Você está na página 1de 71

The rheology of dilute solutions of flexible polymers: Progress and

problems
Ronald G. Larson
Citation: J. Rheol. 49, 1 (2005); doi: 10.1122/1.1835336
View online: http://dx.doi.org/10.1122/1.1835336
View Table of Contents: http://www.journalofrheology.org/resource/1/JORHD2/v49/i1
Published by the The Society of Rheology

Related Articles
Nonlinear signatures in active microbead rheology of entangled polymer solutions
J. Rheol. 57, 1247 (2013)
A cone-partitioned plate rheometer cell with three partitions (CPP3) to determine shear stress and both normal
stress differences for small quantities of polymeric fluids
J. Rheol. 57, 841 (2013)
Jetting behavior of polymer solutions in drop-on-demand inkjet printing
J. Rheol. 56, 1109 (2012)
Letter to the editor: Cone partitioned plate (CPP) vs circular couette
J. Rheol. 56, 675 (2012)
Response to: CPP vs circular couette
J. Rheol. 56, 683 (2012)

Additional information on J. Rheol.


Journal Homepage: http://www.journalofrheology.org/
Journal Information: http://www.journalofrheology.org/about
Top downloads: http://www.journalofrheology.org/most_downloaded
Information for Authors: http://www.journalofrheology.org/author_information

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

The rheology of dilute solutions of flexible polymers:


Progress and problems
Ronald G. Larsona)
Department of Chemical Engineering, University of Michigan, Ann Arbor,
Michigan 48109-2136
(Received 13 May 2004; final revision received 22 October 2004)

Synopsis
Recent progress toward understanding the rheology of dilute solutions of flexible polymers is
reviewed, emphasizing experimental results from flows imaging single deoxyribonucleic acid
DNA molecules and filament-stretching rheometry of dilute polystyrene Boger fluids, as well as
Brownian dynamics BD simulations of these flows. The bead-spring and bead-rod models are
presented, the range of their applicability discussed, and methods presented for inclusion of
hydrodynamics interactions, excluded volume, and other physical effects within BD simulations.
After reviewing and updating work in the linear viscoelastic regime, the primary focus shifts to the
more complex nonlinear regime. While BD predictions of the conformations of 20 to 100 micron
long DNA molecules in strong shear and extensional flows has been in good to excellent agreement
with the corresponding experiments, predictions of the polystyrene dilute solution rheometry data
have been hit or miss, with poorer results obtained for the higher molecular weights. This may be
due, in part, to the more important roles of hydrodynamic interactions and excluded volume
interactions in the more flexible, and therefore more condensed, polystyrene coils. Inclusion of these
effects in BD simulations has led to improved predictions, but does not lead to the accurate
prediction of the plateau Trouton viscosity for higher molecular weight samples, nor alleviate the
complete failure of simulations to predict measurements of coil distortion by light scattering. Thus,
despite enormous progress in the past decade, some significant gaps in understanding remain.
2005 The Society of Rheology. DOI: 10.1122/1.1835336

I. INTRODUCTION
Since the late 1980s, there has been enormous progress in molecular-level understanding of the rheological properties of dilute solutions of flexible polymers. This
progress is due primarily to four advances: 1 The application of methods of imaging the
conformations of isolated polymer deoxyribonucleic acid DNA molecules by optical
microscopy in well-defined flows Perkins et al. 1995; 1997; 2 the development of the
filament-stretching rheometer, which can impose a high-quality steady extensional flow
history on dilute solutions of polymers in viscous solvents Tirtaatmadja and Sridhar
1993; McKinley and Sridhar 2002; 3 the development of model dilute polymer solutions, in which the polymer is nearly monodisperse and of high molecular weight and the
solvent very viscous Mackay and Boger 1987; Magda and Larson 1988; and 4 the
increase in computer speed and the development of methods that permit simulation of
ensembles of bead-spring or bead-rod chains containing enough beads to describe realistically the conformations of real polymers Liu 1989; Lopez Cascales and Garca de la
a

Electronic mail: rlarson@engin.umich.edu

2005 by The Society of Rheology, Inc.


J. Rheol. 491, 1-70 January/February 2005

0148-6055/2005/491/1/70/$25.00

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

LARSON

Torre 1991; Grassia and Hinch 1996; Somasi et al. 2002. A review of the status of
molecular rheology of dilute solutions of flexible polymers at the end of the 1980s can
ttinger 1992. This
be found in Chapter 8 of Larson 1988 as well as in Bird and O
article updates those reviews, emphasizing results arising from the four advances listed
above. A complementary recent review can be found in Prakash 1999.
Since the polymer solutions used commercially are almost invariably nondilute a
notable exception being the dilute solutions used in turbulent drag-reducing flows Virk
1975, interest in dilute polymer solutions derives primarily from their importance in the
characterization of a polymers molecular weight, stiffness, long-chain branching, and
interaction with solvent, and from a fundamental interest in understanding macromolecular response to hydrodynamic forces, free from the complications of intermolecular entanglements. However, as we shall see, diluteness paradoxically creates added complications not present in concentrated entangled solutions. For purposes of molecular
characterization and fundamental understanding, the most useful dilute solutions are
those for which the polymer molecular weight is as monodisperse as possible. By definition of diluteness, the rheology of any polydisperse dilute solution can always be
deduced by linearly combining the rheological properties of a series of monodisperse
dilute solutions representing the molecular weight distribution of the polydisperse
sample. Hence, nothing fundamental is learned by studying the effects of polydispersity
on dilute solution rheology. For this reason, only nearly monodisperse dilute solutions
will be considered in this review. We will also limit this review to uniform velocity
gradients, especially those of steady shear, extensional, and mixed shear/extensional
flows, since these provide the most clear-cut insight into rheological properties.
In the following, we assume that the reader has some familiarity with polymers and
polymer solutions, such as the random-flight configuration of ideal chains, the polymer
end-to-end vector, etc., and knows basic rheological concepts, such as cone-and-plate
rheometry, and the definitions of the storage and loss moduli, G and G . If needed, the
reader can find these concepts thoroughly described in numerous texts, including Bird
et al. 1987, Larson 1988; 1999, and Tanner 1985. This review is organized as follows. We will present the criteria governing the term dilute solution Sec. II, and
describe the forces and interactions that control the dilute-solution rheology of flexible
polymers Sec. III. Next, we will present the conventional bead-spring and bead-rod
models of polymer rheology and limitations on their validity as well as moment equations
and Brownian dynamics equations for obtaining predictions from them Sec. IV. Their
predictions are then compared to experimental dilute-solution rheology data in the linear
Sec. V and nonlinear Sec. VI regimes. Finally, the results are summarized and the
gaps between predictions and the data are discussed Sec. VII.
II. CRITERION FOR DILUTENESS
Here we take the term dilute to have its usual meaning, that the polymer concentration is low enough that the polymers do not interact with each other either topologically i.e., through entanglements or hydrodynamically, and hence the effect of the
polymers on the rheological properties of the fluid in a fixed flow field is linear in the
concentration of polymer in solution. A different criterion for diluteness arises if one
wishes to measure rheological properties of dilute solutions in a flow without the polymer
molecules disturbing that flow. This criterion can in some cases be even stricter than the
normal criterion for diluteness Feng and Leal 1987, and would, for example, make all
polymer solutions that reduce drag in turbulent flows nondilute. Thus, the stipulation of a
linear effect of the polymer in a fixed flow field is a significant one.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

In the usual definition of a dilute solution, then, a rheological property such as the
viscosity is linear in the mass concentration c of polymer. One can, therefore, define
intrinsic properties such as the intrinsic viscosity of the polymer in dilute solution as
0 lim

c0

cs

where is the solution viscosity and s is the viscosity of the solvent. The intrinsic
viscosity can be obtained by measuring the viscosity of a series of polymer solutions
under increasingly more dilute conditions and extrapolating the ratio ( s )/c s to
zero concentration. The subscript 0 on 0 implies that it is measured at low shear
rates, where it reaches an asymptotic plateau value that is independent of shear rate. From
its definition, it is evident that the intrinsic viscosity has units of volume per unit mass,
and can be thought of as the hydrodynamic volume occupied by a unit mass of the
polymer in dilute solution. Interpreted this way, one deduces that polymer molecules
begin to overlap with each other whenever c 0 exceeds unity or so. Thus, a simple
criterion for diluteness of a polymer solution is that
c

1
0

More precise expressions, and a thorough discussion of the criteria for diluteness of
flexible, semiflexible, and rigid polymers in both good and theta solvents defined in
Sec. III can be found in Bercea et al. 1999.
In a theta solvent defined below, the hydrodynamic volume of a polymer coil scales
with the 3/2 power of the molecular weight. Since the intrinsic viscosity is proportional
to the hydrodynamic volume of a polymer molecule divided by its mass, it follows that in
a theta solvent,
0 KM 1/2,

where K is a constant that depends on the equilibrium mean-square end-to-end distance


of a polymer, R 2 0 , per unit molecular weight M, as
K R20 /M 3/2


C2
m0

3/2

Here, the brackets denote an ensemble average over a large number of polymer
molecules, and the subscript 0 on R 2 0 denotes an equilibrium average, i.e., in the
no-flow limit. C is the characteristic ratio of the polymer, defined as the ratio of the
mean-square end-to-end distance of the polymer to that for an ideal freely jointed chain
with the same number of backbone bonds n and same bond length as the real polymer;
that is Ferry 1980:

R20 Cn2,

where C has a value of around 9.6 for polystyrene Fetters et al. 1994. In Eq. 4, m 0
is the molecular weight per backbone bond of the polymer, which is 52 Daltons for
polystyrene, and 1.54 A for a carboncarbon bond.
The FloryFox parameter should have a universal value in a theta solvent; the
experimental value is around 2.51023 mol1 Bercea et al. 1999. For polystyrene,
with C 9.6, this yields K 7.2102 cm3 g1 (g/mol) 1/2, close to the value of
K 8102 cm3 g1 (g/mol) 1/2 reported for polystyrene by Flory 1969, who took

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

LARSON

C 10.0. Thus, using Florys value for K , we find that for polystyrene of molecular
weight 10 million Daltons, 0 253 cm3 /gm in a theta solvent. Hence, a polystyrene
solution of this molecular weight is dilute only if its concentration is well below about
1/ 0 0.004 gm/cc, or about 0.4% polystyrene by mass or by volume for solvents of
typical density. In general, this criterion for diluteness is not very strict, as it implies that
at the cross-over condition, the viscosity of the solution is more than doubled by the
polymer. Sridhar and co-workers found for polystyrene of molecular weight 10.2 million
in a viscous theta-like solvent that the transition to a non-dilute regime begins at a
concentration of around 777 ppm, as measured by onset of a concentration dependence of
the longest relaxation time and by a failure of the extensional viscosity to scale linearly
with concentration Gupta et al. 2000. This concentration is about 20% of the cross-over
value estimated from 1/ 0 . Thus, one might use 1/5 0 as a stricter criterion for
diluteness, and 1/ 0 as a lenient one. When flexible polymers are highly deformed
under flow, the requirements for diluteness appear to become even more stringent than
under quiescent conditions, with diluteness apparently only being achieved in some cases
at concentrations more than a decade lower than 1/ 0 Keller and Odell 1985; Feng
and Leal 1997; Clasen et al. 2004.
We define for future reference the radius of gyration R g of a polymer coil as the radius
of a sphere of equal moment of inertia to that of the polymer, if we were to put all of the
polymers mass on the surface of that sphere. Mathematically,
R2g

i1j1

Rij 2

where n is the number of monomers and R i j is the spatial distance between monomers i
and j. For a random-walk polymer, R 2g R 2 0 /6.
III. PHYSICAL FORCES AND PHENOMENA
Over the ranges of length scales a few nanometers to a few microns and time scales
s to s of primary importance for rheological properties, the following polymer phenomena are of real or potential importance:
1
2
3
4
5
6
7

Viscous drag,
Entropic elasticity,
Brownian forces,
Hydrodynamic interaction HI,
Excluded-volume EV interactions,
Internal viscosity IV, and
Self-entanglement SE.

We have ordered these phenomena according to their importance, as currently understood. Viscous drag, which is the frictional force that the flowing solvent exerts on the
polymer, is always important, even under flows too weak to excite an elastic response
from the polymer. Entropic elasticity becomes important as soon as the flow is strong
enough to deform the chains away from their equilibrium distribution of conformations,
and is shown by viscoelastic i.e., time dependent rheological effects. In addition,
Brownian motion, due to the bombardment of the polymer by solvent molecules, also
influences the chains distribution of conformations under flow.
HI, which is the disturbance to the flow field produced by one part of the chain that
influences the drag on another part, is always important for high molecular weight flex-

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

ible polymers, but its effect can, in some cases as discussed below, be subsumed into
the effective drag properties of the chain. EV interactions are the repulsive forces between monomers that prevent their overlap, leading to a tendency of the chains to expand
beyond the ideal random-walk conformations they have in the melt. EV interactions can
be effectively cancelled out in some solvents at their theta point, which is the temperature at which the weak repulsion the polymer feels toward the solvent is just strong
enough to cancel out the tendency of the coil to expand due to EV. Values for theta
temperatures for many polymer/solvent pairs can be found in Sundararajan 1996. Because the theta point is a single temperature, EV is therefore usually, but not always, of
at least some importance in the dynamics of dilute polymer solutions. Interestingly, in
concentrated solutions or in the melt, both HI and EV interactions are screened out and
so do not usually need to be considered, except perhaps at small length scales, much less
than the chains radius of gyration. In this sense, at least, concentrated solutions are less
complex than dilute solutions !, although in the former one must, of course, cope with
the great complicating factor of entanglements between chains, which are absent in truly
dilute solutions. HI and EV are at least partially to blame for the slowness of development of a detailed understanding of the rheology of dilute polymer solutions.
The last two phenomena on our list, namely IV and SE, are of uncertain significance.
They have been invoked by various authors to explain puzzling rheological data, but to
date their importance or even reality have not been clearly demonstrated. Recent developments, discussed below, may soon shed light on these phenomena.

IV. BEAD-SPRING AND BEAD-ROD MODELS


For years, meaningful comparisons between theories for the rheology of dilute polymer solutions and corresponding experiments were frustrated not only by the lack of
high-quality experimental data especially data that directly measure polymer molecular
conformations, but also by the statistical nature of polymer theories, which give rise to
equations that in the past required some sort of linearization or preaveraging to reduce
them to forms manageable for the computers of the day. Until fairly recently, this meant
that rheological theories had to be simplified to closed-form constitutive equations relating stress or birefringence to flow history before comparisons to experiment were possible. As a result, disagreement between theory and experiment could be blamed, at least
in part, to the mathematical closure approximations of uncertain accuracy that were
needed to bring the molecular theory into a closed form, rather than arising entirely from
limitations of the physical model. Now, however, high-speed computers make possible
the direct solution of stochastic equations containing accurate expressions for the physical forces and interactions, for ensembles of polymer chains large enough to yield accurate averages. Hence, nowadays, physical theories for polymer dynamics and stress can
be tested much more directly than before, without the uncertainties produced by mathematical closure approximations.
However, even using the most advanced computers, atomic-level simulations of longchain polymers are still not close to being feasible since the longest time scales accessible in such simulations are a fraction of a microsecond and, therefore, outside of the
rheologically most interesting range. Hence, there is still a need for coarse-graining
approximations in which only slow variables, representing coarse-grain features, are
tracked, while small-scale, fast, dynamics are assumed to remain at local equilibrium,
slaved to the slow variables. Thus, even with the very fast computers that are now readily
available, it is still important to choose the proper level of coarse graining for a given
flow, so that the simulations track all dynamics that are sluggish enough to be out of

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

LARSON

FIG. 1. Illustration of coarse-grain mapping of real polymer chain, with a carboncarbon backbone containing
fixed dihedral bond angles, onto a bead-rod chain whose configuration is that of a random walk, and further
coarse-grain mapping of the bead-rod chain onto a bead-spring chain.

equilibrium during the flow, while allowing the faster dynamics to remain at equilibrium,
where no computer time need be wasted simulating them.
A. Model definitions
The most commonly used coarse-grained models of polymer chains are the freely
jointed bead-rod and the bead-spring models, depicted in Fig. 1. The physical bases for
these models have been explained in detail elsewhere Doi and Edwards 1986; Bird et al.
1987; Larson 1988. In these models, the beads represent frictional drag centers. A single
rod in the bead-rod model represents a single link, or Kuhn step, in a freely jointed chain
model of a flexible polymer molecule, whose coarse-grained conformation is a random
walk in the absence of excluded volume forces, discussed shortly. A model related to
the bead-rod chain is a chain of freely jointed ellipsoids, where the aspect ratio of each
ellipsoid is chosen to capture the aspect ratio of a piece of the real chain that corresponds
to one Kuhn step of the chain Stigter and Bustamante 1998.
Real polymers are not freely jointed chains, but a freely jointed chain will have the
same equilibrium mean-square end-to-end length R 2 0 and fully extended length L, as
any real polymer in a theta solvent if the freely jointed step length b K and the number of
steps N K of the freely jointed chain are chosen appropriately; i.e, so that b 2K N K
R 2 0 and b K N K L. The fully extended length of a chain whose n backbone
bonds, each of length , are tetrahedrally bonded as is the case for carboncarbon
bonds, is given by
L 0.82n,

where the factor 0.82 enters because the backbone has a zig-zag configuration in the most
extended state that still preserves the tetrahedral backbone bond angle restrictions see
Fig. 1. From this, and the definition of C given in Eq. 5, we obtain the following
formulas for the equivalent freely jointed step length b K and the number of steps N K :

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

bK

C
0.82

NK

0.82 2

n.

For polystyrene molecules, for example, with C 9.6, 1.54 A, and n M /52,
Eq. 8 yields a Kuhn step length b K 1.8 nm, and N K M /742, where M is the
molecular weight in Daltons.
Since the fully extended length of the polymer is proportional to its molecular weight,
the number of rods in a bead-rod model needed to represent the chain grows in proportion
to its molecular weight. And since the computational time grows with the number of rods
raised to a power of 2 or higher, the number of rods one can simulate with current
computers is limited to no more than a few hundred. For polystyrene, this limit corresponds to molecular weights in the range of a few hundreds of Daltons. To reach a
molecular weight in the millions using the bead-rod model would require more than a
hundred-fold greater computational power, and so will remain out of reach for some time
to come.
The bead-rod model is computationally very expensive not only because of the large
number of dynamical variables the bead positions, but also because the motions of
individual rods are very rapid, requiring small simulation time steps. If, however, the
deformation of the polymer molecule is slow enough that a subchain containing, say,
N K,s connected rods fully samples its distribution of conformations in the time required
for the beads to be convected appreciably in the flow, then on this slower time scale of the
convection the rapidly fluctuating force produced by motions of individual rods averages
out to an equilibrium entropic force f s that pulls the two ends of the subchain toward
each other, and the subchain, on these slower time scales, acts as an effective spring, with
spring force f s .
Thus, for slow flows, each spring in the bead-spring model can be chosen to represent
a large number N K,s of Kuhn steps. Within the caveats discussed below, if the molecular
weight of the polymer is increased, one can simply proportionately increase the number
of Kuhn steps represented by each spring, and so the number of springs N s
N K /N K,s in the bead-spring model does not necessarily need to grow with the increasing molecular weight of the chain. The constants in the force law for the spring force
f s (r) depend on the number of Kuhn steps represented by that spring, but rules for
choosing these constants are well established see below.
B. Coarse-graining principles
While coarse graining is necessary to simulate high molecular weight polymers, there
are obviously limits to how far it can be pushed, before physical realism is lost. Generally, finer-scaled models are able to track higher-frequency chain motions, and are valid
up to higher velocity gradients or frequencies, than are coarser-grained models Larson
2004a. It is also worth noting that, contrary to intuition, the bead-rod model might not be
valid at frequencies any higher than that at which the bead-spring model fails because, at
frequencies too high to be described by an entropic spring, a chain of freely jointed rods
might also be inaccurate since the free joints fail to capture the rotational energy barriers
present in the real chain. One can estimate from the energy barrier the frequency where
this might occur and how it compares to the frequency where the bead-spring model fails
Larson 2004b.
It is also possible to incorporate bending and torsional bond rotational potentials into
the bead-rod model, thereby mimicking the flexibility of a real polymer molecule at the
level of individual backbone bonds for synthetic polymers Ryckaert and Bellemans

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

LARSON

1975; Lyulin et al. 1999. A bead-rod model for DNA chains that includes both torsional
and bending stiffness has been developed by Vologodshii and co-workers Jian et al.
1998. To maintain smooth bending of the molecule, the rod length was kept to only
around 1/20 of a Kuhn step, i.e., around 5 nm, or 15 base pairs. Using this model,
Brownian dynamics simulations of supercoiled i.e., twisted like a telephone cord DNA
plasmids circles of length up to 3000 basepairs around 1 m length were possible
over times of around 10 ms, including the effects of HI. If one wishes to simulate at a still
finer level, one can incorporate solvent molecules explicitly, and their intermolecular
interactions with each other and with the polymer can be included in great detail in fully
atomistic molecular dynamics MD simulations; see, for example, Aust et al. 1999.
However, such simulations are at present limited to timescales of less than a microsecond, too slow to capture the dynamics of long polymer molecules. Still, computer power
is now great enough that one could, in principle, capture the full range of dynamics of a
polymer molecule in solution by using multiple methods with overlapping time scales
For example, using: 1 Atomistic MD, 2 Brownian dynamics of rods with torsional
and bending potentials, and 3 Brownian dynamics of bead-spring chains, that could in
principle encompass the entire range of dynamics, from femtoseconds to seconds. Use of
overlapping time and length scales would allow the parameters of the coarse-grained
methods to be assigned by matching their predictions to those of the finer-grained models. It is even possible in principle to refine the modeling all the way to the quantum
level, if one restricts oneself to small enough portions of the macromolecule.
The need for such multiple-method, multiple-scale, modeling efforts is especially
acute in the biological world where very large molecules DNA and proteins interact
with each other via dynamics that vary from slow and long range to very fast and short
range, as occurs, for example, in DNA transcription, which involves both fast local
proteinDNA binding events, and much slower diffusive motions of the bulky DNA and
protein molecules. Some examples of what simulations can tell us about the physicomechanical properties of DNA and protein molecules can be found in Lavery et al. 2002.
A more mundane example of the vast range of time and distance scales that can be
important in the dynamics of dilute polymer solutions is that of flow-induced polymer
scission Keller and Odell 1985; Islam et al. 2004, where a local event bond breakage
governed by quantum-level forces acting on the femtosecond time scale, is driven by
flow-induced unraveling of a long polymer molecule on a time scale of milliseconds or
seconds. Modeling accurately such a process across the full range of time scales remains
an important unsolved, but possibly solvable, problem. Yet another problem area where
multiscale models are probably needed is that of polymers interacting with surfaces,
where both fast locally controlled surface binding events and slow chain motions interact.
A simple example is provided by a polymer confined to a gap comparable to the persistence length or Kuhn length of the chain. In that event, the normal bead-spring model
does not predict the correct entropic spring force owing to the steric restrictions preventing the Kuhn steps in the subchains composing the springs from sampling a full threedimensional ensemble Woo and Shaqfeh 2003. The bead-rod model may be a more
appropriate model for this situation.
At the larger time and length scales of primary interest here, guidance in choosing the
most appropriate level of coarse graining is provided by considering the modes of relaxation captured by bead-spring chains. The inverse characteristic frequencies of these
modes are the discrete relaxation times i . The simplest bead-spring model of a polymer molecule, namely the Rouse model, neglects both HI and EV forces and assumes
Hookean springs; for the Rouse model the mode relaxation times are given by Doi and
Edwards 1986:

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

16k B T s2

sin

N2
42kBTs2i2

coil R 2 0
6 2 k B Ti 2

2N

where is the bead drag coefficient, which for spherical beads is given by
6 s a, with s as the viscosity of the solvent, a as the bead radius, coil N as
the total drag coefficient summed over all the beads, and s2 3/2N K,s b 2K describes the
elasticity of a single spring. The approximation for the sine function used in Eq. 9
becomes accurate when N/i 1 and, in this limit, the relaxation times of the higher
modes decrease as 1/i 2 with increasing mode number i. The relaxation strengths G i are
all equal to each other in the Rouse or Zimm model; i.e., G i G vk B T, where v is
the number of molecules per unit of volume of solution. For any model of polymer chain
dynamics, once the spectrum of relaxation times and corresponding strengths is given, the
linear viscoelastic storage and loss moduli can be computed as
G

Gii2

i 1 2 ;

Gii

i 1 2 .

10

At a given oscillation frequency, , the storage and loss moduli are dominated by the
relaxation times, i , that are of order of, or larger than, the inverse of that frequency .
Modes with relaxation time constants, i , that are much smaller than the inverse frequency, 1 , are not excited by the flow and do not contribute appreciably to the stress.
Experimentally, the longest of these relaxation times, 1 , is related to the intrinsic
viscosity 0 of the polymer by

0M s

S1RT

11

where M is the polymer molecular weight, R N A k B is the gas constant, and N A is


Avogadros number. For the simplest models, such as those of Rouse and Zimm, where
the relaxation strengths are all equal, the coefficient S 1 is related to the distribution of
relaxation times by
S1

ii

12

For the Rouse model, one can show that S 1 2 /6 1.645, while for realistic polymers that are influenced by HI and EV effects, S 1 varies from around 2.369 Zimm
ttinger 1996b in theta solvents to values closer to the Rouse value,
19562.387 O
1.645, in good solvents. We note here that an oft-used characteristic relaxation time 0 is
defined by

0M s

RT

13

From Eq. 11, we find that the time constant 0 is larger by a factor of S 1 than the
longest relaxation time; i.e.,

0 S11 .
14
In general, for a monodisperse flexible polymer molecule described by a spectrum
with equal relaxation strengths, the mode relaxation times are distributed roughly as a

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

10

LARSON

power law in the mode number i: i i p , where by convention the modes are ordered
so that the higher the mode number, the shorter the relaxation time. The exponent p
equals 2 for the Rouse model, but for real polymers, which are affected by HIs and EV,
p ranges between 1.5 and 1.8 as the solvent quality varies from theta to very good. In
general, in an oscillatory flow, if one wishes to capture modes with frequencies up to
some frequency , then the bead-spring model must be able to describe all modes with
relaxation times down to at least 1 . However, a bead-spring chain with N s springs has
only N s modes, at least in the simplest approximate solutions, such as those of Rouse and
Zimm. More accurate solutions to the bead-spring model can have many more modes,
for example as many as N s2 modes, and the relaxation strengths are typically then not all
ttinger 1992. If the highest of these mode numbers must have a
equal Zylka and O
relaxation time of 1 or less, then this sets a minimum number of springs that must be
included in the model. However, there is not only a minimum, but also maximum number
of springs that ought to be used to represent a particular polymer molecule. This is
because the spring force laws f s (r) are derived in the asymptotic limit of many randomwalk steps or Kuhn steps N K per spring. In practice, we cannot increase the number of
springs N s beyond the point where N K,s N K /N s drops below 10 or so Underhill and
Doyle 2004. Thus, if the rate of deformation is fast enough to significantly distort the
molecular configuration at a length scale of only a few Kuhn steps of the chain, then no
simple asymptotic spring law can capture the dynamics.
C. Equations of the bead-spring model
The equations of the bead-spring model have been presented in numerous books and
articles; to fix our notation, and bring readers up to speed with what is needed later, we
review them very briefly here. Since the inertial forces on a polymer molecule are almost
always negligible, a force balance on each bead i in a bead-spring chain includes only the
drag, elastic spring, and Brownian forces, yielding
sp,b
FBi 0.
Fdrag
i Fi

15

1. Drag force

The drag force on a bead is given by


ri v ri ,
Fdrag
i

16

where is the bead drag coefficient per bead, ri is the position vector, ri dri /dt is the
velocity of the bead, and v(ri ) is the velocity of the solvent at the position of bead i. For
a uniform velocity gradient with zero velocity at the origin, v(ri ) ri , where is the
transpose of the velocity gradient tensor; i.e., (v) T . Substituting these expressions
into Eq. 15 and rearranging gives
1
ri r i Fsp,b
FBi .
dt
i
d

17

Because the Brownian force FBi is a random force, the above is a stochastic differential
equation, otherwise known as a Langevin equation. While it is beyond the scope of this
review, we note here that when nonlinearities are present there are subtle issues associated with the rigorous formulation of stochastic differential equations, alternative inter ttinger
pretations of which are those of Ito and of Stratonovich, as is explained by O
1996b.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

11

2. Spring force

The total spring force acting on any bead i, other than an end bead (i 1,N), is
sp
Fsp
Fsp,b
i
i Fi1 ,

18

where Fsp
i is the force that spring i exerts on bead i. For the end beads,
Fsp,b
Fsp
i
1 ,

sp
Fsp,b
FN1
.
N

19

sp
The spring force is a function F sp
i F i (R i ) of the extension R i of the spring, where
R i ri1 ri is the distance between bead i and bead i1; Ri ri1 ri is the
spring vector. The simplest spring force law is that for a Hookean, or linear, spring, which
is

Fsp R HR

20

Hookean spring ,

where the linear spring constant H is given by


H 2kBTs2; with s2

3
2N K,s b 2K

21

A Hookean spring is infinitely extensible; i.e., the force remains bounded for any finite
extension of the spring. A more realistic spring law should produce an asymptotically
large force when the spring is stretched to its full extension L s b K N K,s , where L s is
the fully extended length of a piece of chain represented by a single spring; i.e., L s
L/N s . For a freely jointed chain, a statistical mechanical calculation yields for the
spring force law the inverse Langevin function Bird et al. 1987, L1 ( ):
Fsp R

kBT
bK

L1

Ls

22

Here, the Langevin function is given by


1
L coth ,

23

and coth (ee)/(ee) is the hyperbolic cotangent. The inverse-Langevin


spring law is plotted in Fig. 2. Note that the force grows linearly with extension for small
and modest extensions; i.e., the spring law reduces to the Hookean form for fractional
extension R/L s less than around one-third. However, for large extensions, the force
grows rapidly, approaching a singularity at full extension, R/L s 1.
Since the inverse of the Langevin function is not an analytic function of the spring
end-to-end vector R, analytic approximations to it are frequently used. One such approximation is the Warner spring law:
Fsp R

2 s2 k B T
1 R/L s 2

H
1 R/L s 2

R.

24

A more accurate analytic force law is the Cohen 1991 Pade approximation:
Fsp R

H 3 R/L s 2
3 1 R/L s 2

R,

25

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

12

LARSON

FIG. 2. Elastic spring force versus molecular extension for the freely jointed chain which is given by the
inverse Langevin function, Eqs. 22 and 23, the Warner spring, Eq. 24, the Cohen Pade approximation to
the inverse Langevin function, Eq. 25, and the MarkoSiggia force law, Eq. 26. Note that the inverse
Langevin function and the Cohen Pade function are almost identical. The normalized elongation is the
end-to-end distance R divided by the fully extended distance L.

which much more accurately approximates the inverse Langevin force law over the
whole range of extensions; see Fig. 2.
DNA and many other biopolymers have helical backbones that are bendable but do not
undergo the large torsional bond rotations about individual bonds that many synthetic
polymers do. The DNA double helix can experience torsional strain, however, causing
overwinding or underwinding of the helix. The backbone distance over which a semiflexible rodlike polymer is rigid is called its persistence length p . If the molecular length
L of a semiflexible rodlike polymer is much longer than its persistence length, then the
gross conformation of the molecule at equilibrium is that of a freely jointed chain with
N K L/(2 p ) Kuhn steps each of length of b K 2 p . Locally, semiflexible rod molecules are not well described as links with free hinges, but are better described by the
wormlike chain Yamakawa 1971, which is a flexible thin rod. The statistical mechanics
of the wormlike chain model lead to a force law that is well approximated by the
following MarkoSiggia spring law Marko and Siggia 1995:
Fsp

kBT 1
p

Ls

1 R

4 Ls

2
3

HL s

Ls

1 R

, 26
4 Ls

where H 3k B T/(b 2K N K ) 3k B T/(2L s p ) is again the force coefficient in the linear,


Hookean, regime. Although usage of the MarkoSiggia formula is now well established,
a somewhat more refined approximation can be found in Bouchiat et al. 1999. The
MarkoSiggia force law, like the inverse Langevin force law, has a singularity at full
extension; see Fig. 2. The singularity in force makes the Langevin equation 17 stiff, and
hence small time steps must be taken in fast flows if the simplest, explicit, integration
ttinger 1997 introduced a semi-implicit
schemes are used. However, Hechen and O
predictorcorrector method for solving the stochastic differential equation for the beadspring model that allows much larger timesteps to be taken. This method was refined by
Somasi et al. 2002 and extended by Hsieh et al. 2003 to allow inclusion of HI.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

13

The predictions of the bead-spring model are rather insensitive to the form of the
spring law for steady-state flows or start-up flows, Hur et al. 2000, but are sensitive to
the spring law for relaxation from a highly stretched state Schroeder et al. 2004; Shaqfeh
et al. 2004.
3. Brownian force

The random Brownian force fluctuates extremely rapidly, on the time scale of picoseconds, which is the rate of thermal bombardment of the polymer by the solvent. Over
a somewhat longer time scale of nanoseconds, these bombardments establish a Maxwellian distribution of molecular velocities. Over a somewhat longer, but still very short, time
scale of perhaps microseconds, there are so many random bombardments that the random
force imparted by them averages toward zero, except for a small drift force whose
magnitude gets smaller with the inverse square root of the averaging time. This is the
usual statistical rule for averaging out errors in random processes. Thus, the Brownian
force averaged over a time scale, dt, can be represented by a random number with zero
mean, distributed according to
FB

6k B T
dt

1/2

n,

27

where n is a random three-dimensional vector, each component of which has a magnitude


uniformly distributed between 1,1, generated from random numbers Grassia and
Hinch 1996. The square root of the time step, dt, appears in the denominator of Eq. 27,
as expected from the statistical rule for averaging noise, and the factors of k B T and
appear because of the fluctuationdissipation theorem, which relates Brownian motion to
drag force. A relationship must exist between Brownian motion and drag because the rate
of Brownian motion determines the diffusion coefficient, which is also reflected in the
magnitude of the drag coefficient. A more general expression of the fluctuation
dissipation theorem is Kubo et al. 1985:

FBt 0,
FBtFBt 2kBTtt,

28

where F B (t) or F B (t ) is a one-dimensional component of the random force at times t or


t , respectively, and (tt ) is the delta function. The factor of 2 in Eq. 28 is carried
over into Eq. 27, as is an additional factor of 3, which is the inverse of the average of
the square of a random number that is distributed uniformly over the interval 1,1. The
product of these two factors is the factor of 6 in Eq. 27.
The above provides a stochastic form for the Brownian force. An alternative method
of obtaining the Brownian force is to work with a molecular configuration distribution
function, ( ri ,t). This function is defined such that ( ri ,t)dri is the probability that
at time t a randomly chosen chain has bead coordinates that lie between ri and ri
dri , where dri is a small interval in multidimensional coordinate space. Thus,
c ( ri ,t)dri is the mass concentration of chains whose bead positions lie within the
interval dri at time t. Now the Brownian force, once it is averaged over all chains
whose bead coordinates lie within this interval, will be zero, except for a drift force
B
Fi which is left over because ( ri ,t) is not quite uniform over the interval. The
brackets denote an average over the Maxwell velocity distribution at fixed bead
positions. As a result of this drift force, Brownian motion will carry chains from configurations that are relatively numerous compared to their equilibrium frequency, to those

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

14

LARSON

that are less numerous. A statistical analysis shows that this Brownian drift term is
exactly analogous to the flux, per molecule, produced by the ordinary diffusion of small
molecules; that is, the drift force is given by

Fi k B T

ri

n ,

29

where the analogous term for small molecule diffusion is the gradient in chemical potential, as described in standard textbooks.
Inserting the stochastic form of the Brownian force, Eq. 27, into Eq. 17, we obtain
the simplest form of the Langevin equation for the bead-spring model:
dri
dt

r i

Fsp,b


6k B T
dt

1/2

30

ni .

Integrating this equation numerically, using a random number generator to produce the
sequence of random vectors n, gives the evolution of bead coordinates for a single
realization of a polymer molecule. To obtain the collective behavior of a dilute solution of
many molecules in a transient flow, one must solve this equation many times, each time
with a different sequence of random numbers to represent a different sequence of random
Brownian forces exerted on a different molecule. Then, these multiple realizations of the
chain dynamics can be averaged together to yield the ensemble-averaged behavior. For
steady-state flows, one can instead average a single set of equations representing one
molecule for a long period of time.
Alternatively, substituting the velocity-averaged Brownian force, Eq. 29, into Eq.
17 yields
1
kBT

n .
ri ri Fsp,b
i

ri

31

This equation cannot be solved by itself; we must combine it with the conservation
equation for probability:

ri

ri

ri

1
kBT
ri Fsp,b
.

ri

32

This is the Smoluchowski equation for the probability distribution function ( ri ,t),
a function of all coordinates of all beads, and time. It is impractical to solve for this
function numerically. However, one can take moments of the equation, such as the second
moment, which, after appropriate closure approximations, can be put into a closed form.
The derivation of these moment equations can be found in standard references Bird et al.
ttinger 1996b. These closed-form equations are much cheaper to
1987; Larson 1988; O
solve computationally than the Langevin equation, which requires averaging over ensembles of hundreds or thousands of molecules to obtain accurate results. However, the
Langevin equation readily admits nonlinear phenomena, which in moment equations
require closure approximations of sometimes dubious accuracy. Hence, recent advances
in computer power have spurred the use of the Langevin equation, especially for the
prediction of nonlinear phenomena, where the simplest and cheapest closure approximations are especially dubious. For the prediction of linear viscoelastic phenomena, the
moment equations are often quite accurate, however, especially when Gaussian approxi ttinger 1989; Wedgewood 1989; O
tmations are used to include fluctuation effects O
tinger and Zylka 1992. In fact, in the linear viscoelastic regime, moment equations are

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

15

sometimes preferable to Langevin equations, because the latter require extensive averaging to reduce noise, which in the linear viscoelastic regime is very large relative to the
weak signal. While variance reduction methods can be very effective in reducing noise
ttinger 1996, nevertheless the
in solutions to the Langevin equation Melchior and O
moment equations may, in many cases, still provide the most accurate predictions of
rheology in the linear viscoelastic regime.
4. Hydrodynamic interaction

As mentioned in Sec. II, HI is always important for long, flexible, polymer molecules.
To account for HI using the bead-spring model, the disturbance to the velocity field
produced by the forces that each bead exerts on the solvent must be used to correct the
velocity acting on every other bead. By Newtons second law, the net hydrodynamic
force exerted by bead j on the solvent is equal and opposite the hydrodynamic drag force
Fdj , exerted by the solvent on bead j. But this hydrodynamic force must be balanced by
the nonhydrodynamic spring and Brownian forces exerted on that bead, so that Fdj
(Fsj FBj ). This force produces a disturbance vi to the velocity field at the position
ri of another bead i; this disturbance is a linear function of the hydrodynamic force Fdj
exerted by that bead:
vi i j Fdj i j Fsj FBj ,

33

where i j , the hydrodynamic interaction tensor, is a function of the separation ri r j


between the two beads. The direct measurement of the disturbance velocity created by
movement a bead was recently achieved by Meiners and Quake 2000. When the disturbance velocity given by Eq. 33 is incorporated into the stochastic differential Eq.
17, one can show that Ermak and McCammon 1978:
dri
dt

ri

Di j Fsp,b
j
kBT

j1

1/2

j1

i j n j ,

34

where Di j (k B T/ ) i j I i j is the diffusion tensor. The tensor i j is introduced


into the term for the Brownian motion in the above because of the fluctuationdissipation
theorem relating the diffusivity of a bead to the magnitude of the random forces that act
on it. The overall diffusion tensor D and weighting factor are fourth-order tensors,
which means that each component Di j of the diffusion tensor and each component i j
of the weighting-factor tensor is itself each a 33 tensor. The fluctuationdissipation
theorem implies a square root relationship between Di j and i j :
Di j

l1

il jl .

35

A formula relating the components of i j to those of Di j that satisfies Eq. 35 can be


obtained by a Cholosky decomposition as follows Ermak and McCammon 1978:

D 1
1

1/2

36

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

16

LARSON

0,

Here, , , indicate the row and column positions of the element in the overall fourthrank tensors D and which, for example, are 9 by 9 tensors for three-bead chains. The
calculation of the tensor i j by the Cholesky decomposition is computationally very
expensive. Jendrejack et al. 2000 and Kroger et al. 2000 have implemented a faster
scheme developed by Fixman 1986 that takes advantage of the fact that in Eq. 34 only
i
j 1 i j n j , not i j itself, appears, and the former quantity can be computed more
quickly than the latter, unless the excluded volume interaction exceeds a critical strength,
in which case the Cholesky decomposition is faster; for details, see Jendrejack et al.
2000.
There is also a Smoluchowski form for the force balance with HI, namely:
ri ri

j Di j

1
kBT

Fsp,b
j

rj

n .

37

Because the diffusion tensor Di j depends nonlinearly on the separations between all the
bead coordinates, it is difficult to develop closed-form moment equations from the
Smoluchowski equation that are valid in the nonlinear flow regime, when HI is present.
For the linear viscoelastic regime, Zimm 1956 preaveraged the diffusion tensor over the
known equilibrium distribution of configurations, leading to closed-form moment equations. The solutions to these equations will be discussed shortly. In the nonlinear vis ttinger 1987 introduced consistent averaging of the HI tensor, which
coelastic regime, O
allows it to change as molecules are deformed, but these changes are averaged over the
entire ensemble of molecules, thereby ignoring fluctuations in HI produced by fluctuations in chain deformation. If consistent averaging is combined with a simplified and
apparently very accurate decomposition into normal modes introduced by Fixman
1966, simulations with hundreds of beads can easily be carried out Magda et al.
1988b; however, in these simulations, fluctuations in HI are neglected. Fluctuations can
be included within moment equations using a Gaussian approximation for the distribu ttinger 1989; Wedgewood 1989; Prakash 2002. Thorough comparisons
tion functions O
of moment equations with the Gaussian approximation against Brownian dynamics simulations of Langevin equations for multispring chains with multiple sources of nonlinearity
i.e., HI, EV, and nonlinear springs should be carried out to assess the accuracy and
computational efficiency of the former, relative to the latter, in simulations of real polymers. For shearing flows at least, the Gaussian approximation shows very good agreement with results from Brownian dynamics simulations for Hookean bead-spring chains
with HI Zylka 1991.
Forms for the diffusion tensor. There are several suggested forms for the diffusion
tensor Di j , which are approximations to the HI mediated by the fluid. The simplest of
these, the OseenBurgers tensor Oseen 1927; Burgers 1938; Bird et al. 1987, is derived
by assuming that the beads are well separated enough from each other that they can be
regarded as point sources of drag on the solvent:

Dii

Di j

kBT
8sRi j

kBT
6sa
I

I,

Ri j Ri j
R 2i j

38
,

i j;

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

17

where I is the unit tensor and Ri j is the separation vector between the ith and jth beads,
Ri j r j ri . The OseenBurgers tensor is not suitable for Brownian dynamics simulations, because it becomes nonpositive definite when bead separations are comparable to,
or less than, the bead radius.
A better approximation, that accounts in a crude way for the finite size of the beads, is
the RotnePrager tensor, in which one expression for bead-separation distances larger
than twice the radius of the bead, and another for separations less than this Rotne and
ttinger 1989:
Prager 1969; Zylka and O

Dii

kBT
6sa

I,
39

Di j

kBT

8s Ri j

Rij
2a

2a 2

3R 2i j

I 1

2a 2 Ri j Ri j
R 2i j

R 2i j

R i j Ri j Ri j
8 3Rij

I
,
3 4a
4a R 2
ij

for Rij 2a

i j;

for Rij 2a

Note that as the ratio of bead size to bead separation a/R i j becomes small, the Rotne
Prager tensor, Eq. 39, reduces to the OseenBurgers tensor, Eq. 38. The strength of
the HI between beads can be indexed by the following parameter

h*

s 36 3 k B T

1/2

12 3 1/2R s s

3 a

Rs

40

where H 3k B T/R s2 is the elastic constant of the spring, 6 s a is the drag coefficient of a sphere of radius a, and R s b K N K,s is the root-mean-square end-to-end
vector of a spring at equilibrium. The largest reasonable value for a/R s is around 0.5,
since for higher values the bead radius would be one-half of the average spring length
and the beads would therefore, on average, overlap. Hence, values of h * larger than
0.53/ 0.49 are not physically reasonable. As we shall see later, a value of h *
0.25 gives predictions that are insensitive to the number of beads used and in agreement with experimental linear viscoelastic data for polymers in theta solvents. For example, the experimental value of the FloryFox parameter for theta solvents, 2.5
1023, can be derived from the Zimm theory defined below by setting h * 0.267,
which is close to the special value h * 0.25 Bird et al. 1987.
Note, from the definition of h * , that one can rewrite the RotnePrager tensor as

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

18

LARSON

Dii

Di j

3
4

h*kBT

Rij
2a

kBT
6sa

I,
41

kBT
H

6sa Ri j

2a 2

3R 2i j

I 1

2a 2 Ri j Ri j
R 2i j

R 2i j

R i j Ri j Ri j
8 3Rij

I
,
3 4a
4a R 2
ij

for Rij 2a

i j;

for Rij 2a

In this form, one can choose to reduce or turn off HIs simply by arbitrarily reducing the
value of h * below that given by Eq. 40, or even by setting h * to zero. If one does so,
then the form for the HI included in the Brownian dynamics simulations is no longer
consistent with that expected for spherical beads. Since the assumption of spherical beads
is arbitrary, there might be pragmatic reasons for treating h * as a separately adjustable
parameter. One cannot, however, increase h * above the value given by Eq. 40, for this
would make the tensor Di j nonpositive definite, and its square root could then not be
obtained, thus precluding Brownian dynamics simulations.
5. Excluded volume

For good solvents, a repulsive force Fev between the beads must be added to Eq. 17.
The repulsive force can be given by the gradient of a potential U ev(R); i.e.,
Fev
i

j ri U ev ri rj .

42

For simulations of small molecules, one normally imposes a steep repulsive potential,
such as that of the famous Lennard-Jones potential, whose repulsive part is a twelfth
power in the separation of the centers of mass of the molecules. A Lennard-Jones potential has in fact been used to simulate excluded volume effects in bead-spring models of
polymers Lopez Cascales and Garca de la Torre 1991. However, the steepness of the
repulsive potential should have little effect on the coarse-grained configurations of the
polymer chains. Hence, for simulations of polymers, it suffices to impose softer potentials, which then allow larger simulation time steps than do steep, hard, repulsive potentials, such as that of Lennard-Jones. It is convenient, then, to use an excluded volume
force that remains bounded, so that forces never become singular. A potential that decays
exponentially with separation serves this purpose, namely Jendrejack et al. 2002:
ev

U R

1
2

2
v k B TN K,s

2 R s2

3/2

exp

9R2
2Rs2

43

where, again, R s2 N K,s b 2K is the mean-square end-to-end length of a spring. A similar


ttinger 1996a, b; see also Prakash
potential had been proposed earlier for polymers by O

and Ottinger 1999; and Prakash 2002:

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

Uev R

z*
d*3

k B T exp

HR2

2kBTd*2

19

44

ttinger form we make the identification that the


The two potentials are identical if in the O
dimensionless range of the potential is d * 1/3, and its dimensionless strength is
1/2
3
z * 1/2(3/2 ) 3/2N K,s
v /b K . The interaction volume v is most appropriately defined as
a microscopic volume, proportional to the volume occupied by a single Kuhn step of the
chain, that remains invariant for a given polymer/solvent combination as one changes
either the molecular weight of the polymer and hence N K ) or the number springs used to
represent that polymer and hence N K,s ). The influence of the EV on rheological properties grows with the parameter z z * N s1/2 v /b 3K N 1/2
K , until a saturation is reached at
high values of z Prakash 2002. For shorter chains below the saturation condition, both
parameters z * or, equivalently, v ) and d * affect the bead-spring model predictions
Prakash 2002. The EV force can be incorporated into either the Langevin or the Smoluchowski form of the force-balance equation. As usual, incorporation into the Smoluchowski form necessitates invoking closures to obtain solutions.
The antithesis of EV interactions is polymerpolymer attraction, or, equivalently,
polymersolvent repulsion. If polymerpolymer attraction is weak, then its effect is
typically assumed to be equivalent to a weakening of the EV interaction, and is usually
modeled only in this indirect way. Of course, if the polymerpolymer attraction or
polymersolvent repulsion becomes strong enough, one crosses the theta point, and the
polymer dimension will shrink below the theta size, leading to a collapsed coil. Solvents that are so bad that the coils collapse below theta dimensions usually produce
precipitation of the polymer, unless the polymer molecular weight is low or its concentration is very low, below the overlap concentration. To model such a situation, one must
include net attractive interaction between beads. Simulations with bead attraction yield
collapsed coils that, when stretched, unravel abruptly into stretched filaments under
constant-force conditions and into ball and chain or tadpole configurations under
constant-stretch conditions Halperin and Zhulina 1991; Cooke and Williams 2003. This
behavior is akin to a first-order phase transition, with the ball and chain state resembling
a coexistence of coil and stretched states, but within the same molecule. Stiff polymers in
poor solvents avoid completely collapsed states because of the penalty for molecular
bending, and instead at equilibrium are predicted to form exotic shapes, such as equilibrium torii and nonequilibrium racquets Schnurr et al. 2000; 2002; Montesi et al.
2004.
Experimentally, polymerpolymer attraction can be produced not only by use of a
poor solvent, but also through electrostatic effects, for example, by use of multivalent
counterions in polyelectrolyte solutions. The multivalent ions bridge two like charges on
the polymer chain, causing the chain to attract itself. This phenomenon is used to condense semiflexible DNA molecules into a compact torroidal shapes for transport into cells
for gene therapy, for example. Because of the free energy of the counterions, polyelectrolyte systems show complex behavior even at equilibrium. Such phenomena are beyond
the scope of this review, but the interested reader can pursue this topic through the
literature citations in a recent article by Zherenkova et al. 2003.
Finally, it might also be worth noting that polymerpolymer attraction typically is
longer ranged than excluded volume repulsion, even for uncharged polymers. Hence,
even when the attraction is not strong enough to collapse the chain below theta dimensions, it is in principle possible that under some situations it might not be accurately
modeled simply by dialing down the strength of the EV effect. As an example, one could

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

20

LARSON

in principle generate the theta state in a bead-spring model by including both shortranged beadbead repulsion and long-range beadbead attraction, with magnitudes balanced so as to reproduce the theta state. While the equilibrium statistical properties of
such a chain would hardly differ from one in which both the repulsive and attractive
interactions are neglected, it is possible that this will cease to be the case in highly
nonequilibrium situations. In general, it is fair to say that both EV interactions, and,
especially, polymerpolymer attractions, in highly nonequilibrium states, have received
insufficient theoretical attention.
6. Internal viscosity

It has long been theorized that resistance to deformation of a polymer molecule might
arise not only from the elasticity of the molecule and from the friction between the
polymer and the surrounding solvent, but also from friction within the polymer itself.
That is, there might be a dissipative force generated by bond rotations and other motions
required for the polymer to change its configuration. Such friction is called IV. One
would expect the presence of such a frictional force to be manifested by a contribution to
the stress that is dissipative, that increases with increasing flow rate, and that is independent of the viscosity of the solvent. For the bead-spring model, a simple form for the IV
force that meets these criteria was suggested long ago by Kuhn and Kuhn 1945:
Fi v

R
R
R2

R,

45

where is a phenomenological coefficient that in principle should be independent of


solvent viscosity and only depend on the polymer type and possibly its molecular weight.
This form, when linearly added to the spring force in Eq. 17, treats the connector
between two beads as a spring and dashpot in parallel, so that rapid changes in spring
length produce both a purely elastic and a purely dissipative response. An alternative
formula was proposed by Cerf 1957.
The need for such a term in a bead-spring model has never been conclusively demonstrated; indeed, attempts to find a clear rheological signature of such a term have so far
failed Fuller and Leal 1980; Larson 1988. However, recent efforts to model the dynamics of chromosomes, which are DNA bundles held together by proteins, may revive
interest in internal viscosity models Poirier and Marko 2002. Evidence for internal
viscosity was once thought to be manifested in the high-frequency viscoelastic response
of dilute polymer solutions, where a difference between the experimental dynamic viscosity and the high-frequency prediction of the bead-spring model was found for some
polymers Massa et al. 1971; Ferry 1980. However, further investigation revealed that
the sign of this additional discrepancy was occasionally negative Morris et al. 1988,
which obviously could not be caused by addition of a dissipative and hence positive
contribution to the stress. By examining spectroscopically the motion of the solvent
molecules themselves, it was finally realized that the discrepancy was caused by a
polymer-induced modification to the dynamics of the solvent, which can be thought of as
a modification of the solvent glass transition temperature Morris et al. 1988; Lodge
1993. A polymer such as polystyrene that has a higher glass transition temperature T g
than the solvent in which it is dissolved raises the T g of solvent in its vicinity, making it
more viscous. Polybutadiene, on the other hand, has a T g that is sometimes lower than
that of the solvent it is dissolved in, and the measured high-frequency viscosity in this
case is found to be lower than predicted; i.e., the deviation is negative in sign. Thus,

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

21

high-frequency deviations from the predictions of the ordinary bead-spring model should
not automatically be attributed to viscous contributions arising from motions of the polymer.
Although at this point, there is no good reason for considering internal viscosity to be
of fundamental significance in the dynamics of flexible polymers, one might attempt to
use an IV model to represent crudely the high-frequency effects left out of a coarsegrained bead-spring model. An IV model has been used, for example, to represent the
retarding effect of internal folds or polymer kinks on the unraveling dynamics of
polymers in strong extensional flows Larson 1990.
7. Self-entanglements

Another physical phenomenon sometimes entertained is that of SEs or intrachain


entanglements, in which an isolated polymer forms a knot, which restricts its ability to
unravel in a fast flow. Interchain entanglement is a well-established phenomenon in
concentrated polymers, even though a clear definition of an entanglement between two
or more chains is still lacking. Brochard and de Gennes 1977 have estimated that in a
good solvent knots should almost never occur, while in a theta solvent typical polymer
molecular weights would need to be in the millions before even a single knot becomes
likely at equilibrium. For an isolated chain, entanglements, or knots, can be precisely
defined for ring polymers, and one can use simulations of freely jointed ring polymers to
give an idea about the frequency of knot formation in linear polymers. Michels and
Wiegel 1986 find that the fraction of unknotted random-walk rings drops off roughly
exponentially with the number of Kuhn steps, such that at 300 Kuhn steps less than 40%
of the chains are unknotted. Similarly, Ten Brinke and Hadziioannou 1987, in a lattice
simulation of random-walk chains, found 40% unknotted configurations for rings with
160 steps, and noted that this length corresponds to a molecular weight of 120 000 for
polystyrene. Self-avoiding walks, on the other hand, impose an EV constraint, and are
much less likely to be knotted than random-walk chains. For example, Yao et al. 2001
found only a 0.4% knotting probability in self-avoiding closed walks with 1000 steps on
a cubic lattice, and estimated that knots would only become prevalent for chains containing 2.5105 steps, corresponding to a polymer molecular weight in the hundreds of
millions. The large difference in knotting probability for random-walk versus selfavoiding walks confirms the early estimates of Brochard and de Gennes 1977.
In linear chains, knots are not permanent, and their effect on polymer dynamics is
unknown. Also unknown is the effect of shear or extensional flow on knot formation, and
whether or not tumbling motions in shear might generate many more SEs than occur
under no-flow conditions. In long DNA chains whose ends are attached to beads that can
be manipulated by optical tweezers, self-knots can deliberately be created, and the movement of the knots along the chains have been studied and found to obey simple diffusive
rules Bao et al. 2003. Experiments have not yet provided any direct evidence for any
effect of internal knots on rheology, although the failure to achieve the theoretical high
plateau extensional viscosity in filament-stretching and other extensional flows has been
attributed to the arrest of polymer stretch in these flows due to putative SEs James and
Sridhar 1995.

D. Stress tensor
Once the dynamics of the chain have been solved, one can compute macroscopic
quantities, such as stress, birefringence, scattering, or other observables. The most im-

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

22

LARSON

portant of these is the stress tensor . The polymer contribution to the stress tensor is
given, in general, by the Kramers expression Bird et al. 1987; Larson 1988:
Ns

p v

i1

iv
Fsi Fev
i Fi Ri N s v k B TI,

46

where I is the unit tensor and v is here. The number of polymer molecules per unit
volume. The terms Fev and Fi v can be dropped when excluded volume and internal
viscosity, respectively, are absent. Further discussion of the formulas for the stress tensor
can be found in Larson 1988 and Bird et al. 1987. A method to calculate the birefringence from the bead-spring model can be found in Li and Larson 2000a.
E. The bead-rod model
We have so far concentrated mainly on the bead-spring model, which is the least
expensive computationally for long-chain polymers, and consequently the most frequently used. Within the considerations discussed in Sec. IV B, however, the bead-rod
model is sometimes a more accurate model. Early Brownian dynamics simulations for a
bead-rod polymer chain were presented in a pioneering study by Acierno et al. 1974
and by Liu 1989. Lius solution method, still prevalent Grassia et al. 1995; Grassia and
Hinch 1996; Doyle et al. 1997, relies on the method of Langrange multipliers, which are
in fact tensions in the rods, required to keep their lengths fixed. The method was earlier
used to keep atomic bond lengths fixed in MD simulations of n-alkanes Ryckaert et al.
1977, and in the context of MD simulations is known as the SHAKE algorithm. In a
Langevin equation, the constraint forces replace the spring forces of a bead-spring model,
so that Eq. 17, for example, becomes
d

1
ri ri Fcon,b
FBi ,
dt
i

47

T i ui T i1 ui1 ,
Fcon,b
i

48

with

where the vector u i is the unit vector connecting beads i and i1; i.e., ui Ri /R i
(ri1 ri )/R i . In Brownian dynamics simulations, the tensions T i enter the formula
for stress, necessitating noise reduction schemes to smooth the stress, which would otherwise fluctuate drastically to balance the equally drastically fluctuating Brownian forces
that pull on the rods. Time stepping must also be handled with care, where a midpoint
algorithm for calculating the rod orientation Fixman 1978; Liu 1989; Doyle et al. 1997
has been validated as a safe choice. A new method for maintaining bond length, called
LINCS, which is claimed to be three to four times faster than SHAKE Hess et al. 1997
relies on projection of the bead or atom motion into directions that respect the bondlength restrictions. Application of this method to polymer simulations might be worth
exploring.
The effects of HI on the behavior of bead-rod simulations has been explored by
Neelov et al. 2002. Because in a bead-rod chain, each bead corresponds to a single
Kuhn length of the polymer, and inclusion of HI becomes prohibitively expensive beyond
100 beads, the effective molecular weights that can be studied with the bead-rod chain
and full HI are still quite low. Nevertheless, with full HI, Neelov et al. 2002 were able
to observe close to the expected scaling law c N 1.5 for the dependence on number
of beads N of the critical extension rate c for a coil-stretch transition discussed more

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

23

ttinger 1994 have proposed general methods of


below. Both Fixman 1978 and O
including HI into bead-rod Brownian dynamics simulations. However, there are numerous subtle issues involved in simulations of chains with rigid constraints, as discussed in
detail by Morse 2004.
An initially puzzling observation is that the steady-state chain conformational distribution that emerges from typical Brownian dynamics simulations for the bead-rod i.e.,
Kramers chain in the absence of flow is not quite that of a random walk. For a two-bead,
three-rod Kramers chain, for instance, there is a modest preference for a 90 angle
between the rods over an angle of 0 or 180 Gottlieb and Bird 1976. This strange result
comes about because in statistical mechanics the equilibrium distribution is in general a
distribution not only over configurations but over momenta as well. Although the momentum or velocity distribution equilibrates very rapidly typically on ns time scales as
alluded to earlier in Sec. IV C 3, nevertheless, averaging over this distribution, in the
case of a bead-rod chain, leads to a nonuniform statistical weighting of the different
bead-rod configurations. Brownian dynamics simulations do not keep track of velocity
distributions, but collapse information on momentum into the white noise Brownian
force term. The bead-rod algorithm by Liu, perhaps coincidentally, achieves the correct
equilibrium weighting of chain configurations for the case of beads of equal mass, but
would fail to do so for beads of unequal mass Morse 2004.
While these subtleties may seem worrying, for chains of many rods, the overall configuration retains random-walk statistics perhaps with an effective Kuhn step length that
is slightly different from the rod length and, in reality, no real polymer is a chain of
perfectly rigid links. In fact, one could replace each rod of the bead-rod chain by a very
stiff spring in which the force is zero at a nonzero length equal to the desired rod
length; this is a so-called Fraenkel spring Fraenkel 1952. Then, since the momentum
space for Fraenkel springs is Cartesian, the averaging over momentum variables becomes
trivial, and a random-flight configuration distribution is recovered, so that, surprisingly, a
chain of Fraenkel springs has a slightly different configuration distribution than its counterpart chain of rigid rods, and this difference remains no matter how stiff the Fraenkel
spring becomes and how small the fluctuations in spring length become. Whether a very
stiff Fraenkel spring or a rod better represents the behavior of a real polymer bond or
segment can only be settled by consideration of quantum mechanical effects controlling
bond vibrations Morse 2004. At the coarse-grained level of many bonds, however, such
fine details wash out, and Kramers chain is usually a more sensible choice than is a chain
of Fraenkel springs, because stiff Fraenkel springs require very small time steps to resolve their vibrational modes, and are therefore computationally expensive. Remember
that Hookean or Fene springs differ from Fraenkel springs in that in the former, many
Kuhn steps are subsumed into a single spring, while a Fraenkel spring represents a single
Kuhn step length.
In principle, polymers of moderate molecular weight can be simulated equally well by
either the Fene bead-spring chain or the bead-rod or Fraenkel-spring model. In practice,
for short polymers, with say less than 100 Kuhn steps, accurate predictions cannot be
obtained with the bead-spring model, since a Fene or MarkoSiggia spring should
represent at least 1015 rods, so that the spring force law correctly represents the entropy
of the rod configurations within the spring. On the other hand, polymers containing many
more than 100 Kuhn steps cannot be simulated by the bead-rod model, because of the
computational load required in simulating a chain with so many beads and rods. Chains
with around 100150 Kuhn steps can be simulated by either method, and in these cases,
both the bead-spring model with 1020 springs and bead-rod model give roughly similar

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

24

LARSON

predictions Hur et al. 2000; Somasi et al. 2002. Comparisons of the predictions of
bead-spring and bead-rod models can be found in Somasi et al. 2002.
V. COMPARISON OF PREDICTIONS OF MODELS TO LINEAR
VISCOELASTIC DATA
The predictions of bead-spring models in the linear viscoelastic regime are usually
obtained by solving for the moments of the Smoluchowski equation. The cheapest methods computationally involve performing a normal mode analysis; see below. A normal
mode analysis is much cheaper computationally than Brownian dynamics simulations,
and, in the linear viscoelastic regime, is reasonably accurate. Also, in the linear viscoelastic regime, where deformations and stresses are small, long periods of averaging are
required to obtain accurate results from solutions of stochastic Brownian dynamics simulations, making BD simulations even less attractive. Using normal modes, moment solutions of bead-spring polymer models were already obtained 50 years ago, and since then
thorough investigations of their predictions have been made, with both HI and EV effects
included. In recent years, the emergence of very fast computers has allowed better solutions to be obtained, both through more accurate calculation of moments of the Smoluchowski equation, and through Brownian dynamics simulations of the Langevin equation.
Comparisons of solutions of moments of the Smoluchowski equation with predictions
from Brownian dynamics simulations of the Langevin equation for bead-spring chains in
the linear regime have generally yielded satisfactory agreement between the two Zylka
ttinger 1989; Zylka 1991; Prakash 2002.
and O
A. No hydrodynamic interactions, no excluded volume: The Rouse model
The equations of the bead-spring model take their simplest form when both HI and EV
are neglected. Since chain extension is small in the linear regime, it is usually a good
approximation to consider the chain to be Hookean, at least for long polymers. In this
in Eq. 17 is linear in extension Ri and the entire equation is
case, the spring force Fsp,b
i
then linear in Ri . The equations for the different beads are coupled, since each interior
bead feels the spring forces generated by neighboring beads on the chain, but the equations can be decoupled by matrix methods, resulting in a well-known set of normal mode
equations. These equations and the methods for obtaining them are well described in a
number of books Doi and Edwards 1986; Bird et al. 1987; Larson 1988; 1999 and we
will not rehash these derivations here. We simply point out that the bead-spring model
with no HI, no EV, and with linear springs i.e., no finite extensibility, is the Rouse
model, mentioned in Sec. IV B, and is characterized by a spectrum of relaxation modes,
whose relaxation times i are given by Eq. 9.
B. Effects of hydrodynamic interactions: The Zimm model
The Rouse model is not quantitatively, nor even qualitatively, correct for dilute polymer solutions, even in the linear viscoelastic regime, because it neglects HI, which is
important even for relatively short polymers, and much more so for longer ones. In Sec.
III C 4, we discussed how HI can be included in either the Langevin or the Smoluchowski
equations. Zimm 1956 performed the first calculation of the linear viscoelastic spectrum
of the bead-spring chain containing HI, producing the first quantitatively accurate theory
for the linear viscoelastic properties of very long polymer chains in the absence of
excluded volume effects, i.e., for a theta solvent. The linear viscoelastic spectrum for the
Zimm theory cannot be expressed analytically, but for high mode number i, the relaxation
time scales as i i 3/2, while the mode strength is the same as in the Rouse model:

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

25

FIG. 3. Linear viscoelastic data for dilute polystyrene in two theta solvents. For frequencies greater than 10, G
and G scale as 2/3, in disagreement with the Rouse theory, which predicts a proportionality of 1/2 in this
frequency regime. Here G R G /vk B T and G R (G s )/vk B T, extrapolated to zero concentration see Eq. 50, and 0 0 s M /N A k B T figure from Ferry 1980; data from Johnson et al. 1970, used
with permission from the Society of Polymer Science of Japan.

G i vk B T cRT/M where here R is the gas constant. A consequence of the mode


spacing is that the storage and loss moduli G ( ) and G at a high-frequency scale as
G ( ) G ( ) 2/3, compared to G ( ) G ( ) 1/2 for the Rouse model.
Figure 3 shows that for high molecular weight polystyrene in a theta solvent, the predictions of the Zimm theory are accurate, at least up to dimensionless frequency 0
100.
The above predictions of the Zimm 1956 theory are for the case of no EV and
dominant HI. HI becomes dominant when the motion of one small portion of the chain
induces so much drag on other portions that the flow of solvent within the coil is greatly
suppressed. This leads to nondraining, i.e., the blocking of the imposed flow from penetrating deeply into the coil. Solvent thus tends to flow around the periphery of the coil,
analogous to solvent flow around a porous solid particle, only partially penetrating it; see
Fig. 4. To estimate the drag force on the coil and its scaling with molecular weight, the
coil can then loosely be thought of as a hard sphere. A more accurate picture is that the
solvent flow field matches the imposed macroscopic flow at large distances from the coil,
but as one moves into the coil, the flow gradients, on average, decay continuously toward
ttinger 1996a. In the nondraining or dominant HI
zero near the coils center of mass O
limit, the shape of this solvent velocity gradient profile becomes invariant to polymer
molecular weight and there is an equivalent hard-sphere hydrodynamic radius R H proportional to the radius of gyration R g of the polymer. The hydrodynamic radius is defined
as the radius of a hard sphere that gives the same center-of-mass diffusivity as the
polymer molecule in dilute solution; i.e.,

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

26

LARSON

FIG. 4. Illustration of effects of hydrodynamic interactions. In a, HI among particles is absent; the flow freely
drains through a swarm of very small, widely spaced, particles. In b, the particles are large enough and closely
spaced enough that the solvent mostly is deflected around the entire swarm due to hydrodynamic interactions
between particles from Larson 1988.

RH

kBT
6sDG

49

Because of the StokesEinstein relationship between the diffusivity and the frictional
drag coefficient, the viscosity contributed to a solvent by addition of a polymer coil is
equal to that of a hard sphere of radius R H . For a theta solvent, R H R g /1.479, according to the Zimm normal mode analysis; recent Brownian dynamics simulations give
R H R g /1.33 Kroger et al. 2000. In either case, the scaling with molecular weight of
the zero-shear-rate intrinsic viscosity of a polymer molecule is the same as that of the
radius of gyration, that is, as M 1/2, when HI is dominant. This relationship breaks down
when HI is not dominant, for then the relative penetration of the coil by the solvent
velocity field depends on solvent quality and polymer molecular weight.
Dominant hydrodynamic interaction implies that all viscoelastic properties, when
scaled using the Zimm theory, are independent of the polymer molecular weight, the type

FIG. 5. Intrinsic moduli G R and G R , defined in Eq. 50, as functions of dimensionless frequency 1
in decalin at the theta temperature 18.3 C for polystyrene of molecular weights 1.79105 open symbols,
1.05106 symbol with vertical line, and 5.50106 half-closed symbol. The solid line is from the Zimm
model with h * 0.25 and N 500 beads from Hair and Amis 1989; reprinted with permission from Macromolecules, Copyright 1989 American Chemical Society.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

27

of polymer, and the solvent, as long as it is a theta solvent. Figure 3 shows that curves of
G ( 0 ) and G ( 0 ) are in fact the same for polystyrene in two different theta solvents. Figure 5 shows that the scaled G and G curves are also independent of polymer
molecular weight, and in agreement with the Zimm theory, at least for molecular weights
of 1.79105 or higher. In this plot, the frequency is made dimensionless with the longest
relaxation time, 1 0 /S 1 , where for the Zimm theory, S 1 2.369. From Eq. 13,
and 0 M 1/2, we find that the Zimm theory predicts 1 0 /S 1 M 1.5, while for
the Rouse theory 1 M 2 . The intrinsic moduli G R and G R in Fig. 5 are defined
as
GR

M
RT

lim
c0

G
c

GR

M
RT

lim
c0

G
c

50

where c is the mass concentration of polymer.


But when is HI dominant? Intuitively, in Fig. 4, the strength of the HI will depend
on the number of beads and their size relative to their interparticle separation. In the
bead-spring model, these two factors are N and h * see Eq. 40, respectively. The
overall strength of HI in the bead-spring model is given by the combination parameter
h N 1/2h * . When h is large enough much greater than unity, we approach the limit of
dominant HI for which the linear viscoelastic properties of the chain scale in a simple
way with N; i.e., each relaxation time i is proportional to N 3/2/i 3/2, and neither depend
on h * , nor on h, at least for mode numbers i up to some limit that does depend on N and
h.
For real polymers, the strength of the hydrodynamic interaction will be proportional to
M 1/2, and there will be an effective parameter h whose magnitude is determined by a
mapping of the real chain onto the bead-spring model chain. For polystyrene in the
marginally good solvent Aroclor, it has been found empirically that accurate predictions
of linear viscoelastic properties can be obtained from the bead-spring model by taking a
single spring to represent a submolecule of molecular weight of around 5000 Daltons
Hair and Amis 1989; Amelar et al. 1991; Peterson et al. 2001, so that the number of
springs used in the model equals the chain molecular weight divided by 5000. The reason
for the correspondence between a spring and a particular rather large size of submolecule on a real chain is somewhat mysterious, but has been thought to be due to interference by the solvent in small-scale chain motion, when the solvent relaxation time is
comparable to the relaxation time for small subchain motion Peterson et al. 2001.
Larson 2004b recently offered an alternative explanation that relates the size of submolecule representing a spring to the cross-over point between smaller submolecules
whose relaxation is controlled by internal barriers to bond rotation, and larger ones that
are governed by simple frictional drag from the solvent.
If the number of beads N is not high enough to reach the asymptotic regime of large
h, then the longest relaxation time 1 will no longer be proportional to N 3/2, but will
depend on both N and h * . This is the regime of partial draining. We know from the
Rouse theory that for very small h h * near zero, 1 is proportional to N 2 , and that
the dependence on N weakens as HI is made stronger by increasing h * . It has been found
empirically that for a range of finite values of N up to N 100, the value of h * sets an
effective power-law exponent n in the equation 1 N sn for the dependence of 1 on N s
in the bead-spring model Thurston 1974:
n 21.4 h * 0.78.

51

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

28

LARSON

For h * 0.267 0.25, this formula yields n 1.5, i.e., which corresponds to a
theta solvent, while for larger h * , n 1.5. Thus, for h * 0.25, theta solvent scaling is
nearly attained by the bead-spring model with HI even for small N. For h * 0.25, the
dependence of 1 on N s follows a power law with exponent less than 1.5, while for
h * 0.25, the exponent is greater than 1.5. Of course, contrary to Eq. 51, for large
enough N( 100), the exponent 1.5 must be recovered for any value of h * , as long as
excluded volume is neglected. Experimentally, the value 1.5 is obtained for the exponent
in the dependence of 1 on molecular weight M for essentially all molecular weights
when excluded volume interactions are cancelled out by using a theta solvent. Hence,
h * 0.25 is often taken to be the appropriate value to use in the bead-spring model with
HI for polymers in theta solvents. This special value of h * at which dominant HI
scaling is achieved even for modest values of N can be defined mathematically as the
value at which the coefficient of the N 1/2 leading order corrections to the infinite-chain
limiting behavior of the rheological functions vanish identically. If fluctuations in HI are
neglected, this special value can be calculated exactly as h * 0.2424... Osaki 1972;
ttinger 1987.
O
If the solvent is better than theta, one obtains a power-law exponent greater than 1.5 in
the experimental dependence of 1 on molecular weight M. This occurs because EV
expands the coil, thereby reducing the strength of hydrodynamic coupling between different parts of the coil. This behavior can be mimicked in the bead-spring model without
EV by setting h * to a value less than 0.25. The lower the value of h * , therefore, the
better the quality of the solvent that can thereby be represented by the bead-spring chain.
Incorporating EV effects into the bead-spring model merely by reducing h * is cheating in the sense that while it mimics the change in the longest relaxation time produced
in real systems by EV, it fails to account for any expansion of the coil dimension, which
is obviously the most direct manifestation of EV. Nevertheless, the linear viscoelastic
predictions of the bead-spring chain with h * set less than 0.25, but otherwise no account
taken of excluded volume, have been shown to match well with the linear viscoelastic
properties of real polymers in good solvents. The best-studied example of this is polystyrene in the solvent Aroclor, a moderately good solvent for polystyrene. For this polymer and solvent, the linear viscoelastic properties are very accurately described by the
bead-spring model with no explicit EV, but with h * set to 0.15 Amelar et al. 1991. In
the nonlinear regime, however, this method of accounting for good-solvent effects is not
likely to be successful, especially for time-dependent viscoelasticity, because the rate of
deformation of a coil in a strong flow can be a sensitive function of its initial, equilibrium
size, which is not disturbed from its theta dimensions unless EV is explicitly accounted
for.
C. Effects of excluded volume
To account for the effect of EV on equilibrium coil dimensions, EV interactions must
be included directly in the equation for the bead-spring chain, for example by including
a beadbead repulsion, such as that in Eq. 43 or Eq. 44. The effect of this potential on
equilibrium i.e., no-flow coil size can be estimated using a simple energy minimization
argument first presented by Flory 1949; de Gennes 1979; Graessley 2004. This argument yields R g M v with v 0.6 for chains of high molecular weight M. Sophisticated
renormalization group methods yield a slightly more accurate exponent v 0.589 in
the asymptotic good-solvent limit Li et al. 1995; Graessley 2004.
More generally, the strength of EV interactions, per unit chain length, is controlled by
the EV parameter z * , defined in Sec. III C 5. For the bead-spring model, as the number

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

29

TABLE I. Scaling exponents, high-frequency moduli, and universal ratios.

No HI, no
EV
Rouse 1953
Dominant
HI, no EV
Dominant
HI,
Dominant
EV

GR
at high

GR
at high

R g /R H

R v /R g

0 /1

1.11( 1 ) 1/2

1.11( 1 ) 1/2

N 1/2

N1/6

1.645

3/2

1/2
3v1 0.8

2.09( 1 ) 2/3
Zimm 1956
1.36( 1 ) 5/9
Doi and
Edwards
1986

1.33
RG
1.56
RG

0.84
RG
0.73
RG

2.39

3v 1.8

1.21( 1 ) 2/3
Zimm 1956
1.14( 1 ) 5/9
Doi and
Edwards
1986

of beads N or the excluded volume parameter z * get larger, the good-solvent limit is
eventually reached when the combination parameter z z * N 1/2 becomes large enough.
The parameter z is analogous to the HI parameter h h * N 1/2. At high z, the rheological
properties of the bead-spring chain become independent of z Prakash 2002. Also, analogous to the HI parameter h * , there is a special value of z * for which the asymptotic
large-z scaling behavior is approximated even for modest N. From renormalization group
methods, this value of z * has been found to be z * 0.125 Schafer 1999; Prakash
2002, while Brownian dynamics simulations give a value around 0.26 to 0.29 Kumar
and Prakash 2003.
In general, the intrinsic viscosity 0 and the spacing of relaxation times obey the
power laws:

p lp,

0 M a;

52

where the exponents a and for the various regimes that include either HI, EV, or both,
are given in Table I. From the spacing of the relaxation times, the storage and loss moduli
at low frequency can be obtained by extrapolating Eq. 10 to low frequency, giving
GR 12

i i2;

GR 1

i i.

53

As in Fig. 3, the subscript R means that the moduli have been normalized by
N A k B T cRT/M , where N A is Avogadros number, R is the gas constant, and the
solvent contribution s has been subtracted from G .
Doi and Edwards 1986 have worked out formulas for G ( ) and G ( ) at frequencies well above the terminal regime where there is a power-law dependence on frequency.
Their results are presented in Table I. Also included in Table I are various universal
ratios in the high molecular weight limit. These ratios in Table I are equal to, or related
to, the following quantities:
URD

UR lim

Rg
RH

0s
s 4 R 3g /3

6sDGRg
kBT

2.5


Rv
Rg

54

M 0
4 N A R 3g /3

55

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

30

LARSON

U lim

0s
k B T 1

1
1 i i

S1 ,

56

where R v (3 0 M /(10 N A )) is the radius of a hard sphere that, in a solvent, would


produce the same intrinsic viscosity as the polymer, and here is the number of polymer
molecules per unit volume and c is their concentration in mass/volume. Note that U R
defined in Eq. 55 is related to the FloryFox parameter defined in Eq. 4 by U R
3(6) 3/2/4 N A .
Results for these quantities for polymers in theta solvents in the absence and presence
of dominant HI are given in the first two rows of Table I. The high-frequency behavior of
G and G are taken from the Rouse 1953 and Zimm 1956 theories. The Zimm theory
gives R H R g /1.479, while advanced renormalization group RG methods yield R H
R g /1.33 for no EV with HI dominant. For EV and HI both dominant, RG methods
yield R H R g /1.56 Oono 1983. The dimensionless ratio U R of M 0 to coil volume 4 R 3g /3, has been estimated by Monte Carlo simulations of Bernal et al. 1991 to
be around 1.10.1 in a good solvent dominant HI and EV. Under theta conditions, the
Zimm theory, which preaverages the HI tensor, gives U R 1.66425, while more accurate RG methods, which include fluctuation effects, gives around 1.48. A tabulation of
universal ratios of properties in the linear viscoelastic limit can be found in Kroger et al.
2000 and in Graessley 2004 and are partially reproduced in Table I. The renormalization group values for R g /R H and R v /R g given in Table I have been found to be in
agreement with experimental results within 10% for several polymers in both good and
theta solvents Graessley 2004.
The results in Table I are valid in the infinite chain length limit where universal
properties, insensitive to the local details of the polymer molecule or model of it, are
obtained. However, the universal dominant HI limit is reached even at low molecular
weights M around 10,000 or so for polystyrene, so this limit can be used for almost any
dilute theta solution. On the other hand, the dominant EV limit is reached only rather
slowly, requiring molecular weights in the millions, even in the best available solvents.
However, it has been found empirically that good-solvent behavior for finite molecular
weights can be predicted using universal scaling functions x (z), where x is some
property, and z is a measure of the strength of the excluded volume interactions Graessley 2004. These scaling functions relate the property x in a good solvent to that in a theta
solvent for the same polymer and molecular weight. For example, the radius of gyration
in good solvent is given by R g s (z)R g , where R g is the radius of gyration in a theta
solvent, and s (z) is the coil size expansion factor. For the intrinsic viscosity, there is
a similar scaling law that gives R v v (z)R v see Eq. 55 for a definition of R v ]. In
experiments, the quantity z is proportional to the square root of molecular weight and
to the binary cluster integral, which depends on the interactions between solvent and
polymer; see Yamakawa 1971 or Graessley 2004. In fact, Florys 1949 theory relates
s (z) to z by the formula s5 s3 2.60z. However, since this formula is not particularly accurate and requires evaluation of the cluster integral to calculate z, it is simpler
to use the empirical expression Graessley 2004:
px
x 1M/M
x ,

57

where M
is a cross-over molecular weight at which EV effects emerge and p x is a
x
universal exponent. For s , p s 0.092, while for v , p v 0.079. The cross-over
molecular weights M
vary somewhat from one property to another, and vary a great
x

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

31

deal from one polymersolvent pair to another. For polystyrene in toluene, for example,
M s 6420, while M
12,800. Tabulation of values of M
and p x for various
x
v
properties and various polymer/solvent pairs can be found in Graessley 2004. More
rigorous RG expressions for x are discussed in Schafer 1999 and Brownian dynamics
predictions are compared favorably to these in Kumar and Prakash 2003.
VI. COMPARISON OF PREDICTIONS OF MODELS TO NONLINEAR
VISCOELASTIC DATA
Efforts to compute nonlinear rheological properties of dilute polymer solutions from
bead-spring models date back many years. Reviews of such efforts prior to around 1995
ttinger
can be found in texts by Larson 1988, Bird et al. 1987, Larson 1999, and O
1996b, which in the main cover predictions of moment equations derived from the
Smoluchowski equation through various closure approximations whose accuracy in the
nonlinear regime is uncertain. In more recent years, fast Brownian dynamics simulations
have permitted quantitative comparisons to be made between experimental data and
predictions of the bead-spring model without closures. In the following, we compare
predictions of BD simulations with single-molecule data for fluorescently stained DNA
molecules and for dilute polystyrene Boger fluids, the two well-defined monodisperse
systems for which there is abundant data. While the focus here is on comparisons of
Brownian dynamics simulations of polymer dynamics in flow to experimental data, we
note that, in addition to the papers referred to below, additional physical insight can be
drawn from the Brownian dynamics simulations of bead-spring and bead-rod chains of
Hernandez Cifre and Garca de la Torre 1999; 2001, and Rzehak and Zimmermann
2002.
In Sec. V, we explained that, for long enough chains, there are universal viscoelastic
properties that are independent of polymer and solvent physicochemical properties, and
depend in a universal way on chain length for either good or theta solvents. Since these
universal properties do not depend on local molecular details, they can be captured
quantitatively by bead-spring or bead-rod models, whose local structure e.g., beads and
springs, is not physically realistic. For the linear viscoelastic regime, such universal
properties that are captured by bead-spring models are to be found in Table I. In the
nonlinear regime, universal properties can be expected to hold up to some critical molecular deformation that might be large for very high molecular weight chains, but will
eventually be exceeded as the chains approach full extension Prakash 2002; Kumar and
Prakash 2003 and local details again could become important. At high enough deformations, then, predictions of bead-spring and bead-rod models can depend on the number of
beads and rods, or other ambiguous model parameters, and so the predictions are model
dependent and hence less trustworthy. Nevertheless, we shall find that in many cases
especially for DNA, with reasonable choices of parameters, bead-spring and bead-rod
models can give agreement with experimental data even well outside of the regime of
universal properties.
A. Deoxyribonucleic acid solutions
1. Bead-spring parameters

The first quantitatively accurate predictions of polymer deformation under flow were
achieved for dilute solutions of long, fluorescently stained DNA molecules. While early
imaging of single molecules of DNA was carried out 20 years ago, DNA molecules were
first imaged in well-defined flows in the mid 1990s by Chu and co-workers Perkins
et al. 1995; 1997; Smith and Chu 1998. The simplest flow studied was a uniform flow

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

32

LARSON

with the DNA molecule held in place at one end by an attached bead confined to an
optical trap Perkins et al. 1995. In subsequent work, freely convected DNA molecules
were imaged in well-defined planar extensional, simple shear, and mixed extensional/
shear flows. The DNA molecule most commonly used was -phage DNA, a biologically
derived, and hence perfectly monodisperse double-stranded DNA molecule containing
48,502 base pairs, with a molecular weight of 31.5 million Daltons, and a total length of
L 16.4 m, when unstained by dye. Using direct, single-molecule experiments, unstained double-stranded DNA has been found to follow the MarkoSiggia spring law, Eq.
26, with a persistence length p of around 0.054 m Bustamante et al. 1994. However, when fully stained with an intercalating dye YOYO-1 that inserts itself between
the stacked DNA bases, the DNA length increases to L 21 22 m Perkins et al.
1995; 1997 and the persistence length has been reported to be around 0.066 m Quake
et al. 1997, although this increased value for the persistence length of stained DNA over
that of unstained DNA has not been confirmed.
It is also noteworthy that DNA is a polyelectrolyte with two charges per base pair.
While counterion condensation can reduce the charge, still the charge is high enough to
cause the molecule to expand in salt-free water. The coil size of DNA decreases monotonically with increased electrolyte concentration, reaching a nearly constant value once
the concentration of NaCl reaches 10 mM Hagerman 1988. Consequently, most singlemolecule studies of DNA molecules are carried out in solutions with electrolyte concentrations of at least 10 mM. At these salt concentrations, the Debye double-layer thickness
is very small nanometers so that the DNA chains, plus a thin sheath of water surrounding them, can be considered effectively neutral, and their rheological behavior should be
and apparently is very similar to that of neutral polymers. Staining DNA with positively
charged dye, such as YOYO-1 Larsson et al. 1994, reduces its charge, which could
decrease its effective persistence length to some extent, perhaps offsetting or canceling
the tendency of dye to stiffen the chain. We also note that staining with YOYO oxazole
yellow dimer and TOTO thiazole orange dimer makes DNA susceptible to photocleavage under exposure to light, causing single-strand nicks, which accumulate, leading to
kerman and Tuite 1996. In exdouble-strand breakage during continuous exposure A
periments with stained DNA, shuttering of light must be carefully controlled to avoid
photocleavage artifacts.
DNA coil size in water follows good-solvent scaling laws. Over a range of sizes from
4361 to 309,000 base pairs, corresponding to stained contour lengths of 2 to 140 m, the
center-of-mass diffusivity of TOTO-1-dye-stained DNA in water has been found to range
from 1.94 to 0.14 m2/s, following the law D G L 0.6610.016 Smith et al. 1996.
From this, one can obtain an estimate of the radius of gyration R g using the scaling law
for good solvents Smith et al. 1996. For -phage DNA stained with TOTO dye, the
measured value of D G 0.470.03 m2 /s then yields R g 0.76 m, using U RD
1.56; see Eq. 54 and Table I. Smith et al. obtained R g 0.73 m by using U RD
1.48, a formula for a theta, rather than a good, solvent. Although the good-solvent
scaling law is assumed to apply to DNA, EV effects, like the effects of HI, are relatively
modest for -phage DNA, again because the stiffness of the molecule leads to expanded
coil dimensions, reducing the probability of intermolecular contacts. For purposes of
carrying out simulations over a single, or very narrow range, of molecular lengths, the
modest EV effects in DNA might be absorbed into the value of the DNA persistence
length, leading to accurate modeling of the DNA coil size for a given molecular length,
and this seems to suffice for correct prediction of -phage DNA deformation behavior
Larson et al. 1999; Hur et al. 2000. However, this approach will lead to a perhaps
weak dependence of this effective persistence length on the molecular weight of the

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

33

DNA. For -phage DNA stained with TOTO-1 dye, the value of R g 0.73 m can be
combined with the formula R 2g L p /3 to infer a value of effective persistence length
that includes the effect of EV, yielding p 0.076 m, which is somewhat larger than
the value of 0.066 m inferred by Quake et al. 1997 for DNA stained with YOYO-1. In
general, more study is required of the effects on DNA properties of intercalating dyes
such as YOYOs TOTOs, and others, and of nonintercalating dyes, such as Syber green
Schroeder et al. 2003.
-phage DNA can be modeled by either a bead-spring or a bead-rod chain. Since
highly stained DNA contains around 21/0.066 318 persistence lengths, and the Kuhn
step length is twice the persistence length, around 150160 rods are needed if the molecule is to be modeled accurately by the bead-rod model Hur et al. 2000. For the
bead-spring model, a minimum of around ten beads has been found necessary to obtain
accurate predictions, while around 1520 beads is the maximum that can be used, since
each spring needs to represent around ten or more Kuhn steps if the spring law is to be
accurate Larson et al. 1999. Both bead-rod and bead-spring models have been very
successful in predicting DNA conformations in various flows, as discussed below.
We note that if a given molecule is represented by increasingly greater numbers of
beads and springs, the effective flexibility of the chain is increased. In order to offset this
increased flexibility, and so recover the original forceextension relaxation of the molecule, an increased persistence length, eff
p , which approximately restores the original
flexibility of the polymer, can be used Larson et al. 1997, 1999. Specifically, eff
p is
chosen so that the correct value of the mean squared end-to-end distance ( R 2 0
2.77 m) of the molecule is obtained. For the stained -phage (L 21 m) with
eff
p 0.066 m, for N 5 it has been found that eff
p 0.075 m; for N 10, p
eff
0.082 m; and for N 20, p 0.096 m Chopra and Larson 2002. For a DNA
chain of contour length 21 m, the maximum number of beads that is advisable is
restricted to N 20, so that the subchain length L/(N1) is no less than around 1 m
about 15 times p ), and the adjustment to p required to offset the artificial flexibility
introduced by the beads remains modest. A detailed discussion of optimal choice of the
spring law can be found in Underhill and Doyle 2004.
A peculiarity of -phage DNA is its relatively low extensibility. Although a very long
molecule, it is also rather stiff, leading to a very open coil configuration at rest, and a
large radius of gyration, around R G 0.73 m Smith et al. 1996, implying that the
root-mean-square end-to-end distance of the chain is around R 2 1/2
0 6R g
1.8 m. Hence, the extensibility of -phage DNA, which can be defined as
L/ R 2 1/2
0 , is around 12, a rather modest number compared to that for high molecular
weight synthetic polymers. For example, an extensibility of 12 is attained for polystyrene
in a theta solvent when its molecular weight is only around 100,000.
The relatively low extensibility of -phage DNA implies that HI does not change
much when the molecule is stretched. For extensibilities of order 10 or less, the effects of
deformation-induced changes in HI is relatively modest Larson et al. 1997; Jendrejack
et al. 2002; Hsieh et al. 2003. As a result, we shall see that accurate predictions of
lambda DNA configurations are achievable even when HI is neglected, as long as the
bead drag coefficient is adjusted to achieve an accurate prediction of the longest DNA
relaxation time. This conclusion is reinforced by a recent careful study of short- and
long-time diffusion using fluorescent labeling of an end of the DNA molecule. This study
revealed a mean-square displacement that at short times scales as r 2 t 2/3 for singlestranded DNA ssDNA, and as r 2 t 1/2 for double-stranded DNA dsDNA Shus-

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

34

LARSON

terman et al. 2004. The former scaling is consistent with Zimm theory, in which HI is
dominant, while the latter is consistent with Rouse theory, implying very weak HI, for the
dsDNA molecules studied, which were up to one-half of the length of a -phage molecule. These results show that the greater stiffness roughly 20-fold greater persistence
length of dsDNA, relative to ssDNA, leads to much larger coil dimensions, and much
weaker HI, for dsDNA relative to ssDNA of a given molecular length. For a long enough
dsDNA molecule, however, HI eventually must become important. dsDNA molecules of
order 1 mm length !, have been studied in extensional flow; for these, the strength of HI
is comparable to that of high molecular weight flexible polymers Schroeder et al. 2003.
For stained -phage DNA, the longest relaxation time is frequently measured by
imaging the molecule as it relaxes from a stretched configuration produced, for example,
by imposing a very fast flow. From these images of relaxing molecules, the stretch x of
the chain, which is the projected length of the molecule in the flow direction, is obtained
as a function of time. Since the relaxation of the chain is accelerated when the chain is
stretched beyond the limits of the Hookean-spring approximation, only relaxation data at
times for which the stretch has decayed to 30% or less of the fully extended length L, i.e.,
x/L 0.3, are used in these fits. Within this limit, the square of the stretch for an
ensemble of molecules is plotted semilogarithmically against time. The value of 1 is
then the best fit constant in the expression x 2 x 2i exp(t/1)x20 , where x i is the
initial stretch, x i 0.3L, the brackets represent an ensemble average, and the
subscript 0 denotes the equilibrium value achieved when the chain is completely
relaxed. Simulations show that this method of obtaining 1 gives the nearly the same
relaxation time within 10% as is obtained from the fits of the late-time tail of the mean
square end-to-end vector R 2 versus time; the latter method satisfies the classical definition of longest relaxation time Larson et al. 1999. More typically, rather than ensemble averaging the value of x 2 at each value of time t, values of 1 are obtained for
each molecule by fitting relaxation of x 2 versus time and then the values of 1 for these
individual molecules are averaged together. This method is easier to carry out in practice
and, according to simulations, gives nearly the same result as is obtained from plots of
the ensemble average x 2 (t) Hsieh 2003.
Measurements for -phage DNA indicate that if the electrolyte concentration is high
enough typically 10 mM or so of NaCl; see Hagerman 1988, 1 is proportional to the
viscosity of the solvent, which typically consists of water containing electrolytes, chelating agents, antioxidants, and solvent viscosifiers, such as high concentrations up to 65%
by weight of sucrose. For stained -phage DNA, several measurements from the Chu
group show that 1 roughly follows the formula 1 0.094 s with s in centiPoise and
1 in s. Perkins et al. 1997; Smith and Chu 1998; Fang and Larson 2004. However,
values of 1 twice as large as this were recently reported by the same group Babcock
et al. 2003 and values 50% higher were reported by Li et al. 2004. Most of these
studies were carried out in solutions heavily viscosified with sugar, rendering the solvent
viscosity very sensitive to temperature and small levels of water evaporation, which
might well account for these occasionally measured anomalously large values of 1 .
For conventional polymers, the traditional method for obtaining relaxation times is
through rheometry of bulk, dilute, solutions. Massa 1973 reported intrinsic viscosities
of T2 ( 0 30,000 ml/mg), and B. subtilis ( 0 138,000 ml/mg) DNA molecules. For these latter two molecules, the intrinsic viscosity scales with molecular weight
to roughly the 0.56 power, consistent with a moderately good solvent, and extrapolating
this relationship to -phage DNA, with molecular weight M 3.15107 , we find that
0 for -phage DNA should be around 13,600 cm3/gm. The Zimm prediction, Eqs. 3

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

35

and 4, yields 0 R 2 3/2


0 /M . The FloryFox parameter for a theta solvent can
be obtained from Eq. 55, using the RG value of U R 2.5(R v /R g ) 3 ; this yields
2.551023, which matches the experimental value for a theta solvent, 2.5
1023. We note, however, that, from Table I, in a good solvent dominant HI, dominant
EV, the value of should be 1.671023. For unstained -phage DNA, for which the
contour length is L 16.4 m and persistence length is p 0.054 m, we find
R 2 0 2L p 1.77 m2 . Using 1.671023, we then obtain 0
12,500 ml/gm, consistent with the value of 0 extrapolated from the other unstained DNA molecules cited above. Combining this computed value, 0
12,500 ml/gm, with Eq. 11, yields 1 0.069 s, for -phage DNA in a 1 cP solvent.
This value of 1 is consistent with that for unstained T7 DNA, which is only modestly
shorter than -phage DNA, and has a creep-rheometry-measured relaxation time of 0.046
s in water near room temperature Klotz and Zimm 1972. By using the scaling law 1
L 1.66, one then expects unstained -phage DNA to have a relaxation time of around
0.067 s. This is not only close to the value, 0.072 s, inferred from the intrinsic viscosity
0 13,600 cm/gm extrapolated from the experiments of Massa 1973 cited above,
but also close to the range of values 1 0.058 0.068 s that one infers for unstained
-phage DNA from other intrinsic viscosity measurements, and from birefringence and
dichroism measurements of DNA in water, correcting for the different lengths using the
above scaling law see footnote 8 in Perkins et al. 1997. For example, Thompson and
Gill 1967 used birefringence relaxation to measure a longest relaxation time, extrapolated to zero concentration, of 0.45 s for T2 DNA (M 1.05108 ) in buffered water. A
similar value, 0.45 s, was obtained for T4 DNA (M 1.25108 ) by relaxation of
dichroism, and from the scaling law 1 M 1.6 determined from measurements with
smaller DNA fragments Callis and Davidson 1969. Since T2 and T4 DNA are around
three to four times longer than -phage DNA, the scaling law 1 L 1.6 implies that
unstained -phage DNA should have a relaxation time of around 0.050.07 s. This value
is somewhat smaller than the value 0.094 s that we infer for stained -phage DNA from
the formula 1 0.094 s given above, with s in cP. However, an enlarged value of 1
for stained DNA, relative to unstained DNA, is consistent with the expanded coil size
expected for stained DNA. We note that Shrewsbury et al. 2001 obtained much smaller
values of 0 and 1 using a Vilasic rheometer, but they bypassed the usual extrapolation to zero concentration, because of what was believed to be effects of inadequate
counterion concentration at the low DNA concentration needed to attain the high electrolyte coil dimensions of DNA. The effects of staining, electrolyte concentration, added
viscosifiers, antioxidants, etc., give plenty of scope for variability in the relaxation time,
even beyond the usual experimental uncertainties.
Once the coil size and relaxation time of the DNA molecule are known, the bead-drag
coefficient in the bead-spring model can be set. Using DNA length L 21 m and
persistence length p 0.066 m for YOYO-stained -phage DNA, we obtain R 2 0
2L p 2.77 m2 . If HI and EV are neglected, the longest relaxation time of the
bead-spring model is given by the Rouse formula, 1,R coil R 2 0 /6 2 k B T Eq. 9,
with i 1], which can be made to match the experimental DNA relaxation time say
1 0.1 s] by choosing coil /k B T 2.13 s/ m2 for a 1 cP solvent. This total drag
coefficient is then divided equally among the N beads of a bead-spring chain, so that the
drag coefficient per bead is then coil /N.
However, the bead-spring model without HI and EV does not follow good-solvent or
HI scaling for the dependences of the coil size and longest relaxation time on number of

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

36

LARSON

beads N. This means that if longer DNA chains are used, one cannot simply increase the
number of springs proportionately, leaving the parameters of the model fixed, and expect
to correctly predict R 2 and 1 for other DNA lengths L. Instead, one must adjust p
and separately for each chain length to achieve a match to the experimental values of
R 2 and 1 .
In principle, one might expect that the correct scaling laws could be obtained if EV
and HI are included in the bead-spring model, as has been attempted by Jendrejack et al.
2002. They showed that if a bead-spring model is used with N s 10 springs to represent a 21-m-long stained -phage DNA in a solvent with viscosity 43.3 cP, then
correct predictions for somewhat longer DNA molecules up to around six times longer
are obtained by proportionally increasing the number of springs, using the MarkoSiggia
spring law Eq. 26 with Kuhn length b K 0.106 m, the RotnePrager tensor Eq.
39 with bead radius a 0.077 m, and an exponential excluded volume potential
1/2
Eq. 43 with v 0.0012 m3 . Using R s N K,s
b K 0.472 m for the root-meansquare spring extension at equilibrium, these parameters give h * 3/ a/R s
0.16). If, however, one wishes to adjust the number of springs at fixed DNA molecular weight, then the parameters a and v must be readjusted. Furthermore, Schroeder et al.
2004 studied DNA molecules of lengths up to 1.3 mm 60 times longer than -phage
DNA and found that the best values of h * and v seem to shift somewhat for very
long DNA molecules. Thus, there are still questions about the best method of choosing
parameters of the bead-spring model for DNA, or whether any single set of parameters
will work for all DNA lengths.
One can estimate the translational drag coefficient for uniform flow past a fully extended DNA chain by using the Bachelor formula for the parallel translational drag
coefficient of an extended filament Batchelor 1970; Li et al. 2000:

rod s

ln

L,

58

see Larson et al. 1999. For -phage DNA, with a length L 21 m and a diameter of
2 nm, in water at 24 C ( s 0.95 cP), this gives rod /k B T 3.2 s/ m2 for the drag
coefficient in the extended state, which is only modestly 50% higher than the drag
coefficient near the coiled state, coil /k B T 2.13 s/ m2 , calculated above. Thus,
deformation-induced changes in hydrodynamic interaction are weak for -phage DNA, as
noted earlier. Another way of reaching the same conclusion is to note that the HI parameter h N 1/2h * , with h * defined by Eq. 40, is closely related to the ratio rod / coil
and to the chain extensibility, which is not much greater than unity for -phage DNA.
Hence, in bead-spring or bead-rod simulations of -phage DNA, hydrodynamic interactions can be ignored if the drag coefficient for the beads is chosen to match the longest
relaxation time of the model chain to that of the real DNA molecule. For DNA chains
much longer than -phage, deformation-dependent HI begins to become more important
Larson et al. 1997; Jendrejack et al. 2002; Schroeder et al. 2003, as will be discussed
shortly.
2. Extensional flow

Once the bead-spring model parameters have been chosen as described above, a priori
simulations of DNA deformation under flow can be carried out. In extensional flows, the
streamlines are hyperbolic, and the transpose of the velocity gradient tensor can be

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

37

FIG. 6. Mean fractional stretch x vs Weissenberg number Wi in extensional and shear flows. The extensional
data are from Perkins et al. 1997 and the shear data from Smith et al. 1999 for -phage DNA solutions. The
lines are from Brownian dynamics simulations of Jendrejack et al. 2002; the dashed lines include the effect of
HI, while the dotteddashed lines neglect HI. The solid lines are a guide for the eyes from Jendrejack et al.
2002; reprinted with permission from Journal of Chemical Physics.

written such that only the diagonal components are nonzero. The most commonly explored extensional flows are uniaxial and planar extensional flows for which can be
written as

/2

/2

uniaxial extension

planar extension

59

60

0
0
0
Experimentally, a uniaxial extension flow can be produced by the filament-stretching
device Tirtaatmadja and Sridhar 1993; Orr and Sridhar 1999, while a cross-slot Keller
and Odell 1985 or a four-roll mill produces a planar extensional flow Fuller and Leal
1980. The molecular response should be nearly the same in planar and uniaxial extensional flows, especially at high strain rates , because both flows have an axis the x or
1 axis along which the molecules are strongly stretched at high . In both flows,
below a critical strain rate c , the chains remain only slightly deformed coils, while
above c there is a rather abrupt increase in the steady-state chain deformation. The
critical extension rate is given roughly by c 1/(2 1 ). Figure 6 shows experimental
measurements of the apparent steady-state molecular stretch averaged over an ensemble
of DNA molecules imaged in a crossed-slot device Perkins et al. 1997 along with
predictions from the bead-spring model. Both show onset of strong molecular stretch at
Wi 1 0.5, where we have here defined the Weissenberg number Wi as the product

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

38

LARSON

FIG. 7. Extension x versus Hencky strain t at two different Weissenberg numbers here labeled De for
Deborah number, from repeated experiments with -phage DNA by Smith and Chu 1998, and simulations
of Larson et al. 1999.

of the extension rate and the longest relaxation time. A related dimensionless group is the
Deborah number, which is defined as the relaxation time divided by a flow time that
characterizes the kinematics of the flow. If one chooses as the flow time the inverse of the
strain rate (1/ , say, then the Deborah number would become indistinguishable from the
Weissenberg number, and so sometimes the two dimensionless groups are treated as
synonyms. However, more properly, use of the Deborah number should be restricted to
transient flows or flows where a less trivial definition of the flow time is possible.
Steady-state stretch is achieved in extensional flows only at rather large Hencky
strains, t, of five or higher. The transient stretch at lower strains is highly variable
from one molecule to the next, as is shown in Fig. 7 at two different Weissenberg
numbers. This high molecule-to-molecule variability, even when the flow field and the
molecules are identical, has been called molecular individualism by de Gennes 1997.
Its source is the equilibrium fluctuations in coil size and shape, which are amplified by
the exponential separation of fluid elements in an extensional flow. The heterogeneity in
rates of chain unraveling is also reflected in the different modes of chain unraveling, two
of which, the dumbbell and fold mode, are depicted in Fig. 8. Other modes are
summarized in Fig. 9, along with their one-dimensional mass distributions. The frequencies of occurrence of the different modes in the experiments and simulations at high
extension rates are shown in Fig. 10. There is excellent agreement between the predicted
and the measured frequencies, both at high Weissenberg number Fig. 10 and at Weissenberg numbers close to the critical value for chain stretch, Wi 0.5 Larson et al.
1999. The simulations show that most of this heterogeneity, especially at a high extension rate, is due to the different initial configurations of the chains as they enter the
extensional flow, which predispose the chain to a folded, a kinked, or a dumbbell configuration.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

39

FIG. 8. Illustration of a dumbbell and b fold unraveling modes. The experimental images from Perkins
et al. 1997 and the simulated images from Larson et al. 1999.

For DNA molecules much larger than -phage, with molecular lengths as high as L
1300 m, or 1.3 mm !, the stretch versus Weissenberg number curve steepens, as
shown in Fig. 11, and hysteresis appears between the coiled and the stretch states, so that
the curve of average stretch versus Wi for increasing Wi differs from that for decreasing
Wi over a range of Wi Schroeder et al. 2003. This remarkable behavior was predicted
in 1974 by de Gennes, and is a consequence of the dependence of HI on polymer
deformation, sometimes called deformation-dependent drag. As noted in Sec. VI A 1, the
effective drag properties of -phage DNA change modestly when the coil is stretched
from a coil into a slender filament. In the coiled state, the translational drag coefficient
coil is proportional to the radius of the coil, which increases roughly with the 0.6 power
of molecular length L, while in the extended filament state, the drag coefficient rod is
proportional to L, with a weak logarithmic correction in the denominator; see Eq. 58.
For -phage DNA, the prefactors of these expressions are such that the drag coefficient in
the extended state is only 50% larger than in the coiled state. But, given the scalings just
described, for much larger DNA molecules, say L 1 mm 50 times longer than
-phage phage, this difference in drag coefficient becomes more substantial, so that

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

40

LARSON

FIG. 9. Typical configurations of DNA molecules during unraveling in an extensional flow, along with corresponding brightness profiles. The fold2/3 is distinguished from the fold2/5 in that in the former, at least
2/3 of the DNA molecule is in the folded region which then has twice the brightness of a single stretched strand
from Larson et al. 1999.

FIG. 10. Observed and predicted percentages of DNA unraveling modes during extensional flow at high
extension rate (Wi 1 55). The types of modes are HD: half dumbbell, DB: Dumbbell, K: Kink, FD25:
Fold2/5, FD23: Fold2/3, and Coil, all defined in Fig. 9. The experimental results were obtained by from
videotaped images of DNA, while the predicted percentages were from Brownian dynamics simulations. The
results from the simulations were categorized both by a computer-automated method SIM automated, and by
visual inspection by the experimentalist Smith, using computer-generated images of the simulated chains. In
both experiments and simulations, 100 chains were used from Larson et al. 1999.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

41

FIG. 11. Steady-state fractional stretch x /L vs Weissenberg number Wi 1 , for DNA molecules of
contour length 1.3 mm, showing hysteresis between data for gradually increasing strain rate and those for
decreasing strain rate, obtained in an extensional flow in a crossed-slot device. from Schroeder et al. 2003;
reprinted with permission from Science 301, 15151519, Copyright 2003 AAAS.

rod / coil 5 Schroeder et al. 2003. Once such a long DNA molecule becomes highly
extended, the drag acting on it is so much larger than in the coiled state, that the chain
remains extended even when the extension rate is decreased to a value below that required to stretch the chain in the first place. This phenomenon is consistent with Brownian dynamics simulations that include HI and with solutions to the consistently averaged
moment equations Magda et al. 1988b. A similar effect is predicted to occur in extensional flows of dilute solutions of polystyrene, if the molecular weight is greater than
around 750,000 Hsieh 2004.
3. Simple shearing flow

The velocity gradient tensor for a simple shearing flow contains an off-diagonal term,
namely the shear rate ; i.e.,

simple shearing .

61

The asymmetric off-diagonal term implies that simple shearing flow contains vorticity, or
rotation, as well as stretch. For this reason, the behavior of a polymer molecule in a
simple shearing flow differs profoundly from that in an extensional flow. In a shearing
flow, the polymer rotates, stretches, and collapses aperiodically, as illustrated in Fig. 12
Smith et al. 1999. When the molecule is tipped at a small positive or counterclockwise angle with respect to the flow direction, as shown at the top of Fig. 12, the shearing
flow tends to stretch the molecule and align it in the flow direction. However, if Brownian
motion should cause the molecule to tip to a negative angle, the shearing flow drives the

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

42

LARSON

FIG. 12. Illustration of a polymer molecule in a simple shearing flow. The straining motion in the flow stretches
the molecule and the vorticity rotates it toward the flow direction. Then, Brownian motion can either tip the
molecule back into a direction that permits further stretch bottom left, or into a direction in which the straining
motion causes contraction from Smith et al. 1999; reprinted with permission from Science 283, 1724 1727
Copyright 2003 AAAS.

molecule back toward a coiled configuration. Thus, rather than the monotonic stretching
observed in an extensional flow, in a shear flow a polymer molecule such as -phage
DNA undergoes irregular stretching, tumbling, and coiling, as was directly observed by
Smith et al. 1999 in a simple shearing flow produced by translating a glass plate parallel
to a second stationary plate with fixed uniform gap i.e., a plane Couette flow; see Fig.
13. Typical time dependences of the chain stretch are shown in Fig. 14 at three different
shear rates, made dimensionless, again, using the longest relaxation time.
These irregular DNA deformations can be analyzed by a fast Fourier transform of the
stretch x, which when autocorrelated against itself yields a spectral power density, an
example of which is shown in Fig. 15. Results from bead-spring and bead-rod simulations are also shown in Fig. 15 Hur et al. 2000. This plot shows a power spectrum that
is flat at low frequencies, and then decays with increasing frequency according to a
power-law exponent of 3.76 at intermediate frequencies and 1.73 at high frequency.
In the absence of flow, the high-frequency regime is unchanged, the intermediate regime
disappears, and the low-frequency plateau is shifted downward. Thus, flow enhances the
low-frequency fluctuations in x, and this effect decays away gradually with increasing
frequency, becoming negligible when the frequency equals the shear rate. The power
laws describing the decay observed in the experiments or the bead-spring or bead-rod
models are qualitatively similar to those extracted analytically from a Hookean dumbbell
model, which gives an exponent of 4 in the range 1 f Wi, and 2 for and f
Wi, where f is the frequency in rad/s. Thus, the cross-over frequency separating the
two power laws is set by the shear rate. The power-law exponents of 3.4 and 1.6
obtained from the Rouse model, which has multiple relaxation modes, are closer to those
from the bead-rod and bead-spring simulations, and the experimental data, shown in Fig.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

43

FIG. 13. Images of individual stained -phage DNA molecules in viscosified water in steady shear flow
between parallel glass plates, one of them translating, at a shear rate of 1 s1 , Wi 19. The view is
perpendicular to the plates, along the gradient direction, and so is perpendicular to the view shown in the
cartoon in Fig. 12. Each row of images is a time series, from left to right. The time interval between images is
6 s from Smith et al. 1999; reprinted with permission from Science 283, 1724 1727 Copyright 2003 AAAS.

15, than are the predictions of the dumbbell model. Note the good agreement between the
predicted and measured power spectra, especially for the bead-rod model and the beadspring model with the largest number of beads.
The spectral power density shows how shearing flow enhances polymer fluctuations
over a range of frequencies up to a value equal to the shear rate, this enhancement
resulting from the stretch-tumble-recoil dynamics of the chain in a flow with vorticity.
These tumbling dynamics can be explored in more detail through the autocorrelation
function of the end-to-end vector R x (t)R y (tT) , which correlates the x component of
the end-to-end vector of the chain at time t with the y component at some later or earlier
time, tT. Here x is the flow direction, and y is the gradient direction Chopra and
Larson 2002. Simulations show that fluctuations of the chain in the y direction give rise
at a later time to increased stretch in the x direction, while stretch in the x direction later
leads to suppression of stretch in the y direction, these results being a consequence of the
tumbling dynamics illustrated in Fig. 12. Attempts to directly measure this correlation are
difficult because they require viewing the molecule in the x y plane containing the
gradient direction, that is, from a sideways viewing angle.
The stretching and tumbling dynamics of polymer molecules in a simple shearing flow
are also reflected in the experimental and theoretical histograms of polymer stretch shown
in Fig. 16. At high shear rates, the histogram becomes flat, that is, all values of the stretch
are almost equally probable over a wide range of stretch, from 20% to 80% of full stretch.
Agreement between results from experiments and simulations is excellent. Because of
this wide distribution of stretch, the average polymer stretch, even in fast shearing flow,
seems to saturate at only around 40% of full extension, in reasonable agreement with the
predictions of Brownian dynamics simulations using bead-spring or bead-rod chains; see
Fig. 6.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

44

LARSON

FIG. 14. Extension versus time of -phage DNA, imaged as in Fig. 13, at three different Weissenberg numbers.
For Wi 25.2, data for three different molecules are combined, separated by dashed vertical lines from Smith
et al. 1999; reprinted with permission from Science 283, 1724 1727 Copyright 2003 AAAS.

4. Mixed flow

Recently, the groups of Shaqfeh and Chu et al. Babcock et al. 2003 have imaged
DNA molecules in mixed shear and extensional flows, in which the velocity gradient is
given by

FIG. 15. Power spectra density of -phage DNA molecule in a simple shearing flow at a Weissenberg number
of 38, compared to predictions of bead-spring bead-rod chain. Here is the longest relaxation time from Hur
et al. 2000.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

45

FIG. 16. Probability distribution of molecular stretch x of -phage DNA molecules in shearing flow, determined from a single-molecule imaging and b simulations using bead-rod chains containing 150 beads, at
three different Weissenberg numbers. The experimental data are from Smith et al. 1999 from Hur et al.
2000.

1
2

62

Here, G is the magnitude of the velocity gradient and is the flow type parameter,
with 1 for planar extensional flow, 0 for simple shearing flow, and 1 for
pure rotation. Note that for the case of simple shear 0, Eq. 62 reduces to

1
2

63

which does not match the tensor given for simple shear in Eq. 61. The reason is that Eq.
63 is expressed in a frame oriented such that the flow direction is along a diagonal in
the x y coordinate system; i.e., rotated 45 with respect to the frame in which Eq. 61
is expressed. If Eq. 63 is rotated into a frame in which the velocity is in the x direction,
we recover Eq. 61, and can identify G as the shear rate . Similarly, in the limit of a
planar extensional flow, for which 1, the flow is in a frame rotated to that for which
Eq. 60 applies. Experimentally, the mixed flow was generated in a cleverly designed
device that uses flexible fibers pulled through guiding channels whose curvature estab-

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

46

LARSON

FIG. 17. Mean fractional extension x along the extensional eigenvector vs a Weissenberg number Wi
G 1 and b effective Weissenberg number Wieff G1 for -phage DNA in a mixed flow described by
Eq. 63, for 0 and Wieff 0.6 from Babcock et al. 2003; reprinted with permission from Macromolecules, Copyright 2003 American Chemical Society.

lishes the shape of the streamlines in the flow, all in a device thin enough to permit
imaging of individual stained -phage DNA molecules at high optical resolution. Babcock et al. 2003 found that for 0.005, the average stretch x is a function only of
the combined variable G ; see Fig. 17. G is the largest eigenvalue of the velocity
gradient tensor, and is also the extensional component of the flow, if the velocity gradient
describing the flow is decomposed into a sum of an extensional and a shear part. Thus, at
steady state, the average stretch of the polymer depends only on the extensional portion
of the flow, and is oblivious to the addition of a shearing component. This occurs because
the chain orients along the stretching axis of the extensional flow and the shear gradient
can then only act on fluctuations of the chain away from this orientation, which are
negligible at large flow strengths G. As G decreases, however, a window of small values
of opens up over which an effect of the shearing component of the flow on the stretch
can be observed Woo and Shaqfeh 2003. In general, these new results with DNA
molecular imaging confirm and extend earlier results obtained from birefringence measurements on mixed flows for polystyrene solutions Fuller and Leal 1980, to which we
turn in the next section.
In summary of the work on DNA single-molecule imaging, we can say that for

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

47

modest-sized DNA molecules such as -phage DNA there is excellent agreement between experimental measurements of polymer stretch and very simple bead-spring or
bead-rod BD simulations that do not explicitly account for either excluded volume or
hydrodynamic interactions, although these phenomena must be implicitly included in the
choice of the bead-spring parameters. For much longer DNA molecules, whose extensibility is comparable to that of high molecular weight flexible polymers such as polystyrene, HI is important, and leads to coilstretch hysteresis, for example. Modeling of
these much longer DNA chains is still not well developed.

B. Polystyrene solutions
1. Bead-spring parameters

To model polystyrene molecules using the bead-spring model, we again need to assign
parameter values. The simplest case to consider is that of polystyrene in a theta solvent,
where excluded volume effects can be neglected. In this case, we can use the freely
jointed chain model and thereby obtain the constants for the spring force law from known
parameters for polystyrene, namely the characteristic ratio C , the number of backbone
bonds n M /52, where M is the polystyrene molecular weight, and the length of the a
backbone bond 1.54 A. The parameters of the freely jointed chain model, namely L,
b K , and N K , are then given by Eqs. 7 and 8. With these, the parameters of the
spring law are established. For the Cohen approximation to the inverse Langevin
function Eq. 25, the constant H is given by Eq. 21 and the fully stretched spring
length L s is just L s L/N s , where N s is the number of springs.
The frictional properties of a polymer are represented by the bead-drag coefficient,
which is proportional to the solvent viscosity s . In simple low-viscosity solvents, such
as benzene and toluene, polystyrene molecules have short relaxation times; i.e., 1 is
around 1 ms or so. Thus, probing the linear and nonlinear response of polystyrene and
other synthetic polymers in the linear and nonlinear regimes requires very rapid strain
rates, on the order of 103 s1 or higher. With the exception of capillary rheometers,
which only measure steady-state viscosity, most rheological equipment cannot probe this
very fast regime, in part because of inertial effects at these high rates. These problems can
be minimized if the solvent viscosity is greatly increased, so that the longest relaxation
time reaches an order of a second or so. Dilute polymer solutions for which the solvent
is made very viscous are frequently called Boger fluids Boger 1977; Mackay and Boger
1987. These high-viscosity solvents are typically formulated by dissolving low molecular weight and, hence, rather inelastic polymers, in a low-viscosity solvent. The viscous
solvents are therefore frequently mixtures, whose solvent quality toward the dilute high
molecular weight polymer can be uncertain or rather complex. Simulations and theory
suggest that such mixed solvents cause polymer molecules at rest to shrink below their
theta dimensions, even though the solvent quality remains good in the sense that the
polymer is far from a point of phase separation Magda et al. 1988a.
Whatever the choice of solvent, assuming that the elastic properties of the chain are
modeled properly, the gross frictional properties of the bead-spring chain at equilibrium
can be made to mimic those of the real polymer by matching the longest relaxation time
of the bead-spring chain 1 with that of the real polymer. For a theta solvent, the latter
can be obtained from Eqs. 3 and 11, yielding 1 K M 1.5 s /S 1 RT, where K
8104 dl g1 (g/mol) 1/2 for polystyrene. If the bead-spring simulations neglect
HI, the longest relaxation time is given by Eq. 9, where i 1 for the longest relaxation

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

48

LARSON

time 1 , that is 1 totR20/6 2 k B T. Equating this value with 1


K M 1.5 s /S 1 RT yields the following formula for the bead-drag coefficient , for
N 1,

62KM 1.5 s
S 1N ANR 20

6 2 m 0 K M 0.5 s
S 1 N A C N 2

64

This choice guarantees a match between the theta solvent experimental value for 1
and the simulated value, even though the simulations neglect explicit treatment of HI.
Hence, this matching procedure includes HI implicitly in the choice of the value for .
Since the value of is set using the terminal linear viscoelastic relaxation time, the effects
of HI that are included in the bead-spring model are those that govern slow relaxations
near the equilibrium state. Without explicit inclusion of HI, the shape of the spectrum of
relaxation times in the bead-spring model will match the Rouse, rather than the more
accurate Zimm, spectrum; see Sec. V B. Also, without explicit HI, the drag properties of
the bead-spring chain under large deformations may diverge from those of the real chain
because of deformation-dependent drag, i.e., the dependence of HI on chain extension de
Gennes 1974; Hinch 1974; 1977; see Sec. VI A 1.
For -phage DNA, however, we have seen that good predictions of molecular stretch
in extensional and shear flows are obtained even in the highly nonlinear state without
inclusion of explicit HI in the simulations. As alluded to earlier, explicit HI can be
neglected in simulations of -phage DNA because its low extensibility, which we defined
earlier as the ratio of the fully stretched chain length L to the root-mean-square end-toend length at equilibrium R 2 1/2
0 . The higher extensibility of most polystyrene and other
synthetic polymer molecules implies that changes in drag due to HI are likely to be much
more pronounced for synthetic polymers than for -phage DNA, increasing the importance of including explicit HI in Brownian dynamics simulations, at least for theta solvents.
Explicit HI can be included in Brownian dynamics simulations of polystyrene dilute
solutions using the methods described in Sec. III C 4. However, the HI parameter h *
must be specified. As described in Sec. V B, the linear viscoelastic properties of dilute
polymer solutions in a theta solvent are correctly modeled using a value h * 0.25 and
the number of beads set to M /5000, where M is the molecular weight of polystyrene. This
rule would imply that to simulate a polystyrene molecule of molecular weight M 2
million would require using 400 beads, while for M 20 million, 4000 beads would be
needed! This is well beyond current computational capabilities, especially when explicit
HI is included in the simulations.
An alternative approach for describing nonlinear viscoelastic properties is to choose
bead-spring parameters such that the drag force is correctly modeled both when the chain
is an equilibrium coil and when it is fully extended. We have described how the frictional
properties of the coil in the unstretched state can be modeled by choosing the bead-drag
coefficient so that the longest relaxation time of the bead-spring chain matches that of the
real coil, which is given by the Zimm theory for polymers in theta solvents. In the fully
extended state, on the other hand, the polymer resembles a slender filament of length L
and diameter d, where for polystyrene d is a nanometer or so. The drag force on a
straight, slender filament in a uniform flow with velocity V relative to the filament that
parallel to its axis is given by a formula of Batchelor 1970, namely Eq. 58. To apply
this to an extensional flow in which the center of the filament is moving with the flow, we
can divide the filament into small segments and apply the Batchelor formula to each

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

49

segment. Since the velocity at position x downstream of the center of the filament is x,
the drag force on a piece of the filament of length dx at this position is
dFdrag

2 s xdx

ln

65

This drag force will produce tension in the filament that is zero at the end of the filament,
and accumulates as one moves toward the center. Integrating Eq. 65 from the center of
the filament to its end gives the tension at the center of the filament, which is the total
drag force on one-half of the filament:
Fdrag

L/2

ln

s xdx

1
8

2
ln

L 2.

66

The parameters of the bead-spring model can be chosen to make the distribution of
drag or tension along a fully extended bead-spring chain match that of the slender
filament, derived above from the Batchelor formula. For the fully extended bead-spring
chain, with Brownian forces neglected, the force-balance equation for the chain, Eq. 34,
in a steady-state extensional flow, reduces to
N

sp,b

j 1Di j F j

xi 0,

kBT

67

where the diffusion tensor Di j can be evaluated explicitly, since for the fully extended
chain, the positions of all beads are known. From Eq. 41, we obtain:
Dii,xx

Dij,xx h*kBT
4

kBT
H

6 s a i j L s

kBT
6sa

2a 2
1

3 i j 2 L s2

2a 2
2

i j 2 L s

68
,

for i j,
where L s is the fully stretched length of a spring. Solving Eq. 67 gives the tension on
each spring. Requiring that the tension on the center spring match the accumulated drag
force on half the chain, given by Eq. 66, then yields a matching condition that can be
imposed on the parameter values of the bead-spring model.
If one specifies the number of beads N, one can then use this matching condition,
combined with Eq. 40, to obtain the value of the bead radius a and the value of h * . For
a ten-bead chain, the values of a and h * for polystyrene molecules of various molecular
weights in a theta solvent are given in Table II, along with the corresponding values for
-phage DNA. The values of h * are given as h *
max , because these are the maximum
values of h * that are allowed in the Brownian dynamics simulations for the given value
of bead radius a, since higher values would lead to a nonpositive-definite diffusion tensor,
as noted earlier. For polystyrene of molecular weight 10 million, the chain length and
bead radius are comparable to the values for -phage DNA, but the value of h *
max is much

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

50

LARSON

TABLE II. Molecular size, estimated bead radius, and maximum h * value for -phage DNA and polystyrene
molecules in a theta solvent for a 10-bead chain (N s 9).

L ( m)
R ( m)
R g ( m)
a ( m)
h*
max

-phage
DNA

10M
PS

2 M PS

0.5 M PS

0.2 M PS

21.2
1.67
0.68
0.073
0.13

24.285
0.21
0.086
0.076
1.07

4.7355
0.092
0.038
0.0179
0.56

1.2142
0.047
0.019
0.0055
0.35

0.4857
0.030
0.012
0.0025
0.25

Note: PS Polystyrene.

larger for this polystyrene molecule than it is for -phage DNA, because the coil size of
the polystyrene is so much smaller than that of DNA, due to the higher persistence length
or Kuhn step length of the latter. As a result, for polystyrene, using N 10 beads, the
coil size is scarcely bigger than the bead size! Thus, simulation of a high molecular
weight polystyrene chain in a theta solvent with only ten beads is physically unrealistic,
since the beads highly overlap. Note that for much lower molecular weight such as M
0.2 million, the bead size for the polystyrene chains becomes much smaller than the
radius of gyration, and the value of h *
max becomes more reasonable.
For high molecular weight polystyrene chains, the bead size can only be made much
smaller than the coil size by using more beads to represent the chain. Table III gives the
values of h *
max obtained when a polystyrene chain of molecular weight 1.95 million is
represented by bead-spring chains containing from 10 to around 300 beads. Again, the
values of bead radius a and hydrodynamic interaction parameter h *
max were obtained by
matching the drag on the fully extended chain to that given by the Batchelor formula, and
using Eq. 40 for h *
max . As the number of beads is increased, the longest relaxation time
of the chain 1 decreases, where 1 is obtained by simulating the chains relaxation from
a modestly stretched state. The decrease occurs because of the increased strength of the
HI that accompanies an increasing number of beads remember that the strength of the HI
effect is proportional to h * N 1/2). For the example in Table III, if N 166 beads are
TABLE III. The effect of number of beads N on h * and sim for 1.95 million molecular weight polystyrene
(N K 2626) with Batchelors formula 1970 imposed.

a
m

sim
1
s

Wi

Extension rate
l/s

0.568
0.475
0.386
0.312
0.25

10
20
40
80
166

1.79E02
1.03E02
5.84E03
3.32E03
1.84E03

2.3297
1.7293
1.4402
1.2393
0.9637a

1.32
1.32
1.32
1.32
1.32

0.5666
0.7633
0.9165
1.0651
1.3697

Zimm 1956 results


1.95

0.291

N extrap
101

2.75E03

Z1
1.12

Experimental results
1.95

0.21

N extrap
309

1.13E03

exp
1
0.8

1.32

1.65

M
million
1.95
1.95
1.95
1.95
1.95

L
m

h*
max

4.73569
4.73569
4.73569
4.73569
4.73569

The longest relaxation time of the 166-bead chain is obtained by extrapolating the longest relaxation time of
fewer-bead chains.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

51

FIG. 18. The symbols are experimental extensional viscosity data from the filament-stretching rheometer for a
dilute (c 0.0747 wt %) solution of 1.95 million polystyrene in a solvent consisting of 25% low-molecularweight polystyrene in dioctylphthalate, at the extension rates shown. The lines are predictions from Brownian
dynamics simulations without HI, with the bead drag coefficient chosen using Eq. 64 so that the longest
relaxation of the bead-spring chain matches that for a polystyrene molecule of this molecular weight in a theta
solvent, namely 1 1.12 s. The Trouton ratio Tr is defined as the transient extensional viscosity, divided by
the zero shear viscosity 0 results obtained using methods described in Li et al. 2000.

chosen, one can obtain h * 0.25, the value corresponding to a theta solvent for the
linear viscoelastic properties. Somewhat fewer beads, N 101, are needed to match the
simulated longest relaxation to that given by the Zimm theory, while many more beads
are need to bring the relaxation time down to the value measured experimentally for a
dilute solution of 1.95 million polystyrene studied by Sridhar and co-workers Orr and
Sridhar 1999; Li et al. 2000. Thus, for high molecular weight polystyrene solutions in a
theta solvent, large numbers of beads appear to be required if the drag properties of the
real polymer are to be matched by the simulated polymer in both the fully extended state,
where the Batchelor formula should apply, and in the relaxed state, where the relaxation
time is given by the Zimm theory. Carrying out Brownian dynamics simulations with a
realistic representation of HI is therefore very challenging for high molecular weight
chains with contracted dimensions, such as polystyrene in a theta solvent. However, for
stiffer polymers with more expanded coils, such as -phage DNA molecules, or synthetic
polymers in a good solvent, or for lower molecular weights, HI effects are much weaker
and easy to simulate with modest numbers of beads Hsieh et al. 2003.
2. Extensional flow

We now consider experiments and corresponding simulations of high molecular


weight polystyrenes in a thetalike solvent in an extensional flow. For dilute polymer
solutions, the best-defined extensional flows are those produced by the filamentstretching rheometer of Sridhar and co-workers, as mentioned earlier. Transient extensional viscosities from this device for a dilute solution of polystyrene of molecular weight
2 million are plotted in Fig. 18. The ordinate of this plot is the Trouton ratio, which is
the time-dependent uniaxial extensional viscosity u ( zz xx )/ divided by the
zero-shear viscosity 0 , where zz xx is the tension in the filament divided by its
instantaneous cross-sectional area. The solvent for this dilute Boger fluid is dioctylphthalate DOP, a theta solvent for polystyrene at 22 C Johnson et al. 1970, vis-

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

52

LARSON

cosified with a 25% by mass of low molecular weight (M 50,000) polystyrene Orr
and Sridhar 1999; Gupta et al. 2000. This mixed solvent is believed to be roughly
thetalike with respect to polystyrene, although, as indicated earlier, the coils may be
slightly smaller than in a true theta solvent.
Using the methods described in the first paragraph of Sec. VI B 1 to obtain the spring
constants, and Eq. 64 to get the bead-drag coefficient, simulations in the absence of
explicit HI can be carried out Li et al. 2000, yielding the predictions shown in Fig. 18.
These a priori predictions are in qualitative agreement with the extensional-flow data
from the filament-stretching device in both the overall stress levels and in the strain at
which the Trouton ratio Tr begins to rise rapidly above the low-strain limiting value of
Tr 3. Experiments and simulations both show that when Wi exceeds a value of around
5, the curves of Tr versus Hencky strain roughly collapse onto a single curve.
However, this qualitative agreement between simulations without HI and experimental
data shown in Fig. 18 is not perfect, and degrades as the polymer molecular weight
increases. Figure 19 shows that as the molecular weight increases, the predicted upturn in
the Trouton viscosity shifts toward higher strains, while the measured response shows
almost no change. In addition, the predicted high-strain plateau increases with molecular
weight, while the experimental plateau apparently stays constant. Notice that while the
data in Fig. 18 span a range of Weissenberg numbers, Wi 1 , from 1 to 15, in Fig.
19, Wi exceeds six for all curves, and so the curves for different extension rates are nearly
identical.
If HIs are included in the Brownian dynamics simulations, with the bead diameter and
HI parameter h *
max set by the requirement that the distribution of drag on the bead-spring
chain be consistent with the Bachelor formula, Eq. 66, then the simulations are in much
better agreement with the experimental data; see Figs. 20 and 21. However, in these
simulations, unless very large numbers of beads are used see Sec. VI B 1, the longest
relaxation time of the simulated chain is larger than either the experimental or Zimm
value, the latter of which was used to set the bead drag coefficient in the simulations that
neglected HI. Fortunately, at strain rates a factor of 5 or more above the critical strain rate
for a coilstretch transition, simulated results for Trouton viscosity are insensitive to the
number of beads above a minimum of around 20 80 beads, because the most significant
deviations are at low stretch, where the errors are swamped by the contribution of the
solvent to the stress Hsieh and Larson 2004. As we shall see, however, in shearing flow,
there is a greater effect of the number of beads used in the simulation.
Note that, while agreement between the simulations with HI and experimental data is
very good for polystyrene of molecular weight 2 million Fig. 20, this agreement again
begins to degrade somewhat as the molecular weight increases Fig. 21. The inclusion of
HI brings the predicted strain at the upturn in Trouton ratio into better agreement with
experiment compare Fig. 21 with Fig. 19, but it greatly worsens the overprediction of
the high-strain plateau in Trouton viscosity at high molecular weight. Both of these
effects of HI result from deformation-dependent drag, i.e., the increase in effective drag
coefficient with increasing chain deformation produced by HI. The existence of
deformation-dependent drag in extensional flows of very large polymers is strongly supported by studies with 1 mm long DNA molecules that show hysteresis in the coilstretch
transition; see Sec. VI A 2.
Figure 21 shows that, in the experiments at Wi 5, when the Trouton viscosity Tr is
normalized by the product of concentration and molecular weight, and plotted against
strain, the results collapse onto a single line. This implies a proportionality of Tr with
concentration c, which is expected, since the solutions are dilute, but the proportionality
of Tr with molecular weight is surprising. The simulations with HI also successfully

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

53

FIG. 19. The same as in Fig. 18, except that the polystyrene molecular weights, concentrations, and longest
relaxation times are a: M 3.9 million, c 0.0444 wt %, 1 3.16 s; b: M 10.2 million, c
0.0088 wt %, 1 13.32 s; and c: M 20 million, c 0.0069 wt %, 1 36.71 s from Li et al.
2000.

predict that Tr/(cM ) is roughly independent of molecular weight at low and medium
strains, but completely fail to capture the experimentally observed superposition of the
plateau values of Tr/(cM ) for all molecular weights. The experimentally observed linear
dependence of the plateau Trouton viscosity on molecular weight is very surprising, and
contrary to predictions of bead-spring or bead-rod models, whether or not they include
HI.
Sridhar and co-workers observed that at very high extension rates, the Trouton viscosity actually decreases with increasing extension rate; see Fig. 22. Brownian dynamics
simulations with bead-spring or bead-rod models in a continuum solvent show a constant
Trouton viscosity at high extension rate, rather than a decrease. In addition, if the chain

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

54

LARSON

FIG. 20. Predictions of Brownian dynamics simulations that include HI lines, compared to measured symbols Trouton ratio, for a dilute solution of 1.95 million Dalton polystyrene, described in Fig. 18, at the strain
rates shown from Hsieh et al. 2003, reprinted from J. Non-Newt. Fluid Mech., 113, 147191, Copyright 2003,
with permission from Elsevier.

is a fully extended filament at high extension rates, Batchelors suspension theory for
slender filaments predicts that the viscosity should be a constant, with no decrease.
Interestingly, however, in Monte Carlo simulations of a lattice model of a polymer, Li
and Denn 2004 showed not only the expected coilstretch transition at a Weissenberg
number of around 0.5, but also a maximum, and then a decrease in extensional viscosity
at high extension rate, as the elastic stress in the polymer reaches a well-defined saturation value. The saturated stress in the lattice simulation may arise because entropic stress
on a lattice is inherently bounded by the finite number of microstates possible for a
coarse-grained lattice polymer. Thus, while the lattice simulations might provide a clue to

FIG. 21. Predictions of Brownian dynamics simulations that include HI lines compared to measured symbols for the Trouton ratio, normalized by the product of concentration c and molecular weight M, for dilute
solutions of polystyrene described in Figs. 18 and 19. In the simulations, a 20-bead chain, with h * 0.475 was
used for 1.95 million polystyrene, while 80 beads were used for both 3.9 million (h * 0.394) and 10.2
million (h * 0.5565) polystyrene from Hsieh and Larson 2004.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

55

FIG. 22. Extensional viscosity data for start-up of steady extension at high Weissenberg number for a dilute
Boger solution of 69 ppm 20 million molecular weight polystyrene in the thetalike solvent described in Fig. 18.
The strain rates/Weissenberg numbers are given in the figure from Gupta et al. 2000; reprinted with permission
from Physics of Fluids.

the observed viscosity maximum, it is not clear how in a continuum or molecular


solvent an upper bound on the tension in a polymer might come about, and a satisfying
explanation of the viscosity maximum is not yet available.
The failure of Brownian dynamics simulations to predict the experimentally observed
plateau stress in extensional flow of dilute polystyrene solutions might seem especially
odd, since simulations of -phage DNA molecules proved to be so successful. The poorer
predictions in polystyrene solutions might be due to some combination of the following
factors. 1 For polystyrene, we compared experimental to computed stresses, rather than
individual chain conformations. Since the stress in extensional flows is dominated by
contributions from molecules that are nearly fully extended, and the force in these chains
becomes singular at high extension, a very small deviation in the predicted chain stretch
could produce a huge difference in stress. 2 The DNA solutions can be made exceedingly dilute, since single molecules only are observed in the experiments, while the
polystyrene solutions, although dilute in the sense that c c * , might still be influenced
by chainchain interactions when the chains become highly extended in a strong extensional flow. 3 HI is much more important in polystyrene than in -phage DNA, because
of the tight coil dimensions of the former, relative to the latter. Strong extensional flows
deform the coil into highly kinked or folded conformations see Fig. 9, which in highly
flexible high molecular weight polymers can contain multiple folds Acierno et al. 1974;

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

56

LARSON

Larson 1990; Hinch 1994; Agarwal 2000. HI has been shown to slow the unraveling of
the chain Agarwal 2000; Hsieh and Larson 2004. However, HI among these folded
filaments may be poorly represented by a bead-spring chain, with its point-force approximation to HI. 4 The polystyrene polymers might undergo partial chain scission at high
stresses in the filament-stretching device Hsieh and Larson 2004.
Chain scission has been directly observed in high-speed extensional flows of polystyrene and other polymers in the crossed-slot device Keller and Odell 1985; Odell and
Keller 1986; Islam et al. 2004, and chain scission is achieved at lower strain rates as the
polymer molecular weight increases. In fact, Keller and Odell 1985 find that for polystyrene in low-viscosity solvents that polystyrene molecules with molecular weight
higher than around 30 million will undergo tension-induced chain scission whenever a
coilstretch transition occurs, and so these molecules cannot be unraveled without breakage. Polystyrenes of lower molecular weight polystyrenes can be stretched without breaking, but they too will break if the Weissenberg number is high enough. Since chain
scission is thought to be controlled mainly by stress, differences in solvent viscosity
should affect both the inverse relaxation time and the critical strain rate for scission to the
same degree, and so should apply to Boger fluids as well as long as the Weissenberg
numbers are comparable. One caveat is that in the low-viscosity solvents used by Odell
and Keller, inertia may affect the flow field and therefore the chain scission Islam et al.
2004. Still, their observations suggest that chain scission could provide an explanation
for the anomalously low plateau stress relative to the predictions of Brownian dynamics
simulations; however, confirmation of chain scission in filament stretching flows by size
exclusion chromatography or other methods is lacking so far.
Despite the failure, up until now, to obtain quantitative agreement between Brownian
dynamics simulations and experimental measurements of the plateau Trouton ratio in
start-up of steady extension, the simulations capture many aspects of the rheology of
these solutions. In particular, they capture the effects that arise from molecular individualism, which is reflected in the very heterogeneous population of conformations that
occur in these solutions. An example, shown in Fig. 23, is the peculiar sudden jump in
extensional stress when the extension rate is abruptly increased. In Fig. 23, at a strain of
3.6, the extension rate imposed by the filament stretching rheometer is suddenly doubled,
from 2 to 4 s1, leading to a surprisingly abrupt jump in the extensional stress. In fact,
the stress almost instantly increases to the value that would have been obtained had the
higher extension rate been imposed continuously from the start of the experiment; compare the filled symbols with the open squares in Fig. 23. Thus, while the stress shows a
strong memory of the total imposed strain, it retains only a weak memory of the rate at
which it had previously been deformed. The simulations show almost the same behavior
as the experiments; see Fig. 23. Analysis of the simulations reveals that the origin of this
stress jump lies in the molecular individualism of the polystyrene molecules during the
extensional flow. It is found that at any instant in time, most of the stress is carried by
only a subset of the molecules, namely those that have dumbbell or other configurations
containing a highly stretched section that bears most of the stress. These portions of the
molecules cannot stretch much further when the extension rate is increased. As a result,
a sudden doubling of the strain rate simply doubles the load on these taut load-bearing
elements, resulting in a sudden doubling of stress, as seen in Fig. 23. Continued straining
increases both the fraction of molecules containing taut load-bearing portions, and the
lengths of the taut portions, so that the overall stress continues to grow as additional
strain accumulates.
Other phenomena that are captured by the simulations include sudden jumps in stress
after restarting an interrupted extensional flow Li et al. 2000, and stress-birefringence

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

57

FIG. 23. Comparison of experimental symbols and simulated lines for a dilute Boger solution of a polystyrene of molecular weight 10.2 million in extensional flow at steady rates of 2 and 4 s1 and in an extensional
flow with a step change in strain rate from 2 to 4 s1 at a lagged strain 0 of 3.6. Because these
simulations were carried out without HI, there is a significant offset along the strain axis between experimental
and simulation results; see Fig. 19. Here, this offset is arbitrarily corrected by shifting the simulation results
along the strain axis by 0 0.60 strain units, so that the simulation results at a constant strain rate match the
experimental results from Li et al. 2000.

hysteresis Doyle et al. 1998; Li and Larson 2000a; Ghosh et al. 2001, which is seen
when time-dependent stress is cross plotted against time-dependent birefringence during
both stress growth after start-up of steady extensional flow and stress relaxation after
cessation of this flow. Again, these phenomena result from molecular individualismthe
wide distribution of molecular configurations produced during these flows. The effects of
molecular individualism are not readily predicted by simple constitutive equations such

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

58

LARSON

as the so called Fene-P equation that result from the most obvious methods of preaveraging of the moment equations Lielens et al. 1999; Ghosh et al. 2001.
Another unusual phenomenon that might be explained, in part, by molecular individualism, is the very low deformation of polymer chains measured by light scattering in
strong extensional flows, such as the flow near the stagnation point in a crossed-slot
device, despite the high birefringence measured there Armstrong et al. 1980; Menasveta
and Hoagland 1991; Li and Larson 2000b. This result could perhaps be partly explained
if the chain is folded into a kinked state Larson 1990, Hinch 1994; Li and Larson
2000b, in which chain segments are oriented, but not many are highly stretched, because
the strain required to stretch most chains is not attained everywhere in the finite region
around the stagnation point sampled by the laser beam used to scatter light. However,
anomalously low deformation is also seen in uniform steady-state shearing flows in
which the molecules can be strained indefinitely Cottrell et al. 1969; Lee et al. 1997,
and so this explanation does not seem to be the whole story.
In principle, coilstretch hysteresis, observed in DNA solutions see Sec. VI A 2
ought to be present also in extensional flows of dilute polystyrene solutions. In DNA
solutions, the critical ratio rod / coil needed to see coilstretch hysteresis seems to be
around 3 4. For polystyrene in a theta solvent, we can estimate that the critical molecular
weight for hysteresis should be around a million or so, and recent Brownian dynamics
simulations predict hysteresis in solutions of polystyrene of molecular weight greater
than around 750,000 Hsieh 2004. Moment equations derived from the Smoluchowski
equations with self-consistent averaging of HI see Sec. IV C 4 were solved already
15 years ago with as many as 500 beads, and these showed the signature of coilstretch
hysteresis Magda et al. 1988b; albeit with fluctuations in HI neglected. Most experimental extensional flows appropriate for polystyrene solutions, such as that produced by
the filament-stretching apparatus, probably do not provide a long enough residence time
for hysteresis to be observed.
We also remark that while coilstretch hysteresis may be experimentally inaccessible
in polystyrene solutions, a related phenomenon that can be measured, and is of practical
significance, is the effect of preshearing on the ease with which chains are subsequently
stretched in an extensional flow. Such temporally mixed flows, in which for example
extension follows a region of shearing, are common in polymer processing, including
contraction or entry flows, as well as turbulent boundary layer flows. If chains that are
extended in shear suffer a large increase in drag coefficient, they could undergo a coil
stretch transition at a lower rate of extension than would otherwise be the case. Evidence
for this phenomenon has been cited by James et al. 1987, and a discussion can be found
in Larson 2000.
Finally, experiments using a better-than-theta solvent show that the upturn in extensional stress occurs more rapidly when the initial coil size is swollen than when it has its
theta dimensions. Figure 24 compares the growth of the Trouton ratio versus strain for the
same polystyrene concentration and molecular weight in two solvents, a theta solvent
consisting of the mixture of dioctylphthalate and low molecular weight polystyrene discussed earlier, and piccolastic, which is an oligomeric styrene thought to be a good
solvent for high molecular weight polystyrene. Intrinsic viscosity measurements by Anna
et al. 2001 for polystyrene in piccolastic indicate that R g M v with v 0.520.015,
implying that piccolastic is only slightly better than theta in its quality. Measurements of
intrinsic viscosity and light scattering of high molecular weight polystyrene in a mixed
DOP/low-molecular weight polystyrene solvent showed an exponent v 0.44 0.45, indicative of shrunken coil dimensions relative to theta Solomon and Muller 1996. Thus,
while the piccolastic is only marginally better than a theta solvent for polystyrene, the

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

59

FIG. 24. Comparison between Brownian dynamics simulations lines and experimental data symbols for the
growth of the Trouton ratio after start-up of steady unixial extensional flow at an extension rate of around 5 s1,
for dilute 0.05 wt % polystyrene of molecular weight 1.95 million in a thetalike solvent DOP/low molecular
weight polystyrene and a good solvent piccolastic, whose viscosity at room temperature is 87.5 Pa s. For the
simulations of polystyrene in the good solvent, EV interactions between beads were introduced using a truncated Leonard-Jones potential, with the strength of the potential chosen so that at equilibrium, the polystyrene
molecules are swollen to the size observed for polystyrene molecules of this molecular weight in very good
solvents, such as toluene from Li and Larson 2000a, used with permission from Rheol. Acta.

equilibrium coil dimension in this solvent is significantly larger than in the mixed solvent
DOP/low molecular weight polystyrene. Note in Fig. 24 that the upturn in extensional
stress occurs at a significantly lower strain for the polystyrene in the piccolastic solvent
than in the nominally theta-mixed DOP/low molecular weight polystyrene solvent. This
is not surprising, since the initial coil size is larger, and therefore the molecules will
become high stretched more quickly, in piccolastic, than DOP/low molecular weight
polystyrene. The simulation results agree nicely with the experimental data, but the
simulations assumed very good solvent conditions for polystyrene in piccolastic, equivalent to that of polystyrene in benzene, for example, while piccolastic does not seem to be
as good a solvent as this. Likewise, the good agreement obtained for polystyrene in
DOP/low molecular weight polystyrene was obtained by assuming this solvent to yield
theta dimensions for the polystyrene molecules, which seems to be larger than is really
the case as will be borne out in the discussion on shear properties below. It will be
important to perform experiments with solvents whose quality is varied systematically
before we can say whether the effect of solvent quality on the nonlinear rheology of
dilute polymer solutions is really understood.
3. Shearing flow

The predictions of the bead-spring model for the steady-state first normal stress coefficient 1 in shearing flow as a function of shear rate are compared to experimental data
for a 3.9 million molecular weight polystyrene in DOP/low molecular weight polystyrene
solvent in Fig. 25. This polystyrene Boger fluid is the same one for which extensional
flow properties in the filament stretching rheometer were discussed in Sec. VI B 2, and

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

60

LARSON

FIG. 25. Predictions of Brownian dynamics simulations with and without HI lines with symbols compared to
experiments closed circles for the first normal stress coefficient vs Weissenberg number for a dilute solution
of 3.9 million Dalton polystyrene, as described in Fig. 19 from Hsieh and Larson 2004.

the parameters of the bead-spring model are the same as those used for the predictions of
extensional flow properties in Figs. 21 and 23. In Fig. 25, predictions are given from the
bead-spring model both without, and with, HI. Without HI, the 1 ( ) curve shows a
zero-shear plateau, followed by shear thinning. The shear thinning is produced by finite
extensibility of the springs. The Rouse model, with infinitely extensible Hookean springs,
shows 1 1,0 , independent of the shear rate.
When HI is included, for highly stretchable chains, as the shear rate increases, the
predicted 1 ( ) curve shows a region of weak shear thinning, followed by shear thickening, followed finally by shear thinning again Kishbaugh and McHugh 1990; Hsieh and
Larson 2004. The final shear-thinning region is again a consequence of finite extensibility, and follows a power law, 1 ( ) 4/3 that is insensitive to the model, and also
insensitive to whether HI or EV is included, as long as the chain is finitely extensible
Lyulin et al. 1999. Doyle et al. 1997 find a similar power law 1 ( ) 14/11 in
bead-rod simulations, and are able to derive this power law from a scaling analysis.
The shallow shear-thinning regime at low shear rates, and the shear-thickening regime
at higher shear rate are both produced by HI. HI produces both thinning at low rates and
thickening at higher rates because of the opposing effects of HI on pairs of beads that are
near each other, versus those that are on the opposite ends of the chain. In a frame
moving with the center of mass of the chain, the beads on one side of the molecule tend
to be pulled in a direction opposite that of beads on the other side. Since HI creates a
coupling between bead motions that reduces the stress when beads are moving in the
same direction, but increases it when the move in opposite directions, HI between beads
on opposite ends of the chain tend to increase the net drag force while HI between beads
that are on the same side of the molecule tend to decrease the drag. The weakening of HI
that occurs as the chain is deformed therefore leads to a net decrease in drag contributed
by HI from beads on opposite ends of the chain, and an increase in net drag contributed
by HI between nearby beads Larson 1988. The first effect decrease in drag dominates
at low shear rates where the overall conformation of the chain is deformed but there is

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

61

FIG. 26. Curves of intrinsic relative shear viscosity versus dimensionless shear rate 0 , where 0 is the
relaxation time defined in Eq. 13, for dilute solutions of poly-methylstyrene with molecular weights of 1
690,000, 2 1,240,000, 3 1,460,000, 4 1,820,000, 5 7,500,00, 6 13,600,000 in toluene a good solvent,
and 7 13,600,000 in decalin a theta solvent from Noda et al. 1968; reprinted with permission from Journal
of Physical Chemistry, Copyright 1968 American Chemical Society.

locally little chain deformation, while the second effect increase in drag dominates at
higher shear rates where local chain deformation becomes important. Hence, the effect of
HI on 1 ( ) is nonmonotonic.
Experimentally, the first shear thinning regime and the thickening regime can just
barely be detected for high molecular weight polystyrene; see Fig. 25. These regimes,
especially the thickening regime, are much weaker than predicted. Note also that the
experimental first normal stress coefficient is considerably lower than the predicted one.
Similar results are found for the shear viscosity as a function of shear rate, which is
predicted to have the same shape as the 1 ( ) curve, but with weaker shear-rate dependence, since the shear viscosity contains a large solvent contribution, which is a constant
with shear rate, and therefore tends to swamp the contribution from the polymer. The
discrepancy between the predicted and observed zero-shear values of 1 and of ; not
shown supports the intrinsic viscosity and light-scattering results cited in Sec. VI B 2 that
in the solvent DOP/low molecular weight polystyrene, high molecular weight polystyrene
chains have smaller than theta dimensions.
Shear viscosity versus shear rate curves have also been measured for monodisperse
polystyrenes in pure, small-molecule solvents, such as toluene and decalin, using capillary rheometry. Figure 26 shows an example for high-molecular-weight polystyrenes in
theta and good solvents. Note that in the theta solvent, a weak shear-thinning region is
present, similar to the observations and predictions for the Boger fluid shown in Fig. 25.
The shear rate evidently does not reach a high enough value in these experiments to
encounter the expected shear thickening region.
In a good solvent, the shear thinning becomes more pronounced, as is predicted by the
bead-spring model when excluded volume interactions are included. For example, Fig. 27
shows the shear-rate dependencies of the polymer contribution to the viscosity p and of
the first normal stress coefficient 1 in the limit of dominant EV, with no HI, for infinitely extensible, chains with Hookean springs. Note that EV, by itself, produces substantial shear thinning in both p which is the polymer contribution to the shear viscosity
and in 1 . This shear thinning is evidently the result of the nonlinearity in the excluded
volume forces, which die out exponentially as the beads become separated from each
other. In the coiled state, the excluded volume repulsions are strong, making it relatively

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

62

LARSON

FIG. 27. Dimensionless ratio of polymer contribution to viscosity p / p,0 and first normal stress difference
ratio 1 / 1,0 vs reduced shear rate 0 , where 0 is the relaxation time defined in Eq. 13, in the limit
of strong excluded volume z , for infinitely extensible bead-spring chains with no HI obtained by solving
moment equations from Prakash 2002.

easy to deform the chain. But as the chain is deformed, these repulsive interactions
diminish, making further deformation harder. One might think that this should lead to
shear thickening, rather than thinning, since the force required to obtain a given deformation of the coil increases nonlinearly with increasing deformation. However, the coil
simply deforms much less than it would if there were a linear relationship between
deformation and force, and this leads to shear thinning. The net effect is similar to that
produced by a nonlinear spring that stiffens with increasing deformation, and also leads
to shear thinning.
The effects of both excluded volume and hydrodynamic interactions on shear thinning
and thickening were considered in early Brownian dynamics simulations of Lopez Cascales and Garca de la Torre 1991 for bead-spring chains with infinitely extensible,
Hookean, springs. They also found that EV enhances shear thinning at low and moderate
shear rates, and that HI produces shear thickening at moderate and high shear rates. The
regime of shear thinning at high shear rates was missing in these simulations because of
the choice of Hookean springs. It is expected that as the chain extensibility decreases,
corresponding to decreasing molecular weight, the high shear rate thinning regime caused
by finite extensibility will shift to lower shear rates, destroying the region of shear
thickening shown in Fig. 25. Very recent studies of Liu et al. 2004 incorporate both HI
and EV into a bead-rod model with up to 60 rods. A single, pronounced, shear-thinning
region is found in these simulations, evidently because the chain is not extensible enough
to show the separate regions of thinningthickeningthinning obtained for more extensible molecules.
While the rheological behavior of high molecular weight polystyrene in smallmolecule solvents is in qualitative agreement with predictions of bead-spring models,
quantitative predictions in which parameters of the model are determined a priori have
not yet been attempted for these solutions. There have been, however, efforts to fit

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

63

FIG. 28. Birefringence as a function of the largest eigenvalue of the velocity gradient tensor G for planar
flows in a four-roll mill for dilute solutions of polystyrenes of three different molecular weights shown in
polychlorinated biphenyl solvent. Here, G is the strain rate and is the flow type parameter defined in Eq.
62, where is zero for simple shearing and unity in planar extension. The different symbols correspond to
values of 1.0 circles, 0.8 triangles, 0.5 , and 0.25 diamonds. The curves are the predictions of the FENE
dumbbell model, including the effects of deformation-dependent drag discussed in Sec. VI A 2 from Fuller
and Leal 1980, used with permission from Rheol. Acta.

bead-spring model predictions to dilute solution rheological data, using the bead-spring
model parameters as adjustable constants; see Ahn et al. 1993, for example. Additional
efforts to predict these data a priori using the bead-spring model seem to be called for.

4. Mixed flow

Figure 28 shows steady-state birefringence data for dilute polystyrene solutions in


mixed planar flows, generated by the four-roll mill Fuller and Leal 1980. These data can
be collapsed onto a single line, when plotted against the maximum eigenvalue of the
velocity gradient tensor, which is equivalent to plotting them against only the extensional
component of the velocity gradient. Results similar to these were obtained recently with
single-molecule DNA imaging experiments, and were discussed in Sec. VI A 4; see also
Fig. 17. The birefringence data for polystyrene solutions in the four-roll mill were obtained around 1980, and were modeled at that time by the best numerically tractable
theory, that of the finitely extensible dumbbell, which gave good agreement with the data,
after the effects of the nonuniformity of the flow, and small levels of polydispersity of the
polymer were accounted for. It might be interesting to redo these calculations using
modern Brownian dynamics methods. Doing so would allow us to assess whether or not
the deviations observed between simulations and filament-stretching experiments for
polystyrene Boger fluids are also evident when similar simulations are applied to the
mixed flow data of Fig. 28.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

64

LARSON

VII. SUMMARY OF PROGRESS AND UNSOLVED PROBLEMS IN DILUTE


SOLUTION RHEOLOGY
The last few years have seen the emergence of a very thorough understanding of, and
ability to predict, the deformation of single molecules of -phage DNA in extensional,
shear, and mixed flows, thanks to DNA imaging methods, and powerful computers. This
predictive ability has not required explicit inclusion of the effects of hydroadynamic
interaction or excluded volume, as long as they are included implicitly in the choice of
constants characterizing the spring force and the viscous drag. The predicted profound
differences between extensional flows, which lack vorticity, and shearing flows, have in
fact been observed. In addition, the single molecule experiments and accompanying
simulations have highlighted the importance of molecular individuality in the nonlinear
regime, and the need in theoretical predictions to avoid preaveraging, if quantitative
predictions are to be obtained. The molecular tumbling expected in flows with vorticity,
such as simple shearing flows or mixed flows, has been observed and its temporal characteristics are in essentially quantitative agreement with simulation results.
When the computational methods that have been so successful in predicting -phage
DNA single-molecule deformation are applied to the prediction of the rheological properties of dilute solutions of high molecular weight polystyrene, there is good qualitative,
but not always quantitative, agreement between simulations and experiments. This may
be due, in part, to the much greater role played by HI and perhaps of EV in polystyrene
solutions, but even when HI is included in the simulations, and thetalike solvents used to
minimize EV, very significant disagreements remain between simulations and experiments. These disagreements include:
1 An overprediction of the plateau Trouton ratio Tr at high strains, which becomes
progressively more severe for higher molecular weight polymers.
2 A failure of the predicted curves of Tr/(cM ) versus strain for different molecular
weights to collapse onto a single line, when the Trouton ratio Tr is divided by
concentrations c and molecular weight M.
3 A failure to predict the observed decrease in plateau Trouton ratio when the extension
rate becomes very high.
4 Only qualitative, but not quantitative, prediction of the shear thinning and thickening
behavior of the steady-state shear viscosity and first normal stress coefficient as
functions of shear rate.
5 Failure to predict the apparently small coil deformation inferred from scattering
experiments in extensional and shearing flows of flexible polymers in small-molecule
solvents.
In addition to these discrepancies between theory and experiment that remain to be
resolved, there are a number of gaps in theoretical, computational, and experimental
studies that remain to be filled. These include evaluation of the role of both hydrodynamic interactions, and especially excluded volume, on dilute-solution rheology in shear,
extensional, and mixed flows. Appropriate methods of choosing the HI, and especially the
EV, parameters for given experimental solvent/polymer pairs need to be definitively
established. The roles, if any, of self-entanglements and IV are still uncertain. It is also
not certain whether or not the rheology of dilute solutions in Boger-type viscous solvents
can be superposed, after appropriate shifting of time scale, onto the rheology of equivalent small-molecule solvents. Filling these gaps will require inclusion of both HI and EV
effects in simulations of well-characterized dilute solutions of polystyrene or other polymer with all parameters determined a priori, and comparison of these simulation results

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

65

to carefully executed experiments on very well-defined dilute solutions with controlled


solvent quality. Clearly, while progress in the last decade has been substantial, much
remains to be done.

ACKNOWLEDGMENTS
Many colleagues have helped shape this review, directly or indirectly. Particularly
influential were Eric Shaqfeh, Ravi Prakash, Michael Graham, and David Morse, as well
as former and current graduate students Lei Li and Chih-Chen Hsieh. The author also
owes a debt of gratitude to the members of the Society of Rheology for continual encouragement and stimulation.

References
Acierno, D., G. Titomanlio, and G. Marrucci, Dilute solution rheology of flexible macromolecules bead-rod
model, J. Polym. Sci., Polym. Phys. Ed. 12, 21772187 1974.
Agarwal, U. S., Effect of initial conformation, flow strength, and hydrodynamic interaction on polymer
molecules in extensional flow, J. Chem. Phys. 113, 33973404 2000.
Ahn, K. H., J. L. Schrag, and S. J. Lee, Bead-spring chain model for the dynamics of dilute polymer solutions.
II: Comparisons with experimental data, J. Non-Newtonian Fluid Mech. 50, 349373 1993.
kerman, B., and E. Tuite, Single- and double-strand photocleavage of DNA by YO, YOYO, and TOTO,
A
Nucleic Acids Res. 24, 10801090 1996.
Amelar, S., C. E. Eastman, R. L. Morris, M. A. Smeltzly, T. P. Lodge, and E. D. von Meerwall, Dynamic
properties of low and moderate molecular weight polystyrenes at infinite dilution, Macromolecules 24,
35053516 1991.
Anna, S. L., G. H. McKinley, D. A. Nguyen, T. Sridhar, S. J. Muller, J. Huang, and D. F. James, An
interlaboratory comparison of measurements from filament-stretching rheometers using common test fluids, J. Rheol. 45, 83114 2001.
Armstrong, R. C., S. K. Gupta, and O. Basaran, Conformational-changes of macromolecules in transient
elongational flow, Polym. Eng. Sci. 20, 466 472 1980.
Aust, C., M. Kroger, and S. Hess, Structure and dynamics of dilute polymer solutions under shear flow via
nonequilibrium molecular dynamics, Macromolecules 32, 56605672 1999.
Babcock, H. P., R. E. Teixeira, J. S. Hur, E. S. G. Shaqfeh, and S. Chu, Visualization of molecular fluctuations
near the critical point of the coil-stretch transition in polymer elongation, Macromolecules 36, 4544 4548
2003.
Bao, X. Y. R., H. J. Lee, and S. R. Quake, Behavior of complex knots in single DNA molecules, Phys. Rev.
Lett. 26, 265506 2003.
Batchelor, G. K., Slender-body theory for particles of arbitrary cross section in Stokes flows, J. Fluid Mech.
44, 419 440 1970.
Bercea, M., C. Ioan, S. Ioan, B. C. Simionescu, and C. I. Simionescu, Ultrahigh molecular weight polymers in
dilute solutions, Prog. Polym. Sci. 24, 379 424 1999.
Bernal, J. M. G., M. M. Lopez, M. M. Tirado, and J. Freire, Monte-Carlo calculation of hydrodynamic
properties of linear and cyclic polymers in good solvents, Macromolecules 24, 593598 1991.
Bird, R. B., C. R. Curtiss, R. C. Armstrong, and O. Hassager, Dynamics of Polymeric Liquids, 2nd Ed. Wiley,
New York, 1987, Vol. 2.
ttinger, Transport properties of polymeric liquid, Annu. Rev. Phys. Chem. 43,
Bird, R. G., and H. C. O
371 406 1992.
Boger, D. V., A highly elastic constant-viscosity fluid, J. Non-Newtonian Fluid Mech. 3, 8791 1977.
Bouchiat, C., M. D. Wang, J. F. Allemand, T. Strick, S. M. Block, and V. Croquette, Estimating the persistence
length of a worm-like chain molecule from force-extension measurements, Biophys. J. 76, 409 413
1999.
Brochard, F., and P. G. de Gennes, Dynamical scaling for polymers in theta solvents, Macromolecules 10,
11571161 1977.
Burgers, J. M., Second report on viscosity and plasticity, Amsterdam Academy of Sciences, Amsterdam,
1938, Chap. 3.
Bustamante, C., J. F. Marko, E. D. Siggia, and S. Smith, Entropic elasticity of lambda-phage DNA, Science
265, 15991600 1994.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

66

LARSON

Callis, P. R., and N. Davidson, Hydrodynamic relaxation times of DNA from decay of flow dichroism
measurements, Biopolymers 8, 379390 1969.
Lopez Cascales, J. J., and J. Garca de la Torre, Shear-rate dependence of the intrinsic viscosity of bead-andspring chains: hydrodynamic interaction and excluded-volume effects, Polymer 32, 33593363 1991.
Cerf, R., La macromolecule en chaine dans un champ hydrodynamiqueTheorie generaleProprietes dynamo optiques, J. Polym. Sci. 23, 125150 1957.
Chopra, M., and R. G. Larson, Brownian dynamics simulations of isolated polymer molecules in shear flow
near adsorbing and nonadsorbing surfaces, J. Rheol. 46, 831 862 2002.
Clasen, C., M. Verani, J. P. Plog, G. H. McKinley, and W.-M. Kulicke, Effects of polymer concentration and
molecular weight on the dynamics of viscoelastocapillary breakup, Proc. XIVth Int. Congr. On Rheology
The Korean Society of Rheology, Seoul, Korea, 2004.
Cohen, A., A Pade approximant to the inverse Langevin function, Rheol. Acta 30, 270273 1991.
Cooke, I. R., and D. R. M. Williams, Stretching polymers in poor and bad solvents: Pullout peaks and an
unraveling transition, Europhys. Lett. 64, 267273 2003.
Cottrell, F. R., E. W. Merrill, and K. A. Smith, Conformation of polyisobutylene in dilute solution subjected
to a hydrodynamic shear field, J. Polym. Sci., Polym. Phys. Ed. 7, 14151434 1969.
de Gennes, P. G., Coilstretch transition of dilute flexible polymers under ultrahigh velocity gradients, J.
Chem. Phys. 60, 50305042 1974.
de Gennes, P. G., Scaling Concepts in Polymer Physics Cornell University Press, Ithaca, 1979.
de Gennes, P. G., Molecular individualism, Science 276, 1999 1997.
Doi, M., and S. F. Edwards, The Theory of Polymer Dynamics Clarendon Press, Oxford, 1986.
Doyle, P. S., E. S. G. Shaqfeh, and A. P. Gast, Dynamic simulation of freely draining flexible polymers in
steady linear flows, J. Fluid Mech. 334, 251291 1997.
Doyle, P. S., E. S. G. Shaqfeh, G. H. McKinley, and S. H. Spiegelberg, Relaxation of dilute polymer solutions
following extensional flow, J. Non-Newtonian Fluid Mech. 76, 79110 1998.
Ermak, D. L., and J. A. McCammon, Brownian dynamics with hydrodynamic interactions, J. Chem. Phys.
69, 13521360 1978.
Fang, L., and R. G. Larson, DNA configurations and concentration in shearing flow near a glass surface in a
microchannel, J. Rheol. in press.
Feng, J., and L. G. Leal, Numerical simulations of the flow of dilute polymer solutions in a four-roll mill, J.
Non-Newtonian Fluid Mech. 72, 187218 1997.
Ferry, J. D., Viscoelastic Properties of Polymers Wiley, New York, 1980.
Fetters, L. J., D. J. Lohse, D. Richter, T. A. Witten, and A. Zirkel, Connection between polymer molecular
weight, density, chain dimensions, and melt viscoelastic properties, Macromolecules 27, 4639 4647
1994.
Fixman, M., Polymer dynamicsNon-Newtonian intrinsic viscosity, J. Chem. Phys. 45, 793 803 1966.
Fixman, M., Simulation of polymer dynamics. I: General theory, J. Chem. Phys. 69, 15271537 1978.
Fixman, M., Construction of Langevin forces in the simulation of hydrodynamic interaction, Macromolecules 19, 1204 1207 1986.
Flory, P. J., The configuration of real polymer chains, J. Chem. Phys. 17, 303310 1949.
Flory, P. J., Statistical Mechanics of Chain Molecules Wiley, New York, 1969.
Fraenkel, G. K., Viscoelastic effect in solutions of simple particles, J. Chem. Phys. 20, 642 647 1952.
Fuller, G. G., and L. G. Leal, Birefringence of dilute polymer-solutions in two-dimensional flows, Rheol.
Acta 19, 580 600 1980.
Ghosh, I., G. H. McKinley, R. A. Brown, and R. C. Armstrong, Deficiencies of FENE dumbbell models in
describing the rapid stretching of dilute polymer solutions, J. Rheol. 45, 721758 2001.
Gottlieb, M., and R. B. Bird, A molecular dynamics calculation to confirm the incorrectness of the randomwalk distribution for describing the Kramers freely jointed bead-rod chain, J. Chem. Phys. 65, 24672468
1976.
Graessley, W. W., Polymeric Liquids and Networks: Structure and Properties Garland Science, New York,
2004.
Grassia, P., E. J. Hinch, and L. C. Nitsche, Computer simulations of Brownian motion of complex systems,
J. Fluid Mech. 282, 373 403 1995.
Grassia, P., and E. J. Hinch, Computer simulations of polymer chain relaxation via Brownian motion, J. Fluid
Mech. 308, 255288 1996.
Gupta, R. K., D. A. Nguyen, and T. Sridhar, Extensional viscosity of dilute polystyrene solutions: Effect of
concentration and molecular weight, Phys. Fluids 12, 1296 1318 2000.
Hagerman, P. J., Flexibility of DNA, Annu. Rev. Biophys. Biophys. Chem. 17, 265286 1988.
Hair, D. W., and E. J. Amis, Intrinsic dynamic viscoelasticity of polystyrene in and good solvents, Macromolecules 22, 4528 4536 1989.
Halperin, A., and E. B. Zhulina, On the deformation-behavior of collapsed polymers, Europhys. Lett. 15,
417 421 1991.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

67

ttinger, A detailed comparison of various FENE dumbbell models, J. NonHechen, M., and H. C. O
Newtonian Fluid Mech. 68, 17 42 1997.
Hernandez Cifre, J. G., and J. Garcia de la Torre, Steady-state behavior of dilute polymers in elongational
flow. Dependence of the critical elongational rate on chain length, hydrodynamic interaction, and excluded
volume, J. Rheol. 43, 339358 1999.
Hernandez Cifre, J. G., and J. Garca de la Torre, Kinetic aspects of the coil-stretch transition of polmer chains
in dilute solution under extensional flow, J. Chem. Phys. 115, 9578 9584 2001.
Hess, B., H. Bekker, H. J. C. Berendsen, and J. G. E. M. Fraaije, LINCS: A linear constraint solver for
molecular simulations, J. Comput. Chem. 18, 14631472 1997.
Hinch, E. J., Mechanical models of dilute polymer solutions for strong flows with large polymer deformations in Polymeres et Lubrification Brest, 223 mai, 1974 Colloques Internationaux du C.N.R.S. No
233, 241247 Editions du CNRS Paris 1975, ISBN 2-222-01735-1.
Hinch, E. J., Mechanical models of dilute polymer solutions in strong flows, Phys. Fluids 20, S22S30
1977.
Hinch, E. J., Uncoiling a polymer molecule in a strong extensional flow, J. Non-Newtonian Fluid Mech. 54,
209230 1994.
Hsieh, C. C., L. Li, and R. G. Larson, Modeling hydrodynamic interactions in Brownian dynamics: simulations of extensional flow in dilute solutions of DNA and polystyrene, J. Non-Newtonian Fluid Mech. 113,
147191 2003.
Hsieh, C. C., and R. G. Larson, Modeling hydrodynamic interactions in Brownian dynamics: simulations of
extensional and shear flows of dilute solutions of high molecular weight polystyrene, J. Rheol. 48, 995
1021 2004.
Hsieh, C. C., private communication 2003.
Hsieh, C. C., private communication 2004.
Hur, J. S., E. S. G. Shaqfeh, and R. G. Larson, Brownian dynamics simulations of single DNA molecules in
shear flow, J. Rheol. 44, 713742 2000.
Islam, M. T., S. A. Vanapalli, and M. J. Solomon, Inertial effects on polymer chain scission in planar elongational cross-slot flow, Macromolecules 37, 10231030 2004.
James, D. F., B. D. McLean, and J. H. Saringer, Presheared extensional flow of dilute polymer solutions, J.
Rheol. 31, 453 481 1987.
James, D. F., and T. Sridhar, Molecular conformation during steady-state measurements of extensional viscosity, J. Rheol. 39, 713724 1995.
Jendrejack, R. M., M. D. Graham, and J. J. de Pablo, Hydrodynamic interactions in long chain polymers:
application of the Chebyshev polynomial approximation in stochastic simulations, J. Chem. Phys. 113,
2894 2900 2000.
Jendrejack, R. M., J. J. de Pablo, and M. D. Graham, Stochastic simulations of DNA in flow: Dynamics and
the effects of hydrodynamic interactions, J. Chem. Phys. 116, 77527759 2002.
Jian, H. J., T. Schlick, and A. Vologodskii, Internal motion of supercoiled DNA: Brownian dynamics simulations of site juxtaposition, J. Mol. Biol. 284, 287296 1998.
Johnson, R. M., J. L. Schrag, and J. D. Ferry, Infinite-dilution viscoelastic properties of polystyrene in
-solvents and good solvents, Polym. J. Tokyo, Jpn. 1, 742749 1970.
Keller, A., and J. A. Odell, Extensibility of macromolecules in solutionA new focus for macromolecular
science, Colloid Polym. Sci. 263, 181201 1985.
Kishbaugh, A. J., and A. J. McHugh, A discussion of shear-thickening in bead-spring models, J. NonNewtonian Fluid Mech. 34, 181206 1990.
Klotz, L. C., and B. H. Zimm, Size of DNA determined by viscoelastic measurementsresults on bacteriophages, bacillus-subtilis and Escherichia-Coli, J. Mol. Biol. 72, 779 800 1972.
ttinger, Variance reduced Brownian simulation of a beadKroger, M., A. Alba-Perez, M. Laso, and H. C. O
spring chain under steady shear flow considering hydrodynamic interactions effects, J. Chem. Phys. 113,
4767 4773 2000.
Kubo, R., M. Toda, and N. Hasitsume, Statistical Physics II Springer, New York, 1985.
Kuhn, W., and H. Kuhn, Bedeutung beschrankt freier Drehbarkeit fur die Viskositat und Stromungsdoppelbrechnung von Fadenmolekellosungen I, Helv. Chim. Acta 28, 15331579 1945.
Kumar, K. S., and J. R. Prakash, Equilibrium swelling and universal ratios in dilute polymer solutions: Exact
Brownian dynamics simulations for a delta function excluded volume potential, Macromolecules 36,
78427856 2003.
Larson, R. G., Constitutive Equations for Polymer Melts and Solutions Butterworths, Boston, 1988.
Larson, R. G., The unraveling of a polymer chain in a strong extensional flow, Rheol. Acta 29, 371384
1990.
Larson, R. G., The Structure and Rheology of Complex Fluids Oxford, New York, 1999.
Larson, R. G., The role of molecular folds and pre-conditioning in the unraveling of polymer molecules
during extensional flow, J. Non-Newtonian Fluid Mech. 94, 37 45 2000.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

68

LARSON

Larson, R. G., Principles for coarse-graining polymer molecules in simulations of polymer fluid mechanics,
Mol. Phys. 102, 341351 2004a.
Larson, R. G., An explanation for the high-frequency elastic response of dilute polymer solutions, Macromolecules 37, 51105114 2004b.
Larson, R. G., T. T. Perkins, D. E. Smith, and S. Chu, Hydrodynamics of a DNA molecule in a flow field,
Phys. Rev. E 55, 1794 1797 1997.
Larson, R. G., H. Hu, D. E. Smith, and S. Chu, Brownian dynamics simulations of a DNA molecule in an
extensional flow field, J. Rheol. 43, 276 304 1999.
Larsson, A., C. Carlsson, M. Jonsson, and B. Albinsson, Characterization of the binding of the fluorescent dyes
YO and YOYO to DNA by polarized light scattering, J. Am. Chem. Soc. 116, 8459 8465 1994.
Lavery, R., A. Lebrun, J. F. Allemand, D. Bensimon, and V. Croquette, Structure and mechanics of single
biomolecules: experiment and simulation, J. Phys.: Condens. Matter 14, R383R414 2002.
Lee, E. C., M. J. Solomon, and S. J. Muller, Molecular orientation and deformation of polymer solutions under
shear: a flow light scattering study, Macromolecules 30, 73137321 1997.
Li, B., N. Madras, and A. D. Sokal, Critical exponents, hyperscaling, and universal amplitude ratios for
two-dimensional and three-dimensional self-avoiding walks, J. Stat. Phys. 80, 661754 1995.
Li, L., R. G. Larson, and T. Sridhar, Brownian dynamics simulations of dilute polystyrene solutions, J. Rheol.
44, 291322 2000.
Li, L., and R. G. Larson, Excluded volume effects on the birefringence and stress of dilute polymer solutions
in extensional flow, Rheol. Acta 39, 419 427 2000a.
Li, L., and R. G. Larson, Comparison of Brownian dynamics simulations with microscopic and light-scattering
measurements of polymer deformation under flow, Macromolecules 33, 14111415 2000b.
Li, L., H. Hu, and R. G. Larson, DNA molecular configurations in flows near a glass surface, Rheol. Acta in
press.
Li, X., and M. M. Denn, Monte Carlo simulation of steady extensional flows, J. Rheol. 48, 805 821 2004.
Lielens, G., R. Keunings, and V. Legat, The FENE-L and FENE-S closure approximations to the kinetic theory
of finitely extensible dumbbells, J. Non-Newtonian Fluid Mech. 87, 179196 1999.
Liu, T. W., Flexible polymer chain dynamics and rheological properties in steady flows, J. Chem. Phys. 90,
5826 5842 1989.
Liu, S., B. Ashok, and M. Muthukumar, Brownian dynamics simulations of bead-rod-chain in simple shear
flow and elongational flow, Polymer 45, 13831389 2004.
Lodge, T. P., Solvent dynamics, local friction, and the viscoelastic properties of polymer-solutions, J. Chem.
Phys. 97, 14801487 1993.
Lyulin, A. V., D. B. Adolf, and G. R. Davies, Brownian dynamics simulations of linear polymers under shear
flow, J. Chem. Phys. 111, 758 771 1999.
Mackay, M. E., and D. V. Boger, An explanation of the rheological properties of Boger fluids, J. NonNewtonian Fluid Mech. 22, 235243 1987.
Magda, J. J., G. H. Fredrickson, R. G. Larson, and E. Helfand, Dimensions of a polymer chain in a mixed
solvent, Macromolecules 21, 726 732 1988a.
Magda, J. J., R. G. Larson, and M. E. Mackay, Deformation-dependent hydrodynamic interaction in flows of
dilute polymer solutions, J. Chem. Phys. 89, 2504 2512 1988b.
Magda, J. J., and R. G. Larson, A transition occurring in ideal elastic liquids during shear flow, J. NonNewtonian Fluid Mech. 30, 119 1988.
Marko, J. F., and E. D. Siggia, Stretching DNA, Macromolecules 28, 8759 8770 1995.
Massa, D. J., Flow properties of high-molecular-weight DNA solutions: viscosity, recoil, and longest retardation time, Biopolymers 12, 10711081 1973.
Massa, D. J., J. L. Schrag, and J. D. Ferry, Dynamic viscoelastic properties of polystyrene in high-viscosity
solventsExtrapolation to infinite dilution and high-frequency behavior, Macromolecules 4, 210214
1971.
McKinley, G. H., and T. Sridhar, Filament-stretching rheometry of complex fluids, Annu. Rev. Fluid Mech.
34, 375 415 2002.
Meiners, J.-C., and S. R. Quake, Femtonewton force spectroscopy of single extended DNA molecules, Phys.
Rev. Lett. 84, 5014 5017 2000.
ttinger, Variance reduction simulations of polymer dynamics, J. Chem. Phys. 105,
Melchior, M., and H. C. O
3316 3331 1996.
Menasveta, M. J., and D. A. Hoagland, Light-scattering from dilute polystyrene solutions in uniaxial elongational flow, Macromolecules 24, 34273433 1991.
Michels, J. P. J., and F. W. Wiegel, On the topology of a polymer ring, Proc. R. Soc. London, Ser. A 403,
269284 1986.
Montesi, A., M. Pasquali, and F. C. MacKintosh, Collapse of a semiflexible polymer in poor solvent, Phys.
Rev. E 69, 021916 2004.
Morris, R. L., S. Amelar, and T. P. Lodge, Solvent friction in polymer-solutions and its relation to the
high-frequency limiting viscosity, J. Chem. Phys. 89, 6523 6537 1988.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

DILUTE SOLUTIONS OF FLEXIBLE POLYMERS

69

Morse, D. C., Theory of constrained Brownian motion, Adv. Chem. Phys. 128, 65189 2004.
Neelov, I. M., D. B. Adolf, A. V. Lyulin, and A. V. Davies, Brownian dynamics simulation of linear polymers
under elongational flow: bead-rod model with hydrodynamic interactions, J. Chem. Phys. 117, 4030 4041
2002.
Noda, I., Y. Yamada, and M. Nagasawa, Rate of shear dependence of intrinsic viscosity of monodisperse
polymer, J. Phys. Chem. 72, 28902898 1968.
Odell, J. A., and A. Keller, Flow-induced chain fracture of isolated linear macromolecules in solution, J.
Polym. Sci., Polym. Phys. Ed. 24, 18891916 1986.
Oono, Y., Crossover-behavior of transport properties of dilute polymer-solutionsRenormalization-group
approach 3, J. Chem. Phys. 79, 4629 4642 1983.
Orr, N. V., and T. Sridhar, Probing the dynamics of polymer solutions in extensional flow using step strain rate
experiments, J. Non-Newtonian Fluid Mech. 82, 203232 1999.
Osaki, K., A revised version of the integrodifferential equation in the Zimm theory for polymer solution
dynamics, Macromolecules 5, 141144 1972.
Oseen, C. W., Neuere Methode und Ergebnisse in der Hydrodynamik Akad. Verlag, Leipzig, 1927.
ttinger, H. C., Generalized Zimm model for dilute polymer solutions under theta conditions, J. Chem. Phys.
O
86, 37313749 1987.
ttinger, H. C., Gaussian approximation for Rouse chains with hydrodynamic interaction, J. Chem. Phys. 90,
O
463 473 1989.
ttinger, H. C., and W. Zylka, On the relaxation spectra for models of dilute polymer solutions, J. Rheol. 36,
O
885910 1992.
ttinger, H. C., Brownian dynamics of rigid polymer chains with hydrodynamic interactions, Phys. Rev. E
O
50, 2696 2701 1994.
ttinger, H. C., Velocity field in nondraining polymer chains, J. Chem. Phys. 35, 134 138 1996a.
O
ttinger, H. C., Stochastic Processes in Polymeric Fluids Springer, Berlin, 1996b.
O
Perkins, T. T., D. E. Smith, R. G. Larson, and S. Chu, Stretching of a single tethered polymer in a uniform
flow, Science 268, 83 87 1995.
Perkins, T. T., D. E. Smith, and S. Chu, Single polymer dynamics in an elongational flow, Science 276,
2016 2021 1997.
Peterson, S. C., I. Echeverra, S. F. Hahn, D. A. Strand, and J. L. Schrag, Apparent relaxation-time spectrum
cutoff in dilute polymer solutions: An effect of solvent dynamics, J. Polym. Sci., Part B: Polym. Phys. 39,
28602873 2001.
Poirier, M. G., and J. J. Marko, Effect of internal friction on biofilament dynamics, Phys. Rev. Lett. 88,
228103 2002.
Prakash, J. R., The kinetic theory of dilute solutions of flexible polymers: Hydrodynamic interaction, in
Advances in the Flow and Rheology of Non-Newtonian Fluids, D. A. Siginier, D. De Kee, and R. P. Chabra,
Eds. Elsevier, New York, 1999b.
Prakash, J. R., Rouse chains with excluded volume interactions in steady simple shear flow, J. Rheol. 46,
13531380 2002.
ttinger, Viscometric functions for a dilute solution of polymers in a good solvent,
Prakash, J. R., and H. C. O
Macromolecules 32, 2028 2043 1999.
Quake, S. R., H. Babcock, and S. Chu, The dynamics of partially extended single molecules of DNA, Nature
London 388, 151154 1997.
Rotne, J., and S. Prager, Variational treatment of hydrodynamic interactions in polymers, J. Chem. Phys. 50,
4831 4837 1969.
Rouse, P. E., A theory of the linear viscoelastic properties of dilute solutions of coiling polymers, J. Chem.
Phys. 21, 12721280 1953.
Ryckaert, J. P., and A. Bellemans, Molecular dynamics of liquid normal-butane near its boiling point, Chem.
Phys. Lett. 30, 123125 1975.
Ryckaert, J. P., G. Ciccotti, and J. Berendsen, Numerical integration of the Cartesian equations of motion of a
system with constraintsMolecular dynamics of n-alkanes, J. Comput. Phys. 23, 327341 1977.
Rzehak, R., and W. Zimmermann, Dynamics of strongly deformed polymers in solution, Europhys. Lett. 59,
779785 2002.
Schafer, L., Excluded Volume Effects in Polymer Solutions Springer, Berlin, 1999.
Schnurr, B., F. C. MacKintosh, and D. R. M. Williams, Dynamical intermediates in the collapse of semiflexible polymers in poor solvents, Europhys. Lett. 51, 279285 2000.
Schnurr, B., F. Gittes, and F. C. MacKintosh, Metastable intermediates in the condensation of semiflexible
polymers, Phys. Rev. E 65, 061904 2002.
Schroeder, C. M., H. P. Babcock, E. S. G. Shaqfeh, and S. Chu, Observation of polymer conformation
hysteresis in extensional flow, Science 301, 15151519 2003.
Schroeder, C. M., E. S. G. Shaqfeh, and S. Chu, The effect of hydrodynamic interactions on DNA synamics
in extensional flow: simulation and single molecule experiment, Macromolecules in press.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

70

LARSON

Shaqfeh, E. S. G., G. H. McKinley, N. Woo, D. A. Nguyen, and T. Sridhar, On the polymer entropic force
singularity and its relation to extensional stress relaxation and filament recoil, J. Rheol. 48, 209221
2004.
Shrewsbury, P. J., S. J. Muller, and D. Liepmann, Effect of flow on complex biological macromolecules in
microfluidic devices, Biomed. Microdevices 3, 225238 2001.
Shusterman, R., S. Alon, T. Gavrinyov, and O. Krichevsky, Monomer dynamics in double- and single-stranded
DNA polymers, Phys. Rev. Lett. 92, 048303 2004.
Smith, D. E., T. T. Perkin, and S. Chu, Dynamical scaling of DNA diffusion coefficients, Macromolecules 29,
13721373 1996.
Smith, D. E., and S. Chu, Response of flexible polymers to a sudden elongational flow, Science 281,
13351340 1998.
Smith, D. E., H. P. Babcock, and S. Chu, Single-polymer dynamics in steady shear flow, Science 283,
1724 1727 1999.
Solomon, M. J., and S. J. Muller, Study of mixed solvent quality in a polystyrene-dioctyl phthalatepolystyrene system, J. Polym. Sci., Part B: Polym. Phys. 34, 181192 1996.
Somasi, M., B. Khomami, N. J. Woo, J. S. Hur, and E. S. G. Shaqfeh, Brownian dynamics simulations of
bead-rod and bead-spring chains: numerical algorithms and coarse-graining issues, J. Non-Newtonian
Fluid Mech. 108, 227225 2002.
Sridhar, T., D. A. Nguyen, and G. G. Fuller, Birefringence and stress growth in unixial extensional flow of
polymer solutions, J. Non-Newtonian Fluid Mech. 90, 299315 2000.
Stigter, D., and C. Bustamante, Theory for the hydrodynamic and electrophoretic stretch of tethered B-DNA,
Biophys. J. 75, 11971210 1998.
Sundararajan, P. R., Theta temperatures, in Physical Properties of Polymers Handbook, J. E. Mark, Ed. AIP
Press, New York, 1996.
Tanner, R. I., Engineering Rheology Oxford University Press, New York, 1985.
Ten Brinke, G., and G. Hadziioannou, Topological constraints and their influence on the properties of synthetic
macromolecular systems. I: Cyclic macromolecules, Macromolecules 20, 480 486 1987.
Thompson, D. S., and S. J. Gill, Polymer relaxation times from birefringence relaxation measurements, J.
Chem. Phys. 47, 5008 5017 1967.
Thurston, G. B., Exact and approximate eigenvalues and intrinsic functions for gaussian chain theory, Polymer 15, 569572 1974.
Tirtaatmadja, V., and T. Sridhar, A filament stretching device for measurement of extensional viscosity, J.
Rheol. 37, 10811102 1993.
Underhill, P. T., and P. S. Doyle, On the coarse-graining of polymers in bead-spring chains, J. NonNewtonian Fluid Mech. 122, 331 2004.
Virk, P. S., Drag reduction fundamentals, AIChE J. 21, 625 656 1975.
Wedgewood, L. E., A Gaussian closure of the second-moment equation for a Hookean dumbbell with hydrodynamic interaction, J. Non-Newtonian Fluid Mech. 31, 127142 1989.
Woo, N. J., and E. S. G. Shaqfeh, The configurational phase transitions of flexible polymers in planar flows
near simple shear, J. Chem. Phys. 119, 2908 2914 2003.
Yamakawa, H., Modern Theory of Polymer Solutions, Harpers Chemistry Series Harper and Row, New York,
1971.
Yao, A., H. Matsuda, H. Tsukahara, M. K. Shimamura, and T. Deguchi, On the dominance of trivial knots
among SAPs on a cubic lattice, J. Phys. A 34, 75637577 2001.
Zherenkova, L., P. Khalatur, and K. Yoshikawa, Self-consistent integral equation theory for semiflexible
polyelectrolytes in poor solvent, Macromol. Theory Simul. 12, 339353 2003.
Zimm, B. H., Dynamics of polymer molecules in dilute solutionviscoelasticity, flow birefringence, and
dielectric loss, J. Chem. Phys. 24, 269278 1956; a corrected version is in Herman, J. J., Polymer
solution properties. II: Hydrodynamics and light scattering, Dowden, Hutchinson, and Ross, Stroudsberg,
1978, pp. 73 84.
Zylka, W., Gaussian approximation and Brownian dynamics simulations for Rouse chains with hydrodynamic
interaction undergoing simple shear flow, J. Chem. Phys. 94, 4628 4636 1991.
ttinger, A comparison between simulations and various approximations for Hookean
Zylka, W., and H. C. O
dumbbells with hydrodynamic interaction, J. Chem. Phys. 90, 474 480 1989.
ttinger, On the relaxation spectra for models of dilute polymer-solutions, J. Rheol. 36,
Zylka, W., and H. C. O
885910 1992.

Downloaded 07 Sep 2013 to 129.173.72.87. Redistribution subject to SOR license or copyright; see http://www.journalofrheology.org/masthead

Você também pode gostar