MODELING AND SIMULATION OF POLYMERIZATION OF LACTIDE TO

POLYLACTIC ACID AND CO-POLYMERS OF POLYLACTIC ACID

USING HIGH VISCOSITY KNEADER REACTORS
Boyd T. Safrit, LIST USA, Inc., Charlotte, NC
George E. Schlager, LIST USA, Inc., Charlotte, NC
Ziang Li, University of Texas at Austin, Austin, TX
Abstract
Polymerization of lactide to polylactic acid (PLA)
can be performed using conventional reactor technology
such as stirred tank reactors, but the conversion and/or
final molecular weight may have to be controlled to a
lower level. At higher conversion and/or molecular
weight, the reaction mass will become very viscous,
which limits the ability of conventional reactor technology
to provide adequate mixing, minimize mass transfer
effects on reaction kinetics, remove exothermic heat of
reaction and ensure proper heat transfer in order to
eliminate hotspots/thermal degradation.
Kneader reactor technology has been used over 60
years in many high viscosity applications such as
reactions and polymerization, devolatilization, and drying.
This technology can handle the higher conversion and
molecular weight polymerizations of lactide and other
copolymers of lactide, while also providing the heat
transfer required for proper temperature control. Using
model kinetics and rheology data, a study was performed
that shows the capability of kneader reactor technology
for lactide polymerizations as well as other copolymers.
Kneader reactor technology can also be used to remove
the unconverted monomers from the polymer and
expected results from the continuous operation of a
polymerizer and finisher will be shown.

Introduction
Stirred tank reactors and other conventional reactor
technologies have been successfully used for many years
in the polymerization and copolymerization of a variety of
monomers. It is quite common in such reactions that the
viscosity of the reacting mass will raise with increasing
polymer molecular weight and that they are exothermic.
These two characteristics pose challenges for
conventional reactor technology. Stirred tank reactors, for
example, are limited in their ability to maintain a well
agitated and mixed reacting mass at higher viscosities.
Reaction rates may therefore be reduced by mass transfer
and diffusional limitations.
Poor agitation in stirred tank reactors may also
contribute to low flow velocities near the reactor wall,
resulting in low film heat transfer coefficients (reacting
mass to inner reactor wall). Since the exothermic heat of

reaction in stirred tank reactors is typically removed via

contact heat transfer with the reactor wall, the product
temperature may increase uncontrollably, resulting in a
run-away reaction or degraded product [1].
Kneader reactor technology is well suited for higher
viscosity
processing,
including
exothermic
polymerizations.
Kneading and mixing of viscous
materials is provided by the interaction of the static and/or
dynamic elements mounted on the shaft(s). Kneader
reactors provide a large amount of heat transfer area from
the jacketed reactor casing and the shafts and elements on
the shafts are also actively heated/cooled with heat
transfer media. Many products have a tendency to wrap
around the shaft(s) of processing equipment and may lose
contact with the surface area on the reactor casing. But
kneader reactor technology has the additional agitated
surface area on the shaft and shaft elements that can
continue to provide heat transfer capabilities [2].
The polymerization of lactic acid via the ring opening
reaction of lactide has become an important and
frequently studied process in the last several years. New
applications for polylactic acid (PLA) appear every day
with ever increasing physical properties requirements
[13]. Many medical applications, for example, require a
higher molecular weight polymer than typically used for
standard packaging applications. The monomer, lactide,
is a low viscosity liquid at typical reaction temperatures
but as the reaction proceeds and molecular weight
increases, the viscosity of the reaction mass also
increases. A study was performed on the polymerization
of PLA with an emphasis on heat and mass transfer
limitations encountered at higher molecular weight
polymerizations. Using previously published kinetic
models, a simulation model with simultaneous kinetic,
degradation, molecular weight, viscosity, mass transfer,
and heat transfer effects was developed. This model was
used to simulate and compare the performance of a stirred
tank (STR) and a batch kneader reactor. The same model
was also capable of fully simulating continuous versions
of these two reactor types.

Experimental
Kinetics
Witske (1996) developed a reversible kinetic model
for the polymerization of lactide, catalyzed by tin octoate

in bulk. Equations 1-3 show the general model used in

this study:

M (t ) = M eq + ( M 0 M eq )e( K P It )
K p = Ae

M eq = e

H p S p

R
RT

Ea

RT

(1)

mix = mon +

(3)

(4)

Viscosity
In order to model the reaction mass viscosity, the
effects of monomer content, molecular weight,
temperature, and shear rate had to be investigated.
Dorgan et al. (2004) investigated the melt flow properties
of PLA, including developing a zero-shear viscosity
model based on the Carreau-Yasuda model. Using PLA
zero-shear viscosity measurements taken at 180C,
Dorgan et al. proposed the following zero-shear viscosity
model:

M p 1
0 = 10 M w 1 + w
M c

(5)

Conrad (2009) measured PLA viscosity over a wide

range of shear rates at 180C and fit the data to the
following power law viscosity model:

PLA, pure = 0 n 1

1+ e

x0 x
w

(7)

The data of Vos et al was used to get the best fit of

equation 7, resulting in values of x0=70% and w=6%.

The effect of initiators was not included in this

model. Mantourlias et al. (2008) investigated the effect of
hydroxyl-containing initiator (R-OH) concentration and
reaction temperature on yield. Vos et al. (2009) also show
that the polymerization of lactide in the presence of an ROH initiator is much faster than without the use of an
initiator. It was assumed for our model that the presence
of R-OH had an effect on the activation energy, Ea, only.
Using the data of Mantourlias et a.l (2008), we found that
the effect of different R-OH initators was small on the
activation energy. However, the data of Vos es al. (2009)
show that the pre-exponential constant may be affected by
the choice of R-OH initiator. For our model, we used the
Vos et al data to determine the pre-exponential constant
and the Mantourlias et al data to determine the activation
energy.

PLA, pure mon

(2)

Taking the derivative of Equation 1 and rearranging gives

the basic dynamic kinetic model used in this study:

dM (t )
= K P I ( M Meq)
dt

Equation 7 is a common mixture viscosity model and

was used to model the mixture of monomer (lactide) and
PLA:

(6)

Molecular Weight and Degradation

Aoyagi et al. (2002) modeled the thermal
decompositoin of PLA using a liner relationship between
the degree of polymerization (DP) and time, which is
shown in Equation 8:

1
1
=
+ Kd t
DP DP0
3 * 10 9
Kd =
e
hr

103, 000
mol

RT

(8)

(9)

We assumed for this study that the change in

molecular weight during the polymerization was linear in
the concentration change with time.
Using this
assumption of molecular weight change, the definition of
DP, and Equation 8, the change in PLA molecular weight
as a function of time due to degradation was:

dM w (t )
= K d MWmon
dt

(10)

Equation 10 was combined with Equation 4 to give the

overall dynamic kinetic polymerization model.
Mass Transfer
During most polymerizations, the reaction mass
viscosity increases. At higher viscosities, most chemical
reactors run into mass transfer limitations as these reactors
are unable to provide adequate mixing of the viscous
material and therefore the reaction kinetics become
negatively impacted. At these higher viscosities, the
effective reaction rate approaches the diffusional mass
transfer rate. We assumed a resistance based model for
mass transfer in order to estimate an effective kinetic rate
constant, Keff:

1
1
1
=
+
K eff K p K diff

(11)

Keff would replace Kp in Equation 1. In order to estimate

the diffusional mass transfer coefficient, Kdiff, we assumed
as many models do (e.g. Stokes-Einstein and WilkeChang models) that the mass transfer coefficent was
inversly proportional to the reaction mass viscosity.
Using the data from [5] at 50ppm catalyst concentration,
we fit a correlation for the mass transfer coefficient:

K diff =

250Pa s hr 1

mix

(12)

This data was chosen in order to minimize any

temperature effects due to heat transfer limitations that
may occur at higher catalyst loadings. While mass
transfer coefficients are largely dependant on the
processing equipment in use, in this study we assumed
that the mass transfer coefficient was the same for all
reactor types.
Heat Transfer
For the same reasons mass transfer can be an issue in
polymerizations, heat transfer must also be included. In a
STR for example, as the reaction mass viscosity increases,
the ability to maintain a well mixed reactor goes down,
which reduces the heat transfer rate at the walls of the
reactor. Localized hot spots may develop in the reaction
mass, leading to degradation and possible run away
reactions.
It was assumed that the flow regimes within either the
modeled STR or kneader reactor was in the laminar range.
We used the following heat transfer coefficient
correlation, developed for agitated vessels, for the inside
reactor wall [9]:

hinside

1
k
b
= a (N Re ) (N Pr )3 bulk
D
wall

(13)

NRe and NPr were calculated in the typical manner for the
STR. For the kneader reactor, proprietary calculations
were used for these parameters. The heat transfer area for
the STR was assumed to be the entire wall of the reactor.
Kneader reactor technology is never operated 100% full
and is normally operated with a fill level of 50-70%. The
heat transfer area assumed for the kneader reactor was the
total internal heat transfer surface area of the kneader
reactor scaled by the assumed fill level.

Results and Discussion

All model characteristics previously detailed were
incorporated into a dynamic model so that both batch and
continuous reactors could be simulated. The results of a
batch stirred tank reactor and a batch kneader reactor
follow. We assumed the data of [5] was taken on a small,
laboratory stirred flask of 1 liter with a constant heat bath
temperature of 180C.
Small Batch Stirred Tank Reactor
Figure 1-A shows the experimental conversion data
of [5] and the results of the simulation model (no mass
transfer or temperature effects). Figure 1-B shows the
same data but rearranging Equation 1 so that the kinetic
model would form a straight line when plotted against
time. All results assume a constant reaction temperature
of 180C. The data and model agree well for the case of
50ppm catalyst loading. In this case, the polymerization

proceeded slowly, which most likely allowed the

experimental reactor (assumed to be a small laboratory
stirred flask) of [5] to maintain a constant reaction
temperature of 180C. After a polymerization time of 3
hours, there is a deviation in the data and model. We
believe that at this point the reaction mass viscosity has
increased so that mass transfer limitations in the
experimental reactor are negatively impacting the reaction
rate, especially at higher conversions (viscosity). At
higher catalyst loadings, we see conversions higher than
expected during the early part of the polymerization. It is
possible that the experimental reactor could not remove
the exothermic heat of reaction fast enough due to the
higher rate of reactions. In this case, the reaction mass
temperature would increase and the reaction rate would
therefore increase. Figure 2-A shows the conversion
results of the model when temperature and mass transfer
effects were included.
For the 50ppm case, the
conversions are over-predicted, however, for higher
catalyst loadings, the conversions are under-predicted.
Figure 2-B shows the predicted temperature profiles for
several of the different catalyst loadings. The shape and
maximal value of the temperature profiles was very
sensitive to the heat transfer coefficient calculation. The
data of [5] did not include any temperature information so
determining better heat transfer coefficients was not
possible. One other interesting outcome shown on Figure
2-B was the saddle in the temperature profile. The
reaction mass was assumed to be at 180C at t=0 and the
heat transfer medium was also assumed to be 180C.
When the reaction started, the rate of heat transfer is very
low due to a small temperature difference. There was a lag
between the heat transfer out of the system and the heat
added to the system via the exothermic heat of reaction.
The influence of increasing mixture viscosity on the heat
transfer coeffcient also contributed to the saddle in the
temperature profile.
Small Batch Kneader Reactor
A simulation of a batch kneader reactor was carried
out and the size of the reactor was chosen so that the mass
of polymer (~1.2 kg) was the same as the small batch
STR. As can be seen in Figure 3 (A&B), the results of the
kneader reactor are similar to the batch STR. The
maximum temperature for each catalyst loading was less
in the kneader reactor than the STR due to the fact that the
surface to volume ratio of the kneader reactor was
approximately 74m2/m3 compared to the STR of 46. The
heat transfer coefficients were also higher in the kneader
reactor since the kneader reactor operates at a higher shear
rate than the STR. The effect of shear rate on mass
transfer limitations was also very important at higher
viscosities and PLA molecular weights as the kneader
reactor was not affected as much as the STR.
Commercial Scale

The model was used to simulate a commercial PLA

continuous polymerization process of 10,000 ton/year
using a catalyst concentration of 100ppm. The targeted
conversion of lactide to PLA was 95% with a molecular
weight of 150,000 g/mol. A heat transfer medium
temperature of 180C was also assumed. The results of
the continuous stirred tank reactor (CSTR) can be seen in
Table 1. The same process assumptions were used for the
continuous kneader reactor model and are shown in Table
1. The CSTR reactor volume and the kneader reactor
volume were chosen to allow the reaction to reach the
specified conversion. It is interesting to note that the
CSTR required to complete the specified reaction was
125,000 liters, while the required kneader size is only
2,900 liters. Figure 4 shows the temperature profile for
the kneader reactor as a function of axial position in the
reactor. The maximum temperature is 194C, which is
similar to the CSTR. However, the polymer would be at
this temperature for only a fraction of the total 1.7 hr
residence time in the kneader reactor. The polymer in the
CSTR would be subjected to 195C for the entire
residence time of 3.8 hrs. Thermal degradation reactions
are a function of both time and temperature, meaning the
CSTR would be much more susceptible to polymer
degradation.
Table 1: Commercial Reactor Model Results
Volume
Surface area
Reactor temperature
Reactor viscosity
Residence time

L
m2
C
Pa*s
hr

CSTR
125,000
117
195
300
3.8

Kneader
2,900
43
194 (max)
18 (max)
1.7

Table 1 also shows the viscosity for both the CSTR

and the kneader reactor (maximum value.) The difference
in the viscosities of the two reactor types is due to the
shear rate dependancy. The kneader reactor operated at a
higher shear rate, which lowered the effective reaction
mass viscosity. At a viscosity of 300 Pa*s, the ability of
CSTR technology to adequately maintain a well mixed
environment will be limited. This could possibly lead to
local hot spots, degraded polymer, etc. For higher target
molecular weight products and possibly co-polymers of
lactide and other monomers, high viscosity and heat/mass
transfer limitations may require techonology like kneader
reactors that are more suited to high viscosity processing.
Most commercially available PLA has a specification
of residual lactide of <0.5wt%. For the commercial
polymerization process, a devolatilization step must be
added in order to remove the residual monomer (lactide)
down to an acceptable level. As pointed in [11] and [12],
kneader reactor technology is also very suitable to the
devolatilization of temperature sensitive polymers. Figure
5 shows a typical two-stage kneader reactor process for
the polymerization and devolatilization and PLA. As
shown in [12], lactide levels in the final PLA polymer

could be much lower than the typical commercial

specifications of < 0.5wt%.

Conclusions
A model was developed for the polymerization of
lactide to polylactic acid. This model included the effects
of kinetics, heat and mass transfer, viscosity, molecular
weight, and reactor technology (stirred tank reactors and
kneader reactors.) Lactide conversion data of [5] showed
that as catalyst concentration was increased, heat and
mass transfer limitations of the lab reactor negatively
impacted the polymerization process. A kneader reactor
was simulated and these limitations were minimized.
A commercial scale continuous process was also
simulated. A continuous stirred tank reactor required for
this throughput was very large and resulted in a higher
product temperature over the entire residence time, which
could lead to high polymer thermal degradation. The
required kneader reactor however was much smaller in
size and minimized the heat history of the polymer. The
commercial kneader reactor polymerizer could also be
matched with a kneader reactor devolatilizer for removing
the residual monomer (lactide.)

Nomenclature
a
b
A
B
D
DP
DP0
Ea
Hp
hinside
I
k
Keff
Kdiff
KP
Kd
m
MW
MWmon
M(t)
Meq
M0
Mw
Mc
n
NRe
NPr
p

heat transfer correlation coefficent = 0.36 for

STR and 0.74 for kneader reactor
heat transfer correlation coefficent = 2/3
pre-exponential constant
zero shear rate constant = -4.87
reactor diameter
degree of polymerization
initial degree of polymerization at t=0
activation energy = 76,830 J/mol
enthlapy of polymerization = -23,300 J/mol
reactor wall heat transfer coefficient
catalyst concentration in mol%
reaction mass thermal conductivity = 0.289
W/m-K
effective kinetic rate constant
diffusion rate constant
propagation rate constant in catalyst mol%-1hr-1
degradation reaction rate constant
heat transfer correlation coefficient = 0.21
polymer molecular weight
monomer molecular weight = 144.13 g/mol
monomer concentration at time (t)
equilibrium monomer concentration
initial monomer concentration
polymer molecular weight in kg/kg-mole
critical entanglement molecular weight = 9211
kg/kg-mole
power law index = 0.23 for PLA
Reynolds number
Prandtl number
chain packing length = 3.5

Sp
T
x
xo
w

0
PLA,pure
mon
mix
bulk
wall

entropy of polymerization = -22 J/mol-K

temperature
weight fraction of PLA in the mixture
cross over weight fraction = 0.7
band width weight fraction = 0.06
apparent shear rate in sec-1
zero shear viscosity in Pa*sec
pure PLA viscosity at the apparent shear rate
monomer viscosity
mixture viscosity
reaction mass bulk viscosity
reaction mass viscosity at reactor wall

References
1.

P. A. Fleury, Bulk Polymerization of Methyl

Methacrylate in a Kneader Reactor, Presented at
ANTEC 2006 (2006).
2. B. T. Safrit, Prediction of Continuous Kneader
Reactor Processes From Batch Data, ANTEC 2007
Technical Paper # ANTEC-0425-2007 (2007).
3. D. Witzke, Properties and Engineering of
Polylactide Polymers, Thesis, Michigan State
University (1996).
4. T. Mantourlias, A. Seretis, K. Karidi, S. Parouti, C.
Kiparissides, Lactic Acid Based Functional
Copolymers, Bioproducts 2nd Annual Meeting
(2008).
5. S. de Vos, P. Jansen, Industrial-Scale PLA
Production from PURAC Lactides, 8th World
Congress of Chemical Engineering (2009).
6. J. Dorgan, D. Knauss, J. Janzen, J.Randal, B. Braun,
O. Gruber Perfecting PLA Properties: A Successful
University Collaboration, The Global Plastics
Environmental Conference (2004).
7. NatureWorks PLA Injection Molding Guide for
3051D.
8. J. Conrad, The Rheology, Degradation, Processing,
and Characterization of Renewable Resource
Polymers, Thesis, Clemson University (2009).
9. C. Geankoplis, Transport Processes and Unit
Operations (1983).
10. Y. Aoyagi, K. Yamashita, and Y. Doi, Thermal
degradation of poly[(R)-3-hydroxybutyrate], poly[caprolactone], and poly[(S)-lactide], Polymer
Degradation and Stability, Vol 76, pp 53-59 (2002).
11. P. A. Fleury, D. U. Witte, H. Schildknecht,
Comparison of Devolatilization Technologies for
Viscous Polymers, ANTEC 2005 Technical Paper
#100655 (2005).
12. B. T. Safrit, A. E. Diener, Kneader Technology for
the Direct Devolazliation of Temperature Sensitive
Elastomers, Rubber Division of the American
Chemical Society, Inc. 180th Technical Meeting,
Paper #13 (2011).

13. M. H. Naitove, Conference Report: Bioplastics Are

Breaking Out of Their 'Green' Niche, Plastics
Technology (April 2012).

Key Words
kneader reactor, polymerization, devolatilization, high
viscosity, heat transfer, PLA, lactide, polylactic acid, bulk
polymerization, mass transfer.

Figure 1 (A&B): Monomer Conversion (Experimental and Model)

with No Temperature of Mass Transfer Effects

Figure 2 (A&B): STR Monomer Conversion (Experimental and Model)

and Model Temperature Profiles with Temperature of Mass Transfer Effects

Figure 3 (A&B): Kneader Reactor Monomer Conversion (Experimental and Model)

and Model Temperature Profiles with Temperature of Mass Transfer Effects

Figure 4: Model Results for Commercial Kneader Reactor

Figure 5: Two-Stage Kneader Reactor PLA Process