HAL Id: jpa-00246699
https://hal.archives-ouvertes.fr/jpa-00246699
Submitted on 1 Jan 1992
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.
Universal shape properties of open and closed polymer chains: renormalization group analysis and Monte Carlo
experiments
O. Jagodzinski, E. Eisenriegler, K. Kremer
To cite this version:
O. Jagodzinski, E. Eisenriegler, K. Kremer. Universal shape properties of open and closed polymer
chains: renormalization group analysis and Monte Carlo experiments. Journal de Physique I, EDP
Sciences, 1992, 2 (12), pp.2243-2279. �10.1051/jp1:1992279�. �jpa-00246699�
J. Phys. I France 2
(1992)
2243-2279 DECEMBER1992, PAGE 2243Oassification Physics Abstracts
36.20E 64.60A 82.20W
Universal shape properties of open and closed polymer chains:
renormalization group analysis and Monte Carlo experiments
O.
Jagodzinski(~),
E.Eisenriegler(~)
and K.Kremer(~)
(~) Fachbereich Physik, Universitit Gesamthochschule Essen, Postfach 103764, D-4300 Essen I, Germany
(~) Institut fir
Festk6rperforschung, Forschungszentrum
Jfilich, Postfach 1913, D-5170 Jfilich, Germany(Received
2 December 1991, revised 28 August 1992, accepted 2 September1992)
Abstract. We investigate the influence of excluded-volume interaction
(EV)
on the shapeof a long flexible polymer chain. Both open chains and ring polymers are considered. We study probability distributions of shape parameters which are typically ratios of characteristic lengths
(such
as the principal radii ofgyration)
of a given conformation. For a class of shape parameterssuch as the 'asphericity' Ad of a chain in d space-dimensions it is shown how mean values or higher moments of the distributions can be evaluated by field theoretic renormalization group
methods. The universality of these distributions is shown and the mean asphericity
(Ad)
is calculated withinan e = 4 d expansion.
(Ad)
is found to be much more sensitive to the EV than a frequently used asphericity-approximant which avoids the ratic-averaging. This is the first analytical confirmation of a result observed by other groups in numerical simulations.We also investigate the complete distribution of A3 and of another
(prolate
vs,oblate)
shape parameter in d = 3 by means of Monte Carlo simulations. The dependence on chain length iscarefully
investigated.
This improves the accuracy of previous estimates of universal asymptotic shape distributions. Generally the EV makes the shape more aspherical and prolate.Comparing
quantitatively the increase in(A3)
due to the EV as implied by the Monte Carlo data with that by the(appropriately extrapolated)
first order e-expansion one finds good[fair]
agreement incase of ring polymers
[open chains].
1. Introduction.
The
typical
grossshape
ofpolymer
chain conformations is notspherical
[1, 2]. Thisasphericity
seems to
play
animportant
role for flowproperties
ofpolymeric
fluids[3]-[6].
It must also betaken into account if
polymer
solutionproperties
such as the second virial coefficient are calculatedby
means ofphenomenological
'smootheddensity'
theories [7]. Other random fractalobjects
such as the clustersresulting
from kineticaggregation
orpercolation
processesshow also
non-spherical shapes [8]-[10].
2244 JOURNAL DE PHYSIQUE I N°12
Consider a
specified
conformation of Npoints
in d-dimensional space[representing
eitherthe monomer
positions
in apolymer chain,
or theoccupied
sites in apercolation cluster, etc.]
with
corresponding position
vectorsRj
"(X;,I>
X;,2>>X;,d)
) "I,.
,
N
Ii)
The
shape
of this conformation can beconveniently
characterizedby
the deigenvalues
qa of the radius ofgyration
tensor [2, 11, 12] with elementsN
Qop
=~ £ (x;,a x;,aj (x;,p x;,pj (2)
;,;=1
The usual square radius of
gyration RI
isgiven by
the trace ofQ
RI
d=
L
qo +d# (3)
«=1
and characterizes the linear overall extension or size of the conformation. To characterize the
shape
of the conformation one has to consider ratios ofeigenvalues.
Thesimplest
suchquantity
is the
'asphericity' ill,
13]~
~~ =
~~~ ~~~
did -1) j2 ~)
Note that Ad vanishes for conformations with all
eigenvalues
qo = jequal
and takes a maximum [13] value of I for a(completely collinear)
conformation where alleigenvalues
vanish exceptone.
Ad
characterizes theshape's
overall deviation fromspherical
symmetry. Anothersimple
ratio [13]
~~ ~~~ ~~
~)#)~~~
~~~ ~~ ~~~distinguishes prolate
from oblateshapes
in d= 3. Positive
(negative)
valuesbelong
toprolate (oblate) shape [with
oneeigenvalue greater (smaller)
than and the other two smaller(greater)
than
ii. 53
is bounded [13] to the interval[-~,2].
The value 2(-~)
is taken when two4 4
eigenvalues
vanish(one eigenvalue
vanishes and the other two areequal).
The
probability
distribution of the conformations(R;)
in(I) depends,
of course, on the system under consideration. Ofparticular
interest are critical systems such aslong
flexiblepolymers, percolation
clusters near thethreshold,
etc, where correlations exist over macro-scopic lengths.
These systems show the remarkablephenomenon
ofuniversality.
This means that many of theirlarge
scaleproperties [which
are then called 'universalproperties']
are inde-pendent
of most details of themicroscopic
interactions. As a consequence one may usesimple
models for a
quantitative
calculation of universalproperties.
In the present context it is natural to look for
shape properties
which are universal[13].
We are interested in linear
polymers,
both open chains(oc)
andring polymers (rp).
As we shall discuss below the mean values ofAd
and53 land
even thecorresponding probability distributions]
are universal in thiscase
provided
the number N of links(or monomers)
in the flexible chain tends toinfinity.
This means inparticular
that forpolymer
lattice models in d= 3, say, the universal shape
properties
areindependent
of the lattice structure. Here the so-calledpolymer-magnet-analogy [14,
15] isextremely
useful sinceuniversality
in criticalN°12 UNIVERSAL SHAPE PROPERTIES OF POLYMER CHAINS 2245
magnetic
systems or 'field theories'has beeninvestigated
in great detail. Similaranalogies
areuseful in case of
percolation
clusters or lattice animals [9].Polymers
with and without excluded-volume interactionbelong
to dilTerentuniversality
classes
[15], [16].
The main concern of the present paper is toquantitatively
estimate the influence of excluded volume on universalshape properties
bothanalytically by
means of the renormalization group andby
numerical simulations.We
investigate (for
the first timeanalytically)
the meanasphericity
(Ad) (6)
in presence of excluded-volume interaction. Here we use a method to average ratios such
as Ad or
53,
which has beenproposed by
Diehl andEisenriegler [17].
These authors have obtained exact results forarbitrary
d of thecorresponding quantity (Ad)o
for random walkchains. Previous
analytical
treatmentsill, 13,
18] avoided theratio-averaging
and evaluateda
quantity
j
~~~~~~
(~~~ (~~)
which
is,
of course, dilTerent from the meanasphericity (6). Contrary
to(Ad)
thequantity
Id
has no direct relation to theprobability
distribution of theasphericity
parameter Ad- It is theseshape-distributions, however,
whichprovide
the clearest information. Weinvestigate
analytically
how(Ad) begins
to deviate from(Ad)
o when the space dimension d becomesslightly
less than the
'upper
critical dimension' d= 4
[above
which the excluded-volume interaction is irrelevant[15,
16] and(Ad)
becomesequal
to(Ad)o
as N -cc].
An estimate for(A3)
is obtained
by appropriate extrapolation.
We find that(Ad)
is much more sensitive to the excluded-volume interaction than thequantity Id-
This is in line withprevious
numericalestimates
by Bishop
and Saltieljig]
andby Cannon,
Aronovitz and Goldbart [20].We also calculate
exactly
the average(53)o
of theprolate /oblate shape
parameter53
forrp's
and oc's without excluded-volume interaction. Let us mention that the methods of reference [17] and of the present paper could be used to average besides Ad and 53 also other ratios of
interest in
polymer physics.
Oneexample is,
of course, that of the squares of the end to end distance and thegyration
radius. Another one, for a chaingrafted
to aplane surface,
is thatof the squares of
appropriate gyration
tensor components,parallel
andperpendicular
to thesurface.
To
investigate
universalshape properties
we have alsoperformed
Monte Carlo(MC)
sim- ulations forpolymer
on lattice models in d= 3. This allows to compare with our
analytical (renormalization group)
estimates of(A3)
forrp's
and oc's. Besides the meanvalue,
one mayextract from MC-data the
complete probability
distribution for agiven shape
parameter[21].
These distributions are not
sharply peaked
for N - cc and show a non-trivial universal form whichdepends strongly
on the space dimension[24].
Inhigher
dimensionsid
=
4, 5)
themost
probable asphericity (Ad)mp
where the(broad)
distribution attains its maximumis nonzero and
roughly equal
to(Ad)-
In d =2, however, (A2)mp
and(A2)
can bedrastically
dilTerent.
E-g-
forrandom-walk-rp's (A2)mp
" 0 while(A2)
" 0.28[25].
In this paper we shallconcentrate on the distributions of
A3
and53.
As for open chains these distributions havebeen also addressed in reference [20] and reference [24]. For
ring polymers they have,
to ourknowledge,
not been considered before. We extendprevious
Monte Carlo workby carefully investigating
thedependence
on chainlength
N in order to arrive at more accurate numericalestimates for the universal N - cc
properties.
We close the introduction with a
guide
to the rest of the paper. Theanalytical approach
to evaluate
(Ad)
willproceed
in the usual wayby combining
theexpansion
with respect to2246 JOURNAL DE PHYSIQUE I N°12
the excluded volume
(EV)
with the renormalization group[16].
In subsection 2.I we discuss the random walk(or
ideal or Gaussian-)
chain situation which is thestarting point
of theexpansion.
Here weexplain
the method [17] ofratio-averaging,
This leads to somewhatunusual Gaussian mean values and propagators [17] which will appear
again
later asbuilding
blocks in the
perturbation expansion.
We also take here theopportunity
to calculate(53)o.
Details of this calculation are
given
inAppendix
A.The influence of EV-interaction on
(Ad)
isinvestigated
in subsection 2.2 which is divided into two parts. In the first part we discuss thegeneral
structure of the renormalization group and show that the distributions ofshape
parameters such asAd
are universal. Some details of this discussion have been deferred toAppendix
B. In the second part we determine the form of the first order in EV contribution to(Ad),
evaluate the universal result for(Ad) /(Ad)o
tofirst order in
e = 4 d
(8)
and present an estimate for
(A3).
The necessaryexplicit
numbers are extracted from the first orderexpressions
inAppendix
C. There we alsoverify
that these expressionsobey
thegeneral
renormalization structure.
In section 3 we describe our Monte Carlo simulation and present results for
distributions,
mean values and relative variances of A3 and
53
for oc's andrp's
with EV-interaction. Forcomparison
we present also theA3
distribution for oc's without EV, A summary of our results isgiven
in section 4.2.
Analytical
treatment.2.I IDEAL (RANDOM WALK) CHAIN. Consider a
simple polymer
model where the Npoints
R;
in(I)
form a harmonic chain with nearestneighbour interaction,
I-e- with a Hamiltoniand
7iRW(Rj)
"
£ hRW(X;,a) (9)
a=1
~~~~~
hRW
(~j ~2 ~ ~'
~~ ~
~ ~~~~j"2(J)
The summation over
j
starts atj
= 2 if the chain is open, and atj
= I if it is closed. In this lattercase one must
interpret
the variable Xooccuring
in thej
= I term as XN. The Gaussian model(9), (10)
is one of thesimplest
in theuniversality
class of ideal(or
randomwalk, RW)
chains where true many
body interactions,
such as theEV-interaction,
aredisregarded.
The ratios
(4)
and(5)
can be written as [13]and
53 =
27~lll)~ (12)
respectively.
Here ~j means the traceless tensorij
"
Q ~~~
ld
(13)
N°12 UNIVERSAL SHAPE PROPERTIES OF POLYMER CHAINS 2247
and the definition
(3) off
has been used. Note that numerator and denominator in(11)
and(12)
have asimple (polynomial)
form in terms of the basic variables R or X. This would bemore
complicated
for otherinteresting shape
parameters such as the ratioqm;n/qmax
of the minimal and maximaleigenvalue
ofQ.
In order to average the ratios
II), (12)
it isadvantageous
toexponentiate
the denominators[17]
by
means of theidentity
(~~Q)~~
"~~
~~j/~dV V~~~ ~~~~~~ (~~)
and to rewrite the
ensuing
average as1' (Qi
~~~~~io
-i' IQ)
io,~ l~~
~~~lo (15)
where
F(~j)
istr(~j2)
ordet~j
incase of
(11)
or(12).
Here)o
denotes theoriginal
averageover
(R)
conformations with respect to the Hamiltonian~Rw
in(9)
while )o,~ is an average with respect to a formal Hamiltonian~Rw,~
" ~Rw + vtrQ (16)
From the form
(2)
of one concludes that~Rw,~
hasagain
the form(9)
with hRwreplaced by
N ~
hRw
~
(X; )
= hRw(X; )
+ ~~
£ (Xi X;) (17)
' 2N
1,j=J
The
subscript
0 on the above averages indicates that we stilldisregard
the excluded-volume interaction. Since both ~Rw and~Rw,y
arequadratic
forms in the basic variables R orX,
the above averages can beeasily performed,
e-g-by introducing
normal coordinates. In thecontinuum limit t
-
0,
N- cc with
L =
Nt~ (18)
~~~~ ~~~ ~~~~ ~~~
(e~~ ~~~)$~~ sin~~~/2)~
~~~~~
~~
(e~
~~~)i~~
"sint(fl)1
~~~~~~~
for
ring polymers
and openchains, respectively,
wherefl
=2/j. (21)
Note that the
quantities (19), (20)
areclosely
related to the distribution functions for the square radius ofgyration
of random walk chains[26, 27].
The other average(F(4))o,~
can beexpressed
in terms of the propagatorG>,>, =
lx> xii lxj, xii )~ (22)
by applying
Wick's theorem. The continuum limitG(t,t')
= lim
Gj=i/ia,j,=t,/t2 (23)
2248 JOURNAL DE PHYSIQUE 1°12
of the propagator reads [17]
G~»~(tt~)
=
p
inl~p/~~ C°Sh fl II
'~i~")I
C°Shfl II I)I
cosh
fl ~
~)j
+cosh(fl/2) (24)
2 L
and
cosh
fl l
)j
coshfl l ~')
+osh(fl))
L L
(25)
Of course, t and t' both vary in the interval [0,
L].
This formalism has been used [17] to evaluate
(Ad)o,
the meanasphericity
of random walk chains. Here we calculate thecorresponding
mean value(53)o
of theprolate/oblate
parameter(5).
InAppendix
A we show how(det ij)o,~
can beexpressed
in terms of the propagatorG(t,t'). Using (14)
with m= 3 and
(19), (20)
one findslS3lo
=/~°dflz(fl) (26)
with
~~~~~(fl) =
~°~
p2
~°~~(~fl)
6cosh~)p)
~ j~~°Sh(fl)
lo~fl Cosh(3fl/2)
~~~~~~ ~ ~ ~+2fl~ cosh(fl/2)
+ 6fl~sinh(fl/2) 3fl cosh(fl/2)
+ 48sinh(fl/2)j (27)
and
/
~~~~~~~~
~~ cosh(4fl) -~~~(2fl)
+ 3
(~~
~°~~~~~~ ~~~~~~~~~ ~~~
~°~~~~~ ~+12fl~ sinh(fl) 3fl cosh(fl)
+ 24sinh(fl)j (28)
Numerical
integration yields
the final resultsj~~jjoc)
= 0.475,which show that the average form of open as well as closed Gaussian chains is
prolate.
2.2 CHAIN WITH EXCLUDED-VOLUME INTERACTION. To describe a
polymer
chain in agood
solvent we have to add a
repulsive
interaction V between monomers to the Hamiltonian ~Rw.N°12 UNIVERSAL SHAPE PROPERTIES OF POLYAIER CHAINS 2249
Due to
universality
there is some freedom as to the form of V. Foranalytical
calculations it is most convenient to use a continuum model with the formv
jRj
=j f~ dial'&d (R(i) R(i')j (30)
of V. The
perturbation theory
of the continuum model is well defined if theoccuring
unper-turbed chain
partition
functions containonly
'monomernumbers'larger
than somemicroscopic
cutolT
[16]. Actually
we shalladopt
dimensionalregularization
[16] instead which is more con- venient.In the
following
we shallstudy
how the mean value(Ad)
of theasphericity ill)
is influencedby
the excluded-volume interaction.Using
theidentity (14)
with m = 2 one finds(Ad
" ~/ dy
y
(tr (ij~)) (e~~
~~Q(31)
d I
~
~
Here the second mean value on the rhs has to be calculated with the usual Hamiltonian
~ = ~Rw + V
(see Eqs. (9), (10), (30))
while the first mean value involves~~
=~Rw,~
+ V(see Eqs. (16), (17)).
2.2,I Renormalization and
universality. Straightforward perturbation theory
in u is useless forlong
chains. The reason is that inspatial
dimensions d < 4increasing
u powers getmultiplied by increasing positive
powers of the chainlength
L. One may circumvent thisproblem by
means of the renormalization group(RG),
which maps along
onto a short chain.How to set up the RG for the above-mentioned continuum model is well known
[16].
Sincethe statistics of
polymer
conformations isclosely
related [14] to a(Landau-Ginzburg type)
fieldtheory,
one may use field theoretic renormalization[28, 29].
Here one locates the ultraviolet(UV) divergencies
of thetheory (which
appear, within dimensionalregularization,
aspole
terms in e= 4
d)
and absorbs them into areparametrization.
We discuss inAppendix
B how thereparametrization
for the two mean values on the rhs ofequation (31)
can be obtained. The result is(tr (~j~))
=~~~
FjL~,u~,
(32)
~ p
and
(e~~
~~Q) = F2L~,
u~, (33)
~ Here
L =
~~~ (Zt) L~ (34)
and
u =
16x~f(e)~~Zu u~ (35)
with ~ an
arbitrary
inverselength
scale. We will notspecify
thee-dependence
off(e)
= I +Eli
+ e~f2
+(36)
beyond f(0)
= 1, since it must
drop
out in(a systematic e-expansion of)
universalquantities.
The Fi in
equations (32)
and(33)
are dimensionless functions and arefinite
for e-
0,
if theirarguments
arekept
fixed. Alle-poles
are contained in the factorsZg = I +
e~~Zj~) (u~)
+e~~Zj~) (u~)
+(37)
g # t, U
2250 JOURNAL DE PHYSIQUE I N°12
ofthe
reparametrization.
These are indentical with thecorresponding
factors ofa n-component#~ field
theory
as n - 0. For later use we note theexplicit expressions
Zt = I +
(
+ O
(u~)
~Zu = +
~
+ O
(u~)
~(38)
The arbitrariness of p leads to renormalization flow
properties
of theFj,2.
This is shown inAppendix
B and readsFi
L~,u~, Fill~(I),fi~(A),
~
~
e~~'~ (39)
~
(e-Ap)
with a free
parameter
I. HereEj = -4
,
E2 " 0
(40)
and fi~
,
£~ are determined
by
(~iR(>)
=-fl(a~(>))
,
fi~(o)
=uR (41)
and
£~(l)
~=
L~
exp/
dA'1l(fi~(1')) (42)
o
The functions
fl
and 1l are related to the factorsZg
inequation (37),
seeAppendix
B. Fore > 0 the function
fl
=
fl(u~)
has anon-vanishing
zerou~
=
(~
= e + O
(e~)
,
(43)
which is an attractive fixed
point
forpositive
fi~as I - cc.
Equation (39)
can be used to map along
chainL~
- cc onto a short chain
£~
= l. This
requires
- ccsince1l((R)
ispositive. Quantitatively,
increases withLR (in
theasymptotic
limitL~
-
cc)
ase~ -
[D (u~) L~)~
,
(44)
where
v = =
(I
+)j
+ O(e~) (45)
"( "~)
and D is
given
inequation (I II).
Thus in this limitFi
L~,u~, -~bi
e~~e~~'~ (46)
P ~
with I from
equation (44)
and~b;(Y)
= l§(1,(~, Y) (47)
We may now obtain the renormalization behaviour of the mean
asphericity (Ad).
Fromequations (31),(32),(33)
it follows that(Ad)
") /
dzz
Fi (L~, u~, z)
F2(L~, u~, z)
=: a(L~, u~) (48)
~
N°12 UNIVERSAL SHAPE PROPERTIES OF POLYMER CHAINS 2251
is a function of
L~, u~,
which remains finite as e- 0.
Equations (40), (44), (46)
and(47)
tellus that in the
asymptotic
limit L~- cc
lAd)
= aI, uR
,
u~
> 0(49)
~
Equation (49)
shows theuniversality
of(Ad).
Its value forlong enough
chains isindependent
of thestrength
u~ > 0(I.e.
of u >0)
of the excluded-volume interaction. It does of coursedepend
on whether we consider an open
polymer
chain or aring polymer (For given
arguments the functionsFi,~bi,
a have dilTerent values in the twocases).
We show below that these universal(Ad)-values
aredijfeTent
from thecorresponding
values(Ad)o
#ail, 0) (50)
for random walk type chains which have been
given by
Diehl andEisenriegler [17].
InAppendix
B it is also discussed how the two mean values in the
integrand
of(31)
can in theasymptotic
limit be
parametrized
in a universal way.(53)
and the mean values((Ad)~), ((53)~)
ofpositive integer
powers of Ad andSd
are universal in the same sense as(Ad)-
Each of these mean values has arepresentation
withessentially
[30] the form of the rhs of(31). By repeating
theprevious arguments
one realizes that theensuing
flow factorse~,
seeequation (44), always drop
out from they-integral-
As aconsequence the distributions such as
(6(A Ad)
+P(A) (51)
are universal functions in the limit of
long
chains.2.2.2 Perturbation
theory
ande-expansion
results. In this section we presentexplicit
es-timates for the universal mean
asphericity (49)
of aring polymer
and an openchain,
in the swollen state. We shallinvestigate
how thea(I, (Rl's
start to deviate from their random walkcounterparts
a(1,0)
as the dimension d= 4 e of the
embedding
space becomesslightly
less than4,
and try toextrapolate~towards
d = 3.Actually
weperform
asystematic e-expansion
to first order of the ratio
~~~'"~~
a(1, 0)
According
toequations (32)-(38)
thisrequires expressions
for the two mean values on the rhs of(31)
to orderu°
andvi,
which we denoteby subscripts
0 andI, respectively.
Zero orderexpressions
for(e~~
~~Q) havealready
beengiven
inequations (19), (20).
For the second mean value one finds from(2)
and(22)
ltr(ij~))
" ~~ ~((~
~~/~dti dt[dt2dt[ (2j (ii, t[,
t2,t[)
~(52)
o,~ o
Here
~ (il>i~>12>i~) (~1J42)01 ~(il>12)
+G(i~>i~) ~(il>i~) ~(i~>12) (53)
and we have introduced the notation
A; +
x~ji;) x~(ij)
,
i m 1,2
(54)
To arrive at
(53)
we have used that for the Gaussian(0,y)-statistics
Wick's theorem holds and the dilTerent Cartesian components Xa areindependent
andequivalent. Inserting
thepropagators
(24)
and(25)
onefinds, respectively,
forring polymers
(tr Q~)))[~
=L~
)[))(j)) 2
+(
+
( sinh(fl)
2
Cosh(fl)j (55)
2252 JOURNAL DE PIIYSIQUE I N°12
and for open chains
with
fl
from(21).
Of course, theseexpressions appeared already
in the calculation ofa(1, 0) J71.
To evaluate the u~-contributions
(e~~~~Q)~
=(e~~~~Q)~ ((V) (V) (57)
o o,~
an~
jtr (Q2)j
~~
=
ltr (°~)
vlo,~
+l~~ ~°~~10~ 1~1°Y
~~~~we rewrite
(30)
asv
jRj
=
"
/~didi, / ~ik jR~t)-R(i,)j
6 ~~g~
o k '
with
~x~d
~~~~~
The
remaining
averages with the Gaussian Hamiltonian~Rw,~
all follow from Wick's theorem andsimple
combinatorics. The average(V)o,~,
which werequire
for(57),
isparticularly
easyu
~~~~~, / ~-k2sy(t,t')
=
? f~didi' (s~ji,i')j
~~~~~(61)
~~
o,~ 6o k
(2vfi0~~
°Here
S~(t, t')
=j(A~)
=jG(t, t) G(t, t')
+jG(t', t') (62)
o,~
with a notation
A +
Xd(t) Xd(t') (63)
similar to
(54).
The averageIV)
o in(57)
follows from(61) by putting
y = 0. For thequantity (58)
one findsl~~
~~~~i,y
~~4~~~~ /~~ /~
~ ~~~~~~~~~~~~~~
~~~~~~~~~0& ~~2~~0a~~~~2~0w
(k~)~ (lAiAlo,vj
~
(IA~A)o,yj
~
l
(d I)(d
+2)
u(2~6od
4L4 6 ~x
/dS (Sy(t,t')j ~~~~~2j(tj,t[;t,t')2j(t2,t[;t,t')2j(tj,t[;t2,t[)-
js~(i, i,)j
~~+~/~in (ii, ii
i,i,)j
~in (i~, ii
i,i,)j
~164)
with the notation
/dS
+/
~ dtdt'dti dt[ dt2 dt[ (65)
o
N°12 UNIVERSAL SHAPE PROPERTIES OF POLYMER CIiAINS 2253
These
expressions
allow to calculate theuR-expansions
Fi
L~,u~, §,e
=
£ (u~)~F;jlL~, §,e (66)
~ ;-oi ~
(i
= 1,2)
of the rhs's of(32)
and(33)
up to andincluding j
= I. Here we haveexplicitly
noted thee-dependence
of the Fi's which has beensuppressed
in subsection 2.2.I. The first order inu
expressions (e~V~~Q)i
and(tr (~j~))i,y
contain([or
u, Lfixed) e-poles
at e = 0. The reason for thesepoles
is thevanishing
Sy(t,t')
- It
t'[
,
t - t'
(67)
of
Sy
as tapproaches
t'. This leads tosingularities
in theintegrands
of(61)
and(64)
whichare
non-integrable
for e= 0. In
Appendix
C we calculate theleading
e - 0 contributionsc~ ue~~ and ue° to
(e~V~~Q)i, (tr (~j~))j,y.
Weverify
that thepoles
are indeed absorbedby
the
reparametrizations,
I-e- that the functionsFi;i(L~,z,e)
contain noe-poles,
and presentexplicit expressions
forFi;i(L~,
z,0).
These enter the ratio~j)j,(jj~
= i +uR
b + o(uRe, (uR)~j (68)
with
~
b =
~
/
dzz(Fj.j(1,
~,
0)F2 o(I,
z,0)
+Fio(I,z,0)F2.i(I,
z,0)) (69)
a(1>°>°
3o ' ' ' '
We show in
Appendix
C that is indeedindependent
of the form of the non-universal factorf
inequation (36)
and present inequations (166), (167)
and(169)
numerical results for thetwo contributions to the
integral
in(69).
For the final evaluation of b one needs the random walk results~~~'~'~~
"
~~~~° .~~~
)$(
~~~~for
ring polymers (rp)
and open chains(oc) [17].
This leads toSurprisingly
we find within the accuracy shownnearly
the same result for bothring polymers
and open chains.To estimate
(A3)
+ a(I, (R, I)
we useequation (68)
with uR=
(R
andapproximate
its rhsby
the first ordere-expansion.
Thenequations (43)
and(71)
lead toTo arrive at the final numbers we have
again
used random walk results [17]~~ o
1°.246
0.394(rP) (oc) (~3)
Equation (72)
represents our renormalization group estimate for theasymptotic
mean as-phericities
forring polymers (rp)
and open chains(oc),
in the swollen state and in three2254 JOURNAL DE PHYSIQUE I N°12
dimensions. The
prefactor
I +b)implies,
for both types of
chains,
an increase of 5~it
in8 2
the mean
asphericity
due to the excluded-volume interaction.We note that the
quantity ld,
where numerator and denominator in(4)
areaveraged
inde-pendently,
is much less sensitive to the excluded-volume interaction. Within thee-expansion
Aronovitz and Nelson [13] estimated [31] an increase of
l~~
ofonly
I~it
Monte Carlo simu-2 lations in d
= 3 which will be discussed in the next section arrive at the same
qualitative
conclusions.
3. Monte Carlo calculations.
In the
analytical
part we useda continuum
model,
both in the monomer index space and in theembedding
space. For Monte Carlo simulations of course we have to confine ourselves to discretemodels,
at leastconcerning
the monomer index. In this work we will also discretize theembedding
space, I-e- we aregoing
toinvestigate
theshape properties
of latticepolymers
[32]3.I OPEN CHAINS.
3.I.I Model and method. As we are
only
interested in the case d=
3,
thepolymer
con-formations are modelled as N-step walks on a
simple
cubic lattice(lattice spacing
=I).
We denote one end of the monomer Iby
the lattice vector Ri with I < I < N. Theonly
constraint for RWS is that all successive vectors in the set(Ro,
,
RN)
denoteneighbouring
latticepoints.
The excluded-volume condition iseasily implemented by
thesimple requirement
thatno lattice vector R may appear twice in one conformation
(Ro,
,
RN
).
To go from one
single configuration
to properensembles,
we need an efficientalgorithm
whichscans the space of all
possible
conformations with respect toergodicity.
As we are concernedwith
strongly fluctuating
matter, we have to generate at firstlarge samples
of chains of fixedlength,
in a second step the chainlength
itself has to be varied in order toextrapolate
to theasymptotic
limit. Whereas theprocedure
to generate RW lattice chains oflength
N isobviously
clear(make
N steps at randomon the above
lattice),
the choice of thealgorithm
for SAWS needs more care. For our purpose a static method like dimerization [33] seems to be best suited. This method, which respects all above
requirements,
enables us to generatelarge samples
ofstatistically independent
chains(100000 / length)
for fixedlengths
N with 30 < N < 220. This set of data allows a properscaling analysis
for allproperties
inquestion.
The
method,
which is a modification of theorginal
dimerization method of Alexandrowics [33],proceeds
as follows [34]. For a SAW chain oflength
N we have to generate at first a SAW oflength N/2 by simple sampling,
I-e- generate a non-reversal random walk(NRRW)
oflength
N/2,
start at thebeginning
when the chain intersects itself. In a second step another NRRWis started at the end of the first one. If the second half intersects itself
(and perhaps
also the firstone),
one has to startagain
at the end of the firsthalf,
the first part itselfremaining unchanged. Only
in the case that there aremerely
intersections between the first and the secondhalf, being
SAWS oflength N/2
forthemselves,
both parts are cancelled and anew run has to be started. Thismethod,
which iscompletely equivalent
to theoriginal
dimerizationmethod,
has some technicaladvantages
to the latter one,sharing
all their otheradvantages,
e-g-
high efficiency compared
tosimple sampling.
Forlarger
chains this method can be usedrecursively
on several levels[34].
As the excluded-volume condition omits allintersecting
N°12 UNIVERSAL SHAPE PROPERTIES OF POLYMER CHAINS 2255
1.50 '
P(A3) 1.00
0.50
0.00 0.20 0.40 0.60 0.80 1.00
A3
Fig.
I. Probability distribution for the aspheficity A3 for open RWS(loo
coo chains with N=
1024);
(A3)f~~, (d3)f~~
denote exact mean values.configurations
from thesampling,
allpossible
conformations haveequal probability leading
tosimple
arithmetic averages for allproperties.
In the first part of our calculations we will have a short look on the
shape properties
of open RW chains. The main part,however,
is devoted to theproperties
of open SAWchains,
where the size as well as theshape
will beinvestigated.
3.1.2 Results and
scaling analysis.
Besides meanvalues,
which will now be denotedby
((A3))or
((53)),
we alsogenerated
ahistogram
for theasphericity
of RW chains(see Fig. I).
The main purpose for this calculation(100
000 RWS with N=
1024)
is thepossibility
toinvestigate
the influence of the excluded-volume interaction on the
shape properties
notonly
via average parameters, but also via distribution functions.Using
this as a basis ofcomparison
we treated theproblem
of SAW chainsby
means of the above dimerization method. All averagequantities
were calculated for chainlengths
between N = 30 and N = 220. For each N 100 000statistically independent
chains weresampled (see
Tab.
I) giving
an error bar for((R~))
of about 0.3it. In addition we calculatedhistograms
forthe two
shape
parameters A3 and53
from the data for fixedlength
N = 140.Mainly
as a check for our computer program we firstinvestigated
thescaling
behaviour of the end to end distance((R~))
and the radius ofgyration ((RI)).
A doublelogarithmic plot
of thesequantities
versus chainlength (see Fig. 2)
leads to well-known results for the correlationlength
exponentvR2 = 0.590 + 0.005
vRj
= 0.592 + 0.008(74)
As detailed corrections to
scaling
have not been taken into account[34],
the results are consis- tent both with other numerical andanalytical
work.We also
analyzed
the behaviour of the universal ratio((R~))/((Rl)),
whichis,
due tolarge
fluctuations in the
samples,
a rather delicatequantity.
Ascaling plot
of((R~))/((R[))
versusthe
leading
finite size correction N~~IA
= 0.47 in d
= 3
[35],
see alsoFig. 3)
shows that2256 JOURNAL DE PHYSIQUE I N°12
jj~~~j~«)
iiRiii~"~
20 40 80 160 300
N
Fig.2.
Double logarithmic plot of ((R~))1°~> and ((R[))1°~> vs. N.6.
o o
o
jjR2jj<oc> °
iiRiii~"~ ~
0.00 o-lo o.20
11-A
Fig.3.
scaling plot of((R~))1°~)/((R[))(°~)
vs. N~~universal
scaling
behaviour starts at chainlength
about 80. Below N= 80 lattice
dependent
effects dominate. The
extrapolation
to infinite chainlength yields, assuming
the behaviour as indicated in thefigure,
jj
j~2))(oc)jj j~2
jj(oc)
~'~~~ ~ ~'~~~ '(~~)
G
which should
only
becompared
with other simulation data. Shanes and Nickel[36],
who combined Monte Carlo data and exact enumerationresults, got
theasymptotic
ratio((R~))(°~)/((R[))(°~)
= 6.2487 +0.0013,
which is in excellent agreement to our data.Based on
this,
we think that our data are alsoreliable,
whennon-diagonal
elements ofQ
are considered. We calculated the mean values of A3 and 53 as well as ofAl
andS(
for avariety
of chain
lengths.
These averages also contain information about the width of the distributionN°12 UNIVERSAL SHAPE PROPERTIES OF POLYMER CHAINS 2257
Table I. Simulation data
for
open SAW chains in d= 3
(100
000 chains/ length).
N llR~ll~°~~
llRlll~°~~
llA3ll~°~~llAlll~°~~
llS3ll~°~~llslll~°~~
9.86 0.4404 0.2305 0.5570 0.4905
40 88.29 13.92 6.341 0A410 0.2311 0.5586 0.4928
50 lls.34 18,19 6.339 0.4395 0.2297 0.5556 0.4887
60 143.65 22.64 6.345 0A401 0.2301 0.5569 0.4893
70 173.09 27.31 6.339 0.4398 0.2300 0.5566 0.4893
80 202.91 32.00 6.342 0.4387 0.2291 0.5543 0.4868
90 232.50 36.78 6.322 0.4377 0.2282 0.5526 0.4847
100 265.15 41.87 6.333 0.4388 0.2294 0.5550 0.4887
l10 295.98 46.83 6.321 0.4374 0.2282 0.5520 0A851
120 327.56 51.85 6.317 0.4372 0.2278 0.5518 0A831
130 360.95 57.16 6.315 0.4373 0.2279 0.5519 0.4842
140 393.51 62.44 6.302 0.4368 0.2276 0.5508 0A831
160 462.97 73.36 6.311 0.4372 0.2277 0.5515 0A831
180 533.58 84.48 6.316 0.4373 0.2280 0.5523 0.4843
200 602.37 95.62 6.300 0.4356 0.2264 0.5484 0.4790
220 674.57 106.99 6.305 0.4363 0.2269 0.5506 0.4805
function ofA3 and
53, given by
the relative variancesjj
~2))(oc)
jj~~))(oc)j~ fi~~~~~
~
(~~)
~~
llA3ll(°C)
~~~~
~/jjs())(OC)
(jjS3))(OC)j~~S3
ii s~))(OC) ~~~~Scaling plots
similar to that infigure
3yield following asymptotic
limits(see Figs. 4-7)
((A3))~°~~ = 0.431 + 0.002
(78)
A
~)~
= 0.442 + 0.004(79)
((53))~°~~ = 0.541 + 0.004
(80)
At~
= 0.773 ~ 0.005
(81)
The value of ((A3))~°~~ fits very well to the result of
Bishop
and Saltiel[19],
who used apivot algorithm
for two chainlengths (N
=201,401). They
find 0.431+ 0.002 for both N whichseems not to follow
properly
the correction toscaling
curve(of Fig. 4).
Other simulationsby
Cannon et al. [20]
give
0.447 + 0.011 and 0.572 + 0.025 for the mean values of A3 and53, respectively.
These calculations were confined to onesingle
chainlength
N = 100 and no finite chainlength analysis
w,asperformed.
Their(rather large)
error bar is due to statistical scatter of the data and cannot account forsystematic
elTects. As these authors used the same latticetype as we
did,
their result candirectly
becompared
to the data offigures
4 and 6. As can2258 JOURNAL DE PHYSIQUE I N°12
*
0.
o
~
iiA311~"~ °
o o
0.
o.00 0.10 o.20
/f-~
Fig.4.
Scaling plot of ((A3))1°~~ vs. N~~(statistical
error is of the order of the scatter of thedata);
(o) from reference [19],
(*)
from reference [20].~(oc) o
o o
o.oo o.10 o.20
/f-A
Fig.5.
Scaling plot ofa(]~
=
fin
((A(11 ((A3))~vs. N~~
be seen from our
scaling plots (see Figs. 4, 6),
thisextrapolation brings
up corrections which cannot beneglected
in this case.The value of
A~)~
indicates thelarge
width of the distributionfunction,
as the variance is about half the value of the first moment itself.From the
positive
mean value of53
we see that open SAWS are on averageprolate objects.
The relative variance of
53
of about 75~ indicatesagain
a very broad distributionfunction,
which will now be examined.We calculated
histograms
for A3 and53
from asample
of100000 chains of fixedlength
N = 140
(see Figs.
8,9).
The distribution function of A3 for SAWS is even broadercompared
to the
non-interacting
case(see Fig. I).
In addition the distribution is nowquite symmetric,
most
probable
value and mean valuebeing nearly equal.
N°12 UNIVERSAL SHAPE PROPERTIES OF POLYMER CHAINS 2259
o
~ o
(lS3))~°~~
~
~
o.oo o-lo o.20
j/-A
Fig.6.
Scaling plot of ((53))~°~~ vs. N~~o.780
0.770 o
o ~
~(°C) ~
o o
°
0.760 O
o.oo o-lo 0.20
j/-~
Fig.7.
Scaling plot ofa(]~
=
@fi ((S())
((53))~ vs. N~~The distribution function for
53, however,
is ratherunsymmetric
with a mostprobable
valueof zero. The gross
shape
of thehistogram signals
that the vastmajority
ofconfigurations
hasprolate
asymmetry. It should be noted that the maximum of the distribution function at zero does notimply
that theseconfigurations
havespherical
symmetry.3.2 RING POLYMERS.
3.2.I Model and method. In
analogy
to open chains we modelring polymers
asN-step
walks on a
simple
cubic latticeadding
the obviousrequirement
that Roequals
RN-By
thatmeans the ensemble of all