SAND2016-5741J

THE KINETICS OF POLYURETHANE STRUCTURAL FOAM: FOAMING

AND POLYMERIZATION

Rekha Rao, Lisa Mondy, Kevin Long, Mathew Celina, Christine Roberts, Melissa
Soehnel, Nicholas Wyatt, Victor Brunini
Sandia National Laboratories
Albuquerque, New Mexico 87185 and Livermore, California 94550

We are developing kinetic models to understand the manufacturing of polymeric

foams, which evolve from low viscosity Newtonian liquids, to bubbly liquids, finally
producing solid foam. Closed-form kinetics are formulated and parameterized for
PMDI-10, a fast curing polyurethane, including polymerization and foaming. PMDI-
10 is chemically blown, where water and isocyanate react to form carbon dioxide. The
isocyanate reacts with polyol in a competing reaction, producing polymer. Our
approach is unique, though it builds on our previous work and the polymerization
literature. This kinetic model follows a simplified mathematical formalism that
decouples foaming and curing, including an evolving glass transition temperature to
represent vitrification. This approach is based on IR, DSC, and volume evolution data,
where we observed that the isocyanate is always in excess and does not affect the
kinetics. The kinetics are suitable for implementation into a computational fluid
dynamics framework, which will be explored in subsequent papers.

INTRODUCTION
Foams are multiphase materials with a compressible gas dispersed as bubbles in a continuous
phase. We are interested in understanding the manufacturing of chemically blown polymeric
foams, which evolve from low viscosity Newtonian liquids to bubbly liquids via gas generation,
finally producing solid foam through polymerization. In these foams, bubble microstructure can
affect macroscopic properties such as material strength and density. Characterizing foam is often
difficult since foams are opaque and have microstructure that may vary spatially.

We are developing closed-form kinetic models to elucidate the expansion and dynamic filling
process of a moderate density polyurethane structural foam, polymeric methylene diphenyl
diisocyanate called PMDI-10. Here the “10” indicates that the density of the free rise foam will be
roughly 10 lb/ft3 (0.16g/cm3) though it is often over-packed in a mold to twice (or more) of its free
rise density to reach the density of interest. This polyurethane has a fast catalyst, such that filling
of the mold and polymerization occur simultaneously. Here we focus on the kinetics of the
reactions of polymerization and foaming. PMDI-10 is a chemically blown foam, where carbon
dioxide is produced via the reaction of water, the blowing agent, and isocyanate. Simultaneously,
isocyanate reacts with polyol in a competing reaction, producing a glassy polymer. This work is
useful for describing chemically blown foams such as PMDI-10 that have a short pot-life, large
exotherms, and that may vitrify during cure. When coupled with computational fluid dynamics
simulation, the filling process can be optimized to reduce defects.

1
Examples of a free rise sample of PMDI and foam that was injected into a clear mold are shown
in Figure 1. The image on the left is of foam mixed in a cup. It is apparent that the material has
significant solid-like behavior while still foaming and rising. The fast kinetics can lead to
difficulties in characterizing the changing properties. The foam injection defects can easily be seen
in the clear mold such as incomplete filling, bubble coarsening due to exotherms, microstructural
and density variations, and knit lines.

Figure 1. PMDI has a short pot-life: understanding the kinetics of foam

formation can help reduce defects and improve filling process. The picture
on the left is a free rise foam and the one on the right is idealized structural
foam part.

Polyurethane curing has been studied by a number of researchers. Lipshitz and Macosko [1]
examined fast, highly exothermic polymerization of a thermosetting liquid polyurethane system
using a home-made system to measure the adiabatic temperature rise. IR measurements were used
to confirm that the reaction went to completion. They commented that it can be difficult to measure
the steep slopes of the temperature rise curves and thermocouple response time can limit the
usefulness of the method. A fairly large sample is needed and the high temperatures produced by
the exothermic reactions can cause degradation of the material with undesirable reactions.
Hernández-Sánchez and Vázquez-Torres [2] used differential scanning calorimetry (DSC) to
examine polymerization kinetics for a 2,6-tolylene diisocyanate and diethylene glycol system with
the diethylene glycol in excess. Unfortunately, DSC as used in that study is too slow for our fast
reactions; although we use DSC data synergistically to provide an independent test of our model.
Chern [3] developed a model that allowed reactions to be diffusion-controlled after gelation and
explained that the glass transition temperature Tg increases throughout the curing reaction. He
used the Fourier-transform, infrared spectroscopy, experimental data of Yang and Lee [4] to
validate his model. Prochazka et al. [5] used rheology measurements to monitor gelation of a
polyurethane system and investigate the influence of the stoichiometric ratio. However, in their
system, they did not mention the possibility of vitrification, though they did see unexplained
frequency dependence of the moduli. Chemorheological studies of a polyurethane by Rochery and
Lam [6] led to the consideration of vitrification, when the evolving glass transition temperature
becomes equal to the curing temperature and a secondary transition occurs that dramatically slows
down the reaction rate. Dimier and coworkers [7] focused on the competing effects of temperature
2
increase, which initially lowers the viscosity of the mixture but then increases the rate of chemical
cross-linking and, thus the viscosity. Their system has reactions slow enough to use DSC to follow
the kinetics. They described the reaction with condensation chemistry [8,9], so that several
simultaneous reactions can be represented by one global reaction for modeling purposes. They
then assumed the isoconversion principle that states that the reaction rate at a set extent of
conversion is a function of temperature through an Arrhenius law. We will also use this approach
for the polymerization reaction. However, Dimier et. al did not see evidence of vitrification.
Verhoeven and coworkers [10] compared three different methods to measure kinetics: Adiabatic
temperature rise, measurement kneader, and high-temperature measurements. Complexities in
their two polyurethane systems showed up, unlike the Dimier study. In one system, there was a
diffusion limited reaction at low conversions, and the three methods yielded three different kinetics
equations. Papadopoulos and colleagues [11] reported on DSC and rheological measurements in
their investigation of the reaction of PMDI and glycerol. They noted a complex cure process with
primary and secondary hydroxyl reactivity and observed incomplete cure with evidence of
vitrification. More recently, Chiachiarelli et al. [12] distinguished vitrification from gelation
during cure, also using DSC combined with rheology measurements. However, to do this they had
to extend the pot-life of the system using an inhibitor. Their model included the concept of a
maximum conversion that was a function of the temperature and was directly related to
vitrification.

The simultaneous curing and foaming of polyurethane has also been studied by a more limited
number of researchers. Marciano et al. [13] focused on the curing kinetics of a physically blown
(by boiling a chlorofluorocarbon) polyurethane foam using the adiabatic temperature rise method.
Niyogi et al. [14] used Marciano’s kinetics to develop a model of the bubble-size distribution in a
similar physically blown foam. However, our foam system is chemically blown, as described
above. Saha and colleagues [15] studied the cure kinetics for rigid, chemically blown foams
similar to those of interest here and compared it to non-foaming solid polyurethane, but did not
include vitrification. They combined an autocatalytic version of the Kamal model for
condensation chemistry [8,9] with rheology measurements and estimated the degree of cure from
the ratio of instantaneous shear modulus to final shear modulus. They found that the reaction rate
constants followed an Arrhenius dependence on temperature. They also observed that the foam
took a shorter time to start curing than did the solid. Baser and Khakhar developed kinetic
equations and a one-dimensional model to elucidate foam expansion over time for a water and R-
11 blown foam, by assuming first order reaction kinetics for the blowing reaction [16]. Lefebre
and Keunings developed a finite element model of reaction polymeric foam extrusion using
proprietary kinetics that they do not discuss in detail [17]. Haberstroh and Zabold use the
commercial software CFX with a residence time approach to density predictions to understand free
rise foams [18]. Seo et al. used a time-dependent density function fit to data to model foam
expansion [19]. They later extended this work to use Baser and Kharkhar kinetics for the foaming
reaction [20]. Bikard et al. used reduced order foaming kinetics that they implemented directly in
the continuity equation [21]. They use a Piloyan law for the curing reaction.

Rao and coworkers developed an engineering model combining the equations of motion, an energy
balance, a density model to represent the foam blowing reaction, and a rate equation for the
polymerization reaction for polyurethane PMDI-4 encapsulation foam (free rise foam density of
roughly 4 lb/ft3) [22] and an epoxy foam with a FluorinertTM physical blowing agent called EFAR

3
foam [23]. The equations describing the kinetics of the polymerization reaction were of the form
of condensation chemistry and were populated through micro-attenuated total reflection (ATR)
infrared (IR) spectroscopy measurements. Unfortunately, no IR peak is a fingerprint of the gas-
forming reaction. Therefore, the blowing reaction effects were approximated through an
empirical, temperature-dependent, equation for the density as a function of time, following Seo et
al. [19]. However, this approach leads to prediction of a constant density throughout the foam at
any one time. Unfortunately, to really understand the foaming process, we need to be able to
predict both density evolution and local density behavior, which might lead to gradients and affect
the final foam strength, structural response, and aging behavior [24]. For this reason, we have been
working on a detailed kinetic model of the foaming reaction.

Our approach is unique though it builds on the previous literature and our previous modeling work.
We have developed a new kinetic model, which follows a simplified mathematical formalism that
decouples foaming and curing reactions. This approach is thoroughly grounded in experimental
data, where we have seen that the isocyanate is always in excess and does not affect the kinetics
[25]. The model predicts the polymerization extent of reaction via condensation chemistry, where
vitrification and glass transition temperature evolution must be included to correctly predict the
temperature-dependence of the kinetics. The kinetics are fit to fast and accurate IR data, which
uses small sample sizes to ensure an isothermal experiment. A DiBenedetto form of the glass
transition temperature evolution is used and fit to rheology and DSC data [26]. Instead of using an
extent of reaction for the foaming reaction as has been done previously [16], we track the molar
concentration of both water and carbon dioxide using a volume evolution experiment to fit the
reaction kinetics. We include a nucleation time, which ramps up the rate of reaction based on the
approximate time the carbon dioxide dissolves in the polymer and supersaturates before generating
volume. Understanding the thermal history and loads on the foam due to exothermicity and oven
heating is very important to the results, since the kinetics and material properties are all very
sensitive to temperature.

In this paper, we focus on the details of the reaction kinetics model that predicts the density and
state of cure of the polyurethane foam as it evolves, yet is simple enough to incorporate into a
computational model of the overall process. The method established here can be used for a variety
of polymeric foams, although we only demonstrate it on one polyurethane material. The paper is
organized in the following manner. In first section, we discuss the polyurethane formulation,
mixing protocols, and the key reactions of interest. Next we summarize the proposed closed-form
kinetic models for polymerization and foaming. Measurements for the polymerization and foaming
reactions rates and details of how we fit the data and the resulting parameterizations are described
in the third section. A comparison of the kinetic models to the data is given in the results section.
Finally, conclusions and plans for future work are discussed.

KINETIC MODEL FOR POLYURETHANE FOAM

Polyurethane Formulation
Polyurethanes are chemically blown foams that begin as two-part mixtures, with polyisocyanate
in one part (the curative or T-component) and polyol, water, surfactant and catalyst in the other

4
part (the resin or R-component). When mixed together, several competing reactions occur
simultaneously, the most important of which are the polymerization reaction and the foaming
reaction (Figure 2). Polymerization occurs via a condensation reaction of polyisocyanate and
polyol to form a crosslinked polyurethane. The foaming reaction occurs between polyisocyanate
and water, forming carbamic acid which then decomposes to form carbon dioxide (CO2) and an
amine group [26]. The carbon dioxide is a gas, which is responsible for blowing the polymer into
a foam. Several secondary reactions occur, when products of the primary reactions react with
isocyanate to produce polyurea, biuret, and allophanate.

Two key reactions: Isocyanate reaction with polyols and water

H O
Urethane formation,
R1 N C O + HO R2 R1 N C O R2 crosslinking
H O
Foaming reaction
R1 N C O + H2O R1 N C OH CO2 + R1 NH2 yields CO2 and amine

Various follow up reactions: Isocyanate reaction with amine, urea and urethane

H O H
R1 N C O + R1 NH2 R1 N C N R1 Urea formation

H O H H O R1 O H
R1 N C N R1 + R1 N C O R1 N C N C N R1 Biuret formation

H O H O R1 O
R1 N C O R2 + R1 N C O R1 N C N C O R2 Allophanate formation

Figure 2. The two primary chemical reactions and three secondary

reactions involved in producing solid polyurethane foam.

The recipes for components R and T for the structural foam of interest are shown in Figure 2, in
parts by weight (PBW). The R to T mix ratio is 40.3 to 59.7 parts by weight. This results in the
overall formula also listed in Table 1. Densities of those components listed with specifications
are from the manufacturer supplied data [28, 29].

Table 1. PMDI-10 Structural Foam Formulation

Material PBW Density at
25°C (g/cm3)
R-Component (resin)

Voranol 490 100 1.109

(polyol)
Water 0.8 0.997
Dabco DC-197 1.0 1.01
(surfactant)

5
 TMPDA 0.7
(catalyst)

T-Component

PAPI 27 100 1.23

(isocyanate)

Overall formula Mass fraction

polyol 0.3932
water 0.0031
surfactant 0.0039
catalyst 0.0027
isocyanate 0.5970

To model the reactions over time, we follow a kinetic approach. We track moles of water reacted
to produce carbon dioxide gas, which leads to the cellular structure of the foam. As the material
foams, it is also polymerizes, changing from a liquid to a solid.

As discussed above, the gas formation reaction depends on the presence of water in the R
component. We also formulated similar material without the water and added a dessicant to
attempt to eliminate the effect of moisture absorption from the atmosphere. The polymerization
of this dried material could then be compared to that of the full foaming formulation to indicate if
there were significant interactions among the reactions.

Polymerization Kinetics
The reaction is written for the extent of reaction for polymerization ξ. The reaction term is written
in the form of condensation chemistry [8, 9, 30, 31]:
D
 k (b   m )(1   ) n . (1)
Dt

Here b is a fitting parameter, m and n are fitting exponents, and k is an Arrhenius constant, but also
includes the effects of vitrification though a time-temperature superposition (TTS) shift factor
following the form of the William-Landau-Ferry (WLF) equation [32] with fitting parameters, w
and β.

 1  E / RT
k   k0e  (2)
 1  wa   
 T 

where ko is the rate coefficient, Eξ is the activation energy, R is the universal gas constant, w and
 are fitting parameters, and aT is the shift factor for TTS, which includes an evolving glass
transition temperature, Tg, which is dependent on the extent of reaction.
6
 C1 T  Tg 
log10 aT  (3)
C2  T  Tg
 
Ideally C1 and C2 are WLF constants that can be found by fitting the shifted data in the rubbery
regime [33], however we have used them as fitting parameters. For Tg we use the Di Benedetto
form [26]. A nice discussion of some other possible forms of Tg for polymerizing systems can be
found in Bilyeu et al., [34], but we have found the Di Benedetto form works well for our data,
fitting its quadratic shape with only one floating parameter.

Tg 0 1     A  Tg 
Tg  (4)
1    A  
The parameters for the model are Tg0, the glass transition temperature of the unreacted material,
Tg∞, the glass transition temperature of the fully reacted material, and A, a fitting parameter related
to the stoichimetry of the system. Tg0 and Tg∞ can be inferred from experimental data. The De
Benedetto form is a state function of the reaction extent, therefore, it does not depend on material
history and is straightforward to fit to experimental data.

Foaming Kinetics
Polyurethane foam production is a complex process involving reaction of water with isocyanate,
which then forms carbamic acid. The acid then decomposes into carbon dioxide. The carbon
dioxide is soluble in the polymer precursor and concentrates until it reaches super saturation, at
which point nucleation begins. We hypothesize that dissolved air bubbles from mixing act as
nucleating sites since only minimal foaming occurs unless a sufficient amount of air is
incorporated. Once nucleation occurs, the bubbles start to grow.

Our experimental data measure the volumetric expansion of the foam, not the actual concentration
of carbon dioxide. Thus we measure an effective carbon dioxide concentration that leads to bubble
growth. Unfortunately, this method does not account for the CO2 solubilized in the polymeric
continuous phase or the gas that pressurizes the bubbles after the fluid vitrifies. When fitting the
data, we have found that the concentration of carbon dioxide depends only on the concentration of
water and is insensitive to the concentration of isocyanate, which is always in excess while the
foaming reaction is occurring. This makes sense, since only 3.9 molar % of isocyanate is used in
the water reaction. The rest forms urethane and polyamine leaving 15 molar % of the isocyanate
unreacted.

We write concentration in terms of moles/volume liquid and track both H2O and CO2. Because of
the stoichiometry, one mole of water is equivalent to one mole of carbon dioxide, where water is
the tracked reactant, with isocyanate being in such excess that it does not need to be tracked. As
the water is consumed, it produces carbon dioxide at the same rate.

7
 DCH 2O
  Nk H 2O CH 2O q
Dt (5)
DCCO2
  Nk H 2O CH 2O q

Dt

We fit equation (5) to semi-isothermal data for volume evolution with time to determine the rate
of reaction, Arrhenius rate coefficient k H 2 0 , and the rate exponent, q.

 ( EH 2O / RT )
k H 2O  AH 2O e (6)

Here k H 2 0 is fit at each temperature and from this a base coefficient AH2 0 is determined along with
the activation energy EH2 0 . A nucleation time delay, N, must be included to approximate the time
required for the carbon dioxide to dissolve in the continuous phase, and then begin nucleating
bubbles:

  t  tnucleation  
N  0.5 1  tanh   . (7)
  tscale  

Without this effect, the shape of the curve is wrong since it does not take into account the delay
from the onset of the reaction to volume evolution. This will be seen in the material parameter
fitting section. Best fit parameters for the exponent, q, as well as Arrhenius parameters and the
characteristic time for nucleation will be given in the next section.

Density
From the local concentration of CO2 determined by equation (5), we can determine the volume
fraction of gas. We track the molar volume of water and carbon dioxide in the liquid phase only,
and then convert to volume fraction of gas, .

Vgas M CO2 CCO2 v

v   (8)
Vliq  gas 1 v
 

Here the subscript liq refers to the polymer phase and gas refers to the carbon dioxide, the density
of which we define based on the ideal gas law.

PM CO2
 gas  (9)
RT

We can now predict the foam density from the gas density, liquid density, and the gas volume
fraction.

8
  foam   gas  liq (1   ) (10)
 

EXPERIMENTAL METHOD AND FITTING MATERIAL MODELS

Structural parts are foamed under conditions that vary somewhat depending on the mold size,
shape, and features. Typically the two-part material is preheated to 29 - 30°C and the mold is
preheated to 30 - 40°C. The material is mixed for approximately 45 s, in fairly large batches (at
least 250ml, depending on the size of the mold) with an electric mixer equipped with a blade. The
mixed material is immediately poured into the mold, which has been taken out of the oven, and
the foam is allowed to rise for about 10 minutes. The mold is then placed in an oven for four hours
at a higher temperature, typically 121°C, to complete the cure.

Kinetics of Polymerization
Micro-Attenuated Total Reflectance (ATR) Infra-Red Spectroscopy (IR) was used to probe the
kinetics of polyol isocyanate and related reactions at a range of temperatures. We also examined
whether significant differences in spectral features exist that could be used to decouple the polyol
cure and water based foaming reaction which competitively consume the isocyanate.

IR Spectroscopy Used to Determine Reaction Kinetics

Micro-ATR experiments were conducted using a heated diamond micro-ATR attachment in a

Bruker Equinox 55 IR spectrometer with a DTGS (ID301/8) detector operating at room
temperature. Spectra were collected as an average of 64 scans at 4 cm-1 resolution. For rapid cure
conditions at the higher temperatures (>50oC), we used 16 scans to better capture early reaction
kinetics. Repetitive loop macros were used to acquire sequential spectra over multiple hours. The
observable spectral features are related to the dominant reactions, namely consumption of the
isocyanate at 2260 cm-1, the formation of carbonyls either as urethanes (1723 cm-1) or ureas close
to 1700 cm-1, the amide II band at 1521 cm-1 and 1510 cm-1, a band at 1308 cm-1 and the “urethane
ester linkage” at 1218 cm-1. Since the spectra are heavily convoluted, it is not possible to integrate
specific band areas for all of these bands. This creates significant constraints for the quantitative
analysis of spectral data. While a collective carbonyl band area could be established using a range
from 1650-1750 cm-1 for integration, all other peak changes were quantified using peak height
instead. The spectra also show a decrease of hydroxyls at approximately 3400 cm-1 and an increase
of NH absorption at 3320 cm-1. Neither one of these broader bands is suitable for simple
quantification. Multiple IR bands were examined: carbonyl formation at ~1700cm-1, urethane ester
linkage at 1218cm-1, and a non-assigned band at 1308 cm-1. The peak at 1218 cm-1 represents the
pure curing reaction, as it relates to the C=O-O-R linkage (i.e., the ester side of the urethane)
between the polyol and urethane group. Successive IR spectra were acquired over time (Figure
3), and the relative absorbance changes for specific bands provided time-dependent cure-state
information.

9
 Figure 3. IR-spectra at various times showing the isocyanate peak
decreasing as the urethane band and others increase.

Two resin systems were evaluated: The basic foam system consisting of a standard foam system
in which the polyol component contains a pre-mixed surfactant, water and catalyst, and a “dried”
system from which the water had been removed from the polyol component using common
laboratory-grade zeolites as a drying agent. A standard sample mass of two grams was mixed, as
discussed above, for every run and then placed on the preheated temperature-controlled plate for
spectral acquisition. The first spectrum is generated after approximately 64 seconds for 64 scans
and 16 seconds for 16 scans. For all dried system experiments, four small beakers filled with
anhydrous calcium sulfate were placed around the sample and covered with a beaker, and the
sample was covered with a small strip of aluminum foil to minimize any diffusion of water into
the sample. Wet foam experiments were treated in a similar manner.

For the PMDI foam under investigation, the isocyanate reactions on a molar basis split into 94.4%
for the polyol addition to form urethanes, and roughly 5.6% for the isocyanate water foaming
reaction. In principle, approximately 5.6% of urea formation from an amine isocyanate addition
reaction should be observable as a follow-up of the CO2 producing carbamic acid decomposition.

Ideally, a comparison of the spectral features between a dried non-foaming system, and a water-
containing polyurethane foam system should reveal spectral variations in the nature and quantity
of the underlying reactions. The dried system should exclusively display a dominant urethane
formation, while the wet system should involve a combination of urethane and urea products.
However, both systems will be further convoluted by subsequent allophanate and biuret reactions
when isocyanates are used in excess. This mix of reactions implies that a deconvolution of the
foaming and urethane cure reactions based on IR spectral features alone is intrinsically difficult.
A complicating issue is that the carbonyl group of the urea does not display a specific isolated

10
spectral band, but only appears at a slightly lower wave-number than the urethane carbonyl.
Deconvolution of the urethane and urea carbonyls, particularly at low concentrations of urea is
prohibitive, unless chemometrics and sophisticated algorithms are applied. Such approaches were
not the goal of this overview study.

To determine curing rates in the full water-containing foam system, IR data were obtained for the
1218 cm-1 band height as a function of time and temperature (
Figure 4).

Figure 4. Peak height as a function of time for the 1218 cm-1 peak for the
wet PMDI material. Isothermal tests were carried out for various
temperatures ranging from 30oC to 90oC.

The raw peak height from

Figure 4 can be normalized by the maximum peak height, which we assume to be roughly ~0.48cm-
1
. There is some uncertainty in this number, but we use a value consistent with the maximum peak
height at the highest temperature. The normalized height is a direct measure of the extent of
reaction as a function of time and temperature. Shifting the various temperature data to obtain a
master curve (
Figure 5) gives shift factors as a function of temperature, which can then be used to determine the
activation energy for the Arrhenius rate constant.

11
 Figure 5. Shifted extent of reaction for isothermal tests carried out for
various temperatures ranging from 30oC to 90oC. The shift to a master
curve works well up until 1-2 hours.

Figure 6 shows a plot of the natural logarithm of the shift factors versus the inverse of temperature
in Kelvin. The slope of this linear plot gives the activation energy, E/R, for the rate constant.

Figure 6. The natural log of the shift factor plotted versus the reciprocal
temperature in Kelvin, gives the activation energy, E/R = 4288.9 K, for the
Arrhenius rate constant for the polymerization reaction.

12
We also tested the dry material and noted that, although differences in the peak height evolution
did occur, they were slight and the activation energy was quite similar. We took that as evidence
that we could approximate the Tg of the foaming system to be that of the dried, as will be described
below.

Parameter fitting for polymerization kinetics

Once the extent of reaction and activation energy are determined, the reaction kinetics (e.g. rate
constant and order of reaction) can be obtained by fitting equations (1)-(4) for the early time data.
This is done using the extent and the rate extent, formed by numerically differencing the extent
data, and then fitting this data to equation (1) assuming no advection, diffusion, or cure gradients
exist in the small IR samples.

The measured rate of reaction was determined numerically from the extent data by differencing
the extents and dividing them by the time increment. Differentiating data can compromise
accuracy, and there is often more scatter in the rate data than the extent itself. To solve the
polymerization kinetics equation, we integrate the equations numerically in Excel using a simple
forward differencing method with small time intervals.

d
 k (1   ) q ( A   p )
dt
(i  i 1 )
 k (1  i 1 ) q ( A  i p1 ) (11)
(ti  ti 1 )
i  i 1  k (1  i 1 )q ( A  i p1 )(ti  ti 1 )

             
where ξi refers to the data value at time ti. Note that k is temperature and cure dependent, and
includes the effects of vitrification and evolving Tg as defined in equation (2). Then the parameters
were evaluated in light of both the rate and extent as will be shown below.

The evolution of the glass transition temperature must be included to capture the late time data and
the different final extent of reaction for different temperatures. An estimate for Tg is determined
from rheology and DSC of the non-foaming material. This will be discussed in the next sub-
section.

Parameter fitting for evolving glass transition temperature

Between gelation and vitrification the reaction rate decreases because reactants become less
mobile due to solidification of the polymer. We represent this phenomenon using an evolving glass
transition temperature and a shift factor. The evolution of the glass transition temperature with
extent of reaction is fit using viscosity vitrification data and differential scanning calorimetry
(DSC) data. The time that the viscosity became too large to measure was taken as an estimate of
the vitrification time. Because both the rheology and the Micro-ATR experiments were isothermal,

13
the time of the rheology measurement could be directly translated into an extent of reaction in the
Micro-ATR experiment at the same temperature and then fit to equation 10 using estimated and
measured parameters Tg0 and Tg∞.

Tg∞ was estimated from thermomechanical analysis (TMA), using a TA Instruments Q400 TMA,
of the cured dry material. Samples of both the dry material and an actual part of interest were
tested with a method that measured CTE for both cooling from 175°C to -35°C, and then heating
back up from -35C to 175C. Each had been cured with the normal cure protocol (at 121°C) for
parts made in the application of interest. The approximate Tg determined from the intersection of
the slopes of the curves of dimensional change vs. temperature were about 145°C for both samples,
with a range of gelation (the region where the curve was changing slope) from about 120 – 165
°C. [35]. Because we could not guarantee that the pieces were fully cured with the standard
protocol, and higher temperatures could not be tested because polymer degradation could occur,
we assumed that Tg∞ was somewhat higher than 165°C, or 438K, and so chose as an estimate 450K.
Tg0 was used as a fitting parameter, so that the rheology data described below fit well on the
DiBenedetto form for Tg (equation 10).

Rheology measurements were taken on a TA AR-G2 rheometer with parallel plates, dynamically
(in oscillation) at frequencies of 1 to 31.62 Hz, either controlled to a constant strain (0.1%) or a
constant stress (5 Pa). Frequency sweeps were performed to ensure that measurements were within
the linear viscoelastic regime. It is interesting to note that the maximum dynamic viscosity
occurred at critical extents that were somewhat frequency dependent, a common observation for
curing systems, especially those involving vitrification [5, 6, 11,12]. For simplification, we
estimate this critical extent from the data at 1 Hz.

A study showed little effect of varying the distance between plates. Data presented here were
taken with a gap between plates of 1.0 mm. The components were preheated to 30°C, which is the
common procedure for the application of interest. The rheometer plates were preheated to the
operating temperature so that the material would quickly reach that temperature once loaded onto
the bottom plate. Nevertheless, it takes time (up to 90 s) for the material to be loaded and come to
the plate temperature. This is after the 45 s mixing step. Complex moduli were measured at four
temperatures (30, 40, 50, 70°C) and a dynamic viscosity was determined. Results are shown in
Figure 7. The time, for each temperature, that the viscosity reached the approximate limit of the
instrument (shown with the dotted line) was taken as the approximate point at which the reaction
essentially stops, either through gelation or vitrification.

14
 Figure 7. Dynamic viscosity measurements of the dry (non-foaming)
material at 1 Hz and four temperatures.

The glass transition can be very broad for thermosets, encompassing a temperature range instead
of a single temperature. Because of this uncertainty, we use transient DSC to indirectly observe
the onset of the glass transition in the curing dry PMDI material and then integrate the kinetics
according to the temperature history to determine the state of cure. This approach allows us to test
a larger range of temperatures than the isothermal rheology alone and also have an independent
confirmation of the Tg model. The rheology gives us a mechanical measure of Tg whereas the DSC
gives us thermal measurements of Tg. A batch of dry PMDI was prepared, measured into DSC
pans which were each crimped shut, and then immediately put into the deep freeze to minimize
the amount of curing that occurred during specimen preparation. At test time, a specimen was
removed from the deep freeze, placed in the DSC at 25oC, and immediately quenched to -30oC.
This step is important to prevent condensation from forming on the outside of the specimen pan.
After the temperature had equilibrated at -30oC for 5 minutes, the specimen was subjected to a
series of cyclic temperature ramps, hold, and cooling ramps back at -5oC. The temperature rate
and maximum temperature are experimental variables that may be changed from test to test. We
frequently use 10oC/minute for heating and cooling ramps and hold times at the high temperature
target (50oC, 55oC, and 60oC) of 1 minute.

During these temperature ramps, the heat flow relative to the reference pan, which is the standard
definition in DSC, was measured. On heating, we could clearly either observe the transition from
the glassy state to the rubbery state (increase in endotherm per degree oC), which was followed by

15
a significant exotherm due to curing, or we directly observed a transition from the glassy state to
a state with a substantial curing exotherm. (The exotherm is caused by heat generated by the
polymerization reaction.) The onset of additional cure was easily observed because, especially
during the first two cycles, the endothermic curve had an extremum as the reaction suddenly
restarted. Ahead of this extremum, an inflection point may be calculated from the endothermic
curve during specimen heating. Therefore, for a given specimen, the time since the start of the
cyclic DSC test was recorded at the locations of these extrema and inflection points.

Given the cure kinetics and Tg parameters above, we can integrate the non-linear reaction kinetics
ordinary differential equation using the time-temperature history applied to a specimen during the
transient DSC test. We can associate specific time-temperature histories with extent of reactions.
Finally, we can plot these points of Tg onset vs. extent of reaction with the Tg vs. extent of reaction
used in the DiBenedetto fit. If the agreement is good, then the analysis can be considered a
validation of the reaction kinetics fit and DiBenedetto form. However, if the agreement is poor,
then we must reevaluate the DiBenedetto parameters to try to improve the fit. Each time the
parameters are adjusted, the reaction extent must be re-integrated from the time-temperature
histories for each specimen.

An example of the raw DSC data and with some kinetics analysis is shown in Figure 8. The
specimen was subjected to heating and cooling rates of 10oC/min to a target temperature of 50oC.
Then, after 10 cycles, the specimen was heated to 100oC to determine the glass transition onset
after the ten cycles. The first part of the figure shows the raw heat flow versus temperature data
(with the time axis removed). The inflection and extremum for the first heating ramp are marked.
Other thermal histories were run as well including ratcheted temperature ramps (-3040-
3060-3080-30100-30 all at 10 C per minute with 1 minute holds at each extreme
temperature). Extrema and inflection points were extracted in the same way as in figure 10.

16
 Figure 8. An example of raw DSC data from thermal cycling of dry
polyurethane.

Then, the reaction kinetics are integrated with this time-temperature history, which results in a
reaction extent versus time for this non-isothermal history (Figure 9). The locations of the
inflection and extrema are also placed on this curve, and the temperature vs. time is also plotted
for reference. We note that each pair of inflection/extremum points corresponding to the same
upward temperature sweep are slightly different measures of the onset of the glass transition. They
are not independent.

17
 Figure 9. Estimate of Tg with extent of reaction. The locations of the
inflection (red dots) and extrema (blue squares) are also placed on this
curve, and the temperature vs. time is also plotted for reference. The pair
of inflection/extremum points corresponding to the same upward
temperature sweep are slightly different measures of the onset of the glass
transition.

Finally, these Tg onset locations along with the rheological estimates of vitrification extent are
plotted against reaction extent in Figure 10 to test the DiBenedetto fit. Fortunately the DSC and
rheological vitrification data agree reasonably well. Including vitrification is very important since
we are need to understand the post-gelation curing. If vitrification is not included, the reaction will
go to completion for all temperatures. Vitrification quenches the cure and leaves cure gradients in
place.

18
 Figure 10. Glass transition temperature evolution with extent of reaction fit
to a Di Benedetto model using rheology and DSC data.

Foaming Reaction
Unfortunately, no specific peak in the IR spectrum exists for the foaming reaction for two reasons.
First, the carbamic acid generating the CO2 is a transient species, and second, the carbamic acid
decomposition results in urea linkages (carbonyl and amide groups), which overlap with the
urethanes in the spectrum. Therefore, a different method must be used to determine the foaming
reaction rates, as will be discussed in the following subsection.

Experiments to determine foaming kinetics

In order to estimate the rate of gas formation, we observe the height evolution of a foam column
with time and assume that the increase in volume is solely due to the production of CO2. This
method is useful since it predicts the apparent CO2 generation that goes into volume generation.
However, it does not account for all the CO2 produced, since some can be lost to the atmosphere
through the free surface, some remains supersaturated in in the polymer, and some leads to bubble
pressurization that is difficult to measure after vitrification. All of these secondary CO2 pathways
do not produce an increase in volume. Because the reaction is exothermic and the foam precursor
materials cannot be preheated to high temperatures (excessive preheat causes the water to
evaporate so it will not be available for the foaming reaction), these height evolution experiments
cannot be performed under isothermal conditions, unlike the IR spectrophotometry. We also
hypothesize that the blowing reaction could produce pressurized bubbles, especially between
gelation and vitrification of the continuous phase. Therefore, not only volume change, but also
temperature and pressure are measured. The number of moles of CO2 is then estimated through
the ideal gas law:
19
 PVgas  nCO2 RT (12)
where Vgas is volume of gas measured in the foam experiment. The method assumes that the
volume change due to temperature of the liquid phase is negligible, as is any volume change from
reactions not producing gas. Few if any bubbles were observed to pop at the surface, although a
few were observed at 70oC and more at higher temperatures. Therefore, we did not test above
70oC.

In an attempt to measure the amount of CO2 formed in the reaction of isocyanate and water under
more isothermal conditions, we explored separating the polymerizing reaction from the gas
forming one by reacting water and isocyanate in a surrogate continuous phase devoid of polyol.
We mixed part T in the appropriate ratio with water in an epoxy base that would yield a similar
initial viscosity as the part R. We found that the foaming reaction, measured via the height of the
foam, proceeded almost an order of magnitude slower than the foaming in the actual polyurethane
mix. This could be attributed to any combination of effects, including lack of extra heat generated
by the polymerizing reaction, reaction synergism, or instability of bubbles in the surrogate as
compared to the cross-linking polymer. We did not investigate this further, but proceeded to test
the foaming reaction rate in the real system, compensating where possible for the nonisothermal
nature of the experiment in the analysis of the data.

All tests were performed in an oven, where the mold was preheated to the desired testing
temperature. For all tests the precursor foam material was preheated to 30°C. The material was
mixed for 45 s inside the injection syringe with an overhead mixer fitted with a small spatula
rotating at 1800 rpm. Mixing directly in the injection syringe provided significantly faster transfer
of the material into the mold after mixing.

Figure 11 shows frames from a typical front-tracking video at an oven temperature of 30°C. Note
that the foam started expanding in such a short time that it was difficult to capture the initial time
data, even at this relatively low temperature where the reactions are slower.

20
 Figure 11. Video frames showing structural foam PMDI-10 expanding in a
bar mold at 30 °C. The frames correspond to time= 13, 73, 158, and 245 s
after the end of mixing the resin and the curative together for 45 seconds.
Even at this relatively low temperature the foam had expanded significantly
within the 13 seconds that it took to load the syringe injector and get the
material into the mold.

Individual frames are analyzed with image processing software (Image Pro Plus, Media
Cybernetics, Rockville, MD) to determine the height evolution with time, calibrated with the
known dimensions of the channel. Knowing those dimensions of the channel, we can convert the
height of the foam to volume, and knowing the mass injected we can convert the volume to an
overall density. Assuming volume change is caused by the generation of CO2 gas and possible
expansion or contraction of the gas due to temperature; we can calculate the generation rate of the
gas forming reaction.

Example data are shown for the case of an oven temperature of 50°C in Figure 12 and Figure 13.
Note that the temperature peaks before the foam stops expanding. The temperature rise is caused
by the heat of reaction. Note also that the pressure peaks after the foam stops expanding; implying
that the bubbles may become pressurized after the polymerization restrains the bubble expansion.
However, the pressure is measured at the wall, and it is not clear how that pressure is related to
that within the bubbles. For lack of a better estimate, we use this wall pressure and the temperature
nearest the free surface to calculate the moles of gas from equation (12). The pressure data changes
in an abrupt, stair stepping fashion, thus smoothing is used on the pressure to give the data a
continuous slope. Note that the foam expansion stops relatively early (about 125 s after the mixing
step) compared to the time of vitrification as seen in Figure 7 (about 300 s after the mixing step).
The foam expansion stops about the time of the first change in slope of the dynamic viscosity,
which also corresponds to the first change in slope of the viscous modulus. According to Harran

21
and Laudouard [36], this indicates the gel point. The viscosity is obviously large enough here to
slow or stop the foam expansion, despite the gas forming reaction continuing.

Figure 12. Temperature and pressure measurements used to calculate the

number of moles of CO2 produced at a nominal temperature of 50°C.

22
 Figure 13. Measured volume and the moles of CO2 produced calculated
from the ideal gas law.

Parameter fitting for foaming kinetics

In the experiment, 3.59g of foam precursor is injected giving an injection volume of 3.38 cc
assuming an initial density of 1.06 g/cm3. The volume is difficult to estimate because the foaming
begins very quickly. We use a solubility estimate of 0.018 cc STP CO2/(cm3 polymer *P(cm Hg))
[37] giving a maximum solubility of 0.05 mmol/cm3. We make an assumption that roughly 0.01
mmol/cm3 is the non-volume creating carbon dioxide that is still dissolved in the continuous phase
and use this value for all temperatures. This is an assumption, since we know that the solubility is
temperature and pressure dependent. In future studies, we would like to include rate equation both
for the gas-phase carbon dioxide as well as the solubilized carbon dioxide. For now, we assume
that 0.01 mmol/cm3 of water is unavailable to produce CO2 bubbles because this amount is
dissolved in the liquid.

As there is an excessive amount of isocyanate, the limiting reactant to produce CO2 is water. One
mole of CO2 is produced from each mole of water (H2O) in the foam kit formulation. Note that
this assumes that there is no additional water absorbed into the resin components as received from
the manufacturers or that water is not absorbed or lost to the atmosphere during storage or
processing.

Table 2 gives the initial molar concentrations of the reactants, including the molecular weights,
converting from the formulas in Table 1. The maximum amount of carbon dioxide that can be
produced is limited by the initial concentration of water. We assume that the initial concentration
of CO2 is a small but nonzero and include the air only as an effect on the initial density of the

23
mixture, though we estimate the foam precursor contains at least 14 vol% air bubbles before
nucleation begins.

Table 2. Molar Volumes of Reactants (mmol/cm3)

Component Molecular Initial moles (mmol) Concentration

weight (mmol/cm3)
Polyol 460 3.089 0.9128
Water 18.02 0.631 0.1854
Isocyanate 340 6.346 1.875

Carbon 44.01 0.051 0.015

dioxide

Recall from the discussion below equation (6), that the Arrhenius rate coefficient for the foaming
reaction is fit for 30oC, 40oC and 50oC to determine the activation energy and the base coefficient.
This is then used to fit the 70oC test, which ended up being fairly nonisothermal. Figure 14 shows
the Arrhenius fit for the three temperatures for which we have isothermal data, and extrapolates
this value to 70oC. This gives an activation energy for the foaming reaction of EH2O/R = 4741.7 K.

Figure 14. Arrhenius behavior for the carbon dioxide rate equation.

RESULTS

24
Using the kinetic fits from the previous sections for the polymerization and foam generation, we
can compare the analytical rate equations to the experimental data.

Polymerization Kinetics
A rough value of k=0.0014 s-1 at 23°C, combined with the Arrhenius fit from Figure 6, yields a
value for k0 of 1951.8 s-1. This value seems to capture the temperature dependency well. The best
values to fit the orders of reaction are n=2.0 and m=0.2. The heat of reaction, Hrxn , was measured
from DSC for the wet and dry foam, which had similar values of 181 J/g. Combining the best fits
to all of the data described above leads to the parameter values for the epoxy polymerization
kinetics summarized in Table 2.

Table 3. Polyurethane Polymerization Kinetic Parameters

Polyurethane Cure Parameter Value
Rate coefficient, k0 1951.8 s-1
Universal gas constant, R 8.31 J/mol K
Normalized activation energy, Eξ/R 4289 K
Activation energy, Eξ 35.6 kJ/mol
Rate parameter, b 0.6
Rate exponent, m 0.2
Rate exponent, n 2.0
Heat of reaction, Hrxn 181 J/g
Glass transition temperature at zero cure, Tg0 270 K
Glass transition temperature at zero cure, Tg∞ 450 K
Glass transition evolution parameter, A 0.15
WLF parameter, C1 10
WLF parameter, C2 30
WLF scaling parameter, w 0.2
WLF scaling exponent,  0.2

We simultaneously fit the extent and the rate to insure that both give a fairly good representation
of the curing reactions for all temperatures. The extent will be needed to predict viscosity and the
rate for the energy equation to track the heat generation from the reaction, so representing them
both accurately is important to the modeling effort [23]. Comparison of the theory, using the
parameters listed in Table 3, to data is shown in Figure 15 and Figure 16, for the extent of reaction
and rate of reaction respectively. If we focus on getting a good fit for the lower temperatures
associated with processing, then it is hard to capture the higher temperatures.

25
Figure 15. The measured extent of reaction compared to the numerical fit
from equation (7) for seven different temperatures.

Figure 16. The rate of reaction compared to the numerical fit from equation
(7) for seven different temperatures.

26
Foaming Kinetics and Density Predictions
Figure 17 compares the measured water concentration (from the measured change in volume
translated to moles of H2O consumed) to the predictions from the model. The results agree fairly
well with the data at early times, though the reaction slows more at later times. This is probably
due to the fact that bubbles have trouble growing at later times as the material becomes more
viscous, then viscoelastic, and finally vitrifies to a solid. It is difficult to capture this effect without
a bubble-scale model.

Figure 17. Water concentration for the model compared to data for BKC
44306 PMDI-10 foam at an oven temperature of 30°C, 40°C, 50°C and 70°C.

The measured values of the rates of water consumption have more scatter in them but with the
inclusion of nucleation time, we can capture the change in the shape of the curve, allowing us to
capture the early time data when the carbon dioxide formation does not lead to volume change. A
simple rate equation would not capture the complexity of the data. Figure 18 show the rate data
compared to the model predictions for rate for the four temperatures 30oC, 40oC 50oC, and 70oC.
The model for 70oC used the Arrhenius fit determined from the lower temperature data and the
measured temperature at each time, because the foam in the experiment was the most
nonisothermal. The foam rose quickly at this higher mold temperature, and it took the material
almost the whole rise time to heat to 70oC from the injection temperature of 30oC. The Arrhenius
fit proved to extend quite well to 70oC as shown in Figure 14. The scatter in the rate data is difficult
to minimize because we are not directly measuring it but differentiating the measurements of
height vs. time.

27
Figure 18. Comparing rate of water depletion predictions to data for PMDI-
10 foam at an oven temperature of 30°C.

Figure 19. Comparing rate of water depletion predictions to data for PMDI-
10 foam at an oven temperature of 40°C.

28
 Figure 20. Comparing rate of water depletion predictions to data for PMDI-
10 foam at an oven temperature of 50°C.

For 70oC the experiment is nonisothermal since the foam is always mixed at 30oC. For this reason,
we use variable temperature and pressure from the data to determine to estimate the response using
the Arrhenius fit from the other temperatures. The nucleation time is related to the solubility of
carbon dioxide in polymer; it should probably vary with temperature to improve the fit at higher
temperatures, but for simplicity we use on one set of time values. Temperature-dependent
solubility data for carbon dioxide could improve the fit, and this will be a focus of future studies.
The fit and data are seen for 70oC in Figure 21.

29
 Figure 21. Comparing rate of water depletion predictions to data for PMDI-
10 foam at an oven temperature of 70°C.

The predicted density versus time for 30oC through 70oC experiments is compared to the measured
in Figure 22.

30
 Figure 22. Predictions of density using a nucleation time of 40s compared
to measured density with time in the channel for several temperatures. For
70oC, the model uses the Arrhenius form with temperature variations,
whereas the other comparisons are made for the nominal temperature of
the mold.

The coefficients for the gas generation kinetics model are summarized in Table 4.

Table 4. Gas Production Kinetic Parameters

Gas Production Parameter Value

Rate coefficient, AH2O 1.19e5 s [mmol/cm3]1-n
-1

Normalized activation energy, EH2O/R 4741.7 K

Activation energy, EH2O 39.40 kJ/mol
Rate exponent, q 1.5
Nucleation time, tnucleation 40s
Time scale for reaction, tscale 20s

CONCLUSIONS
A kinetic-based numerical model has been developed for structural polyurethane foam with a fast
catalyst, PMDI-10. The model includes the kinetics of polymerization via an extent of reaction
equation with vitrification based on the glass transition temperature evolution. A DiBenedetto
model is used for Tg and populated with data from IR-spectroscopy, DSC and rheology. We have
developed a new kinetic approach to the gas production reaction based on the molar concentration

31
of water, which forms carbon dioxide after reacting with isocyanate. The kinetics are fit to foam
volume experiments. A goal of this work was to produce a model of the primary reaction rates
that could then predict the density of the foam when used in a processing model. The model is
able to capture the densities seen experimentally. Future work will use this kinetic model to predict
density and cure gradients in a complex foaming system using computational fluid dynamics.

NOTATION
A fitting parameter in the DiBenedetto form for the glass transition temperature
AH2O Arrhenius coefficient for the foaming reaction
aT time shift factor
ATR attenuated total reflection
b fitting parameter in polymerization rate equation
CCO2 molar concentration of carbon dioxide per volume liquid
C1 William-Landau-Ferry constant
C2 William-Landau-Ferry constant
DSC differential scanning calorimeter
EH2O activation energy for the foaming reaction
Eξ activation energy for the polymerization reaction
EFAR a physically blown epoxy foam
ΔHrxn heat of reaction
k rate coefficient for an Arrhenius-type rate equation for polymerization
k0 rate coefficient for an Arrhenius-type rate equation for polymerization
kH2O rate coefficient for an Arrhenius-type rate equation for foaming reaction
m fitting exponent in polymerization extent of reaction equation
MCO2 molecular weight of carbon dioxide
n rate exponent for polymerization extent of reaction equation
N bubble nucleation time function
nCO2 number of moles of carbon dioxide
IR infrared
P pressure
q rate exponent for foaming reaction equation
R universal gas constant
R-component resin part of the foam kit
PMDI polyurethane foam based on polymethylene diisocyanate
Qexothermic heat evolved, measured in differential scanning calorimeter
SEM scanning electron microscopy
STP standard temperature and pressure
t time
tnucleation bubble nucleation time
tscale time scale for reaction
T absolute temperature
T-component curative part of the foam kit
Tg glass transition temperature
Tg0 glass transition temperature of the unreacted material
Tg∞ glass transition temperature of the fully reacted material

32
TMA Thermomechanical analysis
Vgas volume of gas
Vliq volume of liquid (polymer phase)
w fitting parameter to account for vitrification effects
WLF William-Landau-Ferry
  fitting exponent to account for vitrification effects
ξ extent of polymerization reaction
ρfoam density of the foam
ρgas density of the gas
ρliq density of the liquid
 gas volume fraction

33
 ACKNOWLEDGEMENTS
The authors would like to thank internal Sandia National Laboratories reviewers Ryan Keedy
and Dan Bolintineanu for their helpful comments. Sandia National Laboratories is a multi-
program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of
Lockheed Martin Corporation, for the U.S. Department of Energy’s National Nuclear Security
Administration under contract DE-AC04-94AL85000.

LITERATURE CITED
1. Lipshitz, S. D. and C. W. Macosko, “Kinetics and Energetics of a Fast Polyurethane Cure,”
J. Appl. Polym. Sci, 21, 2029-2039 (1977).

2. Hernández-Sánchez, F. and H. Vázquez-Torres, “Activation Energy for Urethane-Linkage

Formation. An Improved Calculation from the Differential Scanning Calorimetry Data,” J.
Polym. Sci.: Part A: Polymer Chemistry 28, 1579-1592 (1990).

3. Chern C-S, “Curing Kinetics of Polyurethane Reactions,” J. Appl. Sci., 40, 2189-2205
(1990).

4. Yang, Y. S. and L. J. Lee, “Polymerization of Polyurethane-Polyester Interpenetrating

Polymer Network (IPN),” Macromolecules 20, 1490-1495 (1987).

5. Prochazka, F., T. Nicolai, and D. Durand, “Dynamic Viscoelastic Characterization of a

Polyurethane Network Formation,” Macromolecules 29, 2260-2264 (1996).

6. Rochery, M. And T. M. Lam, “Chemorheology of Polyurethane. I. Vitrification and

Gelation Studies,” J. Polym. Sci: Part B: Polymer Physics, 38, 544-551 (2000). Erratum,
J. Polym. Sci: Part B: Polymer Physics, 38, 812 (2000).

7. Dimier, F., N. Sbirrazzuoli, B. Vergnes, and M. Vincent, “Curing Kinetics and

Chemorheological Analysis of Polyurethane Formation,” Polymer Eng. Sci., 44, 518-527
(2004).

8. Kamal, M. R., and S. Sourour. "Kinetics and thermal characterization of thermoset cure."
Polymer Engineering & Science 13.1 (1973): 59-64.

9. Sourour, S., and M. R. Kamal. "Differential scanning calorimetry of epoxy cure: isothermal
cure kinetics." Thermochimica Acta 14.1 (1976): 41-59.

10. Verhoeven, V. W. A., A. D. Padsalgikar, K. J. Ganzeveld, L. P. M. Janssen, “A Kinetic

Investigation of Polyurethane Polymerization for Reactive Extrusion Purposes,” J. Appl.
Polym. Sci. 101, 370-382 (2006).

11. Papadopoulos, E., M. Ginic-Markovic, S. Clarke, “A Thermal and Rheological

Investigation During the Complex Cure of a Two-Component Thermoset Polyurethane,”
J. Appl. Polym. Sci. 114, 3802-3810 (2009).

34
12. Chiacchiarelli, L. M., J. M. Kenny, L. Torre, “Kinetic and chemorheological modeling of
the vitrification effect of highly reactive poly(urethane-isocyanurate) thermosets,”
Thermochimica Acta 574, 88-97 (2013).

13. Marciano, Jorge H., Alfredo J. Rojas, and Roberto JJ Williams. "Curing kinetics of a rigid
polyurethane foam formulation." Polymer 23.10 (1982): 1489-1492.

14. Niyogi, D., R. Kumar, and K. S. Gandhi, “Modeling of Bubble-Size Distribution in Free
Rise Polyurethane Foams,” AIChE J. 38, 1170-1184 (1992).

15. Saha, M. C., B. Barua, S. Mohan, “Study on the Cure Kinetic Behavior of Thermosetting
Polyurethane Solids and Foams: Effect of Temperature, Density and Carbon Nanofiber,”
J. Eng. Materials and Technology 133, 011015-1 – 011015-6 (2011).

16. Baser, S. A., and D. V. Khakhar. "Modeling of the Dynamics of Water and R‐11 blown
polyurethane foam formation." Polymer Engineering & Science 34.8 (1994): 642-649.

17. Lefebvre, L., and Roland Keunings. "Finite element modelling of the flow of chemically
reactive polymeric liquids." International journal for numerical methods in fluids 20.4
(1995): 319-334.

18. Haberstroh, Edmund, and J. Zabold. "Modelling and simulation of the foam formation
process." Journal of polymer engineering 24.1-3 (2004): 377-390.

19. D. Seo; J. R. Youn; C.L.I. Tucker, “Numerical simulation of mold filling in foam reaction
injection molding," International journal for numerical methods in fluids 2003, 42, 1105.

20. Seo, Dongjin, and Jae Ryoun Youn. "Numerical analysis on reaction injection molding of
polyurethane foam by using a finite volume method." Polymer 46.17 (2005): 6482-6493.

21. Bikard, J., Bruchon, J., Coupez, T., & Silva, L., "Numerical simulation of 3D polyurethane
expansion during manufacturing process." Colloids and Surfaces A: Physicochemical and
Engineering Aspects 309.1 (2007): 49-63.

22. R. R Rao; L. A. Mondy; M. C. Celina; D. B. Adolf; J. M. Kropka; E. M. Russick.

Proceedings of the Polymer Processing Society 26th Annual Meeting, Proceedings, Banff,
2010.

23. R.R. Rao, L.A. Mondy, D.R. Noble, H.K Moffat, D.B. Adolf, P.K. Notz, Int. J. Numer.
Meth. Fluids, 68, Issue 11 ( 2012) 1362–1392,

24. Gibson, L. J.; M. F. Ashby. Cellular Solids: Structure and Properties, 2nd Edition,
Cambridge Solid State Science Series, Cambridge University Press, Cambridge, UK, 1990.

25. L. A. Mondy, R.R. Rao, Bion Shelden, Melissa Soehnel, Timothy O’Hern, Anne Grillet,
Mathew Celina, Nicholas Wyatt, Edward Russick, Stephen Bauer, Michael Hileman,
Alexander Urquhart, Kyle Thompson, David Smith, “Experiments to Populate and

35
 Validate a Processing Model for Polyurethane Foam: BKC 44306 PMDI-10,” SAND2014-
3292, Sandia National Laboratories, 2014.

26. A.T. DiBenedetto, “Prediction of the glass transition temperature of polymers: a model
based on the principle of corresponding states,” J. Polymer Sci., Phys.: Part B: Polymer
Physics, 25, 1949-1969 (1987).

27. R. Tesser, M. Di Serio, A. Sclafani, E. Santacesaria, “Modeling of Polyurethane Foam

Formation,” J. Appl. Polymer Sci., 92, 1875, 2004.

28. Dow Plastics, “Voronol 490” Form No. 109-01280 (Published 8/2001); “Papi 27” Form
No. PA-26-333-0301 (Published 3/2001).

29. SpecialChem,http://www.specialchem4polymers.com/tds/dabco-
dc197/airproducts/195/index.aspx

30. May, C. A., Ed., Epoxy Resins Chemistry and Technology, 2nd Edition, Marcel Dekker:
New York, 1988.

31. Adolf, D. B.1996, Measurement Techniques for Evaluating Encapsulant Thermophysical

Properties during Cure, Sandia National Laboratories, SAND96-1458.

32. Williams, Malcolm L., Robert F. Landel, and John D. Ferry. "The temperature dependence
of relaxation mechanisms in amorphous polymers and other glass-forming liquids,"
Journal of the American Chemical society 77, 3701-3707 (1955).

33. Ferry J.D. Viscoelastic Properties of Polymers 3rd Edition. John Wiley and Sons, Inc., pp
278-279, 1980.

34. B. Bilyeu, W. Brostow, K. P. Menard, “Epoxy thermosets and their applications III. Kinetic
equations and models,” J. Mat. Ed., 23,189-204 (2001)

35. Menard, K. & B. Cassel, Basics of Thermomechanical Analysis with TMA 4000, Perkin
Elmer, Inc., Technical Note, Shelton, CT, accessed 2016.
https://www.perkinelmer.com/lab-solutions/resources/docs/TCH_TMA_4000.pdf

36. Harran, D.; A. Laudouard. J. Appl. Polymer Sci. 1986, 32, 6043.

37. Celina, Mathew, “Approximate values of CO2 solubility and diffusivity based on O2
estimation,” Personal Communication, 2013.

36