HAL Id: jpa-00209521
https://hal.archives-ouvertes.fr/jpa-00209521
Submitted on 1 Jan 1982
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.
Collapse of a polymer : evidence for tricritical behaviour in two dimensions
A. Baumgärtner
To cite this version:
A. Baumgärtner. Collapse of a polymer : evidence for tricritical behaviour in two dimensions. Journal
de Physique, 1982, 43 (9), pp.1407-1411. �10.1051/jphys:019820043090140700�. �jpa-00209521�
Collapse of a polymer : evidence for tricritical behaviour in two dimensions
A. Baumgärtner (*)
IBM Research Laboratory, San Jose, California 95193, U.S.A.
(Reçu le ler
mars1982, révisé le 9 mai, accepté le 3 mai 1982)
Résumé.
2014Le paramètre d’ordre et la chaleur spécifique des polymères
sur unréseau carré sont calculés par
une
méthode de Monte Carlo et analysés
entermes de loi d’échelle de taille finie. Près de la transition
vers unétat
globulaire, ils présentent les caractéristiques d’un point tricritique. Les exposants critiques correspondants 03B1t, 03C5t et 03A6t, déduits de cette analyse sont
enbon accord
avecles prévisions obtenues d’après le développement
ensérie
de
03B5.Des corrections du 1er ordre sont calculées pour le comportement asymptotique.
Abstract.
2014Finite-size scaling of Monte Carlo generated order parameter (intrinsic density of monomers) and specific heat of square lattice polymers
nearthe collapse transition exhibit characteristics of
atricritical point.
The estimated corresponding critical exponents 03B1t, 03BDt and 03A6t
arein good agreement with predictions from 03B5-expan- sions. First-order corrections to the asymptotic behaviour
arecalculated.
Classification
Physics Abstracts
61.40K - 64.70
-68.60
1. Introduction.
-A long-standing problem in the theory of polymer chains in solution is the effect of intramolecular forces on the shape and size of an
isolated chain. The intramolecular forces are usually
assumed to be of van der Waals’ type, consisting of strong, short-range repulsive and weak, long-range
attractive interactions. By suitable changes in tempe-
rature or in solvent composition, the chain crosses over
from an extended coil to a collapsed dense globule. This phenomena has attracted a great deal of attention, both
theoretical [1-20] and experimental [21-25] possibly
because of its connection with protein folding [26, 27],
and it is also of theoretical interest as a first step towards the understanding of micellar structures.
Our present understanding of the collapse transition
within the context of critical phenomena is still incomplete and controversial. Both analytical [1-14]
and numerical attempts [15-20] have been made ‘in
order to elucidate this problem.
Recently de Gennes proposed [4] that the collapse
transition should exhibit characteristics of a tricritical
point [28]. Although several works [5,.6, 8,13,14] have
been devoted to elaborate the theory, no evidence for tricriticality of the collapse has been demonstrated so
far, neither by experiment nor by computer simulation.
In order to verify the theory, one is confronted with
an important difficulty : at the marginal dimensiona-
lity d
=3, the critical properties of the transition are
governed by logarithmic singularities, which are hard
to analyse conclusively. A detailed Monte Carlo
investigation for d
=3 will be published [29].
In order to avoid this difficulty and to extend our understanding of the collapse transition to two-dimen- sional systems, the present work reports on a finite- size scaling analysis of Monte Carlo generated thermo- dynamic functions of square-lattice polymers.
Experimentally the two-dimensional collapse can
be realized if the polymer is strongly adsorbed on a flat
surface [30] or if the solution is confined in a slit.
2. Model and simulation technique.
-The chain is represented by a random walk (of N steps), which at
each step moves from one point of the square lattice
to any of the four nearest-neighbour sites. The effect of the excluded volume is represented by preventing the
walk from visiting any lattice point more than once.
The length of one step will be called
a.Following
Orr [31], the mutual interactions of polymer and
solvent can be included in the above model by intro- ducing
aBoltzmann factor, exp(- slkb T), for every
nearest-neighbour contact of the chain with itself. The energy 8 measures the energy of
apolymer-polymer
(*) Permanent Address : Institut fiir Festkorperforschung
der Kernforschungsanlage Jflich, Postfach 1913, D-5170 Jflich, West Germany.
Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:019820043090140700
1408
contact relative to the energies of polymer-solvent and
solvent-solvent contacts. Thus,
80 describes
apolymer in
apoor solvent, which means that polymer- polymer contacts are favoured, leading to
acollapse
in the chain dimensions at low temperatures.
The total interaction energy, E(X)
=El (X) + E2(X), for
achain configuration X is thus given by E I (X)
=s tj (X), where q(X) is the number of nearest-
neighbour corltacts, and E2(X) = oo, if any lattice site
is occupied twice. An ensemble of configurations dis-
tributed according to the canonical distribution
,P(X) oc exp( - E(X)/kB T) was generated by
a «rep- tation » Monte Carlo dynamics [32]. Starting from
astate X, there one first selects one of the ends of the chain at random and then removes this end link of the chain and adds it at the other end, specifying ran- domly the orientation of the link. The resulting state
X’ is accepted as a new configuration, if according to
the Metropolis criterion [33]
where ( is a random number 0 C 1; otherwise
state X’ is rejected and the old configuration X is
counted once
morein the averaging. This mechanism,
which corresponds to a movement of the chain along itself, produces an approach towards equilibrium.
Denoting ensemble averages by angular brackets,
energy U, specific heat C and intrinsic density p are then obtained by
where
is the mean-square radius of gyration, and rk are the lattice sites occupied by the chain. The unperturbed (e
=0) chain dimension for N >> 1 is given by [34]
3. Numerical results.
-The temperature variation of intrinsic density, internal energy and specific heat is
shown in figure 1 for the entire range of chains studied.
Identifying kB T,(N)IB with the positions of the spe- cific-heat maxima, we examine the size dependence of
the transition temperature T, in figure 2. The asymp- totic behaviour seems described by
where kB TB(oo)/s
=1.31 ± 0.06 and 0, the crossover
Fig. 1.
-Raw data of monomer-density p (upper part),
internal energy U (middle part) and specific heat C (lower part) plotted
versustemperature for chain lengths N
=20, 40, 80, 160. The suggested asymptotic (N - oo ) behaviour
is indicated qualitatively for p and U.
Fig. 2.
-Variation of the critical temperature T with chain length N.
exponent described below. It should be noted that the
reliability of the present Monte Carlo calculations at low temperatures is supported by that fact, that our
estimates of internal energies seem to coincide (by extrapolating the data to T - 0) with the ground-
state energies, which are indicated by short horizontal
lines in the middle part of figure 1. Obviously the ground-state for an infinite chain is highly degenerated,
but for finite N, we found that the
«spiral » configu-
ration has the highest number of contacts il :
where I
4. Finite-size scaling analysis.
-The expected ther- modynamic behaviour of finite systems of interacting particles has been discussed by Fisher [35] in terms of a scaling theory involving the critical exponents of the corresponding infinite system. An assumption of generalized scaling at
atricritical point [36, 37] for an
isolated polymer [4] then leads to a density of the form
where vt and Ot are the correlation length exponent at the transition temperature T, (-r =- T - Tt I/Tt) and
the crossover exponent, respectively. For the consi-
dered case d
=2, approximate values of vt and Ot
have been calculated using s-expansion techniques [6, 38], carried to order S2 and a, respectively :
The scaling function /(x) should asymptotically reproduce for large
x(i.e., T 1, but N - 00) the
infinite-chain critical behaviour [4]
where
vand vc are the well-known correlation length exponents in the coil and in the globule state, respecti- vely :
Thus one obtains finally the asymptotic behaviour :
where B t, B I are the critical (nonuniversal) ampli-
tudes for an infmite chain. A similar scaling ansatz
for the specific heat is given by
where the tricritical exponent a, has been calculated
[38] to first order in
c =3 - d
The scaling function g(x) asymptotically yields the
infinite-chain critical behaviour, g(x) -+ A t x-11,
Using the data presented in figure 1, we can now test the finite-size scaling relations equations (4)-(11).
In figure 3 we examine the chain length dependence
of p for various temperatures. Well above and below the critical temperature kB T,/e -- 1.31, the data are
in good agreement with p - fi (Eq. (8a)) and
p
=const (Eq. (8c)), respectively. In order to distin- guish properly between the behaviour just at the critical point p ’" N - 0.011 (Eq. (8b») and in the globule phase
p
=const, it would be necessary to extend our investi-
gations to much larger values of N, which is reasonably unapproachable with the present generation of compu- ters.
Fig. 3.
-Variation of the monomer-density p with chain
length N for various temperatures.
However, we adopt the given value (5a) of v, - 0.505 5
in order to examine the scaling form of p (Eq. (4)) ; this
shown in figure 4. In good agreement with the scaling
ansatz, the scaled data all lie on two smooth curves, for T > Tt and T T t, respectively. The two solid
lines correspond to the asymptotic forms given by equations (6)-(8) with B + =- 2.4, B - - 1.2 and the
predicted values., p + -- - 0. 7 5, p - - 0.031. However,
Fig. 4.
-Finite-size scaling plot of the
monomerdensity according to equation (4) for various N at temperatures
T
>7t (lower part) and T T, (upper part).
1410
from the (N It i)
-region covered by the Monte Carlo
data, it is rather suggestive to infer an asymptotic approach of the data to their solid lines. The reason
behind this slow convergence is that corrections to the
asymptotic form (Eq. (6)) for x 10 are important.
This is shown in figure 5. Now our data are well-
described for
x> 1 by
with B’ - - 2.4, B- 0.8 and JJi - 1.4, JJl - 0.1 (compare also table I). In figure 5, Asp is
defined as Asp N2Vt-l
=Bt xu’ - p(x) N2vt-1. The
uncertainties in the critical amplitudes and critical exponents are about 30% and 10%, respectively.
Analytical work is required to check the accuracy of
Bi and pl.
Table I.
-Estimates of critical exponents and critical
amplitudes of the square-lattice polymer chain according
to equations (12) and (13).
Fig. 5.
-Finite-size scaling plot for the correction to the
asymptotic behaviour of the monomer-density according to (12) for various N at temperatures T
>T, (upper part) and
T T, (lower part).
Following the above procedure, the scaling analysis
of the specific heat is presented in figure 6 and figure 7.
Since the specific heat diverges so weakly (see Fig. 1),
the divergent portion does not completely dominate
the
«background » except very close to T,. Analog to
Fig. 6.
-Finite-size scaling plot for the singular part of the specific heat according to (9) for various N at temperatures T
>T, (upper part) and T T, (lower part), and for
various
«background
»portions bt.
Fig. 7.
-Finite-size scaling plot for the correction to the
asymptotic behaviour of the specific heat according to (13)
for various N at temperatures T
>T, (upper part) and
T T, (lower part).
numerical investigations on Ising models, [39, 40] we approximate the nonsingular part by the constants b’. As shown in figure 6, we find that the singular portion of the high-temperature specific heat (C + b+ )
scales quite well according to equation (9) using
b+ -- 0.05. For T Tt, however, scaling is not quite convincing for x 8. The solid lines in figure 6 corres- pond to the predicted asymptotic power law (11).
Similar to the density discussed above, corrections to the asymptotic form (11) of the specific heat are impor-
tant, which is demonstrated in figure 7. The data are in good agreement for
x> 1 with
The amplitudes and exponents are summarized in table I. AC is defined as åC N -rztQ>t
=A * x-at -
(C + b+)N-’t"t. Again, the low-temperature specific
heat do not scale accurately : the large-N data lie systematically above the small-N data. This behaviour does not seriously affect the limiting slope a, - - 1.8;
i.e., at each chain length N, AC exhibit approximately
the slope a 1. A probable reason of the deviations from
scaling is the assumed temperature independence of
the nonsingular part bl of the specific heat, which possibly becomes a bad approximation at low tempe-
ratures.
5. Summary and conclusion.
-By the present calcu- lations for isolated square-lattice polymer chains, it
was