POLYDISPERSITY INDEX FROM LINEAR VISCOELASTIC DATA: UNIMODAL

AND BIMODAL LINEAR POLYMER MELTS.

J. Llorens1, E. Rudé1 and R. M. Marcos2

Correspondence to: J. Llorens (E-mail: llorens@angel.qui.ub.es)

Franquès, 1, 08028 Barcelona, Spain

Països Catalans, 26, 43007 Tarragona, Spain

Abstract

This article describes a method for determining the polydispersity index Ip2 =

Mz M w of the molecular weight distribution (MWD) of linear polymeric materials

from linear viscoelastic data. The method uses the Mellin transform of the

relaxation modulus of a simple molecular rheological model. One of the main

features of this technique is that it enables interesting MWD information to be

obtained directly from dynamic shear experiments. It is not necessary to achieve

the relaxation spectrum, so the ill-posed problem is avoided. Furthermore, a

determinate shape of the continuous MWD does not have to be assumed in order

to obtain the polydispersity index. The technique has been developed to deal with

entangled linear polymers, whatever the form of the MWD. The rheological

information required to obtain the polydispersity index is the storage G'( ) and

loss G"( ) moduli, extending from the terminal zone to the plateau region. The

method provides a good agreement between the proposed theoretical approach

and the experimental polydispersity indices of several linear polymers for a wide

1
range of average molecular weights and polydispersity indices. It is also

applicable to binary blends.

Keywords: polymer rheology, polydispersity index, molecular weight

distribution.

1. Introduction

Most of the properties of a polymeric material are related to its molecular weight

distribution (MWD). The conventional analytical methods for determining the MWD

require dissolution of the sample. However, some polymers are either difficult to

dissolve or insoluble in common solvents. Furthermore, if available equipment is

not fitted with a sensor for determining absolute molecular weight, standard

polymers are needed to calibrate the instrument. In contrast, some rheological

techniques do not come up against these problems. The rheological technique

would appear to be a powerful tool for the molecular characterization of polymeric

materials. Indeed, in recent years much research has aimed to define the inter-

relationship between polymer rheological information and molecular weight

distribution [1-10].

In many technical applications, it is not necessary to know the whole molecular

weight distribution in order to characterize the material from a molecular point of

view; knowing an average molecular weight and a polydispersity index is

sufficient. The parameters normally used are the average molecular weight M w

and a polydispersity index M w / Mn or Mz / M w .

2
 It is widely accepted that the rheological parameters directly related to M w

and the polydispersity of the MWD, Mz / Mw , are the zero shear rate viscosity,

η 0 , and the steady-state compliance, J 0e , respectively.

The zero shear rate viscosity is a characteristic parameter for the entanglement

density between chains within a material, and hence it is very susceptible to chain

length. The steady state compliance is a parameter for melt elasticity which

characterizes the chain elastic deformation capacity between entanglements with

regard to their static equilibrium position. It is also a measure of the elastic energy

that is accumulated during flow and recoverable when the stress is removed.

Therefore, it is very susceptible to the high molecular weight fractions of the

material [11, 12].

The dependence of η 0 on the weight average molecular weight can be

described by the following power law:

o kE M w (1)

where kE is the constant of proportionality and generally lies between 3.3 and

3.7, except for extremely broad blends, for which higher values are observed

[13-16]. is often taken as 3.4, since this is the value usually obtained for most

linear polymers investigated.

The dependence of steady state compliance on molecular weight distribution

has also been widely studied because of its importance in polymer processing.

Mieras and van Rijn [17] observed that in the case of polydisperse entangled

polymer melts (polystyrene, PS, and polypropylene, PP) J 0e is not dependent on

molecular weight but is strongly influenced by the width of the MWD,

characterized by the parameter Mz / M w . This finding was consistent with similar

3
measurements in polydimethylsiloxane (PDMS) and PS melts, with 1 <

Mz / M w < 3, taken by Mills [18], who proposed the following empirical

relationship:

3.7
Mz
J e0 (2)
Mw

which has since been widely accepted.

Later, another empirical correlation was proposed by Agarwal [19] to relate

J 0e and molecular weight averages

MzMz 1
J e0 (3)
MnMw

this being derived by testing data from a set of binary blends of polystyrene

reported in the literature. For the particular case of a logarithmic-normal MWD, Eq.

(3) can be rewritten as

4
0 Mz
J e (4)
Mw

The two equations (2) and (4) give practically the same prediction not only for

the case of binary blends but also for the continuous broad distribution [20].

In addition, several blending laws have also been proposed for estimating the

effect of polydispersity on the linear viscoelastic properties of polymers. The

simplest one is Ninomiya’s linear blending law [21] for a mixture of two

monodisperse polymers with molecular weights M1 and M2:

H w1 H1 / 1 w2 H 2 / 2 (5)

where H( ) is the relaxation spectrum of the mixture, H1( ) and H2( ) are those

of the individual pure components, and w1 and w2 are their weight fractions. The

quantity i is a dimensionless shift factor that modifies the relaxation times of

4
molecules of component i relative to their values in the pure state. These shift

factors account for the effect of the molecular surroundings on the relaxation

time of each molecule. The presence of shorter chains speeds up the relaxation

time of longer ones, and vice versa. A simple shift factor that does this is

i M / Mi , where M is an average molecular weight of the MWD and is a

constant.

Assuming that the viscosity i and the steady-state compliance Jei of the pure

components are related to the molecular weight Mi as [20]:

3
i Hi d Mi (6)
0

2 1 6
J ei i 2
Hi d Mi (7)
0
i

this simple blending law predicts that J 0e is proportional to M z M w if i is

2
assumed to be M w M i and proportional to M z 1 M z 2 M w M z if i is

assumed to be M w M i . Kurata [20] proposed a variation of equation (5) based

on a mole-basis additivity instead of the weight-basis:

H x1 H1 / 1 x2 H 2 / 2 (8)

where xi represents the mole fraction of component i in the blend. If this is

combined with i M w M i , it leads to the Agarwal equation of J oe (Eq. 3).

Higher-order blending laws have also been proposed, for example, the

quadratic law of Bogue et al. [22]:

2 2
H w1 H1 / 1 2w1w2 H12 / 12 w2 H 2 / 2 (9)

and the cubic power-law of Kurata [23]. These quadratic and cubic blending

3.4
laws, with the empirical proportionality of η 0 to M w and certain assumptions

5
 2 3
about the cross terms, predict proportionality to M z M w and M z M w

respectively [20, 23].

There are two theoretical approaches to predicting the dependence of J 0e on

polydispersity in the literature. Graessley [12] estimated the effect of

polydispersity on J 0e by working out the equations given by the molecular theory

of viscoelasticity developed by Doi and Edwards for the dynamics of entangled

linear polymeric liquids [24, 25]. This molecular theory is based on the idea of a

tube model to represent the mutual constraints on configurational

rearrangement of the chains. According to this theory, Graessley [12] found the

following J 0e dependence on average molecular weights:

M z 2M z 3M z 4
J e0 (10)
M wM zM z 1

Similarly, Bird et al. [26], using the kinetic theory of polymeric liquids, or

phase-space theory, found the following relationship between steady-state

compliance and average molecular weights

M z 1M z 2
J e0 (11)
M wM z

Equations (10) and (11) show that both theoretical approaches predict that

the elastic behaviour of polymer melts, characterized by J 0e , is independent of

molecular weight for a monodisperse system but is strongly dependent on the

MWD for a polydisperse one, which is qualitatively in accordance with

experimental results. However, despite J 0e being quite sensitive to the presence

of high molecular weight components, the Doi-Edwards prediction is almost

certainly too strong. If one assumes a logarithmic-normal type distribution, the

Doi-Edwards theory predicts that:

6
 9
0 Mz
J e (12)
Mw

which reveals a marked discrepancy with the experimental results. In contrast,

the phase-space theory predicts a fourth power scaling relationship (Eq. 4),

which is in close agreement with experimental findings (Eq. 2).

It should be pointed out that the constant of proportionality is, in all cases, the

reciprocal of the plateau modulus, G No . Similarly, all the above relationships

emphasize the importance of the higher molecular weight averages in

determining the steady-state compliance.

Other determinations of the polydispersity index from rheological

measurements have been successfully used [27-30]. All of them, however, are

based on phenomenological hypotheses or empirical rules and some are only

valid for a particular class of polymer. Moreover, it is necessary to assume a

type of distribution law (Weslau (logarithmic-normal), Schulz, Poisson or Flory

type) for a quantitative analysis, as otherwise only qualitative information is

available.

The present study describes a quantitative rheological method for calculating

the polydispersity index, one that is based on a molecular theory and which can

be applied to any kind of polymer sample, regardless of the form of the

distribution. The rheological information required is the master curves of the

storage (G’- ) and loss (G”- ) moduli, extending from the terminal to the

plateau zones.

2. Theory

7
 Most of the molecular models used to infer the molecular weight of polymer

samples from linear viscoelastic data are based on the tube and reptation

concept of de Gennes [25]. However, it has proved difficult to incorporate

polydispersity into these models as the effect of the polydisperse environment

on the relaxation times of the individual chains has to be taken into account.

Graessley [31] introduced this polydispersity effect by postulating that the

terminal relaxation time i depends on two non-correlated processes: (i) the

reptation of the chain and (ii) the constraint release due to the motion of the

surroundings, which results in the tube renewal effect, a concept originally

developed by both Klein [32] and Daoud and de Gennes [33].

One of the most successful approaches to predicting the MWD of

polydisperse polymer samples is the double reptation mixing rule,

independently derived by both Tsenoglou [34] and des Cloizeaux [35] from the

entanglement concept and the double reptation mechanism, respectively.

Recently, other authors [5, 7] have refined the double reptation theory by

simultaneously taking into account tube renewal and the double reptation

mechanism of the chains.

The molecular model used in the present study was formulated and analysed

in a recent article [8]. Like most of the molecular models described in the

literature, this model is applicable to polymeric systems when all the chains are

long enough to form an entangled polymer network.

The model establishes that the relaxation function is well-approximated as a

simple exponential function, in reasonable agreement with the well-known tube

model for chain reptation in monodisperse polymers [24]. However, the

experimental spectrum for monodisperse polymers is broader than predicted by

8
the tube model. On the other hand, it is assumed that the effects of constraint

release on the relaxation time, i, of a chain of molecular weight Mi, depend on

an average molecular weight, M , that sets the effect of the environment where

the molecules reptate, according to

i = (kE / GNo )· M ·Mi (13a)

In this equation the parameters kE and are the same as previously described

for Eq. (1), which relates η 0 and M w , GNo is the plateau modulus and is a

parameter which determines the contribution of the molecular weight of a chain

in its relaxation time. Given a polymer with a definite MWD, Eq. (13a) can be

rewritten as

i = K’ Mi (13b)

Other models assume that the relaxation function of probe chains can be

expressed as a product of “chain reptation” and “constraint release” relaxation

processes [31].

It is known that probe chains dispersed in a higher molecular weight matrix

have their relaxation slowed, consistent with Eq. (13a). However, there is a

discrepancy between the proposed model predicting that suppression of

constraint release does not broaden the relaxation function and experiments,

which show broadening [36, 37].

From a molecular point of view, polymers are characterized by their MWD,

W(M), whereas the material function of linear viscoelasticity is often taken to be

the relaxation modulus, G(t). In this model the normalized linear relaxation

modulus for a polydisperse polymer is formulated by the following expression

t
G(t ) GNo exp W ( M ) dM (14)
0
i

9
 Here we expand the previous molecular model [8] to derive a polydispersity

index, for any kind of molecular weight distributions, from linear viscoelastic

data. The study uses the integral transforms of the relaxation modulus, derived

from classical viscoelastic theory, and the proposed molecular model. The

expression for a polydispersity index is achieved as follow:

From classical viscoelastic theory [38] the relaxation modulus is related to the

dynamic moduli by the following relationships:

2 G'
G(t ) sin t d (15a)
0

2 G"
G(t ) cos t d (15b)
0

In order to connect aspects of the molecular weight distribution with the

dynamic moduli we combine Eqs. (14) and (15) through their Mellin transforms.

The Mellin transform fM(s) of an arbitrary function f(t) is defined by

f M ( s) t s 1 f t dt (16)
0

t
where the parameter s is the transform variable. If f (t ) e , the Mellin

transform is, for s > 0:

t
f M ( s) ts 1 e dt s
s (17)
0

(s) being the gamma function of parameter s. Interesting properties of the

Mellin transform can be used to solve problems of classical physics [39]. In

particular, the Mellin transform is usefull to calculate moments of time functions

[40]. Thus, the Mellin transform of the relaxation modulus expressed in terms of

t
the relaxation spectrum ( G(t ) H( ) e d ln ) gives:

t
GM (s) H( ) ts 1 e dt d ln ( s) H( ) s
d ln (18)
0

10
 s
where H( ) d ln is the sth moment of the of the relaxation spectrum.

Moreover, polydispersity indexes of the relaxation spectrum can be obtained by

the quotients of these moments. Therefore, if we know the relation between the

relaxation times and the polymer molecular weights, Eq. (13), we can derive

molecular polydispersity indexes. Then, taking into account Eq. (13b), the Mellin

transform of the relaxation modulus (Eq. 14), GM(s), when s > 0 is:

t s
GM ( s) GNo t s 1 exp dt W ( M ) dM GNo s K' M sW M dM
0 0 K'M 0

(19)

where M sW M dM is the sth moment of the molecular weight distribution,

Mβs . Then Eq. (19) can be expressed as:

s
GM (s) = GNo s K' M s
(20)

Similarly, the Mellin transform of the relaxation modulus (Eq. 15) when 0 < s <

1 is:

2 G 2 s G
GM s t s 1 sin t dt d s sin s 1
d (21a)
0 0 2 0

2 G" 2 s G"
GM s t s 1 cos t dt d s cos s 1
d (21b)
0 0 2 0

Eqs. (20) and (21) can be combined to yield an expression for arbitrary

moments of M with respect to the distribution function W(M) as a function of the

experimental dynamic moduli:

s
2 sin
s 2 G
M 0 s s 1
d (22a)
0
GN K

11
 s
2 cos
s 2 G
M 0 s s 1
d (22b)
0
GN K

where 0 < s < 1. Taking s = n, where n is an integer, Eqs. (22a, 22b) yield the

nth moment of the MWD. The moments and the average molecular weights of

the MWD are related as follows:

1
M = Mn (23a)

M 1 = Mw (23b)

M 2 = Mw · M z (23c)

and so forth.

It is known that > 0, so the -1st moment of the MWD, i.e. Mn , cannot be

determined, since s would be negative. However, if > 2 the +1st and +2nd

moments of the MWD can be determined, since in these cases s lies between 0

and 1. Therefore, we can use Eq. (22) to get Mw and Mz if we know K’, G No and

the dynamic moduli. Usually, K’ is an unknown parameter. However, if we focus

2
on the polydispersity index Mz / Mw = M 2 / M 1 the parameter K’ is not

needed, and:

G
sin 2
d
0 1
Mz
GNo 2 2
(24a)
Mw 2
sin G
2 1
d
0 1

G"
cos 2
d
0 1
Mz
GNo 2 2
(24b)
Mw 2
cos G"
2 1
d
0 1

12
 The parameter must fall in the range 2 < < . The lower limit corresponds

to the minimum value from which the polydispersity index M z M w can be

calculated. The upper limit is determined by Eq. (13a) with the condition - >

0, because any relaxation time, i, of a chain of molecular weight Mi, must

increase when M Mi and diminish when M Mi with respect to the relaxation

time in the monodisperse system ( M Mi ).

Therefore, only the master curves of G'( ) and G"( ) are required to

calculate Mz / M w since the plateau modulus is also calculated from Eq. (15b)

by applying the time limit to the relaxation modulus ( lim G(t ) t 0 GN0 ):

2 G"
GN0 d (25)
0

It should be emphasised that the entire master curves, extending from the

terminal zone to the rubbery plateau region, are needed to calculate the

complete integrals. Otherwise, reliable values for the polydispersity index will

not be obtained.

We have analyzed the effect of the experimental broadening of the spectrum,

for monodisperse polymer, on the polydispersity index calculated, in

polydisperse systems, with Eq. (24). As we mentioned above, the spectrum for

a simple exponential function, in the model relaxation function, is narrower than

the experimental spectrum for monodisperse polymers. Eq. (24) calculates the

polydispersity index from a relaxation time distribution (RTD). If a monodisperse

system has a polydisperse RTD, the calculated polydispersity index will be

greater than 1. Therefore, when we calculate, with Eq. (24), the polydispersity

index, in polydisperse systems, we obtain a polydispersity index, which is the

13
product of two numbers: one that is the true molecular polydispersity index and

the other that is related with the RTD of the monodisperse chains. Then, if we

apply Eq. (24) to experimental monodisperse system data, the polydispersity

index obtained will be a factor that can be used to calculate the true molecular

polydispersity index of polydisperse systems. The true molecular polydispersity

index of polydisperse systems can be obtained by dividing their calculated

polydispersity index by this factor. This correction, due to the non-monodipersity

of the RTD in monodisperse systems, seems to be unimportant in the

calculation of molecular polydispersity indexes, as we can see from the

comparison between experimental and calculated molecular polydispersity

indexes obtained with Eq. (24).

As mentioned above, the steady state compliance is strongly dependent on

polydispersity and, according to our theoretical approach, this dependence must

be obtained as follows.

From the proposed relaxation modulus (Eq. 14), the relaxation spectrum,

H( ), can be related to the MWD by [8]

G No
H M W (M ) (26)

where the relaxation time, , is described by Eq. (13b). In addition, the steady

state compliance can be calculated from the relaxation spectrum by [38]

H d
J e0 0
2
(27)
H d
0

Combining equations (13b), (25) and (26), and taking d = K’M -1dM, the

following J 0e dependence on MWD is obtained

14
 2
0 1 M
J e 2
(28)
G No M

Note that Eq. (28) is a general relationship, for which any value of is

possible, provided that it lies in the correct range of 2 – . Therefore, the J 0e

molecular dependence is not restricted to integer moments of W(M). In

particular, it can be observed that Eq. (28) includes the relationships predicted

by Doi-Edwards (Eq. 10) and Bird et al. (Eq. 11), which can be formulated as a

special case for = 3 and = 2, respectively.

For the particular case of a logarithmic-normal type distribution, the model

predicts that J 0e scales with the polydispersity index Mz / M w according to

0 Mz
J e for 2 < < 3.4 (29)
Mw

Given that must be greater than 2, the power exponent with which J 0e

scales with I p 2 M z M w will be greater than 4. This leads to greater

differences with the empirical value of 3.7 [18] as goes up. However, Agarwal

[19] examined several sets of data from the literature and concluded that while

the equation he proposed (Eq. 3) holds for all of them, Eq. (2) does not. Indeed,

scaling exponents higher than 5.3 have been reported.

It is worth noting that the MWD dependence of J 0e predicted by the model

(Eq. 28) is the same as that predicted by a generalized form of Ninomiya’s

blending law (Eq. 5). This generalized blending law is based on the following

equations:

H( ) wi H i / i (30)
i

15
where i M Mi and = - ,

i Hi d Mi (31)
0

1 2
J ei 2
Hi d Mi (32)
0
i

3. Experimental

3.1. Polymer samples

Different kinds of polymers with different degrees of polydispersity were

analysed. They were: polydimethylsiloxane (PDMS), polyisoprene (PI), two

series of polystyrene (PS), polydisperse and nearly monodisperse PS, and a

random co-polymer of ethylene and propylene (EP). Binary blends of PDMS, PI

and nearly monodisperse PS were also studied. Some of the samples analysed

in this study had also been analysed in the previous work [8]. However, binary

blends and nearly monodisperse PS are new. Moreover, the previous study

focused on unimodal log-normal MWD while the present study extends to the

polydispersity index for any kind of MWD, be it unimodal or bimodal.

The rheology of PDMS and PI samples was analysed in our laboratory. They

were purchased from Aldrich, who also supplied the average molecular weight

information. Three bimodal PDMS mixtures were prepared by blending a

sample of low molecular weight (PDMS_L) with another of high molecular

weight (PDMS_H), adding either 75%, 46% or 25% (in weight) of the latter. In

what follows, the letters H and L in the short forms stand for high and low

molecular weight component respectively. Two binary PI blends were also

16
prepared in the same way, with 90% and 80% of the high molecular weight

component (PI_H).

The molecular weight distributions for the three polymers examined (PDMS_H,

PDMS_L and PI_H) were determined by GPC in a Waters liquid chromatograph

with a light scattering detector at 75 °C. The polymer solutions were prepared at a

concentration of 0.05% in toluene and injected into the system at an injection

volume of 50 L. A Styragel HR 5E column was used with toluene as the mobile

phase at a flow rate of 0.35 mL/min. The average molecular weights of the blends

were calculated from the known composition of the mixture [41].

Rheological data for EP and the polydisperse PS samples were taken, with

permission, from the work of Wasserman and Graessley [42]. The PS sample

was prepared by mixing nearly monodisperse polystyrenes in order to get a

precise logarithmic-normal distribution (referred to as sample M2 in the work of

Wasserman and Graessley [42]).

Other PS rheological data are also reported in the present study, and these

were generously provided by Maier et al. [43]. They consist of a narrowly

distributed PS sample (PS_L), referred to as PS60 in the work of Maier et al.

[43], and a series of bimodal mixtures of this polymer, made by solution

blending, with 10%, 20% and 40% (in weight) of a high molecular weight PS

(PS_H) (referred to as PS177 in the work of Maier et al. [43]). As before, the

average molecular weights of the blends were calculated from the known

composition of the mixture [41].

The binary blends of the polymers analysed in the present study are

hereinafter referred to by their short form: POLYMER_H00L00, where 00

17
indicates the content (% in weight) of high (H) and low (L) molecular weight

components.

The average molecular weights and the polydispersity indices Ip1 and Ip2 of all

the samples investigated are listed in Table 1.

3.2. Rheological measurements

The rheological data of the samples analysed in our laboratory were obtained

by dynamic oscillatory tests in a controlled stress rheometer (HAAKE RS100)

with parallel plates of 20 mm diameter (gap = 0.4 mm), except for the PDMS_L

sample for which a cone-and-plate sensor of 20 mm diameter and 2º angle was

used. All measurements were taken in the linear viscoelastic region. The PDMS

samples were analysed at three different temperatures: 50 ºC, 0 ºC and -40 ºC

(-50 ºC for the PDMS_L sample). Those of PI were analysed at 80 ºC and 110

ºC. All PI samples were studied under nitrogen atmosphere in order to prevent

oxidative degradation. The isotherms were shifted in order to build up the

master curves at a reference temperature of T0 = 0 ºC and 110 ºC for PDMS

and PI, respectively, using the time-temperature superposition principle.

Details of the rheological characterisation of the other polymers used in this

study can be found in the original papers [42, 43].

Figures 1 to 3 show the dynamic moduli G' and G" master curves for the

PDMS, PI and PS mixtures series. Figures 4 and 5 show respectively the

dynamic modulus master curve data for EP and polydisperse PS samples.

4. Results and discussion

18
 The aim of the proposed method was to calculate the polydispersity index,

Ip2, from the dynamic moduli of linear polymers. This was done using equations

(24a or b) and the plateau modulus ( GNo ) obtained through Eq. (25). To

calculate both parameters, Ip2 and G No , several integrals must be calculated

from zero to infinity with respect to frequency. Therefore, experimental

frequency window must be large enough to calculate these integrals accurately.

4.1. Frequency window requirements

The whole frequency evolution of the dynamic moduli, G' and G", from the

terminal to the rubbery plateau region, is needed to achieve accurate

polydispersity indices. This means that G' must evolve from a two scaling law
2
with (G' ) at the low frequency range (terminal zone) to a constant value

of G', namely G No , at the high frequency range (plateau region). In addition, G"

must evolve from a scaling law behaviour with a power exponent equal to one

(G" ) to a scaling behaviour with a power exponent characteristic of the

terminal region are more readily obtained by observing the scatter of the

storage dynamic shear compliance data points, J’ = G’’/(G’2+G’’2), as J 0e is

asymptotically approached. In the terminal zone only functions not influenced by

viscous deformation or relaxation help ascertain the level of precision of

dynamic properties [44]. On the other hand, since polymers with narrow MWD

have J 0e ’s which are 2.5 to 3.0 times larger than their entanglement plateau

compliances, we have to take care that the false plateau in G’ which reflects the

reciprocal of J 0e was not interpreted as the genuine rubbery entanglement

19
plateau, GNo , [44]. In these cases, we need to explore higher frequencies than

those leading to the plateau where G’ reflects the reciprocal of J 0e until GNo is

finally reached. Therefore, the experimental data must extend over a wide

frequency range and this can be achieved for most polymeric materials with the

aid of the time-temperature superposition principle. Figures 1 to 5 show that the

whole evolution from terminal to plateau region is reached in all cases.

However, in some cases it is possible to get practical results by extrapolating

the experimental data in both frequency limits (low and high), according to the

rules given above. An example of the resulting dynamic moduli after the

frequency window extension by extrapolation is shown in Figure 4.

The polydispersity index can be calculated independently from storage or

loss data. Obviously, however, both sets of data must reach the same Ip2

parameter, and this was successfully tested for several of the analysed

systems. However, from a practical point of view, it is simpler to use the storage

modulus G'( ) (Eq. 24a), since the extension of this parameter at high

frequencies is easier ( G' (ω) ω G No ). At the same time, this material

parameter can be found in the literature for most common polymers at different

temperatures. Given this, we recommend using the G' modulus for the I p2

determination, focusing attention on the integrals being completed.

It should be pointed out, however, that the molecular model used is only

applicable from the terminal to the rubbery plateau region [8]. Therefore, glassy

modes, that could appear at high frequencies in the dynamic moduli master

curve data, must be removed.

4.2. Choice of the optimal parameter

20
 The polydispersity index was calculated by fitting the model described by Eq.

(24) to the measured dynamic data, leaving as an adjustable parameter. In all

samples it is adjusted into the valid range of 2 < < 3.4 by trial and error in

order to give the known experimental value of the parameter Ip2. The Ip2

parameter values were determined by SEC analysis for pure components and

by calculations from the known composition of the mixture [41] for the polymer

blends. These values are listed in Table II, together with parameter adjusted

for all the analysed samples in order to have the correct Ip 2 for the rheological

method presented in this study.

To choose the optimal value valid for all the polymers examined, the

evolution of the relative error in Ip2 with was analysed. Figure 6 shows the

results for the five different polymer series. It should be noted that the curves

corresponding to the PDMS, PI and nearly monodisperse PS series show

weighted mean values of the individual relative errors. The optimal values for

these five series are: 1 = 2.5 (for PDMS), 2 = 2.7 (for PI), 3 = 2.5 (for EP), 4

= 2.3 (for polydisperse PS) and 5 = 2.5 (for nearly monodisperse PS), also

listed in Table 2. The weighted mean value of for all the polymer series is

= 2.50 0.14

which is indicated as a vertical dashed line in Fig. 6. Therefore, 2.50 was

chosen as the best value for calculating Ip2 = Mz M w with the proposed

rheological method.

The polydispersity indices were then recalculated with = 2.5 and the

obtained values are listed in Table III. The greatest discrepancies are observed

for high polydisperse samples. However, except for these systems the relative

21
error never exceeds 12%, regardless of the polydispersity or distribution type

form (unimodal or bimodal). The small discrepancies could be reduced if

different values were assigned to the different polymer series, although this

seems unnecessary from a practical point of view. All complex relaxation

modulus data were established with standard experimental error. Each

experimental datum was determined by conventional replication until error does

not exceed 5%. Special emphasis was laid on the use of the complete

frequency evolution of the dynamic moduli, taking into consideration the factors

pointed out in the terminal and the rubbery plateau regions.

5. Conclusions

This article has described a theoretical model for determining the

polydispersity index, Mz / Mw , of linear polymer melts directly from experimental

rheological dynamic data. The model has been shown to be useful for predicting

the polydispersity of different kinds of polymer samples and has certain

advantages over other rheological techniques used for this purpose. As it is

based on a molecular theory, rather than empirical rules, and predicts

polydispersity in an absolute mode, no calibration is needed. Furthermore, it

can be applied to any entangled linear polymeric material, whatever its nature

and distribution form type (uni- or bimodal). A MWD form does not have to be

assumed a priori; only the storage G' and loss G" moduli need to be known.

However, for an accurate calculation of the polydispersity index, the

experimental dynamic data must extend across the frequency window from the

terminal to the rubbery plateau region.

22
 The model was successfully tested on several polymers, using both data

from our laboratory and experimental data already reported in the literature.

This confirms the model’s validity and suggests it can be applied more widely, in

particular to monodisperse samples.

The model uses the parameter , which determines the contribution of the

molecular weight of a chain in its relaxation time. = 2.50 was found to be

adequate for the different polymer series analysed. Only minor discrepancies

were found for polydispersity indices higher than 3.

Acknowledgements

The authors are grateful to the Spanish “Comisión de Investigación

Científica y Técnica” (projects QUI999-0533, PPQ2001-0671) for the funding

that enabled this work to be carried out.

References

1. Wu S. Polym Eng Sci 1985;25:122-128.

2. Tuminello WH. Polym Eng Sci 1986;26:1339-1347.

3. Mead DW. J Rheol 1994;38:1797-1827.

4. Wasserman SH. J Rheol 1995;39:601-625.

5. Carrot C, Guillet J. J Rheol 1997;41:1203-1220.

6. Maier D, Eckstein A, Friedrich Cr, Honerkamp J. J Rheol 1998;42:1153-1173.

7. Léonardi F, Majesté JC, Allal A, Marin G. J Rheol 2000;44:675-692 and

Léonardi F, Allal A, Marin G. J Rheol 2002;46:209-224.

8. Llorens J, Rudé E, Marcos RM. J Polym Sci Part B: Polym Phys

2000:38:1539-1546.

23
9. Thimm W, Friedrich Cr, Marth M. J Rheol 2000;44:429-438.

10. Nobile MR, Cocchini F. Rheol Acta 2001;40:111-119.

11. Ferry JD. Viscoelastic Properties of Polymers, 3rd ed. New York: Wiley,

1980. p.19.

12. Graessley WW. J Polym Sci: Polym Phys Ed 1980;18:27-34.

13. Ferry JD. Viscoelastic Properties of Polymers, 3rd ed. New York: Wiley,

1980 (Chapter 13).

14. Montfort JP, Marin G, Monge Ph. Macromolecules 1984;17:1551-1560.

15. Montfort JP, Marin G, Monge Ph. Macromolecules 1986;19:1979-1988.

16. Carrot C, Revenu P, Guillet J. J Appl Polym Sci 1996;61:1887-1897.

17. Mieras HJMA, van Rijn CFH. Nature 1968;218:865-866.

18. Mills NJ. Nature 1968;219:1249-1250.

19. Agarwal PK. Macromolecules 1979;12:342-344.

20. Kurata M. Macromolecules 1984;17:895-898.

21. Ninomiya K. J Colloid Sci 1959;14:49-58 and 1962;17:759-767.

22. Bogue DC, Masuda T, Einaga T, Onogi S. Polym J 1970;1:563-572.

23. Kurata M, Osaki K, Einaga Y, Sugie T. J Polym Sci: Polym Phys Ed

1974;12:849-869.

24. Doi M, Edwards SF. J Chem Soc. Faraday Trans II 1978;74:1789-1801;

74:1802-1817; 74:1818-1832.

25. Doi M, Edwards SF. The Theory of Polymer Dynamics. Oxford: Clarendon

Press, 1986.

26. Bird RB, Curtis CF, Armstrong RC, Hassager O. Dynamics of Polymeric

Liquids, Vol. 2, 2nd ed. New York: Wiley-Interscience, 1987 (Chapter 19).

27. Couarraze G, Grossiord JL, Puisieux F. Rheol Acta 1986;25:494-500.

24
28. Grossiord JL, Couarraze G. Eur Polym J 1988; 24:259-264.

29. Tzoganakis C. Can J Chem Eng 1994;72:749-754.

30. Hatzikiriakos SG. Polym Eng Sci 2000;40:2279-2287.

31. Graessley WW. Adv Polym Sci 1982;47:67-117.

32. Klein J. Macromolecules 1978;11:852-858.

33. Daoud M, de Gennes PG. J Polym Sci, Polym Phys Ed 1979;17:1971-1981.

34. Tsenoglou C. ACS Polym Prep 1987;28:185-186.

35. des Cloizeaux J. Macromolecules 1990;23:4678-4687.

36. Schroeder MJ, Roland CM. Macromolecules 1999;32:2000-2003.

37. Adachi K, Wada T, Kawamoto T, Kotaka T. Macromolecules 1995;28:3588-

3596.

38. Ferry JD. Viscoelastic Properties of Polymers, 3rd ed. New York: Wiley,

1980 (Chapter 3).

39. Sneddon IN. The Use of Integral Transforms, New York: McGraw-Hill, 1972.

p.262-297.

40. Bracewell RN. The Fourier Transform. and Its Applications, 3rd ed. Boston:

McGraw-Hill, 2000. p. 343-347.

41. Comstock KJ. ACS Symposium Series 1982;197:45-64.

42. Wasserman SH, Graessley WW. J Rheol 1992;36:543-572.

43. Maier D, Eckstein A, Friedrich Cr, Honerkamp J. J Rheol 1998;42:1153-

1173.

44. Plazek, D.J. J Rheol 1992;36:1671-1690.

25
Tables

Table 1. Average molecular weights (in g·mol-1) and polydispersity indices for

the polymers investigated.

Sample Mn ·10-3 Mw ·10-3 Mz ·10-3 Ip1a Ip2a

PDMS_H 380 630 890 1.7 1.4

PDMS_H75L25 159 496 1027 3.1 2.1

PDMS_H46L54 96 341 934 3.6 2.7

PDMS_H25L75 74 228 786 3.1 3.5

PDMS_L 59 94 140 1.6 1.5

PI_H 180 800 4800 4.4 6.0

PI_H90L10 76 724 3533 9.5 4.9

PI_H80L20 48.5 648 3510 13.4 5.4

PI_Lb 12.3 40 130 3.3 3.3

EP 175 350 700 2.0 2.0

PS 155 398 1040 2.6 2.6

PS_Hb 172 177 182 1.03 1.03

PS_H40L60 79 107 142 1.35 1.33

PS_H20L80 67 84 113 1.25 1.35

PS_H10L90 62 72 92 1.16 1.28

PS_L 58 60.4 63 1.04 1.04

a
Ip1 = Mw Mn and Ip2 = Mz M w

b
This polymer was not analysed with the rheological method presented in this

research, but was used to prepare the corresponding polymer blends.

26
Table 2. Values of parameter , for all the experimented polymer samples, at

which there was a fit between polydispersity indices calculated with the

proposed rheological method and experimentally-derived ones.

Measured Ip2a b c
Sample j i

PDMS_H 1.4 2.7

PDMS_H75L25 2.1 2.5

PDMS_H46L54 2.7 2.6 2.5

PDMS_H25L75 3.5 2.1

PDMS_L 1.5 3.1

PI_H 6.0 2.7

PI_H90L10 4.9 2.7 2.7

PI_H80L20 5.4 2.7

EP 2.0 2.5 2.5

PS 2.6 2.3 2.3

PS_H40L60 1.33 2.7

PS_H20L80 1.35 2.3

PS_L 1.04 2.8

a
Ip2 = Mz M w

b
The optimal value for each examined sample.
c
The optimal value for each family of polymers analysed.

27
Table 3. Measured and calculated polydispersity indices for all the experimental

polymer samples. The calculated polydispersity indices were obtained by taking

the optimal value of 2.5 for the parameter.

Measured Calculated
Sample
Ip2a Ip2a

PDMS_H 1.4 1.5

PDMS_H75L25 2.1 2.0

PDMS_H46L54 2.7 2.8

PDMS_H25L75 3.5 2.4

PDMS_L 1.5 1.6

PI_H 6.0 7.9

PI_H90L10 4.9 6.9

PI_H80L20 5.4 7.0

EP 2.0 2.0

PS 2.6 2.3

PS_H40L60 1.33 1.39

PS_H20L80 1.35 1.26

PS_H10L90 1.28 1.17

PS_L 1.04 1.05

a
Ip2 = Mz M w

28
Figure legends (or captions)

Figure 1. Dynamic moduli G' and G" master curves at a reference temperature T0

= 0 ºC for the PDMS mixtures series with 0% ( ), 25% ( ), 46% (+), 75% ( ) and

100% ( ) of the high molecular weight polymer (PDMS_H). The dotted line shows

the plateau modulus for polydimethylsiloxane.

Figure 2. Dynamic moduli G' and G" master curves at a reference temperature T0

= 110 ºC for the PI mixtures series with 100% ( ), 90% ( ), and 80% (+) of the

high molecular weight polymer (PI_H). The dotted line shows the plateau modulus

for polyisoprene.

Figure 3. Dynamic moduli G' and G" master curves at a reference temperature T0

= 170 ºC for the PS mixtures series with 0% ( ), 10% ( ), 20% (+) and 40% ( ) of

the high molecular weight polymer. The dotted line shows the plateau modulus for

polystyrene.

Figure 4. The dynamic modulus master curve data for EP [G' ( ) and G" ( )].

Figure 5. The dynamic modulus master curve data for PS [G' ( ) and G" ( )].

Figure 6. The relative error in Ip2 with dependence on the parameter for the five

kinds of polymers analysed: (+) and full line for PDMS, (x) and full line for PI, full

line for EP, dashed line for polydisperse PS and ( ) with full line for nearly

monodisperse PS. The vertical dashed line shows the chosen optimal value of

(= 2.5).

29
30
Figure 1

1E+06

1E+03

G' [Pa]
1E+00

1E-03
1E-04 1E-01 1E+02 1E+05
aT [rad/s]

1E+05

1E+03
G" [Pa]

1E+01

1E-01
1E-04 1E-01 1E+02 1E+05
aT [rad/s]

31
Figure 2

1E+07

1E+04

G' [Pa]
1E+01

1E-02
1E-07 1E-02 1E+03 1E+08
aT [rad/s]

1E+05

1E+03
G" [Pa]

1E+01

1E-01
1E-07 1E-02 1E+03 1E+08
aT [rad/s]

32
Figure 3

1E+06

1E+04

G' [Pa] 1E+02

1E+00
1E-02 1E-01 1E+00 1E+01 1E+02
aT [rad/s]

1E+05

1E+04
G" [Pa]

1E+03

1E+02
1E-02 1E-01 1E+00 1E+01 1E+02
aT [rad/s]

33
Figure 4

1E+08

G', G" [Pa]

1E+00

1E-04
1E-05 1E-01 1E+03 1E+07
aT [rad/s]

34
Figure 5

1E+07

G', G" [Pa]

1E+01

1E-02
1E-07 1E-03 1E+01 1E+05
aT [rad/s]

35
Figure 6

0,20

Relative error on Ip2

0,10

0,05

0,00
2,0 2,2 2,4 2,6 2,8 3,0 3,2 3,4

36