Available via license: CC BY-NC
Content may be subject to copyright.
RESEARCH ARTICLE
www.mre-journal.de
Influence of Poisons Originating from Chemically Recycled
Plastic Waste on the Performance of Ziegler–Natta Catalysts
Daniel Christian Pernusch,* Gunnar Spiegel, Christian Paulik, and Wolfgang Hofer
Chemically recycled monomer feed streams from postconsumer plastic
packaging waste are expected to contain impurities that act as catalyst
poisons. This work investigates the impact of such impurities on a
commercial fourth-generation Ziegler–Natta catalyst system. A recycled
monomer feed stream in two different procedures by adding representative
catalyst poisons (pyridine and n-butanol) in various quantities to the reactor
setup is simulated. Measuring catalyst activity and molecular weight
distribution (MWD) and performing kinetic and MWD deconvolution
modeling, the impact of the catalyst poisons on polymer product properties at
the microstructural level is evaluated. The results demonstrate that, beyond a
certain concentration (120 ppm), catalyst poisons have a substantial impact
on the activity levels of the catalyst system tested. MWD deconvolution
modeling shows an influence of the poisons on the composition of the
resulting polymer product in the form of a shift toward lower or higher
molecular weights (depending on the procedure). This microstructural
analysis highlights the importance of purifying chemically recycled monomer
feed streams.
1. Introduction
Moving toward a circular economy is of increasing importance to
industry. In the petrochemical sector, postconsumer materials—
such as plastic packaging waste—are of particular interest. While
the destination of most collected plastic packaging is land-
fill (18.5%) or energy recovery (39.5%), an increasing num-
ber of technologies are being investigated to improve recycling
(currently 42%) and chemically upgrade postconsumer plastic
D. C. Pernusch, G. Spiegel, C. Paulik
Institute for Chemical Technology of Organic Materials
Johannes Kepler University Linz
Altenbergerstraße 69, 4040, Linz Austria
E-mail: christian.paulik@jku.at
W. Hofer
OMV Refining & Marketing GmbH
Mannswörther Str. 28, 2320, Schwechat Austria
The ORCID identification number(s) for the author(s) of this article
can be found under https://doi.org/10.1002/mren.202100020
© 2021 The Authors. Macromolecular Reaction Engineering published
by Wiley-VCH GmbH. This is an open access article under the terms of
the Creative Commons Attribution-NonCommercial License, which
permits use, distribution and reproduction in any medium, provided the
original work is properly cited and is not used for commercial purposes.
DOI: 10.1002/mren.202100020
packaging waste.[1] Until 2018, plastic
packaging waste was recycled mainly
mechanically. However, due to particular
difficulties with this method, such as inten-
sive pretreatment requirements, chemical
recycling has become a magnet for new
technologies.[2] In contrast to mechanical
recycling, where the chemical structure
is maintained, chemical recycling depoly-
merizes plastic packaging waste to obtain
smaller hydrocarbons. This is often done
in multiple steps that include catalytically
or thermally driven processes.[3] Depoly-
merization of polymers into mixtures of
oligomers and monomers provides a feed
stream either for fuel components or for
further processing by steam cracking.
Polyolefin waste is particularly suitable for
recycling, as the oil fractions obtained are
rich in linear paraffin and alpha olefins,
resulting in compounds with high cetane
and octane numbers.[3] The advantage of
chemical recycling is that, compared to
mechanical recycling, pretreatment is less
problematic. However, not all types of packaging plastics are suit-
able for all recycling technologies, and the composition of the
plastic waste has a significant influence on product outcome. For
example, Angyal et al. showed that the degradation of polypropy-
lene from plastic waste is affected by variations in the polystyrene
content.[4]
Chemical recycling of plastic waste requires focusing on the
products obtained by steam cracking of recycled oil fractions:
Depending on the type of plastic waste, chemical recycling can
result in a variety of product compositions. In the context of
repolymerization, impurities that can potentially have a nega-
tive effect on the catalyst system (i.e., catalyst poisons) are of
particular interest. Olefin polymerization typically uses fourth-
generation Ziegler–Natta catalysts, which are very susceptible to
deactivation by impurities.[5] Since both the catalyst itself (usually
a TiCl4/MgCl2species) and the cocatalyst (an aluminum alkyl,
most commonly triethylaluminum, TEA) are sensitive to humid-
ity, air, and electronegative heteroatoms, the concentration of im-
purities in the monomer feed stream must be low. Hence, it is of
interest how impurities originating from chemical recycling of
plastic waste affect the performance of Ziegler–Natta catalysts in
olefin polymerization and consequently the properties of the re-
sulting products.
Catalyst poisoning is a type of chemical deactivation and is
caused by strong chemisorption of the poisoning compound onto
Macromol. React. Eng. 2021, 2100020 2100020 (1 of 10) © 2021 The Authors. Macromolecular Reaction Engineering published by Wiley-VCH GmbH
www.advancedsciencenews.com www.mre-journal.de
the active site of the catalyst, which is thus rendered dormant.
Whether a given compound in the reaction mixture acts as a
poison can be determined by the strength of its adsorption to
the catalyst relative to that to the other compounds present. This
has operational implications. While adsorption to the active site
is merely a physical way of deactivating a catalyst, the adsorbed
poison can also change the electronic and geometric properties
of the catalyst surface. Poisoning can be either reversible or ir-
reversible. Most known polyolefin catalyst poisons tend to react
irreversibly with the catalyst components. The effect of the poi-
son on catalyst activity depends on whether the poison is “selec-
tive,” “antiselective,” or “nonselective.” A selective poison pref-
erentially adsorbs to the most active sites of the catalyst at low
concentrations. If the poison preferentially adsorbs to less active
sites, it is antiselective, and if the activity loss is proportional to
the poison concentration, the poison is nonselective. Further, the
poison’s influence on catalyst activity depends on the preferred
place of reaction and/or poisoning (i.e., outer shell).[6–8 ]
Compared to regular crude oil, a chemically recycled crude oil
contains various kinds of impurities, more specifically, higher
quantities of nitrogen- and oxygen-containing compounds and
lower quantities of sulfur-containing compounds. In this study,
selected catalyst poisons that are expected to occur in chem-
ically recycled monomer feed streams were investigated in a
one-stage homopolymerization of ethylene using a commercial
fourth-generation Ziegler–Natta catalyst.
2. Experimental Section
2.1. Chemicals
For the polymerization, the following gases and liquids were
used: nitrogen (5.0, Linde), hydrogen (5.0, Linde), propane (3.5,
Linde), ethylene (3.0, Linde), and n-heptane (97%, Roth). Prior to
introduction to the reactor system, all components were purified
over a purification unit containing oxidizing/reducing catalysts
and using molecular sieves.[9,10 ] A commercial fourth-generation
Ziegler–Natta catalyst was used with TEA (98%) as cocatalyst. Cat-
alyst and cocatalyst were stored and prepared in a glovebox under
inert conditions (nitrogen atmosphere, <1 ppm O2and <1 ppm
H2O).
Pyridine (absolute over molecular sieve, Sigma-Aldrich) and n-
butanol (99.5%, Sigma-Aldrich) were used as model catalyst poi-
sons because a study by Toraman et al. found higher quantities
of nitrogen- and oxygen-containing compounds in steam-cracked
pyrolysis oil from recycled plastic waste.[11]
2.2. Size-Exclusion Chromatography
The polymer samples produced were analyzed by high-
temperature size-exclusion chromatography (HT-SEC) to deter-
mine the following molecular properties: molecular weight dis-
tribution, average molar mass Mw, number average molecular
weight Mn, and dispersity index Ð. The HT-SEC device was
equipped with an IR5 detector from PolymerChar with three
PLgel Olexis mixed-bed columns (7 300 mm) from Agilent.
The samples were dissolved at 160 °C for 120 min in 1,2,4-
trichlorobenzene (TCB).
Tabl e 1 . Experimental design.
No. Poison type Poison
concentration
[ppm]
Procedure
1— 0 A
2 Pyridine 120 A
3 Pyridine 60 A
4— 0 B
5 Pyridine 120 B
6 Pyridine 60 B
7— 0 A
8n-Butanol 180 A
9n-Butanol 120 A
10 n-Butanol 60 A
11 — 0 B
12 n-Butanol 180 B
13 n-Butanol 120 B
14 n-Butanol 60 B
2.3. Procedure
Before each experiment, a purification step consisting of flush-
ing with nitrogen and applying vacuum multiple times was car-
ried out. The reactor was then filled with liquid propane (≈80 g)
and heated to 100 °C for 30 min. Subsequently, the reactor was
flushed, and the temperature set to 30 °C. For the experiment,
propane (165 g) was filled into the reactor. Hydrogen (50 mg)
and ethylene (8 g) were fed into the reactor using mass-flow con-
trollers. Poison (0–180 ppm, calculated based on the total ethy-
lene uptake during a standard reference experiment) was injected
with heptane via a pneumatic injection system. The cocatalyst
(Al/Ti (mol/mol) =100) was injected either after (procedure A)
or prior to (procedure B) introduction of the poison. In proce-
dure A, the cocatalyst was injected together with the catalyst,
while in procedure B cocatalyst and catalyst were injected sep-
arately. Catalyst and cocatalyst were injected into the reactor by
means of a manual injection system. After injection of the cata-
lyst into the reactor, the system was heated to the polymerization
temperature of 70 °C. Upon reaching this temperature, a poly-
merization period of 45 min at constant reactor pressure began.
The ethylene uptake was used to track the activity rate of the ex-
periment.
2.4. Experimental Design
To prevent left-over poison from influencing the experiments
over the course of the runs, the experiments were conducted from
high to low poison concentration. Additionally, reference experi-
ments were conducted in between the runs to check reactor per-
formance. Table 1 lists the experiments and their parameters.
2.5. Kinetic Modeling
To investigate the kinetic behavior under the influence of cata-
lyst poison, the activity rates obtained from the polymerization
Macromol. React. Eng. 2021, 2100020 2100020 (2 of 10) © 2021 The Authors. Macromolecular Reaction Engineering published by Wiley-VCH GmbH
www.advancedsciencenews.com www.mre-journal.de
0 5 10 15 20 25 30 35 40 45
0
0.1
0.2
0.3
0.4
0.5
0.6
Ethylene Uptake / mmol s
-1
Time / min
Experimental Data
Model Graph
k
d
Figure 1. Ethylene uptake representing the activity rate of the catalyst, and model graph calculated based on the approach of Alshaiban.[16]
data were modeled. The catalyst activity rate was derived from the
ethylene uptake in mol s−1. Modeling was based on the elemen-
tary reactions of activation, propagation, and deactivation[12–15 ]
Activation : C +Al ka
→P0(1)
Propagation : Pr+M
kp
→Pr+1(2)
Deactivation : Pr
kd
→Cd+Dr(3)
Here, Cis the catalyst precursor, Al is the aluminum alkyl (co-
catalyst), kais the activation rate constant in s−1,P0is the acti-
vated monomer-free catalyst site, Mis the monomer, Pris the
living chain with length r,kpis the propagation rate constant in
Lmol
−1s−1,Cdis the deactivated catalyst site, Dris the dead poly-
mer chain with length r,andkdis the first-order deactivation rate
constant in s−1.
Using Equations (1), (2), and (3) allows the rate of polymeriza-
tion to be described by
Rp=kp[M][C]01−e−KAt1−kd∕
KAe−kdt
1-kd∕
KA
(4)
where [M] is the monomer concentration in mol L−1,[C]0is the
initial catalyst precursor concentration in mol L−1,andKAis the
modified activation rate constant, which is expressed by
KA=ka[Al](5)
where [Al] is the cocatalyst concentration in mol L−1. Since the
cocatalyst is injected in large excess, this term is assumed to be
constant. The influence of catalyst poisons on the catalyst system
results in the basic model not being suitable for this setup. To cir-
cumvent this, the model was adjusted to include a poison factor
fp, which affects the initial precursor catalyst concentration [C]0.
By doing so, it is possible to include both interactions of interest
(procedure A and procedure B) in the model, which both effec-
tively result in a lower concentration of initial precursor catalyst
concentration. Based on the proposed mechanisms shown later
in Sections 3.1 as well as 3.2 the poison factor was varied for the
prevalent poison type and concentration. The term [C]0was re-
placed by the term
[C]0,p=fp[C]0(6)
The final model can then be described by
Rp=kp[M][C]0,p1−e−KAt(1−kd
∕
KA)e−kdt
1−kd∕
KA
(7)
Following the approach of Alshaiban and expressing the ki-
netic parameters for activation, propagation, and deactivation al-
lows the behavior of the catalyst to be modeled.[16] Since in cat-
alyst poisoning the parameter of interest is the rate constant of
deactivation, this study focused on deactivation. Figure 1 presents
example experimental activity rates, measured in the form of
ethylene uptake, and the corresponding graph obtained by mod-
eling the experimental data to determine the deactivation rate.
2.6. MWD Deconvolution
Ziegler–Natta catalysts are known to have broad molecular
weight distributions (MWDs), which can be explained by their
so-called multisite nature:[17] The catalyst has multiple active sites
(typically between four and six) which are responsible collectively
for the total amount of polymer produced. With analytic meth-
ods such as HT-SEC, the MWD of the whole polymer can be
Macromol. React. Eng. 2021, 2100020 2100020 (3 of 10) © 2021 The Authors. Macromolecular Reaction Engineering published by Wiley-VCH GmbH
www.advancedsciencenews.com www.mre-journal.de
33.544.555.566.57
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
dwt/dlogM
logM/gmol
-1
AS 1
AS 2
AS 3 AS 4
AS 5
AS 6
SEC Data
Figure 2. MWD deconvolution for a given homo-PE sample with six deconvoluted active sites.
measured, but not the proportion of each active site individually.
MWD deconvolution is a method for determining the minimum
number of active sites required to accurately represent a given
MWD obtained from SEC analysis. The MWD of a given active
site can be described by the Flory–Schultz distribution[18,19 ]
WlogMW =2.3026
n
j=1
wjMW
Mn,j2
exp −MW
Mn,j
(8)
under the constraint that
n
j=1wj=1(9)
where wjis the weight fraction of polymer made at site type j,n
is the total number of site types, MW is the polymer molecular
weight, and
Mn,jis the number-average molecular weight of poly-
mer made at site j. For a given homopolyethylene, the deconvolu-
tion model estimates the reactivity ratios of each individual active
site. The MWD measured by SEC is deconvoluted into most prob-
able Flory–Schultz distributions to determine the number of site
types and the mass fractions of polymer produced by them. The
parameters for the deconvoluted active sites (i.e., number of site
types, mass fraction, and Mn) are subsequently optimized to min-
imize the standard deviation between experimental SEC data and
model data obtained by the sum of all deconvoluted active sites.
Standard deviation minimization of the experimental data and
calculated model was performed by varying the three starting pa-
rameters: number of site types, mass fraction at each site type,
and Mnof each site type. Deconvolution was performed in GNU
Octave using the Levenberg–Marquardt minimization method,
which solves the nonlinear least-squares to find a minimum. An
example graph of a deconvoluted MWD of a PE sample is shown
in Figure 2.
After successful deconvolution and minimization of the stan-
dard deviation, Mnand Mwcan be obtained for each individual
active site by
Mn=1
n
j=1
wj
Mn,j
(10)
MW=
n
j=1
Wj
Mw,j(11)
The benefit of MWD deconvolution is that the otherwise prob-
lematic complex multisite parameters can be reduced to several
simplified single-site type parameters. As described above, the
method determines the minimum number of active sites required
to represent the experimental data. It is possible that additional
active sites are present that overlap with other Flory–Schultz de-
convoluted sites because they produce polymers with similar
MWD (peak superposition). However, MWD deconvolution fol-
lows an approach of simplicity and does not require additional as-
sumptions for the individual active sites, which makes it a strong
tool for the analysis of experimental MWD distributions.[20]
3. Results and Discussion
3.1. Pyridine Poisoning
Pyridine is expected to bind to the aluminum atom in the cocat-
alyst to form an Al-N complex.[21] This decreases the reducing
power and thus the activation power of the cocatalyst, and con-
sequently the activity of the catalyst is reduced. In this study, the
effects of pyridine concentration on catalyst activity (and deacti-
vation behavior) and MWD of the product (SEC and MWD de-
convolution) were investigated.
Macromol. React. Eng. 2021, 2100020 2100020 (4 of 10) © 2021 The Authors. Macromolecular Reaction Engineering published by Wiley-VCH GmbH
www.advancedsciencenews.com www.mre-journal.de
Tabl e 2 . Activity rates observed at various pyridine concentrations.
No. Poison
type
Poison
concentration
[ppm]
Procedure Activity rate
[kgPE gcat−1h−1]
Activity
rate [%]
1 — 0 A 13.1 100
3 Pyridine 60 A 12.3 93.9
2 Pyridine 120 A 1.2 9.2
4 — 0 B 21.1 100
6 Pyridine 60 B 20.7 98.1
5 Pyridine 120 B 8.7 41.2
Tabl e 3 . Deactivation constants calculated using the fitted model graph
based on Equation (7).
No. Poison
type
Poison
concentration
[ppm]
Procedure Poison
factor
fp
kd
[mol s−1]
1— 0 A 1 0
3 Pyridine 60 A 0.94 0
2 Pyridine 120 A 0.89 1.3 ×10−4
4— 0 B 1 0
6 Pyridine 60 B 0.94 0
5 Pyridine 120 B 0.89 4.2 ×10−4
3.1.1. Influence of Pyridine on the Activity Rate
Varying the poison concentration allowed the impact on cata-
lyst activity to be determined, as summarized in Table 2:Atlow
poison concentrations, catalyst activity remained largely unaf-
fected. Apparently, the remaining cocatalyst concentration was
sufficiently high to activate the catalyst despite the presence of
a catalyst poison. However, at 120 ppm poison concentration, a
significant reduction in activity was observed, amounting to ab-
solute losses in activity of around 11.9 and 2.4 kgPE gcat−1h−1
compared to 0 ppm poison concentration in procedures A and B,
respectively. While the absolute impacts of a higher poison con-
centration on the total activity of the catalyst were similar in both
procedures, the relative impacts differed. At 120 ppm poison con-
centration, the activity reached 9.2% in procedure A and 41.2%
in procedure B. This significant difference is explained by the
procedures affecting the general activity rates: the activity level in
procedure B is generally higher (also in the absence of a poison).
3.1.2. Deactivation Behavior (Pyridine)
Ethylene uptake was used to monitor the activity profile of the
catalyst during the experiments, and the data were then fitted
by kinetic modeling using Equation (7) to obtain model graphs.
The introduced poison factor fpfor the model was set to 1 for
the reference experiment (no poison). As the poison can be ex-
pected to interact with the catalyst system (3.1), the poison factor
was adjusted based on expected reaction mechanism as well as
molar ratio of poison and cocatalyst used. The results of the ex-
periments are summarized in Table 3, and the model graphs ob-
tained are shown in Figure 3. The deactivation rate constant was
0mols
−1for lower poison concentrations, which indicates that
the poison does not cause kinetic deactivation of the catalyst at
low concentrations. Given the indicated error bars (calculated for
15% variance for experiments) in Figure 3, the models for poi-
son concentration 0 and 60 ppm show no significant influence
on kinetical parameters and activity levels. At 120 ppm catalyst
poison concentration, deactivation rate constants of 9.8 ×10−5
and 3.9 ×10−4mol s−1were observed for procedures A and B, re-
spectively. These results are in accordance with the general trend
of activity reduction.
For both procedures, the same trend was observed, but in pro-
cedure B the deactivation rate was higher at 120 ppm poison con-
centration. Since greater activity generally results in greater de-
activation, it is plausible that the deactivation constant is higher
for procedure B, where the activity rates are generally higher.
3.1.3. HT-SEC (Pyridine)
To investigate the influence of pyridine on the MWD, samples
were chosen for SEC analysis, the results of which are listed in
Table 4.
The SEC data in Table 4 show shifts in the polymer molecular
weights that depend on the procedure used: In procedure A,
a shift toward lower molecular weights (both Mnand Mw)is
apparent for increasing poison concentration. This indicates
a stronger interaction of the poison with active sites that pro-
duce higher-molecular-weight polymer than with those that
produce lower-molecular-weight polymer. Since in procedure A
the poison interacted directly with the catalyst, it is possible
that the poison preferentially blocks active sites that produce
higher- molecular-weight polymer. The SEC data for procedure B
shows the opposite trend. Here, the poison interacted with the
cocatalyst (TEA) and formed a Lewis acid-base complex.[21]
The cocatalyst-pyridine complex has a lower activating and
reducing power than the poison-free cocatalyst. As catalytic
olefin polymerization is exothermal, the lower activating power
of the cocatalyst might result in a lower temperature at the
individual active site. This may influence the kinetic parameters
of the active site. The kinetic propagation rate constant becomes
proportionally higher than the kinetic deactivation rate constant,
which causes higher-molecular-weight polymer to form at the
active site, as shown by the Mnand Mwvalues in Table 4.[21] The
overall dispersity index (Ð) of all samples measured does not
exhibit a clear trend, which indicates that the poison does not
influence the dispersity index of the polymer product.
3.1.4. MWD Deconvolution (Pyridine)
For all samples, the software (GNU Octave) detected a minimum
number of six active sites to accurately represent the experi-
mental data obtained from SEC analysis. For the Flory–Schultz
deconvoluted sites obtained, the area percentages and Mnwere
calculated and analyzed. For a basic calculation the program
uses estimated starting parameters for number of site types, Mn,
and molar fraction at each site type. However, fixed Mnvalues
were chosen for all deconvoluted active sites in the analysis of
the influence of increasing poison concentrations on the area
Macromol. React. Eng. 2021, 2100020 2100020 (5 of 10) © 2021 The Authors. Macromolecular Reaction Engineering published by Wiley-VCH GmbH
www.advancedsciencenews.com www.mre-journal.de
Figure 3. Model graphs of catalyst activity at various concentrations of pyridine obtained by fitting experimental data to Equation (7).
Tabl e 4 . SEC results for polymer samples exposed to pyridine poisoning.
No. Poison
type
Poison
concentration
[ppm]
Procedure Mn
[g mol−1]
Mw
[g mol−1]
Ð
1 — 0 A 38 500 298 600 7.8
3 Pyridine 60 A 36 600 273 800 7.5
2 Pyridine 120 A 32 400 269 600 8.3
4 — 0 B 43 000 288 300 6.7
6 Pyridine 60 B 46 200 310 700 6.7
5 Pyridine 120 B 65 800 440 700 6.7
percentage of active sites. To achieve this, the Mnvalues obtained
by deconvolution of the reference experiment (without poison)
were used as starting parameters for Mn.Figure 4 illustrates
the results: For polymer in procedure A, we observed a weak
trend of mid to high molecular weights shifting toward lower
molecular weights. This is confirmed by the area percentage of
AS 1 increasing from 2.6% at 0 ppm pyridine to 6.9% at 120 ppm
pyridine. It can be seen that the area percentages of AS 3, AS 4,
and AS 6 decreased, while AS 2 and AS 5 remained unaffected.
This trend matches the found experimental results from the
SEC analysis shown in Table 4.
In procedure B, the active sites behaved differently. Here,
AS 2–3 and AS 4–6 exhibited a reduction and an increase in area
percentage, respectively. This resulted in a shift of lower polymer
molecular weights toward higher polymer molecular weights,
which also accords with the experimental results obtained from
SEC analysis (Section 3.1.3.). In addition to an area percentage
analysis, where the Mnvalues for the deconvoluted active sites
were fixed, an Mnvalue analysis was performed to investigate
any potential shifts. Figure 5 shows the calculated Mnvalues for
each deconvoluted active site, plotted for procedures A and B in-
dividually.
Figure 4. Area percentages of deconvoluted active sites at various pyridine concentrations for two different procedures and fixed Mnvalues.
Macromol. React. Eng. 2021, 2100020 2100020 (6 of 10) © 2021 The Authors. Macromolecular Reaction Engineering published by Wiley-VCH GmbH
www.advancedsciencenews.com www.mre-journal.de
Figure 5. Calculated Mnvalues of Flory–Schultz deconvoluted active sites at various pyridine concentrations for two different procedures.
The Mnvalues of deconvoluted active sites shifted toward
lower values in procedure A and toward higher values in proce-
dure B. This accords with the experimental SEC data and with
the analysis of the area percentages.
3.2. n-Butanol Poisoning
Unlike pyridine, n-butanol is expected to react vigorously with the
cocatalyst according to[22]
ROH +AlEt3→Et2AlOR +C2H6(12)
The diethylaluminum alkoxide can react further to form the
trialkoxide and eventually forms aluminum oxide Al2O3. Alu-
minum oxide is unreactive, which means that the cocatalyst is
destroyed by the catalyst poison and rendered unavailable for cat-
alyst activation. As for pyridine, the effects of n-butanol on activity
rates, deactivation behavior, and MWD were investigated.
3.2.1. Influence of n-Butanol on the Activity Rate
In Table 5, the activity rates at various poison concentrations are
shown. It can be seen that in procedure A catalyst activity did
not change significantly at poison concentrations between 0 and
120 ppm. The tolerance for n-butanol seems to be higher than for
pyridine (which caused a marked reduction in activity at a con-
centration of 120 ppm). Hence, the n-butanol concentration was
increased to 180 ppm, which resulted in a significant activity re-
duction in both procedures. Note that in procedure B the activity
reduction was much more dramatic than in procedure A, as il-
lustrated by respective activity values of 0.6 and 6.9 kgPE gcat-1 h-1 .
Considering both procedures, no other significant influence on
catalyst activity was observed.
3.2.2. Deactivation Behavior (n-Butanol)
Using the modeling approach described in Section 2.6., deacti-
vation of the catalyst system due to n-butanol poisoning was in-
Tabl e 5 . Activity rates obtained at various n-butanol concentrations.
No. Poison
type
Poison
concentration
[ppm]
Procedure Activity rate
[kgPE gcat−1h−1]
Activity
rate [ %]
7 — 0 A 12.9 100
10 n-Butanol 60 A 13.3 103.1
9n-Butanol 120 A 11.5 89.1
8n-Butanol 180 A 6.9 53.5
11 — 0 B 14.4 100
14 n-Butanol 60 B 15.9 110.4
13 n-Butanol 120 B 12.2 84.7
12 n-Butanol 180 B 0.6 4.2
Tabl e 6 . Deactivation rate constants calculated from the fitted model graph
based on Equation (8).
No. Poison
type
Poison
concentration
[ppm]
Procedure Poison
factor
fp
kd
[mol s−1]
7— 0 A 1 0
10 n-Butanol 60 A 0.94 0
9n-Butanol 120 A 0.88 0
8n-Butanol 180 A 0.82 1.9 ×10−4
11 — 0 B 1 0
14 n-Butanol 60 B 0.94 0
13 n-Butanol 120 B 0.88 0
12 n-Butanol 180 B 0.82 7.1 ×10−5
vestigated. The kinetic deactivation rate constant was determined
from the model graphs obtained by Equation (7). In Table 6, the
calculated deactivation rate constants are listed, and in Figure 6
the model graphs obtained are shown.
For both procedures A and B, no increase in deactivation was
observed at n-butanol concentrations of 0, 60, and 120 ppm. This
Macromol. React. Eng. 2021, 2100020 2100020 (7 of 10) © 2021 The Authors. Macromolecular Reaction Engineering published by Wiley-VCH GmbH
www.advancedsciencenews.com www.mre-journal.de
Figure 6. Model graphs of catalyst activity at various concentrations of n-butanol obtained by fitting experimental data to Equation (7).
Tabl e 7 . SEC results for polymer samples exposed to n-butanol poisoning.
No. Poison
type
Poison
concentration
[ppm]
Procedure Mn
[g mol−1]
Mw
[g mol−1]
Ð
7 — 0 A 39 100 275 000 7.0
10 n-Butanol 60 A 39 500 284 100 7.2
9n-Butanol 120 A 38 400 259 200 6.7
8n-Butanol 180 A 38 900 307 000 7.9
11 — 0 B 44 400 271 000 6.1
14 n-Butanol 60 B 57 300 353 900 6.2
13 n-Butanol 120 B 36 400 223 900 6.2
12 n-Butanol 180 B 71 800 491 700 6.9
indicates that the catalyst system has a higher tolerance to n-
butanol poisoning, since for pyridine an increase in deactivation
was already observed at 120 ppm. At 180 ppm n-butanol concen-
tration the deactivation parameters increased to 1.9 ×10−4and
7.1 ×10−5mol s−1in procedures A and B, respectively. It is no-
ticeable, that the variance in procedure B is higher compared to
procedure A, albeit the overall activity levels being slightly higher
for procedure B. Especially the experiment of 60 ppm n-butanol
shows unexpectedly high activity levels. Given the high variance,
though, it cannot be declared as significant increase in activity
compared to other concentrations like 0 or 120 ppm. A signifi-
cant difference in activity levels, despite the variance, was found
for poison concentration of 180 ppm in procedure B where also
the deactivation changed significantly compared to lower poison
concentrations.
3.2.3. HT-SEC (n-Butanol)
As with the pyridine-poisoned PE samples, SEC analysis of se-
lected n-butanol-affected PE samples were performed, the results
of which are shown in Table 7. The overall MWDs of the samples
appear to be largely unchanged and exhibit no clear trend. The
dispersity index Ðbased on the SEC data also shows no clear
trend, although Ðof the procedure B samples shows narrower
MWDs.
3.2.4. MWD Deconvolution (n-Butanol)
Deconvolution of the n-butanol poisoning results was performed
analogously to that of the pyridine sample results (Section 3.1.4.).
Figure 7 plots the area percentages. Procedures A and B exhibit
opposite trends: In procedure A, the area percentages of AS 1–
3 increased with increasing poison concentration, while those of
AS 4–6 decreased. This indicates a shift of higher polymer molec-
ular weights toward lower polymer molecular weights with in-
creasing poison concentration. Since procedure A shows the in-
teraction of the poison with the TiCl4catalyst species, the results
Figure 7. Area percentages of deconvoluted active sites at increasing n-butanol concentrations for two procedures and fixed Mnvalues.
Macromol. React. Eng. 2021, 2100020 2100020 (8 of 10) © 2021 The Authors. Macromolecular Reaction Engineering published by Wiley-VCH GmbH
www.advancedsciencenews.com www.mre-journal.de
Figure 8. Calculated Mnvalues of Flory–Schultz deconvoluted active sites at increasing n-butanol concentrations for two different procedures.
indicate selective poisoning of active sites that produce higher-
molecular-weight polymer. This means that, in addition to a re-
duction in total activity, the composition of the polymer product
also changed.
In procedure B, in contrast, the poison interacted with the
AlEt3cocatalyst species. The area percentages of AS 1–3 and
AS 4–6 decreased and increased, respectively. The polymer
molecular weights therefore shifted from lower to higher values,
which indicates stronger, selective poisoning of active sites that
produce lower-molecular-weight polymer. This finding is sup-
ported by the change in Mnwith increasing poison concentra-
tion (Figure 8). In procedures A and B, an overall decrease and
an increase in Mnwere observed, respectively.
4. Conclusion
We have shown that impurities originating from chemical recy-
cling of polyolefins can have a severe impact on the overall per-
formance of the catalyst system when the recycled monomers are
repolymerized. Pyridine and n-butanol as catalyst poisons were
investigated at a range of concentrations in two different proce-
dures (procedure A: interaction of the poison with the TiCl4cata-
lyst species; procedure B: interaction of the poison with the AlEt3
cocatalyst species).
Pyridine—since it can form a Lewis acid-base complex with
the cocatalyst—lowered the polymerization rate at a concentra-
tion of 120 ppm. At lower concentrations, no significant change
in activity was observed. The composition of the polymer prod-
uct obtained changed depending on the procedure. Although the
overall activity did not change significantly, deconvolution of the
active sites based on our modeling approach showed shifts in
polymer molecular weights toward lower and higher masses for
procedures A and B, respectively. In the case of procedure A, it
is worth mentioning that the poison might act as a chain trans-
fer agent, as the SEC data show results similar to typical chain
transfer agent behavior.
n-Butanol, which is expected to react vigorously with the co-
catalyst according to Equation (12), lowered the overall polymer-
ization rate at a concentration of 180 ppm. Due to high variance,
the SEC results exhibited no significant trend in total product
composition for either procedure. However, MWD deconvolution
modeling showed a change in product composition depending
on procedure and poison concentration. The trends observed are
the same as for pyridine poisoning: In procedure A the molecular
weight shifted toward lower masses and in procedure B toward
higher masses. Like pyridine poisoning, for procedure A it is pos-
sible that the poison might have an affect like is expected from
chain transfer agents, however, as the variance of data is high, the
explanation cannot be proven.
Our results indicate that the polymerization process can yield
products with different properties depending on whether the
input material does or does not contain chemically recycled
monomers. This represents a potential limiting factor in the pro-
cessability and applicability of chemically recycled plastics and
addressing these limitations will require a stronger focus on pu-
rification of chemically recycled monomers.
Supporting Information
Supporting Information is available from the Wiley Online Library or from
the author.
Conflict of Interest
The authors declare no conflict of interest.
Data Availability Statement
The data that supports the findings of this study are available in the sup-
plementary material of this article.
Keywords
catalyst poisoning, chemical recycling, microstructural modeling,
polyethylene (PE), Ziegler–Natta catalyst
Received: May 28, 2021
Revised: August 16, 2021
Published online:
Macromol. React. Eng. 2021, 2100020 2100020 (9 of 10) © 2021 The Authors. Macromolecular Reaction Engineering published by Wiley-VCH GmbH
www.advancedsciencenews.com www.mre-journal.de
[1] Plastics Europe: Plastics – The Facts 2020, https://www.
plasticseurope.org/de/resources/publications/4312-plastics-facts-
2020 (accessed: February 2021).
[2] S. M. Al-Salem, P. Lettieri, J. Baeyens, Waste Manage. 2009,29, 2625.
[3] J. Scheirs, W. Kaminsky, Feedstock Recycling and Pyrolysis of Waste Plas-
tics, Wiley, Weinheim 2006, p. 383.
[4] A. Angyal, N. Miskolczi, L. Bartha, J. Anal. Appl. Pyrol. 2007,79,
409414.
[5] K. Ziegler, E. Holzkamp, H. Breil, H. Martin, Angew. Chem. 1955,67,
426.
[6] L. L. Hegedus, R. W. McCabe, Stud. Surf. Sci. Catal. 1980,6, 471505.
[7] D. M. Argyle, C. H. Bartholomew, Catalysts 2015,5, 145.
[8] C. H. Bartholomew, Stud. Surf. Sci. Catal. 1987,34, 81104.
[9] L. Mayrhofer, C. Paulik, Macromol. React. Eng. 2013,3, 194.
[10] M. Ruff, C. Paulik, Macromol. React. Eng. 2012,8, 302.
[11] H. Toraman, T. Dijkmans, M. Djokic, K. Van Geem, G. Marin, J. Chro-
matogr. A. 2014,1259, 237.
[12] Y. Kissin, J. Catal. 2012,292, 188.
[13] K. Chen, S. Mehdiabadi, B. Liu, J. B. P. Soares, Macromol. React. Eng.
2016,10, 551.
[14] A. Alshaiban, J. B. P. Soares, Macromol. React. Eng. 2012,6, 243245.
[15] A. Alshaiban, J. B. P. Soares, Macromol. React. Eng. 2014,8, 723735.
[16] A. Alshaiban, Propylene Polymerization Using 4th Generation Ziegler-
Natta Catalysts: Polymerization Kinetics and Polymer Microstructural
Investigation, University of Waterloo, Canada 2011.
[17] W. Kaminsky, M. Hoff, S. Derlin, Macromol. Chem. Phys. 2007,208,
13411348.
[18] M. A. Al-Saleh, J. B. P. Soares, T. A. Duevar, Macromol. React. Eng.
2011,5, 587.
[19] V. Touloupidis, A. Albrecht, J. B. P. Soares, Macromol. React. Eng.
2018,12, 1700056.
[20] J. B. P. Soares, T. McKenna, Polyolefin Reaction Engineering,Wiley-
VCH, Manheim 2012.
[21] T. R. Crompton, Comprehensive Organometallic Analysis, Plenum
Press, New York, 1987, p. 155.
[22] T. Mole, E. A. Jeefery, Organoaluminium Compounds,Amsterdam,
London 1972.
Macromol. React. Eng. 2021, 2100020 2100020 (10 of 10) © 2021 The Authors. Macromolecular Reaction Engineering published by Wiley-VCH GmbH
