This is an open access article published under an ACS AuthorChoice License, which permits

copying and redistribution of the article or any adaptations for non-commercial purposes.

Article

Cite This: Ind. Eng. Chem. Res. 2019, 58, 886−896 pubs.acs.org/IECR

Insight into the Synthesis Process of an Industrial Ziegler−Natta

Catalyst
Antoine Klaue,† Matthias Kruck,‡ Nic Friederichs,‡ Francesco Bertola,‡ Hua Wu,*,†
and Massimo Morbidelli*,†
†
Institute for Chemical and Bioengineering, Department of Chemistry and Applied Biosciences, ETH Zurich, 8093 Zurich,
Switzerland
‡
SABIC, Technology & Innovation, P.O. Box 319, 6160 AH Geleen, The Netherlands
*
S Supporting Information
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

ABSTRACT: In Ziegler−Natta catalysis, the catalyst particle

size has a strong influence not only on catalyst performance
but also on the morphology and particle size distribution of
Downloaded via 165.73.13.76 on November 7, 2022 at 08:56:10 (UTC).

the final polymer particles. Fundamental insight into the

catalyst particle formation process is therefore of industrial
importance when addressing specific requirements in the final
products. In the present work, we fully characterize a single-
step catalyst preparation process, which comprises a reactive
precipitation of a MgCl2-supported Ziegler−Natta catalyst,
through decomposition of the hetero-bimetallic complex,
Mg(OR)2·Ti(OR)4, by addition of ethyl aluminum dichloride
(EADC). We track the evolution of both of the concentrations
of the metals (Mg, Ti, Al) as well as Cl in the liquid phase and the size of the formed catalyst particles. It is observed that the
liquid-phase composition is governed by the EADC feed rate under fully Cl-starved conditions. The process can be divided into
two stages: The first stage is dominated by the precipitation of the Mg-based support, and the second stage involves complex
adsorption−precipitation of the Ti species. The growth of the catalyst particle size occurs only in the first stage and is controlled
by the aggregation and breakage events during the MgCl2 precipitation. It follows that the hydrodynamic stress in the reactor
plays the essential role in controlling the catalyst size. In the second stage, no further particle growth occurs, not only because of
the depletion of Mg in the liquid phase but also because the adsorbed Ti complex stabilizes the particles against aggregation.
Finally, we have performed polymerization tests with the prepared catalysts and found that the size distribution of the polymer
particles indeed closely replicates the one of the used catalyst particles.

■ INTRODUCTION
Within the polymer industry, polyolefins make up more than
half of the global plastics demand,1 and their worldwide
production is expected to keep increasing along with the gross
world production.2 The production processes of these
materials in the late 1930s were run without catalyst by free-
radical polymerization of ethylene to form low density
polyethylene, requiring 200−300 °C and several thousands
of bars.3 It was only in the 1950s with the discovery of the
heterogeneous coordination catalysts4,5 that the process could
be run at milder conditions. The most used catalyst system in
polyolefin industry remains the MgCl2-supported Ziegler−
Natta catalysts (ZNC).6 They are composed of Ti complexes
of the form TiX3−4, with X being a chloride or an alkoxide
group, which are chemisorbed onto a MgCl2 matrix.7 Usually, a
co-catalyst is used, comprising an aluminum alkyl of the form
AlR3, which results in the activation of the catalyst by
alkylation of the Ti species.8
The chain growth during the polymerization occurs on the
solid-bound active species, which leads to the replication of the
morphology of the parent catalyst particle size distribution
(SD) within the produced polymer particles.9 It has been
shown theoretically by Floyd et al.10 and Kanellopoulos et al.11
that the catalyst size impacts the catalyst activity by affecting
mass and energy-transfer limitations. This is supported by
experimental data from Schnecko et al.12 and more recently
from Philippaerts et al.13 during slurry polymerization.
However, this conclusion remains in contrast with the studies
of Dall’Occo et al.,14 who observed an insignificant effect of
particle size on catalyst activity. On the other hand, they
pointed out that the morphology of the catalyst particle is of
paramount importance in the polymer particle production
processes based on slurry or gas-phase technologies.15 As a
matter of fact, the size and morphology of the ZNC particles
are of primary importance in the production of polyolefins as
they can affect the plant operation and the catalyst
productivity. Up to now, although many preparative methods
Received: October 26, 2018
Revised: December 21, 2018
Accepted: December 21, 2018
Published: December 21, 2018

© 2018 American Chemical Society 886 DOI: 10.1021/acs.iecr.8b05296

Ind. Eng. Chem. Res. 2019, 58, 886−896
Industrial & Engineering Chemistry Research
for ZNCs have been developed as reviewed by Karol,16 only a
very basic understanding of the phenomena controlling the
particle size during the synthesis is available.
One important preparation method relies on the use of
Mg(OEt)2 as precursor for the formation of the support.
Chumachenko et al.17 studied the formation of a ZNC starting
with solid Mg(OEt)2 and following the chemical composition
of the solid products through different reactions including
chlorination, followed by donor addition and titanation. The
authors reported that the ZNC morphology was dominated by
the breakage of the initial solid Mg(OEt)2 leading to a broader
particle size distribution of the final catalyst. As an alternative
to solid Mg(OEt)2, soluble Mg sources may be used. These
can either take the form of Grignard reagents18,19 or other
alkoxides.20 For the latter case, Redzic et al.20 studied the
precipitation of MgCl2 through the addition of AlEtCl2 to a
toluene solution of magnesium 2-ethylhexanolate with focus on
the MgCl2 crystallite size within the produced solid support.
Soluble hetero-bimetallic alkoxides21 involving both Ti and Mg
complexes may be used. The addition of a chlorinating agent
to such compounds leads to the decomposition of the initial
metal complex and the formation (precipitation) of the
catalyst. Studies on the decomposition reaction of such
compounds are very scarce, mainly due to the low solubility
of magnesium alkoxides in hydrocarbon solvents, and a proper
description of the involved mechanisms is still elusive or very
limited.22,23
In general, the particle size obtained through fed-batch
precipitation processes is the result of the primary processes of
nucleation and growth, but very often, secondary processes
such as aggregation and breakage also affect the morphological
characteristics of the final product.24 This was observed by
Söhnel and Mullin25 for the crystallization of strontium
molybdate and by Ilievski26 in the case of aluminum hydroxide.
The interplay between shear forces and the solute concen-
trations27 (e.g., agglomerate bond formation) might also play a
significant role. Clearly, the understanding of the particle
formation mechanism requires the combination of contingent
information about the chemical reactions and the evolution of
particle size.
In the present work, we propose a methodology to study and
fully characterize the particle formation process during ZNC
preparation using a single-step process28 as an example. This
proceeds through the decomposition of a hydrocarbon soluble
hetero-bimetallic complex of the type Mg(OR)2·Ti(OR)4 by
addition of AlEtCl2 in fed-batch mode. This process requires
only one reactive step and is of primary importance in the
frame of the production of polyolefins.6 The identification of
the reactions affecting the particle size is a challenge, and the
decomposition of the precursor complex is not described in the
open literature. The identification of the relevant precipitation
reactions and the evaluation of their kinetics would indeed
contribute to understand the catalyst formation process. With
the proposed methodology, this is achieved by monitoring the
composition of the liquid phase during the ZNC preparation
via ICP-OES and by monitoring the size growth of the ZNC
particles through small-angle static light scattering. The
combination of these two techniques provides a powerful
tool for the understanding of this process and opens new
insight into the actual control over the catalyst characteristics
and consequently on the final properties of the corresponding
final polymer particles.
887
Article
■ MATERIALS AND METHODS
All reactant manipulations, reactions, and measurements were
done under nitrogen atmosphere by using a glovebox or the
Schlenk technique.
Materials. Commercial magnesium diethoxide (Fluka) was
used without any further purification. Titanium tetrabutoxide
(Fluka) was purified by distillation before use (140 °C, 1 ×
10−3 mbar according to Armarego and Chai29). Commercially
available ethyl aluminum dichloride (EADC) in solution (50
wt % in hexane) was used (Chemtura). All reactions and
particle size measurements were done in hexane (HPLC grade,
Fisher Scientific). Hexane was dried over molecular sieves (4
Å, Acros) to maintain a water content below 3 ppm. The water
content of the solvent was checked by means of a Karl Fisher
titrator (Titrino Coulometer, Metrohm). To minimize oxygen
concentration, nitrogen was bubbled through the solvent tank
for 24 h before each reaction, and the tank was stored under
nitrogen atmosphere.
Preparation of the Mg-Supported Ziegler−Natta
Catalyst. The preparation of the catalyst is based on the
procedure described by Friederichs et al.28 Briefly, magnesium
diethoxide, titanium tetrabutoxide, and ethyl aluminum
dichloride are reacted in two steps. In the first one, magnesium
diethoxide and titanium tetrabutoxide are mixed to form a
hydrocarbon soluble precursor complex. In the second step,
EADC is fed continuously to this precursor under intense
agitation. Finally, the formed dispersion is put under reflux
conditions to form the final catalyst.
Preparation of the Catalyst Precursor Complex. The
hydrocarbon soluble Mg- and Ti-containing complex is
synthesized following the procedure described in ref 28 using
73 g of Mg(OEt)2 powder, 105 g of Ti(OBu)4, and 733 g of
hexane to reach a weight fraction of the complex of about 20%.
The resulting solution was clear, with a small amount of gray
precipitate that was filtered out or decanted before use.
Synthesis of the ZNC. The synthesis of the ZNC particles
was carried out in a thermostated 1 dm3 glass reactor, which
was equipped with a condenser, three baffles, and a paddle-
type agitator as depicted in Figure S1. To keep track of the
temperature within the reactor, a K-type thermocouple was
fitted to the setup. All reactions were carried out under
nitrogen atmosphere at constant pressure. Before each
synthesis, the reactor inner surface was heated up with the
help of a heat gun to eliminate traces of water. Subsequently, it
was closed and put under nitrogen flow.
Each experiment consisted of the following steps: The Mg−
Ti precursor solution (134 mL) was placed in the reactor and
diluted with 289 mL of hexane. A 98 mL portion of the original
EADC solution (50 wt %) were placed in a 250 mL Schlenk
tube and diluted with 56 mL of hexane. The diluted EADC
solution was fed continuously to the diluted precursor solution
under agitation with the help of a syringe pump (Lambda)
equipped with a stainless steel syringe (Harvard Apparatus)
and a feed system composed of stainless steel tubing and a
three-way valve (Swagelok). After all EADC had been added,
the reactor was heated to 69 °C to reach reflux conditions. The
dispersion was kept under reflux conditions for 2 h. Finally, the
particles were washed over a filter frit (P10 type, 10 μm pores)
using 2 L of hexane.
SEM Micrographs. The SEM micrographs were collected
on a Leo 1530 (Zeiss). In order to avoid contact with ambient
air, the samples were plated on a doped silicon wafer support
DOI: 10.1021/acs.iecr.8b05296
Ind. Eng. Chem. Res. 2019, 58, 886−896
Industrial & Engineering Chemistry Research
and subsequently protected by a rubber cover. The microscope
was equipped with a prechamber, where the pressure could be
reduced before introducing the samples under the electron
beam. The rubber covers could be removed by using the
pressure difference between the prechamber and the sample
below the rubber cover. This was done to ensure minimal
exposure of the samples to ambient air.
The micrographs were acquired with an acceleration voltage
of 5 kV using an aperture of 30 μm. For the detection, both an
in-lens and SE2 detectors were used.
Composition Analysis. Liquid-Phase Composition anal-
ysis. The liquid-phase composition was measured over time
during the process. For this purpose, aliquots of 5 mL of
dispersion were sampled with the aid of a syringe and using a
stainless steel needle. Each sample was immediately weighed
and filtered over a PTFE syringe filter with 0.2 μm pores
(Macherey Nagel), whereby the solid was discarded. The
liquid phase was poured into a 20 mL glass vial, which was
immediately closed and weighted. The solvent in the liquid
samples was evaporated under nitrogen, and a 5 M sulfuric acid
solution (Fluka) was added to the remaining precipitate to
completely dissolve it. The obtained solutions were diluted
with deionized water to reach an approximate analyte
concentration of 20 ppm. The measurements were performed
on an Arcos ICP-OES (Arcos, Spectro Analytical Instruments,
Inc.).
Final Catalyst Composition Analysis. The composition of
the final catalyst was assessed by XRF according to Bichinho et
al.30 for Mg, Ti, Al, and Cl. For this analysis, the catalyst
samples were filtered and dried under high vacuum (1 × 10−3
mbar). The same protocol was implemented for the
determination of the residual Ti content in the polyethylene
particles subsequently produced in the polymerization reactor.
The content of ethanol and butanol was assessed by GC−FID
after quenching the samples.
Particle Size Measurements. The particle size measurement
during and after their preparation was performed off-line by a
small-angle static light scattering (SASLS) instrument, Master-
sizer 2000 (Malvern), using a flow-through cell. After cleaning,
the flow channel was heated through the cell casing with a heat
gun in order to minimize the presence of water, and the cell
window was put in an oven at 140 °C. Subsequently, the flow-
through cell was reassembled and flushed with nitrogen before
being filled with dry hexane. The sampling was performed with
a polypropylene syringe (Braun Injekt) fitted with a stainless
steel needle (length = 25 cm; inner diameter = 1.2 mm). The
sample volume was typically in the range from 0.1 to 1 mL,
depending on the ZNC concentration, and then diluted in a
Schlenk flask equipped with a stirrer bar and filled with 100 mL
dry hexane to minimize the effect of multiple scattering. The
dilution is also needed to prevent possible aggregation of
catalyst particles. The total time required for sampling, dilution
and injection into the flow cell was less than 30 s.
As discussed later, the catalyst can be considered as compact
fractal-like clusters composed of primary particles of ∼100 nm
in diameter. Thus, the scattered intensity, I(q), can be
expressed as follows31
I(q) = I(0) ·S(q) ·P(q) (1)
where I(0) is the scattered intensity at q = 0, q is the
magnitude of scattering wave vector, defined as
888
Article
4πn ji θ zy
sinjj zz
k2{
q=
λ (2)
with n the refractive index of the solvent, λ the wavelength, and
θ the scattering angle. In eq 1, P(q) is the form factor of the
primary particles, and S(q) is the structure factor, representing
the arrangement of the primary particles inside the cluster.
When the ZNC particles are considered to be fractal objects,
their average morphology is characterized by two quantities,
the average radius of gyration, ⟨Rg⟩, and the fractal dimension,
ÄÅ É
ÅÅ (q⟨R ⟩)2 ÑÑÑ
df. The ⟨Rg⟩ value was estimated using the Guinier plot:
ÅÅ ÑÑ
= expÅÅÅ− ÑÑ for q⟨R ⟩g < 1
ÑÑ
ÅÅ ÑÑÖ
I(q)
ÅÇ
g
I(0) 3 (3)
The df value could be estimated directly from the power-law
region of I(q) because of the substantially large size of the
ZNC particles with respect to that of the primary particles:
I(q) ≈ S(q) ∝ q−d f for q⟨R ⟩g ≫ 1 (4)
In addition, for the ZNC particles as well as the
corresponding polymer particles obtained after polymerization,
we have used the SASLS data to estimate their size distribution
based on the Mie theory of spherical particles.
Catalyst Tests. Polymerization tests were performed with
the produced catalyst particles to evaluate their activity and the
particle size distribution of the corresponding polymer powder.
To this purpose, semibatch slurry polymerization of ethylene
was carried out.28 These reactions were carried out using
hexane as solvent. Further operating conditions are summar-
ized in Table 1. To obtain the polymer yield, the residual Ti
concentration inside the polymer mass was measured by XRF.
Table 1. Summary of the Polymerization Conditions of the
Catalytic Tests
hexane volume (L) 5
catalyst (mg) 20
co-catalyst TIBA (mmol/mL) 0.8
H2/ethylene (mol/mol) 3
T (°C) 85
t (min) 120
pethylene (bar) 1.2
■ RESULTS AND DISCUSSION
To monitor the catalyst formation, we sampled the reaction
mixture at fixed time points during the precipitation reaction,
and the particle size, as well as the elemental composition of
the liquid phase was determined. With this approach,
elemental information on the decomposition reactions of the
precursor complex could be linked to the evolution of the
ZNC particle size during the process.
Time Evolution of the Liquid-Phase Composition.
Effect of the EADC Feeding Rate. The concentrations of
species in the liquid phase were determined as a function of
time at two EADC feed rates, Q = 5 and 10 mL/min, for a total
feed time of tf = 30 and 15 min, respectively. For Q = 10 mL/
min, the concentrations of the soluble Mg and Ti species
normalized by their initial concentration are shown by the
empty symbols in Figure 1a. It is seen that the Mg
concentration in the liquid phase decays rapidly and drops
below the detection limit after about 6 min. This process is
DOI: 10.1021/acs.iecr.8b05296
Ind. Eng. Chem. Res. 2019, 58, 886−896
Industrial & Engineering Chemistry Research
Figure 1. Normalized concentrations of Mg (triangles) and Ti (squares) (a and b) and Al (squares) and Cl (triangles) (c and d) measured by ICP-
OES at EADC feed rates of 5 mL/min (full symbols) and 10 mL/min (empty symbols), respectively.
characterized by the appearance of white precipitates soon
after the first droplets of the EADC solution enters into the
reactor, whereas the filtrate remains colorless. The Ti
concentration decays much slower than the Mg concentration,
and it reaches a local minimum at approximately 10 min, after
which the Ti concentration starts increasing until the EADC
feed is complete, i.e., at about 15 min. The corresponding Al
concentration during the feeding is shown in Figure 1c, which
is normalized by its maximal hypothetical concentration, i.e.,
based on total fed EADC in the system, as given in Table S2. It
increases with time, following a linear trend, and then reaches
unity, i.e., its maximal hypothetical concentration. This
indicates that during feeding the Al consumption in the solid
phase is negligible. The Cl concentration shown in Figure 1c is
also normalized by the maximal hypothetical concentration of
Al.
The results corresponding to the EADC feed rate of Q = 5
mL/min are shown by full symbols in Figure 1, and they are
similar to those at Q = 10 mL/min but with a delayed
development of the behavior described above. When replotting
all data as a function of time, τ = t/tf, renormalized using the
total feed time, tf, all data overlap as shown in Figure 1b,d. This
result indicates that the process is controlled by the feed rate
and not by the chemical reaction. In order to better investigate
this behavior, a further experiment was done with an EADC
feed rate, Q = 5 mL/min, but the feed was stopped after 10
min or τ = 0.31, after which samples were continuously taken
for analysis. The results can be found in Figure S2. It is seen
that after the EADC feeding is stopped, the concentrations of
all species in the liquid phase do not change with time. This
further supports that the process in the reactor is fully
controlled by the EADC feed rate or, in other words, the
889
Article
characteristic time of involved reactions is much smaller than
the feeding time.
It is worth noting in Figure 1,d that initially the Cl
concentration is practically equal to the Al concentration, and
after 10 min, it becomes larger and reaches a maximum at the
end of the feeding stage. This means that Cl is always available
in the liquid phase, while the process is controlled by the
EADC feeding rate. This allows us to conclude that only one
Cl atom in EADC is reactive and participates in the
precipitation. This phenomenon can be understood when
considering that EADC is a Lewis acidic and its structure can
be stabilized by forming dimers. According to a comparison of
experimental and calculated IR spectra,32 the dimer has the
forms depicted in Figure 3. It appears plausible that among the
four Cl atoms in the dimer only the nonbridging two can
participate to the Cl transfer. The Cl transfer leads to the
eventual formation of MgCl2, which is lipophobic and
insoluble in hexane. Furthermore, it is seen that Al remains
soluble throughout the process, since it undergoes a ligand
exchange reaction with soluble Mg complexes as follows:
Mg(OR)2 + AlEtCl 2 → Mg(OR)2 − x Cl x + AlEt(OR)x Cl 2 − x
(5)
Peculiar Evolution of the Ti Concentration. The behavior
of Ti concentration in the liquid phase shown in Figure 1
needs some more discussion. Unlike MgCl2, chlorotitanium
alkoxides, i.e., [TiCl(OEt)3] and [TiCl(OBu)3], as well as their
dichlorinated analogues are known to be viscous liquids and
soluble in hexane.33 This has been further verified by reacting
the same EADC solution with only [Ti(OBu)4]. Although the
reaction mixture turned yellow, no formation of any solid
could be observed, thus confirming negligible precipitation of
Ti-bearing species during the EADC addition. On the other
DOI: 10.1021/acs.iecr.8b05296
Ind. Eng. Chem. Res. 2019, 58, 886−896
Industrial & Engineering Chemistry Research
Figure 2. Normalized concentrations of Ti (a) and Cl (b) during the stage of heating to reflux conditions (69 °C), starting from room temperature
and reaching 69 °C within 10 min.
hand, it was already demonstrated in the literature that Ti(IV)
complexes such as TiCl4 can adsorb on sufficiently Lewis acidic
crystal cuts of MgCl2, more specifically on the (110)
face.7,34−36 Thus, in Figure 1, the initial decay of Ti in the
liquid phase with time can be explained by adsorption of
chloro titanium alkoxides to the MgCl2 crystallite faces. It
follows that both Mg and Ti species contribute to the
disappearance of Cl from the liquid phase during the initial
stage.
After decreasing to a local minimum at τ = 0.7 in Figure 1b,
the liquid phase Ti concentration starts to increase upon
further feeding of EADC. Simultaneously, the Cl concentration
in the liquid phase increases and becomes larger than the Al
concentration. This behavior can be related to what is
commonly referred to as catalyst leaching,15,35 where the
active Ti-based complex is detached from the surface upon
extended contact time with a Lewis acid, which in the present
case is further fed EADC. The formation of soluble Ti(IV)
complexes upon further addition of EADC can be explained by
considering that alkoxychloro complexes of Ti(IV) can
undergo complexation equilibria with EADC.36 Through a
molar balance of Cl and Ti getting soluble in this stage, the
dissolution of Ti and Cl species occurs with a molar ratio of
1:2. This indicates that a complex with a possible formula of
[TiCl2(OR)2] has been released from the solid to the liquid
phase.
Stage after Completing the EADC Feeding. After
completion of the EADC feed, the system was equilibrated
for 10 min. Subsequently, the reactor was heated to 69 °C
within 10 min, which was held for 2 h. At this stage, as shown
in Figure 2a,b, both the Ti and Cl concentrations in the liquid
phase decrease with time. In particular, the Ti concentration
reduces practically to zero after about 15 min, while meantime
the normalized Cl concentration reaches unity, i.e., equal to
that of Al. It is obvious that now the soluble Ti species react
with the Cl species to precipitate to the solid phase.
Furthermore, a color change to red was observed, which
indicates the reduction of Ti(IV). Commonly, the Ti(IV)
reduction occurs through transfer of an alkyl group to the Ti
center and subsequent elimination of an alkyl radical, leading
to the reduced form of Ti(III), Ti(OR)3−xClx, which are
insoluble in hydrocarbons.36 In the study of Cucinella et al., it
is concluded that under large excess of EADC, solid TiCl3 is
formed. According to our liquid-phase concentration measure-
ments, the molar ratio of the precipitated Cl to the precipitated
Ti during the heating step has a value of 3.75, which is in
agreement with the study by Cucinella et al.36
890
Article
Figure 3. Structure of EADC according to Tarazona et al.32
Time Evolution of the ZNC Particle Size. Particle
Growth Mechanism. The particle size evolution in terms of
the root-mean-square gyration radius (⟨Rg⟩) is shown in Figure
4 as a function of time at a stirring speed of 1400 rpm and
three different EADC feed rates, 2.5, 5.0, and 10.0 mL/min.
Let us first consider the case at the EADC feed rate of 5.0 mL/
min in Figure 4a. The EADC addition was completed after 30
min, followed by subsequent heating. It shows that the ⟨Rg⟩
value is around 1 μm after 1 min and very quickly reaches a
maximum value within 10 min, and then in the remaining
EADC feed time the particle size remains constant. In the
subsequent time, no measurable change of the catalyst particle
size is observed, suggesting that all the reactions responsible for
the particle growth are occurring during the feed of the EADC
solution.
When the results at the EADC feed rate of 5.0 mL/min are
compared to those at feed rates of 2.5 and 10.0 mL/min in
Figure 4a, it becomes clear that the faster the feed rate, the
quicker the particles reach their final size. One may also
conclude that the feed rate has no significant effect on the final
particle size under the tested conditions. In the same figure, it
can be noticed that the particle size increases dramatically at a
certain point in the process. In analogy to the composition
analysis, this size evolution is also reported against the
dimensionless time τ in Figure 4b. It is found again that the
three curves practically overlap. This indicates that the particle
growth in the three cases follows the same mechanism, which
is closely related to the measured liquid-phase composition
and, therefore, to the EADC feeding rate.
To gain further insight into the particle growth mechanism,
we have assessed the colloidal stability of the particles along
their growth. For this, we kept the sample for 1 h in a Schlenk
tube under gentle agitation by a magnetic stirring bar and then
measured again the ⟨Rg⟩. The measured values are compared
in Figure 5a. It is seen that 1 h after sampling, the ⟨Rg⟩ value
increases quite a bit for samples taken at short times. Such a
difference decreases, however, as time proceeds and becomes
insignificant after about τ = 0.5. This indicates that the initially
formed particles have poor stability in hexane, and significant
aggregation occurs during the 1 h of storage. In Figure 5b, the
normalized scattering intensity curves corresponding to the
DOI: 10.1021/acs.iecr.8b05296
Ind. Eng. Chem. Res. 2019, 58, 886−896
Industrial & Engineering Chemistry Research Article

Figure 4. ⟨Rg⟩ value as a function of time (a) and dimensionless time (b) for the EADC feed rate at 2.5 mL/min (circles), 5.0 mL/min (triangles),
and 10.0 mL/min (squares), respectively. The stirring speed is set at 1400 rpm.

Figure 5. (a) ⟨Rg⟩ values as a function of the dimensionless feed time, measured immediately after sampling (triangles), after 1 h of storage
(circles), and after 1 h of storage and then 5 min of sonication (squares). (b) Normalized scattering intensity I(q)/I(0) as a function of q,
corresponding to the points at τ = 0.16 in (a). EADC feed rate = 10 mL/min; stirring speed = 1400 rpm. The continuous curves in (b) are fitting
using the Fisher−Burford expression.31

first measurement point in Figure 5a are compared. With a 1 h
delay for the measurement, in addition to the much larger ⟨Rg⟩,
the value of the fractal dimension (df = 2.5) is also much
smaller than that (df = 2.9) without delay. The smaller df value
indicates a more open morphology and/or a broader particle
size distribution, thus further supporting the occurrence of
aggregation among the formed catalyst particles during the 1 h
storage. It is evident that the stability of the particles improves
along the EADC feeding (i.e., along the particle growth)
because the ⟨Rg⟩ increase during the 1 h storage tends to
vanish.
In addition, to explore the strength of the clusters formed
during the 1 h storage, we have sonicated the samples after the
storage in an ultrasonic bath (Branson) for 5 min, and the ⟨Rg⟩
values were remeasured. The results, shown in Figure 5a by
squares, show that the ⟨Rg⟩ value reduces and approaches very
close the original one before the 1 h storage. This indicates
that the bonding among the particles in the aggregates formed
during the 1 h storage is not as strong as that formed within
the reactor and can be easily destroyed with a short ultrasonic
treatment. In addition, the fact that there is no significant
difference in the ⟨Rg⟩ values after τ ≈ 0.5 for the three cases
indicates that the finally formed particles in the reactor are
stable in hexane. The final ZNC particles are in fact stable even
after storage for several months, as reported in Table S3.
To explain the stability of the particles along their growth,
we should consider that the very first particles formed during
the (precipitation) reaction are very small and mainly
composed of lipophobic MgCl2, as discussed in the previous
891
section. For these two reasons, the particles are not very stable
and exhibit a strong tendency toward aggregation. Thus, the
particles formed inside the reactor at the initial times result
from both precipitation and aggregation. This is also the
reason why aggregation takes place during the storage of the
samples after sampling and dilution. The aggregation during
storage is only due to van de Waals attraction, and the formed
physical bonds are therefore weak and thus easy to break, as
shown by the sonication experiment. On the other hand, the
aggregation inside the reactor leads to stronger bonds because
the continuously fed EADC leads to precipitation inside the
aggregates, thus strengthening the cohesive forces among the
particles.37 This process is described by Mumtaz27 based on
the formation of a crystal neck between two surfaces of
aggregating particles and received some validation by
Hounslow et al.38 for calcium carbonate and calcium oxalate
monohydrate. The increased stability of the particles in the
later stage of the precipitation could be explained by
progressive formation of a lipophilic surface composition. It
is found, from quenching experiments of the final ZNC
particles with water, that significant quantities of ethanol and
butanol are formed, as reported in Table S3. This confirms that
alkoxy ligands are indeed present on the surface of the final
ZNC particles, and these are responsible for the improved
stability.
Effect of the Shear Rate on the Particle Size. As discussed
above, the initially formed particles inside the reactor result
from both precipitation and aggregation. Then it is expected
that the shear rate generated by the stirring system would affect
DOI: 10.1021/acs.iecr.8b05296
Ind. Eng. Chem. Res. 2019, 58, 886−896
Industrial & Engineering Chemistry Research
strongly the particle size because it not only interferes with the
aggregation but also breaks up the formed aggregates. Figure 6
Figure 6. ⟨Rg⟩ values as a function of the dimensionless EADC feed
time at the EADC feed rate of 2.5 mL/min and at the stirring speed,
1400 rpm (triangles), 400 rpm (squares), and 1400 rpm until τ = 0.33
and then 400 rpm for the remaining time (diamonds).
shows runs at the EADC feed rate of 2.5 mL/min but at two
different stirring speeds, 400 and 1400 rpm, respectively. From
the results, it is evident that the ⟨Rg⟩ value reaches its
maximum at the same dimensionless time (around τ = 0.33)
for the two runs at different stirring speeds, but the maximum
⟨Rg⟩ value reached at 400 rpm is much larger than that at 1400
rpm. This clearly shows that breakage, which sets in at lower
size for larger shear, plays a major role. Considering the
knowledge gained above on the liquid-phase composition, a
further experiment was carried out to better identify the
growth mechanism at different times. In particular, we set
initially the stirring speed to 1400 rpm, and then at the EADC
feed time of τ = 0.33 the stirring speed was reduced to 400
rpm. We selected τ = 0.33 because this is the typical location
where ⟨Rg⟩ reaches its maximum and the depletion of soluble
Mg is complete (see Figures 1 and 5). The obtained results are
shown again in Figure 6. It is seen that there is no significant
difference between the ⟨Rg⟩ evolutions obtained at 1400 rpm
all the time and at 1400 rpm before τ = 0.33 and then 400 rpm.
Two conclusions may be drawn from this experiment.
First, it is further confirmed that the particle growth is
strongly linked to the MgCl2 precipitation, and once this
process is completed, the particles do not grow any more, as
shown in the scheme in Figure 7. In this period the stirring
speed does affect the particle size through breakage. Second, as
discussed above, the Ti species do not precipitate when in
contact with EADC, but instead, they adsorb in the form of
chloro titanium alkoxides to the MgCl2 crystallite surface.
Figure 7. Scheme showing the particle formation process, which involves the initial precipitation, the interplay between aggregation and breakage,
the subsequent precipitation on the formed aggregates, stabilization, and TiCl3 precipitation after completion of the EADC feed.
892
Article
Thus, the results in Figure 6 confirm that it is the adsorbed
chloro titanium alkoxides on the MgCl2 surface that stabilize
the particles such that no further aggregation occurs inside the
reactor in the later time, and therefore, no influence of the
stirring speed is observed.
In addition, it should be mentioned that the formation of
crystal necks within the formed aggregates along the MgCl2
precipitation also contributes to the particle growth because
this permits the particles to better withstand the hydrodynamic
stress caused by the stirrer. A similar mechanism was already
proposed by Ilievski et al.26 for the case of Al(OH)3
crystallization in caustic aluminate solutions. Furthermore,
the present observation is in line with those by Hartel et al.39
and Wójcik et al.,40 where a dependency of agglomeration and
breakage rate on the applied supersaturation was observed in
the case of calcium oxalate and calcium carbonate, respectively.
As discussed above, during the particle growth, breakage
induced by the hydrodynamic stress of the stirring system plays
a dominant role. It is known that whether an aggregate is
broken or not is determined by two opposite factors, the
hydrodynamic stress and the cohesive strength within the
aggregate. The breakage rate is a monotonically increasing
function of the particle size,41 such that a given hydrodynamic
stress is associated to a critical aggregate size, beyond which
the aggregate would break.42 It is evident that the critical
aggregate size decreases as the hydrodynamic stress increases.
In the literature, the relationship between the critical aggregate
size and the hydrodynamic stress is typically expressed in terms
of a power law, ⟨Rg⟩ ∝ ε−γ, where ε is the energy dissipation
rate and the exponent, γ, has a typical value around 0.3.43−47
For example, Soos et al.42 found a scaling exponent of 0.26 in
breaking aggregates formed from destabilized polystyrene
nanoparticles. For our system, we have estimated the mean ε
values, ⟨ε⟩, based on the power input
P
ε=
ρV (6)
where ρ and V are the density and volume of the fluid and P is
the power input rate, which can be estimated using the
following correlation:48
P = NPN3D5ρ (7)
Here, N is the stirring speed, D the paddle diameter, and NP
the power number. For the specific stirrer used in this work,
considering that the applied stirring speeds are in the turbulent
regime (Reynolds number between 36000 and 130000), we
can use a value of 2 for NP.49 Therefore, we have prepared the
final ZNC particles at the EADC feed rate of 10 mL/min at
three stirring speeds, 400, 900, and 1400 rpm, about which the
DOI: 10.1021/acs.iecr.8b05296
Ind. Eng. Chem. Res. 2019, 58, 886−896
Industrial & Engineering Chemistry Research Article

Figure 8. (a) ⟨Rg⟩ values of the final ZNC particles, prepared at the EADC feed rate of 10 mL/min and three stirring speeds, 400, 900, and 1400
rpm, plotted as a function of estimated average energy dissipation rate, ⟨ε⟩. (b) Corresponding normalized scattering intensity curves, where the
continuous curves are fit from the Fisher−Burford expression.31

Figure 9. SEM micrographs of the ZNC particles obtained at a stirring speed of 1400 rpm (a, b), 900 rpm (c, d), and 400 rpm (e, f).

⟨Rg⟩ evolutions with time are reported in Figure S3. The ⟨Rg⟩
values of the final ZNC particles are plotted in Figure 8a as a
function of ⟨ε⟩ in a log−log plane, and the estimated value for
γ from the plot is 0.27, which is consistent with the typical
values obtained in the literature mentioned above for breakage
processes. This result further confirms that the formation of
the ZNC particles is controlled by the breakage process. It may
be of practical interest to mention that a γ value of 0.27 implies
multiplication of the energy dissipation rate by a factor of 8 in
order to decrease the particle size by a factor of 2.
Figure 8b shows the normalized scattering intensity curves
corresponding to the prepared final ZNC particles. It is seen
893
that the slope of the power-law regime, i.e., the fractal
dimension (df) of the particles, increases from 2.5 to 2.9 as the
stirring speed increases from 400 to 1400 rpm. This means that
the compactness of the ZNC particles increases as the stirring
speed in the reactor increases. This result is different from the
typical results obtained from breakage of aggregates formed
from polymeric nanoparticles.42 In the latter cases, the fractal
dimension of the aggregates is independent of the shear rate
(or stirring speed). This difference may be related to the
continuous precipitation occurring in our system, which leads
to continuous strengthening of the cohesive forces in the
aggregates, thus affecting both breakage and restructuring.
DOI: 10.1021/acs.iecr.8b05296
Ind. Eng. Chem. Res. 2019, 58, 886−896
Industrial & Engineering Chemistry Research Article

Figure 9 shows the SEM micrographs of the ZNC particles
prepared at the stirring speed of 400, 900, and 1400 rpm, and
they confirm the SASLS results that the catalyst size decreases
as the stirring speed increases. On the other hand, no
significant difference in the compactness of the catalyst
particles can be observed, probably because the fractal
dimension is relatively high in all cases.
As a further test of the role played by aggregation and
breakage, we have also assessed the effect of the reactant
concentration by preparing the catalyst particles by diluting
both the catalyst precursor complex and the EADC feed by a
factor of 1, 5, and 25. The corresponding ⟨Rg⟩ value of the final
ZNC particles is 4.6, 3.7, and 3.0 μm, respectively, i.e., the size
of the final ZNC particles decreases as the concentrations of
the reactants decrease. This is in contrast to classical
crystallization systems, where a lower supersaturation is
associated with an increased growth rate as reviewed
extensively by Slund et al.50 for the crystallization of benzoic
acid. Conversely, this is the typical behavior observed for
systems exhibiting strong aggregation and agglomeration such
as in the case of calcium oxalate39 and calcium carbonate.40 In
a stirred vessel, aggregation and breakage typically occur
simultaneously. However, since aggregation is a second-order
process, while breakage is first-order, dilution reduces the rate
more for aggregation than for breakage, thus shifting the
aggregate size toward a smaller value. Furthermore, the
formation of crystal necks within the formed aggregates
becomes faster at higher supersaturation, leading to mechan-
ically stronger agglomerates.51
Polymerization Test of the Prepared ZNC Particles.
The semibatch slurry polymerization of ethylene was used to
test the prepared ZNC particles under the conditions
summarized in Table 1. As a measurement of the catalyst
productivity, the XRF measurements of the residual Ti mass
fraction in the polymer mass are provided in Table 2. Because
Table 2. Concentration of Residual Ti in Polymer Measured
by XRF, Which Was Produced Using the ZNCs Prepared in
This Work at Different Stirring Speeds, as Well as the
Catalyst Yield, CY, and Replication Factor, Φ
residual Ti
stirring speed used in the concentration in
ZNC preparation (rpm) polymer (ppm) CY (kg/g)
400 9.9 10.5 ± 0.2
900 10.7 9.81 ± 0.09
1400 9.7 10.8 ± 0.2
the mass of catalyst initially injected is orders of magnitude
smaller than the final polymer mass, the catalyst yield (CY) can
be computed from the residual Ti mass fraction within the
polymer xPTi and the one in the initial catalyst xCTi with
C
x Ti
CY = P
x Ti (8)
As reported in Table 2, the CY value remained constant for
all three ZNCs prepared using three different stirring speeds.
This indicates that the productivity of these ZNCs is less
sensitive to their size variations than for other systems studied
in the literature,13 where a higher yield was obtained for the
ZNC with smaller particles in a slurry polymerization process.
This may be due to the fact that the ZNCs were synthesized
using a different synthetic route and tested under different
polymerization conditions. In addition, as shown in Figure 8b,
all the ZNCs formed in this work are typical fractal objects.
Though the fractal dimension increases as the stirring speed
increases, the accessibility to the catalytic sites is always
excellent for fractal objects.
Figure 10a and 10b show the catalyst SD and the
corresponding polymer SD, respectively, estimated based on
the SASLS measurements and the Mie theory by assuming
spherical shape. It is seen that the size of the polymer particles
scales strictly with that of the corresponding catalyst particles.
Since the catalyst yield is constant, it follows that the larger the
catalyst size, the larger the polymer particle size. Second, the
shape of the polymer SD reflects the catalyst SD. The variance
of the distributions scales with the mean value, as the broadest
SD is obtained from the largest catalyst. If one defines a
replication factor, according to Casalini et al.,52 as the ratio of
the polymer to the catalyst particle size, Φ = dP50/dC50, from the
Φ values reported in Table 2, it appears that the polymer
particle size dP50 does not correlate linearly to the catalyst size
dC50. Similar behavior was also found by Philippaerts et al.13 A
reason could be the stronger tendency of breaking up for larger
catalysts during or at the very beginning of the polymerization
process, as already pointed out elsewhere.53 In this case, the
larger the initial catalyst particles, the larger the importance of
the phenomenon, leading to a nonlinear relation that should
lead to a maximum polymer particle size for given hydro-
Φ
dynamic conditions.47 A further reason for this nonlinear
14.6
correlation in size between catalyst particle and polymer
20.9
particle at equal productivity could be related to the differences
22.7
in the polymer porosity, related to the porosity of the catalyst
prepared at different stirring speeds.54

Figure 10. (a) SDs of the final ZNC particles prepared at an EADC feed rate of 10 mL/min and a stirring speed of 1400 rpm (squares), 900 rpm
(circles), and 1400 rpm (triangles). (b) Corresponding SD of the polyethylene particles.

894 DOI: 10.1021/acs.iecr.8b05296

Ind. Eng. Chem. Res. 2019, 58, 886−896
Industrial & Engineering Chemistry Research
■ CONCLUDING REMARKS
The decomposition of the hetero-bimetallic complex, Mg-
(OR)2·TiOR)4 (with R being ethyl and n-butyl groups), by
addition of EADC to form the Ziegler−Natta Catalyst (ZNC),
has been investigated by measuring both the liquid-phase
composition via ICP-OES and the solid particle size via
SASLS. The results from the liquid-phase characterization
show that the initial precipitation reaction involves a ligand
exchange with EADC leading to the decomposition of the
complex characterized by the complete precipitation of the
initially charged Mg in the form of MgCl2. In this stage, about
30% of the initially fed Ti moieties leave the liquid phase as
well. Further addition of EADC causes part of the Ti in the
solid phase to resolubilize, but the subsequent increase of
reactor temperature leads to the complete precipitation of all
Ti left in solution in the form of TiCl3 through its reduction.
The SASLS measurements during the formation process of
the ZNCs show that the size of the ZNC particles increases
with the addition of EADC up to the point where the liquid
phase does not contain soluble Mg anymore. It is
demonstrated that the particle growth is controlled not only
by the precipitation reactions but also by the aggregation-
breakage events. For this reason, the catalyst particle size can
be effectively controlled by changing the shear rate applied in
this initial stage. After this, the further reactions do not affect
the final particle size because the steady Ti complex adsorption
on the particle surface stabilizes the particles against
aggregation. Due to the observed close match in the shapes
of the SDs of the catalyst and polymer particles, it is concluded
that the quality of the final polymer particles can be controlled
by the shear rate applied in the growth stage of the ZNC
particle preparation process.
■
ASSOCIATED CONTENT
*
S Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/acs.iecr.8b05296.
Reactor setup; tables showing concentrations, ⟨Rg⟩, and
XRF-based composition (PDF)
■
AUTHOR INFORMATION
Corresponding Authors
*E-mail: hua.wu@chem.ethz.ch.
*E-mail: massimo.morbidelli@chem.ethz.ch.
ORCID
Antoine Klaue: 0000-0002-5119-1500
Hua Wu: 0000-0002-2805-4612
Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS
Financial support from the Swiss National Science Foundation
(Grant No. 200020_165917) is greatly acknowledged. The
authors would like to thank Dr. Anna Beltzung and Dr.
Baptiste Jaquet for support and fruitful discussions. For the
polymerization tests, the authors thank Erik Janssen for his
valuable assistance.
■
REFERENCES
(1) MarketResearchGroup. World Plastics Materials Demand 2015 by
Types; PlasticsEurope, 2015.
895
Article
(2) Jeremic, D., Polyethylene. In Ullmann’s Encyclopedia of Industrial
Chemistry; Wiley-VCH: Weinheim, 2014.
(3) Fawcett, E. W.; Gibson, R. O.; Perrin, M. W.; Paton, J. G.
Improvements in or relating to the polymerisation of ethylene.
GB471590A, 1936.
(4) Hogan, J. P.; Banks, R. L.; Lanning, W. C.; Clark, A.
Polymerization of light olefins over nickel oxide−silica-alumina. Ind.
Eng. Chem. 1955, 47 (4), 752−757.
(5) Ziegler, K.; Holzkamp, E.; Breil, H.; Martin, H. Das Mülheimer
Normaldruck Polyäthylen Verfahren. Angew. Chem. 1955, 67 (19−
20), 541−547.
(6) Böhm, L. L. The ethylene polymerization with Ziegler catalysts:
Fifty years after the discovery. Angew. Chem., Int. Ed. 2003, 42 (41),
5010−5030.
(7) D’Amore, M.; Credendino, R.; Budzelaar, P. H. M.; Causá, M.;
Busico, V. A periodic hybrid DFT approach (including dispersion) to
MgCl2-supported Ziegler−Natta catalysts − 1: TiCl4 adsorption on
MgCl2 crystal surfaces. J. Catal. 2012, 286, 103−110.
(8) Cossee, P. Ziegler-Natta catalysis I. Mechanism of polymer-
ization of α-olefins with Ziegler-Natta catalysts. J. Catal. 1964, 3 (1),
80−88.
(9) Galli, P.; Vecellio, G. Technology: Driving force behind
innovation and growth of polyolefins. Prog. Polym. Sci. 2001, 26
(8), 1287−1336.
(10) Floyd, S.; Heiskanen, T.; Taylor, T.; Mann, G.; Ray, W. H.
Polymerization of olefins through heterogeneous catalysis. VI. Effect
of particle heat and mass transfer on polymerization behavior and
polymer properties. J. Appl. Polym. Sci. 1987, 33 (4), 1021−1065.
(11) Kanellopoulos, V.; Dompazis, G.; Gustafsson, B.; Kiparissides,
C. Comprehensive analysis of single-particle growth in heterogeneous
olefin polymerization: The random-pore polymeric flow model. Ind.
Eng. Chem. Res. 2004, 43 (17), 5166−5180.
(12) Schnecko, H.; Dost, W.; Kern, W. Olymerisation von Propylen
mit Al haltigen und Al freien Titantrichloriden verschiedener
Korngrößen. 8. Mitt.*. Zur Polymerisation mit metallorganischen
Mischkatalysatoren*. Makromol. Chem. 1969, 121 (1), 159−173.
(13) Philippaerts, A.; Ensinck, R.; Baulu, N.; Cordier, A.; Woike, K.;
Berthoud, R.; De Cremer, G.; Severn, J. R. Influence of the particle
size of the MgCl2 support on the performance of Ziegler catalysts in
the polymerization of ethylene to ultra high molecular weight
polyethylene and the resulting polymer properties. J. Polym. Sci., Part
A: Polym. Chem. 2017, 55 (16), 2679−2690.
(14) Dall’Occo, T.; Zucchini, U.; Cuffiani, I. Industrial Aspects of
the Production of Catalysts for Ethylene Polymerization. In Transition
Metals and Organometallics as Catalysts for Olefin Polymerization;
Kaminsky, W., Sinn, H., Eds.; Springer-Verlag: Berlin, 1988; pp 211−
222.
(15) Bashir, M. A.; Monteil, V.; Boisson, C.; McKenna, T. F. L.
Experimental proof of the existence of mass-transfer resistance during
early stages of ethylene polymerization with silica supported
metallocene/MAO catalysts. AIChE J. 2017, 63 (10), 4476−4490.
(16) Karol, F. J. Studies with high activity catalysts for olefin
polymerization. Catal. Rev.: Sci. Eng. 1984, 26 (3−4), 557−595.
(17) Chumachenko, N. N.; Zakharov, V. A.; Bukatov, G. D.;
Sergeev, S. A. A study of the formation process of titanium-
magneisum catalyst for propylene polymerization. Appl. Catal., A
2014, 469, 512−516.
(18) Nikitin, V. A.; Mikenas, T. B.; Zakharov, V. A. Catalyst
preparation process and ethylene polymerization and ethylene-alpha-
olefin copolymerization processes utilizing this catalyst.
RU2257263C1, 2005.
(19) Mikenas, T. B.; Koshevoy, E. I.; Cherepanova, S. V.; Zakharov,
V. A. Study of the composition and morphology of new modifications
of titanium magnesium catalysts with improved properties in ethylene
polymerization. J. Polym. Sci., Part A: Polym. Chem. 2016, 54 (16),
2545−2558.
(20) Redzic, E.; Garoff, T.; Mardare, C. C.; List, M.; Hesser, G.;
Mayrhofer, L.; Hassel, A. W.; Paulik, C. Heterogeneous Ziegler-natta
DOI: 10.1021/acs.iecr.8b05296
Ind. Eng. Chem. Res. 2019, 58, 886−896
Industrial & Engineering Chemistry Research
catalysts with various sizes of MgCl2 crystallites: Synthesis and
characterization. Iran. Polym. J. 2016, 25, 321−331.
(21) Berger, E.; Derroitte, J.-L. Polymerization of olefins.
US3901863A, 1975.
(22) Yanovskaya, M. I.; Kotova, N. M.; Golubko, N. V.; Turova, N.
Y. Reactions of magnesium and titanium alkoxides. Preparation and
characterization of alkoxy-derived magnesium titanate powders and
ceramics. J. Sol-Gel Sci. Technol. 1998, 11, 23−29.
(23) Turova, N. Y.; Turevskaya, E. P.; Kessler, V. G.; Yanovskaya, M.
I. The chemistry of metal alkoxides; Kluwer Academic Publishers:
Boston, 2002.
(24) Kucher, M. Vom Keim zum Kristall - Ü ber die Partikelbildung bei
der Fällung schwerlöslicher Feststoffe; Universität Karlsruhe, 2008.
(25) Söhnel, O.; Mullin, J. W.; Jones, A. G. Crystallization and
agglomeration kinetics in the batch precipitation of strontium
molybdate. Ind. Eng. Chem. Res. 1988, 27 (4), 1721−1728.
(26) Ilievski, D.; White, E. T. Agglomeration during Precipitation:
Agglomeration mechanism identification for Al(OH)3 crystals in
stirred caustic solutions. Chem. Eng. Sci. 1994, 49 (19), 3227−3239.
(27) Mumtaz, H. S.; Hounslow, M. J.; Seaton, N. A.; Paterson, W. R.
Orthokinetic aggregation during precipitation: A computational
model for calcium oxalate monohydrate. Chem. Eng. Res. Des. 1997,
75 (2), 152−159.
(28) Friederichs, N. H.; Bilda, D. H.; Gerlofsma, R. Process for the
production of ultra high molecular weight polyethylene.
WO2009127410, 20090414, 2009.
(29) Armarego, W. L. F.; Chai, C. L. L. Purification of inorganic and
metal-organic chemicals. In Purification of Laboratory Chemicals, 6th
ed.;Armarego, W. L. F., Chai, C. L. L., Eds.; Butterworth-Heinemann:
Oxford, 2009; Chapter 5, pp 445−576.
(30) Bichinho, K. M.; Pires, G. P.; Dos Santos, J. H. Z.; Wolf, C. R.
Determination of Mg and Ti in Ziegler-Natta catalysts by x-ray
fluorescence spectrometry. X-Ray Spectrom. 2005, 34 (2), 131−134.
(31) Sorensen, C. M. Light scattering by fractal aggregates: A review.
Aerosol Sci. Technol. 2001, 35 (2), 648−687.
(32) Tarazona, A.; Koglin, E.; Buda, F.; Coussens, B. B.; Renkema,
J.; VanHeel, S.; Meier, R. J. Structure and stability of aluminum alkyl
cocatalysts in Ziegler-Natta catalysis. J. Phys. Chem. B 1997, 101 (22),
4370−4378.
(33) Cullinane, N. M.; Chard, S. J.; Price, G. F.; Millward, B. B.
Organic derivaties of titanium. III. Chlorotitanium alkoxides. J. Appl.
Chem. 1952, 2 (5), 250−251.
(34) Busico, V.; Causa, M.; Cipullo, R.; Credendino, R.; Cutillo, F.;
Friederichs, N. N.; Lamanna, R.; Segre, A.; Van Axel Castelli, V.
Periodic DFT and high-resolution-magic-angle-spinning (HR-MAS)
1H NMR Investigation of the active surfaces of MgCl2-supported
Ziegler-Natta catalysts. the MgCl2 matrix. J. Phys. Chem. C 2008, 112
(4), 1081−1089.
(35) Copéret, C.; Allouche, F.; Chan, K. W.; Conley, M. P.; Delley,
M. F.; Fedorov, A.; Moroz, I. B.; Mougel, V.; Pucino, M.; Searles, K.;
Yamamoto, K.; Zhizhko, P. A. Bridging the gap between industrial and
well-defined supported catalysts. Angew. Chem., Int. Ed. 2018, 57,
6398−6440.
(36) Cucinella, S.; Mazzei, A.; Marconi, W.; Busetto, C. Reactions
between AlRCl2 and Ti(OR)4 and activity in diolefin polymerization.
J. Macromol. Sci., Chem. 1970, 4 (7), 1549−1561.
(37) Thoma, S. G.; Ciftcioglu, M.; Smith, D. M. Determination of
agglomerate strength distributions Part 3. Application to titania
processing. Powder Technol. 1991, 68 (1), 71−78.
(38) Liew, T. L.; Barrick, J. P.; Hounslow, M. J. A micro-mechanical
model for the rate of aggregation during precipitation from solution.
Chem. Eng. Technol. 2003, 26 (3), 282−285.
(39) Hartel, R. W.; Gottung, B. E.; Randolph, A. D.; Drach, G. W.
Mechanisms and kinetic modeling of calcium oxalate crystal
aggregation in a urinelike liquor. Part I: Mechanisms. AIChE J.
1986, 32 (7), 1176−1185.
(40) Wójcik, J. A.; Jones, A. G. Experimental investigation into
dynamics and stability of continuous MSMPR agglomerative
896
Article
precipitation of CaCO3 crystals. Chem. Eng. Res. Des. 1997, 75 (2),
113−118.
(41) Spicer, P. T.; Pratsinis, S. E. Coagulation and fragmentation:
Universal steady-state particle-size distribution. AIChE J. 1996, 42
(6), 1612−1620.
(42) Soos, M.; Moussa, A. S.; Ehrl, L.; Sefcik, J.; Wu, H.; Morbidelli,
M. Effect of shear rate on aggregate size and morphology investigated
under turbulent conditions in stirred tank. J. Colloid Interface Sci.
2008, 319 (2), 577−589.
(43) Bouyer, D.; Liné, A.; Do-Quang, Z. Experimental analysis of
floc size distribution under different hydrodynamics in a mixing tank.
AIChE J. 2004, 50 (9), 2064−2081.
(44) Coufort, C.; Bouyer, D.; Liné, A. Flocculation related to local
hydrodynamics in a Taylor-Couette reactor and in a jar. Chem. Eng.
Sci. 2005, 60, 2179−2192.
(45) Kobayashi, M.; Adachi, Y.; Ooi, S. Breakup of fractal flocs in a
turbulent flow. Langmuir 1999, 15 (13), 4351−4356.
(46) Coufort, C.; Liné, A. Forces on spherical particles in terms of
upstream flow characteristics. Chem. Eng. Res. Des. 2003, 81 (9),
1206−1211.
(47) Bäbler, M. U.; Morbidelli, M.; Baldyga, J. Modelling the
breakup of solid aggregates in turbulent flow. J. Fluid Mech. 2008, 612,
261−289.
(48) Hassan, I. T. M.; Robinson, C. W. Stirred-tank mechanical
power requirement and gas holdup in aerated aqueous phases. AIChE
J. 1977, 23 (1), 48−56.
(49) Nienow, A. Q.; Miles, D. Impeller power numbers in closed
vessels. Ind. Eng. Chem. Process Des. Dev. 1971, 10 (1), 41−43.
(50) Slund, B. L. A.; Rasmuson, A. C. Semibatch reaction
crystallization of benzoic acid. AIChE J. 1992, 38 (3), 328−342.
(51) Brunsteiner, M.; Jones, A. G.; Pratola, F.; Price, S. L.; Simons,
S. J. R. Toward a molecular understanding of crystal agglomeration.
Cryst. Growth Des. 2005, 5 (1), 3−16.
(52) Casalini, T.; Visscher, F.; Janssen, E.; Bertola, F.; Storti, G.;
Morbidelli, M. Modeling of polyolefin polymerization in semibatch
slurry reactors: Experiments and simulations. Macromol. React. Eng.
2017, 11 (1), 1600036.
(53) Casalini, T.; Visscher, F.; Tamaddoni, M.; Friederichs, N.;
Bertola, F.; Storti, G.; Morbidelli, M. The effect of residence time
distribution on the slurry-phase catalytic ethylene polymerization: An
experimental and computational study. Macromol. React. Eng. 2018,
12 (3), 1700058.
(54) Alizadeh, A.; McKenna, T. F. L. Particle growth during the
polymerization of olefins on supported catalysts. Part 2: Current
experimental understanding and modeling progresses on particle
fragmentation, growth, and morphology development. Macromol.
React. Eng. 2018, 12 (1), 1700027.
DOI: 10.1021/acs.iecr.8b05296
Ind. Eng. Chem. Res. 2019, 58, 886−896