HAL Id: jpa-00246950
https://hal.archives-ouvertes.fr/jpa-00246950
Submitted on 1 Jan 1994
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 tube diameter in polymer melts, its existence. and its relation to the quantum Hall effect
Arkady Kholodenko, Thomas Vilgis
To cite this version:
Arkady Kholodenko, Thomas Vilgis. The tube diameter in polymer melts, its existence. and its relation to the quantum Hall effect. Journal de Physique I, EDP Sciences, 1994, 4 (6), pp.843-862.
�10.1051/jp1:1994232�. �jpa-00246950�
Classification Physics Abst?.acts
61.25H 46.30J 64.70D
The tube diameter in polymer melts, its existence and its
relation to the quantum Hall effect
Arkady
L. Kholodenko ('> 2) and Thomas A.Vilgis (')
(') Max-Planck-Institut fur
Polymerforschung,
P-O- Box 3148, D-55021 Mainz, Germany (2) 375 H. L. Hunier Laboratories, Ciemson University, Ciemson, SC 29634-1905, U-S-A-(Received 20 August 1993, ?.ei<ised 23 Feb?.uaiy 1994, accepted 2 Maich 1994)
Abstract. The self consistent single tube modei which
plays
themajor
rote in thereptation
theory of de Gennes, Doi and Edwards is reanaiyzed with heip of the discovered analogy with quantum Haii effect (QHE). Westudy
the existence and thestabiiity
of a given tube caused by ils interactions with other tubes. The interaction is of atopoiogicai
nature due to the effects of entangiements. The formalism developed aliows us tu explain the observed crossover from the reptation tu the Rouse regime in ternis of meiting transition of a quasi-iattice of de Gennes tubes.Some of the
major
resuits of the existing packing and scaiing theories reiating the monomer densitytu the tube radius are obtained as
iimiting
cases of a more generai theory. The resuits of Dur caiculations are supported by the most recent Monte Carlo simulations.1.
Introduction,
basic formulation and definitions.Recently
thetopological theory
ofreptation
has beenproposed [il.
Thereptation theory
wasoriginally suggested by
de Gennes[2]
andsubsequently
refinedby Doi,
Edwards[3]
andothers. From the very
begmning
of thedevelopment
of thetheory
ofreptation,
it was clear thattopology
shouldplay
an important rote in thisproblem.
In the same time, theexistmg developments (except
thatpresented
in Refs.[1, 4])
usetopology only indirectly,
e-g-by considering
theproblem
of motion of apolymer
chain in the array of obstacles, usmg the concepts of tubes andprimitive paths [2, 3]. Although
the concept of a tube tutus out to be decisive for thedevelopment
and success of the reptationtheory,
it is introduced in thistheory axiomatically.
Indeed tochange
the lime scale ofpolymer
motion from the Rouse timer
~N~ (N being
the chainlength
ordegree
ofpolymerization)
to r~N~
characteristic forreptation,
a newlength
scale had to be introduced. De Gennes used anunderlying
method witha certain mesh size whereas Doi and Edwards confined the chain into a
(harmonic)
tube ofdiameter a.
Analogous
to the tube model is theprimitive path
model[3],
which is alsoinstructive. It fixes the
primitive path
steplength
a(1.e.
the tubediameter)
to beforger
than the Kihn steplength f.
It has been difficult to find atheory
whichprovides
this newlength
scale from atopological
point of view. The aim of this paper is todevelop
a formalism and to find ananalogy
to relate the tube diameter to a moregeneral physical
picture. Inspite
of the fact thatreptation
is adynamical phenomenon,
the tube is introduced in atraditionally thermodynamic
way
(to
be discussed below in this Sect. and in Sect.2). Although tubes, apparently,
canappear and
disappear,
at time scales shorter than the terminal relaxation time[5]
r~ N ~.~
(where
N is thelength
of apolymer chain)
the melt offully
flexiblepolymers
could be viewed as a porous continuum with tubes(pores) being randomly
distributed in it[6].
To characterize the porous
material,
thepartition
coefficient X is used[7].
For thefully
flexible chains it isgiven by
X " exp
(- As/k~
m exp(-
a(R~/Rp)~) l.1)
where As is the
configurational
entropy decrease causedby
the confinement of the chain.As
depends
on the characteristic size of the poreRp,
its geometry(reflected
in the numerical coefficienta)
and the character of interactions between thepolymer
and the pore watts.R~
=Nf~
is the size of the unconstrained macromolecule and f is of the order of an effectivemonomer size.
It is important to notice that In X can also be
remterpreted
as a decrease in the entropy causedby
thermal fluctuations. Such fluctuations take the system away fromequilibrium
and,therefore, Rp
should be considered,rigorously speaking,
as a random variable. From this pointof view. the occurrence of tubes in
polymer
melts is a direct extension of the vacancy modelsof self-diffusion m
simple liquids [8].
These models are also based on theassumption
ofsurface tension-limited spontaneous vacancy formation which facilitates the diffusion of host
atoms
through
theliquid.
In the case ofsimple liquids
theLaplace-Young equation
2 «H
=
AP
(1.2)
where « is surface tension,
ù
is themean curvature while AP is the pressure difference, admits
only
onesolution,
asphere.
In the case ofpolymer
solutions(1.2)
also canprovide
acylindrical (tube-like)
solution. This could be understoodsimply by posing
aquestion
how many surfaces of genus zero can(1.2)
describe ? Asimple analysis [9]
shows that there areonly
twopossibilities
: asphere
and an infinitecylinder.
Insimple
fluids animagmary cavity
could be created
by taking
round-like molecule out from the bulk to the surface while mpolymer
solutions suchimaginary procedure
would create acylindrical-like
tube.As was
already
notedby
Casassa andTagami loi,
for the case ofstrongly repulsive
porewatts
(1.1)
is not very sensitive to the actualshape
of the pore : for pores ofspherical, cylmdrical
and slab geometry the differences in X are very small(e.g.
seeFig.
3 of Ref.loi).
Because both in a porous medium and in the
polymer
meltRp
is a random variable(restricted
to non-negativevalues), (1.1)
should be understood in somepre-averaged
sense, i-e-Rp
in(1.1)
should be understood as some representativenon-fluctuating (1.e. averaged) quantity.
The situation here is similar to that which isbeing developed
in the free volumetheories of the
liquid-glass
transition[1ii.
Similar arguments are alsobeing
used in the theories ofliquid-liquid polymorphic
transitions[12].
Unlike the abovetheories,
in the present casetopology plays
a crucial rote as we shall demonstrate below.Topology
determmes the veryexistence of the tubes and is
responsible
for the different viscoelastic behaviour of thepolymer
melts as a function of the
polymer weight (e,g.
seeFigs.
7.6, 7.7 of Ref.il 3]).
Because of the above mentionedinsensitivity
of(1.1)
to the actualshape
of the pore, it wasrecognized
along
time ago
[14]
that forR~
wRp
the entropy decrease for apolymer
chain confined in the tube could be calculated with thehelp
of asimple
harmomc oscillator model.Following
Doi and Edwards[3],
let us mtroduce the end-to-end distribution functionG(r,
r'; N which satisfies the equation « of motion»
l~ V)
+
"~
(.<~ +
y~)
G(r,
r' N=
0
(1.3)
éN
supplemented
with theboundary
condition :G(r, r'; N)j~_~
-ô(r r'). (1.4)
Here
w~
represents thestrength
of the harmonic-likepotential
which models the presence of the tube around the chain. The partition function Z for such confinedpolymer
chain can be writtenas
Z
=
ldr~
dr[
dz dz' G(i~
r[,
z z' , N ),(1.5)
where we have taken into account the fact that the « horizontal » 1 and the « vertical »
degrees
of freedom are
completely decoupled. Solving (1.3)
and using(1.5)
the free energy F can be written asF
=-kBTlnZm-TAS, (1.6)
where As has the same meaning as in
(1.1), k~
is the Boltzmann constant and T is thetemperature. The
quantity
of primary mterest in thetheory
of reptation is not F but the average(x~ +y~)
= =
a~ (1.7)
w
which
produces
the tube cross sectiona~.
Ifa~
is given, thenevidently
w is fixed as well For the purpose of this paper we introduce anarbitrary plane through
the three-dimensionalpolymer
melt. The melt and itsdensity
can be visualizedby
the number of crossmgs which are obtained when thepolymers (tubes)
cross theplane.
Let us concentrate our attention on the distribution ofjust
defined tube cross sections m thisimaginary plane.
Because of the natural cut-offgiven by
the Kuhnlength Î
we require thata~
~
Î~.
Let A be an area of the surface(plane)
and let n~ be the number of « tubes »(cross
sections) which cross thisplane.
Introducenow the surface
density p
i,iaà
=
1 (1.8)
and,
following
references[15, 16],
define theWigner-Seitz
radiusi~_~
viaarr'$-z
=
(1.9)
p
Evidently,
that 2 r~_z represents an average distance between the centres of twoneighbouring
tubes. The
filling
fraction v can now be defined as~ 2
v = =
ara~ à (Î.ÎÙ)
'w-z
Obviously, by construction,
v w 1.Because both
a~
andà ultimately depend
on thepolymer (monomer)
concentrationp m the melt we rewrite
(1.10)
asv t
ara2(p)à(p) (i.ii)
JOURNAL DE PHYSIQUE T 4 N' 6 JU~E lQ94 ~I
where both
a~
and#
are mgeneral
unknown functions of p. Weexpect, however,
thatà (p
cc p in view of(1.8). Suppose
now that there is some interaction between the «Wigner-
Seitz discs »
(cross sections)
in theplane
so that the gas of discs cannotoverlap
and may form alattice or a disordered structure. In the latter case the «
liquid
» of uniformdensity might
occur so that theinitially
well defined tube radiusa~
may loose itsmeaning.
Like m thetheory
ofliquid-gas
transitions(based
on theIsing model),
weanticipate
that, if thesolid-liquid
transition indeed occurs, it will be characterized
by
v * which isjust
someparticular
model-dependent
number, 0 w v* w1. If this number is known, then we can write v*=
ara~(p *) à(p*) (1.12)
which
implicitly
determines the criticalpolymer (monomer) density p* According
toexperiments [13],
the viscoelastic behaviour ofpolymer
solutionsdramatically changes
fromRouse-like to
reptation-like beyond
some «critical»polymer
molecularweight.
Such achange,
in view of thepicture just described,
could be associated with themelting
transition : when the tubes(cross sections)
form a quasi solid lattice, we expect to have the reptationbehaviour,
while whenthey
form aliquid,
we expect to hâve Rouse-like behaviour(no
tubesare
required
to exist for Rouse~hkeregime).
To make the above statements more
precise,
the «magnetic
»analogy presented
in section 2 ishelpful. By noticing
that the harmonic oscillatormodel,
givenby (1.3), formally
coincideswith that for an electron in a constant
magnetic field,
we translate our tubeproblem
intomagnetic language
used in thetheory
of quantum Hall effect(QHE).
This then allows us togeneralize
one tube(one electron)
model to the case of many interactmg tubes in section 3 while m section 4 we reduce theproblem
ofphase
transitions in theassembly
ofinteracting
tubes to that of
phase
transitions m one-component two-dimensional Coulombplasma (CCP).
This then enables us to
study phase
transitions in such system based on available Monte Carlo andanalytical
results. Dur resultsstrongly support
the conclusions ofpacking [17]
andscaling [18] reptation
models and are m agreement with recent Monte Carlo simulations[19].
In section 5 we discuss the obtained results andprovide
some additional arguments m favour of thetopological
origin of theviscosity exponent
3.4 observed in the reptation regime. In this section we also consider some extensions of our results topolymer
networks. In the case ofnetworks,
the presence of crosslinksplays
a similar role in our model as Coulombicinterelectron interactions in the case of
QHE.
To make the paper accessible to a broadreadership
we introduce the basic concept in bothlanguages,
i e, usmg thepolymer
notations and that of theQHE.
Therefore similar formulae will appear several times for the two cases.We nevertheless think, that this
parallel
introduction of the concepts is useful and will make the paper more readable.2. Landau
diamagnetism
andpolymers
confined into tubes.We would like to remmd the reader about some facts from the
theory
of Landaudiamagnetism
w order to make connections with the
existing
hterature.The Hamiltonian
Ù
foran electron in a
magnetic
field H=
Vx A can be written as
(h
= c = e =
1)
Ù
=
V)
+ A V~ +~~
,
(2.1)
where A is vector
potential
of the field H. In the case of constantmagnetic
fieldH we may choose
[20]
~~
Î~~~ ~~'~~
or
A
=
(- Hy, Hx, 0)
,(2.3)
2 2
1-e- the constant field H is
being perpendicular
to thexy-plane.
The Blochequation
for thedensity
matrix p(r,
r' ;p )
for the electron in themagnetic
field H can be written m a standard way as~ p =
Ùp (2.4)
éP
The above
equation
issupplemented by
the initial conditionp (r, r'
p
- 0+)
= à
(r r'), (2.5)
where
p
=
(k~ T)~'.
The combined use of
(2,1)-(2.4) produces
thefollowing equation
for p[20]
:lÎ_ v2_
H é éj~2
~éô 2 m ~ 2 mi ~
éy
Y£
~$ (~
+ Y~))
P = 0(2.6)
The
corresponding pjlymej problem
is obtainedby making
thefollowing replacements
; p # N,=
~~
,
~
# '~ and
considering only
states with the totalangular
monomentum2 m 6 8 m 6
zero. After these
replacements,
and in view of(2.5),
we obtainequation (1.3)
asrequired.
Thepartition
function Z isgiven by
z
=
dr p
(r
t r' ;
p (2.7)
Evidently,
both(1.5)
and(2.7) (after replacements)
willproduce
the same results for As w the limit N- m.
Equations (1.5)
and(2.7)
are m fact somewhat different : while the first is in accord with thegeneral
definitionsgiven
in Doi and Edward's book, the second is inaccord with reference
[3],
page1816, equation (A.6).
This could beeasily
understood if we recall that thedensity
matrix p can beexpressed
in the formP
(r, r'Î P )
=
z
e~ ~~~~Piii(r) ji~(r'), (2.8)
where w (n
)
and #i~ arerespectively
theeigenvalues
and theeigenfunctions
of the HamiltonianÙ (2.1).
Forlarge p (or N)
theground
state dominance[14]
leavesonly
the lowestn-terni
contributing
to(2.8).
In the present
study,
m view of(1.7),
we are interested in the mean cross sectional size of a tube in theplane.
Togeneralize
our results for manyinteracting
tubes, the use ofcomplex
variables is most convement.
By introducing complex
variables z= x + ty, and f
= x iy, we rewrite the Hamiltonian
Ù
givenby (2.1)
in the form[21] (for
the constant magnetic fieldcase)
Ù= ~àé+£[z[~-~(zé-Ià) (2.9)
where à
=
),
à=
~
,
[z
~ = zf, etc.z ai
The
angular
momentumoperator
J = zéfà
con beignored
if we are interestedonly
in thestates with zero
angular
momentum. For such states theground
stateeigenfunction
ofÙ
isgiven by
4ro(z, 1)
= N exP(
lz1~ (2.10)
or,
taking
into account ourreplacements (after Eq. (2.6)),
weobtam, 4io(z, i)
=
N exp
) jz2
,
(2.
ii i where N is the normalization factor.Equation (1.7)
can now be rewritten asa~
=
d~z[#fo(z,
f)]~ (z(~
m
d~zà(z, I)[z[~ (2.12)
Thus,
the existence ofa~ depends
on the convergence of themtegral
w the r-h-s- of(2.12).
In the limit ofnon-interacting
tubes thisintegral
is well defined, while if tubes(disks)
areinteractwg,
the conclusion about the convergence of such anintegral
is much more difficult to make as we willexplicitly
demonstrate below in section 4. For thebeginning,
however, we have to find the mechanism causwg the tubes to wteract.3. Statistical mechanics of
polymers
m the presence oftopological
constraints.The
study
of statistical mechanics ofpolymers
w the presence oftopological
constraints wasmitiated a
long
time agoby
Edwards[22].
Todevelop
his ideas further, it ishelpful
to recallbriefly
some of hisfindings.
The
winding
number w for a fine w aplane encirclmg
aninfinitesimally
small puncture inthis
plane
isgiven by
l
l~
~~xi,
yx(3.1)
~ 2
ar o x~ + y~
where
j
=
~Y
,
.i
=
~
and L is the
length
of the lwelywg
in theplane.
In writing the abovedr dr
equation we
had,
without loss ofgenerality,
chosen location of the puncture as the origin of coordinate system. The result(3.1)
isconveniently
rewritten w thefollowmg
way[23]
w=
j~dr)8(r(r)). (3.2)
" o ~
where
&ir(r)j
= tan~ '
fi (3.3)
and
r(r
=
(x(r ), y(r )),
The unconstrawed propagator for the
polymer lying
in theplane
is givenby
r(L)=r~
lj
LG
(rj,
r21 L)
=
D
[r(r
exp/ dré~
,(3.4)
r(0) r 0
while w the presence of a hole we can write instead,
G
(ri,
r~,
L, w
=
r(L) r~ j L ~ L
r(0>
~r
~ ~~~~~~ ~
l~ "
~~~ ~~~~~~l
~~~10
~~~~~~~
so that
G
(ri,
r~ ; L=
ldwG (ri,
r~ ;L, w) (3.6)
Equation (3.5)
can beequivalently
rewritten asG(rj,
r~ ;L)
=
7r Î~
~~~"~ ~~~_
~~ ~ ~~~~~~ ~~~Î)
~~ ~~ ~~ÎÎ /r ~~~
~~~~ ~~ ~~At the classical level, the action S in the exponent of the functional
wtegral (3.7)
givenby
S
=
j~
dré~
+£ ) 8(r(r))) (3.8)
o " ~
could be
simplified
because the second term is the total« lime » derivative
and, whence,
can bedropped.
At the quantum level we cannotsimply drop
the second term but some importantsimplifications
still can be made as we shall demonstrate below.The
generalization
to no punctures isstraightforward (see
also[23])
j L ~ "o
w =
m
drp jj
&(r(r
r,(3.9)
o
, =1
To make the
important
connection withtopology,
consideragain
theproblem
of a«
particle
»(polymer)
«moving
» in theplane
in the presence of infinitesimal hole(puncture).
The
setting
of theproblem
will remain the same if instead of a hole we would have anotherpolymer
piercingthrough
theplane.
Consider now these twopolymers
m 3-dimensional space.Following
theargument by
Delbrück[24] (see
also Ref.[l]),
we consider the ratioR~/N
N~ ~'~ For N- m, i e. for
sufficiently long polymer chains,
the distance between thepolymer
ends is much smaller than the totalpolymer length
so that the chains could be considered aseffectively
closed. For limes less than r~, defined in section 1, the twoentangled polymers
could be considered asforming
a link(1,e,
two mterlockedrings)
e,g, seefigure
1,The field
theory
which generates thelmking
numbers is known to be as the Abelian Chem-Simons
theory
which we havebriefly
discussed and used earlier w reference[1].
It isrelatively
easy to
consider,
mstead ofjust
two mterlocked « rings », say, n interconnected rings.Indeed, following
references[21, 23],
we consider the functionalintegral
of the formG
=
fi
Djr (ri )i
DjA~
eXp
j- 50 ir (T )i Sl~ iA i
SiniiA, ri j (3. ÎÙ)
where
~ ,,
Li
~°
~2f~~~
~
~~~~~~~~~~
'
~~'~~~
Fig,
l,Fig.
2.Fig. l. -Link made of two rings dissected by an
arbitrary plane.
Fig.
2. Braid-type representation of the lmkdepicted
infigure
1.and
Sj~iAj
= i
1 d~x e"~PA~ (x) ô~4~(x), (3.12)
*
n L~
S~~~IA, r
j
=
£ dT~
lj(T
~) /Î~(r(T~ ))
,
(3.13)
1=1 0
and summation over the
repeated
Greek indices is assumed from to 3.The functional
integral
over A-fields can beeasily
calculated(by completing
thesquare)
because it is Gaussian. Astraightforward
calculationproduces [1, 23]
:G
=
fl Djr(r~)j exp(- soir(r)j
-14S£ ijr(r~), r(r~)j) (3.14)
where
]
~ ~ (r(T, ) r(Tj ))~
I
[r(r~), r(r~)]
=dr~ dr~ e~~~1." (r~
~
iP(r~), (3.15)
4 "c, ,~
(r(r~) r(r~)(
and c~(c~
)
denotes thecontour(s)
of1 th ~jth) chain(s) respectively. Equation (3.15)
represents the sum oflinking
(1 #j
and selflinking
(1=
j
numbers while ~tl in equations(3.12), (3.14)
is to be determmed below. As we have
briefly
demonstrated in reference[Ii (see
also Refs.[25, 26]
and Sect.5),
the selflinking
numbers mayeffectively
contribute to therigidity
of the individual
polymer
chains while thelinking
numbers could be rewritten after somealgebra (e.g,
see p,198,
199 of Ref.[23]
and Sect. 7.2 of Ref.[55])
as followsj
L
~I
jr(r~ ), r(r~ )]
=
dr
8(r~ (r
r~(r ))
+I~. (3.16)
2 "
0
~~
It was
argued
in reference[23]
that, for closedpaths, I~
vanishes, while for openpaths
it does not contribute todynamics appreciably.
This can beeasily
understood if we consider one of therings
(or quasi-rings, ifthey
are notchemically closed)
aslying
m someplane
while the otherone as piercmg the
plane just
m two points.Then, locally,
the picture will lookexactly
thesame as described
by equations (3.1), (3.3)
in accord with reference[22].
Obviously,
the Gausslinking
number(3,15) (for
#j)
isonly
the lowest ordertopological
invariant and other
topological
mvariantsdescribing
links and knots shouldbe,
inprinciple,
considered. Because of the skein relations between the different knots
(links) [27],
we canalways
unknot the knot[link]
thusreducing
morecomplicated
knot(link)
to lesscomplicated.
It was shown in references
[28, 29]
that thequantum
Hall effectpicture
remains in most cases almost the same ifhigher
ordertopological
invariants arebeing
used instead of thelinking
numbers. Therefore we restrict ourselves to the
simplest linking
numberpicture.
In order to utilize the results
given by
equation(3.16),
consider the situationdepicted
infigure
1. Infigure
we havedepicted
anarbitrary plane
which « dissects » our link while infigure
2 we have made abraid-type projection
of the link on thearbitrary plane
sothat, according
tofigure
2, our «particles
» start «moving
» at the same « time »(1.e.
their mtemal« times » are
synchronized). Examples
of suchsynchronized polymers
are known inpolymer
physics
as directedpolymers [30].
The directedpolymers
differ from the usual flexiblepolymers [31] only by
the choice of a « gauge »[32].
For the flexiblepolymers
the contourlength
is chosen as a naturalparametrization
for thespatial position
of segmentsalong
thepolymer
chain while for the directedpolymers
one of thespatial
coordinates, e.g. z-direction,is
subject
to the additional constiaint :~j~~
= l for 0 w r w N. Such a choice of constraint isr
motivated
by
thephysics
of theproblems
which involve the use of directedpolymers.
The choice of the additional gauge condition introduced aboveusually
leads to thesynchromzation
of « lime » for different
polymer
chains as discussed in some detail m theAppendix.
We consider the
polymer
chains asasymptotically
closed, and thesynchronization
isalways possible
if wemitially require
that the individual action(1.e,
that for asingle chain)
to bereparametrization-invariant [32].
The addition ofmagnetic
fieldonly
thickens the fines thusconverting
them mto tubes. From thispoint
ofview,
the first term w the r-h-s- of(3,16)
represents a mutualwinding
number for such tubes. For suchsynchronized
« motion» of tubes
we can
write,
instead of(3,14),
thefollowing
final result which takes into accountonly
thecross sectional « motion »
(in
theplane)
:-
i,i
Dir (T)i xP1 Il dT1,f i ri
+ i
i,i i &(rJ
(T))11(3.17)
where r,~
(
r)
= r~ r r~ r).
The form of
equation (3.17)
isjust
that used in thetheory
ofQHE [21, 23],
except that we usehere the Euclidean « time » mstead of Minkowski which amounts of
considering
theQHE
atfinite temperatures. In view of the discussion
presented
in reference[23],
thisreplacement by
the Euclidean time
produces practically
nochanges
m thesubsequent development
of theformalism. The constant «
magnetic
field » causes our tubes to have a fimte thickness. This is alsobeing
used in thetheory
of theQHE. Although
the distribution function givenby (3.17) (with
added constant magneticfield),
represents a quantum mechanicalmany-body problem,
its solution is rather
simple
if andonly
if we shall use the «synchromzed
time »(directed) polymer
formalism(e.g.
seeAppendix).
To illustrate theidea, followmg
reference[33],
let usconsider a system described
by
the collection ofsmgle partiale
HamiltoniansÙ
=
£ Ù (r,, Î
m H(r,,
~(3.18)
,
ér, ér,
Topological
interactionschange Ù
intoÙ~ =
H
(r,, ) £ £
V~,
8(r,
r~~
(3.19)
~, "
, #j
so that the
Schrôdinger-like equation
can be written now as) #i((r,),
t)= Ù~
#i((r~), t) (3.20)
where, as in quantum
mechanics,
the wave function #i isrequired
to besingle
valued. This requirement makes calculation of #i nontrivial.Fortunately,
there is another way ofsolving
the abovemany-body problem. If1( jr, ), r)
is multivalued wa~e function such that1((r,), r)
=
exp(-) £ 8(r, -r~)j #i((r,), r) (3.21)
"1<j
and, m addition, we require
that,
1
=
Ù#i
,
(3.22)
then the
single
valued function #i in(3.21)
evolvesaccording
to(3.20). Thus,
use of themultivalued function
1
allows us to solve the above quantummany-body problem exactly.
With the
help
ofcomplex
variables z and f, we can rewrite1
as follows
[23, 33]
~
i
=
fl (z, z~)~
"j(z,, 2, ). (3.23)
Taking
equations(2.8), (2.10)
and(2.11)
into account we obtain thefollowing
result for theground
state wave function#io.
4ro(z~, f,
= Nfl
(z~ z~)2
« exp1' jj
(z~(~(3.24)
, ~j
~
,
In the
theory
ofQHE
the wave function m the form of(3.24)
was firstproposed by Laughlin [34]. Taking (2.12)
into account the calculation ofa~
isgiven
now asa~
=
N
d~z
z ~
#,~~
(z,
2),
(3.25)
where
ni
àint(Z, 1)
~=
1, fl
d~Zi 4~0(Z,,1,)j (3.26)
=?
Thus, the
completely
solved quantummany-body problem
leads to the classicalmany-body problem
which admits an exact solution for à~~~only
mexceptional
cases as we shallexplaw
below.
Moreover,
as we shall demonstrate, theexponent Polyakov spin
factor ~which so 2 7r
far is left undetermmed will be
specified
based on theanalogy
with two-dimensional one- componentplasma.
4. Classical statistical mechanics of one~component two dimensional Coulomb
plasma.
The
equation
of state for classical one and two-componentplasmas
are derived in refer-ence
[35] by
standardscaling arguments.
It is remarkable that bath one and two-componentplasmas
have the same equation of stategiven by
P
= p (kB T
g~/4
,
(4.1)
where p, as
before,
isgiven by
p=
NIA
(e.g.
see(1.8))
while g is themagnitude
of the« electric »
charge.
Let T~ =g~/4 k~,
then for T~ T~ the system is m a « gas »
phase
while for T ~ T~ it is in a «liquid
»phase.
In the case of a two-componentplasma,
aliquid
state consists of clusters ofdipoles, quadrupoles,
etc. formedby charges
of oppositesigns.
The transition tosuch a state is known as Kosterlitz-Thouless
(K-T) type
of transition. In the case of CCP thereare no
explicit charges
of opposite sign, except those which contribute to the uniformbackground (to
insure the overall electricalneutrality).
WhenT~T~
the pressure inequation (4.1)
becomesformally negative
and the systemundergoes
aphase
transition which is not of the K-T type so that the standard RGanalysis [36]
cannot beapplied
to the CCP. Recall that the Hamiltonian for CCP is givenby [15, 21]
H
=
g~ ~j
In[z,
z~[ +"~~~
~j [z,
[~(4.2)
,<j
2
,
The combined use of
equations (3.24)-(3.26) produces
as wella~
=
l,~j d~z,
1zi ~ exP1- ùl
,(4.3)
where
ù
is givenby
ù
=
Î 1
inlz,
z/1
+~/ i lzi
l~
(4.4)
,<j ,
and Z is the
partition
function.Comparison
betweenequations (4.2)
and(4.4) produces,
in view of(1.8),
f
'~~'~~
and
g2
=
~
(4.6)
Combining
these two equations andtaking (1.7)
into account we obtam~
=
7ra~ à
,
(4.7) g~
which
produces,
in view of(1.10),
v=
~=~" (4.8)
g ~°
Whence,
the requirement of the overallelectroneutrality permits
us to fix the value of~tl with
help
of(4.8).
Thisproduces
the final form of the Hamiltoman which we are going to useù
=
Z
inlzi
zj1 +~/ i lzi
l~
(4.9)
, <j
Notice
that, by
construction, Mm1 so that the coefficient in front of the first term is~ 4. Denote m
=
~
and consider the
partition
function Z forarbitrary
non-negative values ofv m. We obtain
z
=
là d~z
exP
(+
m1
in 1zizj1 ~/ i zi1~~ (4. io)
, i <j ,
As a side remark we mention that the
partition
function in the form givenby (4.10)
wasconsidered in the
theory
of random matrices[37],
where the exact solution was found[38, 39]
for the
special
value of m= 2, and in the
theory
ofstrings,
formulated as random matrix models[40],
forarbitrary
m. In the latter case no exact solution was found but theSchwinger- Dyson equations
which connect correlation functions of variouscomplexities
were obtained andanlysed. Alternatively,
m thetheory
ofQHE
thepartition
function Z was considered from the point of view of thetheory
ofintegral equations
forliquids [41],
while m reference[15]
the extensive Monte Carlo calculations combined with variousanalytical approximations
werediscussed. In view of such
already existing approaches leading
towards approximate solution for Z, we have not to go mto much of the formal details m this paper thusrestricting
ourselves to several illustrative butimportant examples.
Consider first the
exactly
soluble case m = 2. Eventhough
it cannot bedirectly
used for our case, smce m isarbitrary,
it isphysically
veryillummating
and willprovide
an interesting limit later on.Let
l~ni(Zl,
22, , ZJ,~, Zl, 22,., Z,,~) "
Î~J,i( (Zi)
,
(Z,)
)~ COnSt eXp
(- H) (4,
ii)
where H is given
by (4.9)
with ~= m. The
n-point
correlation function is defined asv
R~(z~,
,
z,,
,
Ii,
..,
f~)
=
~~~
ij
d~z~
P,,,(
(z~ ;(2~ ). (4,12
)~~t '~~'
, =,I +
It is convenient to rescale z' s i z
- z and 2
- l. This
rescaling
will contribute to~2
w
Î2
w
the constant term m equation
(4.11).
It is important to realize that[37]
fl Îz,
zjÎ~"
fl (zi
z~)
(f,
2~(4.131
1<j , ~j
and
d~z l~
f' zJ e~
= 7r à
~~
j (4. 14)
It can be shown
[37]
that,,1-1 ~t
à,m(z, 1)
ozRi (z, 1)
= ar
' e- ~
£ Î~ (4,15)
(-o
where #~~~ was defined in
(3.26).
In the
thermodynamic
hmit n~- m and
( ÎzÎ~~
_ ~+
lzl~
(4.16)
~~o
~!
The use of
equations (3.25), (3.26), (4,15)
and(4.16)
indicatesthat,
for m=
2,
we would havea~
- m.
Hence,
for m=
2 the CCP model is in «
liquid
» state. As m the case of normalliquids,
thedensity Ri
isuniform,
i-e- constant in thethermodynamic
limit. Thesimilarity
withthe usual
liquids
ends if we realize(based
on Monte Carlo and the approximateanlytical
data
[15])
that for small(1) degrees
offilling (1.e,
for v-
0)
the CCP could form acrystal.
Athigher filling
fractions it is m a«liquid»
state. Inreference[15]
it was shown thatcrystallisation happens
for ~~140
(or
v ~=
0.0286).
v 140
To see what
happens
to theoriginal
tube in the presence ofinteractions,
it is instructive toconsider a
simple
variational calculationusing
thethermodynamic inequality [42]
:F w
Fo
+(ù Ùo) (4.17)
Here the free energy F is defined as m
(1.6)
while for the trialFo (or ùo)
thenon-mteracting
case is chosen.
Ho=(1-A)£[z~[~,
Am0.(4.18)
It is convement to rescale z-variables both in
(4,10)
and(4.18) by mtroducing
w=
~
/
Equation (4.10)
willacquire
thefollowing
formnj
Z
= const
fl d~w
exp(-
n~à [w
,
(4.19)
i
where
à[w]
is given
by ù lwl
=
1
1W, ~) z
in1W, WI1~ ,
(4.20)
, 1~
~~
and, accordingly,
inÙo
wereplace [z,[~ by [w~[~
Using equation (4.14),
we obtain :~~
~~ ~~ ~ ~~ ~ ~~~ ~ ~ ~°~~~ l À
~°~~~
,
(4.21)
which leads to the
following
result forFo.
n~
Fo
= n~ In (1 A + const'
(4.22)
For
(Ù Ùo)
,
we obtain as well
o
(Ù Ùo)
= ~~ ~ +'~~~ In
(1
À)
+ const.(4.23)
o À 4
~
Minimizmg,
we obtain :1 [Fo
+(Ù Ùo)
= 0(4.24)
which is
equivalent
to theequation
À mn~ 4 n~ n~
À 4 v
'4
v
~~'~~~
This
produces
:nt(1
À n~=
(4.26)
nt
~ v
For n~ w we obtain :
n~(1
Àw v
~4.27)
This
produces
for the average size of the tube result~2 n
~~'~ ~ ~~~~
l À
~~
'
~~'~~~
i-e- for fixed
v and n~
- cc we would obtain
again a(~~
- m in agreement with the result
(4.16).
The above calculations servedonly
to illustrate how the interaction between tubes maydestroy
their existence. At the same time, for n~- m the
partition
function(4.19)
can becalculated with the
help
of the saddlepoint
method. In view of(4.20), by
minimizingÙ
we obtainrespectively
1<~ =
~
~j
,
(4.29)
n~ w~ w~
Ju #,>
and
w~ =
~
~j (4.30)
lli
j~ ~,
fil,
lij
To solve these
equations, multiply
equation(4.29) by
w~ while(4.30) by
fiJ~ and suppose that w~ =Rexp(2 7rik/n~),
le- we assume that solution represents«particles
»sitting
on thering [43]
of radius R in thecomplex plane.
For this case we obtain R~=
~
~j
~t
J
e~"~'~'
~ J~~ m 2
(4.31)
"
~
2 2 v 'where m the second fine we have used the sum rule given m reference
[43] (e,
g. seeAppendix
of this reference and
Eq. (5)). Thus,
the ringconfiguration always provides
an extremum forÙ[w,].
This does notimply,
that the solution which wejust
found is theonly
one. There are other stable solutions which arebriefly
described in references[43, 44].
All other solutions are obtamednumerically. Using
equation(4.31)
we can obtain an average distance r between theparticles
(1)
=
~ "~
=
~ "
j~ (4.32)
nt nt "