You are on page 1of 8

Brownian dynamics of a flexible polymer.

Internal modes and quaiselastic scattering


function
Antonio Rey, Juan J. Freire, and Jos Garcia de la Torre

Citation: The Journal of Chemical Physics 90, 2035 (1989); doi: 10.1063/1.456046
View online: http://dx.doi.org/10.1063/1.456046
View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/90/3?ver=pdfcov
Published by the AIP Publishing



















Advertisement:

This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
14.139.227.196 On: Mon, 21 Oct 2013 10:15:29
Brownian dynamics of a flexible polymer. Internal modes and quaiselastic
scattering function
Antonio Rey and Juan J. Freire
a
)
Departamento the Quimicas F/sica, Facultad de Ciencias Quimicas, Universidad Complutense, 28040
Madrid, Spain
JOSe Garcfa de la Torre
Departamento de Quimicas F/sica, Facultad de Ciencias Quimicas y Matematicas, Universidad de Murcia,
30001 Murcia, Spain
(Received 10 June 1988; accepted 26 October 1988)
Brownian Dynamics trajectories have been simulated by means of the Ermak and McCammon
algorithm for Gaussian polymer chains of different lengths with hydrodynamic interactions.
The results for relaxation times and dynamic scattering functions of chains with 2-20
statistical units are close to those calculated from the preaveraged Rouse-Zimm theory. The
differences between both sets of results are slightly higher than the range of uncertainties of the
simulation values (about 5% in most typical cases). Moreover, the simulated scattering
functions have been fitted to sums of exponentials in order to extract translational diffusion
coefficients and first relaxation times, following the Pecora procedure. The values obtained this
way are in satisfactory agreement with the results directly calculated from the trajectories.
INTRODUCTION
The dynamics of a flexible polymer chain in solution
constitutes a noncompletely solved theoretical problem. The
most popular description of the low frequency dynamics is
provided by the Rouse-Zimm theory 1. However, this theory
is based in an approximation, the preaveraging of hydrody-
namic interactions, whose validity is still an open question.
Since more rigorous treatments based on the generalized
Kirkwood theory can only be approached by means of per-
turbation techniques 1 or renormalization group theory,
2
simulations studies are currently among the more promising
choices. The use of Monte Carlo techniques to perform aver-
ages over nonpreaveraged results corresponding to instanta-
neously rigid conformations proposed by Zimm
3
yields re-
sults in good agreement with most experimental data,3-5 but
a possible coupling of hydrodynamic interactions with flexi-
bility6,7 may pose some doubts on the procedure accuracy.
In past years Brownian dynamics has become a very
useful alternative. It has been applied to flexible polymer
chains to investigate the fluctuating hydrodynamic interac-
tion problem
6
,7 and also to describe properties in shear
flow.
8
,9 Though some of these applications require rather
specialized algorithms, many general features of the Rouse-
Zimm theory that can be obtained with simpler procedures
have not been yet reproduced by dynamics simulation. Thus,
very recent simulations studies for the first relaxation times
only lead to qualitative conclusions since they have been per-
formed through Monte Carlo algorithms 10 or neglecting hy-
drodynamic interactions along the Brownian trajectories. 11
In this paper we employ the simple Ermak and McCam-
mon algorithm 12 to obtain equilibrium and hydrodynamic
a' Author to whom the correspondence should be addressed,
properties, the longest relaxation times and also the dynamic
light scattering function of flexible polymer chains with hy-
drodynamic interactions. The simplicity of the algorithm al-
lows us to obtain explicit trajectories from which different
properties can be easily calculated for any type of polymer
model, though more specialized methods can be much faster
for particular problems.
6
,7 The scattering function can be
related with translational and internal motions according to
the Pecora theory,13 also based in the Rouse-Zimm model.
The same algorithm has been previously utilized by Allison
and McCammon 14 for semiflexible, wormlike, models in cal-
culations of fluorescence depolarization (that depends
strongly on rapid motion) and dynamic light scattering.
However, the latter property was investigated for scattering
conditions that cannot monitor the internal motions. In our
present calculations, both translational and internal motions
contribute to the total scattering intensity. This way, our set
of results allows us to (a) ascertain the capability of the
algorithm to reproduce the different chain properties that
can be correctly (or at least with known reasonable accura-
cy) predicted by the theory for simple flexible chain models
and, therefore, to obtain in the future quantitative results for
other systems (branched polymers or chains with excluded
volume) for which the theoretical predictions can only be
currently expressed in terms of qualitative scaling laws
l5
;
(b) check the Rouse-Zimm and Pecora theories for condi-
tions and properties that could be significantly affected by
the preaveraging approximation.
METHODS
We have modeled the flexible polymer as a linear chain
constituted by N + 1 units, each one connected with its first
neighbors by means of harmonic springs so that their dis-
tances follow a Gaussian distribution whose rms is equal to b
(statistical segment length). From the hydrodynamic point
of view we consider the units as spheres, or beads, with iden-
tical friction coefficients S. Hydrodynamic interactions are
J. Chern. Phys. 90 (3), 1 February 1989 0021-9606/89/032035-07$02.10 1989 American Institute of Physics 2035
This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
14.139.227.196 On: Mon, 21 Oct 2013 10:15:29
2036 Rey, Freire, and Garcia de la Torre: Dynamics of a flexible polymer
introduced by a modified version of the Oseen tensor. We use
the diffusion tensor proposed by Rotne and Prager
l6
and
Yamakawa.
17
The diagonal terms of the tensor are given by
Dij = (kBT Is)1 (1)
(I is a 3 X 3 unit tensor). For i #-} the elements are
Dij = (3k
B
T 14s)(o/b) [I + RijRijlR t.
+ (2cr/3R t)(I-3RijRijIRt>1, ifRij>2CT (2)
or
Dij = (kBT Is) [(1 - 9Rij132CT) 1
+ (3/32CT)RijRijIRij]' if Rij<2CT, (3)
where Rij is the distance between beads i and), kB T is the
Boltzmann factor and CT is the Stokes radius of the beads,
whose value is determined from the parameter
h * = (3hr)I/2(CTlb) (4)
commonly employed in polymer hydrodynamics. We set
h * = 0 when hydrodynamic interactions are neglected (so
that 0 is diagonal in this case). Otherwise we use h * = 0.25,
a value convenient to describe the hydrodynamic behavior of
Gaussian units.
3
,5 Equation (3) serves to describe properly
the hydrodynamic interactions between overlapping beads.
The polymer motion is governed by the stochastic dif-
ferential equation of Langevin that we solve through the sim-
ple Ermak and McCammon algorithm. 12 Then, the position
of unit i, f;, can be calculated from his value after the preced-
ing step according to the equation of motion
f;(t+ at) =f;(t) +
j
i,)= 1,"'N + 1, (5)
where the superscript 0 refers to the instant at which the time
step begins. According to our model of spring connections,
the force FJ exerted by the neighbors of unit} is given by
F
j
= K(AR)j' (6)
where K = 3k
B
T Ib
2
is the adequate spring constant, R con-
tains the different position vectors fiO and A is defined as
Au=l, ifi=}=lorN+l,
Au = 2, if i =}#- 1 or N + 1,
Aij = - 1, if Ii - }I = 1,
(7)
Aij = 0, if Ii - }I#-O, 1.
The gradient term on the right-hand side of Eq. (5) is zero
when hydrodynamic interactions are ignored or when these
interactions are represented by the Rotne-Prager-Ya-
makawa tensor. Finally, p? is a Gaussian random vector
with zero mean and a variance-covariance matrix given by
(8)
In our calculations, we have used scaled dimensionless,
or "reduced," quantities based on the following basic units: b
for the length, S for the friction coefficient, and k B T for ener-
gy. This way, the time unit is given by sb
2
IkB T.
The time step should be small enough to ensure practi-
cally constant forces and diffusion tensor during the step,
but should be also long enough to allow us to achieve a con-
veniently long trajectory. To perform an empirical deter-
mination of at we have studied the averages of b, b
2
, b4, and
b
6
for chains of several units, using the criterion for trajec-
tory lengths that will be described below. We have found
results in good agreement with the expected values for a
Gaussian chain when the reduced step time, at *, is within an
interval whose upper limit is about at * = 0.01. Consequent-
ly, we have chosen this value as our time interval. Taking
T - 300 K, 710 = 1 cp, CT-l A, and b-l0 A we find that the
equivalent real time units for such a value is at - 5 X 10 - 12 s.
In order to know the maximum length for which the
trajectories are not affected by roundoff errors we have ob-
tained the results over three different trajectories generated
from different random number seeds. When the trajectories
are too long, the results calculated from the different sam-
ples diverge in a very remarkable way, perhaps as a short-
coming of the simplicity of the algorithm. We have extended
our calculations to 40 000 steps per trajectory, though the
final part of these trajectories were neglected since they ex-
ceeded the adequate length. Our final results are given as
statistical means over the three samples.
We have obtained trajectories for chains with
N + 1 = 2, 3, 4, 5, 6, 8, 11, 15, and 20 units (some results for
N + 1 = 35 have been also obtained, but only for h * = 0).
Then, we have investigated the following properties as aver-
ages over the results obtained for different values of t along
each trajectory:
(a) End-to-end distance: The quadratic average of this
property is calculated as
(R 2) = f
l
- fN+ 1)2). (9)
(b) Translational diffusion coefficient: It is obtained
from the quadratic displacement of the center-of-masses po-
sition vector, FCM (1'), according to the Einstein formula
FCM (1') = 1 2([If; (t) - If; (t + 1') ]2)
(N + 1); ;
= 6D,1'. (10)
Since extensive work has been already performed on this
property,6,7 our results obtained from the Einstein formula
are mainly aimed as another check on the correctness ofthe
algorithm.
(c) Internal modes: They describe approximately the
normal internal motions of the chain, as obtained from the
time-correlation functions PR and Pk as
PR (1') = (R(t)"R(t + 1',
(11)
(12)
where Uk is the k th Rouse coordinate, that for a finite chain
of N + I units is precisely given bylS
N+I
Uk(t) = L {2/(N + 1)}112
j= I
xcos[ U - 1/2)1Tk I(N + 1) ]R
j
(t). (13)
R
j
(t) is the position vector ofunitjwith respect to an inter-
nal frame centered in the first unit of the chain. (If h * = 0,
J. Chern. Phys., Vol. 90, No.3, 1 February 1989
This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
14.139.227.196 On: Mon, 21 Oct 2013 10:15:29
Rey, Freire, and Garcia de la Torre: Dynamics of a flexible polymer 2037
TABLE I. Quadratic mean end-to-end distance obtained from the trajector-
ies of chains of different lengths.
N+ 1
(R 2)/b
2
4 3.00 0.05
6 4.9 0.1
8 6.8 0.3
11 1O.00.7
15 14.80.1
20 18.8 0.9
these coordinates represent the normal modes of the chain.
When hydrodynamic interactions are present, there is not an
apparent way to define normal motions exactly, but the pre-
averaging approximation shows that the Rouse coordinates
can be considered normal as a very good approximation. I)
(d) Dynamic scattering/orm/actor It is obtained as
l3
S(q,1') = 1 2(Ni' Ni'e-;q[ri(,)-r}'+T)), (14)
(N + 1) ;= I j= I
where q is the scattering vector, depending on .-t, the radi-
ation wavelength, and 0, the scattering angle
Iql = (41TI.-t)sin(OI2).
RESULTS AND DISCUSSION
End-to-end vector
(15)
Our results for (R2 ) are summarized in Table I. It can
be observed that they reproduce the theoretical prediction,
(R2) = Nb
2
(within their error ranges), except in the case
N + 1 = 15, where we obtain a slightly higher value.
(Slightly abnormal results for this particular chain will be
also observed in the rest of properties.) It should be re-
marked that the dimension averages obtained by simulation
procedures present higher statistical deviations than other
properties, e.g., diffusion coefficients, since they have
broader distributions.
4
Then, the similarity between our re-
sults and the theoretical predictions can be considered as
satisfactory and confirms the validity of the choices for tra-
jectory parameters described in the preceding section.
Translational diffusion tensor
Our results for FCM (1') vs l' for each trajectory were
least-squares fitted to a straight line and the values of D, so
obtained were averaged over the three different samples.
These averages for the different chains corresponding to the
simulations with hydrodynamic interactions are shown, in
reduced units, in Table II. They are compared to the values
obtained from two different approximate theoretical ap-
proaches: (a) the preaveraged expression for D, according
to the Kirkwood-Riseman theory, 19 that can be also derived
from the Rouse-Zimm theory'S
Dr:z/(slkBn =
where, for the Gaussian model
H;;=l,
Hij = 21/2h *Ii - jl-l/2 fori=j.
(16)
(17)
(b) The values obtained by means of Monte Carlo nonpre-
averaged calculations performed on rigid structures
S
ac-
cording to the Zimm method, that should be considered as
lower bounds of the actual values.
20
It can be shown that our
results from the trajectories are in good agreement with
these sets of predictions for the general purposes explained in
the Introduction. Moreover, from a log-log plot of our re-
sultsofD, vsN + 1 we obtain the slope ( - 0.50 0.02), in
very good accordance with well-known scaling laws. IS Since
the preaveraged and rigid body values differ about 10% it
would be interesting to use the trajectory values to check
their validity. It can be observed that the trajectory and
Monte Carlo results are very similar for some of the chains,
while nonsystematic differences are found for the other
cases. However, our simple numerical methods yield results
with an estimated error range of about 5% and, therefore,
they cannot be safely used for such a discrimination. In fact
the correction to preaveraging for this property can be isolat-
ed in terms of a time correlation function. An exhaustive
investigation of this point has been performed by Fixman,6,7
using a considerably more especialized and elaborate algo-
rithm to investigate these particular differences for long
chains. His conclusions are too subtle and complex to be
summarized here.
Internal modes
An exponential behavior of the functions PR (1')
and P d 1') is theoretically predicted I, 13 for long values of l'
[e.g., see Eq. (8.8.12) of Ref. 13]. We have extracted the
relaxation times 1'R and 1'k corresponding to such behavior.
Thus, we have performed least-squares fittings of In [p R
TABLE II. Reduced translational diffusion coefficients of chains of different lengths calculated from the Rouse-Zimm theory (RZ), Monte Carlo averages
over rigid conformations (MC), the Einstein formula applied to our trajectories (BD-E), and fittings of the trajectory scattering functions (BD-LS).
Df = sD,/kBT.
N+I
D , 2 4 6 8 11 IS 20
RZ 0.667 0.468 0.380 0.328 0.279 0.239 0.207
MC 0.442 0.001 0.354 0.001 0.302 0.001 0.254 0.001 0.216 0.001 0.185 0.001
BD-E 0.64 0.02 0.43 0.02 0.31 0.02 0.30 O.o1 0.26 0.01 0.25 0.01 0.20 0.01
BD-LS(x= I) 0.67 0.01 0.45 0.01 0.33 0.04 0.30 0.01 0.26 0.02 0.24 0.01 0.19 0.03
BD-LS(x= 5) 0.740.09 0.52 0.04 0.28 0.08 0.4D 0.02 0.27 0.06 0.26 0.01 0.22 0.04
J. Chern. Phys., Vol. 90, No.3, 1 February 1989
This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
14.139.227.196 On: Mon, 21 Oct 2013 10:15:29
2038 Rey, Freire, and Garcia de la Torre: Dynamics of a flexible polymer
('T)/PR (0)] and In [Pk ('T)/Pk (0)] vs 'Tto a straight line and
obtained an estimation of ( - lIT;) from the slope values.
The fittings have been carried out for an intermediate time
range where short time effects are not present but the trajec-
tory numerical errors are still small. These intervals have
been selected by analyzing the first and second derivatives of
the numerical logarithmic functions calculated from the tra-
jectories and Eqs. (11 )-( 13). The results for the relaxation
times (in reduced units) obtained for h * = 0 and 0.25 are
contained in Table III. Means and error ranges correspond
to averages of the values obtained independently for the
three different trajectories.
'T R can be related to the orientational diffusion of the
end-to-end vector, and the 'Tk should very approximately
correspond to the relaxation times defined in the Rouse-
Zimm schemel
(18)
whereA
k
is the k th eigenvalue of matrix HA. The relaxation
times extracted from the trajectories should be identical to
those obtained from Eq. (18) if hydrodynamic interactions
are neglected, since in this case the chain moves according to
the scheme of Rouse normal modes. In this case (h * = 0),
the eigenvalues can be obtained from the simple analytical
formula
Ilk =Ak(h*=O) =4sin2[1rk/2(N+ 1)]. (19)
It can be observed in Table III that the results obtained for
k = 1 and k = 2 from the trajectories with h * = 0 are in
general good agreement with the theoretical values. This
confirms again the validity of our numerical procedures.
Moreover, log-log fittings of'TR and 'TI obtained from the
trajectories versus N + 1 yield slopes of 1.90 0.01 and
1.93 0.01, close to the slope 2 predicted by scaling laws
that should be considered as the long chain limit. For k = 2
we obtain the value 1.80 0.01 (the long chain behavior
should be expected for longer chains as k increases, as it can
be also shown from the theoretical values).
The results for 'Tk obtained with hydrodynamic interac-
tions can be compared with results obtained with Eq. (18)
and values of Ak calculated numerically from the diagonali-
zation of matrix HA with h * = 0.25. These results are also
included in Table III. The agreement with the trajectory
values is still good and, considering the uncertainty ranges of
the latter values, it is not possible to discern remarkable dif-
ferences that can be clearly attributed to preaveraging ef-
fects, though smaller values are systematically obtained with
the preaveraged theory for the longest chains. Log-log fit-
tings of 'T
R
, 'T
I
, and 'T
2
yield slopes of 1.46 0.01,
1.49 O.oI, and 1.49 0.01, in good accordance with the
slope 3/2 provided by scaling laws, asymptotic behavior
which is reached sooner than when hydrodynamic interac-
tions are not present.
DynamiC scattering form factor
We have calculated numerical values of the function
S(q,'T) according to Eq. (14) and using the position vectors
obtained along the trajectories. q is defined from the param-
eter x, related also with the mean square radius of gyration
IN b
2
2
X=j) a q,
where
(20)
Na=N[I+1/(N+l)]. (21)
We have employed the values x = 1 and 5. Hydrodynamic
interactions have been included in these calculations. The
form factor should be a real function and, consequently, the
imaginary term in Eq. (14) serves as an indicator of the
uncertainties involved in our calculations.
In Figs. 1-3 we present plots ofln [S(x,'T)/S(x)] VS'T
for three chains of different lengths. S(x) =S(x,O) is the
TABLE III. Reduced relaxation times for different chains with and without hydrodynamic interactions. Trajectory results obtained directly from the Rouse
coordinate positions (BD) and extracted by fittings of scattering function (BD-LS) are included, together with results calculated from the Rouse-Zimm
theory (RZ). 71 = k8TTj/tb2.
N+I
2 4 6 8 11 15 20 35
h*=O
~ B D ) 0.159 0.003 0.597 0.005 1.20 0.07 1.82 0.09 3.6 0.3 7.6 0.1 10.4 0.5 37 4
rT(BD) 0.163 0.004 0.586 0.004 1.232 0.007 2.05 0.05 4.1 0.4 8.00.1 14.2 0.8 43 3
rT(RZ) 0.167 0.569 1.244 2.19 4.11 7.63 13.5 41.4
'11 (BD) 0.161 0.001 0.332 0.002 0.598 0.006 1.11 0.03 1.69 0.01 3.5 0.1
'I1(RZ) 0.0833 0.167 0.333 0.569 1.05 1.93 3.41
h* =0.25
~ B D ) 0.245 0.005 0.708 0.004 1.1 0.1 1.79 0.01 3.2 0.4 5.5 0.1 7.21 0.04
rT(BD) 0.247 0.004 0.71 0.01 1.23 0.05 1.90 0.09 3.06 0.05 5.04 0.06 8.1 0.3
rT(RZ) 0.258 0.672 1.225 1.885 3.04 4.85 7.47
1"/'(BD-LS,x = 1) 0.18 0.02 0.60 0.02 1.3 0.2 1.7 0.3 2.6 0.3 41 8.3 0.2
1"/'(BD-LS,x = 5) 0.14 0.02 0.22 0.Q2 0.5 0.2 0.27 0.05 0.7 0.2 0.83 0.08 1.6 0.9
'11 (BD) 0.230 0.005 0.425 0.009 0.67 0.Q1 1.07 0.01 1.67 0.01 2.67 0.07
'I1(RZ) 0.247 0.416 0.622 0.986 1.557 2.386
'11 (BD-LS,x = 5) 2.40
a
B With theoretical estimates of the other fittings parameters (see the text).
J. Chem. Phys., Vol. 90, No.3, 1 February 1989
This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
14.139.227.196 On: Mon, 21 Oct 2013 10:15:29
Rey, Freire, and Garcia de la Torre: Dynamics of a flexible polymer 2039



-2

N+l = 3
...... -4
FIG. 1. In [S(x,'T)/S(x) 1 vs 'T for a
chain of N + 1 = 3 with hydrodynam-
ic interactions. (0) Results obtained
from the trajectories; (X) result ob-
tained previously (Ref. 18) from Eq.
(23).
toO_
X
-6
x=S
\

0.4 0.8 1.2 1.6

static form factor of the chain, that can be also evaluated
from the simple formula
The results obtained from Eq. (22) are in very good agree-
ment with the trajectory values, see Table IV. In Figs. 1-3 we
have also included numerical values calculated from the
preaveraged formula given by Pecora 13
1
N+IN+I
S(x) = 2 L L e-xji-JI/No.
(N+ 1) i=1 j=1
(22)

N+l=ll
-2
N
x
x = 1
r::'
x
If)
-
toO.
-4
FIG. 2. As Fig. 1 but with N + 1 = 11.
x
If)
1--1
.=
-6
x
x=5

2 4 6 8
KaT t/tb
2
J. Chern. Phys., Vol. 90, No.3, 1 February 1989
This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
14.139.227.196 On: Mon, 21 Oct 2013 10:15:29
2040 Rey, Freire, and Garcia de la Torre: Dynamics of a flexible polymer
0
--........
--........
-,----

N +1 = 20
-2
x
N

x=5
)( x
en
-
x
.... -4
)(
en
FIG. 3. As Fig. 1 butwithN + 1 = 20.

x
.E
-6

4 8
1 N+IN+I [ 1 N+I
S(q,r) = 2 I I exp - _q2 I b
2
p;: I
(N+l) i=1 j=1 6 k=1
X + QJk - 2QikQjke-TITk)], (23)
where P k is the k th eigenvalue of matrix A, calculated from
Eq. (19), Qij is an element of the transform matrix whose
columns contain the eigenvectors of HA, and the r k are cal-
culated from the actual eigenvalues A
k
, according to Eq.
( 18). These numerical values were evaluated in previous
work.
18
Comparing the trajectory and the Pecora (or
Rouse-Zimm) results for S(x,r) it can be observed that dif-
ferences between them are very small, though a systematic
downward deviation of the preaveraged results is observed
16
for the highest values of x and r. It should be remarked that,
though the range of r values required by actual experiments
should be wider for longer chains, the scattering function
shapes should not vary much from those exhibited in Figs. 2
or 3, when the time is conveniently scaled
l8
as rlq
2
D,.
According to Pecora,13 the scattering functions can be
expanded as
S(q,r) = So(x)e - q'D,T + S2(X)
Xexp[ - (q2D, + 2Irl)r) + .... (24)
We have employed a Gauss-Newton nonlinear least-squares
method to fit our numerical results for S(q,r) of each trajec-
tory to the equation
TABLE IV. Static scattering functions S(x): translational and principal intramolecular contributions. So(x) and S2(X), for chains of different lengths.
Results obtained for h = 0.25 from fittings of the trajectory scattering functions (BD), calculated from Eq. (22) (when indicated) and tabulated by Perico
etal.' (N + 1 = 100, h* =0.2).
N + 1
2 4 6 8 11 15 20 100
x=1
S(x) (BD) 0.753 0.003 0.738 0.003 0.741 0.005 0.740 0.009 0.731 0.009 0.739 0.006 0.736 0.009
S(x), Eq. (22) 0.7567 0.7401 0.7376 0.7368 0.7364 0.7361 0.7359
So(x) (BD) 0.717 0.005 0.708 0.005 0.69 0.02 0.700.02 0.71 0.01 0.714 0.006 0.68 0.06 0.730"
S2(X) (BD) 0.036 0.007 0.029 0.005 0.05 0.02 0.04 0.01 0.02 0.01 0.020.01 0.07 0.05 0.015'
x=5
S(x) (BD) 0.518 0.001 0.365 0.003 0.343 0.005 0.333 0.008 0.320 0.008 0.323 0.007 0.322 0.006
S(x), Eq. (22) 0.5178 0.3685 0.3417 0.3324 0.3268 0.3239 0.3224
So(x) (BD) 0.3 0.1 0.260.02 0.18 0.06 0.30 0.01 0.240.D4 0.28 0.01 0.26 0.05 0.208'
S2(X) (BD) 0.2 0.1 0.1D 0.02 0.16 0.05 0.04 0.Dl 0.08 0.04 0.042 0.002 0.060.04 0.080"
, Reference 21.
J. Chern. Phys., Vol. 90, No.3, 1 February 1989
This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
14.139.227.196 On: Mon, 21 Oct 2013 10:15:29
Rey, Freire, and Garcia de la Torre: Dynamics of a flexible polymer 2041
S(1") =ae-
bT
+ce-
dT
, (25)
obtaining this way estimations ofS
o
(x),S2 (x),Dpand 1"1 by
comparison of the parameters a, b, c, and d with Eq. (24).
These estimations (expressed as the averages over the results
obtained for three different trajectories) are contained in
Tables II-IV, where they can be compared with the numeri-
cal values calculated for these properties according to the
theories and methods described previously for D and 1"
, I'
while So(x) and S2 (x) have been previously evaluated by
Perico et a/.
21
Our estimations for So(x) and S2 (x) for the longest
chains are in good agreement with the theoretical values,
except in the case of S2 (x = 1), whose very small value can-
not be accurately estimated through a fitting procedure. The
estimations of D, and 1"\ from the scattering functions with
x = I are also in good agreement with the results obtained
more directly from the motion of the center of masses or the
Rouse coordinates, and also with the preaveraged values cal-
culated according to the Rouse-Zimm theory. All these co-
herent sets of data indicate the legitimacy of the Pecora pro-
cedure to extract both translational diffusion coefficients
and first relaxation times from fittings of the type given by
Eq. (25) in conditions similar to those we have simulated.
Also it shows the capability of our numerical algorithm to
obtain accurate estimations of physical characteristics of
chains even from fittings of complex properties as the scat-
tering form factor. However, the estimations corresponding
to scattering functions calculated with x = 5 are not good for
D, and clearly poor for 1" This disagreement with the cor-
rect values should be attributed to the influence of higher
terms in the expansion of S(q,1") given by Eq. (24), since this
function exhibits now a clear multiexponential behavior [see
the differences between S(x) and So(x) in Table IV]. A
more adequate expansion formula is, in this case,
S(q,1")

= e - q- ,T[ So(x) + SI,2 (x)e - TiT, + S2,II (x)e - 2TiT,
+
S (x)e - 2TIT, + S (x)e - 4TIT,]
2,22 4,1111 ,
(26)
where we have included the most significant preexponential
contributions according to the tabulated values of Perico et
a/.
21
Then, we have introduced these tabulated values in Eq.
(26), together with our estimations for D, and 1"1 obtained
with x = I, and have calculated 1"2 by numerical fitting to
the equation established this way. We have performed such a
fitting for our longest chain, N + 1 = 20. The result so ob-
tained, see Table III, is in good agreement with the preaver-
aged value and the evaluation performed from the motion of
the Rouse coordinates. However, it seems clear that such an
elaborate method cannot be easily employed in the general
study of multiexponential scattering functions obtained with
high values of x.
In summary, the Brownian dynamics algorithm used
through this work leads to dimensions of flexible chains in
agreement with the theoretical predictions. The transla-
tional diffusion coefficients are reasonably close to those ob-
tained from the preaveraged theory. Exhaustive studies of
the precise amount of these differences have been previously
reported by Fixman.
6
,7 Our work shows that the first and
second relaxation times obtained from the motion of the
Rouse coordinates along the trajectories are fairly close to
those predicted by the preaveraged Rouse-Zimm theory.
The scattering form factors predicted from the preaveraged
Pecora formula are also reproduced, though small but sys-
tematic preaveraging effects are detected. Of course, differ-
ences can be higher for other types of chains (e.g., branched
polymers). Finally, the values of translational diffusion co-
efficients and first relaxation times estimated from fittings of
these functions, according to the Pecora scheme, are also
satisfactory, except in cases where the scattering conditions
impose a multiexponential behavior with more than two
contributions. Therefore, the preaveraged theory is shown to
provide a remarkable quantitative description of the longest
internal motions and the quasielastic scattering function of
linear Gaussian chains with hydrodynamic interactions.
ACKNOWLEDGMENTS
This work was supported by Grant No. PB86/0012
(J.J.F.) and PR84/0561 (J.G.T.) from the Direccion Gen-
eral de Investigacion Cientifica y Tecnica. A.R. acknowl-
edges a Fellowship from the Plan de Formacion del Personal

IH. Yamakawa, Modern Theory of Polymer Solutions (Harper and Row,
New York, 1971).
2S._Q. Wang, J. F. Douglas, and K. F. Freed, J. Chern. Phys. 87 1346
(1987). '
3B. H. Zimm, Macromolecules 13, 592 (1980).
4J. Garcia de la Torre, A. Jimenez, and J. J, Freire, Macromolecules 15 148
(1982). '
5J. J. Freire, J. Pia, R. Prats, andJ. Garcia de la Torre, Macromolecules 17
1815 (1984). '
6M. Fixman, J. Chern. Phys. 78,1954 (1983).
7M. Fixman, J. Chern. Phys. 84,4085 (1986).
8H. H. Saab and P. J. Dotson, J. Chern. Phys. 86, 3039 (1987).
I:F. G. Diaz, J. Garcia de la Torre, and J. J. Freire (submitted).
J. P. Downey, C. C. Crabb, and J. Kovac, Macromolecules 19 2202
(1986). '
11M. van Waveren, M. Bishop, and J. P. J. Michaels, J. Chern. Phys. 88
1326 (1988). '
L. Ermak and J. A. McCammon, J. Chern. Phys. 69,1352 (1978).
B. Berne and R. Pecora, Dynamic Light Scattering (Wiley New York
1976). ' ,
::S. A. Allison and J. A. McCammon, Biopolymers 23, 363 ( 1984).
P. G. de Gennes, Scaling Concepts in Polymer Theory (Cornell Universi-
ty, Ithaca, 1979).
:? Rotne and S. Prager, J. Chern. Phys. 50, 4831 (1969).
H. Yamakawa, J. Chern. Phys. 53, 436 (1970).
18J. A. Escudero and J. J. Freire, Polym. J. 14,277 (1982).
Fixman and A. Horta, J. Am. Chern. Soc. 90, 3048 (1968).
M. Fixman, J. Chern. Phys. 78, 1588 (1983).
21A. Perico, P. Piaggio, and C. Cuniberti J. Chern. Phys. 62, 1269 (1975).
J. Chern. Phys., Vol. 90, No.3, 1 February 1989
This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
14.139.227.196 On: Mon, 21 Oct 2013 10:15:29

You might also like