Upload
independent
View
0
Download
0
Embed Size (px)
Citation preview
Unconventional Phase Transitions in Constrained Single Polymer
Chain
L. I. Klushin∗
Department of Physics, American University of Beirut,
P.O. Box 11-0236, Beirut 1107 2020, Lebanon
A. M. Skvortsov†
Chemical-Pharmaceutical Academy, Prof. Popova 14, 197022 St. Petersburg, Russia
1
AbstractPhase transitions were recognized among the most fascinating phenomena in physics. Exactly
solved models are especially important in the theory of phase transitions. A number of exactly
solved models of phase transitions in a single polymer chain are discussed in this review. These are
three models demonstrating the second order phase transitions with some unusual features: two-
dimensional model of β-structure formation, the model of coil-globule transition, and adsorption a
polymer chain grafted on the solid surface. We discuss also models with first order phase transitions
in a single macromolecule which admit not only exact analytical solutions for the partition function
with explicit finite finite-size effects but also the non-equilibrium free energy as a function of the
order parameter (Landau function) in closed analytical form. One of them is a model of mechan-
ical desorption of a macromolecule which demonstrates an unusual first order phase transitions
with phase coexistence within a single chain. Features of first and second order transitions become
mixed here due to phase coexistence which is not accompanied by additional interfacial free energy.
Apart from that, there exist several single-chain models belonging to the same class (adsorption of
a polymer chain tethered near solid surface or liquid-liquid interface, and escape transition upon
compressing a polymer beween small pistons) that represent examples of a highly unconventional
first order phase transition with several inter-related unusual features: no simultaneous phase coex-
istence, and hence, no phase boundary; non-concave thermodynamic potential and non-equivalence
of conjugate ensembles. Analysis of complex zeroes of partition functions upon approaching the
thermodynamic limit is presented for models with and without phase coexistentce.
∗Electronic address: [email protected]†Electronic address: [email protected]
2
I. CONTENTS
1. Introduction
(a) Specifics of phase transitions in a single macromolecule
(b) Exactly solved models of the phase transitions
(c) Main features of the first and second order phase transitions
2. Second order phase transitions in a single chain
(a) Zwanzig-Lauritzen model of 2d crystallization
(b) Coil-globule transition
i. General considerations
ii. Mean field theories of the coil-globule transition
iii. Comparison with the Landau theory of phase transitions
iv. Comparison with the Landau theory of phase transitions
v. Experiments
vi. Computer simulations
vii. Coil-globule transition in higher dimensions
viii. Collapse vs. freezing
(c) Adsorption a single polymer chain grafted on a solid surface
i. Adsorption of ideal lattice chain
ii. Adsorption of macromolecule with excluded volume interactions
A. Scaling ansatz
B. Crossover exponent φ
iii. Continuum model of adsorption of an ideal chain
A. Exact Green’s function
B. Partition function as realization of crossover ansatz
C. Comparison with the Landau theory
D. Mapping of continuum and lattice parameters
3
E. Analytical partition function for finite lattice chains
iv. Adsorption on curved surfaces
3. First order phase transitions with phase coexistence within a single chain
(a) Mechanical desorption of a macromolecule in f -ensemble
i. The Green’s function and the partition function
A. Adsorption-force symmetry and the phase diagram
B. Partition function around the transition line f = c
ii. Unconventional features of the first-order transition with phase coexistence
iii. Landau free energy
(b) Mechanical desorption of a macromolecule in z-ensemble
i. Analogy with gas-liquid transition in (N,P, T ) and (N, V, T ) ensembles
ii. The partition function in the z-ensemble
iii. Evidence of phase coexistence
iv. Local order parameter
v. Comparison of two ensembles
4. First order phase transitions without phase coexistence within a single chain
(a) Polymer chain tethered near an adsorbing solid surface
i. Model and partition function
ii. Landau function
iii. Stretching force
iv. Microcanonical ensemble and the entropy gap
(b) Polymer chain end-tethered near an absorbing penetrable interface (near step
potential)
i. Chain attached near a liquid-liquid interface
ii. Chain attached at a liquid-liquid interface (rolling transition)
(c) Escape transition
4
i. Escape transition in 3d geometry
A. Partition function for ideal chain in D-ensemble
B. Compression curve (average force vs separation)
C. Conjugate ensemble (average separation vs force)
D. Nenequvalence of ensemble
E. Order parameter and Landau function
F. Escape of chain with excluded volume
ii. Escape transition in 2d space
iii. Another escape transition setup: dragging a polymer chain in tube
5. Complex zeroes of partition functions for a single-chain.
(a) Complex zeroes in transition with phase coexistence (mechanical desorption of
ideal chain)
(b) Complex zeroes in transition without phase coexistence (adsorption of ideal chain
tethered near solid surface)
6. Conclusions
7. Acknowledgments
8. References
A. Specifics of phase transitions in a single macromolecule.
Phase transitions were recognized among the most fascinating phenomena in physics since
the days of van der Waals, Boltzmann and Gibbs. Historically, the question of whether sta-
tistical mechanics is strictly applicable to phase transitions was much debated [1]. Onsager’s
solution of the 2d Ising model showed [2] that the full description of the phase transition is
indeed contained in the exact partition function. A lot of progress has been made since then
in understanding both physical and mathematical aspects of the phase transitions phenom-
ena. However, as the phase transition concepts are applied to broader areas new insights
and new problems arise. The aim of this review is to discuss phase transitions in a single
macromolecule and to demonstrate their unconventional features.
5
Phase transitions in macromolecular systems can be fundamentally divided into two types.
The first type occurs in the condensed bulk matter, involves a macroscopically large number
of molecules, and is basically similar to that in ordinary fluids or solids. Crystallization,
segregation of incompatible polymer liquids, and liquid-crystalline ordering of polymers can
serve as examples. The number of molecules in a typical experiment is of the order of the
Avogadro number and finite-size effects are completely negligible unless specific geometry
with the large surface-to-volume ratio is examined. The size of a single macromolecule plays
only minor roles in the nature of these transitions at least as far as equilibrium aspects are
concerned.
The second type of phase transitions belongs exclusively to the realm of polymers since
it is realized at the level of a single macromolecule and does not have any analogies in the
physics of low molecular mass systems. Note, that even a single polymer coil in dilute solu-
tion is characterized by anomalously large size fluctuations analogous to critical phenomena.
The best-known examples of phase transitions in a single chain are the transition from a
loose coil to a compact globule, as well as adsorption-desoption transition of a polymer at a
solid or liquid interface.
The notion of a phase transition in a single chain requires a conceptually clear understand-
ing. The number of the repeat units in a single macromolecule, N , is not really macroscopic,
usually 102 − 104, (for DNA 109 − 1010), while the traditional concept of a phase transition
requires taking the thermodynamic limit, N → ∞ . It turns out, however, that the frame-
work of the phase transition theory including the study of finite-size effects (which are the
rule rather than exception for the single-molecule transitions) proves to be very fruitful. The
approach to the thermodynamic limit for single chains is in principle accessible to experi-
mental studies since many polymers can be synthesized with their molecular weight being
varied over several decades without changing the intra-chain interactions. In many cases
these finite-size effects are completely different from those in bulk matter: for a droplet,
the finite-size effects are dominated by the surface contribution to the free energy, while a
polymer taking a typical coil conformation does not have a well-defined surface at all.
The interest in single-chain phase transitions has been spurred by recent advances in
experimental manipulations techniques. Nowadays, the existing potential includes various
powerful experimental methods such as AFM, magnetic levitation, photon microscopy, op-
tical tweezers, etc.
6
Manipulations of single chain have become available since 1986 [3] and the most impres-
sive progress was achieved in the physics of nanoscale objects. It possible now to measure
the elasticity of single chain [4–6], estimate the intensity of the receptor-ligand bond [7], to
study the physical characteristics of fundamental biological objects (DNA, proteins, molec-
ular motors) [8–13], and to determine the energy dissipation in the case of friction when a
single molecule moves directionally on a solid surface [14]. The forces required for unfolding
of DNA were measured [15, 16], the strength of the specific interaction between antigen-
antibody molecules was estimated [17], the conformational transitions in polysaccharides
were observed [18] as well the transitions during reversible protein folding [19]. Force-strain
curves for individual polymer chains have been studied extensively [5, 6, 20–22] , for several
synthetic polymers including both polyelectrolytes and uncharged polymers, biopolymers
(polymerase, proteoglucans, and xanthane), etc. Evident achievements in this direction
have triggered the development of a new science coined the “nanomechanics of polymers”. A
typical AFM experiment involves measuring the average force while controlling the tip-to-
sample separation. There is another AFM operation mode in which the force is fixed while
the separation is adjusted and measured. These two modes correspond to two different sta-
tistical ensembles. In principle, a combination of the two types of measurements would allow
to address the question of ensemble equivalence experimentally. There are two aspects in
the problem of relationship between different ensembles. First, although standard theorems
of statistical mechanics state their equivalence in the thermodynamic limit this does not
apply to finite-size systems. Second, single chain phase transitions include unique examples
of ensemble non-equivalence even when the thermodynamic limit is taken, as disscussed in
detail below.
Theoretical description of single-chain phase transitions may include the notion of phase
coexistence within one macromolecule, which we demonstrate in Section IV. On the other
hand, first order phase transitions in a single chain may also include unconventional examples
when no phase coexistence is possible in principle, see Section V.
B. Exactly solved models of phase transitions
Exactly solved models revealing details of the phenomena without making a priori con-
jectures or approximations are especially important in the theory of phase transitions. The
7
thesaurus of exact solutions is rather modest [23? , 24]. First, there are two-dimensional lat-
tice models (Ising, Potts, eight-vertex, three-spin and segnetoelectric). Two somewhat more
artificial models have also played a significant role: the one introduced by Kac [25, 26] with
infinitely weak, infinitely long-range interaction, and the spherical model corresponding to
the infinite number of spin components. The models mentioned admit exact solution in the
thermodynamic limit N →∞, where N is the number of particles in the system; finite-size
corrections are available for the Ising model. Only some of these models are solved in the
presence of the external field.
A number of models from polymer physics can be added. Surprisingly, even most simpli-
fied models exhibit non-trivial behavior which could be absent in their low-molecular mass
counterparts. Exact solutions describing second-order phase transitions are known for the
model of two-dimensional β-structure formation [27] and for some closely related models of
directed polymer adsorption [28–30]. Another inherently related set consists of the ideal
model of DNA melting [31] (see [32] for a review), and the model of adsorption of ideal
lattice chain on planar and cylindrical surfaces [33–39]. For all the models mentioned exact
solutions are obtained in the grand canonical formalism, and finite-size effects are analyzed
numerically or by scaling considerations.
There is yet another class of single macromolecule models which admit exact solutions
for the partition function for finite number of segments, N , and demonstrate true phase
transitions in the N → ∞ limit. These are ideal continuum chains interacting with some
external potential and described by the Edwards Hamiltonian
H(r(s))/T =
Nbˆ
0
ds
[3
2b2
(dr
ds
)2
+ U(r(s))
](1)
where b is the segment length, the segments are specified by the contour length distance s
counted from one end, and is the vector specifying the position of the corresponding segment
in space. The partition function for a chain with two ends fixed is
G(r0, rN , N) =
ˆD(r(s)) exp [−H(r(s))] δ(r(0)− r0)δ(r(Nb)− rN) (2)
where integration is over all paths r(s) connecting the two end positions .
It is demonstrated by direct verification that the partition function satisfies the equation
8
∂G(r0, r, s)
∂s=Nb2
6
∂2G(r0, r, s)
∂2r− U(r)G(r0, r, s) (3)
Eqs (1) and (2) establish the well-known analogy between the statistical mechanics of
ideal polymers and the path-integral formulation of quantum mechanics of a particle in
potential U(r) [40]. The wavefunction, ψ(r, t), for a particle of mass m in a potential U(r),
is given by the path integral
ψ(r, t) =
ˆD(r(t)) exp [iS/~] δ(r(t)− r), (4)
where
S =
ˆdt
[m
2
(dr
dt
)2
− U(r(t))
](5)
It is equivalent to the time-dependent Schrödinger equation
∂ψ(r, t)
∂t= − ~2
2m
∂2ψ(r, t)
∂2r+ U(r)ψ(r, t) (6)
A mapping between the quantum-mechanical problem and Gaussian chain in an external
field requires the associations: t ↔ s,ψ(r, t) ↔ G(r0, r, s),r(t) ↔ r(s),m ↔ 3/(Nb2),~ ↔
i,S ↔ H(r(s)),U(r(t))↔ −U(r).
Exactly solved models based on the continuum ideal chain description include examples
of first-order and continuous phase transitions with some very unconventional features that
are discussed below in Sections I, II, and III.
It is worth noting that the mathematical techniques employed in solving different classes of
models are quite distinct. Classical solutions for 2d lattice models are based on the formalism
relating the partition function to the eigenvalues of a transfer matrix. The ideal lattice
chain models are solved by finding the grand canonical partition function and analyzing its
singularities. The grand canonical formalism hinges on the superposition principle expressing
the ideal nature of the model whereby different parts of the chain are non-interacting except
for the connectivity condition. Finally, for the continuum chain model the path-integral
representation of the partition function leads to an equivalent formulation in terms of a
Green’s function of a partial differential equation (analogue of the Schrödinger propagator).
Again, the superposition principle plays a fundamental role.
9
The exactly solvable continuum models with analytical description of finite-size effects can
be pushed beyond the equilibrium partition function to obtain exact solutions for the Landau
free energy as a function of the order parameter[41]. This description covers configurations
that are far away from equilibrium and gives the full free energy landscape defining the first-
order transition kinetics including the possible metastable states. Metastable states as well
as the barriers separating them from stable equilibrium states are described analytically
for finite N . This is in contrast to non-polymeric models, where the Landau free energy
is normally postulated on the basis of symmetry considerations, while its exact analytical
calculation is not possible.
In conclusion we point out yet another fascinating feature of some exactly solved polymer
models. A very distinct approach to describe phase transitions was introduced by Yang and
Lee [42, 43], and is based on representing the partition function in terms of its zeros in the
complex plane of fugacity (or temperature, as suggested later by Fisher [1]. The theory is
mathematically rigorous which brings about a hefty price to be paid: the problem of finding
the actual distribution of zeros for the partition function of even a simple model turns out
to be formidable. Yang and Lee proved a circle theorem, which states that for ferromagnetic
Ising models and for the attractive lattice gas models, the zeros are located on a circle.
Later the theorem was further extended to other ferromagnetic lattice systems. However,
the densities of zeros are exactly known only for the simplest models (ferromagnetic Ising
models in 1d [43], on the Bethe lattice [44], on fractal lattices [45], and for an ideal Bose-gas
in an external potential [46]).
A number of exact results concerning temperature zeros have been obtained for the Ising
model with spin 1/2 [47–55] and with higher spin [56–59].
There exists a phenomenological approach relating certain characteristics of the transition
(amplitudes and critical indices) to the parameters of the distribution of zeros assumed to
be known [60], as well as scaling predictions for this distribution [51]. In practice, for a given
model the zeros are calculated numerically for small samples and then some extrapolations
are employed [61, 62].
There are at least two polymer models [63, 64] that admit analytical solution for the
distribution of complex zeros of the partition function in the cases when first-order and
continuous phase transitions are involved. Exact analytical expressions appearing in the
thermodynamic limit substantiate the scaling considerations existing in the framework of
10
the Yang-Lee-Fisher approach. We discuss this in last part of this review.
C. Main features of the first and second order phase transitions.
The conceptual framework of theory of phase transitions cannot be applied automati-
cally to a single macromolecule without some re-thinking and re-interpretation. We start
be recalling some basic facts about the standard analysis of phase transitions. The earliest
formal classification goes back to Ehrenfest [65]. If the first derivatives of the thermody-
namic potential (entropy, volume per particle, etc.) experience a jump with the change in
a control parameter the transition is first order. If these are changing continuously and the
singular behavior (jumps or divergences) appears only in higher derivatives the transition is
continuous or second order (in the contemporary language these terms are used as equiva-
lents irrespective of which particular higher derivative exhibits a singularity). Importantly,
the above classification applies in the N → ∞ limit, and only to specific statistical ensem-
bles with two fixed intensive parameters, such as the Gibbs NPT ensemble or its magnetic
NHT counterpart, or the grand canonical µV T ensemble. Here, the equilibrium state of
the system as a whole is a homogeneous phase before and after the transition point; phase
segregated states can exist only at the first-order transition points, and their relative phase
composition is not independently controlled.
On the other hand, the Helmholtz NV T ensemble is perfectly suited to study well defined
phase coexistence as we learn from textbook sections on van der Waals fluid[65]. However,
naïve application of the Ehrenfest criteria to ordinary gas-liquid condensation in the NV T
ensemble turns out to be confusing and counterproductive. Identification of the appropriate
ensemble is an important aspect that may be far from obvious in the context of polymer
models as shown by examples discussed below.
The basis for conceptual understanding of the essential background of phase transitions
was provided by the theory due to Landau [41, 66]. The central idea is to identify an order
parameter and then to analyze the non-equilibrium free energy as a function of the order
parameter. Historical examples of superfluidity and superconductivity show that defining a
suitable order parameter may be a non-trivial task by itself.
Landau’s original suggestion was to expand the free energy in powers of the order pa-
rameter, s, assuming analyticity at s = 0. Although it was recognized early enough that
11
F
s
a
T > Tc
T = Tc
T < Tc
F
s
h = 0h > 0
h < 0b
Figure 1: Nonequilibrium Landau free energy Φ as a function of the order parameter s for an
Ising-type model at second-order (a), and the first-order (b) phase transitions. In the first case the
transition is driven by a decrease in temperature T , while in the second case - by changing the
external field h coupled to the order parameter.
the original version is equivalent to a mean-field approximation (and therefore generally
incorrect) the Landau-Ginzburg formulation which incorporates interacting fluctuations of
the local order parameter field is the cornerstone of the critical phenomena theory. For the
sake of qualitative discussion the simple original version is quite sufficient. Two possible
scenarios are illustrated in Figure 1a,b showing the change in the shape of the Landau free
energy as a function of the order parameter, Φ(s).
In the first scenario Φ(s) = A (T − Tc) s2 + Bs4, always has only one minimum corre-
sponding to the true equilibrium state; in the high temperature range T > Tc the minimum
stays at s = 0 and represents the stable disordered phase, while for T < Tc the minimum
gradually moves away signifying a continuous onset of spontaneous ordering. Here, tem-
perature is taken as a particular example of a control parameter: other control parameters
will be discussed below in specific models. In fact, there may be two (or many) equiva-
lent minima representing the symmetry of the original Hamiltonian but only one is relevant
once the symmetry is spontaneously broken. Importantly, in the vicinity of T = Tc the
minimum is anomalously shallow which implies large fluctuations in the order parameter
and correspondingly, anomalous susceptibility to a conjugate field. Thus three features are
intrinsically linked together in the scenario of a second order (continuous) phase transition:
the order parameter starts growing at the transition point without a jump; the order pa-
rameter distribution always remains unimodal; and the fluctuations in the order parameter
12
display anomalous growth in the vicinity of the transition point.
Figure 1b shows the other scenario corresponding to a generic first-order transition. Here,
at T < Tc , the Landau free energy Φ(s) = −hs−A (T − Tc) s2+Bs4 has two minima meaning
that the order parameter distribution is bimodal. The deeper minimum corresponds to the
true equilibrium whereas the other minimum describes a metastable state. The equilibrium
transition temperature is defined by the condition that both minima are of equal depth. At
this point, the two different states are equally probable which would normally imply phase
coexistence. Upon crossing the transition point, the equilibrium state switches from one
minimum to another; this is accompanied by a jump in the order parameter and in the
first derivatives of the equilibrium thermodynamic potential. Again several distinct features
are packaged together in this scenario: a jump in the order parameter accompanied by a
δ-peak in the order parameter fluctuations and in the heat capacity (thus giving rise to
latent heat); bimodal distribution of the order parameter and metastability effects but no
anomalous pre-transitional fluctuation growth; phase coexistence at the transition point. A
more detailed picture incorporating spatial distribution of the order parameter leads to the
notions of an interface, nucleation, etc [66].
Note, that when the Landau free energy applies to the system as a whole, scenarios shown
in Figs1a,b are relevant only in the NPT or similar ensemble where macroscopic phase
segregation is restricted to the transition point only. If the order parameter and the density
of the Landau free energy are understood as local quantities, the choice of ensemble is largely
irrelevant since even in the NV T ensemble the system is always approximately homogeneous
locally. Traditionally, a useful criterion for identifying the nature of the transition when
studying finite systems is the shape of the distribution in the order parameter (or energy):
bimodal in the case of first-order transitions or unimodal in second order transitions.
Some very unconventional behavior in a class of single-macromolecule models demon-
strating first order transitions will be discussed in detail in the present review. Classically,
phase coexistence is a necessary attribute of first-order phase transitions. This always goes
together with metastability effects in the vicinity of the equilibrium transition point and a
nucleation mechanism for the appearance of new phase. The models to be discussed below
defy this common wisdom. In one case, phase coexistence can be properly identified but
nucleation mechanism is absent and no metastable states are to be found; in the other case,
metastable states do exist but phase coexistence is precluded by the very nature of phases.
13
aT*
n*
b
0 5 10 15 20 25
0.100
0.050
0.030
0.150
0.070
¶E*
¶T* Γ
T*
c
0 2 4 6 8 10
0.5
1.0
1.5
2.0
Figure 2: (a) Zwanzig-Lauritzen model of 2d cristallization; (b) the average number of bends per
monomer as a function of reduced temperature T∗ = kT/ε; (c) square-root singularity of the heat
capacity vs. T∗. (According to [27]).
We show examples when the convexity of equilibrium thermodynamic potentials is violated
leading to nonequivalence of different statistical ensembles in the thermodynamic limit and
in one instance, to an equilibrium negative compressibility.
II. SECOND-ORDER PHASE TRANSITIONS OF SINGLE-CHAIN.
A. Zwanzig-Lauritzen model of 2d crystallization.
One of the earliest exactly solved models demonstrating a non-trivial critical behavior in
a single macromolecule was introduced by R. Zwanzig and J.Lauritzen [27, 67]. The model
is relevant for crystallization, β-sheet formation and surface roughening phenomena. The
chain on a 2d plane is folded into a sequence of straight segments (not necessarily of the
same length) densely stacked in one direction, see Figure 2 a.
There are two parameters in this model: the attraction energy per unit length ε and the
bending energy u . The energy is due to short-range attraction between adjacent stacked
segments and for each pair of segments is proportional to their overlap length. Each bend
carries additional energy due to deformation and the surface effects. The grand canonical
partition function as a function of the chemical potential µ is evaluated exactly
Ξ(µ) = −1 +2
σ
Jν(σν)
Jν−1(σν)(7)
where Jν(x) is the Bessel function, ν = εµ−ε , σ = 2
√e−βµ
βε, and β = 1/(kT ).
If bends are energetically disadvantageous (u > 0), the crystal is a perfect (infinite)
square at T = 0, and the number of bends per monomer n∗ is zero in the L→∞ limit. The
14
number of bends per monomer serves as an order parameter and can be used to distinguish
between the low-temperature crystalline phase and the high-temperature phase which is
entropy dominated and is characterized by finite order parameter (the number of bends
being proportionate to the chain length), see Figure 1a. Since the model does not allow
random coil configurations, the high-T phase can be loosely described as a crystal destroyed
by extreme roughening. Above the critical temperature the average order parameter grows
as n∗ = A(T − Tc)1/2.
Although the critical index for the order parameter growth is 1/2 coinciding with the
mean-field value, other characteristics deviate from mean-field predictions. In particular,
the heat capacity displays a square-root singularity rather than a finite jump, Figure 1c.
The exact partition function of the Zwanzig-Lauritzen model was used to extend the
theory and include 3d coil configurations [68]. This changes the nature of the transition and
makes it first order with bimodal distribution in energy, and infinite heat capacity. Finite-
size effects for this system was analyzed numerically [69]. Later the theory was subsequently
extended to include the possibility of α-spiral formation and selective adsorption in a chain
with α-helix, β-sheet and random coil sequences [70, 71]. This theory combined with MC
simulations of finite chains with N = 20 was used to explain experimental data on structural
changes in beta-coil transition and alpha-beta-coil transitions in polypeptides induced by
adding low molecular mass detergents and to the surfactant-induce transitions in globular
proteins [72]. In all of these theories and calculations, interaction between different structural
elements of the chain as well as excluded volume interactions within coiled sub-chains was
ignored.
B. Coil-globule transition.
Interest in the coil-globule transition and in the features of globular conformations in
homopolymers arouse in connection with the problem of denaturation of globular proteins,
protein folding, and DNA packing. Fundamentally, the coil-globule transition is an extension
of the gas-liquid condensation for particles connected into a linear chain. Chain connectivity
brings two important distinctions as compared to the standard gas-liquid transition: first,
the collapse of an isolated chain takes place at zero external (osmotic) pressure; second, the
translational entropy is replace by configuration entropy due to internal degrees of freedom
15
in the macromolecule.
1.General considerations. The standard description of the collapse transition is a tricrit-
ical point related to a field theory for an n-component vector field with ϕ4−ϕ6 interaction,
in the limit n → ∞ [73]. One might then expect that above the upper critical dimension
(d = 3) some type of self-consistent mean-field theory would give a full description of the
transition [74].
For finite polymer length N the transition is rounded. In the language of the finite-size
scaling analysis this means that in the behavior of various thermodynamic quantities in the
vicinity of the Θ-point follows certain scaling laws. In particular, the average monomer
density in the chain ρ = Na3/ 〈R〉3 (where 〈R〉 is the gyration radius of the chain at a given
temperature) is given by
ρav(N, τ) = ρav(N, 0)Φ(τNφ), (8)
where τ = (T − Θ)/Θ is the relative deviation from the Θ-point, ρav(N, 0) ∼ N−1/2 is the
density at the Θ-point. The crossover exponent φ describes how the model rescales as the
tricritical point is approached from different directions [75], and Φ(z) is the crossover scaling
function. Some properties of the scaling function can be deduced on general grounds. At
large negative values of the argument the scaling function must be a power law Φ(z) ∼ |z|x
with the index x determined by the condition that the density of a well-formed globule is
independent ofN , which gives x = 1/(2φ) . For large and positive values of the argument, the
scaling function describes the approach to self-avoiding limit with 〈R〉 ∼ N ν and ρav ∼ N1−3ν
, where ν = 0.5878 is the Flory index for the gyration radius of a self-avoiding chain in 3D.
It follows that Φ(z) ∼ zy , with y = (3/2− 3ν)/φ. Altogether,
Φ(z) ∼
|z|1/(2φ) z � 1
1 z = 0
z3(1−2ν)/(2φ) z � 1
(9)
and the average density is
16
ρav(N, τ) ∼
|τ | τ � 1
N−1/2 τ = 0
τ 3(1−2ν)N1−3ν τ � 1
(10)
Mean-field type theories give a linear growth of density with |τ | which implies x = 1 and
φ = 1/2 (see below). This is all consistent with the general conclusions of the field-theoretical
approach for d=3 since the space dimensionality is just equal to the upper critical dimension
(though renormalization-group arguments suggest logarithmic corrections).
2.Mean-field theories of the coil-globule transition. We discuss briefly the existing
mean-field theories of the coil-globule transition. The interaction free energy can be written
as a virial expansion series, and the configuration entropy is evaluated separately using the
ideal chain statistics. In the simpler version of the theory due to Birshtein and Pryamitsyn
(BP) [76], one deals only with the global monomer density ρ = Na3/R3 defined by the
gyration radius of the chain. Here R is a parameter characterizing the state of the chain and
may be quite different from the equilibrium average value. The entropy is also expressed in
terms of the gyration radius R. Equilibrium size of the chain is found by minimizing the
total free energy as a function of R following the idea proposed by Flory [77] to describe
swelling under good solvent conditions .
The interaction free energy is written keeping the two dominant terms in the virial ex-
pansion:
Fint(R,N)
kT=R3
a3
(A2ρ
2 + A3ρ3)
(11)
The virial coefficients appearing in this expressions should be understood as renormalized
coefficients [78, 79] in order to account for the distinction between local and global densities.
In the standard approach A2(T )is taken as a linear function of temperature in the vicinity
of the Θ-point, A2 = bτ , and A3(T ) - as a constant: A3 = c. The entropy of an ideal
chain with a given radius of gyration, S(R), was studied by Fixman [80]; combining the full
function S(R) with the interaction free energy (11) provides a unified description of swelling
for τ ≥ 0 and compactification for τ ≤ 0. A simple analytical form
S(R)/k = 5 ln(R0/R)− 9R20/4R
2 (12)
17
can be used to describe compact configurations with R < R0 where R = (N/6)1/2a is the
equilibrium average gyration radius of an ideal walk.
Minimization of the total free energy F (R,N) = Fint−TS(R,N) with respect to R leads
to the following equation for the expansion factor α = R/R0:
α3 − α = BN1/2τ + C
(1
α3− 1
), τ < 0, (13)
where B and C differ from b and c, respectively, only by numerical coefficients. In the
thermodynamic limit N → ∞ this equation leads to a simple picture of the transition,
consistent with the crossover scaling form discussed above. Below the Θ-point, (τ < 0, α�
1) Eq.13 gives the swelling coefficient as α ∼(
CBN1/2|τ |
)1/3
. At the Θ-point itself, τ = 0 and
α = 1. Above the Θ-point (τ > 0, α � 1)the configurational entropy has to be modified
appropriately, giving the classical Flory picture of chain swelling with α ∼(BN1/2τ
)1/5. If
one considers the average density as the order parameter, its variation with temperature can
be summarized as:
ρ =
τ−3/5N−4/5 τ > 0
N−1/2 τ = 0
bc|τ | τ < 0
(14)
coinciding with the scaling prediction of Eq.10 with the Flory value of the index ν = 3/5.
The equilibrium free energy in the globular state counted from that at the Θ-point is F/kT =
−b2Nτ 2/(4c). This implies that the transition is second order with a finite jump in the heat
capacity, as one expects in the mean-field approach.
3.Comparison with the Landau theory of phase transitions. The phenomenological
theory of phase transitions by Landau [41] is based on consideration of the non-equilibrium
free energy Φ as a function of the order parameter, s. For Ising-class systems with a scalar
order parameter and mirror-reflection symmetry, Φ is expanded in powers of s in the vicinity
of the second-order transition point, only the terms with even powers being present due to
symmetry considerations:
Φ(s) = Φ0 + A(T − Tc)s2 +Bs4 (15)
Analysis shows that the entropic contribution to the total free energy is negligible in the
well-formed globular state [76]. Then the virial expansion (15), re-written as
18
Φ(ρ,N)
NkT= bτρ+ cρ2 (16)
can be interpreted as the non-equilibrium Landau free energy as a function of the order
parameter. Since the order parameter ρ is non-negative by its physical meaning, a complete
analogy with the original Landau theory is achieved by the mapping ρ = s2. At or above
the Θ-point the minimum is located at ρ = 0; as the temperature drops below the Θ-point,
the minimum shifts linearly with τ . This is a classical picture of a second-order transition
within the tenets of the Landau theory. For finite N the transition is smoothed over the
temperature range δτ ∼ N−1/2 (this can be detected, e.g.. in the shape of the heat capacity
jump), consistent with the crossover exponent value φ = 1/2.
A more sophisticated version of the mean-field theory due to Lifshits, Grosberg, and
Khokhlov (LGKh) [78, 79] is based on the density-functional approach and historically ap-
peared earlier. The total free energy is written as functional of the local density profile;
minimization yields the equilibrium density profile, and all the other characteristics are ex-
pressed in terms of it. Both versions (BP and LGKh) give the same results for all global
chain characteristics and the fundamental features of the transition. We have concentrated
on the simpler BP version since it allows a straightforward analysis in the framework of the
Landau theory and is easily extended to arbitrary space dimension.
Interestingly, the two versions differ in their starting points as far as the transition itself
is concerned. The LGKh version is based on the ground-state dominance approximation
which implies that the globule is already well-formed. The transition is therefore approached
from the globular side of the phase diagram, and the entropic contribution gets a natural
interpretation as the surface free energy. The starting point of the BP approach is a weakly
perturbed ideal coil at the Θ-point, and the entropy term just accounts for compactification
with respect to the ideal coil size.
4.Experiments. Most experimental studies of the coil-globule transition have been carried
out using various solutions of homopolymers in organic solvents: polystyrene in cyclohexane
[81–87] and poly(methyl-metacrylate) in appropriate theta-solvents [88–93]. Monitoring an
individual chain in the process of coil-to-globule collapse and observing fully collapsed ther-
modynamically stable single chain globules is extremely difficult since, on one hand, van der
Waals interaction driving the transition in organic solvents is very weak and on the other,
multi-chain precipitant is formed even at low polymer concentrations. Experimental difficul-
19
ties have been overcome only recently. Fairly monodisperse very high molar mass polymer
samples (up to N ∼ 104 − 105 segments) of temperature responsive poly(N-isopropyl acril-
amide) and poly(N,N-diethylacrilamide) in aqueous solutions were investigated at extremely
low concentration in a wide temperature range near the phase transition and reversible coil-
to-globule transition was demonstrated [94–103].
Theoretical analysis of the collapse of chains of finite length is based on expression (8)
which turns out to be quite useful for interpreting experimental data. In particular, it sug-
gests that the data points presented in the scaling coordinates (α−α3)/(α3−1) vs. |τ |N1/2
should collapse onto a straight line, its slope and intercept giving the coefficients B and (-C),
respectively. Experimental data for coil-to-globule transition in organic solvents show good
agreement with the predictions of the mean-field theory for the expansion factor, see Fig-
ure 3a. Temperature-responsive polymers in aqueous solutions demonstrate a considerably
sharper collapse, Figure 3b.
Thus the coil-globule transition seems to be established as a universal phenomenon at least
for a broad class of typical flexible polymers, including those that demonstrate anomalous
temperature behavior with the collapse induced by increase in temperature.
5. Computer simulations. The first computer simulation of the coil-globule transition
of a flexible model chain on a cubic lattice was undertaken over 40 years ago [105]. Later
on this transition was studied extensively using both Monte Carlo [109–111] and Molecular
Dynamics [112–114] techniques. Earlier computer simulations were restricted to relatively
short chains with N ∼ 102 − 103 where profiles with a well-formed core of constant density
and a diffuse boundary could not be obtained. However, very recent MC simulations with
N up to 104 [115] demonstrate density profiles with compact core of constant density that
are qualitatively very similar to those predicted by the functional-density theory [116–119],
see Figure 4. Systematic quantitative comparisons are not available yet.
6. Coil-globule transition in higher dimensions. Although the overall picture of
the collapse transition for flexible polymers as presented above seems to be quite clear and
satisfying some intriguing questions are left unanswered. Field-theoretical approach to the
collapse phenomenon is far from being fully consistent and clearly understood. Confusion is
revealed most convincingly when the transition is studied in 4-dimentional space. Standard
theoretical arguments suggest that finite-size scaling with the mean-field exponent φ = 1/2
should be exact since one is above the upper critical dimension. However, other approaches
20
æææ
æ
æ
æ
æ
æ
à
à
à
à
à
à
ì
ì
ì
ìì
ì
ì
ì ì
ΤN1�2
1-Α-3a
Α3-Α
1-Α-31 2
3
0.0 0.5 1.0 1.5 2.0 2.5 3.0
0.05
0.10
0.15
0.20
0.25
0.30
0.35
ææ
ææ
ææææææææææææææææææ
ààà
à
ààààà
àà
ààààààààààà
N = 108 107 106
105à PDEAM
æ PNIPAM
0.96 0.98 1.00 1.020.0
0.5
1.0
1.5
Θ�T
ΑS
Figure 3: (a) Experimental data for PS in cyclohexsane (1) [104] , PS in dioctyl phthalate (2)
[81], and results of MC simulations for lattice chain (3) [105, 106] in the coordinates suggested by
the BP theory [76]. (b) Temperature dependence of the expansion factor αs = 〈Rg〉T / 〈Rg〉Θ for
poly(N,N-diethylacrylamide) (PDEAM) with M = 1.7 · 107 g/mol in water [103] and for poly(N-
isopropylacrylamide) (PNIPAM) with M = 1.0 · 107g/mol in water [96, 97, 107]. The Flory Θ-
temperatures are 28.5oC for PDEAM , and 30.5oC for PNIPAM. Solid lines represent calculations
on the basis of yet another version of the mean-field theory [108] with several values of the number
of monomers, N . (The data are taken from [76], and [103])
lead to completely different and contradicting values of the crossover exponent (φ = 0 [120]
or φ = 1 [121]. MC simulations by PERM kinetic growth algorithm with N up to 16384
give clear evidence that close to the Θ-point the crossover scaling is not applicable at all.
It is destroyed by pre-transitional behavior with clear features of a first-order transition:
bimodal distribution of energy and a sharp drop in the chain size [122] .
Quite unexpectedly, this unusual behavior was first suggested on the basis of the mean-
21
Figure 4: Segment density distribution as a function of the distance from the center of mass of the
lattice model chains for various chain lengths N (indicated in the Figure) and energy interaction
parameters between units χ = 0.37 (a) and χ = 0.45 (b), both in the poor solvent regime. For this
model, χ = 0.28 corresponds to Θ-conditions. (c) Swelling parameter α = R/R0 as a function of
τN1/2: simulation data for different N fall on a universal master curve. (According to [115])22
field theory [123] appropriately generalized for arbitrary space dimension d . This can be
easily done within the approach presented in this section taking the Landau free energy as
a function of the average monomer density. The virial expansion terms (16) remain the
same; the entropy of compact conformations still has the form of Eq.12 with the dominant
term S(R) ∼ N/R2 although numerical pre-factors will change. The simplest modification
that adequately describes the (nearly) unperturbed ideal coil at τ = 0 consists of adding the
expansion branch of the entropy so that the ideal Gaussian size is recovered at b = c = 0[124].
The resultant Landau free energy as a function of the global density is given by
F (ρ,N)
NkT= bτρ+ cρ2 + γ
(ρ2/dN−2/d + ρ−2/dN−2+2/d
)(17)
where γ is a numerical coefficient related to d. For d > 3 and large N the Landau free energy
(12) displays all the features that lead to the collapse scenario articulated by Owczarek
and Prellberg [121]. This includes two competing minima at a certain temperature shift
|τ ∗| ∼ N−1/(d−1) causing a sharp drop in the chain size from N1/2 to N1/(d−1). Upon further
decrease in temperature the chain compactifies smoothly and the density grows linearly with
τ . Within this scenario the higher is the space dimensionality, the worse is the failure of the
finite-size scaling ansatz, Eq. (14). The case of 3d is marginal: bimodality can occur only
when the third virial coefficient, c, is numerically very small, and even then the drop does
not sharpen with the increase in N .
It seems that the difficulties in the systematic field-theoretical treatment of the Θ-point
stem exactly from the dramatic asymmetry between the two directions of approach to the
Θ-point which appears in high-dimensional spaces. For τ > 0 the Landau free energy always
has one minimum, while for τ < 0 bimodality appears and undermines the very basis of the
renormalization-group approach.
7. Collapse vs. freezing. Another intriguing aspect of the single polymer chain collapse
was found recently in MC simulations of the bond-fluctuation model [125]. Upon lowering the
temperature, the coil-globule transition is followed by freezing and formation of an ordered
crystalline state which proceeds as a discontinuous first-order transition. This by itself
would not be surprising. However, extrapolation to N → ∞ has lead to a conclusion that
the abrupt freezing transition happens at the Θ-point thus preempting the much smoother
coil-globule transition.
This was in contrast to a big body of data from earlier simulations based on simple cubic
23
lattice walks that display only one collapse transition. It is well known, however, that the
lattice gas on the simple cubic lattice is equivalent to the Ising model and there is only
one condensation transition in this case. Bond-fluctuation model turns out to be closer to
off-lattice models where two distinct transitions (gas-liquid and liquid-solid) are found for
simple fluids.
A much broader point of view has been formulated very recently by Taylor et al. [126,
127]. Off-lattice simulations has shown that the relative position of the two transitions
depend on the range of attraction between the repeat units of the chain. For potentials with
hard repulsive core and a square-well attractive part the relevant parameter is the ratio
λ = r1/r0 where r0 is the radius of the repulsive core, and r1 is the outer radius of the
attractive well. It was shown that a smooth change in λ leads to shifts in both transition
temperatures, and for λ sufficiently close to 1 the coil compactifies directly to an ordered
solid state without an intermediate liquid-like globule (see Figure 5 ). Gas-solid transition
bypassing the liquid state is not specific to the collapse of a single flexible macromolecule
and has been known in the context of phase transitions in colloidal systems. For colloids
(in contrast to atoms and molecules) the ratio of the attraction/repulsion distance can be
varied in a broad range. Phase diagrams of colloidal suspensions include all three phases
(gas, liquid and solid) for large values of λ but the range of existence of the liquid phase
shrinks and eventually disappears with the decrease in λ.
One may speculate that real polymers composed of very bulky repeat units connected
by short flexible links will show abrupt collapse from a random coil directly into an ordered
solid-like state. In particular, the findings of Taylor et al open up intriguing venues of
research on the role of the building block size in the specifics of proteins folding.
C. Adsorption a single polymer chain grafted on the solid surface.
The study of polymer adsorption from a solution onto a solid surface has a long his-
tory and is important for many applications such as adhesion, coating of surfaces, wetting,
adsorption chromatography, etc. A fundamental issue is the adsorption-desorption second-
order transition of an isolated long polymer chain grafted on a planar surface. There is
an important analogy with the melting transition of a double-stranded DNA because the
conformational features in these two cases are similar, consisting of an alternation of trains
24
Figure 5: Temperature-range interaction T−λ phase diagram for the N=128 chain with square-well
potential between units with depth −ε. The filled circles locate the continuous coil-globule collapse
transition and the filled squares locate the discontinuous globule-crystal or coil-crystal freezing
transition. The connecting dashed and solid lines are simply interpolations between these points
and uncertainties are smaller than the symbol size. The simulation snapshots show representative
chain conformations coexisting with different energy -E/ε=404 and 89 at the freezing transition for
interaction ranges λ=1.05 and with−E/ε=379 and 299 at λ=1.15 , as indicated. For short-range
interactions λ <1.05 the chain freezes directly from an expanded coil state while for longer range
interactions λ >1.06 the chain freezes from the collapsed globule state.(From the work [127])
(or native helix pieces) and single (or double) loops.
1. Adsorption of infinite ideal lattice chain
Historically the adsorption of a flexible polymer chain on a solid surface was described by
ideal random walks on regular lattices [33, 128, 129]. The interaction of each step with the
surface is described by a square-well potential with a depth ε and a width b. We take ε and
b as energy and length units and use the dimensionless interaction parameter χ = βε. Exact
partition function was calculated by Rubin in the N →∞ limit [33, 128, 129]. In particular,
for a chain on a six-choice cubic lattice the free energy per segment (in the adsorbed state,
this coincides with the chemical potential) is given by
25
βµ (χ) =
ln[
6(eχ−1)eχ
(√4 + 1
eχ−1− 2)]
χ ≥ χc
−N−1 ln[(e−χ − e−χc)
√6πN
]χ ≤ χc
(18)
The critical value χc below which a chain does not adsorb is χc = ln(6/5) . Close to
the critical point βµads(χ) ≈ (χ − χc)2. The quadratic dependence implies a second-order
adsorption/desorption transition with a finite jump in the heat capacity. The adsorption
order parameter is defined as the fraction of segments, θ, in contact with the surface. The
average order parameter is given by 〈θ〉 = ∂(βµads)/∂χ. Its initial growth is linear in the
deviation from the critical adsorption point: 〈θ〉 ∼ χ− χc.
2. Adsorption of macromolecule with excluded volume interactions.
Scaling ansatz. Great efforts were put into elucidating the excluded volume effects in
the adsorption transition. Since the adsorption point in the grand canonical formulation is
of a tricritical nature [130], the crossover scaling picture presented in the previous section
is expected to hold. For the canonical partition function in the near vicinity of the critical
point this corresponds to the following scaling ansatz
QN(χ) = QcΨ(τNφ) (19)
where QN(χ) is the canonical partition function, Qcis the partition function at the critical
adsorption point, τ = χ/χc−1 is the deviation from the critical point, and φ is the crossover
index. Analysis of the scaling function Ψ behavior similar to that presented in the previous
section gives predictions about its asymptotic behavior:
ln Ψ(x) ∼
x1/φ x > 0
0 x = 0
− ln(−x) x < 0
(20)
The average bound fraction (order parameter) is found from〈θ〉 = 1N∂ ln Ψ∂χ
:
26
〈θ〉 ∼
τ−1+1/φ τ > 0
Nφ−1 τ = 0
(Nτ)−1 τ < 0
(21)
Note that in contrast to the previous section, the choice of τ is such that positive τ
corresponds to adsorbed (condensed) phase; this convention facilitates the comparison to the
exactly solved continuum ideal chain model. The mean-field value φ = 1/2 is exact for ideal
Gaussian chain. For real interacting chains the index φ is evaluated from renormalization-
group calculations or from the numerical simulations data.
Crossover exponent φ. It is clear that the crossover exponent defines not only the near-
critical behavior but also the weak adsorption of asymptotically long chains. De Gennes
[131] originally proposed that φ = 1 − ν ≈ 0.41, where ν = 0.5878 is the Flory exponent.
This was based on the assumption that the density profile of a critically adsorbed chain is
essentially the profile of an unperturbed coil cut in the middle by a fictitious plane, as is
the case in ideal chain adsorption where, indeed, φ = 1 − ν = 1/2. It was soon recognized
that the relation is not valid for adsorption of real chains at a solid substrate, although
it was suggested that it may still hold for polymer adsorption at liquid-liquid interface
[132]. From a theoretical point of view, the mean-field value of φ = 1/2 applies to higher-
dimensional spaces with d ≥ 4 where excluded volume effects are negligible. On the other
hand, conformal-invariance approach [133] in d = 2 also gives the exact value of φ = 1/2
rather unexpectedly. It was also found that φ = 1/2 applies rigorously to adsorption of lattice
animals (randomly branched self-avoiding polymers) [134, 135]. This prompted an exciting
suggestion that φ = 1/2 is a superuniversal index. However, field-theoretical calculations
for d = 3 based on two different methods (ε-expansion with Pade-Borel summation [136],
and massive theory in fixed 3d space [136, 137] seem to contradict the superuniversality
hypothesis giving the estimates of φ = 0.483 and φ = 0.52 , respectively. In any case,
theoretical considerations strongly suggest that φ must be at least quite close to 1/2.
Numerical studies using different Monte-Carlo algorithms were employed to attack the
problem of adsorption (see [138] with references). A traditional way of evaluating φ from
numerical data involves plotting various scaled chain characteristics vs. the finite-size scaling
parameter τNφ in hope of obtaining the data collapse on two universal curves corresponding
27
to adsorbed and desorbed regimes. In this method, χc and φmust be adjusted simultaneously
until the best data collapse is achieved. Earlier studies produced estimates φ = 0.588± 0.03
[36] and φ = 0.53± 0.007 [139]. It was shown that a slight change in the estimated χc leads
to large effect on the estimate for φ: parameter pairs (χc = 1.01, φ = 0.59) or (χc = 0.98,
φ = 0.50) give almost the same overall quality of the data collapse [140]. A more refined
analysis was suggested by van Rensburg and Rechnitzer [141] who studied adsorption of
SAWs in 2d and 3d obtaining extremely accurate data for N up to 300. They demonstrated
that analysis of the specific heat data is fraught with difficulty. A more robust procedure
is based on studying the energy ratios En+1/En for walks of different lengths. The values
reported were φ = 0.501 ± 0.015 (in accordance with the exact theoretical value 1/2) and
χc = 0.565 ± 0.010 for the square lattice and φ = 0.5005 ± 0.0036 and χc = 0.288 ± 0.020
for the cubic lattice.
All the works mentioned above started with the scaling crossover ansatz (19). Explicit
corrections to scaling were introduced by Grassberger [142] in his analysis of the exten-
sive numerical data obtained by a chain growth Monte-Carlo algorithm (pruned-enriched
Rosenbluth method for self-avoiding walks on simple cubic lattice with adsorbing surface)
for N up to 8000. It was shown that corrections to scaling persist even for very large
N beyond 1000, masking the true value of φ which was estimated to be somewhat below
1/2: φ = 0.484 ± 0.002 ; χc = 0.28567 ± 0.0001. An earlier estimate obtained by Hegger
and Grassberger [130] (φ = 0.496 ± 0.004 ; χc = 0.286 ± 0.003) and consistent with super
universality hypothesis seems to be just an apparent value for N ≤ 2000.
It is worth mentioning some results for off-lattice models. MC simulation of an off-lattice
model with N up to 512 demonstrated that the heat capacity extrapolated to N → ∞
experiences a finite jump which is consistent with the mean-field value of φ = 0.5 [143]. MC
simulations of a bead-spring model reported φ = 0.5± 0.02 [144].
One can summarize that the exponent φ is quite close to φ = 0.5 although its exact
value is still being debated. Very importantly, this means that the thermodynamics of real
self-avoiding polymers is to a large extent described by ideal chain models. Quantitatively,
excluded volume effects lead to a shift in the transition point, a change in the numerical
values of the entropy of a strongly adsorbed state etc., but these effects are of the same
nature as the difference between various lattice and off-lattice ideal chain models. This is
in contrast to geometrical characteristics (the size of the chain in the lateral and normal
28
directions) which are governed by the Flory index ν for 2d and 3d cases, respectively, and
therefore quite different for ideal and real chains. There is, therefore, a strong motivation to
consider adsorption of ideal chains in more detail and in particular, finite-size effects. For
lattice model, finite-size effects can be evaluated only by numerical calculations. The only
exactly solved model that includes N explicitly is the adsorption of a continuum Gaussian
chain model.
3. Continuum model of adsorption of an ideal chain.
Exact Green’s function. Chain configurations in the continuum model are described
as Brownian trajectories in the presence of a planar surface situated at z = 0. The total
statistical weight, G0(z, n), of all configurations for a walk of n steps (of unit rms step length)
starting at the surface and ending at position z satisfies the Edwards diffusion equation [145]:
∂G0
∂n=
1
6
∂2G0
∂z2(22)
The initial condition for a grafted chain is G0(0, z) = δ(z). Interaction with the surface is
introduced through a boundary condition G−10 ∂G0/∂z|z=0 = −c , equivalent to a pseudopo-
tential.
Subscript 0 refers to the absence of an external force and will be useful for the later
discussion. The critical point corresponds to c = 0, while the adsorption region to c > 0;
in the latter case, c−1 is the average thickness of the adsorbed layer. For strong adsorption
(large positive c) this thickness becomes smaller than the step length, which is unrealistic
but inherent to the continuum model neglecting the finite monomer size.
The solution of (22) is well known [35, 36] and may be written as:
G0(z,N, c) =(R√π)−1
e−z2/4R2 [
1 + cR√π Y (z/2R− cR)
](23)
Here R =√N/6 is the rms gyration radius and Y (x) = exp(x2)erfc(x), where erfc(x) is
the complimentary error function. Integrating G0 over the position z of the free end gives
the partition function:
Q0(N, c) = Y (−cR) (24)
29
The average number of adsorbed segments in the continuum model is 〈m〉 = ∂ lnQ0/∂c.
Partition function as realization of crossover ansatz. The asymptotes of the partition
function are:
Q0(N, c) = Y (−x) ≈
2ex
2x� 1
1− 2x/√π |x| � 1
(√πx)−1 −x� 1
(25)
Comparing these results with the general crossover scaling ansatz, we see that the argu-
ment x = c√N/6 , of the partition function in the continuum model, is nothing but the
crossover scaling parameter τN1/2 for Gaussian chain. Thus the scaling function Ψ(τN1/2
)is obtained explicitly as Y (−cR). This is a unique example of the scaling crossover func-
tion known exactly. All the strong and weak points of the continuum model stem from
this fact. In particular, the free energy per monomer in the adsorption region (cR � 1)
is βµads = −N−1 lnY (−cR) ≈ c2/6 and the average fraction of surface contacts 〈θ〉 = c/3
grows linearly without any saturation. Saturation effects in the strong adsorption regime
necessarily involve corrections to the crossover scaling.
Comparison with the Landau theory. For a Gaussian chain adsorbing onto a planar
surface, the Landau free energy will be a function of the fraction of adsorbed segments
θ = m/N . The partition function for a given fraction of contacts P (θ) was obtained in [146]
and has the form
P (θ) =
(πN
6
)−1/2
exp
(−Ncθ − 3
2Nθ2
)(26)
This leads to an exact expression for the Landau free energy
Φ(θ) = −N−1 lnP = Φ0 − cθ +3
2θ2 (27)
Note that the Landau free energy as a function of the adsorption order parameter is valid
not only in the thermodynamic limit, but for finite chains as well and is in complete analogy
with the coil-globule transition discussed in the previous section, see Eq. (15).
Mapping of continuum and lattice parameters. The continuum model is exactly solv-
able and allows analytical solutions for any chain length but it has a few inherent drawbacks.
For large values of c the fraction of adsorbed segments shows unlimited growth. There are
30
no explicit temperature effects because the boundary condition corresponds to an effective
pseudopotential. A lattice model in which a chain is described as a walk on a regular lattice
does not suffer from these deficiencies and takes into account many features of real poly-
mers and sorbents except for the interaction of monomers distant along the chain contour.
However exact analytical solutions for the partition function of lattice models exist only in
the infinite chain limit. There is a simple approach [147, 148] that allows to incorporate
finite-size effects into lattice models and to obtain analytical expressions for the partition
function Q(N, T ) .
The basic idea is to enforce the exact correspondence between the asymptotic free energy
expressions in the lattice and continuum models, µcont(c) = µads(χ). This gives the following
mapping between the interaction parameters in the continuum model and the chain on cubic
lattice:
c(χ) =
√
6 ln(
eχ
eχ−1
)− ln
(√4 + 1
eχ−1− 2)
χ > χcr
6 (e−χcr − e−χ) χ < χcr
(28)
Analytical partition function for finite lattice chains. The obtained mapping allows
one to write the partition function of a finite lattice chain in terms of the lattice interaction
parameter χ
Q0(N, c) = Y (−√N/6c(χ)) (29)
This leads to closed form analytical expressions for the equilibrium average characteristics
of finite chains as functions of temperature. In the strong adsorption limit the correct lattice
behavior is guaranteed automatically, while in the vicinity of the critical point the scaling
crossover ansatz is recovered. The two branches (28) match at the critical adsorption point
together with their first derivatives. Higher derivatives are discontinuous, which results
in a small jump of order N−1in the heat capacity at the critical point . To avoid of these
discontinuities a more complicated N -dependent mapping can be used [148]. This procedure
can be considered as an implicit introduction of sub-leading corrections to non-analytic
scaling function at the critical point. Figure 6 shows how the average bound fraction < θ >,
and the heat capacity C for finite lattice chains in the absence of a force depend on χ/χcr.
The solid curves give analytical results for N = 100, 300, and for the N →∞ limit.
31
<Θ>
aΧ
Χcr
12 3
1 2 3 4
0.2
0.4
0.6
0.8
Χ
Χcrb
1
2
3
C
1 2 3 4
0.05
0.10
0.15
0.20
0.25
Figure 6: The average bound fraction < θ >(a), and heat capacity C (b), for chains on a 6-
choice lattice in the absence of an external force as a function of χ/χc, for N = 100, 300, and
the N → ∞ limit (curves 1, 2, and 3, correspondingly). The symbols are the exact numerical
results.The critical adsorption point for chain on six-choice cubic lattice is χc = ln (6/5) w 0.182.
A small discontinuity of order N−1appears in the analytical curves for heat capacity at the critical
point which is an artifact of the mapping.
Figure 6 demonstrate clearly how the second-order adsorption transition develops when
the chain length is varied. In the thermodynamic limit the heat capacity shows a finite jump
at the transition, which is in contrast to the diverging behavior seen in most second-order
transitions for low-molar-mass systems[149]. A finite heat-capacity jump is predicted by the
Landau mean-field theory [41], and adsorption of a polymer chain is one of the few examples
(along with the transition to a superconducting state [150]) which is perfectly consistent
with this framework.
32
4. Adsorption on curved surfaces
Ideal chain. Exact solutions for the grand canonical partition function were obtained
for adsorption of an ideal lattice chain onto a thin thread [34], and on a cylinder of finite
radius R [37]. It turns out that the free energy has an essential singularity at the critical
point of adsorption so that all the derivatives of the order parameter vanish at this point.
In particular, the order parameter in the adsorbed phase vanishes upon approaching the
critical temperature as
〈θ〉 ∼ exp(−b/τ)
where b is a model-dependent numerical constant. According to the Ehrenfest classification
this indicates a phase transition of “infinite order”. The most general result based on a
continuum Gaussian model was presented by Eisenriegler et al[151] for adsorption on a
“generalized cylinder”. This object is embedded in the space of dimension d, has an infinitely
extened axis of dimension d − d1 and a curved surface of constant curvature radius R in
the co-dimension d1. An ordinary cylinder obtains for d = 3, d1 = 2, a normal spherical
object - for d = d1 = 3. A closed-form grand canonical partition function is expressed in
terms of the modified Bessel functions Kα and Kα+1 with α = d1/2−1. Due to separation of
degrees of freedom for a Gaussian chain, the value of the embedding space dimension , d, is
irrelevant. It follows that the scaling ansatz 19 generally holds with the crossover exponent
given by
φ = |d1 − 2|/2 (30)
for 1 ≤ d1 < 4 and d1 6= 2. For d1 = 4 the order parameter at the critical point scales as
〈θ〉 ∼ N/ lnN , while for d1 > 4 one obtains φ = 1 implying a first-order transition. It is
clear that the cylinder case is also special, and formally results in φ = 0. The true finite-size
scaling at the critical point is, however, logarithmic: 〈θ〉 ∼ lnN/N .
It was also proved[152][153] that the adsorption of a Gaussian chain onto a rigid rod
is strictly equivalent to the unzipping transition of two flexible Gaussian chains A and B
if monomer s of chain A can only interact with monomer s of chain B reflecting the key-
lock principle of complementary pairs in natural DNA. This can be shown by a standard
transformsation familiar from the two-body problem reduction in classical mechanics: r(s) =
rA(s)−rB(s), rCM(s) = 12
(rA(s) + rB(s)) and recognizing that the center of mass coordinates
rCM(s) and the relative coordinates r(s) are decoupled due to the special properties of
33
Gaussian chains.
Chain with excluded volume interactions. General scaling results for adsorption
of a self-avoiding walk on “generalized cylinders” were obtained by Hanke [153] by using
renormalization group arguments in conjunction with available results for quantum field
theories with curved boundaries. An estimate for the crossover exponent in the case of a
thin rigid rod in 3-dimensional space was obtained by assuming that φis a smooth function
in the (d, d1)plane. The crossover exponent is known on two lines in this plane, namely
the Gaussian line d = 4 where Eq.30applies and the so-called “marginal” line at which the
unperturbed self-avoiding walk intersects with the generalized cylinder at a manifold of zero
dimension, and as a result, one expects φ = 0. Linear interpolation gives the estimated
value φ ' 1/6 at the point (d = 3, d1 = 2) representing the rigid rod.
III. FIRST ORDER PHASE TRANSITIONS WITH PHASE COEXISTENCE
WITHIN A SINGLE CHAIN.
A. General remarks. In this section we present a unique model which allows exact
analytical investigation of first-order transitions with finite-size effects: a polymer chain
end-grafted to a solid adsorbing substrate with a normal force applied to the free end. The
transition is referred to as mechanical desorption and may be described in two different
statistical ensembles [154]. In the force ensemble the external end-force is fixed while the
fluctuating position z of the free end. The force ensemble with only one extensive variable
N conforms to the traditional analysis of phase transitions. The first-order transition in the
force ensemble has some unusual features: order parameter fluctuations and the heat capac-
ity demonstrate an anomalous pre-transitional growth according to a power law, metastable
states are completely absent, and instead of a bimodal curve, the distribution of the order
parameter has a flat region which becomes more pronounced with increasing chain length.
We demonstrate also that the origin of this anomaly lies in phase segregated configurations
providing the dominant contribution in the transition region. Coexistence of the adsorbed
and the stretched parts within the same macromolecule is quite peculiar due to the absence
of an excess surface energy at the boundary (this boundary consists of one segment only).
Near the transition point, strong fluctuations arise since all the conformations with different
phase composition have approximately the same free energy. To demonstrate clearly the
34
intra-chain coexistence of adsorbed and stretched phases we analyze mechanical desorption
in the conjugate z-ensemble and discuss an analogy with the gas-liquid transition.
A. Mechanical desorption of a macromolecule in f-ensemble.
The first analytical theory for mechanical desorption of a grafted polymer chain from a
solid surface by external end-force was constructed a long time ago [155]. It was based on an
ideal lattice model and treated only the infinite chain length limit. Later, a theory for finite
chains within a continuum Gaussian model was proposed [63, 146, 156] and subsequently
used to interpret AFM experiments [157]. The response of an adsorbed polymer chain
to a pulling force with a lateral component was also analyzed [14]. Temperature effects
and the reentrant nature of the phase diagram in the force-temperature variables at fixed
adsorption energy were discussed in [158, 159]; Monte-Carlo simulations were performed
for mechanical desorption of an off-lattice chain with excluded volume interactions [143,
160, 161]. The authors also proposed a theory incorporating excluded volume effects in the
limit of asymptotically long chains. Importantly, the unusual features of the mechanical
desorption mentioned above are completely unaffected by excluded volume interactions.
The Green’s function and the partition function. This is a generalization of the
standard adsorption-desorption problem discussed in Section II. A constant force is applied
to the free end of the chain in the direction normal to the surface. Within the continuum
model of ideal Gaussian chain the short-range interaction with the surface is described by the
adsorption parameter c (the pseudopotential amplitude). The Green’s function G0(z,N, c)
of the macromolecule with one end fixed at height z is given by Eq.(23). The presence of the
external end-force f changes the statistical weight of every configuration by the Boltzmann
factor ezf , where the temperature T is taken as unity and incorporated into f . Positive f
corresponds to stretching the chain while negative force means pressing the free chain end
down to the surface. The force-modified Green’s function becomes:
G(z,N, c, f) = G0(z,N, c)ezf =(R√π)−1
ezf−z2/4R2 [
1 + cR√π Y (z/2R− cR)
](31)
Integrating G over the position of the free end gives the partition function :
35
Q(c, f) =cY (−cR)− fY (−fR)
c− f(32)
The average height of the free end is 〈z〉 = ∂ lnQ/∂f . The average number of adsorbed
segments is 〈m〉 = ∂ lnQ/∂c.
Adsorption-force symmetry and the phase diagram. The partition function of an
adsorbing chain with the external field is remarkably symmetric with respect to an inter-
change of the adsorption parameter c and the external force f . Hence all the moments of the
distribution functions for the number of adsorbed segments P (m|c, f) and for the height of
the chain end P (z|c, f), as well as the functions themselves, coincide under the interchange
of the conjugated parameters c↔ f [156] . In particular, a restricted partition function for
configurations with a given number of adsorbed segments, m, follows from Eq.(31)
P (m|c, f) =(πR2
)−1/2ecm−(m/2R)2
[1 + fR
√π Y (m/2R− fR)
](33)
The fraction of adsorbed segments θ = m/N is the standard choice for the order parameter
in the adsorption problem [63]. A second order parameter conjugated to the end-force can
be defined as the stretching degree ζ = z/N . In the thermodynamic limit N →∞ one can
speak of a definite phase state for the macromolecule.
Possible states of the chain for various values of the adsorption interaction parameter c
and the force f are expressed by the phase diagram, Figure 7. Phase diagram in terms of
the parameters f and c is remarkably symmetric.The solid line c = f for positive f and c in
the phase diagram is the first-order transition line between adsorbed and stretched states.
The vertical dashed line at c = 0, f < 0 and the horizontal dashed line at f = 0, c < 0
correspond to second-order transitions involving single - loop conformation. The point at
u = 0, c = 0 is the bicritical point.Typical conformations of a chain in stretched phase , and
in adsorbed phase are shown on Figure 7.
It is clear from the diagram that the classical problem of adsorption with no force applied
corresponds to moving along the horizontal dashed line of second order transitions through
the bicritical point. An equivalent dual situation is realized when the adsorption parameter is
fixed at c = 0 and the force applied is changed from pressing down to stretching. Adsorption
of a chain with the end pressed down (constant f < 0) is equivalent to stretching of a
chain anchored to an inert or repulsing surface, and is also a second order phase transition.
36
ISOTROPIC
PHASE
ADSORBED
PHASE
Second order transitions
Second
order
transitions
First
order
transitions
c
f
Bicritical
point
STRETCHED PHASE
Figure 7: Phase diagram in terms of the parameters f and c. The solid line for positive f and c
in the phase diagram is the first-order transition line between adsorbed and stretched states. The
vertical dashed line at c = 0, f < 0 and the horizontal dashed line at f = 0, c < 0 correspond
to second-order transitions involving single - loop conformation. The point at f = 0, c = 0 is the
bicritical point.Typical conformations of a chain are shown in stretched phase (a), and in adsorbed
phase (b).
Adsorption of a chain under constant stretching force f > 0 applied is equivalent to tearing
off an adsorbed chain under the condition of c = const.
Partition function around the transition line f = c. The coexistence line can be
crossed by changing the force f , the adsorption parameter c, or both. Experimentally, a
natural way to induce desorption is to increase the force keeping the adsorption parameter
constant. The detachment point is then at ftr = c. For small deviations (f − ftr) and rela-
tively large total free energy of a pure phase at the transition line, F ∗ = −Nc2/6 = −Nf 2/6
, |F ∗| � 1 , the partition function in the vicinity of the transition can be approximated as
Q = 2cec
2R − fef2R
c− f= Qtr e
t sinh t
t(34)
where t = τ |F ∗|, τ = (f/c) − 1 is relative deviation from the transition point, and
Qtr = −4F ∗ exp(−F ∗) is the partition function exactly at the coexistence line [159]. In
this approximation, F ∗ is to be treated as a constant and all the dependence on the control
parameters f and c resides in the universal parameter t which is negative on the adsorption
37
side and positive on the stretched side of the line.
It follows from (34) that the change of the reduced order parameter 〈ζ〉 /ζtr in the vicinity
of the transition is described by a universal function of a single parameter t:
〈ζ〉ζtr
= 1− t−1 + coth t, (35)
where ζtr = ftr/6.
Unconventional features of the first-order transition. As discussed in the Introduc-
tion, the standard picture of a first-order transition includes a bimodal distribution of the
order parameter with a corresponding two minima in the Landau free energy. Two dis-
tinct competing minima imply also that there is no pre-transitional growth of fluctuations.
Contrary to that, fluctuations in both order parameters in mechanical desorption display
power-law growth upon approaching the transition point. The approximate form of the par-
tition function, Eq.(34) predicts that the reduced mean-square fluctuations for both order
parameters are described by one and the same universal function of the composite parameter
t:
〈δζ2〉〈δζ2〉tr
=〈δθ2〉〈δθ2〉tr
= 3(sinh−2 t+ t−2) ≈
3t−2 |t| � 1
1− t2/5 |t| � 1, (36)
where t = (f/c−1)Mc2/6 and 〈δζ2〉 = 〈δθ2〉 = 13(ctr/6)2 [159]. The mean-square fluctuations
of the adsorption order parameter are directly related to the heat capacity which is commonly
discussed in more conventional phase transitions. The heat capacity per segment is C/N =
(βε)N 〈δθ2〉 where ε is the energy per contact. A way to express the c parameter in terms of
βε was discussed above: this is a smooth function not affecting the anomalous growth near
the transition point. It follows from (34) that the heat capacity demonstrates a power-law
growth with index α = 2 . This is in contrast to the δ-peak singularity in the heat capacity
typical for “normal” first-order transitions. Figure 8 displays the reduced heat capacity per
segment (βε)−2C/N calculated from the full partition function for several chain lengths
N =500, 750, 1000 at fixed adsorption strength c = 1 as a function of force f .
Anomalous fluctuation growth is consistent with a unimodal distribution function of the
order parameter that broadens upon approaching the transition. Figure 9 displays the
distribution of the stretching order parameter near and exactly at the transition point.
38
N<∆Θ 2 >
f1
3
2
0.85 0.90 0.95 1.00 1.05 1.10 1.15
2000
4000
6000
8000
10 000
Figure 8: Fluctuations of the adsorption order parameter N⟨δθ2⟩vs. force f for lattice ideal
chains at fixed adsorption parameter with ftr = c = 1 for several chain lengths N = 500, 750,
1000 (curves 1, 2, and 3, respectively). The dashed line corresponds to the power-law asymptotics
C ∼ (f − ftr)−2.
P(z)
z
f =0.99 c
f =1.01 cf = c
0 100 200 300 4000.0
0.1
0.2
0.3
0.4
Figure 9: Distribution of the free-end position of a lattice ideal chain with length N = 1000 at
external end-force around the transition point ftr = c = 1
Unimodality in the order parameter distribution and pre-transitional fluctuation growth
are normally associated with continuous second-order transitions.
Landau free energy. The most important insight about the nature of the transition comes
from the detailed analysis of the Landau free energy as a function of the order parameter.
For simplicity we consider only the stretching order parameter, ζ = z/N . Due to the
symmetry discussed above the same results apply to the fraction of contacts, θ The Landau
free energy per segment is Φ(ζ) = N−1 lnG(z, c, f). In the thermodynamic limit N → ∞,
the function Φ(ζ) consists of two branches:
39
Ζ
F
f<c
f=c
f<c
0.2 0.4 0.6 0.8 1.0
-0.2
-0.1
0.0
0.1
0.2
0.3
Figure 10: Landau free energy per segment Φ as a function of the stretching order parameter
ζ = z/N at mechanical desorption of a Gaussian chain under external end-force. Three cases are
shown: the value of the force smaller then adsorption interaction parameter c (low curve); the force
is equal to c (middle curve); and the force is stronger then adsorption interaction (upper curve). In
two last cases the dependence Φ (ζ) is nonanalytic.
Φ(ζ) =
(c− f)ζ ζ ≤ c/3
(c− f)ζ + 32(ζ − c/3)2 ζ ≥ c/3
(37)
The function Φ(ζ) is explicitly non-analytic since the second derivative is discontinuous
at the junction point. It is clear from Figure 10 that the Landau free energy has always only
one minimum so that no bimodality in the order parameter distribution is ever present. At
the transition point itself, part of the Φ(ζ) curve is strictly flat, which brings about a finite
jump in the average order parameter. On the other hand, the gradual flattening of the Φ(ζ)
curve in the vicinity of the transition leads also to anomalous pre-transitional fluctuation
growth. The origin of this discrepancy with the classical picture is due to configurations
that consist of two coexisting phases within one macromolecule but without any interfacial
contribution to the total free energy of phase-segregated state. In order to see this more
clearly we consider the polymer chain desorption in the conjugated ensemble, where the
position of the free chain end is fixed at height z above the plane while the reaction force
acting on the end monomer is fluctuating.
40
B. Mechanical desorption of a macromolecule in the z-ensemble.
In the previous sub-section we considered desorption of a single end-grafted chain by
applying an external force to its free end. We demonstrated that there is abrupt transition at
the critical value of the force with some very unusual features. To undestand more clearly the
nature of the transition we discuss the situation when the free end of the end-grafted chain
is fixed at some distance z above the adsorbing solid surface, and consider the equilibruim
characteristics of the chain as functions of z. We show below a certan critical distance the
chain can be viewed as comprised by two coexisting phases, namely, the adsorbed and the
stretched parts. Above the critical distance, the adsorbed phase disappears and the chain
exists in a pure stretched phase. It is important that in a rather broad range of z values the
two phases coexist and their local properties do not change.
Analogy with gas-liquid transition in (N,P, T ) and (N, V, T ) ensembles. It was
mentioned in the Introduction that the standard classification and the qualitative picture
of phase transitions commonly taken for granted applies only to statistical ensembles with
only one extensive variable, such as the Gibbs (N,P, T ) ensemble for fluids. Gas-liquid
condensation in the canonical (N, V, T ) ensemble displays some very unorthodox features if
common-wisdom criteria are applied without proper caution. We specifically analyze some
details of gas-liquid transition in the two conjugate ensembles using the simplest textbook
model of van der Waals fluid. We feel that this discussion is very instructive as it pro-
vides a simple analogy and enriches intuition for understanding single-macromolecule phase
transitions.
Gas-liquid transition is a classical example of a first order phase transition [27, 31, 67].
In the Gibbs ensemble, the phase state is uniquely defined by the two intensive parameters
P and T , independently of the sample size. At fixed P and T the Gibbs free energy is
proportional to the number of particles, N . In the P − T phase diagram the coexistence of
liquid and gas phases corresponds to a certain line Pcoex(T ), which terminates at the critical
point. Upon crossing the coexistence line volume, entropy, average energy, etc. change
abruptly. Each point on the coexistence line corresponds to many phase-segregated states
with different ratios of relative phase volumes. In order to study more closely the states with
phase coexistence canonical ensemble is particularly useful. At a fixed temperature and a
fixed number of particles, volume changes may also induce the liquid-gas phase transition,
41
Figure 11: Thermodynamic analogy between gas-liquid system and an adsorbed polymer chain at
several positions of the end. Coexistence of two phases under piston and inside a single shain are
shown in two situations. The difference in interfacial effects is dicussed in the text.
T=0.8Tc
a
V
Vc
P
Pc
2 4 6 8
0.2
0.4
0.6
0.8
1.0
1.2V
Vc
U
NTc
b
2 4 6 8
-3
-2
-1
1
Figure 12: (a) Reduced isothermal pressure P/Pc (a), and the average potential energy per particle
(b) as a function of reduced volume V/Vc are shown for the van der Waals fluid at fixed temperature
T/Tc = 0.8; coexistence (binodal) curve is shown by the dashed line.
but this transition looks quite different due a wide interval of phase coexistence. Following
an isotherm along the V axis one passes from a homogeneous liquid state at low V into the
region where liquid and saturated vapor coexist in equilibrium (Figures 11b, 11c).
Upon further increase in the volume the liquid phase disappears (Figure 11d) and then
the vapor becomes unsaturated (Figure 11e). For a real gas, the isotherm or the P (V )
dependence at a fixed temperature shows a horizontal step at P = Pcoex, which corresponds
to the coexistence range of the two phases Figure 12 a.
The average potential energy per particle is shown in Figure 12 b as a function of the vol-
ume (kinetic energy of 3/2kT is not included to make a closer comparison with the polymer
42
system where kinetic energies are typically omitted). Note that the average potential energy
per particle in a particular phase depends on the local density and has to be calculated
separately for the two phases. The global average is then found by weighting according to
the relative sizes of the phases.
In the region of phase coexistence, the fraction of particles belonging to the liquid phase
changes linearly as a function of volume. This leads to a linear portion in the range of
phase coexistence as shown in Figure 12 b. It is clear from Figure 12 that in the (N, V, T )
ensemble, the phase transition is not abrupt but extends over the wide interval of changes in
volume V . Jump-wise changes typical of the first order phase transitions are observed only
in the P ensemble, where the region of phase coexistence collapses into a single point. The
V (P ) isotherm is obtained via rotation (Figure 12 a) and shows an abrupt volume change
at P = Pcoex. At this pressure, abrupt changes in energy and entropy likewise occur and the
fraction of molecules existing in the condensed phase changes from unity to zero.
The canonical ensemble allows yet another setting for inducing the gas-liquid transition:
one can change the temperature of the sample under fixed volume conditions. Cooling an
initially homogeneous fluid one would observe the onset of phase segregation below certain
temperature, and phase coexistence will persist for all lower temperatures. Figure 13 a shows
the change in the average potential energy per particle with temperature at two values of
the volume V/Vc = 1; 3, where Vc is the critical volume.
Again, instead of a jump only a discontinuity in the slope is observed. Without proper
awareness of the ensemble involved one may be wrongly tempted to interpret this as a sign
of a second-order transition. However, this is just a hallmark of the underlying extended
phase coexistence. It is worth noting that in this setup the coexistence line Pcoex(T ) is
never crossed (but only followed) . The process of cooling a fluid at fixed volume can be
conveniently illustrated in the (T, n) diagram, Figure 13 b, n = V/Vc is the reduced average
density (n = 1 at dense packing). The binodal curve in the (T, n) diagram gives the boundary
of the region unstable with respect to phase segregation and consists of two branches joining
at the critical point. The branches describe the local densities of the coexisting liquid and
gas phases as functions of temperature. The path followed by the sample upon cooling at
fixed average density is indicated by arrows: for an initially homogeneous fluid the path is
vertical; upon reaching the binodal line the sample splits into two phases with different local
densities.
43
T
Tc
U
NTc
V = Vc
V = 3 Vca
0.6 0.7 0.8 0.9 1.1
-2.5
-2.0
-1.5
-1.0
-0.5
T
Tc
n
b
0.2 0.4 0.6 0.8 1.0
0.5
1.0
1.5
Figure 13: (a) The average potential energy per particle (in units of Tc) of the van der Waals fluid
vs. reduced temperature T/Tc at fixed volume V = Vc and V = 3Vc. (b) Paths followed by the
fluid upon cooling at fixed average density n: moving down the temperature axis, an initial state
of uniform density separates into two phases with different local densities described by the two
branches of the binodal line; dotted lines represent cooling at fixed critical volume (V = Vc), solid
lines - at V = 3Vc.
Interestingly, the shape of the 〈ε〉 vs. T curve for the van der Waals model remains
qualitatively the same even if the system actually passes through the critical point (see the
curve with V = Vc). In sharp contrast to the Gibbs ensemble, the temperature dependence
of the average energy〈ε〉 vs. T measured in the (N, V, T ) ensemble is completely insensitive
to the distance from the critical point, although other characteristics (e.g. compressibility)
would signal the onset of critical phenomena.
An advantage of the (N, V, T ) ensemble in studying phase coexistence is that if one
analyses, e.g., the local density profile instead of quantities averaged over the whole sample,
the inhomogeneity of the phase-segregated sample becomes obvious.
The partition function of the desorbing chain in the z-ensemble. The partition
function in the z-ensemble coincides with the Green’s function of the original continuum
model of adsorption:
Q(z,N, c) = G0(z,N, c) =(R√π)−1
e−z2/4R2 [
1 + cR√π Y (z/2R− cR)
](38)
The average reaction force, 〈f〉 = ∂ lnQ/∂z, at the end-monomer at point z and the
44
z
a
<f>
c =0.6
c =1
0 100 200 300 400 500
0.2
0.4
0.6
0.8
1.0
1.2
1.4
< E >
N
z
c =1
c =0.6
b
100 200 300 400 500
-0.4
-0.3
-0.2
-0.1
Figure 14: The average force (a), and the everage energy (b) of the ideal chain with N = 1000
vs. the distance z of the nongrafted end above the surface at two different values of the adsorption
interaction parameter c=0.6, and c = 1.
average bound fraction 〈θ〉 = ∂ lnQ/∂c for asymptotically long chains are simplified to:
〈f〉 =
c, z/N ≤ c/3
3z/N, z/N ≥ c/3(39)
〈θ〉 =m
N=
c3− z/N, z/N ≤ c/3
0, z/N ≥ c/3(40)
One can relate the bound fraction to the average potential energy of interaction with the
adsorbing surface per monomer, 〈E/N〉 = −ε 〈θ〉 , to be directly compared to Figure 14b
Evidence of phase coexistence. The fact that in the same range of z the average
reaction force is constant, similar to the constant pressure portion of the P (V ) isotherm, is
a strong indication that phase coexistence is involved. In the same interval of z values the
average energy per particle increases linearly in the same way it changes with V in the phase
segregated fluid. Both features are consequences of the change in the phase composition with
increasing z, see the cartoon of Figure 11 b.
In the adsorbed part of the chain the probability for a segment to be in contact with
the plane is not affected by the end coordinate z and remains equal to c/3; in the torn
off tail, this probability is zero. Weighting these two contributions with the relative phase
size results in the curve shown in Figure 14 b. Beyond ζ = z/N = c/3 the adsorbed part
disappears and the chain consists of a single stretched phase. Simultaneously, the average
force starts growing with further stretching. The free energy of the polymer in the range
ζ ≤ c/3 can be understood on the basis of the standard discussion of phase coexistence.
45
The total free energy is a sum of two contributions:
F (N, c, z) = Fads(N − n, c) + Fstr(n, z), (41)
where the first term Fads(N − n, c) = −(N − n)c2/6 describes the adsorbed part containing
(N − n) segments, and the second term Fstr(n, z) = 3z2/(2n) stands for the free energy of
the stretched tail of length n with no contacts with the plane. Both terms are written in the
dominant asymptotic form although finite-size corrections are available. The equilibrium
distribution of segments between the phases is determined by the condition that the two
chemical potentials be equal: µads = µstr with µads = −c2/6 and µstr = −(3/2)(z/n)2.
This imposes the local stretching parameter in the tail in terms of the adsorption strength,
〈ζtail〉 = c/3 irrespective of the value of z as long as phase coexistence persists. The condition
of mechanical equilibrium (the equality of the normal forces) gives 〈ftail〉 = c coinciding with
the equation of the coexistence line in the f -ensemble. Using the coexistence conditions one
obtains the total free energy of the phase-segregated state as
F (N, c, z) = −Nc2/6 + cz , 0 ≤ z ≤ Nc/3, (42)
the same as the asymptotic form obtained from the full partition function (Eq.32). This
form of the free energy in the z-ensemble gives rise to the linear branch of the Landau free
energy in the f -ensemble discussed in the previous section:
Φ = N−1(F (N, c, z)− fz) = −c2/6 + (c− f)ζ (43)
Note that the form of the total free energy (43) does not include any interfacial contribu-
tions since the boundary between the adsorbed and the stretched parts in the flexible chain
is just one segment. This is directly linked to the linear structure of the polymer chain and
is eventually responsible for the absence of bimodality and of metastable states.
Local order parameter. One can still feel very much confused when the curve of the
bound fraction 〈θ〉 vs. the end coordinate z, Figure 15, is recognized as the change in the
average order parameter with the control parameter of the ensemble. It goes against the
accepted conventions to associate this smooth curve with a first order transition.
The way out of the dilemma is to realize that the order parameters 〈ζ〉 and 〈θ〉 as
introduced in the context of the f -ensemble are defined for the chain as a whole and do
46
<Θ>
z
c =1
N = 1000 a
0 100 200 300 400 500
0.1
0.2
0.3
æææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææææ
æ
æ
æ
àààààààààààààààààààààààààààààààààààààààààààààààààààààààààààààààà
à
à
à
à
à
à
à
à
à
àààààààààà
ìììììììììììììììììììììììììììììììììì
ì
ì
ì
ì
ì
ì
ì
ì
ì
ì
ììììììììì
òòòòò
ò
ò
ò
ò
ò
ò
ò
ò
ò
ò
ò
òòòòòòòòò
ô
ô
ô
ôôôôôôôô
<Θ>loc
k
z=5z=100
z=200
z=300
z=400
b
0 200 400 600 800 10000.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
Figure 15: The average order parameter (average bound fraction) vs the end coordinate z of the
long chain at the fix (a), and the average local order parameter as a function of local place k
along the chain at several z (b). The value of chain length N = 1000, and the value of adsorption
interaction parameter c = 1 are the same for both cases.
not allow to distinguish between the two coexisting phases. However, it is possible to define
local order parameters 〈θ〉loc = d〈m(k)〉dk
and 〈ζ〉loc = d〈z(k)〉dk
where z(k) is the z-coordinate of
k-th segment and m(k) is the number of contacts in the sub-chain from the zeroth (grafted)
segment to the k-th segment. The average number of contacts in the sub-chain (0, k) depends
on the position zk of the k-th segment and can be obtained from the Green’s function
G(zk, k, c) given by Eq.(23). Then, the result 〈m〉 (k, zk) has to be averaged over zk with
the weights W (zk) = G(zk, k, c)G(zk, z, N − k, c) where the second term is the Green’s
function of the (N − k)-segment sub-chain with the two ends fixed at zk and z. For the
continuum Gaussian model this Green’s function is well-known and is given by Eq. (23).
Hence calculating the local order parameter profiles is a straightforward exercise. Figure 15
a,b compares the behavior of the global order parameter 〈θ〉 as a function of z for chain of
length N = 1000 at fixed adsorption parameter c = 1 with the series of local order parameter
profiles 〈θ〉loc (k) for the same system. The profiles are calculated at several values of z = 5,
100, 200, 300, 400.
It is clear that with the increase in z the boundary between the adsorbed and the stretched
phase moves closer to the grafted end but the local properties of the adsorbed part of the
chain do not change. Thus the gradual decrease in the global order parameter shown in
Figure 15 a does not contradict the fact that the transition is first order and reflects the
specifics of the z-ensemble allowing extended phase coexistence.
A major difference between the gas-liquid transition and the mechanical desorption is
47
related to the interfacial effects. It is the interfacial free energy that leads to metastable states
and the bimodal distribution of the order parameter in the gas-liquid transition. Tearing
off an adsorbed flexible macromolecule by one end is a unique example of phase coexistence
without interfacial effects. The boundary between the two phases is essentially nominal and
consists of one segment. As a result, in the force ensemble close to the detachment point
fluctuations involving vastly different conformations with different phase composition (from
purely adsorbed to completely torn off chain) without any barrier separating them. This is
the ultimate origin of large pre-transitional fluctuations and of unimodal order parameter
distribution. The argument based on phase coexistence without a physical interface is valid
not only for ideal chains but for chains with excluded volume interactions as well. That
is why the nature of mechanical desorption transition is qualitatively unaffected by these
interactions.
Comparison of two ensembles. Although mechanical desorption looks quite differently
in the f - and z-ensembles, the fundamental theorem of statistical mechanics stating that all
ensembles are equivalent in the thermodynamic limit, holds. Indeed, ensemble equivalence
means that the equations of state are the same irrespective of which set of parameters is
independently controlled and which are considered as a fluctuating response to be measured
or evaluated in the framework of statistical mechanics. For the system under consideration,
〈z〉 vs f represents the force-extension curve (equation of state) in the f - ensemble, while
〈f〉 vs z corresponds to z-ensemble. Ensemble equivalence would mean that one function is
the inverse of the other. Figure 16 displays the force-extension curves obtained in the two
ensembles for a finite but fairly long chain (N = 1000). It is clear that there is still some
difference between the curves but both tend to the simple limiting shape described by
〈zf〉Na
=
0 βf < c
βfa3
βf > c(44)
A more subtle consequence of the ensemble equivalence is that other thermodynamic
averages (e.g. the bound fraction〈θ〉 expressed in terms of z or 〈z〉 should also coincide.
This is indeed correct in the proper N →∞ limit [154].
Ensemble equivalence does not extend to the description of fluctuations. The fact that
fluctuations in the canonical and grand canonical ensembles are different is well recognized
[65]. The difference in the character of fluctuations for mechanical desorption in f - and
48
z/N ,<z>/N
f
<f>
z-ensemble
f-ensemble
0.1 0.2 0.3 0.4 0.5 0.6
0.6
0.8
1.2
1.4
1.6
1.8
Figure 16: Force-extension curves at mechanical desorption of a Gaussian chain in the f -ensemble
(〈z〉 vs f , dashed line) and in the z- ensemble ( 〈f〉 vs z , solid line) at fixed adsorption parameter
c = 1; N = 1000.
z-ensembles is quite dramatic as illustrated below.
The change in the distribution function for the number of contacts is shown for the case
of c = 1 and N = 500 in Figure 17. In the f -ensemble (Figure 17 a), the distribution is
anomalously broad exactly at the transition point f = 1 and much narrower both above
and below the transition (f = 0.9; 1.1). In the z-ensemble (Figure 17b), the values of z are
chosen to coincide with the average chain extension values〈z〉 of Figure 17a. In contrast to
the f -ensemble the width of the distribution does not change appreciably with the change
in z.
We conclude Section III by re-stating the main qualitative results. First, the notion
of phase coexistence withing a single macromoecule provides a close analogy between the
phenomenon of mechanical desorption and standard gas-liquid transition. It is exactly this
analogy that gives a natural explanation of the apparently different pictures observed in the
f- and z-ensembles. Whereas in the force ensemble coexistence is restricted to a line in the
phase diagram (or to a single transition point when one of the control parameters is fixed)
in the conjugate z-ensemble it can be observed in a wide range of change in z, similar to the
gas-liquid coexistence in the NVT ensemble. These ideas are not yet commonly accepted, see
for example the discussion in [160, 161]where the transition in both ensembles was referred to
as “dichotomic” without any phase coexistence. Unconventional behavior of all the common
thermodynamic functions in the z-ensemble can be quite confusing if one does not realize
that this is just a result of re-distribution of matter between the two coexisting phases: the
49
Figure 17: Distribution of the number of contacts during for mechanical desorption in the f -
ensemble (a) and in z- ensemble (b) at c = 1 and N = 500. The curves correspond to the three
values z = 1, 85, and 200 in the z-ensemble and to the same three values of 〈z〉 in the f -ensemble
at f = 0.9, 1.0, and 1.1, respectively.(From [154])
same effect is observed in the van der Waals fluid in the Helmholtz ensemble. If the order
parameter is defined as a local quantity this becomes unambiguously clear. Although the
basic analogy is qualitatively correct, mechanical desorption still possesses some very unusual
specific features that include pre-transitional divergence of fluctuations and the absence of
bimodal distributions. These are due to the absence of the interfacial free energy at the
boundary between the two phases.
IV. FIRST ORDER PHASE TRANSITIONS WITHOUT PHASE COEXISTENCE
WITHIN A SINGLE CHAIN.
In Section IV a we consider the conformational properties of an ideal polymer chain
tethered by one end at distance z0 near a solid adsorbing surface. The short-ranged interac-
tion between the segment and surface are describes by a pseudo-potential with adsorption
interaction parameter c. The exact expression for the partition function and the Landau
free energy are available for this model. A first-order conformational transition occurs from
homogeneous coil state to a nonhomogeneous state composed of a strongly stretched stem
and a pancake that collects the remaining adsorbed segments. We discuss the impossibility
of simultaneous coexistence between two states (phases) and the related abnormal behavior
50
Figure 18: Illustration of the systems of interest. A liquid-solid interface is a plane . An isolated
Gaussian chain is fixed by one end at a position z0 as indicated by the black dot, the other end is free.
When the grafting point is sufficiently from the solid surface the chain may take an unperturbed coil
conformation or an inhomogeneous flower-like conformation with a stretched stem and the adsorbed
crown (pancake) .(From [162])
in the microcanonical ensemble.
In Section IV b we consider a similar situation where the chain is end-tethered at some
distance from a penetrable interface (instead of a solid adsorbing surface). The interface is
modeled as an external potential of a Heaviside step-function form. Special attention will
be paid to exactly solvable model with an unusual phase transition: the rolling transition of
a polymer chain tethered exactly at the penetrable interface.
Escape transition of a polymer chain compressed between pistons is considered in Section
IV c.
A. Polymer chain tethered near an adsorbing solid surface.
Model and partition function. An isolated chain near end-fixed at some distance
from an adsorbing surface as schematically presented in Figure 18.
Experimentally, such a situation can be realized with the help of an atomic force micro-
scope (AFM). The chain would be chemically attached by one end to the AFM tip. One
has to ensure that the chain does not adsorb onto the tip which can be achieved by using a
separate probe particle glued to the tip and used as a grafting substrate. This chain is then
brought near the adsorbing surface. When the tip is still far from the adsorbing surface the
chain will be in a mushroom conformation. However, the free energy of a mushroom con-
formation is approximately the same as that of an unperturbed coil, and the simple model
51
illustrated in Figure adequately represents the experimental situation. With the AFM appa-
ratus it is possible to measure the force on the chain as soon as it takes the partially adsorbed
flower conformation. In effect, the flower bridges the gap between the tip or probe, and the
adsorbing surface. The force which is easily picked up by the AFM apparatus is expected
to be independent of the separation and is only a function of the adsorption parameter that
characterizes the affinity of the chain for the adsorbing surface. In AFM experiments it is
possible to control the time of contact or the time of close proximity of the chain to the
adsorbing surface. Therefore, one should expect to observe the hysteresis effects discussed
bellow in full glory.
The exact partition function Q(z0, N, c) of an ideal polymer chain was obtained in [162]
and has a form
Q(z0, N, c) = Qcoil(z0, N) +Qfl(z0, N, c) = erf(z0/2R) + e−z2/4R2
Y (z0/2R− cR) (45)
The two terms correspond to two distinct states of the macromolecule. The first one
describes the state of a weakly perturbed ideal coil, while the second describes an inhomo-
geneous conformation consisting of an adsorbed part and a stretched stem connecting the
grafting point and the adsorbing plane
The intersection of the two branches of the free energies gives the line of first-order
transitions. For z0 � R and c� 1/R the transition line is defined by
( z0
Na
)tr
=ac
6(46)
Along the transition line the average number of segments in the stem is N/2, i.e. exactly
one half of the chain. The remaining half constitutes the adsorbed pancake.
Landau function. It was suggested [163, 164] that the chain stretching can be used as
the order parameter. For the coil states, this parameter refers to the chain as a whole,
s = (z − z0)/Na . The maximum value of the order parameter in the coil state is achieved
when the free end is touching the surface - at z = 0. For the flower states, only the stem is
stretched, and the order parameter is defined as s = z0/na where n is the number segments
in stem. The following simple analytical expressions for the two branches of the Landau free
energy were obtained in thermodynamic limit:
52
s
F
c = 0.6 0.9
1.2
1.5a
-0.2 0.2 0.4 0.6 0.8 1.0
-0.05
0.05
0.10
0.15
s
F z0 /N = 0.4
0.3
0.2
0.15b
-0.2 0.2 0.4 0.6 0.8 1.0
-0.1
0.1
0.2
0.3
Figure 19: The Landau free energy Φ as a function of the order parameter s for Gaussian chain
at fixed reduced distance z0/N = 0.2 from the adsorbing solid surface at several values of the
adsorption parameter c (a), and for fixed adsorption parameter c = 1.2 and several values of the
reduced distance z0/N (b). N = 1000 in both cases.(According the work [64]).
Φ(s, c) =
32s2, s ≤ s0
32ss0 − c2
6
(1− s0
s
), s ≥ s0
s0 = z0/Na (47)
The two branches of the Landau function match each other at s = L/Na.
The Landau free energy Φ(s) calculated according to (Ref:eq:53) is presented in Figure
19 as a function of the order parameter. A set of typical curves Φ(s) for several values of the
adsorption parameter c at a fixed reduced distance z0/N = 0.2 (s0 = 0.2) is shown in Figure
19 a . The coil state branch and the point where two branches meet do not change with the
adsorption parameter c, while the flower state branch is clearly affected by it. With affinity
for the surface, the minimum in the flower branch becomes more pronounced.
In Figure 19 b the other control parameter, z0, is varied, and the adsorption parameter
is kept constant. Again, the Landau function of the coil state remains the same, but the
matching point is shifting with z0.
The binodal condition, when the two minima are equally deep, leads to equation (46)
found from the analysis of the partition function. The equation of the spinodal line has a
form(z0Na
)sp
= ac3. The barrier height counted from the coil state is ∆coil = 3z0/2N , and
corresponds to the free energy of stretching the chain Fstr. The barrier height with respect
to the flower minimum is given by the combination of the free energy of stretching and that
of adsorption
53
∆fl =
(√3z2
0/2Na2 − ca
√N/6
)=(√
Fstr −√Fads
)2
(48)
Along the spinodal line the average number of segments in stem is equal to N , which means
that the pancake just has not developed. We conclude that metastable flower state has more
than N/2 segments in stem.
Stretching force. The average stretching force acting at tethering point is given by relation:
f = Q−1e−3z20/2Na2
cY (z0/2R− cR) (49)
and in the thermodynamic limit of N →∞ and finite z0N
has a simple behavior:
f =
0, c < ccr
c, c > ccr
(50)
In the vicinity of the transition point the force the force decreases linearly with z0:
f ≈ ftr −Nf 2
tr
2
( z0
Na− ca
6
)(51)
The dependence of the force on the position of the fixed end for several values of the
adsorption parameter c is displayed in Figure 20 a. As the chain end is moved away from
the adsorbing surface, the force remains constant until we approach closely the binodal
distance.
At larger distances, the chain is effectively torn away from the adsorbing surface and
the force is practically zero. In the coordinates f vs z0/N used in the phase diagram and
Figure 20 a this results in a sharp drop of magnitude ∆f = c. The sharpness of the drop is
proportional to N(ftr)2 ∼ Nc. The description in terms of the Landau function allows us to
introduce the force associated with metastable states and demonstrate the hysteresis effect
on Figure 20b,c.
The influence of the excluded volume effects on the transition point and spinodal equation
was discussed in [164, 165] for a very similar situation of the escape transition, see subsection
C below. Due to the model equivalence established in [162] this applies equally to the chain
attached near an adsorbing surface considered here.The binodal and spinodal lines come
close to each other than in the case of ideal chains. The region where metastable flowers
54
Figure 20: (a) The force f needed to keep the tethered chain with the end at z0 versus the reduced
distance z0/N . The sign of the force is in reality negative because the chain is attracted to the
surface. The number of segments is N = 1000 and the c parameter is varied as indicated. (b) The
force versus the adsorption parameter c for various values of the chain length N as indicated and
for z0/N = 0.2. The dashed lines with the arrows represent the hysteresis effects. (c) Example of
a reduced force as a function of the reduced distance z0/N is presented for c = 0.6 and N = 1000,
with special attention to the hysteresis effect indicated by the dashed lines with the arrows.(From
[162])55
exist is relatively smaller, since the adsorbed part at the coexistence line comprises only
2N/5 segments.
Microcanonical ensemble and the entropy gap.
A very unusual feature of the first order transition described above is that, at the tran-
sition point, the two states cannot coexist simultaneously due to their very nature. Both
of them, the coil and the “flower” (stem + adsorbed pancake) are essentially defined for a
macromolecule as a whole, but not for constituent sub-chains. It is therefore impossible
to construct the mixed state in which both phases coexist in arbitrary proportions. This
peculiarity is closely linked to some abnormal thermodynamic properties of the same system
considered now under conditions of the microcanonical ensemble. The role of the appropri-
ate thermodynamic potential would be played by the entropy considered as a function of
energy and number of segments, S(E,N). In the ideal chain model the energy is only due
to the contacts made by chain segments with the adsorbing surface, E = −cm, where m is
the number of contacts. The entropy can be found from the restricted free energy calculated
under condition that the number of contacts is fixed. The corresponding partition function
is represented as a sum [63]
Q(N, c,m, z0) =
N̂
m
Q(k, c,m, 0)G(N − k, 0, z0)dk (52)
where the first term in the integrand represents the partition function of the adsorbed
pancake consisting of k segments with both ends in contact with the surface, and the second
term is the standard Green’s function of the stem with two ends at z = 0 and z = z0,
correspondingly. The restricted partition function Q(k, c,m, 0) is found from Eq. 33 by
taking the negative (pressing down) force. Evaluation of the integral by steepest descent
method yields the following asymptotic result for the restricted free energy F (c,N,m) =
−Ncm + 3/2m2 and S(c,N,m, s0) = −Ncm + 3/2m2 for the entropy as a function of the
number of contacts. All of the above considerations do not apply to a special case of zero
energy (no contacts with the surface). Asymptotically, the entropy of such a state is just
zero since a Gaussian coil serves as a reference state. Figure 21a displays the entropy per
segment as a function of the number of contacts, m, for a chain of length N = 500 at two
different grafting distances. The special state with m = 0 is indicated by a big dot. A very
unusual feature of the entropy as a function of energy is that it is not just non-analytic
56
but discontinuous. One can say that the coil state with no contacts is separated from all
the other states by an entropy gap. The magnitude of the gap increases with the reduced
grafting distance z0Na
. We stress that the standard jump in the entropy associated with
a first order transition in the canonical ensemble (where the entropy is a first derivative
of the thermodynamic potential) should not be confused with the discontinuous behavior
shown in Figure21since the microcanonical entropy is the thermodynamic potential by itself.
Physically the origin of the entropy gap is linked to the polymeric nature of the system and
to the effect of grafting: in order to form the first contact with the surface, the chain has to
deform anf lose configurational entropy.
The second unconventional feature of the microcanonical entropy with far-reaching con-
sequences is that S(E) is not convex everywhere. Figure 21b shows S(E) together with
its convex hull. Concavity of the microcanonical entropy will lead to inequivalence of the
microcanonical and canonical ensembles. The states corresponding to the entropy branch
lying below the convex hull are not thermodynamically stable in the canonical ensemble
(they are only metastable). In the canonical ensemble the system would undergo a jump
between the two states indicated by dots connected by a dashed segment.
Ensemble inequivalence will be disussed in more detail below in the subsection dealing
with the escape transition. There the pair of conjugate ensembles (analogs of NVT and
NPT ensembles) is more closely related to experimentally accessible situations; however, the
underlying physical reason for this inequivalence lies in the existence of two configuration
subsets separated by an entropy gap.
B. Polymer chain end-tethered near a penetrable interface (near step potential)
A liquid-liquid interface is modeled as a Heaviside step function of the external dimen-
sionless (reduced by kT ) potential u(z): it assumes a value u > 0 for positive z and zero
otherwise. An isolated Gaussian chain is fixed with one of its ends at a position z0. When
the grafting point is at any negative z0or at positive z0 far enough from the interface the
chain is approximately in a Gaussian conformation indicated by the big spheres. When the
grafting end is situated in the high potential region (z0 > 0)and close enough to the interface
a flower may form consisting of a stretched stem and a coiled crown, as presented in Figure
22. Comparing Figures 18 and 22one can clearly see the similarity between the two models.
57
ææ
ææææææææææææææææææææææææææææææææææææææææææææææææææ
àààààààààààààààààààààààààààààààààààààààààààààààààà
m
S
z0 = 250
z0 = 150
a
0 10 20 30 40 50 60
-0.5
-0.4
-0.3
-0.2
-0.1
0.0
0.1
æ
S
Ε
z0
N= 0.3, c = 1
b
-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
0.2
Figure 21: (a) The entropy per segment S as a function of the number of contacts, m, for a chain of
length N = 500 at two different grafting distances z0above the solid surface at contact interaction
parameter c = 1.The values of z0 are shown on Fugure. The special state with m = 0 is indicated
by a big dot. (b) The microcanonical entropyS(E) at the fixed values z0/N = 0.3, and c = 1. The
states lying below the convex hull (below dash line) are metastable.
This similarity can be put on a firm mathematical basis [162] by comparing the Landau free
energy given by Eq. 53 and the corresponding Landau free energy of the chain grafted near
the step potential [166]:
Φ(s, u) =
32s2 + u, s ≤ s0
32ss0 + u s0
s, s ≥ s0
(53)
It follows immediately that all the results for the binodal and spinodal lines of the previous
subsection apply with the change ca =√
6u , namely(z0Na
)tr
=√
u6and
(z0Na
)sp
=√
2u3.
The phase diagram for lattice chain with finite extensibility is presented in Figure 23 while
the average fraction of contacts for the lattice chain with N = 200 vs the reduced grafting
distance z0/N for several values for the external potential, u, demonstrating the first order
transition is displayed in Figure 24.
1. Rolling transition: Chain attached at a liquid-liquid interface.
We consider now the special situation when a polymer chain attached directly at a liquid-
liquid penetrable interface at z0 = 0. In this case the coil is sitting with most of its segments
in the region where the potential u is most favorable. It does not leave the interface as it is
grafted with one of its ends to the boundary .
58
Figure 22: A liquid-liquid interface is modeled as a Heaviside step function of the external potential
u(z): it assumes a value u for positive z and zero otherwise. An isolated Gaussian chain is fixed
with one of its ends at a position z0 as indicated by the black dot. When the grafting point is at any
negative z or at a very positive z the chain is approximately in a Gaussian conformation indicated
by the big spheres. When the grafting end is at z0 > 0 and near the interface, a flower may form
which consists of a stem from of stretching blobs and a crown as perturbed coil; the dashed spheres
are drawn to help understand the flower structure.(From [166])
Figure 23: Phase diagram for chain end-tethered at distance z0 near step-potential u. The nor-
malized distance z0/N is plotted against the square root of the potential field. The binodal line is
shown as solid and the spinodal is displayed as a dotted line. Regions of stability and metastability
of the coil as well as the flower conformations are indicated. In the thermodynamic limit the second
spinodal line coincides with the abscissa and not indicated.(From [166]
59
Figure 24: The average fraction of contacts numerically evaluated for the lattice model withN = 200
plotted against the reduced grafting distance z0/N for several values for the external potential felt
by each segment of the polymer chain, u, as indicated.(From [166])
If in addition the molecular nature of the components that are responsible for the presence
of the interface is ignored, and replaced by an artificial external potential field felt by the
polymer units, one arrives at a model which is in fact quite general. In the following it is not
necessary to specify whether the step in the external potential has an entropic origin e.g.,
when the interface is the boundary between two polymer gels which differ only in polymer
density and not in chemical composition or is enthalpic in nature e.g., in a liquid/liquid
interface.
The exact partition function for this model was obained in [167, 168], and has a form:
Q(u,N, ) = e−uN/2 I0(uN/2), (54)
where I0(x) is the modified Bessel function. The free energy, which is counted from that of
an unrestricted Gaussian chain taken as a reference point, is given by
F (u) = − lnQ(u) ≈
uN2
[1− uN/8 + (uN/8)3] , uN � 1
12
[ln(πuN)− 1/(2uN)− 1/(2uN)2] , uN � 1(55)
From Eq. (55) is may seem odd that the free energy only grows logarithmically with N
for large uN . However, one should keep in mind that Eq. (55) is only the correction of the
free energy on the reference state of Gaussian coil which is proportional with N . The free
energy is symmetric, F (uN) = F (−uN).
60
Figure 25: (a) Average fraction of segments 〈s〉 in the negative half-space, (b) the fluctuations in
the fraction of segments in the negative half-space, as a function of the external potential u for
Gaussian chain attached at a liquid-liquid penetrable interphase . The degree of polymerization is
indicated.(From [167]).
The free energy is continuous at u = 0, but the derivative with respect to u is not. According
to the classification of Ehrenfest this discontinuity indicates that the system passes a first-
order phase transition at u = 0 - see Figure 25.
The thermodynamic quantities of the system: the full free energy, entropy, energy are
considered in [167].The behavior of the Landau free energy at rolling transition was discussed
in [168]. From the analysis of the free energy it was determined that the transition is of
the first order type. The energy, however, remains continuous at the transition, but its first
derivative jumps. This is indicative of a second-order phase transition. Finally, the entropy
in the system, and its derivatives, remain continuous at u = 0. From the exact analytical
equation for the Landau function it is found that there are no metastable states associated
with the transition . In addition the structure of the coil does not change dramatically during
this process (the end-point distributions remain a function with a single maximum).Roughly
speaking, when the step potential changes its sign, the polymer coil rolls from one region to
the other. These features show that the system discussed above is rather special.
C. Escape transition.
In this section we consider a phase transition in a single macromolecule that has received
much attention recently, namely the escape transition observed for an end-tethered chain
compressed between two pistons of circular cross section - Figure 26. At weak compressions,
the chain is deformed uniformly forming a relatively thick pancake. In this “coiled state” the
61
Figure 26: Typical snapshots showing a configuration of an imprisoned (a) and an escaped (b)
polymer chain of length N = 200, grafted under a cylindrical piston of radius L = 40 at hight
D = 12(a), and D = 4(b) over the substrate. A repulsive Lennard-Jones potential , smeared on all
surface , makes them impenetrable for the chain.The piston edge is rounded so as to preserve the
continuety of the force when a bead moves around it.(From [164]).
resistance force due to the compressed chain increases monotonously as the distance between
two pistons, D, decreases. Beyond certain critical compression, the chain conformation
changes abruptly to an inhomogeneous “escaped state” consisting of a “stem” (a stretched
string extending from the anchoring site to the piston border) and a “tail” outside of the
confining pistons. The resistance force decreases abruptly, indicating a first-order transition.
There is a very close relation between the escape transition and the conformational phase
transition of a polymer chain grafted at some distance from a solid adsorbed surface con-
sidered in the previous section. In both cases one state (the coil or the compressed coil)
is homogeneous. The other phase is inhomogeneous and consists of a stretched stem and
the tail (adsorbed or escaped). There are several reasons to discuss the escape transition in
some detail.
First, the setting of the escape transition is close to what is currently experimentally
feasible. One can envision a force spectroscopy experiment, wherein a single polymer chain
of contour length of ∼ 100− 200 nm and of gyration radius Rg ∼ 10− 15 nm is tethered to
an inert, nonadsorbing surface and compressed by the probe tip with the radius of curvature
of ∼ 20 − 50 nm. In this example it is relevant that the radius of curvature is larger then
the radius of gyration. Currently, it is difficult to make a tip in the form of a cylindrical
piston with a flat surface; however, the model of a macromolecule compressed by a piston
62
includes all physical attributes of a real system.
Second, the escape transition allows considerable modifications in geometry without af-
fecting qualitatively the underlying physics. From a theoretical point of view, an intriguing
question is whether the transition persists for chains with excluded volume interactions in 2d
space where the compressed state is not coiled but already stretched into a quasi-1d string
[165]. A similar situation occurs when a 3d chain is trapped inside a tube of diameter D .
This setting is very relevant to biophysical applications including DNA fragments confined
inside membrane channels . For both cases analytical theory can be created which is in good
agreement with the simulation results.
Last, but not least, the escape transition can be considered theoretically and (at least
in principle) realized experimentally in two conjugate ensembles. It was demonstrated
[154, 164, 169] that in the D-ensemble where the distance D between two pistons serves
as an independent parameter the force-separation curve has a loop with a negative com-
pressibility region. In contrast to that, in the conjugate f -ensemble the force-separation
curve is monotonic with a flat portion resembling the normal gas-liquid coexistence . We
analyze both situations and demonstrate that the escape transition is a unique example of
a highly unconventional first order transition with negative compressibility region and non-
equivalence of conjugate ensembles which persists even in the thermodynamic limit. We
address the general question of nonequivalence of statistical ensembles and the underlying
physical reasons leading to this nonequivalence in the case of a single polymer chain.
Escape transition in 3d geometry. In this section we present an analytical theory of the
escape transition for Gaussian chains in two conjugate D- and f - ensembles, demonstrate
qualitative differences in the behavior of thermodynamic properties of the system in these
two ensembles and discuss the origin of this nonequivalence. We analyze the origin of the
negative compressibility region as obtained from the equilibrium canonical partition function.
We also visualize the metastable and the coexisting states in escape transition on the base of
Landau approach. In the conclusion we discuss the escape transition in chains with excluded
volume interactions in good solvent.
Escape transition in 3d dimension has been studied by scaling arguments [170, 171], via
numerical methods [157, 172], and by computer simulations for good solvents and theta
solvent conditions [173–175] .
Numerical methods have been used to study the effect of the curvature of pistons
63
[176, 177], adsorption of a chain on piston walls [178], and compression of a star shaped
macromolecule [179] and a molecule based on a diblock copolymer [180]. A rigorous ana-
lytical theory addressing equilibrium and kinetic aspects of an ideal chain compressed by
pistons has been constructed in [162, 163]. The effects of metastability and negative com-
pressibility and the behavior of a chain in the conjugated ensembles have been described
in [169, 181, 182]. The theory taking into account the excluded volume between monomer
units has been compared with the Monte Carlo calculations and presented in [165]. We start
from the main result of the rigorous analytical theory of escape transition for a Gaussian
chain .
Partition function for ideal chain in D-ensemble. The full partition function in D-
ensemble was obtained in [163] and has a form
Q(D,N) = Q1 +Q2 = erf
(L
2R
)exp
[−(πR2
D
)]+ exp
(−πLD
)[1− erf
(L
2R− πR
D
)](56)
The first term takes into account all coil configurations, when the chain is confined in the
inter-piston volume; the second describes escaped configurations. Two asymptotic branches
of the free energy F (D,L,N) = − lnQ(D,L,N) are:
F (D,L,N)
N≈
π2
6D2 , D > Dtr
πLDN
, D < Dtr
(57)
The phase transition point occurs at the compression distance Dtr which depends on the
ratio of the contour length of the chain N to the piston radius L:
Dtr =πN
6L, (58)
At the transition point the tail escaping outside the inter-piston volume comprises exactly
1/2 of the chain.
As discussed later, both branches of the free energy can be continued into the metastable
regions. One of them terminates at the spinodal point D = D∗, the other persists down to
arbitrary small D (which is a specific feature of the phantom Gaussian chain). Figure 27
shows the Helmholtz free energy F as a function of the separation distance D for L = 90;
64
0
50
100
150
200
3 4 5 6 7 8
D
F
Figure 27: Free energy of a Gaussian chain at escape transition as a function of the piston separation,
D, for a cylindrical piston’s radiusL = 90, and chain length N = 800. The equilibrium branches
are shown by solid lines and the metastable branches by dotted lines. At the transition point, the
free energy has a discontinuity in the slope. The region where F (D) is not concave is limited by
two triangles.(According to [169])
N = 800. The equilibrium branches are shown by solid lines and the metastable branches
by dotted lines.
The free energy curve F (D) has a very unconventional feature: is not concave within
the range of 33/25 < D/Dtr < 32/23 . This contradicts a general statement of statistical
mechanics on the concavity of thermodynamic potentials. In the standard discussion of
statistical mechanics, a non-concave region must be replaced with the help of the double-
tangent construction that corresponds to phase coexistence and lowers the total free energy.
However, in the present case phase coexistence is impossible due to the very nature of the
phases, as discussed later. Here we just stress that the free energy displayed in Figure 27 is
obtained from a complete analytical partition function that includes all possible configura-
tions. Indeed, in the previous section concerning mechanical desorption we were able to see
that phase-segregated states can be naturally included in the exact partition function.
Compression curve (average force vs. separation). The average compression force
〈f〉 can be obtained by differentiating the free energy F (D,L,N) with respect to D. The
two asymptotic branches are given by:
65
0
10
20
30
40
50
0 2 4 6 8 10
<f>
D
1
2
3
Figure 28: Average compression force < f > for Gaussian chain compressed under piston
with radius L as a function of compression distance D for (L;N) = 40; 400(1), 80; 800(2), and
160; 1600(3).(According to [169])
β 〈f〉 ≈
πLD2 , D < Dtr
π2Na2
3D3 , D > Dtr
(59)
With progressive squeezing, the average force jumps down from 72L3/πN2a4 in the impris-
oned state to one half of this at the escaped state.
In Figure 28 the average compression force < f > vs. compression distance D is given
for three sets of parameters (L;N) = (40, 400); (80; 800), and (160; 1600) at fixed ratio
L/N = 0.1. At fixed ratio L/N , the transition becomes sharper with increasing chain
length while the region of negative compressibility narrows down and eventually shrinks to
a point as the limit N → ∞ is approached. We stress that negative compressibility is a
strictly equilibrium result that follows from the exact partition function and is not related
to unstable states.
Conjugate ensemble (separation vs. force). The force ensemble wherein the external
compression force plays the role of the independent variable is conjugate to the D-ensemble.
The partition function
66
Qf =
∞̂
0
QD exp(−fD) dD (60)
in the f -ensemble is the Laplace transform of the partition function QD. The transition
point in the f -ensemble is ftr = 16π
(43
)4 L3
N2 . The Gibbs free energy is globally concave as a
function of the compression force f . The average compression distance 〈D〉 = −∂ lnQf/∂f
has two asymptotic branches
〈D〉 ≈
(π2N3f
)1/3
, f < ftr(πLf
)1/2
, f > ftr
(61)
and experiences a downward jump at f = ftr .The region of negative compressibility is
absent.
Non-equivalence of ensembles. To compare the above two ensembles we present on
Figure29 reduced average compression force 〈f〉 / 〈f〉tr vs reduced distance D/Dtr as well
reduced compression force f/ftr vs. reduced average distance 〈D〉 / 〈D〉tr . In all cases, the
ration of piston radius to chain contour length, L/na, is kept constant, while the number
of chain units, N , and the piston radius, L, are varied. For small system with N = 100,
D = 15 (Figure 29 a) the phase transition is very wide and in the force ensemble it is
hardly identifiable. At N = 1600, D = 240 (Figure 29 b) the phase transition becomes well
pronounced in both ensembles and demonstrate the different behavior. In D-ensemble this
transition is accompanied by abrupt change f vs 〈D〉, while 〈f〉 vs D has the appearance of
a van der Waals loop. In contrast to a liquid-gas transition, all points of this loop correspond
to the equilibrium states of the system. Therefore the negative compressibility close to the
transition point is not related to any metastable effects. As follows from Figure 29, the
difference between the force-strain curves in two conjugated ensembles do not decrease with
N but become even more pronounced. This scenario disagrees with well-known theorem of
statistical mechanics that predict that pressure is a monotonic function of volume [154, 164].
To understand more clearly the situation of nonequivalence of two ensembles we note that
the curve of the free energy from compression distance F (D) is convex exactly in the region
where nonequivalence occurs - see Figure 29 and Figure 27. A general theorem of statistical
mechanics states that a convex region of the free energy signals that a homogeneous state
is unstable with respect to phase segregation. The convex region is then replaced by the
67
0,6
0,8
1
1,2
1,4
0,7 0,8 0,9 1 1,1 1,2 1,3 1,4
af / f*< f >/ f*
N=100L=15
<D>/D*D/D* 0,6
0,8
1
1,2
1,4
0,7 0,8 0,9 1 1,1 1,2 1,3 1,4
вf / f*< f >/ f*
< D>/D*D/D*
N=1600L=240
Figure 29: Average separation < D > vs. force f (solid lines) and average force < f >vs. separation
D (dash lines) for Gaussian chain compressed under piston with radius L in two conjugate ensembles
forN = 100;L = 15 (a) andN = 1600;L = 240 (b). Nonequivalence of the two conjugate ensembles
becomes more pronounced as system size is increased. (According to [154])
standard double-tangent construction that lowers the total free energy and represents phase-
segregated states. However, in the escape transition simultaneous coexistence of the two
phases is impossible due to the nature of the states. Both of them, the compressed coil and
the partially escaped “flower” (stem + tail) are essentially defined for a macromolecule as a
whole, but not for constituent sub-chains. It is therefore impossible to construct the mixed
state, in which both phases coexist in arbitrary proportions. This is the main difference with
the typical low-molecular mass systems and with the macromolecule undergoing mechanical
desorption, see Section III.
In the f -ensemble the Laplace transform of Eq. (60) automatically creates the convex
envelope of the F (D) function which is effectively equivalent to the double-tangent construc-
tion: hence the flat portion of the f vs.〈D〉 curve in Figure 29. We stress that the results
discussed are not artifacts of the simple Gaussian model. Their origin is quite general and
is eventually linked to the “entropy gap” discussed in Section 4.1. Indeed, Monte-Carlo
[174, 175] and Molecular Dynamics [164] simulations of the escape transition in a chain with
excluded volume interactions demonstrated qualitatively the same negative compressibility
loop in the D-ensemble. Anomalous behavior is also independent of other details such as the
exact position of the grafting point, the geometry and the precise alignment of the piston
68
(within certain limit).
One might ask whether examples of ensemble non-equivalence are known in other ar-
eas of statistical physics. Indeed, self-gravitating systems also fall outside the ensemble-
equivalence class and the difference between microcanonical and canonical pictures of finite
self-gravitating clusters has been discussed recently [183]. Experimental verification of this
effect is, however, at present hardly within reach. On the other hand, experimental study
of the escape transition is feasible and would be highly warranted.
Ensemble equivalence is rigorously established for a broad class of systems composed of
indistinguishable particles with the interaction vanishing at infinity. Existence of a well-
defined thermodynamic limit with a finite energy per particle is also required. All known
examples in the physics of condensed low molecular weight matter seem to fall into this
class. The last condition is not satisfied in self-gravitating systems. A single polymer chain,
considered as a statistical system, allows taking a proper thermodynamic limit in which
the total free energy is extensive. However, the first condition concerning the nature of
interactions is violated. The topological connectivity of linear chain is provided by bonding
interactions that do not satisfy the conditions stipulated above. This does not mean that
phase transitions in the single chain should always violate the ensemble equivalence; rather
that these are promising candidates for research in the present direction. The details of
intrachain interactions seem to be irrelevant since the ensemble non-equivalence is present
even for ideal Gaussian chain. On the other hand, we note that end-grafting is essential in
setting the stage for the escape transition. At the fundamental level, grafting removes the
translational degrees of freedom as well as the more subtle translational invariance along the
chain contour.
Order parameter and Landau function. It was suggested in [162, 164] that the param-
eter that characterizes the chain stretching can serve as the order parameter of the escape
transition. For the imprisoned coil states, this parameter refers to the chain as a whole,
s = r/N where r is the end-to-end distance. For the partially escaped states, only the
stem is stretched, and the order parameter is defined as s = L/n, where n is the number
of segments in the stem. The stem is the sub-chain composed of n imprisoned segments,
starting from the fixed chain end and ending with the first segment that reaches the edge of
the pistons. The two definitions of the order parameter match smoothly for the border-line
configuration in which the stem includes all the segments with r = L and n = N . The
69
following simple analytical expressions for the two branches of the Landau free energy were
obtained in the asymptotic limit :
Φ(s,D) ≈
32s2 + 1
6
(πD
)2, s ≤ L
N
3L2Ns+ L
6N
(πD
)2s−1, s ≥ L
N
(62)
Comparing this result to Equations 47 and 53 it becomes obvious that the three models
discussed in this Section (chain tethered near an adsorbing surface, near a step potential, and
compressed between pistons) are equivalent in the thermodynamic limit with the following
mapping between parameters: z0 ↔ L; and ca ↔√
6u ↔ πaD
[162]. The shape of the
Landau free energy as a function of the order parameter is typical for standard first-order
transitions with two minima separated by a barrier located at s = L/N , similar to the curves
of Figure 19. This means that as far as the behavior of the average order parameter and
its fluctuations is concerned, the escape transition is perfectly normal. In the vicinity of
the transition metastable states are possible and can be easily analyzed from the shape of
the Landau free energy. In particular, the spinodal point where the escaped state becomes
unstable is given by the simple relation: Dsp/Dtr = 2.
We stress again a close analogy between the escape transition and the adsorption tran-
sition for a polymer chain end-fixed at a distance from the solid surface or a liquid-liquid
interface. Although these are two distinct phenomena that are quite different at first glance
their fundamental similarity is revealed in the shape of the Landau free energy. In the
asymptotic limit, there is a simple mapping between the three partition functions.
The new escaped (or partially adsorbed) phase emerges via a single nucleus when the
free end of the chain reaches the piston edge (or adsorbing surface/interface) and forms a
seed crown of a few or one segments; its appearance involves a global change in the chain
conformation with a finite jump in the entropy, see Figure 21. The nucleation barrier height
is determined by the entropy gap ∆ = 3L2/2Na2 (or 3z20/2Na
2) only.
Escape transition (as well as the two other eqivalent models discussed in the current
section) is remarkable since not only the exact partition functions with finite-size effects
are known for them but also the exact expressions for the Landau free energy, containing
all the information on the metastable states. It can also serve as a pedagogical example
allowing a simple visual interpretaion of a state at the spinodal point. In classical systems
undergoing first order phase transitions it is not easy to form a simple picture of spinodal
70
states. In contrast to this, spinodal conformation in the escape transition is just a uniformly
stretched stem that contains all the segments and a vanishingly small crown. More detailed
characteristics of the binodal and spinodal states were dicussed in [182].
The classical spinodal decomposition occurs via growth of long-wave fluctuations of the
order parameter. The escape transition, as well as the eqivalent models, is adequately
described in terms of a single global order parameter that allows one to construct a simple
theory for both equilibrium and kinetic aspects of this phenomenon. The nature of the
order parameter also dictates the basic mechanism leading to a decay of an unstable state.
Note that not all single-chain phase transitions are naturally described in terms of a global
order parameter. For instance, in the coil-globule transition the order parameter is the local
monomer density. Correspondingly, kinetics of the coil collapse is governed by growth and
coalescence of multiple nucleation centers, although its initial stage does not necessarily
involve any change in the large-scale coil conformation. For a discussion of local order
parameter, see also mechanical desorption in the z-ensemble, Section 3b.
Escape transition in chain with excluded volume. The theory of escape transition for
flexible polymer chain under good solvent conditions was developed in [164] using Landau
free energy approach based on the order parameter defined above. The approach allows a
straightforward incorporation of finite chain length effects. The main results obtained in
[164] are: 1) the line of first-order transitions is given by equation
Dtr = A(N/L)ν/(1−ν) (63)
where the numerical pre-factorA = 0.834 is linked to the non-universal pre-factors in the con-
finement free energy and the lateral end-to-end distance and was evaluated for the Kremer-
Grest model. 2)The spinodal line where the barrier maintaining a metastable state vanishes
is Dsp/Dtr = (4/3)ν/(1−ν). The second spinodal equation is L/N = 0 in the limit N → ∞.
3)The average order parameter in the transition point has a jump from zero to 4L3Na
, while
the average fraction of imprisoned monomers jumps from unity to 34. Thus at the transition
point the escaped tail always comprises 14of the total monomers irrespective to the L/Na
ratio.
A comprehensive study of both static and dynamic properties using Molecular Dynamic
simulations for Kremer-Grest bead-spring model demonstrated good agreement with quan-
titative theoretical predictions. Therefore we can speculate that all the unusual features of
71
the escape transition (as well as the two other eqivalent models ) qualitatively apply to real
polymer chains.
Escape transition in 2d space. The escape transition for a two-dimensional polymer with
excluded volume interactions was analyzed in [165, 184] using the model of self-avoiding
walks on a square lattice. The inter-piston volume becomes a strip in the 2d-setting. Self-
avoidance of the chain confined inside a strip means that the confined coil conformation is
transformed into a linear string of blobs. Common-sense intuition suggests that with the
progressive compression of the chain the blob size decreases and the position of the free end
moves monotonically towards to the edge of a strip. Upon further compression the extra
monomers are squeezed out of the strip and form a coiled tail outside. No phase transition
occurs in this “toothpaste” scenario.
Both simulation results and the analytical theory indicate however that in the thermody-
namic limit of large N and L but finite L/N , there is a weak first-order transition with small
but perfectly identifiable jump the order parameter and in the average number if imprisoned
units.
Simulations and scaling arguments show that the free energy has two distinct branches
depending on whether the chain is completely confined of part of it forms the escaped tail.
The best fit for the free energy as evaluated in simulations is given by
F (N,L,D) =
Fimp ≈ 1.944(2)ND−4/3 D > Dtr
Fesc ≈ 2.03(3)L/D D < Dtr
(64)
The two braches join at transition point Dtr ≈ 0.889(Na/L)3.There is a discontinuity in
the slope when F is plotted as a function of D. The fraction of segments that are escaping
the strip at the transition point is quite small: (Nimp − Nesc)/N = 1 − 3/25/3, instead of14in 3d case (see previous section).The accurate estimation of these values needs precise
calculations for chain length up to 50000 with the help of the pruned-enriched Rosenbluth
method. Since two distinct phases separated by a high barrier are involved introducing a
special bias is required: otherwise, misleading spurious results are very likely [184].
The physical reasons of a weak first-order escape transition in 2d case as opposed to the
smooth “toothpaste” scenario are quite subtle. Indeed, the form of the free energy given by
Eq. 64 suggest that there is also a discontinuity in the slope when the free energy is consid-
ered as a function of L. The derivative ∂F/∂L has the meaning of the average force applied
72
to the fixed chain end. For the fully confined state this force is zero and the blob picture
is applicable without reservations. However, for conformations that include a tail outside
the strip the force at the grafting point is non-zero. Thus the escaped tail is equivalent to a
constant stretching force and therefore, the confined stem experiences additional stretching,
as illustrated in the cartoon Figure 30. The naïve blob picture in the “toothpaste” scenario
fails to take this additional stretching into account while a more rigorous approach based
on the Landau free energy incorporates it automatically and correctly predicts a jump-wise
transition. The blobs in the stem are deformed due to the tail effect as compared to the
blobs in the fully imprisoned state.
Another escape transition setup: Dragging a polymer chain into a tube. A
situation when a 3d chain is confined in a tube of diameter D is qualitatively similar to the
2d escape. An important qualitative result of the analytical theory of the escape transition
is that the transition can be induced by changing any of the three parameters: the chain
length, N,the tube diameter, D, and the distance between the fixed end and the tube
opening. In the case of a tube, changing D (i.e. squeezing the chain) may not be easy to
arrange experimentally unless the tube is made of a soft compressible material. Changing
N would imply a process in which the chain is grown slowly in a controlled way within
the restricted tube geometry. Experimentally this may be feasible although quite difficult
to realize. Finally, the distance between the end-monomer inside the tube and the tube
opening can be changed. This can be realized by slowly dragging a polymer chain by one
end into a narrow tube e.g., by using optical tweezers. Well calibrated nanochennels were
produced recently in fused silica substrates by lithography methods with the width in the
range of 30 to 400 nm, which were used to study the confinement of single λ-phage DNA
molecules driven electrophoretically into these nanochennels [185]. The persistence length
of DNA under conditions used in these experiments is about 50 nm while its contour length
was about 1000 times larger. This means that except for the case of the narrowest channels
DNA behaved essentially as a long flexible macromolecule on the relevant length scales.
Cartoons of a flexible polymer chain with one end dragged into a nanotube in a quasi-
static process are shown in Figure 30. The chain end inside the tube is characterized by its
coordinate x counted from the tube opening. It experiences a reaction force f that balances
the pull of the undeformed swollen coil outside the tube. At the transition point x∗, the
remaining tail is sucked into the tube abruptly by a uniform shrinking of all blobs in the
73
Figure 30: Schematic drawings of a flexible chain confined partially (above) and fully (below) in a
nanotube.In the first case, there is an ejecting force due to the outside tail, and external force is
needed to keep the chain from ejection; no force appears in second case.
stem, and the reaction force becomes zero . As long as the chain is fully confined in the
tube, no reaction force appears at the fixed chain end .
The properties of a single macromolecule confined in a tube have been studied extensively
for decades, both by analytical theory and by numerical simulations for various models of
flexible and semi-flexible chains [186–191] . For a fully imprisoned homogeneous state there
are scaling predictions concerning various chain characteristics which were tested by MC
simulations. We focus here in the nonhomogeneous state where the confined part of the
chain inside the tube forms a stretched stem and the free tail still in solution forms a coiled
crown. This type of conformations appears in a variety of situations including translocation
through a thick membrane [192, 193].
The results of the analysis are qualitatively similar to those of the 2d escape although
the Flory exponent involved is changed from ν2 = 3/4 to ν ≈ 0.587 . Again, there are two
branches of the free energy:
F (N,L,D) =
Fimp = 5.4ND−1/ν x > xtr
Fesc = 1.46 x/D x < xtr
(65)
where x is the coordinate of the chain end counted from the tube edge (the variable to
be changed in this setup) and the numerical prefactors refer to the walks on simple cubic
lattice. The transition occurs at the right end position
74
Figure 31: (a) MC data for the average fraction of imprisoned units, Nimp/N , plotted against
the reduced end coordinate x/x∗ for relatively long chains or various values of tube diameter D
at fixed number of blobs (nb = 60 ). The solid line shows the predicted linear growth below the
transition point. At the transition point x/x∗ = 1, Nimp/N jumps up from 0.76 to 1.(b) Nimp/N
vs the reduced end coordinate x/x∗ for tube diameter D = 21 and for different number of blobs
nb, displaying the rounding of the transition due to finite-size effects.(From [194])
xtrNa≈ 1.26 (D/a)1−1/ν (66)
At the transition point the outside tail comprises about 24% of all segments which is
much larger that the corresponding jump in the 2d geometry (5.5%) and remarkably close
to the jump in the traditional 3d setting (exactly 1/4).
The theory predicts that upon gradual dragging of the chain end further into the tube the
average fraction of imprisoned units, Nimp/N , plotted against the reduced end coordinate,
x/xtr, increases linearly before the transition point is reached and then jumps up from 0.76
to unity. These predictions are plotted on Figure 31a together with simulation data for four
different values of D. Another set of numerical data is shown on Figure 31b, displaying the
rounding of the transition due to the finite chain length.
The analysis of the Landau free energy allows identifying two spinodal points, see Figure
32 b. If the chain end is moved at a finite speed one would expect a standard hysteresis
loop illustrated in Figure 32 a. In that sense, the transition conforms to what is expected
of a standard first-order transition. However, the abnormality associated with the “entropy
gap” and the impossibility of simultaneous phase coexistence still exists and would show up
75
Figure 32: (a)The hysteresis loop of the average fraction of imprisoned units, Nimp/N , associated
with metastable states of a self-avoiding chain in a nanotube. (b) Landau free energy of the chain in
a nanotube of diameterD = 17 as a function of the order parameter at the transition point x = xtr
and at two spinodal points x(1)sp ≈ 0.76xtr and x(2)
sp ≈ 1.30xtr.(From [194])
as a non-concave region in the free energy, Eq. (65) considered as a function of D.
V. COMPLEX ZEROES OF PARTITION FUNCTIONS FOR SINGLE-CHAIN.
At a phase transition point, thermodynamic functions have a singularity, meaning that
they or their derivatives have a finite or infinite discontinuity. On the other hand the
partition function Q =∑
i exp(−Ei/kT ) is just a sum of exponentials and thus has no
singularities as a function of external parameters. Since the free energy is the logarithm
of the partition function, and the logarithmic function has a singularity at zero argument,
the only possibility left is that the partition function should vanish (or at least, in some
sense, tend to zero) when the system approaches the transition point. It is obvious though,
that the partition function is positive and cannot be zero at any real values of external
parameters. It was also realized very early that a true mathematical singularity of the free
energy can develop only in the thermodynamic limit when the number of particles tends
to infinity. The approach pioneered by Yang and Lee [42, 43] related these singularities to
complex zeroes of the partition function.
Their original papers dealt with the liquid-gas transition induced by the change in fugac-
ity, so these were the analytical properties of the grand partition function in the complex
plane of fugacity that were the object of investigation. Similarly, the temperature-induced
76
transitions should be described in terms of the zero distribution for the canonical partition
function in the complex plane of temperature (or β = 1/kT ). These are commonly called
Fisher zeros in order to distinguish them from the Yang-Lee zeroes in the fugacity plane.
Obviously, the general approach is applicable to phase transitions induced by changing any
other external parameter as well.
Yang and Lee showed that for finite N the partition function can have only complex
conjugated zeroes but no zeroes on the real positive axis. The only possibility for a phase
transition to appear is that as the number of particles increases, the complex zeroes come
closer to the real positive axis, and eventually, in the N → ∞ limit, they pinch upon the
real axis at the transition point. While the Yang-Lee theorem states that the zeroes in the
fugacity plane have to be located on a unit circle, there are no general results known for
Fisher zeroes. Empirical regularities show, however, that they tend to fall on smooth arcs
that cross the real axis at a certain angle. In the thermodynamic limit, the free energy and
its derivatives can be represented as integrals over the continuous distribution of zeroes that
could be characterized by some limiting density function. But the problem of finding the
actual distribution of zeros for the partition function of even a simple model proves to be
formidable. In practice, for a given model the zeros are calculated numerically for small
samples and then some extrapolations are employed [61, 195].
There exists a phenomenological approach relating certain features of a phase transition
(amplitudes and critical indices) to the characteristics of the distribution of zeros assumed
to be known [60], as well as scaling predictions for this distribution [51].
Assuming that the zeroes concentrate on two symmetric support lines which cross the
real axis at the point β = βc making an angle ω with it, Grossmann and Rosenhauer [60]
were able present an extensive classification of phase transitions and to express the main
characteristics of a transition (jump magnitudes, critical indices and amplitudes) through
the parameters of the linear density of zeroes g(y), where y is the coordinate along the
imaginary axis. They assumed that the linear density of zeroes is a power function of the
coordinate y along the imaginary axis: g(y) ∼ y1−α. Using this assumption several cases
can be distinguished. Two of them are summarized below.
(1) If α = 1 and the density of zeroes g(y) tends to a constant at small y, then the zeroes
necessarily approach the real axis at a straight angle ω = π/2 and the energy has a finite
jump ∆E = 2πg(0) upon crossing the β = βc point. This obviously corresponds to a first
77
order transition with a δ-peak singularity in the specific heat.
(2) If α = 0 , the density of zeroes grows linearly with y , and the support lines cross
the real axis at ω = π/4, the energy is continuous but the specific heat has a finite jump
discontinuity, as in a classical mean-field second order transition.
As we discussed in the Introduction there are two different types of the first-order phase
transitions in a single chain: transitions with phase coexistence and transitions without phase
coexistence. It is naturally to put forward the question: is it possible to discriminate these
transitions by analyses of complex zeroes of the partition functions these systems. Below
we present complex zeroes for two exactly solvable models. The first model is an ideal chain
desorbing from solid surface by external end force. It was demonstrated in Section III that
this model undergoes the first order transition with phase coexistence. The second model
is an ideal chain tethered be one end near adsorbing surface. This model is also exactly
solvable and undergoes the first order transition without phase coexistence as was discussed
in Section IV.
a) Complex zeroes in transition with phase coexistence (mechanical desorption of
ideal chain). In the model of Gaussian chain desorbing by external end-force , the governing
parameter c (or f) is analogous to the inverse temperature β, while the conjugated variable,
the number of adsorbed segments (or the free end height) would be the analog of the energy
E. The analysis of the partition function Q(c, f0) in the complex plane of the variable c
was presented in [63]. It was shown that the limiting density of zeros is g(ρ) = ρ/6π. At
the point of crossing the real axis ρ = f0, and g tends to the finite value g0 = f0/6π. This
falls into category 1) according to the above classification, and indeed, gives the first order
transition with a finite jump in the order parameter. The magnitude of the jump was derived
in ref. [60] to be 2πg0 = f0/3, in accordance with the value obtained by direct differentiation
of the partition function.
On the asymptotic wings of the hyperbola the distance from the transition point along
the curve, s, is the same as ρ, and the density of zeros increases linearly with it. Taking s ' ρ
and solving for ρ as a function of k one obtains the position of the k-th zero ρk ' (12πk/N)1/2
When one crosses the line of the first order transitions closer and closer to the bicritical
point the focus distance of the hyperbola, f0, decreases and so do accordingly the density
g0 and the magnitude of the jump in the order parameter. Eventually, at f0 = 0, the
curve degenerates into two straight lines at an angle ω = π/4 with the real axis crossing
78
it at the origin (see Figure 33) . Here, the transition becomes second order: changing the
adsorption parameter c one passes exactly through the bicritical point. Strictly speaking,
the asymptotic representation of the Y function used in Eq. (#) is not valid in this case
any more. However, the limiting picture of the distribution of zeros does apply.
Numerical results shown in Figure 33 suggest that zeroes indeed condense on a support
line at an angle ω = π/4 with the real axis. The limiting density of zeros has a form
g(|c|) = |c| /6π. This case falls into category 2 according to the above classification, and,
indeed, results in a second order transition of a mean-field type with a finite jump in the
order parameter fluctuation squared (see Section IV). The magnitude of the jump is 1/3.
Of course, this coincides with the result obtained directly by differentiating the free energy,
Eq. 24.
The limiting density and the slope of the line of zeros of the partition function upon
crossing the line of second order transitions away from the bicritical point (f0 = const < 0)
are exactly the same: ω = π/4, g(|c|) = |c| /6π. This is consistent with the fact that the
Landau free energies coincide for both cases in the thermodynamic limit (see section IV).
Numerical data for a few first zeros of the exact partition function (27) with N = 100 and
N = 500 for several values of f0 are displayed in Figure 33 and support the analytical results
for the distribution of zeros.
It is worth noting that in contrast to the lattice models where the partition function of a
finite system is a polynomial and therefore has only a finite number of zeros, the model of
the adsorbing chain that we discuss results in a partition function with an infinite number
of zeros for any value of N. The reason for this is that we deal with a continuum model as
opposed to lattice models with finite sets of discrete states.
b) Complex zeroes in transition without phase coexistence ( adsorption of ideal
chain tethered near a solid surface). The exact partition function Q(z0, c) of an ideal
chain tethered at the distance z0 near surface with adsorption interaction parameter was
presented before, Eq (45). The complex zeroes this partition function were analyzed in [64].
For zeros close to the real axis (i.e. with small χ) was obtained: χk = (2k+1)π6Ns2
.The closest
zero is characterized by the polar angle χ0 = (π/6N)(N/z)2. For a fixed value of the ratio
s = z/N and increasing N , χ0 approaches the real axis as N−1 . At distances of order
z ∼ N1/2 the deviation of polar angle from zero is on the order of unity and one certainly
cannot speak of a first order transition.
79
Figure 33: Distribution of zeros of the partition function for ideal polymer chain at mechanical
desorption under external end-force f in the complex plane of the adsorption parameter c for chain
length N = 100 (a), and N = 500 (b). Various values of the end force magnitude, f , indicated in
the plot. The analytical limiting curves are shown by solid lines.(From the work [63])
At the point of crossing the real axis (χ → 0 ; ρ → 6s) , the density of zeros tends
to a constant value g0 = s/2π. According to Grossman and Rosenhauer, this falls into
the category of first order transitions with a finite jump in the energy (for our adsorption
model- in the number of contacts). The magnitude of the jump must be 2πg0 = s . Direct
analysis of the partition function shows that the jump of the number of contacts is, indeed
〈m〉 /N = z/N .
On the asymptotic wings of the hyperbola χ→ π/4, the density of zeros increases linearly
80
Figure 34: The distribution of zeros of the partition function in the complex plane of the adsorption
parameter for ideal polymer chain tethered at the distance z0 near solid surface . The chain length
N = 400 and three values of the tether point z0/N = 0; 0.1; 0.2 are indicated.The lines are drawn
to guide the eye.(From the work [64])
with the distance from the transition point along the curve: g ' ρ/6π. Along these wings
the position of the k-th zero for large enough k, k/N � s2, is ρk ≈ (2πk/N)1/2.
When the distance from the adsorbing plane to the grafting point, z, decreases, so do
accordingly the density g0 and the magnitude of the jump in the order parameter. Eventually,
at z = 0, the curve degenerates into two straight lines at an angle ω = π/4 with the real
axis crossing it at the origin. The density of zeroes turns out to be a linear function of
the distance from the critical value c = 0. Here, the transition becomes second order, as
mentioned before The distribution of the partition function zeroes for chain with both end
fixed can be easily understood now. The transition is again second order with the critical
value of the adsorption parameter being c∗ = 3z/N .
Comparing Figures 33 and 34 one concludes that the distribution of the complex zeroes
does not allow to discriminate between transitions with and without phase coexistence.
CONCLUSIONS
We have considered several polymer models exhibiting phase transitions which admit
exact analytical treatment in the calculation of the partition function. Moreover, almost
all models allow us to obtain exact expression for the Landau free energy as a function of
the order parameter both for first- and second-order transitions. For some models, it is also
possible to study analytically the distribution of complex zeros of the partition function.
This rigorous treatment can be applied not only in the thermodynamic limit, but also to
systems of finite size. The fact that all these solutions are available in closed form is very
81
remarkable; they can serve as good examples for discussing various methods and approaches
in statistical physics and be a useful teaching aid.
Now, there is a natural question: why such a treatment is possible, and what distinguishes
these polymer models from the classical non-polymeric exactly solvable lattice models.
We can point to at least three special features of our polymer models. First, the interac-
tions between individual elements of the systems (the repeat units of the polymer chain) are
taken into account from the outset in writing the initial differential equation for the random
walk chain. By assumption, each unit interacts only with its closest neghbours, while there
are no long-range volume interactions. Second, successful treatment was possible when the
order parameter was not defined locally, but rather characterized the system as a whole.
Because of this, there are no correlation of fluctuations of the order parameter. Third, the
external field is applied to the free end of the chain, and does not affect the loops nor the
adsorbed parts of the molecule.
Almost all the models discussed display some unconventional behavior. Even the well-
studied coil-to-globule transition with its mean-field character becomes very unorthodox in
higher dimentions. For the mechanical desorption model features of first and second order
transitions become mixed due to phase coexistence not accompanied by interfacial free en-
ergy. Models belonging to the escape transition class represent a unique example of a highly
unconventional first order phase transition with several inter-related unusual features: no si-
multaneous phase coexistence, and hence, no phase boundary; non-concave thermodynamic
potential; non-equivalence of conjugate ensembles and negative compressibility region.
All the unconventional features of the phase transitions mentioned above apply not only
to the idealised models that admit closed-form exact solutions. Excluded volume interac-
tions don not change the basic picture which is rooted either in the fact that the interface
between the two phases within a single macromolecule consists of a single segment, or in
the entropy gap between the two classes of configurations. Thus the qualitative conclusions
about the unconventional features of the transitions should apply to real macromolecules.
In particular, the macromolecules capable of escape-class transitions are speculated to be
practically utilized as trigger systems .
Ensemble equivalence is rigorously established for a broad class of systems composed of
indistinguishable particles with the interaction vanishing at infinity. Existence of a well-
defined thermodynamic limit with a finite energy per particle is also required [1? ]. All
82
known examples in the physics of condensed non-polymeric matter seem to fall into this class.
However, a single polymer chain, considered as a statistical system, does not. The topological
connectivity of linear chains is provided by bonding interactions that do not satisfy the
conditions stipulated above. Of course, this does not mean that phase transitions in a
single chain should always violate the ensemble equivalence; rather that these are promising
candidates for research in the present direction. The details of intrachain interactions seem
to be irrelevant since the ensemble non-equivalence is present even for ideal Gaussian chains.
On the other hand, we note that end-grafting is essential in setting the stage for the escape
transition. At the fundamental level, grafting removes the translational degrees of freedom
as well as the more subtle translational invariance along the chain contour. Presumably
it is also important that one of the phases has to be nonhomogeneous and consist of two
mictrophases like stem and and crown in the escaped configuration.
Examples of ensemble non-equivalence are not completely unknown in other areas of sta-
tistical physics. Indeed, non-extensive self-gravitating systems also fall outside the ensemble-
equivalence class and the difference between microcanonical and canonical pictures of finite
clusters with long-range interactions has been discussed recently [183]. Experimental ver-
ification of this effect is, however, at present hardly within reach. On the other hand,
experimental study of the escape transition is feasible even now and would be highly war-
ranted.
Acknowledgments
We are grateful to Prof. Kurt Binder who conceived of the idea of this review and stim-
ulated us throughout this work by his deep insights and thoughtful discussions. Technical
help provided by Dr. A. Polotsky is greatly appreciated. This work is supported by DFG
grant 436 RUS 113/863/0-2 and by the Russian Foundation for Basic Research (RFBR
grants No.09-03-91344-a , and No.08-03-00402-a).
[1] M E Fisher. Lectures in Theoretical Physics. University of Colorado Press, Boulder, 1965.
[2] L Onsager. Phys. Rev., 65:117, 1944.
[3] G K Binnig, C F Quate, and Ch Gerber. Phys. Rev. Lett., 56:930, 1986.
83
[4] V T Moy, E L Florin, and H E Gaub. Science, .257:.266, 1994.
[5] H J Kreuzer and M Grunze. Europhys. Lett., 55:640, 2001.
[6] T Hugel, M Rief, M Seitz, H E Gaub, and R R Netz. Phys. Rev. Lett., 94:048301, 2005.
[7] E L Florin, V T Moy, and H E Gaub. Science, .264:.415, .1994.
[8] S B Smith, L. Finzi, and C Bustamante. Science, .258:.1122, 1992.
[9] M Rief, M Gautel, F Oesterhelt, J M Fernandez, and H E Gaub. Reversible Unfolding of
Individual Titin Immunoglobulin Domains by AFM. Science, 276:1109–1112, 1997.
[10] L Tskhovrebova, J Trinick, J A Sleep, and R M Simmons. Nature, 387:308, 1997.
[11] A D Mehta, M Rief, J A Spudich, R A Smith, and R M Simmons. Science, 283:1689, 1999.
[12] T R Strick, V Croquette, and D Bensimon. Nature, 404:901, 2000.
[13] A D Mehta, R S Rock, M Rief, J A Spudich, M S Mooseker, and R E Cheney. Nature,
400:590, 1999.
[14] A. Serr and R R Netz. Europhys. Lett., 73:292, 2006.
[15] B Essevaz-Roulet, U Bockelmann, and F Heslot. Proc. Nat. Acad. Sci. USA., 94:11935, 1997.
[16] U Bockelmann, Ph Thomen, B Essevaz-Roulet, V Viasnoff, and F Heslot. Biophys. J.,
.82:1537, 2002.
[17] U Dammer, M Hegner, D Anselmetti, P Wagner, M Dreier, W Huber, and H.-J. Gunterdort.
Biophys. J., 70:2437, 1996.
[18] P E Marszalek, A F Oberhauser, Y P Pang, and J M Fernandez. Nature, 396:661, 1998.
[19] M Kessler, K E Gottschalk, H Janovjak, D J Muller, and H E Gaub. J. Mol. Biol., 357:644,
2006.
[20] F Kuhner, M Erdmann, and H E Gaub. Phys. Rev. Lett., 97:21831, 2006.
[21] G Neuert, T Hugel, R R Netz, and H E Gaub. Macromolecules, 39:789, 2006.
[22] N B Holland, T Hugel, G Neuert, D Oesterhelt, L Moroder, M Seitz, and H E Gaub. Macro-
molecules, 36:2015, 2003.
[23] R J Baxter. Exactly Solved Models in Statistical Mechanics. Academic Press, 1982.
[24] C Domb and M S Green, editors. Phase Transitions and Critical Phenomena. Academic
Press, N.Y., 1972.
[25] M Kac, G E Uhlenbeck, and Hemmer P C. J. Math. Phys., 4:216, 1963.
[26] T H Berlin and M Kac. Phys. Rev., 86:821, 1952.
[27] R Zwanzig and J I Lauritzen. Exact calculation of partition function for a model of 2-
84
dimensional polymer cristallzation by chain folding . J. Chem. Phys., 48:3351, 1968.
[28] V Privman, G Forgacs, and H L Frisch. New solvable model of polymer-chain adsorption at
a surface. Phys. Rev. B, 37:9897, 1988.
[29] F Igloi. Phys. Rev. A, 43:3194, 1991.
[30] F Igloi. Europhys. Lett., 16:171, 1991.
[31] M E Fisher. J. Chem. Phys., 45:1469, 1966.
[32] A A Vedenov, A A Dykhne, and M D Frank-Kamenetsky. Sov. Phys. Uspekhi, 105:479, 1971.
[33] R J Rubin. Random-Walk Model of Chain-Polymer Adsorption at a Surface. J. Chem. Phys.,
43(7):2392–2407, 1965.
[34] R J Rubin. Random-walk model of adsorption of a chain polymer molecule on a long rigid
rod molecule . J. Chem. Phys., 44(7):2130, 1966.
[35] Y Lepine and A Caille. CONFIGURATION OF A POLYMER-CHAIN INTERACTING
WITH A PLANE INTERFACE . Can. J. Phys., 56:403, 1978.
[36] E Eisenriegler, K Kremer, and K Binder. Adsorption of polymer chains at surfaces: Scaling
and Monte Carlo analyses. J. Chem. Phys., 77(12):6296–6320, 1982.
[37] A A Gorbunov, E B Zhulina, and A M Skvortsov. Polymer, 23:1133, 1982.
[38] T M Birshtein and O V Borisov. Theory of adsorption of polymer chains at spherical surfaces:
1 Partition function: diagram of states. Polymer, 32(5):916–922, 1991.
[39] T M Birshtein and O V Borisov. Theory of adsorption of polymer chains at spherical surfaces:
2 Conformation of macromolecule in different regions of the diagram of states. Polymer,
32(5):923–929, 1991.
[40] M W Matsen. Soft Matter, v.1: Polymer melts and mixtures, page 87. 2006.
[41] L D Landau and E M Lifshitz. Statistical Physics. Butterworth-Heinemann, 3rd edition,
1980.
[42] C N Yang and T D Lee. Statistical theory of equations of state and phase transitions.1.Theory
of condensation. Phys. Rev., 87:404, 1952.
[43] C N Yang and T D Lee. Statistical theory of equations of state and phase transitions. 2.
Lattice gas and Ising model. Phys. Rev., 87:410, 1952.
[44] G A Baker and P Moussa. J. App. Phys., 49:1360, 1978.
[45] B W Southern and M Knevevic. Phys. Rev. B, 35:5036, 1987.
[46] X Z Wang. Physica A, 308:163, 2007.
85
[47] S Katsura. Progr. Theor. Phys., 38:1415, 1967.
[48] S Ono, Y Karaki, M Suzuki, and C Kawabata. J. Phys. Soc. Jpn., 25:54, 1968.
[49] A J Guttmann. J. Phys. C, 2:1900, 1969.
[50] C Domb and A J Guttmann. J. Phys. C, 3:1652, 1970.
[51] C Itzykson, R Pearson, and J B Zuber. Distribution od zeros in ising and gauge models.
Nucl. Phys. B, 220:415, 1983.
[52] G Marchesini and R Shrock. Nucl. Phys. B, 318:541, 1989.
[53] I G Enting, A J Guttmann, and I Jensen. J. Phys. A, 27:6987, 1994.
[54] V Matveev and R Shrock. J. Phys. A, 28:1557, 1995.
[55] V Matveev and R Shrock. J. Phys. A, 28:5235, 1995.
[56] C Kawabata and M Suzuki. J. Phys. Soc. Jpn., 27:1105, 1969.
[57] P F Fox and A J Guttmann. J. Phys. C, 6:913, 1973.
[58] V Matveev and R Shrock. J. Phys. A, 28:L533, 1995.
[59] V Matveev and R Shrock. Phys. Lett. A, 204:353, 1995.
[60] S Grossmann and W Rosenhauer. Zeit. Physik, 207:138, 1967.
[61] E Marinari. Complex zeros of the d=3 ising model - finite-size scaling and critical amplitudes.
Nucl. Phys. B, 235:123, 1984.
[62] R B Pearson. Phys. Rev. B, 26:6285, 1982.
[63] L I Klushin, A M Skvortsov, and Gorbunov A A. Adsorption of a macromolecule in an
external field: An exactly solvable model with bicritical behavior. Phys.Rev. E, 56:1511,
1997.
[64] L I Klushin, A M Skvortsov, and F A M Leermakers. Exactly solvable model with stable
and metastable states for a polymer chain near an adsorbing surface. Phys.Rev.E, 66:036114,
2002.
[65] R Balescu. Equilibrium and Nonequilibrium Statistical Mechanics. N.Y.-London-Sydney-
Toronto: Wiley-Interscience Publication, 1975.
[66] Chaikin P M and Lubensky T C. Principles of Condensed Matter Physics. Cambridge
University Press, 1995.
[67] J I Jr. Lauritzen and R Zwanzig. Exact Calculation of the Partition Function for a Gener-
alized Model of Two-Dimensional Polymer Crystallization by Chain Folding. The Journal of
Chemical Physics, 52(7):3740–3751, 1970.
86
[68] T M Birshtein, A M Elyashevich, and A M Skvortsov. Mol.Biol. (USSR), 5:78, 1971.
[69] F Kano and H Maeda. Mol. Simul., 16:261, 1996.
[70] V G Adonts, T M Birshtein, A M Elyashevich, and A M Skvortsov. Biopolymers, 15:1037,
1976.
[71] T M Birshtein, O V Borisov, A K Karaev, and E B Zhulina. Biopolymers, 24:2057, 1985.
[72] H Wakana, T Shigaki, and N Saito. Intramolecular alpha-helix-beta-structure-randon coil
transition in polypeptides.1. Equilibrium case . Biophys.Chem., 16:275, 1982.
[73] J des Cloizeaux and G Jannink. Polymers in Solutions: Their Modeling and Structure. Oxford:
Claredon, 1990.
[74] I D Lawrie and S Sarbach. Phase Transitions and Critical Phenomena, volume 9. Academic,
London, 1984.
[75] van Rensburg E J J and Rechnitzer A R. Multiple markov chain monte carlo study of
adsorbing self-avoiding walks in two and in three dimensions. J. Phys. A, 37:6875–6898,
2004.
[76] T M Birshtein and V A Pryamitsyn. Macromolecules, 25:1554, 1991.
[77] S. Flory, P.J.;Fisk. J. Chem. Phys., 55:4338, 1971.
[78] I M Lifshits, A Y Grosberg, and A R Khokhlov. Rev. Mod. Phys., 50:68, 1978.
[79] I M Lifshits, A Y Grosberg, and A R Khokhlov. Usp. Fis. Nauk, 127:353, 1979.
[80] M Fixman. J. Chem. Phys., 36:306, 1962.
[81] S T Sun, I Nishio, G Swislow, and T Tanaka. J. Chem. Phys., 73:5971, 1980.
[82] I H Park, Q.-W Wang, and B Chu. Macromolecules, 20:1965, 1987.
[83] B Chu, R Xu, and J Zhuo. Macromolecules, 21:273, 1988.
[84] B Chu, J Yu, and Z L Wang. Prog. Colloid Polym. Sci., 91:142, 1993.
[85] J Yu, Z Wang, and B Chu. Macromolecules, 25:1618, 1992.
[86] B Chu, Q Ying, and A Y Grosberg. Macromolecules, 28:180, 1995.
[87] S. Picarra, J. Duhamel, A. Fedorov, and J. M. G. Martinho. Coil-globule transition of pyrene-
labeled polystyrene in cyclohexane: Determination of polymer chain radii by fluorescence .
J. Phys. Chem. B, 108:12009, 2004.
[88] M Nakata. Phys. Rev. E, 51:5770, 1995.
[89] M Nakata and T Nakagawa. Phys. Rev. E, 56:3338, 1995.
[90] E E Gürel, N Kayaman, B M Baysal, Karasz, and F E J. Polym. Sci. Part B: Polym Phys.,
87
37:2253, 1999.
[91] N Kayaman, E E Gürel, B M Baysal, and F E Karasz. Macromolecules, 32:8399, 1999.
[92] N Kayaman, E E Gürel, B M Baysal, and F E Karasz. Polymer, 41:1461, 2000.
[93] Y Nakamura, N Sasaki, and M Nakata. Macromolecules, 34:5992, 2001.
[94] K Kubota, S Fujishige, and I Ando. J. Phys. Chem., 94:5154, 1990.
[95] M Meewes, J Richa, M de Silva, R Nyffenegger, and Th Binkert. Macromolecules, 24:5811,
1991.
[96] CWu and S Zhou. LASER-LIGHT SCATTERING STUDYOF THE PHASE-TRANSITION
OF POLY(N-ISOPROPYLACRYLAMIDE) IN WATER .1. SINGLE-CHAIN . Macro-
molecules, 28:8381, 1995.
[97] C Wu and S Zhou. First observation of the molten globule state of a single homopolymer
chain . Phys. Rev. Lett., 77:3053, 1996.
[98] C Wu. Polymer, 39:4609, 1998.
[99] C Wu and X Wang. Globule-to-coil transition of a single homopolymer chain in solution .
Phys. Rev. Lett., 80:4092, 1998.
[100] X Qiu and C Wu. Macromolecules, 31:2972, 1998.
[101] X Wang and C Wu. Macromolecules, 32:4299, 1999.
[102] J Xu, Z Zhu, S Luo, C Wu, and S Liu. First observation of two-stage collapsing kinetics of a
single synthetic polymer chain. Phys. Rev. Lett., 96:027802, 2006.
[103] K Zhou, Lu Y, J Li, Shen L, G Zhang, Z Xie, and C Wu. The Coil-to-Globule-to-Coil Tran-
sition of Linear Polymer Chains in Dilute Aqueous Solutions: Effect of Intrachain Hydrogen
Bonding. Macromolecules, 41:8927, 2008.
[104] P Stepanek, C Konak, and B Sedlacek. Macromolecules, 15:1214, 1982.
[105] A K Kron, O B Ptitsyn, A M Skvortsov, and A K Fedorov. Molek. Biol. USSR. 1967, 1, 576,
1:576, 1967.
[106] A M Eljashevich and A M Skvortsov. Molek. Biol. USSR, 5:204, 1971.
[107] C Wu and S Zhou. THERMODYNAMICALLY STABLE GLOBULE STATE OF A SIN-
GLE POLY(N-ISOPROPYLACRYLAMIDE) CHAIN INWATER . Macromolecules, 28:5338,
1995.
[108] I C Sanchez. Macromolecules, 12:980, 1979.
[109] I Szleifer, E M O’Toole, and A Z Panagiotopoulos. J. Chem. Phys., 97:6802, 1992.
88
[110] W Hu. J. Chem. Phys., 109:3686, 1998.
[111] 110 10212 Liang, H. J Chem Phys 1999. J. Chem. Phys., 110:10212, 1999.
[112] Y Zhou, C K Hall, and M Karplus. Phys. Rev. Lett., 77:2822, 1996.
[113] H Noguchi and K Yoshikawa. J. Chem. Phys., 109:5070, 1998.
[114] J M Polson and N E Moore. J. Chem. Phys., 122:024905, 2005.
[115] A N Rissanou, S H Anastasiadis, and I A Bitsanis. Monte Carlo study of the coil-to-globule
transition of a model polymeric system. J. Pol. Sci. Part B: Pol. Phys., 44:3651, 2006.
[116] A Y Grosberg and D V Kuznetsov. Quantitative theory of the globule-to-coil transition .1.
Link density distribution in a globule and itsradius of gyration. Macromolecules, 25:1970–
1979, 1992.
[117] A Y Grosberg and D V Kuznetsov. Quantitative theory of the globule-to-coil transition .2.
Density-density correlation in a globule and the hydrodynamic radius of a macromolecule.
Macromolecules, 25:1980–1990, 1992.
[118] A Y Grosberg and D V Kuznetsov. Quantitative theory of the globule-to-coil transition .3.
Globule-globule interactions and polymer-solution binodal and spinodal curves in the globule
range . Macromolecules, 25:1991–1995, 1992.
[119] A Y Grosberg and D V Kuznetsov. Quantitative theory of the globule-to-coil transition .4.
Comparison of theoretical results with experimental data . Macromolecules, 25:1996–2003,
1992.
[120] A D Sokal. Europhys. Lett., 27:661, 1994.
[121] A L Owczarek and T Prellberg. First-order scaling near a second-order phase transition:
Tricritical polymer collapse . Europhys. Lett., 51:602, 2000.
[122] T Prellberg and A L Owczarek. Four-dimensional polymer collapse: Pseudo-first-order tran-
sition in interacting self-avoiding walks . Phys. Rev. E, 62:3780, 2000.
[123] A R Khokhlov. Physica A, 105:357, 1981.
[124] A Yu Grosberg and A R Khokhlov. Statistical Physics of Macromoecules. AIS Press, New
York, 1994.
[125] F Rampf, W Paul, and K Binder. On the first-order collapse transition of a three-dimensional,
flexible homopolymer chain mode. Europhys. Lett., 70:628, 2005.
[126] M P Taylor, W Paul, and K Binder. All-or-none proteinlike folding transition of a flexible
homopolymer chain . Phys.Rev. E, 79:050801(R), 2009.
89
[127] M P Taylor, W Paul, and K Binder. Phase transitions of a single polymer chain: A Wang-
Landau simulation study . J.Chem.Phys., 131:114907, 2009.
[128] R J Rubin. A random walk model of chain polymer adsorption at a surface. 2. Effect of
correlation between neghboring steps. J. Res. Natl. Bur. Stand. Sec. B, 69:301, 1965.
[129] R J Rubin. A random walk model of chain polymer adsorption at a surface. 3. Mean square
end-to-end distance . J. Res. Natl. Bur. Stand. Sec. B, 70:237, 1966.
[130] R Hegger and P Grassberger. J. Phys. A: Math. Gen., 27:4069, 1994.
[131] P-G de Gennes. J. Phys. (France), 37:1445, 1976.
[132] E Bouchaud and J Vannimenus. J. Phys. (France), 49:2931, 1989.
[133] T W Burkhardt, E Eisenriegler, and I Guim. Nuc. Phys. B., 316:559–572, 1989.
[134] H K Janssen and A Lyssy. Phys. Rev. E, 50:3784, 1994.
[135] H-P Hsu, W Nadler, and P Grassberger. J. Phys. A: Math. Gen., 38:775, 2005.
[136] H-W Diel and M Shpot. Phys. Rev. Lett., 73:3431, 1994.
[137] H-W Diel and M Shpot. Nucl. Phys. B, 528:595, 1998.
[138] M Luo. The critical adsorption point of self-avoiding walks: A finite-size scaling approach .
J.Chem.Phys., 128:044912, 2008.
[139] H Meirovitch and S Livne. J. Chem. Phys., 88:4507, 1988.
[140] R Descas, J-U Sommer, and A Blumen. J. Chem. Phys., 120:8831, 2004.
[141] E J Janse van Rensburg and A R Rechnitzer. J. Phys. A: Math. Gen., 37:6875, 2004.
[142] P Grassberger. J. Phys. A: Math. Gen., 38:323–331, 2005.
[143] S Bhattacharya, H-P Hsu, A Milchev, V G Rostiashvili, and T A Vilgis. Macromolecules,
41:2920–2930, 2008.
[144] S Metzger, M Müller, K Binder, and J Baschnagel. Macromol. Theory Simul., 11:985, 2002.
[145] P.-G. de Gennes. Scaling Concepts in Polymer Physics. Cornell University Press, Ithaca and
London, 1979.
[146] A A Gorbunov and A M Skvortsov. J. Chem. Phys., 98:59610, 1993.
[147] A A Gorbunov and A M Skvortsov. Polymer Sci. (Russia), 31:1244, 1989.
[148] A A Gorbunov, A M Skvortsov, J van Male, and G J Fleer. J. Chem. Phys., 114:5366, 2001.
[149] K Binder, editor. Applications of the Monte-Carlo Method in Statistical Physics. Springer,
1987.
[150] P. G. De Gennes. Superconductivity Of Metals And Alloys. Westview Press, 1999.
90
[151] E Eisenriegler, A Hanke, and S Dietrich. Polymers interacting with spherical and rodlike
particles. Phys. Rev. E, 54:1134–1152, 1996.
[152] M S Causo, B Coluzzi, and P Grassberger. Simple model for the DNA denaturation transition.
Phys. Rev. E, 62:3958, 2000.
[153] A Hanke. Adsorption transition of a self-avoiding polymer chain onto a rigid rod. J. Phys.:
Condens. Matter, 17:S1731–S1746, 2005.
[154] A M Skvortsov, L I Klushin, and T M Birshtein. Stretching and Compression of a Macro-
molecule under Different Modes of Mechanical Manipulations. Polym. Sci. Ser. A. (Russia),
51:469–491, 2009.
[155] Zhulina E.B. Birshtein T.M. Skvortsov A.M., Gorbunov A.A. Influence pf mechanical forces
on adsorption and adhesion of macromolecules . Polymer Sci. (USSR), A20(2):278, 1978.
[156] A M Skvortsov, A A Gorbunov, and L I Klushin. Adsorption-stretching analogy for a polymer
chain on a plane- symmetry property of the phase diagram . J. Chem. Phys., 100:2325–2334,
1994.
[157] B J Haupt, J Ennis, and E M Sevick. Langmuir, 15:3886, 1999.
[158] A M Skvortsov, L I Klushin, G J Fleer, and Leermakers F A M. Temperature effects in
the mechanical desorption of an infinitely long lattice chain: Re-entrant phase diagrams.
J.Chem.Phys., 130:174704, 2009.
[159] A M Skvortsov, L I Klushin, G J Fleer, and Leermakers F A M. Analytical theory of finite-size
effects in mechanical desorption of a polymer chain . J.Chem.Phys., 132:064110, 2010.
[160] S Bhattacharya, H-P Hsu, A Milchev, V G Rostiashvili, and T A Vilgis. Phys. Rev. E,
79:030802(R), 2009.
[161] S Bhattacharya, H-P Hsu, A Milchev, V G Rostiashvili, and T A Vilgis. Macromolecules,
42:2236, 2009.
[162] A M Skvortsov, L I Klushin, and F A M Leermakers. Exactly solved polymer models with
conformational escape transitions of a coil-to-flower type. Europhys. Lett., 58:292, 2002.
[163] L I Klushin, A M Skvortsov, and F A M Leermakers. Partition function, metastability, and
kinetics of the escape transition for an ideal chain. Phys. Rev. E, 69:061101, 2004.
[164] D I Dimitrov, L I Klushin, A M Skvortsov, and Binder K. The escape transition of a polymer:
A unique case of non-equivalence between statistical ensembles . Europ.Phys.J., 29:9–25, 2009.
[165] H-P Hsu, K Binder, L I Klushin, and AM Skvortsov. What is the order of the two-dimensional
91
polymer escape transition? . Phys. Rev. E, 76:021108, 2007.
[166] A M Skvortsov, L I Klushin, J van Male, and Leermakers F A M. First-order coil-to-flower
transition of a polymer chain pinned near a stepwise external potential: Numerical, analytical,
and scaling analysis. J.Chem.Phys., 115:1586–1595, 2001.
[167] A M Skvortsov, L I Klushin, J van Male, and Leermakers F A M. The rolling transition of a
Gaussian chain end-grafted at a penetrable surface. J.Chem.Phys., 112:7238–7246, 2000.
[168] A M Skvortsov, J van Male, and Leermakers F A M. Exactly solvable model with an unusual
phase transition: the rolling transition of a pinned Gaussian chain source: Physica a-statistical
mechanics and its applications volume: 290 issue: 3-4 pages: 445-452 published: Feb 15 2001.
Physica A - Stat.Mech., 290:445–452, 2001.
[169] A M Skvortsov, L I Klushin, and F A M Leermakers. On the escape transition of a tethered
Gaussian chain; Exact results in two conjugate ensembles. Macromol. Symp., 237:73, 2006.
[170] G Subramanian, D R M Williams, and P A Pincus. ESCAPE TRANSITIONS AND FORCE
LAWS FOR COMPRESSED POLYMER MUSHROOMS. Europhys. Lett., 29:285, 1995.
[171] G Subramanian, D R M Williams, and P A Pincus. Interaction between finite-sized particles
and end grafted polymers . Macromolecules, 29:4045, 1996.
[172] J Ennis, E M Sevick, and D R M Williams. Phys. Rev. E., 60:6906, 1999.
[173] E M Sevick and D R M Williams. Macromolecules, 32:6841, 1999.
[174] A Milchev, V Yamakov, and K Binder. Escape transition of a polymer chain: Phenomeno-
logical theory and Monte Carlo simulations. Phys. Chem. Chem. Phys., 1:2083, 1999.
[175] A Milchev, V Yamakov, and K Binder. Escape transition of a compressed polymer mushroom
under good solvent conditions. Europhys. Lett., 47:675–680, 1999.
[176] D R M Williams and F C MacKintosh. J. Phys. II, 9:1417, 1995.
[177] J Abbou, A Anne, and K Demaille. J. Phys. Chem., 110:22664, 2006.
[178] F A M Leermakers and A A Gorbunov. Polymer-surface interactions in bridging escape and
localization transitions . Macromolecules, 35:8640, 2002.
[179] E M Sevick. Macromolecules, 33:5743, 2000.
[180] J Ennis and E M Sevick. Macromolecules, 34:1908, 2001.
[181] F A M Leermakers, L I Klushin, and A M Skvortsov. Negative compressibility for a polymer
chain squeezed between two pistons going through the escape transition . J. Stat. Mech.,
Theory Exp., page 10008, 2004.
92
[182] A M Skvortsov, L I Klushin, and F A M Leermakers. Negative compressibility and nonequiv-
alence of two statistical ensembles in the escape transition of a polymer chain. J. Chem.
Phys., 126:024905, 2007.
[183] D.H.E. Gross. Microcanonical Thermodynamics. Singapore, 2001.
[184] H-P Hsu, K Binder, L I Klushin, and Skvortsov A M. Escape transition of a polymer chain
from a nanotube: How to avoid spurious results by use of the force-biased pruned-enriched
Rosenbluth algorithm. Phys. Rev. E, 78:041803, 2008.
[185] W Reisner, K J Morton, R Riehn, Y M Wang, Z Yu, M Rosen, JC Sturm, .S Y Chou, E Frey,
and R H Austin. Statics and dynamics of single DNA molecules confined in nanochannels.
Phys. Rev. Lett., 94:196101, 2005.
[186] P Cifra. Channel confinement of flexible and semiflexible macromolecules. J.Chem.Phys.,
131:14 2009, 2010.
[187] A Milchev, W Paul, and K Binder. Polymer-chains confined into tubes with attractive walls-a
Monte-Carlo simulation. Macromol.Theory Simul., 3:305, 1994.
[188] P Sotta, A Lesne, and JM Victor. Monte Carlo simulation of a grafted polymer chain confined
in a tube. J. Chem. Phys., 112:1565, 2000.
[189] P Sotta, A Lesne, and JM Victor. The coil-globule transition for a polymer chain confined in
a tube: A Monte Carlo simulation. J.Chem.Phys., 113:6966–6973, 2005.
[190] Y Yang, TM Burkhardt, and G Gompper. Free energy and extension of a semiflexible polymer
in cylindrical confining geometries. Phys. Rev. E, 76:011804, 2007.
[191] K Kremer and 81 6381. Binder, KJ Chem. Phys. 1984. Dynamics of polymer chains confined
into tubes- scaling theory and Monte-Carlo simulations. Chem. Phys., 81:6381, 1984.
[192] H Loebl, R Randel, C P Goodwin, and CC Matthai. Simulation studies of polymer translo-
cation through a channel. Phys. Rev. E, 67:041913, 2003.
[193] R Randel, H Loebl, and C Matthai. Molecular dynamics simulations of polymer transloca-
tions. Macromol.Theory Simul., 13:387, 2004.
[194] L I Klushin, A M Skvortsov, H P Hsu, and Binder K. Dragging a polymer chain into a
nanotube and subsequent release source: Macromolecules volume: 41 issue: 15 pages: 5890-
5898 published: 2008. Macromolecules, 41:5890–5898, 2008.
[195] R B Pearson. Partition function of the Ising mdel on the periodic 4x4x lattice. Phys. Rev.
B, 26:6285, 1982.
93