
D/kT. (1.5)
This is a necessary relation between
a~Aolaq
aE anda~Aolaq~
at E= 0 that should in
principle
be derivable as a property of the matrix of wuich
Ao(q)
is theleading eigenvalue.
As apurely
mathematicalproblem
associated with this model it is achallenging
one and has not been solved forgeneral N, although
we later(Sect. 4) verify
it for N= 1,
2,
and 3. In sections3, 4,
and 5 we use(1.5)
to obtain D fromV(eE)
even for N~
3,
on thegrounds
that as aphysical,
as distinct from
mathematical, principle (1.5)
is unassailable and we are confident that it is satisfiedby
this model for all N.There is another as
yet
unsolved mathematicalpuzzle
in tillsmodel,
to which we areagain
confident of the answer onphysical grounds. Physically,
the driftvelocity V(eE) (hence
alsoD)
must beindependent
of which x;(I
=
I,
,
A§
is chosen to be the variable x thatspecifies
the location of the chain as a whole in the x direction. The same is
probably
true, moregenerally,
of all theeigenvalues A~(q).
Thequestion
arisesonly
forq#0;
whenq=0
the matrixA(q)
of which theA~(q)
are theeigenvalues (Sect. 2)
ismanifestly
independent
of thatchoice,
butexplicitly depends
on it when q # 0.Also,
thequestion only
comes up for N m 3 ; for N
= I there is
only
one x;, while for N=
2, although
there are two,they
areequivalent by
symmetry. In section 4 weverify
thatV(eE)
for N=
3 is indeed
independent
of which x, is chosen to be x. In section 3 weexploit
thecertainty
thatV(eE)
isindependent
of that choice for any N(although
it remains unprovenmathematically)
to
simplify
thegeneral
formula forV(eE) by removing
from it reference to thearbitrary
choice of marker.
In section 3 we obtain
formally
exact formulas for the driftvelocity
V and diffusion coefficient D based on(1.2)-(1.5)
and other results in section 2.Then,
in section4,
weexplicitly
calculate the dimensionlessV(eE)law
for N=1,
2,3,
and4,
andDla~w
for N=1, 2, 3, 4,
and 5.We deduce in section 5 that
Dla~w
vanishesproportionally
toI/N~
as N- ~x~, which is characteristic of
reptation [2],
but we have not evaluated the coefficient. In section6,
we present the results of simulations for N up to 150 we make some connections with earlierwork,
andverify
theI/N~
law. In section 7 we summarize our results and referbriefly
tocurrent work on
scaling properties
of the model forlarge
N and small E.2,
p~,~(t);
identification of Vand D.In section I we chose an
arbitrary
one of the repton coordinates x; and called it x, we indexed the3'~
internalconfigurations
of the chainby
y, and we defined p~,~(t
to be theprobability
that the chain is at x in internal
configuration
y at time t. The move rules and transition rates described in section I may berepresented symbolically by time-independent probabilities
perunit time W~,
~, ~ ~ of the transitions x, y
-
x', y'.
Thendp~,~(t)/dt
=
it W~,~,~,,~, p~,,~,(t) W~,,~,,~,~ p~,~(t)j (2.1)
From the rules of section I we know that W~, ~,.~
~
for
given
y andy' depends
on x and x'only
via the difference x' x and vanishes iftiis'difference
is not 0or ± I. We then define 1k~7~~,
lf~i),
andlf~,§~ by
w(~
) ~'_ ~= j
y y '
w(f)
~'_~ ~ ~
w,
~
y y'
~ ~~ "~
lljfy~,
x'- x #0
,
otherwise
(2.2)
By
thedynamical
rules in section I many of theIk~T~),
W)I),
and lk~f~~ vanish. Theonly lljT~~
that do not vanish are those for which the transition y -y'
in the internalconfiguration
occurs
by
an allowed move( by
the markerrepton (the repton
whose x coordinate we haveadopted
as the x variable for thechain),
and then thatnon-vanishing llj§~
is wB-[with
B defined in(I.I)]
theonly lljf~)
that do not vanish are those for which the transitiony -
y'
occursby
an allowed moveI by
the marker repton, and then thatnon-vanishing
ll~§~
iswB;
and theonly W)I)
that do not vanish are those for which the transitiony -
y'
occursby
an allowed moveby
somerepton
other than the markerrepton,
and then thatnon-vanislling lk~i)
is wB if therepton
move that causes the transition isI
andwB-'
if it is(.
The solution of
(2.I)
is then of the formJ'x,
y(t)
~
£ j~ Cn(q) fj~~(q)
~~~~~~~~~~dq (2.3)
where the
A~(q) (n
=
0,
1,2,.., 3'~~
lare the
eigenvalues
of the3'~
x3'~~
transition-rate matrix
A(q)
with elementsA~~,(q)
definedby
~YY' ~~iY'~~~~
~
~~~
~~~i~
~ ~~ ~Y' # Y~Ayy =
Z (Hl~y~
+Wlfl
+Wlty~), (2.4)
y,<~y>
where the
fj~~(q)
are the y-components of the associatedright eigenvectors fl~~(q),
~(~) f~~~(~)
~
~
n(~) f~~~(~)
,
(2.5)
and where the functions
c~(q)
are determinedby
the initial p~~(0). Formally,
if the lefteigenvectors
ofA(q)
are calledg~~)(q),
g~~~
(q)
A(q)
"
An (q ) g~~>(q )
>
(2.6)
and if we choose the normalization
g~~~(q) f~~~(q)
= ,
(2.7)
thenCn(~)
~(2 ")~ z z
~~ ~~~~j~~(~)Px, y(0) (2.8)
We note from
(2.4)
that thediagonal
elements A~~ ofA(q)
areindependent
of q and arerelated to the
off-diagonal
elements at q= 0
by
A~~ =
£ A~,~(0) (2.9)
v>(#y>
It then follows from
(2.5)
and(2.6)
that when q = 0 there isnecessarily
aneigenvalue A(0)
=0 and the associated lefteigenvector g(0)
is a constant;I-e-,
itscomponents
g~(0)
areindependent
of y. We index thiseigenvalue by
n= 0 :
Ao(0)
= 0.
(2.10)
At all other q this
Ao(q)
has anegative
realpart,
as do all the otherA~(q)
at all q ; if anyA~(q)
had apositive
realpart p~,~(t)
would increase without bound as t- ~x~, which cannot be.
The
long-time
behavior ofp~,~(t)
is dominatedby
the n= 0 term in
(2.3),
andby
q near0,
because of(2.10). Thus,
Px,y(t) j~ co(q) f)°~(q) e'~~~
~°~~~dq (t
- ~x~), (2.i i)
with the
major
contributioncoming
from theneighborhood
of q= 0. Then if
Ao(q)
isexpanded
as in(1.2),
we find that afterlong times,
@Px,y/°t
V@Px,y/@(aX)
+ 5~@~Px,y/@(ax)~
+(2.12)
(Recall
that x isdimensionless;
it is ax that measures distance in thex-direction.)
Wecompare this with the
phenomenological equation aplat
=
Vapla(ax)+
5~a~pla(ax)~
appropriate
to thelong-time, small-gradient limit,
with V and 5~ the(field-dependent)
driftvelocity
and diffusioncoefficient,
and thus make the identificationsanticipated
in sectionI, Equations (1.2)-(1.4).
We refer to the
long-time
limit in(2. II)
as the «steady
state », and we then see that theeigenvector #°)(0)
associated with theleading eigenvalue o(0)
at q = 0 isproportional
to thesteady-state
distribution among the internalconfigurations
y of thedrifting
anddiffusing
chains. Its
components f)°)(0)
may then be taken to bepositive
real numbers. We remarked above that the componentsg)°)(0)
of the associated lefteigenvector
areindependent
of y. Ifthe common value of these
components
is chosen to beI,
and if we continue toadopt
thenormalization
(2.7),
then thesteady-state
distributionf)°)(0)
would be normalized tounity, g)°)(0)
m I
,
£ f)°)(0)
=
(2.13)
y
When the extemal field
vanishes,
so that B=
I,
theprobability
per unit time of any transition and its inverse areequal.
Thesteady-state
distribution among internal states y is then theequilibrium
distribution in which all states areequally populated.
With thenormalization
(2.13),
andrecalling
that there are3'~~
internal states,lim
f)°)(0)
w
1/3'~
,
(2.14)
B ~ l
independently
of y. In this same E=0 limit theeigenvalues A~(0)
atq=0
forn # 0 are all
negative
real numbers.They
are thenegatives
of the relaxation rates(reciprocal
relaxation
times)
in the modes n= 1,
2,..., 3'~
lby
which the distribution among internal states relaxes to theequilibrium
distribution(2.14)
when there is no extemal field.When E # 0
(B
# I),
thef)°)(0)
dodepend
on y ; thesteady-state
distribution is not a uniform one. Notonly
is such asteady-state
distribution different from the uniformequilibrium distribution,
but it is alsopartly
maintainedby cyclic
processes,which,
atequilibrium,
would be ruled outby
detailed balance. Infigure
2a is anexample
with N=
3,
in which three of the nineconfigurations ~labeled
y=
0,
1, and3, following
the nomenclature we shall introduce in Sect.4)
areshown,
with arrowsI
and( indicating
themoves allowed
by
the rules of section I. If eE~
0,
say, so that B~ l
by (I.I),
the transitions y = 3- - 0
- 3 are all « downhill »
(transition
rate wB~
w)
while y=
0
- - 3
- 0 are
all «
uphill
»(transition
rate wB~ <w),
as shownschematically
infigure
2b. That each of theseconfigurations
is bothuphill
and downhill from itself is aparadoxical
circumstance that~ ~~~ @
§q~ Y~' wB~' ~ ~ wi'
~~~§ Y~3 f ~@
wB"
(a) (b)
Fig.
2.(a)
Three of the nineconfigurations
of a repton chain with N= 3, and
(b)
the transition ratesconnecting
thoseconfigurations
inpairs.
cannot occur at
equilibrium
but can and does occur in thesteady
state. For thesteady-state
distribution to be an
equilibrium
one theproduct
of the rates of the transitionsgoing
clockwise around anycycle
mustequal
theproduct
of the counterclockwise rates,by
detailedbalance. In the present
example, by
contrast, those twoproducts
arew~B~
andw~B~~,
which areequal only
when E= 0
(B
=
I
).
When
llji)
in(2.2)
and(2.4)
is wB(or wB~~)
thenlljf~)
iswB~~ (or wB)
andlk~i)
=lf~()
= 11~7~) =W)))
=0,
whereas when the transitions y ++y'
do not involve themarker repton then
lljj
= W)T~) =
ll~)
)=
lf~f~)
=0 while
ll~(I
=
wB
(or
wB~ ~) andll~i)
= wB
(or wB). Thus,
except when B=
I,
the matrixA(q ),
the elements of which aregiven
in(2.4),
is not Hermitian forgeneral
q norsymmetric
at q = 0.Because,
as we saw, there are unbalancedcycles
in the transitions amongconfigurations
when B #I,
the matrixA(0)
is notonly
notsymmetric
when B # I but is then not evensymmetrizable I-e-,
there isno vector f
[and,
inparticular, f1°)(0)
is not such avector]
such thatf~A~,~(0)m
f~,A~~,(0);
which is another way ofstating
the failure of detailed balance. SinceA(0)
for B # I is neithersymmetric
norsymmetrizable
any of theeigenvalues A~(0)
may becomplex [except
forAo(0),
whichby (2.10)
isalways 0].
When~(0)
iscomplex
the relaxation to thesteady
state via mode n isoscillatory
in timealthough
the oscillations areexponentially damped
since the real part ofA~(0)
remainsnegative. Complex A~(0)
for B # I are illustrated for N= 3 in section 4.
The sum
lljj )+ W)(I
+W)))
is the transitionprobability
per unit time of the transitiony'
- yirrespective
of whether that transition occursby
move of the marker repton orby
moveof any of the N I non-marker
reptons.
That sum is wB if the transitiony'
- y occurs when
some repton moves
I,
it is wB if the transition occurs when some repton moves(,
and it is 0 if nosingle
allowed moveby
anyrepton
canbring
about that transition.By (2.4)
the matrixA(0)
is thenindependent
of thearbitrary
choice ofmarker,
and so then also are all theeigenvalues A~(0).
But that is not true ofA(q)
when q # 0 : some of theoff-diagonal
elements are wB~ ~, some are wB e~~, some arewB,
and some are wB e~'«,
and which onesare in which
positions
in the matrixdepends
on thearbitrary
choice of marker.Nevertheless,
since that choice isarbitrary
nophysical quantity
candepend
on it. Inparticular,
the driftvelocity V(eE),
hence also the diffusion coefficientD,
must beindependent
of that choice.As we remarked in section
I,
we shall later(Sect. 4) verify
thatindependence
forN
=
3,
but elsewhere(starting
in Sect.3)
wejust
assumeit,
eventhough
it has not yet been provenmathematically
for N~ 3.
3. Formulas for V and D.
It is apparent from the
expansion (1.2)
and from the identification of the coefficient ofiqla
in thatexpansion
as the driftvelocity
V(Sect. 2)
that V forgeneral
eE may be obtained fromA(q) by
first-orderperturbation theory
withA(0)
treated as theunperturbed
matrix.There are two distinct routes to the diffusion coefficient D. One is via
(1.2)-(1.4),
from whichit is seen that D may be obtained from
A(q) (at
E=0) by
first- and second-orderperturbation theory [terms
O(q)
in theperturbation
to be treated in second order and thoseO(q~)
to firstorder],
withA(0) (at
E=
0)
theunperturbed
matrix. The second route to D is via(1.5),
onceV(eE)
has first been foundby
first-orderperturbation theory
as describedabove.
By
this route,then, differentiating V(eE)
to obtainV'(0)
isequivalent
togoing
to the next order ofperturbation theory
at E= 0.
From
(2.4),
they'y
element of theperturbation
isAy~
y(q) Ay~y(0)
" H§~y~ Wj~y~)
iq
(11~~7y~ +Wjfy~) q~
+ O(q~) (3.1)
(The configuration changes
if anyrepton
moves,including
the markerrepton,
sonecessarily lJ§p
) =W)(
) = 0 for anyy.)
Then with theright eigenvector #°)(0)
normalizedby (2.13),
and
recalling (2.10), Ao(q)
"
iq z z ( Hj~y~ Wjt~~) /j°~(0)
+ O(q~) (3.2j
We also recall that if the transition y
-
y'
results from the allowed move( by
the marker repton thenlljT~)
= wB andlJjt~)
= 0 ; if that transition results from the allowed moveI by
the marker repton then
lJ~t~)
= wB andlljT~)
=0 and
llj,j)
=
lljf~)
= 0 otherwise.Therefore
Ao(q)
=
iqw B~
~if)°~(o)
B~( fj°)(o)j
+ o(q2), (3.3)
<->
where
£
means sum over allconfigurations
y from which a moveI by
the marker repton is~ i+
allowed and
£
means sum over allconfigurations
y from which a moveI by
the markery
repton is allowed. Then from
(1.2),
Vlaw
= B£ f)°~(0)
B~jj f)°)(0) (3.4)
The formula
(3.4)
for the driftvelocity
V is exact for allE,
but it still makes reference to anarbitrarily
chosen markerrepton.
Weknow, however,
that it isreally independent
of thatchoice
(as
we illustrate for N =3 in Sect.4). Hence,
we may choose eachrepton
I
=
I,
...,
N in turn to be the marker repton, write
(3.4)
for each suchchoice,
add all theresults,
and then divideby N,
thusobtaining
forVlaw
a formula that nolonger
refers to anymarker :
Viaw
= N
z (Br~
Bs~) /j°~(0)
,
(3.5)
where r~ is the total number of
I
arrows on all of thereptons
of the chain inconfiguration
yand s~ is the number of
(
arrows.
(For example,
in theconfiguration
y illustrated forN
= 15 in
Fig. I,
r~=
5 and s~ =
3.)
Equation (3.5)
forVlaw
is exact for the model. Ityields
the driftvelocity
V in fields E ofarbitrary strength.
Itrequires knowing
thefight eigenvector #°~(0)
associated with theeigenvalue Ao(0)
= 0 of the transition-rate matrix
A(0).
From(2.4) (with
the known values of W)7~),etc.), (2.5)
and(2.10),
and with the same r~, s~ notation as in(3.5),
theequations
for the componentsf)°~(0)
areB
£ f)?~(0)
+ B£ fj?~(0)
=(Br~
+ B 's~) fj°~(0
,
(3.6)
~Y.
I)
I<y'£)1
where the notation is meant to indicate that the first sum is over all
configurations y'
from which theconfiguration
y may be achievedby
an allowed moveI by
any repton of thechain,
and likewise for the second sum with(
instead ofI.
The condition on the first sum couldequally
well have been written y-y'(, meaning
a sum over all
configurations y'
achievable from theconfiguration
yby
an allowed move( by
any repton of the chain. Thereare therefore as many terms in that sum as there are
(
arrows on all thereptons
of the chain inconfiguration
y ;viz.,
s~ and there are likewise r~ terms in the second sum.Note, then,
thaton the left of
(3.6)
the factor Bmultiplies
s~ terms and the factorB~~ multiplies
r~ terms, whereas on the
right
it is theopposite.
For use in(3.5),
the solutionf)°~(0)
of(3.6)
is to be normalized as in(2.13).
The first route to the diffusion coefficient D as described earlier is via
(1.2)-(1.4).
Thisrequires
the terms O(q~)
in(3.2),
at E=
0.
They
are obtainedby taking
account of the termsO(q)
andO(q~)
in(3.I)
in second- and first-orderperturbation theory, respectively.
TheO(q)
term inAo(q) (Eq. (3.2))
vanishes at E= 0 because
f)°)(0)
isindependent
of y when E=
0, by (2.14),
and because£ £
W)T~)=
£ £
W)f~~,by
symmetry.(Thus,
the driftvelocity
vanishes at E=
0,
as itmust.)
Taking
account of(2.13)
and(2.14),
we see that the contribution to the terms O(q~)
inAo(q)
that come from the termsO(q~)
in(3.I) by
first-orderperturbation theory
is~~
i~- I ~'Y~
~~~~~
~~$~ ~~'~~
Y Y
where M is the number of distinct
configurations
y in which the markerrepton
carries anI
arrow
(=
the number in which it carries a(
arrow,by symmetry).
The contribution to theterms
O(q~)
inAo(q)
that come from the termsO(q )
in(3.I) by
second-orderperturbation
theory
is~2 ~2
~
z j-
A~(0)i~ iv f~~~(0)i~ (at
E=
0
,
(3.8)
3 n<#o>
where the components v~ of the vector v are
v~ = I if in
configuration
y the markerrepton
carries anI
arrow but not a(
arrow,v~ = I if it carries a
(
arrow but notan
I
arrow, andv = 0 otherwise
(either
no arrow or both anI
and a( arrow)
; and with theeigenvectors
/~(0)
forn # 0 normalized
by
j#~~(0)
~=
l
(n
# 0) (3.9)
[This
is in contrast with(2.13),
which is the more convenient normalization for n=
0,
whenthe
components f)°)(0)
areproportional
to thesteady-state populations
and may thus be taken real andpositive.
Since E= 0
here,
the matrixA(0)
issymmetric,
so except forarbitrary
normalization constants the lefteigenvectors g~~)(0)
arejust
theright eigenvectors
#~)(0) transposed.
For convenience we takeg)~~(0)
=f)~~(0) here,
for n #0,
and then(2.7)
becomes
(3.9).J
The sum of
(3.7)
and(3.8)
is theO(q~)
term in(3.2)
at E= 0.
Therefore, by (1.2)-(1.4),
the dimensionless diffusion coefficient
Dla~
w isf
=
/ (M
w£ [-
A~(0) J-
[v~~)(0)J~)
,
(3.10)
a w 3
~,~o~ E=o
with M and v defined beneath
(3.7)
and(3.8), respectively,
and with#~)(0)
normalizedby (3.9). By
a convention fornumbering
the states y that we shalladopt
in the nextsection,
thecomponents
v~ of v will be odd in y while thecomponents f)~~(0)
will be either even or odddepending
on n. The summation over n in(3.10)
may then be restricted to those n for whichf)~)(0)
is odd in y. There is still in(3.10)
aseeming dependence
on the choice of marker repton(via
v andM~,
but as we shall seeexplicitly
for N=
3 in the next
section,
and as we know fromgeneral principles
for allN,
thatdependence
isonly
apparent, not real.The second route to D is via
(3.5)
and(1.5). Suppose
for small E(I.e.,
for B nearI)
thefj°~(0),
which at E=
0 are
given by (2.14),
have theexpansion
f)°~(0)
=/ II
+
ay(B
I)
+ OI(B
I)~l (3.I1)
Then from
(3.5), (1.5)
and(I.I),
~~~~
~~
~~2~~N
-1
I
~~Y ~~Y ~~Y ~Y~ ~Y~
(~.l~)
Y
floote
from the definition of s~ and r~ beneath(3.5)
that£ r~
=
£
s~,by symmetry.]
It follows from
(3.6)
and(3.I I),
andby recalling
that there are s~ terms in the first sum in(3.6)
and r~ terms in thesecond,
that thea~ satisfy
£
a~,= 2
(r~ s~)
+(r~
+s~) a~ (3.13)
y>
<y> -y>
where the notation
(y'++ y)
means that the sum is over allconfigurations y'
that may be obtained from or thatyield
theconfiguration
yby
anysingle
moveI
or
( by
any one repton of the chain. The number of terms in that sum is r~ +s~,
the total number ofI
and(
arrows on
the
repton
chain inconfiguration
y. Therefore ifa~
is any solution of(3.13)
so is q~ +A,
where A is anyquantity independent
of y. Thisnon-uniqueness
of the solution of theinhomogeneous
linearequations (3.13)
is a result of thevanishing
of the determinant of coefficients of the a~. To determine the a~uniquely requires adding
a condition to(3.13),
asfollows.
The
configurations
y occur inmirror-image pairs,
y andf,
reflected in what infigure
Iwould be a horizontal
mirror,
so thatI
arrows iny become
(
arrows in
fl
and vice versa. Theconfiguration
in which allreptons
are at the same x, so that xi= x~ = = x~, is its own
image;
in the notation to beadopted
in the next section thatconfiguration
is called y = 0. Recall thatf)°)(0)
with the normalization(2.13)
is thesteady-state
distribution in thefield E. Then
by
symmetry,f)°)(0)
in the field E must be the same asf~°~(0)
in the fieldE;
so from(I.I)
and(3.ll),
a~
=ay,
ao = 0.(3.14)
With these conditions the
3'~ equations (3.13)
may be reduced to(3'~
lequations
2
for that smaller number of
a~'s.
This isagain
a set ofinhomogeneous
linearequations,
butnow with a
non-vanishing
determinant of coefficients. Theunique
solution ofthese, together
with
(3.14), yield
all3'~~
of thea~.
Equation (3.10)
with(3.9),
andequation (3.12)
with(3.13)
and(3.14),
are two altemative butequivalent
formulas for the diffusion coefficient D. Theirequivalence
is assertedby
theNemst-Einstein relation
(1.5).
In the next section we evaluate Dby
both formulas for N=
1, 2,
and 3 andverify
theirequivalence [as
well asverifying
for N=
3 that
(3.10)
isindependent
of the choice ofmarker],
and we evaluate Dby
the second formula for N=
4 and 5.
4. N =
t, 2, 3, 4,
S.Here we
illustrate,
for smallN,
theapplication
of theprinciples
and formulas of sections 1-3.N
= i.
Here there is
only
one intemal state y, call it y=
0,
and thenf)°~(0)
= 1,
by (2.13),
and ao =0, by (3.14).
There are no terms in the sum over n in(3.10).
There is nowonly
onerepton
in the chain offigure
I and it carries both anI
and a(
arrow. Then ro=so = I in
(3.5)
and(3.12),
and M= I in
(3.10).
Then from(3.5),
Vlaw
= B B(4.I)
and from either
(3.10)
or(3.12)
Dla~
w=
(4.2)
We have thus verified the
equivalence
of(3.10)
and(3.12) [I.e.,
the Nemst-Einstein relation(1.5)]
for N= I. The driftvelocity (4.I)
as a function of the dimensionless fieldeEa/kT [=
2 InB, by (I.I)]
is shown as the curve labeled N= I in
figure
3.N
= 2.
There are now
3'~~~
= 3
configurations
y. We show them with their attached arrows infigure 4,
wherethey
are numbered y=
-1, 0,
1. This follows anumbering
convention weadopt
hereafter. Theconfiguration
in which all reptons are at the same x we call y =0,
andconfigurations
that are mirrorimages by
reflection in that one,I-e-, by
reflection ina horizontal
mirror,
are numbered y and y. Such amirror-image pair
was called y andf
in the text above(3.14).
Whichever of the two
reptons
is chosen to be themarker,
the matrixA(q)
of(2.4),
atq =
0,
isI 0
A
(0)
= w
(B
+ B ~) l 2(4.3)
0
o.7 ~~' N=2
O.6
O.5 Vlaw
O.4
O.3
o. 2
O.
-I 2 3 4 5 6 7 8 9 IO
-°.' eEa/kT
Fig.
3.- Driftvelocity
V in units of aw as a function of the field E in units ofkTlea
for N= 1, 2, 3, and 4. V is an odd function of eE.
~-i
~
of~
Fig.
4. The threeconfigurations
y= 1, 0, for N
=
2.
with
eigenvalues Ao(o)
=
o,
i
(o)
= w
(B
+ Bi), ~(o)
=
3 w
(B
+ B-1) (4.4)
and
corresponding eigenvectors
1/3 Ill Ill
f1°)(0)
=
1/3j
,
f~~~(0)
=
0
,
f~~~(0)
=
2/ Qi (4.5)
1/3 1/ / 1/ Qi
Here
f1°~(0)
is normalized as in(2.13)
whilefl~)(0)
andfl~~(0)
are normalized as in(3.9).
In
figure
4 we see r_j = s_ j = rt = sj =
I,
ro= so =
2,
M=
2
(with
either choice of markerrepton),
and v= ±
[1, 0,
J, the(inconsequential)
choice ofsign depending
on which of the two reptons is chosen as marker.From
(3.5)
we then find the driftvelocity, Vlaw
=
~
(B
B~). (4.6)
This is
plotted
as the curve N=
2 in
figure
3.The
only
oddeigenvector
in(4.5)
isfl~~(0),
so in the sum in(3.10)
theonly non-vanishing
term is n
=
I. We then calculate
Dla~
w =1/3 (4.7)
Either from
(3.13), (3.14),
andfigure 4,
or from(3.ll)
and(4.5),
we see that a_j =ao = at =
0 here so from
(3.12)
we calculateDla~w
=
1/3,
inagreement
with(4.7), again verifying
the Nernst-Einstein law for this model.N = 3.
The nine
configurations
y=
4,
,
4 of the repton chain with N
=
3 are
pictured
infigure
5.The matrix
A(q)
of(2.4),
at q =0,
is(B+B~')
B 0 B~' o 0 0 0 0B~'
-2(B+B~')
B B B~' 0 o 0 00 B~' -28 0 0 B~' 0 0 0
B B~' 0
-2(B+B~')
B 0 B~' 0 0A(0)
= W 0 B 0 B~' 2(B + B~ ') B 0 B~ ' 0
0 0 B 0 B~' -2(B+B~') 0 B B~'
0 0 0 B 0 0 -28~' B 0
0 0 0 0 B B~' B~'
-2(B+B~')
B0 0 0 0 0 B 0 B~'
(B+B~')
(4.8)
With the index y as in
figure 5,
in(4.8)
it runs from 4 to 4reading
left toright
and top to bottom.The
leading eigenvalue
ofA(0),
and the associatedright eigenvector,
the latter normalizedas in
(2.13),
are(B+B~~)~
~2
~ ~~~-2 1+B~~+28~~
28~+1+B~~
Ao(0)
=
0, f~°~(0
= ~ ~ ~ ~
(B
+ B )~(4.9)
2(B
+58 +6+58~ +B~~~~
~~~_~
28~+B~+1 28~+1+B~~
(B+B~~)~
The
eight remaining eigenvalues,
evaluated at E=
0
(B
=
I