See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/264542209

Study on LLDPE molecular structure characterization by preparative and

analytical cross-fractionation

Article  in  E-Polymers · December 2010

DOI: 10.1515/epoly.2010.10.1.1596

CITATIONS READS

6 309

3 authors, including:

Age Larsen Jens Kjær Jørgensen

SINTEF, Oslo, Norway SINTEF
25 PUBLICATIONS   296 CITATIONS    15 PUBLICATIONS   239 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Funzionano View project

DAta assisted materials technology View project

All content following this page was uploaded by Jens Kjær Jørgensen on 15 September 2016.

The user has requested enhancement of the downloaded file.

 http://www.e-polymers.org
e-Polymers 2010, no. 143
ISSN 1618-7229

Study on LLDPE molecular structure characterization by

preparative and analytical cross-fractionation

Jens Kjær Jørgensen,1* Åge Larsen,1 Irene Helland 2

1*
SINTEF Materials and Chemistry, P.O.Box 124 Blindern, N-0314 Oslo, Norway,
+47-22067350 Jens.K.Jorgensen@sintef.no, Age.Larsen@sintef.no
2
Borealis AS N-3960 Stathelle, Norway. (Current address: Norner Innovation AS N-
3960 Stathelle, Norway), Irene. tel: +47-35578124; e-mail: Helland@norner.no.

(Received: 12 November, 2008; published: 11 December, 2010)

Abstract: Two grades of LLDPE were analysed by a number of methods for

fractionation and molecular structure determination (molecular weight, branching
density). Methods used were: preparative fractionation according to comonomer
content (TREF) and molecular weight (solvent – non solvent), CRYSTAF, SEC,
SEC-FTIR and NMR. We explore different ways of combining and processing data
to obtain two and three dimensional results. Methods to correct for chain end
effects in and construct bivariate distributions from preparative TREF + SEC +
NMR data is presented. The methods studied have different drawbacks and
advantages. Preparative TREF + SEC + NMR and SEC-FTIR provide most
information relative to the labour requirement. In general, the most complete
information is obtained from combined interpretation of results from different
methods.

Introduction
Polyolefins is and has for many years been the dominant material class on the plastic
materials market. Often it has been predicted that polyolefins would loose market
shares to new high performance plastics. However, this has never happened thanks
to continuous improvements in the performance of polyolefins. Since the constituents
of polyolefins are only hydrogen and carbon, differences between different
polyolefins are solely due to differences in chain architecture as described by
molecular weight distribution (MWD) and branching length, position and distribution.
These molecular structure characteristics are governed by the process and catalysts
used in the polyolefin synthesis. The development of better polyolefins is a result of
improvements in catalyst and synthesis technology, characterization methods,
understanding of rheology and processing and the interaction between these fields.
This paper reports results of an investigation of the performance of a number of
fractionation and analysis methods for characterization of molecular structure in
broad bimodal linear low density polyethylene. The methods used are listed in Tab. 1
and the workflow and combination of methods are illustrated in Fig. 1. In the end of
the paper we briefly illustrate the importance of detailed molecular structure
characterization by correlating results with mechanical properties and crystal
properties.
As briefly mentioned above, molecular structure is quantified by the MWD and the
branching characteristics. The MWD is defined as

1
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
 dw
MWD log M (1)
d log M

Tab. 1. Methods used for molecular structure characterization.

Method Purpose
Preperative TREF [1] CC fractionation
Preparative Holtrup Molecular weight fractionation (MW-
fractionation [2] fractionation) (solvent-non solvent )
NMR [3] Butene content determination of fraction
SEC Molecular weight distribution determination
Analytical CRYSTAF [1] CC fractionation of fractions
SEC-FTIR [4] CCM

where w and M are the cumulative weight fraction and molecular weight respectively.
Note that we will use M for the molecular weight as a parameter and MW as
abbreviation for molecular weight (see also list of abbreviations at the end of the
text). When concerning short chain branching in LLDPE one commonly refers to the
comonomer content distribution (CCD). Comonomer content (CC) is typically
measured in weight or mole fraction. The weight CCD is defined equivalent to MWD

dw
CCD CC (2)
dCC

Also, often results are reported as β = number of end groups pr 1000 carbon which is
almost equivalent. A detailed graphical way to present molecular structure data is as
three dimensional bivariate distributions of molecular weight and chemical
composition (i.e Fig. 12)

dw dw
f3D (log M,CC) (3)
d log M dCC

The bivariate distribution sums up to the MWD or the CCD by integration along the
CC or the logM axis respectively. Another useful type of result is the average
comonomer content as a function of molecular weight (CCM ) some times referred to
as “cross fractionation plot”. CCM is determined from a combination of MW-
fractionation and CC determination or opposite, for example using SEC-FTIR [4].
Nevertheless, though they provide detailed information these distributions and plots
do not provide information on the length of branches or segment statistics which will
require two additional distribution dimensions. This fact emphasizes the complexity of
describing polymer molecular structure.
Experimentally the most complete and accurate results are obtained by preparative
cross-fraction (preparative MW-fractionation followed by preparative CC fractionation
of each fraction or visa versa) followed by analytical molecular weight and CC
determination of all final fractions. This allows for determination of bivariate
distributions [5]. However, preparative cross-fractionation is extremely laborious and
time consuming. The alternative is to replace the second preparative fractionation
step with an analytical equivalent. Fig. 1 shows some different alternatives starting

2
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
with either preparative CC or molecular weight fractionation. Nevertheless, since the
preparative fractions always inherit some polydispersity the results as shown in this
paper and earlier by Aust et al. [5] depend on the analysis sequence.

Fig. 1. Schematic presentation of the experimental methods used, and how they
were combined to generate different types of results.

Preparative molecular weight fractionation of polyolefins is in most cases performed

by some solvent – non solvent method [2, 6]. Note that a consequence of the
thermodynamics governing the MW-fractionation using the solvent – non solvent
principle is, that with increasing MW it becomes increasingly more difficult to obtain
low polydispersity fractions [7].
Preparative comonomer content fractionation is preferably performed by temperature
rising elution fractionation (TREF) [1, 6]. TREF is based on the melting point
depression occurring when irregularities as comonomers are introduced in a linear
polymer. The melting temperature depression is experimentally found to depend
linearly on the end group density [1] which can be related to the comonomer content.
However, for low molecular weight chains, chain end groups contribute significantly
thereby introducing an error as illustrated and discussed later in this paper.

Results and discussion

The materials used, PE1 and PE2, are two pilot grade bimodal LLDPE butene
copolymers synthesized with the Borstar® process. The process results in molecular
weight bimodal materials and from the process conditions we know that the low MW
modes contain only a small amount of copolymer and the high MW mode a high
amount of copolymer. As seen from Tab. 2 and Fig. 2, PE1 and PE2 have very
similar MWD, density and total comonomer content. Hence, differences in their
properties must owe to differences in the distribution of co-monomers within the two
grades.
First step of our analyses is preparative fractionation. Fig. 3 and Fig. 4 show the
MWDs of the obtained fractions of PE1. Results for PE2 show the same features. As
a result of the high solubility of low MW species the fraction with lowest MW from the
MW-fractionation is large and broad compared to the remaining fractions which are
distributed evenly along the molecular weight axis.

3
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
Tab. 2. Properties of PE1 and PE2.

Mw Mw/ Wt% Dart Crystal lammela αT

(kg/mol) Mn (g/mL) butene drop (g) thickness (nm) (oC)
PE1 155 12 0.925 7.2 150 7-9 90
PE2 145 11 0.924 7.4 115 7-11 100

0.6
PE1
0.5 PE2

0.4
dw /dlogM

0.3

0.2

0.1

0
2 3 4 5 6 7
logM (g/mol)

Fig. 2. MWDs of PE1 and PE2.

0.6

0.5

0.4
dw /dlogM

0.3
1
34
0.2 2 5 6 7
8
0.1

0
2 3 4 5 6 7
logM (g/mol)

Fig. 3. MWD of PE1 and its fractions after MW-fractionation. Areas of MWDs
correspond to their sizes.

In fact all fractions are rather broad and overlap each other. When concerning CC
fractionation the MWDs are not a direct measure of the quality of the fractionation but
ads an additional dimension to the data. Note that the CC fractions with highest
comonomer content contain a low MW fraction which as shown later in this text has a

4
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
low comonomer content and hence distort the data. Fig. 5 shows the butene content
of all fractions as determined by NMR. Not surprisingly the overlapping of the MWDs
of the MW fraction “smears out” the variation in their butene content, while in
comparison the CC fractions shows a large variation in CC. Thus the CC fractions
contain more information of the CCD than the MW fractions.

0.6 1.4
frac 1, most butene
0.5 1.2 frac 4
frac 8, least butene
1
0.4
dw /dlogM

0.8
0.3
0.6
0.2
0.4

0.1 0.2

0 0
2 3 4 5 6 7 2 3 4 5 6 7
logM (g/mol) logM (g/mol)

Fig. 4. MWD of PE1 and its fractions after CC fractionation. Areas of MWDs in left
view correspond to fraction sizes. Right view shows normalized MWDs of selected
fractions.

0.35
Butene weight fraction from NMR

PE1 CC-fractionated
0.3
PE2 CC-fractionated
0.25 PE1 MW-fractionated
PE2 MW-fractionated
0.2

0.15

0.1

0.05

0
0 2 4 6 8
Fraction number

Fig. 5. Butene weight fractions of all fractions determined by NMR.

CCM (Comomoner content as a function of MW)

From the MWDs and the CC we can calculate estimates of CCM (Fig. 6) of the
materials by weighting the contribution from each fraction according to the formulas

5
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
 hi MWDi (logM )fw ,i (logM )
i
fw (logM ) ,
hi MWDi (logM )
i

where index i denotes each fraction (4)

dw fraction massi
fw : comonomer weight fraction , MWD , hi
dlogM fraction massk
k

Fig. 6 shows that CCM from the two fractionation approaches differs considerably
(ignore for the moment the corrected data), but both approaches show that PE1
contains more butene at high MW and less at low MW than PE2. The differences in
the two approaches are due to the polydispersity in both MW and CC of the fractions
and can only be reduced by obtaining more monodisperse fractions which is
challenging. This is also an indication of the precision of the methods. At low MW
CCM from MW-fractionation shows a low butene content opposite to the data from
CC fractionation that shows very high butene content. However, data from MW-
fractionation in the low MW region originates to a large extend from only one fraction
and are therefore reliable. In addition, from knowledge of the synthesis of the
materials we expect low butene content at low MW. It is therefore clear that the high
values obtained with CC fractionation are only a result of co-dissolution of highly
soluble low MW species together with highly branched high and moderate MW
species. As we argued from the values in Fig. 5, CCM values from MW-fractionation
are due to fraction MW-polydispersity highly “averaged” at most M values. Thus CC
fractionation provide more detailed results and if corrected for the low MW chain end
effect more correct results.

0.25
PE1 MW-frac.
PE1 CC-frac. uncorrected
0.2 PE1 CC-frac. corrected
CC, Weight fraction butene

PE2 MW-frac.
PE2 CC-frac. uncorrected
0.15
PE2 CC-frac. corrected

0.1

0.05

0
2.5 3 3.5 4 4.5 5 5.5 6 6.5 7
logM (g/mol)

Fig. 6. CCM obtained from preparative fractionation followed by SEC and NMR.

Correction of end groups

We have tested a simple procedure to correct for the contribution of end groups. Our
starting point is result from a study on CRYSTAF by Nieto et. al [8]. Fractionation
using CRYSTAF depends as for TREF, that we used for CC fractionation, on the
crystallization temperature. In the referred study it was found that end groups had the

6
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
same quantitative effect on crystallization temperature regardless whether they
belonged to a branch or a main chain end.

Fig. 7. Iterative scheme to calculate end group corrected co-monomer content as a

function of molecular weight from CC fractionation, SEC and NMR. The procedure is
performed independently on each fraction.

7
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
By adopting this conclusion we assume that our CC fractionation occurs according to
end group density β and that the variance of β within each fraction is smaller than the
variance of comonomer content. Thus, the fractions are more precise represented by
an average β value than average comonomer content.

PE2
0.35 Frac1, [87]

0.3

Annutation:
0.25 Fraction number,
[number of end groups pr 1000 C]

0.2
wt% butene

0.15
Frac2, [36]

0.1
Frac3, [22]

0.05 Frac4, [13]

Frac5, [9]
Frac6, [7]
Frac7, [4]

0
3 4 5 6 7 8
logM

Fig. 8. Molecular weight dependence of comonomer content of CC fractions of PE2,

calculated using the procedure in Fig. 7.

β and the comonomer weight fraction fw are related through the chain molecular
weight M and the comonomer molecular weight Mcom. With a few reasonable
assumptions (M >>2, only comonomers forming SCB are counted in fw) the relations
are simple.

28 14 fw 28 Mcom
1000 fw - (5)
M Mcom 1000 M 14

8
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
As a consequence, if β is nearly constant for all M, fw varies as a function of M. This
effect is significant at low M values but loses significance as M increases.
NMR measures the weight average comonomer mole fraction which is converted to a
weight average comonomer weight fraction <fw >. Assuming that the end group
density β is the same for all molecular weights in a given fraction we calculate fw as a
function of M under the constrain that <fw> equals the value determined by NMR.
This is performed according to Eq. 5 and the iterative scheme in Fig. 7. Fig. 8 shows
calculated molecular weight dependence of comonomer content of CC fractions for
PE2.
Fig. 6 shows the CCM from CC fractionation corrected for end group effects together
with the uncorrected and CCM from MW-fractionation. As expected the correction
reduces considerably the comonomer content values in the low MW region. In case
of PE1 the correction brings the low MW values close to the ones obtained with MW-
fractionation. For PE2 the values at low MW are reduced considerably but are still
higher than those found with MW-fractionation. Since the correction procedure do not
change the total amount of comonomer it results in an increase in CC for the
remaining MWs. It is hard to draw strong conclusions on the reliability of this
correction approach without knowing the true CCM. In theory it is sound but the
polydispersity in CC of the fractions introduce uncertainties. Also, small uncertainties
in the low molecular weight tails of MWDs originating from baseline determination in
SEC analysis can influence the number of chain end groups and thereby influence
the extent of correction due to chain ends.

SEC-FTIR
Another more facile way to determine CCM is to use SEC-FTIR [4]. In SEC-FTIR the
sample is MW-fractionated in the SEC columns and concentration and average
methyl content of each eluting fraction is measured with a FTIR spectroscope. From
these measurements and a SEC calibration CCM is found. The drawbacks of SEC-
FTIR compared to preparative fractionation followed by SEC and NMR is that it
measures methyl groups while NMR selectively measures short chain branches of
different length.

0.2
PE1 TREF uncorrected
PE1 TREF corrected
Weight fraction butene

0.15 PE1 SEC-FTIR uncorrected

PE1 SEC-FTIR corrected

0.1

0.05

0
3 4 5 6 7
logM (g/mol)

Fig. 9. Comparison of CCM from preparative TREF – SEC – NMR with CCM from
SEC-FTIR.

9
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
If all chain ends contain methyl groups this can be corrected with relations like Eq. 5,
else this requires additional knowledge of the chain ends (i.e. chain termination
mechanism). Also, the quality of SEC-FTIR data relies on high enough concentration
at the detector which compromises the SEC resolution.
Fig. 9 shows CCM of PE1 found with SEC-FTIR together with data from Fig.6 found
with preparative CC fractionation, SEC and NMR. Considering the complexity of the
two methods the two dataset matches very well. SEC-FTIR data shows more scatter
at the extremes and as expected for SEC-FTIR chain ends give a strong contribution
to the methyl content at low MW. Correction for the contribution of chain ends as
described by DesLauriers et al. [4] has an effect similar to the correction procedure
we introduced above for CCM from preparative CC fractionation. We believe that
since the SEC-FTIR CCM is based on narrow MW fractions it gives the most reliable
data as long as the concentration of the eluting fraction is high enough. For the
materials considered here that approximately includes data in the interval [3 < logM <
6] in Fig. 2. and Fig. 9. Contrary to this, preparative fractions extending the interval
[2.5 < logM < 7] (Fig. 4.) can be analysed at high concentrations and thereby provide
CCM data in a broader MW interval as shown in Fig. 9.

Bivariate distributions
We have constructed bivariate distributions using two approaches. Both approaches
are based on fractions from preparative CC fractionation analysed with SEC and
NMR. In the first approach CRYSTAF was in addition used to determine the CCD of
each fraction. We name data from this approach 3D CRYSTAF. In the second approach
CCD information was calculated from the fraction sizes and end group content. We
name data from this approach 3Dfast since it requires less analysis. Intuitively
3DCRYSTAF should have better resolution, but as is shown in the following, limitation of
CRYSTAF makes 3Dfast better at some points.
CRYSTAF profiles of PE2 and its fractions are shown in Fig. 10. The rectangular low
temperature parts of the profiles represent a “soluble fraction” of species that did not
precipitate above 30 ˚C. PE2 has a large “soluble fraction”. In fact all of fraction 1 is
“soluble”. Theoretically the weighted sum of the fraction CRYSTAF profiles should
equal the profile of the parent material but as seen from Fig. 11 this is not the case.
The fraction profiles appear to be too narrow to describe the parent material profile
resulting in wavy weighted sum profile. We consider co-crystallization during the
cooling step a possible explanation for this discrepancy.
A well fitting linear CRYSTAF calibration curve (T → β) were obtained by correlating
β values of the fractions and CRYSTAF profiles peak temperatures (β values as
determined by NMR and the end group corrections procedure). Bivariate distributions
of each fraction were calculated by assigning the determined β distribution to each
molecular weight in the MWD (Eq. 3). The bivariate distribution for the whole material
was obtained by weighted summation of the fractions bivariate distributions. A
requirement for this approach is that the β distributions are sufficiently narrow so that
the assumption of equal β distributions for each M is not too strong. The CRYSTAF
profiles (Fig. 10) that we convert to β distributions are not very narrow. On the other
hand since they are monomodal and quite symmetric the sum of the overlapping
parts of the profiles probably describe an imaginary intermediate fraction well. Fig. 12
shows the obtained bivariate distributions of PE1 and PE2. As could be foreseen
they inherit a “wavy” shape due to inconsistency between the sum of fraction

10
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
CRYSTAF profiles and parent material CRYSTAF profile, and also a considerable
“soluble” fractions. Nevertheless, they still reveal important differences that we
discuss later in this text.
The 3Dfast approach utilizes that when performing preparative TREF a rough CCD
can be calculated. First step is to calculate a cumulative CCD as shown in Fig. 13.
The cumulative CCD is constructed by plotting wi versus βi , where wi is given by
i 1
h1 hi
w1 , wi hn , i 2,3,..... (6)
2 2 n 1

Here h is weight fractions of the fractions.The CCD (Fig. 13 b right view) is obtain by
differentiation of the cumulative CCD.

0.3
frac1
0.25 frac2+3
frac4 7
0.2 1 frac5 6
5
frac6
dw /dT

0.15 frac7
4
frac8
0.1 3
Parent

0.05 2

0
20 30 40 50 60 70 80 90
o
T ( C)

Fig. 10. CRYSTAF profiles of PE2 and its fractions from preparative TREF.

0.05 PE2 parent

Sum CC fractions
0.04

0.03
dw /dT

0.02

0.01

20 40 60 80 100
o
T ( C)

Fig. 11. Comparison of the CRYSTAF profile of PE2 and the weighed sum of
CRYSTAF profiles of its fractions (Fig.10).

11
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
 PE1 PE2

0.04 0.04
dw/dlogM*dw/d

dw/dlogM*dw/d
0.03 0.03

0.02 0.02

0.01 0.01
8 8
0 6 0 6
60 60
40 4 40 4
20 20
0 0
-20 2 logM (g/mol) -20 2 logM (g/mol)
(methyl/1000 C) (methyl/1000 C)

Fig. 12. Bivariate distributions of PE1 and PE2 obtained from preparative CC
fractionation, and analytical MWD and CCD determination (SEC and CRYSTAF).
1 0.06

0.8
cum. weigth frac., w

0.04
0.6
dw/d

Differentiate
0.4
0.02
0.2

0 0
0 20 40 60 0 20 40 60
(methyl/1000 C) (methyl/1000 C)

Fig. 13. CCD from PREP-TREF: Left view: Cumulative CCD obtained using Eq.6.
Right view: Differential CCD distribution obtained by derivation of left view data.

PE1 PE2

0.04 0.04
dw/dlogM*dw/d

dw/dlogM*dw/d

0.03 0.03

0.02 0.02

0.01 0.01
8 8
0 0
6 6
60 80
40 4 60 4
20 40
2 20 2
0 logM (g/mol) (methyl/1000 C) 0 logM (g/mol)
(methyl/1000 C)

Fig. 14. 3Dfast plot: Bivariate distributions obtained from PREP-TREF, NMR and SEC.

The 3Dfast bivariate plots (Fig. 14) are constructed by assigning the corresponding
MWD to each point on the CCD. The MWD for intermediary values on the β axis are
found by linear interpolation. The assumptions of this approach are that each fraction
is well described by one average β value and a MWD. Compared to this the

12
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
3DCRYSTAF approach includes a CCD of each fraction. Nevertheless, if these
CRYSTAF based CCDs are symmetric peaks, as in this case, their weighed
averages will not differ much from interpolation between discrete values as in the
3Dfast approach. An advantage of 3Dfast is that it includes fractions with high end
group concentration (including branches) that appear as “soluble fractions” in
CRYSTAF and 3Dfast is not bulky like 3DCRYSTAF. Comparison of Fig. 12 and Fig. 14
shows that the two approaches provide similar results but 3Dfast is smoother and
covers a larger end group content range. Thus, in conclusion, in this investigation we
find 3Dfast to be a better approach despite its simplicity.

Structure property relations

PE1 and PE2 have similar average branch content (Tab. 2) but comparison of their
bivariate distributions shows that the branches are differently distributed. The
bivariate distribution of PE2 has significant fractions on a larger range of the β axis,
signifying that co-monomers are more evenly distributed in PE1 than in PE2. This
together with the observation that PE1 has a higher average co-monomer content in
the high MW tail (Fig.6.) is probably the most important differences between the two
materials.
Tab. 2 list results from impact measurement (drop dart) and crystallinity related
experimental observations using transmission electron microscopy (TEM) and
dynamic mechanical thermal analysis (DMTA). PE1 has higher dart drop impact
strength than PE2. This observation is in accordance with the well known fact that an
appropriate amount of short chain branches in the highest molecular weight fraction
makes polyethylene less brittle. Investigation of morphology using TEM reveals that
in both materials most crystal lamellae have thickness in the range 7.8-8.5 nm, but
PE2 contains some individual lamellae with thickness 10.5-11 nm. The observation of
thicker lamellae in PE2 is in accordance with a higher α transition temperature in PE2
than in PE1 (PE1:90 °C, PE2:100 °C) [9]. We correlate the high lamella thickness
observed for PE2 to the low and broadly distributed co-monomer content in its high
molecular weight fraction that likely contains chains with long crystallizable
segments.

Conclusions
Our investigation has shown that molecular structure distributions of the studied
LLDPE can be found from both preparative MW fractionation (solvent-non solvent)
and preparative CC fractionation (TREF) in combination with analytical methods.
However, both preparative methods inherit limitations and generate errors under
certain conditions. Chain end effects influence CC fractionation and fraction MWD
overlap limits the resolution of results from MW fractionation. A combined
interpretation provides most reliable information. A method to correct for chain end
effects in analysis with prep-TREF combined with NMR and SEC was tested with
good results. Corrected CCM-data (co-monomer content versus molecular weight)
was in accordance with CCM-data from SEC-FTIR.
Bivariate distributions were obtained by two approaches: 3DCRYSTAF using CRYSTAF
profiles to predict the CC component and 3Dfast using fraction sizes and average end
group content to predict the CC component. Despite its simplicity 3Dfast gave the best
results. The problems we encounter with CRYSTAF could possible be eliminated by
using the more precise but also more time consuming analytical TREF method. If

13
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
lower accuracy is acceptable the 3Dfast approach can be speeded up further by
performing CC analysis using FTIR instead of NMR.
Our results shows, as observed in other studies [1, 5, 10] that cross fractionation
disclose property relevant details, that are not disclosed by either CC or MW
fractionation alone. We reveal critical differences in distribution of comonomer at high
MW that correlates with differences in impact strength and crystal structure. This
information is valuable in material development and is not disclosed by one
dimensional fractionation analyses.
Many of the experimental procedures used in this investigation are labour intensive.
The results of this work suggest SEC-FTIR and the 3Dfast approach as the methods
delivering most valuable information compared to the required labour.

Experimental part

Prep-TREF
Preparative TREF were performed using a Polymer Char Prep MC2+ automated
fractionation instrument with xylene as solvent. The temperature profile for
dissolution, stabilization, crystallization and fractionation are listed in Tab. 3.

Tab. 3. Temperature profile used in preparative TREF.

Dis Stab Crys Fr.1 Fr.2 Fr.3 Fr.4 Fr.5 Fr.6 Fr.7 Fr.8
Rate (ºC/min) 20.0 20.0 0.2 20.0 20.0 20.0 20.0 20.0 20.0 20.0 20.0
Temperature
(ºC) 130 95 30 30 48 64 72 80 84 88 130
Time (min) 60 45 15 30 30 30 30 30 30 30 30

Prep-MW
Preparative MW fractionation was performed on a Polymer Char Prep MC2+
automated fractionation instrument using a coacervate extraction process based on
solvent non-solvent partitioning [2, 6]. Xylene was used as solvent and DEGMBE
(diethylene glycol monobuthyl ether equivalent to (2,2-butoxationyethoxy)ethanol) as
non-solvent. The fractionation was performed at 130 C with the non-solvent
percentages listed in Tab. 4.

Tab. 4. Non-solvent percentage used in molecular weight fractionation.

Fr1 Fr2 Fr3 Fr4 Fr5 Fr6 Fr7 Fr8

Non-solv. % in
mixture 60.0 55.0 50.0 46.5 43.0 39.0 35.0 0.0

SEC
SEC for MWD determination was performed with a procedure based on ISO 16014-
4:2003. A Waters 150CV plus instrument, equipped with refractive index detector and
online viscosimeter was used with 3 x HT6E styragel columns from Waters (styrene-
divinylbenzene) and 1,2,4-trichlorobenzene (TCB, stabilized with 250 mg/L 2,6-Di tert
butyl-4-methyl-phenol) as solvent at 140 °C and at a constant flow rate of 1 mL/min.
500 μL of sample solution were injected per analysis. The column set was calibrated

14
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
using universal calibration with 15 narrow MWD polystyrene (PS) standards in the
range of 1 kg/mol to 12 000 kg/mol. Mark Houwink constants were used for
polystyrene and polyethylene (K: 9.54 x10-5 dL/g and a: 0.725 for PS, and K: 3.92
x10-4 dL/g and a: 0.725 for PE). All samples were prepared by dissolving 0.5 - 3.5
mg of polymer in 4 mL (at 140 °C) of stabilized TCB (same as mobile phase) and
keeping for 3 hours at 140 °C and for another 1 hours at 160 °C with occasional
shaking prior sampling into the SEC instrument.

CRYSTAF
CRYSTAF was performed on a Polymer Char CRYSTAF work station. 30 mL of
0.1%w/v sample in TCB was introduced into the crystallization vessel. This was
followed by 45 min dissolution step at 160 oC and a 45 min stabilization step at 95 oC.
The analysis was performed by cooling from 95 oC to 30 oC at 0.1 oC/min. CRYSTAF
curves were obtained by continues monitoring of concentration using IR detector.

NMR
Approximately 90-95 mg of each sample was accurately weighed into separate NMR
tubes. 0.7 ml of ortho-di-chloro-benzene (ODCB) was added to each tube. Nitrogen
was bubbled through the solvent and the NMR tubes were sealed by melting. The
tubes were placed in an oven for two days at 130C. During that period the tubes
were inverted several times to ensure a homogeneous solution. NMR was performed
at 130 0C. For fractions 1 to 3, 5000 transients (16 h) were sampled. For fractions 4
to 8 18000 transients were needed (58 h) to improve the signal to noise ratio. Details
of how the NMR data was processed have been described previously by Hansen et
al [3].

SEC-FTIR
SEC-FTIR were performed on a PL-GPC 220 equipped with 2 x PLgel Olexis (300 x
7.5mm) coupled to a PE Spectrum One with MCT detector by the PL-HTGPC/FTIR
interface. The system was calibrated using Polystyrene EasiCal narrow standards
and polyethylene short chain branching standards. Analyses were performed with
500 μl samples nominally 2.0 mg/ml, flow rate 1mL/min at 160 °. Data was collected
over 3000 to 2700 cm-1 at 8 cm-1 resolution with 16 scan accumulations over the 22
min run time. Chromatograms were then generated from the time-resolved spectra
from the root mean square intensity plots, which were read directly into PL Cirrus
GPC-FTIR SCB software.

Density measurements
Densities were measured in a density column after ASTM D 1505.

Dart drop
Dart drop tests were performed according to ISO 7765-1.

DMTA
DMTA measurements were performed with a PerkinElmer DMA-7e, from
compression-moulded plaques. The analyses were run at 2,5 °C/min under a helium
atmosphere in the temperature scan mode from -60 °C up 100 °C at a frequency of 1

15
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
Hz. The storage modulus (E´), the loss modulus (E´´), and the ratio of E´´ to E´ (tan
δ) were recorded.

TEM
To improve electron density contrast between crystalline and amorphous phases
samples for TEM were treated with chlorsulphonic acid (increases electron density in
amorphous phase). Thin electron- transparent foils for TEM were obtained by an
ultramicrotome. TEM samples were examined in a Phillips CM30 TEM operated at
100 keV. The thickness of crystal lamellae was evaluated directly on the negatives
using an ocular and a light table.

Acknowledgements
This work was financially supported by the Norwegian Research Council. The
authors thank Dr. Christian Piel and Dr. Siw Bodil Fredriksen for discussions of the
results and Polymer Laboratories for providing SEC-FTIR analysis.

List of abbreviation and acronyms

LLDPE Linear low density polytethylene
TREF Temperature rising elution fractionation
CRYSTAF Crystallization analysis fractionation
SEC Size exclusion chromatography
SEC-FTIR SEC coupled to FTIR via a flowcell
NMR Nuclear magnetic resonance spectroscopy
CC Comonomer content
CCD Comonomer content distribution
CCM Comonomer content as a function of molecular weight
SCB Short chain branching
TEM Transmission electron microscopy
MW Molecular weight (concept)
MWD Molecular weight distribution
M Molecular weight (parameter)
β Number of end groups pr 1000 carbon
w Cumulative weight fraction
Mw Weight average molecular weight
Mcom Comonomer molecular weight
fw Comonomer weight fraction
3D Bivariate distribution

References
[1] Anantawaraskul, S.; Soares, J.B.P.; Wood-Adams, P. Adv Pol Sci. 2005, 182, 1
[2] Holtrup, W. MakrChem. 1977,178, 2335.
[3] Hansen, E.W.; Redford, K.; Oysaed, H. Polymer. 1996, 37, 19.
[4] DesLauriers, P.J.; Rohlfing, D.C.; Hsieh, E.T. Polymer. 2002, 43, 159.
[5] Aust, N.; Beytollahi-Amtman, I.; Lederer, K. Int J Pol Anal Char. 1995, 1, 245.

16
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
 [6] Soares, J.B.P. “Fraction” In:. Encyc Pol Sci Tech 3. ed. Mark, H.F. editor,
Hoboken: John wiley and Sons; 2004.
[7] Koningsveld, R.; Kleintjens L.A.; Geerissen, H.; Schuetzeichel, P.; Wolf, B.A.
Fractination. In: Comprehensive polymer science. Oxford: Pergamon Press; 1989.
Editor: Allen,G.
[8] Nieto, J.; Oswald, T.; Blanco, F.; Soares, J.B.P.; Monrabal, B. J Pol Sci B Pol
Phys. 2001, 39, 1616.
[9] Boyd, R.H. Polymer. 1985, 26, 323.
[10] Faldi, A.; Soares, J.B.P. Polymer, 2001, 42, 3057.

17
Unauthenticated | 78.91.103.24
Download Date | 4/29/14 6:53 AM
View publication stats