The Reactive Extrusion of Thermoplastic Polyurethane

The Reactive Extrusion of Thermoplastic Polyurethane
Vincent Verhoeven
RIJKSUNIVERSITEIT GRONINGEN
The Reactive Extrusion of Thermoplastic Polyurethane
Proefschrift
ter verkrijging van het doctoraat in de Wiskunde en Natuurwetenschappen aan de Rijksuniversiteit Groningen op gezag van de Rector Magnificus, dr. F. Zwarts, in het openbaar te verdedigen op vrijdag 24 maart 2006 om 16:15 uur
door
Vincent Wilhelmus Andreas Verhoeven geboren op 24 mei 1973 te Waalre
Promotor: Beoordelingscommissie:
Prof. dr. ir. L.P.B.M. Janssen Prof. dr. A.A. Broekhuis Prof. dr. S.J. Picken Prof. dr. A.J. Schouten
ISBN 90-367-2520-8 ISBN 90-367-2521-6 (Electronic version)
1 1.1 1.2 2 2.1 2.2 2.3 2.4 2.5 3 3.1 3.2 3.3 3.4 3.5 3.6 3.7 4 4.1 4.2 4.3 4.4 4.5 4.6 4.7 5 5.1 5.2 5.3 5.4
INTRODUCTION POLYURETHANE EXTRUSION SCOPE OF THE THESIS AN INTRODUCTION TO EXTRUSION AND POLYURETHANES EXTRUSION THE CLOSELY INTERMESHING COROTATING TWIN-SCREW EXTRUDER POLYURETHANES LIST OF SYMBOLS LIST OF REFERENCES RHEO-KINETIC MEASUREMENTS IN A MEASUREMENT KNEADER INTRODUCTION EXPERIMENTAL SECTION THEORY OF MEASUREMENT OF THE KINETICS RESULTS CONCLUSIONS
7 7 8 9 10 9
30 32 35 35 37
19
43 52
39
LIST OF SYMBOLS LIST OF REFERENCES
51 53
A COMPARISON OF DIFFERENT MEASUREMENT METHODS FOR THE KINETICS OF POLYURETHANE POLYMERIZATION INTRODUCTION REACTION KINETICS EXPERIMENTAL RESULTS CONCLUSIONS LIST OF SYMBOLS LIST OF REFERENCES THE REACTIVE EXTRUSION OF THERMOPLASTIC POLYURETHANE INTRODUCTION THE MODEL RESULTS EXPERIMENTAL SECTION 55 55 57 59 65 80 79
81 83 83 84 92
96
5.5 5.6 5.7 6 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9
CONCLUSIONS REFERENCES
LIST OF SYMBOLS
115 117
114
THE EFFECT OF PREMIXING ON THE REACTIVE EXTRUSION OF THERMOPLASTIC POLYURETHANE INTRODUCTION MIXING EXPERIMENTAL SETUP MATERIALS RESULTS ADIABATIC TEMPERATURE RISE ANALYSIS CONCLUSIONS LIST OF SYMBOLS REFERENCES 119 119 120 123 122 123
133 135 134
125
6.10 APPENDIX 1 7 8 8.1 8.2 8.3 8.4 CONCLUSIONS APPENDIX SUMMARY SAMENVATTING DANKWOORD LIST OF PUBLICATIONS
136 139 143 143 155 149
157
1
1.1
Introduction
Polyurethane extrusion
Polyurethanes are mostly known for their widespread usage as building foam (PURfoam). However, their applications extend much further than just foam. Polyurethanes are in fact a broad class of polymers with the urethane bond as a common element. As for foam, thermoplastic polyurethane (TPU), the key player in this thesis, forms an important subclass in the field of polyurethanes. Thermoplastic polyurethane (TPU) is a versatile elastomer that is used in automotive products, electronics, glazing, footwear and for industrial machinery. For all these applications thermoplastic polyurethanes show a good performance regarding resistance to chemicals and hydrolysis, tear and abrasion resistance, lowtemperature flexibility and tensile strength. Thermoplastic polyurethane is a block copolymer that owes its elastic properties to the phase separation of so-called hard blocks and soft blocks. Hard blocks are rigid structures that are physically crosslinked and give the polymer its firmness; soft blocks are stretchable chains that give the polymer its elasticity. By adapting the composition and the ratio of the hard and the soft blocks, polyurethane can be customized to its application. As for most polymers, further tailoring of the material properties occurs through additives. TPU can be produced in several ways. The most common production method for thermoplastic polyurethane is reactive extrusion. For slow reacting systems, batch processes are used. An alternative process for extrusion is to cure premixed monomer pellets on a conveyor belt. Space requirements in combination with longer reaction times make the latter process less favorable. For the reactive extrusion process, the monomers are separately fed to the extruder by a precise metering system. In the extruder, reaction and transport take place, and the polymer formed is peletized at the die. These TPU-extruders are, to the best of our knowledge, mainly operated based on experience. This empirical approach is caused by the fact that flow and reaction are directly connected in an extruder, which makes the prediction of the outcome of a reactive extrusion process a difficult task. Moreover, the fact that numerous combinations of monomers and catalysts are used to produce a variety of TPUs does not improve the situation. Therefore, to control the extrusion process, a reliable extruder model in combination with reliable knowledge of the kinetics of the system used is highly desirable.
Chapter 1
1.2
Scope of the thesis
In the introduction the two key components of this thesis, extrusion and polyurethane, are discussed. An elaboration on these subjects is presented in the second chapter, giving more insight into the basics and relevant areas regarding polyurethane extrusion. Subsequently, the kinetics of the polyurethane reaction is addressed. The emphasis of this part of the thesis lies on the effect of mixing and temperature on the kinetics of the reaction. For many polyurethane applications, low-temperature no-mixing kinetic measurements suffice. However, considering the working range of an extruder, this approach may be insufficient. Due to the immiscibility of the monomers, the reaction will initially take place at the interface. Depending on the temperature and the mixing conditions, diffusion limitations may predominate. Because of this competition between diffusion and reaction, the measurements of the kinetics for TPU polymerization are best performed at the temperature and the mixing situation of the application for which the investigation is intended. To bring this idea into practice, a new kinetic measurement method is introduced in the third chapter, based on torque kneader experiments. In the fourth chapter, the results of these kneader experiments are compared with other kinetic methods. The attention then shifts to the extruder. A reactive extrusion model is presented in chapter 5, in which the relevant effects for polyurethane extrusion are taken into account. Special emphasis is put on the depolymerization reaction, which is an important factor in polyurethane extrusion. The effect of premixing on the extruder performance is presented in chapter 6. Finally, the conclusions of this thesis are presented in chapter 7.
2
2.1
An Introduction to extrusion and polyurethanes

Extrusion
Extruders have a widespread application in food and polymer technology. In general, extruders find their use in processing of medium to high viscosity materials that do not need a long processing time. Compounding of polymers, production of powder coatings and hot melts, paper pulp processing, and cooking extrusion of pasta, chips, pet food, and cereals are among others the working area of extruders. Extruders are even found to be useful for more exotic applications such as for production of explosives, ice cream manufacturing, and metal extrusion. The general working principle of an extruder is straightforward: a screw rotates in a closely fitted barrel; material is transported through the rotating action of the screw in the downstream direction. Extruders come in different forms, each with their own advantages. The classification of extruders is straightforward. First, there is the difference between single and twin-screw extruders. Based on costs, a single screw extruder is always first choice. However, for several applications single screw extruders are less suitable, which only leaves the choice for a twin-screw extruder. The most predominant inconvenience of a single screw extruder is the transport mechanism. Transport is only based on drag flow, which makes a single screw extruder sensitive to viscosity changes and slippage. Twin-screw extruders have this disadvantage to a much lesser extent. Twin-screw extruders come in different varieties; several types of extruders are shown in figure 2.1. More details on the benefits and limitations of every type of extruder can be found in Janssen (1), Rauwendaal (2), and Todd (3). For reactive processing, a closely intermeshing corotating twin-screw extruder is often the preferred choice. Due to the self-wiping action, the transport of material is largely independent of the viscosity of the material. Of course, this is an advantage for a reactive system, since the viscosity rises exponentially along the screw. Moreover, the high average shear-rate promotes a well-mixed reaction mass, and the diversity in screw build-up make a twin-screw extruder a versatile reactor, which can be tailored to its application.
Chapter 2
Figure 2.1
Different types of extruders, a) single screw, b) tangential extruder, mixing emphasis, c) tangential extruder, transport emphasis, d) closely intermeshing counterrotating, e) conical closely intermeshing counterrotating, f) closely intermeshing corotating (1).
In general, if we look at the extruder as a polymerization reactor, the benefits and disadvantages are well known. The high investment costs (expensive reactor volume), in combination with the unsuitability for time-consuming processes compete with a narrow residence time distribution, a fair heat transfer, no need of solvents and good mixing properties. Most important, in an extruder a one-shot polymerization and pellet forming process can be carried out. For several high-end polymers, as for polyurethane, the extruder is the preferred reactor.
2.2
2.2.1
The closely intermeshing corotating twin-screw extruder

Working principle
In a closely intermeshing corotating twin-screw extruder, material is transported from the feed zone to the die. The conveying mechanism in this type of extruder is similar to a single screw extruder. However, for the twin-screw extruder the seconds screw wipes the first screw, which prevents slippage and guarantees forward conveying (figure 2.2). Because of the requirement that one screw wipes the other, the screw cross section has a unique shape for a given diameter, pitch, centerline distance, and number of tips (parallel channels).
10
An introduction to extrusion and polyurethanes
Figure 2.2
Two closely intermeshing corotating screws.
Booy (4, 5) derived the mathematical expressions from which the geometry of fully wiped corotating twin-screw extruders can be calculated. Due to the constraints on the screw geometry, the screw has a relatively large channel width compared to the flight width. As a result, hardly any decrease of the channel area is found in the intermeshing zone between the two screws. Roughly speaking, a screw channel continues from one screw to the next, giving one continuous channel. Due to the multiple thread starts that are common practice for corotating extruders, several parallel channels exist; the number can be calculated from the number of thread starts (1).
Figure 2.3
Parallel channel representation of a corotating closely intermeshing twin-screw extruder (6).
A common way to represent the flow in a screw channel is related to the idea of an infinite channel. As shown in figure 2.3, the flow in a corotating intermeshing extruder can be envisaged as several parallel channels, with the barrel wall sliding
11
Chapter 2
as a infinite plate over the channels. In figure 2.3, the curvature of the channels is ignored, and the flow in and the geometry of the intermeshing zone is not captured completely in this way. The route the material travels in a channel is shown in figure 2.4.
barrelwall
b,z
z x
b,x
Figure 2.4
The helical flow pattern in a single channel.
Near the barrel wall, material flows in the positive x-direction (due to the movement of the wall) until it meets the upcoming flight. The material is then forced to the bottom of the channel (negative y-direction); at the bottom, the material flows back in the x-direction. This time, the presence of the flight-wall pushes the material upwards (y-direction) and this completes the cycle. Due to the z-component of the barrel wall velocity, the net flow of material is in the downstream direction of the channel; the material therefore follows a helical path. Experiments and 3D-simulations (7) confirm this flow pattern and show that at 2/3th of the channel height a stagnation point exist. 2.2.2 Energy considerations
This helical flow pattern has clear consequences for the temperature gradient in the channel. The material that resides at the center of rotation does not come close to the barrel wall, while other material passes the barrel wall regularly, exchanging heat with the barrel. Therefore, temperature gradients in the channel are inevitable, especially, since viscous dissipation and reaction heat have a dominant effect in the heat balance. This effect is particularly important for larger extruder diameters (D 5 cm). Still, due to the helical flow pattern, the heat transfer is much better than what would be expected for flow between two moving plates. To obtain a first estimate of the effect of reaction, viscous dissipation and heat transfer through the wall on the energy balance, a dimensionless number analysis can be made. Three
12
dimensionless numbers are relevant: DamkhlerIV (DaIV) number, the Brinkmann (Br) number, and the Graez (Gz) number (equation 2.1).
Da IV =
HR Q T D
heat of reaction conductive transport of heat
Br =
N2 D 2 T
viscous dissipation conduction of heat
( 2.1 )
Gz =
a L Q
conductive heat transfer convective heat transfer
For the reactive extrusion of polyurethane (for the system and extruder used in this thesis), an evaluation of these numbers shows that the heat of reaction is lower than the viscous dissipation (DaIV / Br < 1). Moreover, the extruder operates somewhere in between isothermal and adiabatic conditions (Gz 1). For more specific information, the energy balance of the extruder has to be solved. Due to the complicated flow pattern, only a fully developed three-dimensional flow model can take care of all effects. However, a more simple approach will give reasonable insight. Commonly, a one-dimensional heat balance over short sections of the extruder is used (chapter 5). 2.2.3 Flow behavior
As for the heat balance, the three-dimensional flow pattern in the screw channel must be condensed to a more simple equation, in order to estimate the filling degree and pumping characteristics of a corotating intermeshing extruder. A basic approach is to express the throughput of an extruder in a drag and a pressure flow term (8):
Q = Q drag + Q pressure = A N
B dP sin dL
( 2.2 )
Equation 2.2 states that the net throughput in an extruder equals the maximum drag flow capacity (AN) minus the pressure flow, which occurs in the opposite direction. The pressure flow is proportional to the pressure build-up capacity
13
Chapter 2
divided by the viscosity. Equation 2.2 can be derived from a momentum balance over a screw channel. The constants A and B are specific for an element type and represent the curvature of the channel. The A and B terms can be obtained from a geometrical analysis (1, 9, 10, 11) or through an experimental approach (12). Several effects are not taken into account by applying Equation 2.2: 1. 2. 3. 4. The leakage flows The effect of the intermeshing zone Non-Newtonian flow behavior The effect of radial temperature gradients (and resulting viscosity gradients). These phenomena cause a deviation of the linear dependence of the pressure drop on the rotation speed. Several measures can be taken to obtain a more precise description. 1. The leakage flows
The leakage flows can be taken into account by adapting equation 2.2:
Q = A N
B dP QL dL
( 2.3 )
Of all the leakage flows present in a corotating intermeshing extruder (1), the leakage over the flight predominates. The other leakage gaps, which are located near the intermeshing zone, are less important, due to the smaller leakage area, and because in this part the two screws rotate in the opposite direction, giving no net flow through the leakage gaps. The leakage over the flight can be introduced using a pressure and a drag flow term (13): Drag Pressure
v 3 P w+e R sin Q L = u 0 sin cos R + 2 L tan 12 flight e u = 2 (D + 2 R ) ( )
( 2.4 )
Due to the small gap size R, the pressure driven leakage flow would seem to be a small contribution to the leakage flow. However, for non-Newtonian fluids, the
14
leakage over the flight is of importance; the high shear rate over the flight results in a low apparent viscosity causing the pressure driven leakage flow to become important. 2. The effect of the intermeshing zone
To refine equation 2.3, the flow in the intermeshing zone can also be introduced. The intermeshing zone forms a local restriction in the channel; the material undergoes no net drag flow since it is not in contact with the barrel wall. Moreover, a small contraction of the channel area is present in the intermeshing area. Michaeli et al. (11) and Vergnes et al. (14) each came up with a solution to account for the intermeshing zone, based on a pressure driven flow through this zone. 3. Non-Newtonian flow behavior
To take into account the non-Newtonian behavior of the material in the screw channel, the average or the local shear rate must be known. For a one-dimensional approach, the average shear rate in the channel can be expressed as in equation 2.5:
& =
ND H
( 2.5 )
so that equation 2.3 becomes:

n
ND Q = A N H
B dP sin Q L k dL
( 2.3a )
A better estimate of the average shear rate can be obtained through a twodimensional analysis of the flow (x and z direction in figure 2.3), taking into account the actual channel geometry as for example was done by Michaeli et al. (11). However, no analytical equation appears in that case. With the approach of Potente et al. (15), based on single screw calculations of Tadmor and Gogos (16), this disadvantage is not present. 4. The effect of radial temperature gradients
A further refinement of the flow model, for example by taking into account the temperature gradients, results in two- or three- dimensional models.
15
Chapter 2
2.2.4
Kneading paddles
So far, all emphasis has been placed on the regular transport elements. One of the benefits of a corotating intermeshing extruder is its flexibility. Not only transport elements, but also numerous types of other elements can be applied in endless combinations to tailor the extrusion process. A survey of possible elements is for example presented by Todd (3). For this thesis, the most important class of elements (besides the transport elements) are the kneading blocks (figure 2.5). The main function of the kneading blocks is to enhance mixing. Kneading blocks consist of staggered kneading paddles. The stagger angle of the successive paddles and the width of the paddles can be varied. With a larger stagger angle, the forward conveying capacity diminishes, but the kneading action improves at the cost of more energy dissipation. The forward conveying capacity diminishes with a larger stagger angle because the leakage gaps between two paddles (figure 2.5) increases. Kneading paddles reorient the fixed flow lines that are present in the regular transport elements, which give a distributive mixing effect. Moreover, going from paddle to paddle, further distributive mixing takes place due to the staggering of the paddles (extra reorientation of the flow lines) and the backflow through the leakage gaps. Besides distributive mixing, the kneading paddles also promote dispersive mixing. Material is squeezed between two neighboring paddles, giving large extensional flow rates compared to the normal transport elements. In general, by widening the paddles width, the mixing emphasis shifts from distributive to dispersive mixing. By using wider kneading paddles, the material has less possibilities to escape when it is squeezed together, giving larger elongational and shear forces.
Figure 2.5
A kneading block for a corotating intermeshing twin-screw extruder.
16
Many investigations have been directed towards understanding and describing the flow and the mixing behavior in kneading blocks (7, 10, 13, 14, 17-23). The most straightforward way is to consider the kneading blocks as modified transport elements. In that case, the equations that are used to calculate the transport capacity of a transport element can be used, with some modifications:
Q = A N
B dP sin QL QL,stag dL
( 2.6 )
Compared to equation 2.2, an extra leakage flow is introduced due to the staggering of the kneading paddles. This extra leakage flow can be defined in different ways, as done by Potente et al. (10) or Meijer et al. (9). 2.2.5 The filling degree and residence time
Through residence time distribution measurements or modeling efforts, the residence time in an extruder can be determined. Especially for reactive extrusion, the residence time is an important parameter, since it is directly related to the yield that is obtained with the extrusion process. Corotating extruders are usually starved fed. Consequently, sections of the extruder are not completely filled. To calculate the residence time, the filling degree of the partially filled zones and the length of the fully filled zones must be determined.
Figure 2.6
A typical screw profile for a corotating intermeshing twin-screw extruder (1).
A typical extrude profile is shown in figure 2.6. As pointed out, partially filled zones alternate with fully filled zones along the screw. The filled regions are created by
17
Chapter 2
upstream elements that form local restrictions and create backpressure. Examples are reverse elements, 90 (non-conveying) kneading blocks, or, as a special case, the die. The length of each fully filled zone is dependent on the pumping characteristics of both the backpressure and forward-pressure creating screw elements. The pumping characteristics can for example be calculated using equation 2.2, or a modified version of this equation, depending on the desired accuracy and the element type under consideration. For the die, a different approach must be taken. The pressure over the die is very dependent on the die geometry. For cylindrical dies, the most straightforward equation is based on the flow in a tube:
P =
128 Q d4
L die
( 2.8 )
The second parameter that is important for calculating the residence time is the filling degree in the partially filled zones. A general expression gives:
f=
Q feed Q max
( 2.7 )
Qfeed is the feed rate of material. For Qmax, the AN-term of the right side of equation 2.2 may be used. In case other types of elements are taken into consideration, the A-factor for Qmax changes.
18
2.3
Polyurethanes
As explained in paragraph 1.1, polyurethanes are a group of polymers that have the urethane bond in common. Polyurethane can be regarded as a linear block copolymer as shown in figure 2.7. This segmented polymer structure can vary its properties over a wide range of strengths and stiffness by modification of its three basic building blocks: polyol, diisocyanate, and chain extender (diol). Essentially, the hardness range covered is that of soft jelly-like structures to hard rigid plastics. Material properties are related to segment flexibility, chain entanglement, inter chain forces, and cross-linking.
Figure 2.7
The basic unit in a urethane block-copolymer (24).
Evidence from X-ray diffraction, thermal analysis and mechanical properties strongly support the view that these polymers can be considered in terms of long (100 200 nm) flexible segments and much shorter (15 nm) rigid units which are chemically and hydrogen bonded together (24). The structure becomes oriented via extension as indicated in figure 2.8. The stretching of an elastomer proceeds by the stretching of the coiled flexible polyol segments while the hard segments stay bonded to each other.
19
Chapter 2
Figure 2.8
Flexible and rigid segments in a polyurethane elastomer.
Modulus-temperature data usually show at least two definite transitions, one below room temperature, related to segmental flexibility of the polyol and one above 100C due to dissociation of the inter chain forces in the rigid units. Multiple transitions may also be observed if mixed polyols and rigid units are present in the polymer structure. 2.3.1 Isocyanates
O C N C H
Figure 2.9
N C O
O C N C H H N C O
Structure of 4,4-MDI (left) and 2,4-MDI (right).
Only the diisocyanates are of interest for linear urethane polymer manufacturing, and relatively few of these are used commercially. The most important ones in elastomer manufacturing processes are 2,4- and 2,6-toluene diisocyanates (TDI), 4,4-diphenylmethane diisocyanate (MDI) and its aliphatic analogue 4,4dicyclohexylmethane diisocyanate. Also 1,5-naphtalene diisocyanate (NDI) and 1.6 hexamethylene diisocyanate (HDI) are used. The diisocyanates used in this research
20
are 4,4-diphenylmethane diisocyanate (4,4-MDI) and a mixture of 50% 4,4diphenylmethane diisocyanate (4,4-MDI) and 50% 2,4-MDI. The structures of these compounds are shown in figure 2.9. 2.3.2 Polyols
Although diisocyanates are the intermediates responsible for chain extension and the formation of urethane links, much of the ultimate polymer structure is dependent on the nature of the components carrying the groups with which the isocyanates react. An example component can be a simple short diol, as such was employed in the early work on linear polyurethanes (24). Linear polyurethanes of this type are crystalline, fiber-forming polymers but have a lower melting temperature than the corresponding polyamides, and none have become of real importance either as a synthetic fiber or as a thermoplastic material. However, replacement of the simple diols by polymeric analogues has resulted in an extensive commercial development. This arose from the finding that linear polyesters or polyester-amides, of molecular weights of about 2000 and carrying terminal OH groups, can react with hexamethylene diisocyanate (HDI) and toluene diisocyanate (TDI). Through a chain lengthening process, tough elastomeric or plastic materials can be formed, which can be cross-linked by using additional isocyanate. The original polyols used in PU elastomer synthesis are structurally simple and three classes have been recognized, namely polyesters, polyethers and more recently polycaprolactones. For elastomer synthesis, these are available in various molecular weights, and products in the range of 600-2000 g/mol are commonly used industrially. The polyol used in this research was a polyester-based polyol of the type P765 (Huntsman Polyurethanes), based on an ester of mono-ethylene glycol, di-ethylene glycol and adipic acid. The influence that different polyester backbones have on the properties of polyurethane elastomers is large. Tensile strengths and moduli depend largely upon the presence of a side chain in the polyester. For example, polyesters that contain methyl side chains give elastomers that have significantly lower tensile strengths than those from the linear polyesters. 2.3.3 Diols (chain extenders)
The flexible (polyol) blocks primarily influence the elastic nature of the product. In addition, they make important contributions towards the hardness, tear strength, and modulus. But chain extenders for example a diol like butanediol particularly
21
Chapter 2
affect the modulus, hardness and tear strength, and determine the maximum application temperature by their ability to remain associated at elevated temperatures. Rigid segments are usually formed by the reaction of diisocyanate with a glycol or a diamine. In this research mainly glycol is used as a chain extender, namely methyl-1,3-propanediol. 2.3.4 Polyurethane chemistry
In figure 2.10, the most common reactions that occur when making polyurethanes are shown (25). Figure 2.10 shows overall reaction schemes so no details on the order of the reaction can be concluded. For the production of thermoset polyurethane foam (PUR) reaction 5 is indispensable. For thermoplastic polyurethane (TPU) production, water is excluded, so that only reactions 1, 2, 3 and 4 can take place. For normal condensation polymerization, in which always a small molecule (mostly water) is formed, equilibrium between the forward and the reverse reaction can be prevented by removing this small molecule (e.g. evaporation of water). For all isocyanate reactions, this option is not present; therefore, the reverse reaction can have a substantial impact. For the polyurethane formation reaction (reaction 1), an equilibrium state has been demonstrated. Dissociation of the polyurethane bond has been observed with DSC and rheology (26). In addition, it was shown by Ando (27) that for a bulk system without catalyst and at temperatures between 180 and 220 C the molecular weight decreases with polymerization temperature. Ando (27) attributes this effect to the depolymerization reaction (i.e. the reverse of reaction 1). Which of the reactions shown in figure 2.10 take place during polyurethane production depends on the temperature, and the presence and the type of solvent and catalyst used. Solvent and catalyst can greatly enhance the rate of one (or sometimes more) reactions. Moreover, the temperature affects the reaction rate and the equilibrium of each of the reactions specified. Normally, the type and ratio of monomers and the type of catalyst is chosen in such a way that the polyurethane reaction will dominate. However, even the occurrence of a limited amount of side reactions may interfere with the final material properties. In the literature, some articles have been published that take the side reactions during polyurethane formation into account. However, most of the publications on polyurethane kinetics use the kinetics as input for modeling purposes (e.g. for reactive injection molding), and the side reactions are neglected. Moreover, for these systems the kinetics are very fast which makes a detailed analysis of the reaction difficult. In the next paragraphs, a short overview of the relevant reactions will be presented.
22
H (1) Urethane Formation: N C O + OH

Possibly catalyzed
N C O O O
(2) Isocyanurate Formation:
N C O O
N C
N C O
H (3) Allophanate Formation: N C O O + N C O
H O N C N C O O O C
(4) Uretidione Formation:
N C O
N C O
H (5) Urea Formation: H2 O + N C O N C O OH
H O
H
N C O
H N H + CO2
N C N
Figure 2.10
The most commonly occurring isocyanate reactions.
2.3.5
Reaction 2: Isocyanurate formation
At lower temperatures (up to 50C) and with N,N,N-pentamethyl dipropylene triamine (PMPT) as a catalyst, it was shown that up to 30% isocyanurate can be formed (28). HPLC measurements showed that allophanate appears as an intermediate during this reaction. A second publication of these authors (29) shows that the type of tertiary amino catalyst determines if and at what speed isocyanurate is formed. A mechanism for isocyanurate formation is proposed by Kresta et al. (30). A catalyst-isocyanate complex is formed in an equilibrium step;
23
Chapter 2
subsequently two isocyanate units are added. During the last step, a fourth isocyanate replaces the trimer that is formed at the catalyst site. However, this mechanism does not concur with the observations of Wong and Frish (28, 29) that allophanate acts as an intermediate for isocyanurate formation. Vespoli and Albetino (31) have fitted adiabatic temperature rise data for a MDI-polyol system with this mechanism. They assumed that only at a higher ratio of isocyanate to alcohol isocyanurate is formed. Sun et al. (32) used the mechanism of Kresta et al. (30) for modeling a RIM-process for thermoset polyurethane production. They observed during their ATR experiments that the isocyanurate activation energy is higher and the polyurethane reaction is slower. Therefore, at higher temperatures the isocyanurate formation predominates. Sun et al. (32) concluded further that urethane oligomers cause a diffusion limitation for the isocyanurate formation. A free-volume model was used to consider this effect. Isocyanurate formation is sometimes desirable because it enhances thermal and dimensional stability and decreases the combustibility and smoke production of the resulting polymer. Conditions that favor isocyanurate formation are a high isocyanate to alcohol ratio and the presence of certain types of catalyst (for instance tertiary amino catalysts like PMPT enhance isocyanurate formation). If these factors are not present, as is the case for the extrusion process presented in this thesis, isocyanurate formation will not be of importance. 2.3.6 Reaction 3: Allophanate formation
In contrast to the isocyanurate bond, which is still remarkably stable at 200C, allophanates dissociate more readily. Malwitz et al. (33) took a computational chemistry approach to calculate the rate of allophanate formation. They found an equilibrium temperature of 165C. According to their calculations, the rate of allophanate formation is slow without catalyst, but is quite considerable in the presence of catalyst. Generally, it is assumed that formation in bulk and without catalyst occurs only at temperatures higher than 120C (34, 35). Jhnson and Flodin (36) showed with NMR-study that in a non-catalyzed system at temperatures lower than 100C no allophanate is formed. They also stated that allophanate formation would only happen at higher temperatures. Imawaga et al. (37) measured reaction products of a bulk system without catalyst at 85C. No side products were found though it was stated that at higher temperatures side reactions may well occur. Dorozhkin et al. (38) reported a second order kinetic constant for allophanate formation: ln (k2) = 19 60 / RT.
24
The short list of publications on allophanate formation indicates that there is limited knowledge on this subject. Based on the publications as presented above, allophanate formation does not take place below 120C, but above this temperature, the formation rate can be substantial. For polyurethane extrusion, allophanate formation may therefore interfere with the polyurethane reaction. Allophanate formation interferes with the stoichiometric ratio of alcohol and isocyanate, resulting in a lower final molecular weight. Moreover, allophanate will give branched polymer chains at low concentrations and at high concentrations, even a cross-linked polymer network would result. In fact, for polyurethane production at high temperatures (> 150C, for example during extrusion), a constant amount of isocyanate will be present due to the reverse reaction. These free isocyanate groups can choose between a relatively low concentration of alcohol groups and a relatively high concentration of urethane groups. Depending on the reaction rate constants and the equilibrium constants of the urethane and the allophanate reaction, a gradual increase of allophanate groups may therefore occur when keeping polyurethane at a high temperature for a longer time. Hentschel and all showed this effect indirectly by rheological experiments (26). 2.3.7 Reaction 4: Uretidione formation
Uretidione formation (reaction 3) in most cases does not influence polyurethane extrusion. Because of its low equilibrium temperature, uretidione readily dissociates at normally used reactive extrusion temperatures. The two free isocyanate groups that appear upon dissociation will react further to form polyurethane. Problems with uretidione formation may arise when heating the isocyanate prior to the reactive extrusion. Uretidione is insoluble in isocyanate, so a precipitate will form. 2.3.8 Polyurethane kinetics
Reaction mechanism
O N C O + OH
k1 k2
H
k3
.. N C O: .. H O
-OH
.. .. N C O: .. .. H O
H N C O O + OH
Figure 2.11
Lewis base catalysis for urethane formation.
25
Chapter 2
Several studies have been conducted on polyurethane kinetics. Two reaction mechanisms are used as the basis for a kinetic equation: A Lewis acid catalyzed reaction and a Lewis base catalyzed reaction. Actually, uncatalyzed reactions do not exist for polyurethane formation, since the alcohol group itself works as a Lewisbase catalyst. The mechanism for the Lewis-base catalysis is shown in figure 2.11 (the alcohol group in this case is the base-catalyst), the mechanism for the Lewisacid catalysis is shown in figure 2.12 (35).
H H...A ]
ROH
k3
N C O + HA
k1 k2
.. N C O: ..
N C O O
+ HA
Figure 2.12
Lewis acid catalysis for urethane formation.
Tertiary amino catalysts (for example DABCO), are Lewis-base catalysts. It is clear from literature (39) that the transition metals (Co, Mn) form a complex with the isocyanate group while the post-transition metals (Sn, Sb, Pb) form a complex with the alcohol group. In the literature, if the catalyst complex is taken into account in the kinetic equation a Lewis-base catalyzed reaction is always assumed. The most elaborate kinetic equation (for metal-complex catalysis) has been proposed by Richter and Macosko (40). They used the mechanism in figure 2.11 with an extra equilibrium step: dissociation of the catalyst in Metal observed four limiting cases:
+
and Rest-. The resulting
kinetic equation did not have an analytical solution but Richter and Macosko (40)
d[NCO] = k f [Cat ] [OH] dt d[NCO] 0.5 0.5 = k f [Cat ] [NCO] [OH] dt d[NCO] = k f [Cat ] [NCO] [OH] dt d[NCO] 0.5 = k f [Cat ] [NCO] [OH] dt
Which equation prevails depends on the degree of dissociation of the metalcomplex and the degree of association of the metal+ and the isocyanate group. Of course, the k in these equations is a lump sum k that consists of a combination of
( 2.10 )
26
rate and equilibrium constants. Dissociation of the metal complex has not been mentioned in the literature on polyurethane catalysis. Steinle et al. (41) used the mechanism in figure 2.11 for an analytical rate equation. This equation has a hyperbolic form:
EC 1 1 R T TR
d[NCO] K C1 e = dt
[Cat ] [OH] [NCO] 1+ K C2 [OH]
( 2.11 )
The main assumption Steinle et al. (41) make is that EA of k2 is equal to EA of k3. The rate equation of Steinle et al. (41) is both used for uncatalyzed reactions and reactions with tertiary amines as a catalyst. No decisive evidence has been presented on the exact reaction mechanisms during the polyurethane bond formation. The developed kinetic equations are therefore quite general, without any deep knowledge on which intermediate steps are rate limiting, and what the activation energy of each step is. In practice, this knowledge does not seem to be necessary to describe reactive injection molding processes. However, with reactive extrusion, the experiments on the kinetics are performed at different temperatures than the reactive extrusion process is operated, which may give an incorrect extrapolation of the reaction rate constant. Most authors report that up to 50 % conversion, the kinetics follow a second order trend but at higher conversions different effects are observed. Both acceleration and deceleration of the reaction velocity have been reported. Acceleration is mostly ascribed to the autocatalytic effect of the polyurethane bond. However, this autocatalytic effect has never been quantified. Deceleration is attributed to diffusion effects which may become important (especially in bulk systems) at higher conversions. In case of diffusion limitation, the idea that the reactivity of a functional group is independent of the chain length is no longer valid. For relatively short chain lengths, Krl (42) has shown that a higher molecular weight causes a slower reaction rate, but this effect is only observable up to a carbon backbone of five units. The slowing-down of the reaction at high conversions is therefore not explained by his findings. However, the findings of Krl (42) could mean that for polyurethane polymerization the chain extender reacts faster than the polyol. This difference in reaction rate is hardly ever taken into account for bulk polyurethane polymerization. The underlying reason is that the experimental difficulties related to the tracking of the two species (chain extender OH and polyol -OH) in a fast reacting high-temperature bulk process are hard to
27
Chapter 2
resolve. Moreover, for many applications it is sufficient to be able to predict the overall reaction rate, since longer oligomers are rapidly formed. A further assumption for polycondensation kinetics is that the reactivity of a reactive group on a molecule is independent of whether another reactive group on the same molecule has reacted. With all these conditions in mind, a general rate equation for polyurethane polymerization can be written:
RNCO =
d[NCO] = RNCO, Uncat + RNCO, Cat = k f [NCO]n dt

E A,Uncat RT
( 2.12 )
with k f = A 0,Uncat e
+ A0
E A [cat]m e RT
In equation 2.12 a stoichiometric amount polyol and isocyanate is assumed. For an isothermal batch reactor, the isocyanate balance can be solved to give:
[NCO] = [NCO]0 1+ k f (n 1 [NCO]0 )
n 1
t 1 n
( 2.13 )
Often a second order rate equation is found to be valid for polyurethane polymerization, which gives for the isocyanate concentration:
[NCO] =
[NCO]0 1+ k f [NCO]0 t
( 2.14 )
The number and weight average molecular weight are related to the isocyanate concentration. The increase in number and weight average molecular weight in time for a second order reaction gives (43):
MN = Mrep (1+ [NCO] 0 k f (T,[cat]) t ) M W = Mrep (1+ 2 [NCO]0 k( T,[cat]) t)
( 2.15 )
In this equation Mrep is the molecular weight of a repeating unit and [NCO]0 is the initial isocyanate concentration.
28
As explained in paragraph 2.3.4, the reverse reaction of polyurethane formation occurs at higher temperatures. To incorporate the reverse reaction, the rate equation (equation 2.12) changes:
R NCO = with
d[NCO] = k f [NCO]2 k r [U] dt

E A
k f = [Cat]m A 0 e RT ,
kr =
kf
E A,eq
( 2.16 )
A 0,eq e and [U] = [NCO]0 [NCO]
R T
Depending on the reactor type, equation 2.16 can be solved analytically to give the isocyanate concentration as a function of time. The equilibrium constant can be expressed in several ways (43):
[U] eq MN MN Mrep k = K= f = = A 0,eq e 2 2 kr [NCO] eq Mrep [NCO] 0
E A ,eq R T
( 2.17 )
This equation can be used to calculate the effect of the reverse reaction.
29
Chapter 2
2.4
a A A0 B [Cat] D e EA f H HR k kf kr K L n n N [NCO] [NCO]0 m MN Mrep MW [OH] P/L Q R RNCO t T T u
List of Symbols
Thermal diffusivity Geometrical constant Reaction pre-exponential constant Geometrical constant Catalyst concentration Diameter Flight land width Reaction activation energy Filling degree of a not fully filled element Height of the screw channel Heat of reaction Power law consistency Forward reaction rate constant Reverse reaction rate constant Equilibrium constant Length Reaction order Power law index Rotation speed Concentration isocyanate groups Initial concentration isocyanate groups Catalyst order Number average molecular weight Average weight of repeating unit Weight average molecular weight Concentration alcohol groups Pressure gradient in the axial direction of the extruder Throughput Gas constant Rate of isocyanate conversion Time Temperature Temperature difference Circumference of the eight-shaped barrel 30 m2/s kg mol/kg s kgm mg/g m m J/mol m J/mol Pasn kg/mols 1/s kg/mol m 1/s mol/kg mol/kg g/mol g/mol g/mol mol/kg Pa/m kg/s J/mol K mol/kgs s K K m
[U] v v0 w Greek symbols R
Concentration urethane bonds Velocity Circumferential velocity of the screw Width of the screw channel
mol/kg m/s m/s m
Clearance between barrel and flight tip Shear rate Viscosity Pitch angle Heat conductivity Kinematic viscosity Density Intermeshing angle
m 1/s Pas W/mK m2/s kg/m3 -
&
Subscripts b Cat Die Eq L Uncat
Barrel wall Catalyzed Die Equilibrium Leakage Uncatalyzed
31
Chapter 2
2.5
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34.
List of References
L.P.B.M. Janssen, Reactive extrusion systems, Marcel Dekker Inc., New York, Basel (2004). C.J. Rauwendaal, Polymer extrusion, Hanser, Munich (2001). D.B. Todd, Plastic compounding, Hanser, Munich (1998). M.L. Booy, Polym. Eng. Sci., 18, 973 (1978). M.L. Booy, Polym. Eng. Sci., 20, 1220 (1980). W. Michaeli, A. Grefenstein and U. Berghaus, Polym. Eng. Sci., 35, 1485 (1995). D.J. van der Wal, Improving the properties of polymer blends by reactive compounding, Phd-Thesis, Rijksuniversiteit Groningen (1998). J. Mckelvey, Polymer Processing, John Wiley & Sons, New York (1962). H.E. Meijer, and P.H.M. Elemans, Polym. Eng. Sci., 28, 275 (1988). H. Potente, J. Ansahl and B. Klarholz, Int. Polym. Process., 9, 11 (1994). W. Michaeli, A. Grefenstein and U. Berghaus, Polym. Eng. Sci., 35, 1485 (1995). D.B. Todd, Int. Polym. Process., 6, 143 (1991). W. Michaeli, and A. Grefenstein, Int. Polym. Process., 11, 121 (1996). B. Vergnes, G. Della Valle, and L. Delamare, Polym. Eng. Sci., 38, 1781 (1998). H. Potente, J. Ansahl, R. Wittemeier, Int. Polym. Process., 3, 208 (1990). Z. Tadmor, and G. Gogos, Principles of Polymer Processing, John Wiley & Sons, New York, Brisbane, Chichester, Toronto (1979). T. Fukuoka, Polym. Eng. Sci., 40, 2524 (2000). V.L. Bravo, A.N. Hrymak, and J.D. Wright, Polym. Eng. Sci., 40, 525 (2000). J.L. White, and Z. Chen, Polym. Eng. Sci., 34, 229 (1988). A. Poulesquen, and B. Vergnes, Polym. Eng. Sci., 43, 1841 (2003). H. Werner, Chemie Ing. Techn., 49, heft 4 (1977). M.A. Huneault, M.F. Champagne, and A. Luciani, Polym. Eng. Sci., 36, 1694 (1996). G. Shearer, and C. Tzoganakis, Polym. Eng. Sci., 40, 1095 (2000). C. Hepburn, Polyurethane elastomers, Elsevier Applied Science, London, New-York (1992). J.M. Buist, and H. Gudgeon, Advances in polyurethane Technology, Elsevier (1968). T. Hentschel, and H. Mnstedt, Polymer, 42, 3195 (2001). T. Ando, Polym. J., 11, 1207 (1993). S.W. Wong and K.C. Frisch, Polym. Sci. Part A: Polym. Chem., 24, 2877 (1986). S.W. Wong and K.C. Frisch, Prog. Rub. Plast. Techn., 7, 243 (1991). J.E. Kresta and K.H. Hsieh, ACS Polym. Prep., 21, 126 (1980). N.P. Vespoli and L.M. Alberino, Polym. Proc. Eng., 3, 127 (1995). X. Sun, J. Toth, and L.J. Lee, Polym. Eng. Sci., 37, 143 (1997). N. Malwitz, Cell. Polym. III, int. Conf., Paper 18, 1 (1995). S.D. Lipshitz, and C.W. Macosko, J. Appl. Polym. Sci., 21, 2029 (1977).
32
An introduction to extrusion and polyurethanes 35. 36. 37. 38. 39. 40. 41. 42. 43. J.H. Saunders and K.C. Frisch, Polyurethanes chemistry and technology. Part 1, Chemistry, Interscience publishers (1962). K. Jhnson, and P. Flodin, Brit. Polym. J., 23, 71 (1990). O. Imawaga, F. Ishimaru, Y. Kurahashi and T. Yamada, Polym. React. Eng., 4, 47 (1996). K.J. Dorozhkin, V.J. Kimelblat, and J.A. Kirpikznikov, Vysokomol. Soed. A., 23, 1119 (1981). A. Petrus, Int. Chem. Eng., 11, 314 (1971). E.B. Richter, and C.W. Macosko, Polym. Eng. Sci., 18, 1012 (1978). E.C. Steinle, F.E. Critchfield, and C.W. Macosko, J. Appl. Polym. Sci., 25, 2317 (1980). P. Krl, J. Appl. Polym. Sci., 57, 739 (1995). G. Odian, Principles of Polymerization, John Wiley & Sons Inc., New York (1991).
33
Rheo-kinetic measurements in a measurement kneader
3.1
Introduction
To establish reliable kinetics of thermoplastic polyurethane polymerization is not a straightforward task. The monomers from which thermoplastic polyurethane is produced in general are poorly miscible. Therefore, a combination of diffusion and reaction determines the reaction rate observed for each measurement of the kinetics. Diffusion limitation may be noticeable during the initial part of the reaction and at high conversions. In the early phase of the reaction, mixing will enhance the observed reaction velocity, through improvement of the microstoichiometry and through enlargement of the contact surface of the immiscible monomers. At the end of the reaction, the mobility of the end-groups and of the catalyst is much lower due to the large polymer molecules that have formed. This limited diffusion at high conversions may also have an impact on the observed reaction velocity. As a consequence of the competition between diffusion and reaction, the measurement of the kinetics for TPU polymerization are best performed at the same temperature and the mixing conditions as occur in the application for which the kinetic investigation is intended. For instance, for reactive injection molding the reaction takes place at temperatures between 30C and 120C, the reaction mass initially experiences a high shear and after the injection the reaction mass remains stagnant. Adiabatic temperature rise experiments (ATR), which are performed under the same stagnant conditions, are for that reason best suited to establish the kinetics in reactive injection molding. Applying this requirement to reactive extrusion would mean that measurement of the kinetics should be performed under shear conditions and at high temperatures (150C-225C). These conditions are available in a rheometer and in a measurement kneader. However, both instruments are not specifically designed for measurement of the kinetics. Measurement kneaders, for instance, are mostly used for (reactive) blending of polymers as was done by Cassagnau et al. (1) or for rubber research (2). Both instruments have a drawback if they are used for measurement of the kinetics: in both instruments the extent of the reaction can only be followed indirectly through the increase in torque. In order to correlate the torque to the reaction conversion, a calibration procedure is necessary for which samples must be taken. Simultaneous measurement of conversion in the rheometer or kneader would make
Chapter 3
this sampling procedure superfluous. Unfortunately, no obvious method is available. An adiabatic method as applied by Lee et al. (3) or Blake et al. (4) is not apt, due to the lack of heat production at higher conversions. A combination of rheology with a spectroscopic method, for example with fiber optic IR or Raman spectroscopy, has not been reported yet for polyurethanes. The accuracy at high conversion is not sufficient, and a stagnant polymer layer may form on the measurement cell. If we return to the comparison between a rheometer and a kneader, a rheometer seems more suitable for measurement of the rheo-kinetics, since, in a rheometer, the viscosity can be measured directly. Nevertheless, a measurement kneader is preferred in this research. The reasons for this are: The mixing behavior in a kneader resembles the mixing behavior in an extruder more closely, with both dispersive and distributive mixing action and both simple shear and elongational flow. Highly viscous material can be processed more accurately in a kneader, because in a rheometer, constant shear experiments at shear rates that are comparable to those occurring in an extruder are sensitive to edge failure and demand a high torque. Sampling of a small amount of material does not disturb the measurements in a kneader, whereas rheology measurements are gravely affected by taking (several) samples. Temperature control in a kneader is straightforward. In a rheometer, temperature control becomes complicated at temperatures above 150C since both cone and plate must be heated in that case. There are several studies known in which the kinetics of TPU polymerization is measured under mixing conditions (3 - 8). All of these measurements were performed at relatively low temperatures (<90C) and mostly on cross-linking systems. Therefore, no high conversions could be reached, since the gellation temperature was reached reasonably early in the reaction (around 70% conversion). Methods for measuring the kinetics that do reach high conversions are largely zero-shear methods. As is the case for radical polymerization (9), little attention has been paid to the interaction between mixing and reaction in step polymerization. Often it is expected for step polymerization that shear does not have a major impact on the reaction velocity due to the relatively high mobility of the reactive end groups of a polymer chain. Malkin et al. (10), for instance, state
36
that any observed acceleration of the reaction speed for poly-condensation reactions can usually be ascribed to viscous heating of the reaction mass. Schollenberger et al. (11) performed the only study known to us in a measurement kneader. Unfortunately, no quantitative data were obtained in this study. So no reliable data on the kinetics exist on TPU polymerization in an extruder, although this is a large industrial process. Therefore, this chapter focuses on the acquisition of relevant data on the kinetics for extruder modeling. A new method is presented, which is based on performing experiments in a measurement kneader. In a kneader, the measurement conditions are more similar to those in an extruder in comparison to existing methods for measuring the kinetics. Quantitative kinetics and rheological data can be obtained through this method; moreover, the effect of mixing on the polymerization reaction can be investigated.
3.2
3.2.1
Experimental section
The kneader
The kneader used in this research was a Brabender W30-E measurement mixer. A picture of the non-intermeshing torque mixer is shown in figure 3.1. Two triangular paddles counter-rotate in a heated barrel. The barrel can be closed with a (heavy) plug. The volume of the kneader is 30 cm3.
back plate
kneading paddles plug
front plate
Figure 3.1
The Brabender measurement kneader.
The kneader is driven by a Brabender 650-E Plasticorder. Two heating elements in combination with two control thermocouples (one in the back-plate and one in the kneader section) keep the kneader on the set temperature (Tset). A thermocouple
37
Chapter 3
sticking in the non-intermeshing zone of the kneading chamber is used for the measurement of the temperature of the melt (Tmeasure). The torque and temperature development in the kneader can be followed by means of a data acquisition system. 3.2.2 Experimental method
Preparations before an experiment The TPU system for the experiments discussed in this chapter consisted of: A polyester polyol of mono-ethylene glycol, di-ethylene glycol and adipic acid (MW = 2200 g/mol, f = 2). Methyl-propane-diol (Mw = 90.1 g/mol, f = 2). A eutectic mixture (50/50) of 2,4 diphenylmethane diisocyanate (2,4-MDI) and 4,4 diphenylmethane diisocyanate (4,4-MDI). (Mw = 250.3 g/mol, f = 2). The percentage of hard segments was 24%. The reaction was catalyzed using bismuth octoate. Both the polyester polyol as the methyl-propane-diol were dried under vacuum at 60C and stored with molecular sieves (0.4 nm) prior to use. The isocyanate was used at 50C. Just before an experiment the polyol, diol, isocyanate, and catalyst were weighed in a paper cup and mixed, using a turbine stirrer at 2000 rpm for 15 seconds. Experience showed that this premixing was necessary to obtain reproducible results. About 30 grams of the premixed reaction mixture was transferred to the kneader with a syringe. The exact amount of reaction mixture was determined by weighing the syringe before and after filling the kneader. The kneader measurement was started upon filling. Sampling In order to relate torque to molecular weight, samples were taken and analyzed (see theoretical section). The sampling method consisted of removing the stamp of the kneader, collecting the sample with tweezers, followed by quenching the material in liquid nitrogen. After taking a sample the stamp was put back on the kneader; the whole sampling routine had a negligible influence on the torque during a very short period. In order to inactivate the still reactive isocyanate end-groups the samples were dissolved in THF with 5% di-butylamine. The samples were subsequently dried and used for size exclusion chromatography analysis. 3.2.3 Size Exclusion Chromatography (SEC)
Samples were analyzed for their molecular weight distribution by size exclusion chromatography (Polystyrene calibrated). The chromatography system consisted of
38
two 10 m Mixed-B columns (Polymer Laboratories) coupled to a refractive index meter (GBC RC 1240). The columns were kept at 30C. Tetrahydrofuran (THF) was used as mobile phase and the flow rate was set to 1ml/min. The molecular weight distribution was analyzed using Polymer Laboratories SEC-software version 5.1. About 25 mg of polymer was dissolved in 10ml of THF; the dissolved samples were filtered on 0.45-m nylon filters.
3.3
Theory of measurement of the kinetics
The objective of this study is to determine the reaction rate constant for the formation of the thermoplastic polyurethane under investigation. Therefore, the torque and temperature curves measured in the kneader must be translated into a time-dependent conversion curve. For condensation polymerization conversion, molecular weight (M) and viscosity () are related in a straightforward way. However, it is impossible to derive the conversion (p) directly from the viscosity. This is called the direct rheo-kinetic problem by Malkin (10). The relationship between viscosity and molecular weight has to be established first, before conclusions can be drawn on the reaction pattern (figure 3.2). In addition, there is a complicating factor in a measurement kneader. Due to the complicated flow profile in a kneader it is not immediately clear how the measured torque can be related to the viscosity. Nevertheless, a (simplified) flow analysis can tackle this problem. Subsequently, the relationship between the torque and the molecular weight can be established.
Chemistry
Kinetics
p (t)
M (p)
M (t)
Rheology
Figure 3.2 The rheokinetic scheme (10).
(M)
(t)
39
Chapter 3
3.3.1
Rheology basics
A simplified model of the kneader forms the basis of the flow analysis. The true geometry of the kneader is simplified as shown in figure 3.3.
Figure 3.3
A simplification of the flow geometry in the measurement kneader.
The shear stress can then be calculated using a flat-plate approach for which the paddle is considered stationary and the barrel moves with a velocity Vb. The shear stress () at the wall is then equal to:
ND = app = app M & H
(3.1)
The factor M can be calculated through a flow analysis, for which the height H is a function of the angular coordinate. The viscosity is written as the apparent viscosity (app), since for our polymeric material a Newtonian approach is inaccurate. The value of the torque acting on a paddle is opposite to the torque value experienced by the barrel wall, and is equal to the force acting on the wall times the lever arm.
Torque = ( Area Shear Stress ) Lever Arm = (DW ) (D / 2)

For two paddles, this equals:
(3.2)
Torque =
M 2D3 W N app = C N app H
(3.3)
40
C can be considered as a geometry factor. The manufacturer of the kneader gives a similar equation to correlate torque to viscosity, with the constant C equal to 50. Equation 3.3 shows that for a Newtonian fluid the torque is directly proportional to the viscosity of the material in the kneader. If we consider the polyurethane as a power-law liquid, equation 3.3 can be rewritten to:
Torque = C ' Nn 0
(3.3a)
The next step, necessary for tackling the direct rheo-kinetic problem is to correlate the viscosity of the polymer to its weight average molecular weight. It is well established experimentally as well as theoretically that for an entangled linear polymer:
= A(T) M W
3.4
(3.4)
A(T) is a proportionality-factor that is temperature dependent. For linear amorphous polymers A(T) can be described with a Williams-Landel-Ferry-equation (WLFequation) or with an Arrhenius-type of expression. In general, for a temperature less than 100C above the glass transition temperature (Tg) of the polymer, a WLFequation is preferable. For higher temperatures, an Arrhenius-type expression is best-suited (12). For polyurethanes, the value of Tg is dependent on the specific chemicals used but for most polyurethanes Tg does not exceed 320K (13). An Arrhenius-type of expression should therefore be suitable to describe the temperature dependence of viscosity for the temperature range under consideration (400 - 475K). If equation 3.3a and 3.4 are combined, the following equation results:
1
MW
Torque 3.4 = A ( T)
with
A' ( T) = A( T) C ' Nn
(3.5)
The torque is now related to the molecular weight. If the function A(T) is known, the weight average molecular weight versus time for the TPU-reaction can be calculated from the logged torque and temperature values. By analogy with the temperature dependence of the viscosity the temperature dependence of A(T) can be described using an Arrhenius-type equation:
41
Chapter 3
A ' ( T) = A 0
UA R T e
(3.6)
In order to relate torque to the molecular weight, the flow activation energy (UA) and pre-exponential factor (A0) in equation 3.6 must be known. These constants can be found through a calibration procedure. For this procedure, samples are taken from the kneader and analyzed for their molecular weight with size exclusion chromatography (SEC). Samples of different molecular weights and samples taken at different reaction temperatures are necessary for the procedure. The molecular weight can be calculated from torque and temperature (Tmeasure) values using equation 3.5 and 3.6. The calculated and measured molecular weight can be compared, and the optimal value for A0 and UA can be found through a least-square fitting routine. 3.3.2 Basics of the kinetics
From the molecular weight versus time curve, the kinetics of TPU-polymerization can be obtained. Although the exact reaction mechanism is more complex, the TPU polymerization reaction is often described successfully with a second order rate equation (14), as is described in chapter 2.
M W = Mrep (1+ 2 [NCO]0 k( T,[Cat]) t)
(2.15)
Equation 2.15 shows that the molecular weight increases linearly in time. Since Mrep and [NCO]0 are constants, the slope of the molecular weight versus time curve is proportional to the reaction rate constant k(T,[cat]):
dM W = 2 Mrep [NCO]0 k( T,[Cat]) dt
(3.7)
If A0 and UA are known, the torque versus time graph can be translated into a molecular weight versus time graph (equation 3.5). From the slope of this curve and by applying equation 3.7, the value of the reaction rate constant can be calculated. If experiments are performed at different temperatures and at a constant catalyst level, an Arrhenius-expression can be established for the reaction rate constant.
42
There is just one limitation. According to equation 2.15, the molecular weight will rise to infinity at longer reaction times. In practice, this will not happen. Several phenomena may cause a leveling off the molecular weight and torque values at longer reaction times: The initial ratio of alcohol groups to isocyanate groups will never be exactly unity. This stoichiometric imbalance will limit the maximum conversion. Chain scission. The long molecules that are present at longer reaction times are prone to scission due to shearing. Depolymerization (chapters 2.3.4, 2.3.8) Allophanate formation (chapter 2.3.6). The high concentration of urethane bonds together with the continuous presence of a small portion of free isocyanate groups due to depolymerization can give rise to allophanate formation. Allophanate formation causes branched molecules. Polydispersity will therefore increase but since also the stoichiometry of reactants is affected, the net effect on the molecular weight is not clear. Due to branching the A-factor in equation 4 may change. A last reason why MW will not rise to an infinite value is degradation. This will of course limit the maximum MW. All of these factors gain importance at longer reaction times and at higher molecular weights. Therefore, reliable data for the kinetics using the measurement kneader are best obtained during the initial stage of the reaction.
3.4
3.4.1
Results
A typical kneader experiment
Figure 3.4 shows a typical graph obtained for a kneader experiment. The torque and the temperature are shown as a function of time. As expected, the torque increases over time due to the polymerization reaction. The torque curve in figure 3.4 reaches a steady value after 15 minutes. After an initial drop due to the filling of the kneader, the temperature also rises steadily to a constant value.
43
Chapter 3
5 4 3 2 1 0 0 2 4 6 8 10 Time (min)
190
150
130
Figure 3.4
The torque and temperature versus the time in the measurement kneader. Tset = 175C, 80 RPM.
Clearly, viscous dissipation plays an important role in the kneader; the dissipated heat cannot be completely removed through the walls. In general, the measured temperature exceeds the set temperature (in figure 3.4 Tset = 175C). Analysis of the experimental curves shows that the temperature increase due to viscous dissipation (Tviscous = Tmeasure-Tset) is proportional to the torque value with a proportionality factor of 2 C / Nm. 3.4.2 The determination of the flow activation energy and the preexponential factor The torque-temperature graph can be converted into a molecular weight versus time graph using equation 3.5. To do so, the function A(T) must be known, which means that the flow activation energy (UA) and flow pre-exponential factor (A0) have to be established. To determine these constants, experiments were performed at four different set-temperatures (125, 150, 175, 200C). Every experiment was repeated three times; 4 to 5 samples were taken per experiment at different reaction times. The molecular weights of these samples were determined and obviously, the value of the torque and the temperature at the moment a sample was taken is also known. UA and A0 can now be established by fitting the measured molecular weight to equations 3.5 and 3.6, with UA and EA as the fit parameters. Figure 3.5 shows the resulting parity plot in which the measured molecular weight is plotted against the calculated one. For the whole range of molecular weights, the agreement is good. 44
Temperature (C)
Torque (Nm)
170
100000 90000 80000 Mw calculated 70000 60000 50000 40000 30000 20000 10000 0 0 20000 40000 60000 80000 100000 Mw m easured
Figure 3.5
Parity plot for the calculated and measured molecular weight.
The values obtained for UA and A0 are respectively 42.7 kJ/mol and 7.210-22 Nmmol3.4/g3.4 (see table 3.1). In general, thermoplastic polyurethanes have a much higher flow activation energy (100 - 200 kJ/mol) than is normally expected for linear polymers. The hard segments that are present in thermoplastic polyurethanes cause this effect. Hard segments are associated in hard domains and are physically cross-linked, which gives rise to a higher resistance to flow. Dissociation of the hard domains takes place at temperatures between 150C and 200C, depending on the composition of the polyurethane. Beyond that temperature, the flow behavior will be that of a normal linear polymer. However, for the polymer under investigation, the hard segments will dissociate at a much lower temperature. UA (kJ/mol) A0 (Nmmol /g
3.4 3.4)
42.7 7.210
-22
EA (kJ/mol) k0 (mol/kg K)
61.3 2.1810-6
Table 3.1
The flow and kinetic parameters for the TPU under investigation.
This is caused by the relatively low percentage of hard segments (24%) and the composition of the hard segments. The hard segments are built from a bulky chain extender and an isocyanate blend containing 50% 2,4-MDI. Steric hindrance, therefore, complicates association of the hard segments and improves the compatibility of the hard and soft segments. The flow activation energy found (42.7
45
Chapter 3
kJ/mol) confirms this expectation, as it falls within the expected range for linear polymers (15). This result implies that for the TPU under investigation the hard segments are molten and completely dissolved in the soft segments, already at 125C. 3.4.2 The determination of the reaction rate constant
The torque and temperature versus time curves of figure 3.4 can be translated into a plot of molecular weight versus time by applying equations 3.6 and 2.15. Figure 3.6 shows this plot for three repeated experiments at 175C. The lines represent the molecular weights as calculated from the torque and temperature and the dots are the measured molecular weights. The agreement between the three experiments is reasonably good. In general, the reproducibility was somewhat better at higher temperatures. Long reaction times in combination with higher molecular weights seemed to cause the reproducibility to become worse. The initial slopes are straight, which supports the second order assumption of the rate equation.
120000
90000
Mw (g/mol)
60000
30000
0 0 4 8
Time [min]
Figure 3.6
The weight average molecular weight versus time in a measurement kneader. Tset = 175C, 80 RPM.
The reaction rate constant can be derived from the relation between molecular weight and time by determining the initial slope of the curves (e.g. in figure 3.6 the average slope between 0 and 4 minutes). As stated earlier, the initial slope gives the most reliable information on the kinetics. In table 3.2, the different slopes with their confidence intervals are shown as well as the value for the reaction rate 46
constant k. The reaction rate-constant is calculated using equation 3.7. It increases, as expected, with increasing temperature. The temperature in table 3.2 is the measured temperature, Tmeasure. Since the kneader does not operate completely isothermally, there is always a temperature range over which the slope is determined. The mentioned temperature, Tmeasure, is the average temperature over which the slope is measured. This temperature range never exceeded 5C. Temperature (C) 194.3 Slope (MW/min) 42154 47070 44246 17030 173.4 17684 18316 7134 149.6 6546 7784 2172 123.3 2416 2332
Table 3.2 The slopes and the kinetic results obtained from the kneader experiments.
Average slope / 1000 (MW/min) 44 +/- 6
k (kg/mol s)
0.37 +/- 0.06
17.6 +/- 1.6
0.147 +/- 0.014
7.2 +/- 1.6
0.060 +/- 0.012
2.3 +/- 0.4
0.0192 +/- 0.002
Now, the kinetic constants can be derived from an Arrhenius plot (figure 3.7). The values obtained for EA and k0 are respectively 61.3 kJ/mol and 2.18e6 mol/kg K (see also table 3.1). Three conclusions can be drawn from figure 3.7 and table 3.1. First, the straight line in the Arrhenius-plot is an extra confirmation that the second order rate equation holds for the temperature range considered. Secondly, the value of EA falls within the range reported for TPU-polymerization (30-100 kJ/mol). The scatter in activation energies reported in the literature are caused by the different catalysts and chemicals used. Finally, the plot shows that within the experimental uncertainties that are inevitable for measurement kneader experiments, quantitative kinetic and rheological results can be obtained.
47
Chapter 3
0.002 0
0.0021
0.0022
0.0023
0.0024
0.0025
0.0026
-1 ln(k) (mol/kg s)
-2
-3
-4
-5 1/T (1/K)
Figure 3.7
The Arrhenius-plot for the kneader experiments.
3.4.3
Evaluation of the kinetic model
Model predictions are compared to experimental data in figure 3.8 in order to check the correctness of the obtained kinetic parameters. The slopes of the model prediction and of the experimental results are in good agreement with each other, which is a confirmation of the data on the kinetics. However, both at the start and near the end of the reaction, the model and experiment do not coincide. A closer look at the starting point of the reaction reveals that the initial molecular weight is much higher than anticipated. For this reason, the model equation (equation 2.15) is adapted in figure 3.8, to correct for the initial high molecular weight:
MW = 17000 + Mrep (1+ 2 [NCO]0 k( T,[cat]) t)

The value of 17000 for the molecular weight at t=0 is for all temperatures the same, and is fitted to the experimental curves. This correction is needed, since, at the start of the measurement, the reaction has already started due the premixing procedure. However, the molecular weight at the start of the measurement is unexpectedly high. A calculation learns that, with the kinetic constants obtained in this research, the molecular weight after the premixing procedure should not exceed 1500. The difference corresponds to an observation that other authors (16, 17) have also made for TPU-polymerization. The initial low-viscosity part of the reaction proceeds much faster than the last high-viscosity part of the reaction. This observation has been verified through ATR-experiments for this system (data not
48
shown). With an initial temperature of 50C and the same catalyst level as for the kneader experiments, the reaction reaches a conversion of about 80-90% within 30 seconds.
150000
120000
1 7 5 C
Mw [g/mol]
90000
2 0 0 C
150C
60000 30000
1 2 5 C
0 0 4 Time [min] 8 12
Figure 3.8
The measured and calculated weight average molecular weight versus time.
The explanation of the tremendous decrease in the observed reaction velocity at higher conversions falls under the term diffusion limitation. As soon as high molecular weight material is formed the mobility of the catalyst or the end groups decreases, which causes a decrease in the observed reaction velocity. The exact nature of this phenomenon cannot yet be understood due to the limited range of these experiments. This problem will be the subject of further experimental research. At the end of the reaction, the experimental molecular weight levels off to a steady value. It is improbable that the initial stoichiometric deviation of at most 0.2% is the cause of this. An imbalance of 0.2% in stoichiometry leads to an equilibrium molecular weight of 350,000, which is much higher than the maximum molecular weight reported for this investigation. For the two high temperature runs, it is very probable that depolymerization has a major impact on the last part of the reaction and that, therefore, the reverse reaction is the predominant cause of the leveling of the MW-curve. For the two low temperature runs, the situation is less distinct. Figure 3.8 shows that for the 200C and 175C experiments a higher temperature leads to a lower equilibrium molecular weight. This trend is hardly visible for the 150C run
49
Chapter 3
and not visible at all for the 125C run, because these runs are not completed within the time shown. Longer reaction times are here necessary to get to an equilibrium situation. Unfortunately, the results at long reaction times are less reproducible. The color of the polymer coming out of the kneader is light brown/yellow but deepens at longer reaction times. Degradation, therefore, interferes with experiments that last longer. An obvious indication of allophanate formation has not been found. The kinetic model obtained in this research appears to have a limited validity. Still, an important part of the reaction is captured with this model. The initial, fast reaction takes only five percent of the total reaction time. Therefore, to predict the necessary residence time in an extruder the kinetic model obtained in this study is indispensable. However, an expansion of the model is desirable. At low conversions, the reaction proceeds much faster than the measured data indicate. On the contrary, at very high conversions the reaction stops, while the model for the kinetics predicts a continuous increase of the molecular weight. For the low conversion part, adiabatic temperature experiments need to be performed to get the kinetic constants for this part of the reaction. Subsequently, these data can be combined with the data obtained from the kneader in order to complete the model of the kinetics. For the very high conversion part of the reaction, depolymerization needs to be taken into account. Future experimental work will be directed towards depolymerization and low conversion experiments, to complement the present results.
50
3.5
Conclusions
Investigations on the kinetic of TPU polymerization, performed in a measurement kneader, show that quantitative kinetic and rheological data can be obtained using this method. The method has advantages over other measurement methods since the reactants are mixed during the experiment, mimicking real processing conditions. Therefore, for applications where the reaction takes place under mixing conditions, as is the case for reactive extrusion, the parameters obtained for the kinetics will be more accurate. Besides, the effect of mixing on the polymerization reaction can be investigated using this method. The kinetic data obtained prove that a second order reaction can be used to describe TPU polymerization. The experiments indicated that a fast initial reaction is followed by a slower high conversion part of the reaction. At the end of the reaction, the molecular weight levels off due to depolymerization and degradation. More experiments are necessary to elucidate these effects. Because of the complex geometrical form of the kneader, the viscosity-value obtained with a measurement kneader is not very accurate. Therefore, no attempt has been made to correlate the torque values to viscosity values. Nevertheless, the activation energy of flow could be established. The activation energy of flow falls within the range expected for linear polymers, which indicates that the hard segments are completely dissolved in the soft segments.
51
Chapter 3
3.6
A0 A(T) A(T) [Cat] C, C D EA
List of symbols
Flow pre-exponential constant Empirical constant which relates viscosity to MW Empirical constant which relates torque to MW Catalyst concentration Geometry factor of the kneader Diameter of barrel Reaction activation energy Average distance between barrel and paddle Reaction pre-exponential constant Average weight of repeating unit Weight average molecular weight Power law index Rotation speed Concentration isocyanate groups Initial concentration isocyanate groups Gas constant Temperature Glass transition temperature Measured temperature of material in kneader Set temperature of the kneader Torque Time Flow activation energy Barrel velocity Width barrel Nm(mol/g)3.4 Pas(mol/g)3.4 Nm(mol/g)3.4 mg/g m3 m J/mol m mol/kg s g/mol g/mol 1/s mol/kg mol/kg J/mol K K K K K Nm s J/mol m/s m
H
k0 Mrep MW n N [NCO] [NCO]0 R T Tg Tmeasure Tset Torque t UA Vb W Greek symbols
&
app 0
Shear rate Apparent viscosity Consistency Shear stress
1/s Pas Pasn Pa
52
3.7
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17.
List of references
P. Cassagnau, F. Mlis, and A. Michel, J. Appl. Polym. Sci., 65, 2395 (1997). A. K. Maity, and S. F. Xavier, Eur. Polym. J., 35, 173 (1999). Y.M. Lee, and L.J. Lee, Intern. Polym. Process., 1, 144 (1987). J.W. Blake, W.P. Yang, R.D. Anderson, and C.W. Macosko, Polym. Eng. Sci., 27, 1236 (1987). D.S. Kim, M.A. Garcia, and C.W. Macosko, Intern. Polym. Process., 13, 162 (1998). X. Sun, J. Toth, and L.J. Lee, Polym. Eng. Sci., 37, 143 (1997). J.M. Castro, C.W. Macosko and S.J. Perry, Polymer. Comm., 25, 82 (1984). R. John, N.T. Neelaqkantan, and N. Subramanian, Thermochim. Acta, 179, 281 (1991). M. Cioffi, K.J. Ganzeveld, A.C. Hoffmann, and L.P.B.M. Janssen, Polym. Eng. Sci., 42, 2383 (2002). A. YA. Malkin, and S. G. Kulichikhin, Rheokinetics, Hthig & Wepf, Heidelberg, Germany (1996). S. Schollenberger, K. Dinbergs, and F. D. Stewart, Rub. Chem. Tech., 55, 137 (1981). C. W. Macosko, Rheology, VCH Publishers Inc., New York (1993). J. Brandrup, E. H. Immergut, A. Abe, and D. R. Bloch, Polymer handbook, Wiley, New York (1999). C. W. Macosko, RIM - Fundamentals of Reaction Injection Molding, Hanser, Munich (1989). D. W. Van Krevelen, Properties of Polymers, Elsevier, Amsterdam (1990). T. Hentschel, and H. Mnstedt, Polymer, 42, 3195 (2001). X. Sun, and C. S. P. Sung, Macromolecules, 29, 3198 (1996).
53
A comparison of different measurement methods for the kinetics of polyurethane polymerization
4.1
Introduction
In the previous chapter a method is described which can be used to measure the kinetics of polyurethane polymerization for reactive extrusion purposes. Compared to common methods the method described offers in principle the advantage that measurements can be performed at high temperatures under mixing conditions, mimicking extrusion conditions. In the current chapter, this is investigated by comparing the results of other measurement methods for kinetics with the results of the kneader experiments. Technique Conversion range Low (<98%) Titration FT-ir ATR SEC NMR Fluorescence Rheometry
*
Mixing Yes / No No No Yes

*
Temperature Low <60C + + + + + + + + + + Middle 60140C High >140C
High (>98%)
+ + + + + + +
No No No Yes
Only premixing
Different kinetic measurement techniques.
Table 4.1
Several techniques have been used for the acquisition of data on kinetics (1, 2, 3, 4). Commonly used methods are titration, Fourier-transform infrared (FT-IR), adiabatic temperature rise (ATR) and size exclusion chromatography (SEC). Less common methods are fluorescence and NMR measurements. Unfortunately, the method applied often poses limits to the reaction conditions. In general, the reaction should not be too fast for all of these methods. Therefore, it is often necessary to keep the temperature and catalyst level low for the measurement of the kinetics. This limits the predictive window of the investigation, as reactive processing will usually occur
Chapter 4
at high temperature and catalyst loading. In table 4.1, several techniques are compared. At first, the division of the conversion range in table 4.1 seems a little peculiar since the transition between low and high conversion is set at 98%. However, at this conversion the methods measuring the decrease of reactive groups become imprecise, while the methods that depend on the size of the molecules become more accurate above 98% conversion. If we now look at the extrusion process for polyurethane polymerization, the monomers are fed to the extruder at a temperature of 60 - 80 C. The temperature of the reaction mass increases rapidly in the first part of the extruder, mainly due to the fast exothermic reaction. Heat transfer through the wall and viscous dissipation are still of minor importance. For this part of the reaction, the reaction conditions more or less mimic adiabatic temperature rise measurements, although no mixing is present during adiabatic temperature rise experiments. However, the situation changes as soon as high molecular weight material appears. At that moment, the reaction velocity will have slowed down considerably (due to the second order nature of the reaction) and relatively little reaction heat will be generated. Furthermore, the temperature of the reaction mass will be well over 160 C. In this regime, ATR-experiments will give a poor prediction of the reaction kinetics since the small heat of reaction will give a large error in the ATRmeasurements. Methods based on molecular weight measurements, such as size exclusion chromatography or rheology, are more suitable in this situation. Therefore, two different measurement methods seem necessary to establish the kinetics for the modeling of the polyurethane polymerization in an extruder. Of course, this is only the case if a different reaction temperature and different mixing condition result in a different behavior. In paragraph 4.2, this subject will be discussed in more detail. Nevertheless, in the few studies on thermoplastic polyurethane extrusion that are known in literature (5 - 9) only a single method was used to measure the kinetics. Either adiabatic temperature rise measurements or size exclusion chromatography are used in these studies, inevitably leading to the described errors. The importance of these errors was investigated experimentally; the results are described in this chapter. The different methods will be compared with respect to the Arrhenius-behavior, the influence of the catalyst and the effect of mixing. Three different methods are surveyed: adiabatic temperature rise, size exclusion chromatography and kneader measurements. Two different thermoplastic polyurethane systems are investigated in order to further validate the number of
56
A comparison of different kinetic measurement methods
measurement methods that are required to determine the kinetics of polyurethane polymerization for reactive extrusion purposes.
4.2
Reaction Kinetics
For polyurethane polymerization, a few key phenomena may lead to a change in the observed activation energy and reaction rate with temperature and mixing: Different rate limiting steps may dominate at different temperatures, giving a change in activation energy and reaction rate. As explained in chapter 2.3.8, the multi-step reaction mechanism for the urethane formation is often condensed in a second order rate equation. This simplification may lead to erroneous extrapolation of the kinetics. Miscibility of the monomers. Due to the incompatibility of the monomers, an interfacial reaction will initially take place. If and how this affects the reaction will be described in paragraph 4.2.1. Diffusion limitation at high conversion. The reaction may slow down due to the appearance of large molecules. To what extent this affects the reaction will be described in paragraph 4.2.2. Phase separation. Phase separation of hard and soft segments (as described in paragraph 2.2) may give differences in local concentration of reactive groups. Moreover, the rigid hard segments may slow down the reaction rate by restricting the mobility of the molecules. Depolymerization. When the reverse reaction occurs (paragraph 2.3.4), as is the case at elevated temperatures, the observed reaction velocity will be slower. If this is not accounted for, an extrapolation of the result on kinetics will results in erroneous predictions. 4.2.1 Miscibility of the monomers
Due to incompatibility of isocyanate and alcohol molecules, the reaction will take place on and near the interface, and interfacial effects will influence the reaction. These interfacial aspects of polyurethane polymerization have been investigated in several publications (10 - 12). The starting point of these investigations was to evaluate the effect of impingement mixing, since many polyurethane products are made through a reactive injection molding processes where generally impingement mixing is an important process step. Kolodziej et al. (13) found that impingement mixing gives a dispersion with droplets that are still quite large (> 100 m). An increase of the Reynolds number above 200 did not seem to decrease the droplet 57
Chapter 4
size any further. This droplet diameter is far too high to result in a kinetically controlled reaction. A second process is necessary to overcome these limitations. This second (fast) mixing process seems to be related to surface instabilities. Machuga et al. (14) confirmed the observation of other authors that the polyol disappears more rapidly into the isocyanate than could be explained by pure diffusion. They found that the dimers that are formed on the boundary layer of the isocyanate and the polyol play an important role in this process. Probably, the urethane groups of these dimers undergo H-bond interactions with the isocyanate molecules across the border, resulting in strong surface destabilizing forces. It was found that the initial growth of the interfacial zone was independent of the monomers used. However, the further growth of this zone appeared to depend on the viscosities of the species that were present. Rigid oligomer molecules, a fast reaction, or the use of a crosslinking system limited the growth of the interfacial zone, which results in a diffusion controlled reaction. The effect of catalyst on the interfacial process is not clear. Wickert et al. (12) observed a much finer dispersion with catalyst than without, while Machuga et al. (14) detected no difference between catalyzed and uncatalyzed experiments. 4.2.2 Concept of functional group reactivity independent of molecule size
Another phenomenon that can have an effect on the polyurethane reaction is the concept of functional group reactivity independent of molecule size. For condensation polymerization reactions, it is normally assumed that the reaction rate constant and the reaction mechanism are constant for the entire reaction (15). The size of the molecules attached to a reactive group has no influence on the reaction rate. In other words, possible diffusion limitations will have no effect. To explain this it is assumed that a reactive group can be in two states: colliding with a different reactive group, or diffusing to a next reactive group. If a long molecule is attached to the reactive group, the diffusion time is longer, but the collision time is also longer. A reactive group will switch many times between these states before it actually reacts; therefore, the length of a molecule will not have a net effect on the reaction rate. This hypothesis is applied successfully in many cases. However, the theory does have a limitation; it does not hold for very long molecules or for very fast reactions. The theory has been verified with rather slow reacting systems (t reaction > 100 minutes). The polyurethane reaction is much faster, especially at higher temperatures. Whether this will result in a reaction that is diffusion limited can be verified experimentally.
58
4.3
4.3.1
Experimental
Chemicals used
Two different polyurethane systems were used in this investigation. The difference between both systems is the type of chain extender and the type of isocyanate used. The two systems were selected on the basis of the difference in compatibility of the chain extender and the isocyanate. This difference is expected to give a different behavior upon mixing. Where system 1 is a common TPU system, system 2 is easier to handle due to the liquid state of the isocyanate at room temperature. Both systems have the same amount of hard segments (24.0 %) and use the same catalyst (bismuth octoate). For all experiments, the pre-treatment of the monomers was as described in paragraph 3.2.2. System 1: A polyester polyol of mono-ethylene glycol, di-ethylene glycol and adipic acid (MW = 2200 g/mol, f = 2) 1,4 butanediol (Mw = 90.1 g/mol, f = 2). 4,4 diphenylmethane diisocyanate (4,4-MDI). (Mw = 250.3 g/mol, f = 2).
System 2 (the same system as in chapter 3): A polyester polyol of mono-ethylene glycol, di-ethylene glycol and adipic acid (MW = 2200 g/mol, f = 2). Methyl-propane-diol (Mw = 90.1 g/mol, f = 2). A eutectic mixture (50/50) of 2,4 diphenylmethane diisocyanate (2,4-MDI) and 4,4 diphenylmethane diisocyanate (4,4-MDI). (Mw = 250 g/mol, f = 2). Although the difference is not very large in the chemicals used in both systems, the differences that do exist may well result in a different reaction pattern. The following properties are affected:
59
Chapter 4
The polyol and diol are more compatible for system 2 than for system 1; therefore, the chain extender will dissolve at a lower temperature in system 2.
The hard segments in system 1 will crystallize more readily. Both the differences in chain extender and in isocyanate contribute to that. In system 2 a methyl group on the chain extender will hinder the formation of a layered structure of hard segments. In addition, the non-linear 2,4-MDI that is present in system 2 will also be an obstacle for the crystallization of the hard segments.
The compatibility of hard and soft segments in system 2 is also different to that of system 1. The use of methyl-propane diol as a chain extender in system 2 may influence the solubility of the hard and soft segments in a positive way.
The polymer molecules formed are generally assumed to adapt a different conformation, depending on the system. While system 1 produces a completely linear molecule, the polymer molecules in system 2 will adopt a more staggered/coiled structure, due to presence of non-linear 2,4 MDI.
The reactivity of the end groups of both systems may differ. We expect that the isocyanate group of 2,4-MDI that is placed in the ortho position will have a comparable reactivity to that of an isocyanate group in the para position. However, the approachability of the isocyanate group in the ortho position will be less due to steric hindrance. Therefore, the reactivity of the ortho-positioned isocyanate group may be lower than of the parapositioned isocyanate group. This difference in reactivity may lead to a lower overall reaction velocity.
4.3.2
Adiabatic Temperature Rise experiments
Adiabatic temperature rise (ATR) is a common method to measure the kinetics for polyurethane polymerization. With this method, the polyurethane kinetics at relatively low conversions and relatively low temperatures can be investigated. Many authors have described the experimental procedure for ATR measurements (1). The adiabatic reactor consisted of a paper cup (diameter = 5cm) surrounded by a layer of urethane foam for insulation. The reactor could be closed with a lid. The lid was equipped with a thin Copper Constantine thermocouple that stuck in the middle of the reaction mass when the lid was closed. The reaction mass was stirred with a turbine stirrer with a diameter of 4 cm. 200 grams ( 1 %) of material was used per experiment. To start an experiment, the necessary amounts of polyol and diol were
60
weighed in the reactor and mixed for 60 seconds with a turbine stirrer at 600 RPM. Care was taken to keep the temperature of the mixture above 60 C, since demixing will take place at lower temperatures for both systems. The proper amount of catalyst was added with a syringe, and the polyol mixture was stirred for another 30 seconds. Finally, the proper amount of isocyanate was added with a syringe, and the reaction mass was stirred at 1500 RPM for 15 seconds. The cover was put on top of the reactor and the measurement was started. Analysis of ATR results In order to derive kinetic data from the ATR experiments, a simplified heat balance (equation 4.1) and rate equation (equation 2.12) were solved simultaneously (3, 11). For the heat balance, quasi-adiabatic conditions were assumed, since the reactor was not completely adiabatic for the time period under investigation. Depending on the reaction time, up to 4 % of the total reaction heat generated during the reaction was lost to the surroundings. The heat transfer coefficient h* was obtained by fitting the cooling curves of several experiments, using equation 4.1. We took the density and the specific heat to be constant over the whole measurement range. Although both the specific heat and the density are somewhat dependent on the temperature, the temperature effects of both constants counteract, so that the net effect is negligible (< 5%). A non-linear regression method (error controlled Runge-Kutta) was used to solve the differential equations. With a least square routine, the difference between the model and the measurement was minimized. The calculations were performed with the software program Scientist.
V Cp
dT = V R NCO HR h A (T Troom ) dt with

*
or
( 4.1 )
dT Cp = R NCO HR h* (T Troom ) dt
h A h = V
R NCO =
d[NCO] = R NCO, Uncat + R NCO, Cat = k f [NCO]n dt

E A,Uncat R T
( 2.12 )
with k f = A 0,Uncat e
+ A0
E A [Cat]m e RT
The fit procedure was as follows. Data obtained from the uncatalyzed runs on EA,Uncat and A0,Uncat were used as input parameters for the fit of the catalyzed runs. All the catalyst dependent runs were fitted simultaneously, giving the values for EA, m and A0. HR was taken from the experiment that gave the largest temperature rise.
61
Chapter 4
Representation of the ATR results Often the results of ATR measurements are plotted straightforwardly as the temperature versus the time. These plots give a clear view on Tadiabatic and a global indication of the reaction velocity. A different method of plotting the results is to translate the temperature versus time plot into an Arrhenius-plot. Although it is much harder to visualize Tadiabatic in such a graph, these plots give more information on the course of the reaction. The activation energy and the actual reaction velocity constants are better illustrated. Furthermore, the effect of the catalyst on the reaction velocity is clearly perceptible in these graphs. For an Arrhenius plot, the reaction rate constant must be known as a function of temperature. The reaction rate constant for an n-th order reaction can be calculated from an ATR experiment according to Richter and Macosko (16):
dT Tad kf = n T HR [NCO]0 ad + Tt =0 T dt
Cp
( 4.2 )
To account for the non-adiabatic conditions in our ATR reactor, the temperature versus time curve that is obtained in an ATR experiment is modified. This modified curve then serves as the basis for the calculation of the reaction rate constant (equation 4.2). To modify the curve, the amount of heat lost must be calculated for every time interval, starting at t = 0 (equation 4.3).
Tloss =
h* t (T Troom ) Cp
( 4.3 )
This temperature loss can be added to the measured temperature at that time interval. In this way, a modified ATR curve can be constructed. 4.3.3 High temperature measurements
A method to follow the conversion of a polyurethane polymerization at higher temperatures and conversions is for instance described by Ando et al. (17). In contrast to ATR experiments, this method is based on isothermal measurements. Small reaction flasks filled with premixed monomers are kept in a thermostatted oilbath. The polymer in the flasks is allowed to react for a certain time. Subsequently, the reaction is quenched and the samples are analyzed using size exclusion chromatography. The kinetic constants are then derived from a plot of the number average molecular weight versus time.
62
Two important conditions must be met in order to get meaningful results from this method. First, it is important that the reaction flasks reach the oilbath temperature much faster than the characteristic reaction time. For our experiments, an analysis based on the Fourier number revealed that this condition is met if the reaction time is larger than 15 minutes. This analysis does not take into account the reaction heat generated in the flasks. However, the reaction heat released only helps to reach the oilbath temperature sooner. Moreover, during the relevant part of the measurement, hardly any heat is generated. A second condition that must be met to obtain relevant results is related to the analytical method. The molecular weight that is measured must represent the real molecular weight of the sample. Since our SEC equipment is calibrated with polystyrene samples, this requirement is not obvious. To check for this requirement, the samples of one experiment have been analyzed on a second SEC system. This second system was equipped with a triple detection system, so that the real molecular weight could be determined. A comparison of the results of the two systems revealed that the polystyrene calibrated system underestimated the weight average molecular weights ten to twenty percent. The difference in number average molecular weight was about ten percent. These errors are acceptable, which means that the results obtained on the polystyrene calibrated column can be used for our investigations on polyurethane kinetics. Experimental procedure The premixing procedure for these experiments was similar to that of the ATR experiments. However, the premixing time was extended to 40 seconds to ensure optimal mixing. After premixing, part of the reaction mass was transferred to small 1.5-ml reaction vials using a syringe. Subsequently submerging of the flasks in liquid nitrogen quenched the reaction temporarily. The total premix, fill and quench cycle took about two minutes. In the next step the flasks were capped while they were still frozen, the capping was carried out in a nitrogen atmosphere to prevent intrusion of moisture. The flasks were then submerged in a heated oilbath to restart the reaction. After the desired reaction time, a flask was transferred quickly into a beaker filled with liquid nitrogen. The flasks were broken and the content was dissolved in a 5% solution of di-butyl-amine in tetrahydrofuran (THF). Subsequently, the THF was evaporated. The samples obtained in this way were analyzed through size exclusion chromatography (18). The SEC-procedure used is described in the previous chapter (section 3.2.3). The experiments for system 1 were performed at five different temperatures (150, 160, 170, 180, and 200 C). The effect of catalyst concentration was investigated at 150 C. Furthermore, three different catalyst
63
Chapter 4
levels were investigated (0.005, 0.015, 0.05 mg / g). The experiments for system 2 were performed at seven different temperatures (150, 160, 170, 180, 190, 200 and 210 C). Every experiment was done at least once. The effect of catalyst concentration was investigated at 180 C. Three different catalyst levels were investigated (0.1, 0.17, 0.3 mg / g). Analysis of the experiments The result of a high temperature experiment consists of a plot of the number average molecular weight versus time, an example is shown in figure 4.4. The number average molecular weight is taken as a measure of the conversion in these plots, because this average represents the amount of molecules present. For a second-order step-polymerization reaction, the number average molecular weight increases linearly in time (19):
MN = Mrep (1+ [NCO]0 k f ( T,[cat]) t )
( 2.15 )
Strictly speaking, equation 2.15 is only valid for step-growth homopolymerizations with an A-B type of monomer. For the terpolymerization that we investigated, large deviations of this equation may occur, especially if the reactivities of the chain extender and the polyol are different (20). However, for the conversion range we investigated (> 95 %) the differences are negligible and, therefore, equation 2.15 is still suitable. To derive the reaction rate constant k from an experiment, the initial slope of the curve has to be determined. A least square routine is used to establish this slope for each experiment. As follows from equation 2.15, the initial slope relates to the reaction rate constant according to:
k f ( T,[cat ]) =
dMN 1 1 = slope dt Mrep [NCO]0 Mrep [NCO]0
( 4.4 )
In this way, the reaction rate constants can be obtained at different temperatures. The initial slope is used to derive the reaction rate constant, since the number average molecular weight will not increase indefinitely over time. For each temperature, the equilibrium molecular weight can also be established with high temperature experiments. This value can be used to calculate the equilibrium constant and the reverse reaction rate at that temperature (equation 2.17).
64
K=
[U]eq M (M M0 ) kf = = N2 N = A 0,eq e 2 kr [NCO]eq M0 [NCO]0
E A ,eq R T
( 2.17 )
The experimental graphs (e.g. figure 4.4) show that there is some scatter in the value for the equilibrium molecular weight. Therefore, an average equilibrium molecular weight is taken for every temperature. In figure 4.4, the shaded areas indicate which part of the curve is considered to be in equilibrium. 4.3.4 Kneader experiments
The third method to measure the kinetics of polyurethane polymerizations with a measurement kneader has been described in chapter 3. Experiments were performed at four different temperatures (125, 150, 175, 200 C). The effect of the catalyst concentration was investigated at 175 C. Four different catalyst levels were used (0.25, 0.40, 0.75 and 1.30 mg / g). For system 1 and 2 the same experiments have been performed. All experiments were repeated three times. The results of these experiments will be discussed in the result section.
4.4
Results
The result section is split into different parts. The results of each measurement method are discussed separately, and for each method, the two different urethane systems are compared. Subsequently, the measurement methods are compared for every system, in order to see if they really result in different kinetic data. 4.4.1 Adiabatic temperature rise measurements
Typical graph As discussed in the experimental section, the ATR results are shown in an Arrhenius plot. Figure 4.1 shows the results of a duplicate experiment for system 1 and 2. The same catalyst level is used for both systems. A second order reaction rate equation is adopted to construct figure 4.1. This assumption seems to be valid for both systems. If we compare both graphs, it is clear that system 1 reacts about one and a half times faster than system 2. As expected, the reaction rate does not rise to infinity; the reaction slows down considerably at a certain conversion. In figure 4.1, this is visible at the point where the tangent line deviates from the measurement points. Surprisingly, the conversion at that point is still quite low, for both systems between 65 and 70 %. A comparison of all experiments showed that regardless of the catalyst level, the decrease in reaction velocity starts between 65
65
Chapter 4
and 70 % for both systems. The reaction does proceed after that point, but the reaction rate constant continues to decrease at higher conversions. The reason for the decrease of the reaction rate constant is not immediately clear. The decrease is too large to attribute it to a change in the reaction order.
1/T (1/K) 0.0024 -2 -2.5 -3 ln (k) (kg/mol s) -3.5 -4 -4.5 -5 -5.5 -6 -6.5 0.0025 0.0026 0.0027 0.0028 0.0029 0.003
Figure 4.1
ATR experiments. [Cat] = 0.075 mg/g, / system 1, / system 2.
As explained in the theoretical section, the reaction may slow down due to diffusion effects. However, the average degree of substitution at 70 % conversion is about equal to three, for linear homopolymers this would be too low to give rise to a large diffusion resistance. Nevertheless, for polyurethanes, phase separation of hard and soft segments may be the cause of the drop in reaction velocity. Due to the clustering of the hard segments, the mobility of the molecules decreases considerably, this can decrease the observed reaction velocity. Blake et al. (21) showed that for fast ATR experiments, the onset of the phase separation is dependent on the initial temperature, catalyst level and the hard segment percentage. They found that phase separation occurred between 66 % and 90 % conversion, which is in agreement with our observations. However, contrary to Blake et al. (21), we do not see an effect of the catalyst concentration on the position of the onset point. This can be explained by the fact that our experiments are much slower. In that case, the phase separation kinetics will be much faster than the reaction kinetics, regardless of the catalyst level. In other words, in our case the phase separation rate does not limit the rate of reaction. Surprisingly, also the chemical composition seems to have no influence on the onset point, since both
66
systems show the same effect at the same conversion. Possibly, the structure of the hard segments does not differ largely for our systems, in spite of the difference in chain extender and isocyanate. In many ATR investigations, the effect of phase separation on the reaction velocity has not been observed. However, these investigations often use a higher hard segment percentage, which increases the temperature at which the phase separation takes place (21). Since at higher conversions the reaction becomes difficult to follow (due to the decrease in heat generation at high conversions) the effect of phase separation may be less visible, which would explain the lack of data. In ATR studies using cross-linking polyurethane systems, a decrease in reaction velocity has been observed at higher conversions (21, 22). Contrary to phase separating systems, the mobility of the molecules for these systems is limited due to crosslinking at higher conversions, instead of clustering of the hard segments. Cross-linking already takes place at a conversion of 70%, this makes the effect much easier to detect.
0.0024 -3.5
0.0025
0.0026
1/T 0.0027
0.0028
0.0029
0.003
0.200 0.150 0.100 0.076 0.050 0.038 0.026 uncat
-4.5 ln (k) (kg / mol s)
-5.5
-6.5
-7.5
Figure 4.2
ATR experiments. Catalyst dependence of system 2.
Comparison of different catalyst levels In figure 4.2 and 4.3, the experiments at different catalyst levels are shown. The zero catalyst experiments are much slower than the runs with the lowest catalyst level (3 - 6 times for system 1, 2 - 4 times for system 2).
67
Chapter 4
1/T (1/K) 0.0024 -2 0.0025 0.0026 0.0027 0.0028 0.0029 0.003
-3
ln (k) (mol/kg s
-4
0.150 0.100 0.075 0.050 0.025 uncat
-5
-6
-7
Figure 4.3
ATR experiments. Catalyst dependence of system 1.
However, the reaction path of the uncatalyzed runs can hardly be compared with those of the catalyzed runs. As mentioned in the theory, the isocyanate droplets may disperse much finer in the presence of catalyst. In case the catalyst is absent, the then occurring larger droplets will result in a more pronounced diffusion limitation for the initial part of the reaction. This explains the low initial activation energy of the uncatalyzed runs (20 kJ/mol for system 1, 35 kJ/mol for system 2). Nevertheless, at a certain conversion, the oligomers formed are likely to compatibilize the reaction mass, resulting in a less diffusion-limited reaction and in a higher activation energy ( 100 kJ/mol mol for system 1, 75 kJ/mol for system 2). This would explain the sudden increase in activation energy in figures 4.2 and 4.3. An autocatalytic process might also be responsible for the sudden increase of the reaction velocity. However, repeated experiments showed that the uncatalyzed runs were very sensitive to mixing, which supports the mixing hypothesis. A model fit of the uncatalyzed runs will be imprecise due to this mixing sensitivity, especially since the activation energy increases suddenly during the reaction. Still a fit of the uncatalyzed runs was used in this kinetic study, since the uncatalyzed reaction contributes to some extent to the overall reaction velocity. Now, if we look at the catalyzed runs, the experiments for system 2 show a remarkable behavior. Normally, one would expect the reaction velocity to increase with increasing catalyst level, while the activation energy remains the same.
68
However, if we look at figure 4.2, it seems that the activation energy decreases with increasing catalyst level, whereas the initial reaction velocity increases with catalyst level, as expected. For system 1, this behavior does not occur (figure 4.3). Therefore, the cause for this phenomenon must be found in the structure of the monomers of system 2. The 2,4-MDI in system 2 results in staggered oligomer and polymer molecules (as explained in the theoretical section). Staggered or rigid molecules hinder the formation of a broad interfacial zone of isocyanate and polyol. Only if this layer is present the mixing will be so fast that the reaction is kinetically limited. According to Machuga et al. (16), the initial growth of this zone will be the same for all catalyst levels. In that case, the reaction velocity depends on the catalyst concentration in the interfacial zone, resulting in an initial reaction velocity that is catalyst dependent. However, at higher catalyst levels the growth rate of the intermaterial zone decreases or even stops, due to the faster formation of large, viscous molecules. Therefore, the combination of staggered molecules and high catalyst level may result in incomplete micromixing of the reactants. The resulting diffusion limitation is observable in an Arrhenius plot as a decrease in activation energy with increasing catalyst level. In figure 4.2, the activation energy continues to decrease with higher catalyst concentrations until the maximum in reaction velocity is reached at high catalyst levels (0.15 and 0.20 mg/g). As a result, two different sets of kinetic parameters needed to be determined for system 2. The runs with the lowest four catalyst levels (0.025 0.075 mg / g) were used to establish the kinetic constants for the experiments at a low catalyst level (the fitting procedure can be found in the experimental section). However, due to the inconsistency in activation energy of system 2, a second set of parameters was necessary to model the reaction at high catalyst levels. Therefore, the highest two catalyst level runs were used to establish a catalyst-independent rate equation, since these experiments were found to be equally fast, regardless of the catalyst concentration. The results are shown in table 4.1.
69
Chapter 4
System 1 A0, Uncat (kg/mol s) EA, Uncat (kJ/mol) A0 (kg/mols) (g/mg)m m(-) EA (kJ/mol)
ATR 169.1 35.3 1.25e6 0.61 50.5 ATR ATR high [cat]
High Temperature Experiments
Kneader Experiments
2.69106 0.5 53.6 High Temperature Experiments
5.13105 0.57 52.0 Kneader Experiments
System 2
low [cat]
A0, Uncat (kg/mols) EA, Uncat (kJ/mol) A0 (kg/mols) (g/mg)m m(-) EA (kJ/mol)
5.37e3 45.8 1.09e5 0.92 42.5 0.208 0 9.9 1.49107 0 71.9 2.18106 0 61.3
Table 4.1
The kinetic parameters for system 1 and 2.
In contrast to system 2, the Arrhenius plot for system 1 (figure 4.3) shows a regular behavior. At a higher catalyst level, the activation energy remains constant whilst the reaction rate constant increases, indicating that no diffusion limitations occur unlike system 2. Using figure 4.3, a fit has been made according to the fitting procedure as described in the experimental section. The resulting model parameters are also shown in table 4.1.
70
4.4.2
High conversion experiments
A typical graph for a high conversion experiment is shown in figure 4.4. Two experiments and their duplicates are shown. System 2 was used for these experiments. The solid lines represent the model predictions; the model predictions are based on a fit of all high temperature experiments performed in this research. The resulting parameters describing the kinetics are shown in table 4.1. As expected, the molecular weight increases in time. Initially, the increase is linear. This part of the curve is used to determine the initial slope. At longer reaction times, the molecular weight levels off due to depolymerization. At both temperatures, the reproducibility of the experiments is reasonable.
90 80 70 60
Mn (kg/mol)
50 40 30 20 10 0 0 25 50 75
Time (min)
Figure 4.4
Mn versus time for the high temperature experiments for system 2, experimental results and model prediction. the equilibrium molecular weight. 150 C, 150 C duplicate, 200 C, 200 C duplicate. The points in the grey areas are used to determine
The procedure to derive the kinetic data is described in the experimental section. This procedure is used to obtain the forward and reverse reaction rate at every temperature under investigation.
71
Chapter 4
1/T (1/K) 0.002 0 0.0021 0.0022 0.0023 0.0024
-1 ln (k) (kg/mol s)
-2
-3
-4
-5
Figure 4.5
The forward reaction rate constant as a function of temperature for high temperature experiments. system 1, System 2.
In figure 4.5 an Arrhenius plot of the forward reaction rate is shown for all experiments performed with system 1 and 2. At lower temperatures both systems exhibit a linear relationship between ln(k) and 1/T, which confirms the second order rate assumption. However, for system 2 a deviation from linearity turns up at higher temperatures (200 C, 210 C). This is due to the fact that the slopes of these curves are determined largely during the first 15 minutes of the reaction, when the flasks are still warming up (as explained in the theoretical section). Therefore, the effective flasks temperature will be lower than the oilbath temperature, which explains the downward curvature in figure 4.5. For this reason, the experiments at 200 and 210 C are not used to determine the Arrhenius-parameters for system 2. However, the runs at 200 and 210 C can still be used to determine the kinetics of the depolymerization reaction.
72
11 10.5 10 ln(K) (kg/mol) 9.5 9 8.5 8 7.5 7 0.002
0.0021
0.0022 1/T (1/K)
0.0023
0.0024
Figure 4.6
The equilibrium constant as a function of temperature for high temperature experiments. System 1, system 2.
The equilibrium molecular weights are determined at reaction times larger than 15 minutes, which makes sure that the flasks have reached the oilbath temperature. In figure 4.6 the Arrhenius plot for the depolymerization reaction is shown for system 1 and 2. System 1 Aeq (kg/mol) EA,eq (kJ/mol)
Table 4.2
System 2 0.0393 43.4
0.0110 52.7
The equilibrium parameters for system 1 and 2.
The equilibrium constant for each temperature is calculated by substituting the equilibrium molecular weight in equation 2.17. The plot shows that even at 150 C the effect of depolymerization is noticeable. The resulting parameters for the depolymerization reaction are presented in table 4.2. Besides the effect of depolymerization, the effect of the catalyst concentration was also investigated through high conversion experiments. For system 2, the catalyst level did not have an effect on the reaction rate constant, at least not for the relatively high catalyst levels that were chosen. For system 1, the reaction rate was about proportional to the square root of the catalyst concentration (table 4.1).
73
Chapter 4
4.4.3
Kneader Experiments
The results of the kneader experiments for system 2 were discussed in a previous chapter. However, in this chapter only one catalyst level was used. Therefore, additional experiments were performed to establish the catalyst dependence for both systems. The resulting plots of reaction rate versus catalyst concentration are shown in figure 4.7.
ln[Cat] (mg/g) -2 -1.5 -1 -0.5 0 0.5 0
-0.5 ln(k) (kg/mol s)
-1
-1.5
-2
-2.5
Figure 4.7
The dependence of the Arrhenius pre-exponential constant on the catalyst level. System 1, System 2.
For system 2, the experiments at the highest catalyst level are slightly faster than the other three experiments, indicating that there is a slight influence of catalyst concentration on the reaction velocity. This influence is very small, and since the other three catalyst levels do not show any effect of the catalyst, the reaction velocity is considered independent of the catalyst concentration. For system 1 the catalyst dependence is obvious, the dependency factor m in equation 2.12 is equal to 0.57. More discussion on the effect of catalyst will follow in the next sections. Furthermore, figure 4.7 shows that system 1 reacts faster than system 2 at all catalyst levels. As explained in the theoretical section, the 2,4- MDI that is used in system 2 may cause it to react slower. The effect of the temperature on the reaction velocity for system 1 is shown in figure 4.8. Analogous to system 2, the model predictions and the experimental curves are shown, and the kinetic constants are obtained in a similar way as for system 2 (18).
74
150000
120000
200C
175C
150C
Mw [g/mol]
90000
60000
125C
30000
0 0 4 8 Time [min] 12 16
Figure 4.8
The Mw versus time for the kneader experiments for system 1. 80RPM. Model predictions and experimental results.
4.4.4
Comparison of the different measurement methods, System 1
Table 4.1 shows a comparison of all the kinetic parameters obtained for system 1. Both the catalyst dependence and the activation energy seem to agree fairly well for all experimental methods. This observation indicates that for all measurement methods used, the reaction develops identically for system 1. Neither the activation energy nor the catalyst dependence changes appreciably with the temperaturerange or the conversion-range of the measurement method. The simplified second order assumption seems to hold for all measurement conditions. If we look at the catalyzed urethane reaction, the reaction develops through several equilibrium steps, all related to the catalytic center. Naturally, these equilibriums will shift with temperature, which may change the reaction order and catalyst dependence. Surprisingly, both the reaction order and the catalyst dependence remain constant for the temperature range under investigation. The order of catalyst dependence ( 0.5) falls within the limits reported by other authors (0.5 1). A possible explanation for the value of 0.5 for the order of catalyst dependence is given by Richter et al. (16). They related this value to a simple reaction mechanism. In the first step of this mechanism, the catalyst dissociates and, in a second step, the
75
Chapter 4
cationic catalytic center forms a complex with an isocyanate group. If both steps are thermodynamically unfavorable, the reaction order equals 0.5.
Temperature (C) 50 0 100 150 200
-2 ln(k) (kg/mol s)
-4
-6
-8
Figure 4.9
The model reaction rate constant versus the temperature for the different measurement methods. System 1. The open symbols are extrapolations to areas where no measurements have been carried out. temperature.
ATR,
kneader,
high
The effect of the different measurement methods on the reaction rate constant is shown in figure 4.9. Figure 4.9 shows that the ATR experiments are as fast as the high temperature experiments. Within the experimental error, for these two types of measurements, the method does not seem to have any influence on the observed reaction rate. A priori, one would expect the kneader experiments to be equally fast, or even faster than the high temperature experiments. However, the kneader experiments show reaction rate constants that are 3-4 times slower. No obvious explanation is available to account for this result. The mysterious drop in reaction velocity may be attributed to the materials used. The only difference between the kneader experiments and the other two experiments is the batch of polyol and chain extender used. The acidity and OH-value, which have an effect on the reaction rate, may change per batch of polyol. No corrections were made for these changes. Due to the use of a different polyol batch, no valid comparison can be made between the kneader experiments and the other two types of experiments. However, the other two experiments do give important information on the polyurethane 76
reaction. The observation that the reaction velocity does not depend on the measurement method deployed implies that the theory of functional group reactivity independent of chain length is also valid for polyurethanes. Moreover, it strengthens the belief that for the ATR and high temperature experiments (initial) diffusion limitations do not occur, since the reaction velocity, the reaction order and the catalyst dependence is the same for both experiments. For extruder applications, the result means that the most convenient measurement method can be chosen to obtain the correct data for the kinetics, at least for this polyurethane system. The advantage of ATR experiments is the ease of measurement; on the other hand, the high temperature experiments give essential information on the depolymerization reaction. 4.4.5 Comparison of the different measurement methods, System 2
Similar to system 1, two batches of materials were used for the experiments of system 2. The ATR experiments for system 2 were performed with the same batch of polyol as the ATR experiments and high temperature experiments of system 1. The kneader and the high temperature experiments were performed with the other batch of polyol, which is also used for the kneader experiments of system 1. In table 4.1 the same batches have the same color. Again, possible inconsistency in batches complicates the comparison of the different experiments. Moreover, the specific diffusion limitations that were observed for the ATR experiments make comparison of this method with the other two methods even harder. These diffusion limitations are specific for the ATR method (and for the polyurethane system used), and result in a catalyst-independent reaction velocity at higher catalyst levels, and a lowering of the activation energy. For the high conversion experiments, the lack of catalyst dependence is also observed at higher catalyst levels, but the activation energy is much higher for these experiments. Therefore, the cause of the catalyst independency for these experiments must be different. An obvious explanation is not available, possibly the functional groups of the polymer molecules experience a diffusion limitation that is noticeable at higher catalyst levels. In that case, the catalyst level does not make any difference above a certain threshold concentration. This explanation is not completely satisfactory. First, this type of diffusion limitation does not occur for system 1. However, the difference in monomers for the two systems might give a difference in the polymer structure and therefore in diffusion behavior. The staggered 2,4-MDI groups in system 2 may result in a more coiled polymer molecule which subsequently can result in a more entangled polymer melt in which diffusion limitations occurs more readily. Still, a
77
Chapter 4
second question remains. The activation energies that are found for the high conversion experiments are rather high for a diffusion-limited reaction (60 70 kJ/mol). The flow activation energy for system 2, which can be considered as the activation energy of diffusion, is much lower: 43 kJ/mol (18). The reason for this difference is not clear.
Temperature (C) 50 0 100 150 200
-2
ln(k) (kg/mol s)
-4
-6
-8
-10
Figure 4.10
The
reaction
rate
constant
versus
the
temperature
for
the
different
measurement methods. System 2. The open symbols are extrapolations to areas where no measurements have been carried out. Kneader,
high temperature.
Now, if we look at the difference between the two high conversion methods, mixing seems to have an influence. Both experiments are performed with the same batch of polyol; therefore, a comparison can be made. The kneader experiments show a much higher reaction rate constant than the high temperature experiments (figure 4.10). This observation may support the assumption that the reaction for system 2 is subject to diffusion limitations. Mixing in that case alleviates the diffusion limitation, resulting in a faster reaction.
78
4.5
Conclusions
The reaction kinetics of polyurethane polymerization was studied in this chapter. In particular, the need for different measurement methods for reactive extrusion purposes was investigated. The study shows that ATR and high temperature measurements give the same kinetic constants for a commercial polyurethane system (system 1). Since both methods differ greatly in reaction time, reaction temperature, and analytical method, it can be concluded that for this system both measurement methods can be applied. Unfortunately, an extra validation of this conclusion with a third method (with the measurement kneader) could not be used, since a different batch of polyol had to be used for these last measurements. However, the activation energy, catalyst dependence and reaction order was similar for the kneader experiments, which strengthens the vision that any of the three measurement methods will yield the same kinetic equation for system 1. Therefore, it is probable that the reaction is kinetically controlled and that the reaction proceeds uniformly over a wide range of temperatures and conversions. For a less common polyurethane system (system 2), a completely different result is obtained. The three different measurement methods each result in a different kinetic equation, indicating that for this system a uniform reaction mechanism cannot be adapted. For extrusion purposes, this means that a single kinetic measurement method does not suffice. At least two measurement methods seem to be required; a low temperature, low conversion method as adiabatic temperature rise experiments and a high temperature high conversion method. The cause of these inconsistencies may result from the structure of the monomers used in system 2. As explained in the result section, the presence of 2,4-MDI in system 2 may hinder the expansion of surface instabilities (which are indispensable for good micromixing) and therefore prevent a kinetically controlled reaction (14). However, since this hypothesis is only derived from the kinetic parameters for this system, a further validation would be necessary; for example by following the reaction under a microscope.
79
Chapter 4
4.6
A0 A [Cat] Cp EA HR h h* kf kr m M0 MN MW n [NCO] [NCO]0 R RNCO T Tad t [U] V
List of Symbols
Reaction pre-exponential constant Surface area of ATR reactor Catalyst concentration Heat capacity Reaction activation energy Heat of reaction Heat transfer coefficient Overall heat transfer coefficient Forward reaction rate constant Reverse reaction rate constant Catalyst order Average weight of repeating unit Number average molecular weight Weight average molecular weight Reaction order Concentration isocyanate groups Initial concentration isocyanate groups Density Gas constant Rate of isocyanate conversion Temperature Adiabatic temperature rise Time Concentration urethane bonds Volume ATR reactor mol/kg s m2 mg/g J/kgK J/mol J/mol J/m2sK J/kgsK kg/mols 1/s g/mol g/mol g/mol mol/kg mol/kg kg/m3 J/mol K mol/kgs K K s mol/kg m3
Subscripts Cat Uncat Eq Catalyzed Uncatalyzed Equilibrium
80
4.7
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.
List of References
C. W. Macosko, RIM - Fundamentals of Reaction Injection Molding, Hanser, Munich, (1989). P. Cassagnau, F. Mlis, and A. Michel, J. Appl. Polym. Sci., 65, 2395 (1997). X.D. Sun, and C.S.P. Sung, Macromolecules, 29, 3198, (1996). X. Sun, J. Toth, and L.J. Lee, Polym. Eng. Sci., 37, 143 (1997). P. Cassagnau, T. Nietsch and A. Michel, Intern. Polym. Process., 14, 144 (1999). M.E. Hyun, and S.C. Kim, Polym. Eng. Sci., 28, 743 (1988). A. Bouilloux, C.W. Macosko, and T. Kotnour, Ind. Eng. Chem. Res., 30, 2431 (1991). S. Hoppe, S. Grigis, and F. Pla, Chisa 2002 A53 G. Lu, D.M. Kalyon, I. Yilgr, and E. Yilgr, Polym. Eng. Sci., 388 (2003) S.D. Fields, E.L. Thomas, and J.M. Ottino, Polymer, 27, 1423 (1986). S.D. Fields, and J.M. Ottino, AIChE J., 33, 959 (1987). P.D. Wickert, W.E. Ranz, and C.W. Macosko, Polymer, 28, 1105 (1987). O. Kolodziej, C.W. Macosko, and W.E. Ranz, Polym. Eng. Sci., 22, 388 (1982). S.C. Machuga, H.L. Midje, J.S. Peanasky, C.W. Macosko, and W.E. Ranz, AIChE J., 34, 1057 (1988). P.J. Flory, Principles of Polymer Chemistry, Cornell University Press, Ithaca, N.Y., (1953). E.B. Richter, and C.W. Macosko, Polym. Eng. Sci., 18, 1012 (1978). T. Ando, Polym. J., 11, 1207 (1993). V.W.A. Verhoeven, M. van Vondel, K.J. Ganzeveld, and L.P.B.M. Janssen, Polym. Eng. Sci., 44, 1648 (2004) G. Odian, Principles of Polymerization, John Wiley & Sons Inc., New York (1991). F. Lopez-Serrano, J.M. Castro, C.W. Macosko, and M. Tirrell, Polymer, 21, 263 (1980). J.W. Blake, W.P Yang, R.D. Anderson, and C.W. Macosko, Polym. Eng. Sci., 27, 1237 (1987). Y.T. Chen, and C.W. Macosko, J. Appl. Polym. Sci., 62, 567 (1996).
81
5
5.1
The Reactive extrusion of thermoplastic polyurethane

Introduction
The kinetics of the polyurethane reaction has been discussed in the previous two chapters. Relevant information on the kinetics was obtained under mixing conditions that also occur during the extrusion process. In the current chapter, we leave the kinetics behind and shift our emphasis to the extruder. As stated in chapter 2, extrusion is a relatively expensive process. Comprehension of the reaction in an extruder, coupled with a rational design of the extruder can therefore lead to a cost benefit. Improvement of the extruder efficiency and control of the product quality may be defined as a goal in that perspective. The efficiency of the extruder operation can simply be expressed as the conversion at the end of the extruder. The product quality is more difficult to grasp, it is related to the conversion, the occurrence of side reactions (allophanate formation, oxidation, crosslinking) and the size and morphology of the hard segments. For the current study the emphasis lies on understanding how the conversion in an extruder can be optimized. The use of an extrusion model is essential, because of the complicated processes that take place in an extruder, and the wide diversity in extruder configurations that are possible. A model can be useful for the optimization of an existing process, the implementation of new types of materials, or the conversion of a batch process into a continuous process. In the literature, several studies have been directed towards understanding the TPU production in an extruder (1 - 5), each with their own emphasis. Single screw extruders (4, 5), counterrotating extruders (1, 2), and corotating extruders (3) have been studied. A broad range of subjects was covered; predictive modeling (2, 4), reactive blending (3), mixing efficiency in the extruder (1), and the effect of extrusion on product quality (5) are described. However, a missing subject in this survey is the depolymerization reaction. The depolymerization reaction is inevitably noticeable at temperatures above 150C (6), which is a typical extrusion condition. The presence of this reaction may hinder the extrusion efficiency. In the current study, the effect of the reverse reaction will be investigated. In addition, the ability of the model to capture the reverse reaction will be examined.
Chapter 5
5.2
5.2.1
The Model
Introduction
The purpose of the extruder model is twofold: it should increase the understanding of the complex mechanisms of the reactive extrusion of polyurethane and it should be suitable to optimize the process. In this way, the number of (expensive) experimental trials can be minimized when a new process is designed. To meet this objective, it should be easy to test different extruder configurations and different operating conditions with the model, and the calculation time should be short. In that case, a complex computational fluid dynamics model (CFD-model) is not useful. Therefore, an analytical approach was chosen, which is similar to previous modeling studies (7 -13). Of course, the flexibility and broad applicability of such an engineering model comes with a price. Non-incorporated radial temperature gradients, a simplified approach for the non-Newtonian flow behavior and the complicated flow in the kneading sections may result in a less accurate model prediction. 5.2.2 Reaction
As explained in the introductory section, the main output parameter of the model is the degree of polymerization () or the directly related weight average molecular weight (Mw) at the end of the extruder. The degree of polymerization is governed by the reaction kinetics of the polyurethane system under investigation. For this investigation, polyurethane system 2 as described in chapter 4 was chosen. The kinetics of this polyurethane system can be described by a second order rate equation:
R NCO = with
d[NCO] = k f [NCO]2 k r [U] dt

m E A e R T
k f = [Cat] A 0
kr =
kf
E A,eq
( 2.16 )
A 0,eq e and [U] = [NCO]0 [NCO]
R T
To predict the isocyanate conversion in a reactor, the isocyanate balance is solved for that specific reactor, taking into account the above rate equation (equation 2.16). To solve the isocyanate balance, knowledge on the residence time
84
The reactive extrusion of thermoplastic polyurethane
distribution of the specific reactor is indispensable. Now, for two extreme cases of residence time distribution (no (axial) mixing and complete mixing), the isocyanate balance is solved. Both cases will be used further on in the model. First, the no-mixing situation is evaluated. Practically, a no-mixing situation is present in a plug-flow reactor. For a plug-flow reactor, the solution of the isocyanate balance, using equation 2.16, is as follows:
[NCO]N =
with
(k
+ D C e tN
+ D kr
D
2 k f 1 C e tN
2k f [NCO]N1 + k r D 2k f [NCO]N1 + k r + D
( 5.1 )
D = k 2 + 4 k f k r [NCO]0 d
and
C=
( 5.2 )
These equations are also applicable for the conversion in an ideally stirred batch reactor. It is clear from equations 5.1 and 5.2 that the isocyanate concentration is dependent on the residence time and temperature. The concentration at the end of the reactor, [NCO]N, is expressed as a function of the residence time tN and the inlet isocyanate concentration, [NCO]N-1. For the second situation, a completely mixed reactor, the isocyanate balance of a continuous ideally stirred tank reactor (CISTR) can be solved. Since a polymerization reaction is under investigation, the micro-mixing situation in such a reactor is best considered as micro-segregated (14). The conversion in a micro-segregated CISTR is equal to:
[NCO]N =
1 [NCO]t,batch E(t) dt = [NCO]t,batch e

0 0
t t
dt
( 5.3 )
[NCO]t,batch is the isocyanate concentration for a batch reactor with a residence time t, as can be calculated using equation 5.1 and 5.2. Since no analytical solution is available, equation 5.3 is solved numerically in the model. Obviously, the isocyanate concentration for both reactor types is dependent on the temperature and the residence time. In order to predict the right conversion in the extruder, these parameters must be known. The residence time in the extruder can be extracted from the flow model, while the temperature of the reaction mass along the extruder can be analyzed through the energy balance (which in itself is also
85
Chapter 5
related to the flow model). Furthermore, for every part of the extruder the residence time distribution must be known. All of these issues will be addressed in the following paragraphs. In the concluding paragraph, details on the overall extruder model are given. 5.2.3 Residence time / flow model
In order to be able to predict the residence time, a flow model of the corotating twin-screw extruder is necessary (paragraph 2.2.5). To model the flow behavior in the extruder, it should be taken into consideration that different types of screw elements are used. Most commonly used are the transport elements and the kneading paddles. To introduce the flow behavior in the model, for both element types a simplified approach was chosen, based on the flow between two parallel plates. In this approach, the screw channels are represented as stationary, infinite screw channels, whereas the barrel moves over the channel (paragraph 2.2.1). This approach is similar to previous analytical models (7 - 10, 12, 13). Details on the analysis can be found in these publications. This approach is specifically appropriate for the most commonly occurring elements, the transport elements. For kneading elements, a modification is made to this approach. Both types of elements will be treated separately. Transport elements For the transport elements, the filling degree fT of the not fully filled sections is equal to the ratio of the real throughput (Q) and the maximal obtainable throughput:
fT =
Q Q S,max,T Q L,drag
( 5.4 )
Equation 5.4 is a modified form of equation 2.7. The maximum throughput equals the maximum conveying capacity (QS,max,T) minus the leakage flows (QL,drag) over the flight The different flows are derived from the parallel plate flow model. Their exact definitions are described by Michaeli et al. (12). Besides the filling degree of the not fully-filled zone, the residence time in a transport section is also determined by the length of the fully filled part. This filled length is a result of the pressure build-up capacity of the element concerned. In case of a larger pressure build-up capacity (P/L), the (filled) length needed to overcome a pressure barrier is shorter. The pressure build-up capacity in a transport element is calculated according to Michaeli et al. (12) as well: 86
P = L
(Q S,max,T Q Q L,drag ) k p,T channel

3 i h 3 w ( ) + w + e R u k p,T channel sin 12 e flight tan
( 5.5 )
Q L,drag = u
v0 sin cos R 2
The pressure build up in a transport element is proportional to the viscosity, and to the maximum flow rate minus the real throughput and the leakage flow. The proportionality factor is a function of the channel geometry and a shape factor kP, T. The calculation of the maximum conveying capacity (QS,max,T) takes into account the effect of the intermeshing zone. The leakage over the flight is taken into consideration with as well a drag flow dependent (QL,drag) as a pressure flow dependent term (integrated in equation 5.5). The non-Newtonian behavior of a polymer fluid is taken into account indirectly. The average shear rate in the element is calculated according to Michaeli et al. (10). Both the shear rates over the flight and in the channel are calculated. In order to calculate the shear rate in the channel, a two-dimensional flow analysis is made, for which the actual channel geometry is taken into account. Subsequently, the apparent viscosities are calculated using the appropriate rheological model and the calculated shear rates. This apparent viscosity is used in equation 5.5. In this way, most rheological models can be used. Kneading paddles For the kneading paddles, the flow behavior differs considerably from that of a transport element (paragraph 2.2.4). Still, for the current modeling approach, a kneading block is considered as a modified transport element, with an extra leakage flow (QL,k) due to the staggering of the kneading paddles. The maximal conveying capacity is lowered due to this leakage flow. Therefore, the equation to calculate the filling degree of the partially filled zone fk resembles that of the transport elements:
fk =
Q Q S,max,K Q L,drag Q L,k
( 5.6 )
87
Chapter 5
In equation 5.6, QL,k is a result of the extra leakage flow due to the leakage gaps that exists between the staggered kneading paddles. The pressure build up (or consumption) in a kneading section is calculated with (12):
P (Q S,max,K Q ) k P,K channel = L i h 3 w ( )
( 5.7 )
Again, a similarity exists between the equation for the kneading blocks and for the transport elements. The pressure build up in a kneading element is proportional to the viscosity, and to the maximum flow rate minus the real throughput. In case of a kneading element, the extra leakage term due to the staggering of the kneading paddles is incorporated in the equation for QS,max,K. For the kneading blocks, the apparent viscosity is calculated in the same way as for the transport elements. The approach that is taken in our model for the flow behavior in the kneading blocks suffices for low staggering angles, since in that case, the similarity with the transport elements is still present. However, this model is inadequate at higher staggering angles. In that case, an approach adapted by Verges et al. (13) may give better results. Still, experimental validation of the pressure build up in kneading paddles is scarce. This makes a comparison of the different modeling approaches for the kneading paddles difficult and the present approach sufficient. 5.2.4 Residence time distribution
As explained in the paragraph on the polyurethane reaction, knowledge on the residence time distribution in the extruder is indispensable to calculate the appropriate conversion. In the model, a distinction is made between a kneading block and a transport element, since the flow in a kneading block differs substantially from the flow in a transport element. Moreover, an extra distinction is made between flow in partially filled elements and fully filled elements. In a partially filled element, hardly any axial mixing takes place, because the material is more or less glued to the flank of the screw. Therefore, all the partially filled elements are considered to operate under plug-flow regime, and the isocyanate balance for a plug-flow reactor (equations 5.1 and 5.2) can be used to calculate the conversion in a partially filled zone. Fully filled transport elements However, for fully filled elements, the situation differs. In a fully filled transport element, each particle follows a different (helical) path in the screw channel, which causes a distribution in the residence time. Pinto and Tadmor (15) developed a 88
residence time distribution model, based on this helical flow pattern in the channels of a single screw extruder. This model is also applicable for our self-wiping twinscrew extruder, even though the intermeshing zone will disrupt the flow pattern somewhat. The RTD analysis of Pinto and Tadmor (15) shows that the flow in the fully filled transport elements is neither comparable to plug-flow or flow in a pipe. The actual flow lies somewhere in between. However, as a first approach, the residence time distribution in the fully filled transport elements will be regarded as plug-flow in the current model (equations 5.1 and 5.2). Fully filled kneading paddles The flow in the kneading paddles differs completely from the flow in a transport element. In general, in a kneading zone, the circumferential flow rate is much higher than the axial flow rate (16). Besides, due to the squeezing action of two paddles in the intermeshing zone, the mixing is much better. Moreover, a considerable backward flow will be present between two neighboring paddles, because of a leakage gap between these paddles. For these reasons, the residence time distribution in a kneading zone will have a similarity to that of a cascade of continuous ideally stirred reactors. Therefore, this approach is used for the extruder model (equation 5.3). The kneading blocks are divided in a number of CISTRs. Tentative experiments in a Perspex extruder were performed to establish the length of every reactor; it was found to equal half of the screw diameter. This length is typically two to four times the width of a kneading paddle. 5.2.5 Energy
The temperature is the last essential factor that is needed for calculating the conversion in an extruder. The temperature can be derived from the energy balance. In our model, the energy balance for the extruder or for a part of the extruder is:
Cp Q (TN TN1 ) = h A wall (TWall TN ) & & Q HR [NCO]0 ( N N1) + WC + WF
( 5.8 )
The temperature rise in (a part of the) extruder (TN-TN-1) is a result of the heat transfer through the wall, of the exothermic reaction and of the viscous dissipation in the channel (WC), and over the flight (WF). The energy balance considers the extruder or a part of the extruder to be a continuous ideally stirred reactor (CISTR). Obviously, a more complicated flow situation exists in the extruder, which will result in radial and axial temperature
89
Chapter 5
gradients. The latter can be resolved through applying equation 5.8 on short axial sections of the extruder. Nevertheless, the radial temperature gradient can result in a deviation of the measured and predicted temperature in an extruder. The heat transfer coefficient in the energy balance is adapted from Todd et al. (17). As for the average shear rate in the channel, the viscous dissipation in the channel is calculated using a two-dimensional flow analysis (10). Since the viscous dissipation over the flight is substantial (7), it is integrated in the heat balance according to an equation by Michaeli et al. (10):
v2 e & WF = flight 0 i L R sin

5.2.6 Modeling approach
( 5.9 )
A general modeling scheme has been developed to calculate the conversion in the extruder. For this calculation, the extruder is split up in segments of a quarter of the diameter of the extruder. The output of the first segment is the input of the second segment and so on. The sectioning is necessary due to the large temperature and conversion gradient in the axial direction. The size that is chosen for the segment is a compromise between accuracy and calculation time. The sectioning strategy is not compatible with the continuous laminar flow profile that is present in an extruder. For the sectioning approach, a continuous flow is divided into segments that have closed-closed boundary conditions. For example, in case a fully filled transport zone is divided into segments, the residence time distribution over the whole section can be calculated with the approach of Pinto and Tadmor (15). However, to do so for every segment and applying closed boundary conditions will give an erroneous result. To prevent this error, a plug-flow approach is chosen for the transport zones. A plug-flow reactor can be divided in segments without any problems. Segmental iteration For a segment N, the temperature TN, conversion N, viscous dissipation WN, average shear rate N and the viscosity N are calculated. The equations used for every parameter are described in the previous paragraph. However, it is not possible to solve these equations sequentially; looking at equation 5.10 it is clear that the equations are interrelated.
90
N = f (TN , t res,N ,...) & TN = f (WN , N ...) & WN = f (N ,...) N = f (TN , N ,...)
Therefore, a dichotomy routine is used to solve equation 5.10 for every segment. The convergence criterion for this routine is the viscosity, since the viscosity is the most sensitive parameter in equation 5.10. In general, the dichotomy routine converged within five steps. Extruder iteration Having calculated all variables in segment N, the output of this segment is the input of the next one, segment N+1. However, in an extruder, the situation of this next segment can influence the filling degree of the previous segment. At the start of the extruder iteration, all segments are considered partially filled. Both the die and reverse or neutral screw elements raise a pressure barrier. This pressure barrier needs to be overcome by the previous segment, which fills itself for that reason. A similar mechanism is present in the model (figure 5.1). N-2 N-1 N N-2 N-1 N ( 5.10 )
Pressure
Pressure 91
Figure 5.1
The calculation of the filled length in front of a reverse element.
In case a negatively conveying segment (N) is encountered, the upstream segment (segment N-1) is filled and recalculated using the segmental iteration. Subsequently, the usual calculation order is followed, so the next (in this case the negatively conveying) segment (N) is calculated. In case the pressure at the end of this segment is still below zero, another upstream segment is filled (N-2) and so on, until the pressure at the end of segment N is zero. For the die, a similar routine is followed. The calculation ends if the pressure at the outlet of the die is atmospheric.
Chapter 5
5.3
Experimental section
Several types of experiments were performed to validate the extruder model: coldflow extrusion experiments, non-reactive extrusion experiments and reactive extrusion experiments. In the experimental section, every type of experiment is addressed separately. Two types of extruders have been used for the validation study: a Perspex extruder (D = 50mm, Cl = 39mm, R = 1mm, L/D = 25) and an APVBaker MPF50 twin-screw extruder (D = 50mm, Cl = 39mm, R = 0.8mm, L/D = 24). Both extruders can be equipped with different types of transport elements or with kneading paddles (width = 0.25D) with staggering angles of 30, 45, 60, 90, 120, 135 and 150. For the APV-baker extruder, the temperature of the barrel wall can be regulated through ten independent heating/cooling zones (electric heating, water cooling). 5.3.1 Cold-flow extruder experiments
For the cold-flow experiments, the Perspex extruder was equipped with two-lobed 50/50 (diameter/pitch) transport elements. A calibrated pressure gauge was placed in front of the die and at 22 D. Glucose syrup and a 1.5 % solution of hydroxy-ethyl cellulose (HEC) in water were used for the experiments. Both liquids were rheologically characterized with a constant strain rheometer (TA Instruments, AR 1000-N Rheometer) using a cone and plate geometry. Glucose syrup showed a Newtonian behavior ( = 10 Pas) while the shear dependency of the HEC viscosity obeyed a power-law equation (0 = 76.1 Pas, n = 0.25). A gear pump (Maag) was used as a feed pump for the extruder. The throughput was set to 7.5 kg/hour and the rotation speed of the extruder was varied between 12.5 and 100 RPM. 5.3.2 Non reactive validation
Polypropylene (Stamylan PP, DSM) was used for the non-reactive validation. The rheological behavior of the polypropylene was established on the same rheometer as for the cold flow experiments, the rheometer was operated in the oscillatory mode. The temperature dependency of the viscosity could be described with a Williams-Landel-Ferry (WLF) equation:
( T) C1( T Tr ) log (T ) = C + ( T T ) r 2 r
( 5.11 )
92
With C1 = 2.66, C2 = 305.6, and Tr = 493 K. The shear rate dependency was accounted for using the Williamson model:
app =
322.1 1+ (0.001214 ) 0.625
at 493 K
( 5.12 )
The extruder was equipped with one -45/8/100 kneading zone
(stagger
angle/number of kneading paddles/ length kneading zone) to ensure complete melting of the polypropylene. The kneading zone was placed 20 cm downstream of the inlet zone. The pressure is measured at three locations for establishing the pressure gradient in the fully filled zone. The temperature is measured in two places along the fully filled zone with non-protruding thermocouples. No significant difference in temperature was observed along the fully filled zone, indicating an isothermal fully filled zone. Five different rotation speeds and five different wall temperatures were investigated. For every experiment, the die diameter was adapted to obtain a sufficiently long fully filled zone. The polypropylene is added to the extruder with a hopper (K-tron T-20). A constant feed rate of 15 kg/hour was maintained. 5.3.4 Reactive validation
Equipment The extruder layout for the reactive experiments is shown in figure 5.2. One kneading zone (45/8/100) is placed three diameters from the inlet. Only half of the extruder is used for these experiments to prevent an excessive long residence time. Two feed streams are added to the extruder. These streams come together above the feed pocket of the extruder. To premix both streams, a static mixer of the Kenics type of variable length can be placed in the joint feed line. The first stream consists of the premixed chain extender and polyol; the second stream is formed by the isocyanate. A solution of the catalyst in dioctyl-phtalate is added continuously to the polyol feed line using an HPLC-pump. The static mixer of 32 elements is placed after the catalyst injection point to mix the catalyst evenly in the polyol. The isocyanate supply vessel is kept at 25 C while the polyol supply vessel and feed lines are kept at 80 C. The flows of both streams are controlled in the same way. A gear pump (Maag TX 22/6) is combined with a flow sensor (VSE, VS0.04-E) which sends its signal to a PI controller/flow computer (Contrec 802-A). Through a frequency deformer (Danfoss VLT 2010), the PI controller controls the
93
Chapter 5
gear pump. A throughput of 12.5 kg/hour is maintained for most of the experiments.
Polyol + diol + catalyst
MDI P1 P2 P3
Figure 5.2
The extruder layout for the reactive validation experiments.
Rheo-kinetics System 2 as described in chapter 4 was used for the reactive extrusion experiments. This system consists of: A polyester polyol of mono-ethylene glycol, di-ethylene glycol and adipic acid (MW = 2200 g/mol, f = 2). Methyl-propane-diol (Mw = 90.1 g/mol, f = 2). A eutectic mixture (50/50) of 2,4 diphenylmethane diisocyanate (2,4-MDI) and 4,4 diphenylmethane diisocyanate (4,4-MDI). (Mw = 250 g/mol, f = 2).
A0, Uncat EA, Uncat A0 m EA

Table 5.1
(kg/mol s) (kJ/mol) (kg/mols) (g/mg)m (-) (kJ/mol)
7.4104 52.4 4.53107 2.25 45.18
The kinetic parameters used for the reactive extrusion model.
Adiabatic temperature rise experiments were used to obtain the kinetic parameters, according to paragraph 6.6.4. For these experiments, the monomers were premixed
94
with the static mixer as shown in figure 5.2. The resulting kinetic parameters are shown in table 5.1. As discussed in chapter 4, the reaction rate slows down considerably at high conversions for this polyurethane. Mixing seems to have an effect a high conversions (paragraph 4.4.5), moreover, hardly any catalyst dependence is present (paragraph 4.4.3). Both factors are caused by diffusion limitations. However, the conversion and temperature at which the reaction slows down has not been established. For the extrusion model, a pragmatic approach was chosen to consider the high conversion effects. Above a conversion of 98% (Mn > 31000) the kinetics found with the kneader experiment (table 4.1) were applied. The relationship between viscosity, temperature, and molecular weight has been obtained from the extruder experiments by applying equation 5.13 and 5.14 to the experimental data.
k=
P L Q(3 + 1/ n) + CQ 2 R R3
UA n n
( 5.13 )
and
k = Mw 3.4 A 0,flow e RT
( 5.14 )
In equation 5.14, the consistency of a power-law liquid is given as a function of the molecular weight and temperature (19). Equation 5.13 shows the relationship between the consistency of a power-law liquid k and the pressure drop over the die, with a factor C added for entrance losses. Equation 5.13 can be substituted in equation 5.14. For all different experimental conditions, the pressure drop over the die, the molecular weight, and the temperature of the material coming out of the extruder was measured. In addition, the power law index n of the polyurethane was determined experimentally on a capillary rheometer (Gttfert Rheograph 2003) and found to be equal to 0.61. With these data, a least square fit was performed using a substituted version of equations 5.13 and 5.14, in order to obtain the parameters UA, A0,flow, and C.
95
Chapter 5
5.4
Results
In the result section, a validation study of the extrusion model is presented. Moreover, the effect of several extrusion parameters on the polyurethane extrusion will be discussed and compared with the model. The result section is split into several parts. First, a limited validation study is presented on the pressure build-up capacity of the screw elements. As stated in the theoretical section, a correct prediction of the pressure build-up capacity will contribute substantially to a correct prediction of the residence time and therefore of the end conversion. Moreover, the validity of the approach for the flow model can be tested by checking the pressure build-up capacity. Subsequently the extruder model will be compared to an experimental study on polyurethane extrusion. Measurements on the conversion, temperature, and pressure will be compared with the model predictions. The effect of several extrusion parameters will be discussed and special emphasis will be put on the depolymerization reaction. 5.4.1 Validation of the transport elements, a literature check
As stated in paragraph 2.2.3, the pressure build-up capacity of the transport elements is often expressed as (20):
dP 1 = (A N Q ) dL B
or
Q = A N
B dP dL
( 2.2 )
At first sight, a comparison of equation 2.2 with our pressure build up description (equation 5.5) seems troublesome. However, a closer look reveals that the factor B in equation 2.2 is equal to the denominator divided by the k-factor in equation 5.5, while (QS,max-Ql,drag)/N is equal to the A factor. A few experimental studies (21, 22) have been directed to experimental determination of the A and B factors for Newtonian fluids.
96
D (cm) 5 5 3.07 3.07

Table 5.2
Cl (cm) 3.85 3.85 2.62 2.62
Pitch (cm) 5 1.67 2 4.2
R (mm) 0.4 0.4 0.25 0.25
A (cm3) 36 / 40.8* 12.8 / 12.4 4.9 / 4.6

# # *
B (cm4) 0.12 / 0.118* 0.017 / 0.021 0.015 / 0.013

* #
k (-) 27 27 27 27
0.0038 / 0.0045
#
9.5 / 11.5
A and B factors for transport elements (bold face current model, * (21), # (22)).
In table 5.2, the results of these studies for transport elements are compared with our model. The resemblance is good, considering the engineering purposes of the model. Strictly speaking, equation 2.2 is only valid for Newtonian fluids. Model calculations show that for non-Newtonian fluids the deviation can be considerable (23), however, no supporting experimental data exists for twin-screw extruders. In literature, some experimental results are shown for which the pressure is plotted versus extruder length for non-Newtonian fluids. We compared one of these studies (8) with our model; the comparison shows a satisfactory agreement (table 5.3). dP/dL (bar/mm) 100 rpm 1.30 / 1.08 0.78 / 0.65 dP/dL (bar/mm) 200 rpm 1.67 / 1.59 1.26 / 1.42 dP/dL (bar/mm) 300 rpm 1.88 / 2.0 1.60 / 1.9
D=30mm, 28/28 Polystyrene HDPE

Table 5.3
Pressure build up comparison for non-Newtonian fluids (bold face current model, regular face according to (8))
5.4.2
Validation of the transport elements, an experimental check
As an addition to the literature validation, experiments were performed on a Perspex extruder. In this extruder, the filled length together with the pressure along the filled length can be measured. Sugar syrup was used as a Newtonian experimental fluid. A viscosity of 10 Pas was chosen, in order to prevent gravitational effects to be dominant over the viscous forces (24). In addition, a nonNewtonian fluid was tested, which consisted of a 1.5% solution of hydroxy-ethyl cellulose (HEC) in water.
97
Chapter 5
0.4
0.3 (dP/dL)*
0.2
0.1
0 0 20 40 60 80 100 Rotation speed (RPM)
Figure 5.3
The pressure build up capacity as a function of rotation speed. (dP/dL)*=(dP/dL) / (0(n-1)) D = 5 cm, Cl = 3.85 cm, pitch = 5 cm, = 0.02D, cellulose, Sugar syrup. Hydroxy-ethyl
For both liquids, the experimental and model pressure build up capacity of 50/50 transport elements is shown as a function of rotation speed (figure 5.3). The agreement between model and measurement for the HEC is good, while the pressure build-up for the sugar syrup is overestimated at higher rotation speeds. Air bubbles were inevitably present at higher rotation speeds for the sugar syrup experiments, which may cause a deviation of the flow behavior. The pressure build up capacity (expressed as (dP/dL) / app) for the sugar syrup is considerably higher than for the hydroxy-ethyl cellulose solution, which is according to expectations. For a non-Newtonian fluid, the apparent viscosity over the flight decreases considerably due to shear thinning. The pressure-driven leakage over the flight is therefore substantially higher than for a Newtonian fluid. Therefore, the pressure build up divided by the apparent viscosity is much lower for HEC than for sugar syrup. These experiments emphasize the importance of the leakage flow over the flight.
98
300 150
100
200
50
150
0 0 50 100 150 Rotation Speed (RPM)
100 200
Figure 5.4
The pressure build up and temperature for the polypropylene extruder experiments (barrel wall temperature: 190C, 205C, 220C, 240C, closed symbols: pressure build up, open symbols: measured temperature).
To test the pressure build up capacity for a non-Newtonian polymeric material, experiments were performed with polypropylene in an APV-Baker twin-screw extruder. The material was rheologically characterized with a cone and plate rheometer. The results of the extruder experiments are plotted in figure 5.4. The pressure build-up capacity is predicted accurately, except for the 190C experiment. The temperature of the melt seems to be over predicted for all of the experiments. However, the temperature was measured using a wall thermocouple, which tends to underestimate the melt temperature. Since the pressure prediction is correct for these experiments, we can assume the temperature to be predicted correctly. This observation indicates that the energy balance of the model approaches the actual situation. Considering the validation studies above, the flow in the transport elements is described sufficiently well using the model, at least for the types of transport elements and the extruder diameters that were investigated. 5.4.3 Validation of the kneading elements
Concerning the pressure characteristics of the kneading paddles, less information is present in literature. In an experimental study by Todd (21) the pressure build up characteristics for kneading paddles are expressed in the same A and B factors that are used in equation 2.2. In order to compare these factors with our model, equation 5.7 can be rewritten in a similar manner as was done for the transport
99
Temperature (C)
250
dP/dL (bar/m)
Chapter 5
elements. The factor B in equation 2.2 is equal to the denominator in equation 5.7 divided by the k-factor, while (QS,max)/N is equal to the A factor. A comparison between the experiments of Todd (21) and our model is shown in figure 5.5.
120 100 A-factor (cm^3) 80 60 40
0.65
0.15
-0.1 20 0 0 0.1 0.2 0.3 0.4 0.5 0.6 Paddle Width (Width/D) -0.35
Figure 5.5
A and B-factors for a kneading block as a function of paddle width and staggering angle (D = 5 cm, Cl = 3.85 cm, = 0.008D, open symbols B-factor, closed symbols A-factor. Stagger angle: squares 30, triangles 45, circles 60). The lines represent the model simulations, the symbols are the measured values.
In this figure, the paddle width and stagger angle is varied. Considering the simplicity of the modeling approach, the agreement between experiments and model is remarkable. Only for the 60 kneading paddles, the B-factor is largely underestimated. Due to the nature of the modeling approach for the kneading paddles, this deviation is understandable. In the modeling approach, a kneading block is considered as a modified transport element. For larger staggering angles, this approach deviates largely from the actual situation. For non-Newtonian fluids, hardly any experimental data are present for the kneading elements. With the flow model currently used, Michaeli et al. (12) show a reasonable prediction of the pressure characteristics of the kneading paddles for non-Newtonian fluids. A comparison of the dimensionless pressure build up capacity with a 2-D non-Newtonian model (25) shows an acceptable agreement (figure 5.6). Only for a right-handed 60-stagger angle element, the pressure consumption is overestimated.
100
B-factor (cm^4)
0.4

60 40 20 (dP/dz)* (-) 0 -20 -40 -60 -80 -100 -90 -60 -30 0 30 60 90 Staggering Angle ()
Figure 5.6
Dimensionless pressure gradient (dP/dz)* = (P/L)R / (0(2N)n) as a function of staggering angle. The dimensionless throughput Q* = Q/(2R3N) is equal to 0.05. (solid line = model No, dashed line = this chapter).
5.4.4
Polyurethane extrusion
A reactive validation study has been carried out on an APV-Baker MPV-50 extruder. The experimental details of this study are described in a previous section. Obviously, for every experimental setting, a model simulation is generated in order to compare the model prediction with the experiment. Figure 5.7 shows such a simulation for one specific situation. In this figure, the development of the conversion, temperature, pressure, and filling degree along the extruder is shown. Of course, the reaction proceeds mainly in the fully filled sections, due to the longer residence time in these sections. Furthermore, the reaction approaches an equilibrium situation before leaving the extruder; the increase in molecular weight slows down considerably in the last fifteen centimeters upstream of the die. In this area, a dynamic equilibrium is approached for which the forward and the reverse reaction are equally fast. Due to the link between the flow, energy and the reaction, the equilibrium is specific for this particular operation condition and extruder geometry.
101
Chapter 5
250 200 Temperature (C), Filling degree (%) 150
40 Mw (kg/mol), Pressure (bar)
Mn
30
20 100 50 0 0 0.2 0.4 Length (m) 0.6
Filling degree P
10
Figure 5.7
An example of the pressure, Mn, temperature and filling degree along the extruder. The number average molecular weight ( ), pressure () and melt temperature () at 150 RPM, 12.5 kg/hour, Tbarrel = 185 C , [cat] = 30ppm, and ddie = 4 mm.
A model simulation as shown in figure 5.7 has been carried out for all operating conditions that were experimentally tested. In order to compare the model with the experiments, the temperature and conversion are preferably measured at different locations along the screw. In this way, a complete view of the extruder performance can be obtained. Furthermore, with these data, a detailed comparison can be made between the model and the experiments. However, in an extruder, the measurement of the temperature and conversion is notoriously unreliable. The temperature of the melt can only be measured using protruding thermocouples, which affects the flow situation considerably (26). Sampling ports are sometimes used for conversion and temperature measurements but they are vulnerable to clogging; moreover, the sampling procedure can take too long for a reliable measurement. To overcome these problems, an inventive and promising sampling port design has been described by Carneiro et al. (27). Unfortunately, such geometry could not be adapted to our extruder. Therefore, our validation study takes into account the conversion and temperature at the end of the extruder. In addition, the pressure development along the extruder is followed by three pressure sensors. The temperature is measured by inserting a thermocouple in the melt coming out of the extruder. The conversion is measured by a size exclusion chromatography method. Material coming out of the extruder is immediately
102
quenched in liquid nitrogen. Subsequently the molecular weight of the sample is determined as described in chapter 3. In figure 5.7, the outcome of one extruder experiment is compared with the model. In a similar manner, a wider model validation study has been carried out. For the model validation, the effect of catalyst level, barrel temperature, rotation speed and throughput is investigated. Every variable is discussed shortly in the discussion below. 5.4.5 The effect of the catalyst
In figure 5.8, the effect of the catalyst level on the extruder performance is shown for a barrel wall temperature of 185C and a rotation speed of 150 RPM. The model predictions and measurements agree reasonably well on the end pressure, outlet melt temperature, and the molecular weight. Due to viscous dissipation, the temperature of the melt exceeds the wall temperature considerably.
250
Mw (kg/mol), Temperature (C)
25 20 15 10 5 0 0 100 200 300 400 500 Catalyst Concentration (ppm) Pressure (bar)
200 150 100 50 0
Figure 5.8
Weight average molecular weight ( ), pressure () and melt temperature () at the die as a function of catalyst level. (150 RPM, 12.5 kg/hour, Tbarrel = 185C, ddie = 4 mm)
In figure 5.8, a surprising trend is visible; the molecular weight of both the model simulations and the measurements does not show any catalyst dependence. For all catalyst levels, the end conversion and temperature is more or less the same. This observation is not in agreement with earlier kinetic experiments that were performed with this polyurethane (6). In these experiments, a catalyst dependence
103
Chapter 5
was observed. An explanation for the extrusion results may be that the reaction reaches a depolymerization equilibrium before leaving the extruder. In that case, the catalyst concentration has no effect on the end conversion. To test this hypothesis, a more discriminative working zone was tried. The barrel wall temperature was lowered to reduce the effect of the depolymerization reaction. Moreover, we chose a larger die diameter to decrease the residence time in the extruder. The results of these adjustments are shown in figure 5.9.
250
40 30 20 10 0 -10 0 100 200 300 400 Catalyst Concentration (ppm) Pressure (bar)
200 150 100 50 0
Figure 5.9
Weight average molecular weight ( ), pressure () and melt temperature () at the die as a function of catalyst level. (100 RPM, 12.5 kg/hour, Tbarrel = 160C, ddie = 5 mm)
Clearly, a more profound effect of the catalyst concentration is present for both model and experiment. The agreement between the model predictions and experimental results is reasonable, although a somewhat strange and inexplicable deviation exists at 90 ppm. Nevertheless, the upward trend of molecular weight as a function of catalyst concentration is predicted sufficiently by the model. Remarkably, the catalyst level seems to need a threshold value before having an effect, which can be observed both in the model and experimentally. 5.4.6 The effect of the barrel wall temperature
An increase in the barrel wall temperature is a critical test for the extruder model. Changing the barrel wall temperature has a large influence on the reaction in the extruder, the rheological properties of the polymer and on the heat transfer to the
104
melt. At a higher temperature, the reaction velocity will increase and the depolymerization reaction will gain importance. Viscous dissipation will have a lesser influence due to a decrease of the viscosity.
250
20 15 Pressure (bar) 10
200 150
5 100 0 50 0 150 160 170 180 190 200 210 220 Barrel temperature (C) -5 -10
Figure 5.10
Weight average molecular weight ( ), pressure () and melt temperature () at the die as a function of barrel temperature. (solid line: 150 RPM, 12.5 kg/hour, [cat] = 30 ppm, ddie = 4 mm, dashed line (open symbols): 150 RPM, 12.5 kg/hour, [cat] = 30 ppm, ddie = 5 mm)
The effect of the barrel wall temperature was investigated at two different die diameters. The results are shown in figure 5.10. As can been seen in this figure, the reaction hardly develops at 160 C and 180 C for the larger die diameter. At these conditions, the molecular weight at the end of the extruder remains low and viscous dissipation is not present. The latter is clear if we look at the temperature of the melt, which is about the same as the barrel wall temperature. In contrast, at 210 C, the molecular weight is much higher and approaches its equilibrium value. The combination of residence time and temperature is insufficient to reach a high conversion at lower temperatures. A prolonged residence time or a higher catalyst level will give a better result. The first idea is tested experimentally by decreasing the die diameter. For the current extruder configuration, most of the residence time is generated in the last part of the screw. Therefore, the residence time increases considerably by decreasing the die diameter. As can been seen in figure 5.10, the effect of a longer residence time is considerable at lower temperatures. With a smaller die diameter, the end conversion and temperature of the melt are much
105
Chapter 5
higher at 160C. This effect lessens at a higher temperature. The decline of the effect of a prolonged residence time at higher temperatures can be attributed to the depolymerization reaction. The depolymerization reaction limits the conversion at higher temperatures. In that case, an increase in residence time does not lead to a higher conversion so that the final conversion for both die diameters is about the same. If this situation were translated to a commercial situation, it would mean that expensive extruder volume is not utilized efficiently, due to the depolymerization reaction. This paragraph started by stating that an increase in barrel wall temperature is an interesting test for the extruder model. If we now look at figure 5.10, and compare the model and the experiments, the agreement for the small die diameter is reasonably sound. For the larger die diameter, the model prediction does not follow the experiment well at 180 C. However, the trend going from a low to a high temperature is clearly captured. 5.4.7 The effect of the rotation speed
An increase of the rotation speed has both an influence on the residence time and on the viscous dissipation in an extruder. Due to an increase in the rotation speed, the melt temperature will rise and the residence time will shorten; these effects have an opposite influence on the conversion. Which effect prevails depends on the extruder geometry and the polyurethane under consideration.
250
Mw (g/mol), Temperature (C)
30 25 Pressure (bar) 20 15
200 150 100 50 0 50 100 150 200 250 300 Rotation Speed (RPM)
10 5 0
Figure 5.11
Weight average molecular weight ( ), pressure () and melt temperature () at the die as a function of rotation speed. ([cat] = 30 ppm, 12.5 kg/hour, Tbarrel = 185C, ddie = 4 mm)
106
From figures 5.11 and 5.12, it seems that both effects keep each other in equilibrium for this system. In both figures, the rotation speed does not seem to affect the molecular weight to a large extent. As expected, the temperature rises in both situations slightly with increasing rotation speed. However, this effect is not very spectacular and is obviously counterbalanced by a shorter residence time, since the conversion remains approximately the same, independent of the rotation speed. In contrast, the effect of the rotation speed on the end pressure is more obvious. The end pressure decreases with increasing rotation speed. Presumably, the decrease of the end pressure is caused by the combined effect of an increase in temperature and a somewhat lower molecular weight. Both effects lower the melt viscosity and therefore the pressure drop over the die. If we compare the model with the measurements (figures 5.11 and 5.12), the model follows the experiments well for different rotation speeds. A change in rotation speed gives a change in viscous dissipation and therefore a different equilibrium situation in the energy balance. This means that for the current extruder configuration the viscous dissipation is described sufficiently well.
200
150
2.5 Pressure (bar)
100
50
-2.5
0 75 100 125 150 175 200 Rotation Speed (RPM)
-5 225
Figure 5.12
Weight average molecular weight ( ), pressure () and melt temperature () at the die as a function of rotation speed. ([cat] = 30 ppm, 12.5 kg/hour, Tbarrel = 160C, ddie = 5 mm)
5.4.8
Effect of the throughput
In figure 5.13, the effect of a change in the throughput is shown. Both model and measurement show little effect of the throughput on the molecular weight. A higher throughput will give a higher pressure-drop over the die, which will increase the
107
Chapter 5
filled length in the extruder. However, in this case this increase in reactor volume does not lead to a higher conversion, due to a higher throughput to volume ratio in the extruder. Therefore, the residence time remains more or less constant, giving an equal conversion for all three the throughputs. Figure 5.13 shows that the model prediction for the temperature and conversion is accurate; however, the pressure at the end of the extruder is over-estimated.
250
25 20 15 10 5 0 9 10 11 12 13 14 15 16 Throughput (kg/hour) Pressure (bar)
200 150 100 50 0
Figure 5.13
Weight average molecular weight ( ), pressure () and melt temperature () at the die as a function of the throughput. ([cat] = 30 ppm, 150 RPM, Tbarrel = 160C, ddie = 5 mm)
5.4.9
Depolymerization
Obviously, for all extrusion circumstances, the depolymerization reaction has a severe impact on the extruder performance. Model simulations show that in case the reverse reaction is not incorporated, the simulated molecular weight is a factor ten higher than in case the reverse reaction is incorporated. Likewise, the temperature of the melt is much higher without depolymerization. Of course, for the experiments, the depolymerization reaction cannot be suppressed; the effects of the reverse reaction are always present. In case an experimental parameter is adjusted, the change in molecular weight is dampened by the depolymerization reaction. This effect is very clear for the experiment with different catalyst levels at 185C. In this case, an increase of the catalyst level does not have any impact, because the reverse reaction limits the maximum conversion. For these experiments, the molecular weight at the die is within five percent of the equilibrium molecular weight at the outlet temperature.
108
If we focus on this situation, it is somewhat surprising that the conversion reaches almost the same limiting value for every catalyst level. Both model and experiments show this behavior, and for both the model and the experiments, the filled length decreases with a higher catalyst level.
160 140
Pressure (bar)
200
120 Mw (kDa) 100 80 60 40 20 0 0,3 0,4 0,5 0,6 Extruder Length (m )
150
100
50
0 0,3 0,4 0,5 0,6 Extruder Length (m )
Figure 5.14
Build up of the weight average molecular weight and pressure in front of the die.
If we look at the model, this effect may be explained by the way the filling degree in front of the die is calculated. As described in the theoretical section, the pressure drop over the die must be overcome by the filled length in front of the die. Through an iteration procedure, the filled length is extended from zone to zone until the pressure at the end of the die is atmospheric. If we look at a non-equilibrium
second order reaction, the filled length remains more or less the same, independent
of the catalyst level, only the conversion increases with an increased catalyst level1. In figure 5.14, a conceptual (isothermal) drawing of this situation is shown. If a reverse reaction is introduced, the situation changes in figure 5.14. In that case, the
molecular weight is limited by, for example, the dashed line in figure 5.14. This means that for the low catalyst run, the situation does not change, regardless of the reverse reaction. However, for the high catalyst experiment, the molecular weight in
For a simplified isothermal situation, the pressure build-up capacity in the filled zones is a function of
Mw3.4, and the pressure drop over the die is a function of Mw3.4. In case the Mw in the filled zone increases (for example due to a higher catalyst concentration), the pressure build up capacity increases. However, the pressure drop over the die rises proportionally, giving an equal filled length. Therefore, for this situation, the reaction velocity does not influence the filled length in front to the die.
109
Chapter 5
the last part of the extruder will not exceed the equilibrium molecular weight. This limitation will result in a lower molecular weight at the die, giving a lower pressure drop over the die. A lower pressure drop will give a shorter filled length upstream of the die and therefore a shorter residence time in the reactor. The result is that the molecular weight is more or less independent of the catalyst level for this situation. Only the filled length changes with the catalyst concentration. Exactly this behavior is observed for the experiments with different catalyst levels at 185C. The behavior is somewhat different for the experiments at 160C (figure 5.8). For these experiments, the conversion clearly increases with the catalyst level. Presumably, the conversion for these experiments is lower than the equilibrium molecular weight. In that case, more normal catalyst dependence is observable, which is comparable to the lower curve in figure 5.14. A comparison of the measured weight average molecular weight with the equilibrium molecular weight at the outlet temperature endorses this assumption for the experiments at 160C. The depolymerization reaction has several important consequences: Firstly, the reaction is not finished after a reactive extrusion process. For commercial applications, the polyurethane is pelletized at the die. Due to the equilibrium reaction, the remaining pellets still contain a considerable amount of reactive groups and they will continue to react in the bag. This reaction may continue from hours to days. Secondly, the continuous presence in the extruder of reactive isocyanate groups in a pool of urethane bonds may also lead to undesired allophanate formation. Thirdly, the depolymerization reaction has a stabilizing influence on the extrusion process. In case the depolymerization reaction governs the extrusion performance, as in the example above, small variations in the process parameters will hardly affect the end conversion. The reaction near the die is very slow, and therefore a small disturbance at the entrance of the extruder will not affect the output to a great extent. However, the obtained extruder stability has a price. Due to the depolymerization reaction, the extruder volume is not used efficiently. Hardly any reaction takes place in the last part of the extruder. To improve this situation, the throughput can be increased, in combination with a larger die diameter. To evaluate the best combination of throughput and die configuration, the currently developed model can be of use. In addition, the
110
temperature profile of the extruder may be modified. For example, if the temperature of the zones near the die is lowered, the reaction may proceed to a higher conversion, because the depolymerization reaction is slowed down. Nevertheless, for larger extruders, which operate almost adiabatically, such a measure may not be very effective. Also for this situation, an extrusion model can be helpful to evaluate the net effect. 5.4.10 Pressure build up For all experiments, the pressure build up characteristics have been measured in the last part of the extruder through three pressure sensors (figure 5.2). The pressure build up capacity in front of the die can be monitored in this way; furthermore, the filling degree in front of the die can be estimated. For reasons of brevity, these data were left out in the foregoing comparison. However, if we compare these data with the model predictions, a clear trend is visible; the model seems to overestimate the pressure build-up capacity in all cases (figure 5.15).
300
dP/dL (bar/m), Model
200
100
0 0 100 dP/dL (bar/m), Experiment 200
Figure 5.15
Calculated versus measured pressure build up capacity in front of the die.
Several plausible reasons can be formulated for this phenomenon. However, to assess the exact cause of the overestimation of the pressure drop, it is not sufficient to have only data on the pressure drop. Preferably, the conversion along the extruder should also be known. Since the latter could not be determined, the exact reason of the discrepancy cannot be established. Possibly, the k-factor that is used in equation 5.5 is too high, which will result in an over prediction of the pressure build up capacity. On the other hand, the non-reactive validation studies
111
Chapter 5
did not show an inaccurate prediction of the pressure build-up. Another explanation for the discrepancy between model and measurement may lie in the approach for the residence time distribution. The elements in front of the die are considered as plug flow reactors. A plug flow reactor will give a much higher conversion than a well-mixed reactor for the same residence time (14). Therefore, the filled length needed to reach a certain conversion is much shorter for such a reactor. For our specific extruder configuration, it might be expected that the flow behavior in the filled section in front of the die comes closer to a well-mixed reactor than to a plug flow reactor. The extruder is operated at a low throughput in comparison to its maximum throughput capacity. Therefore, in a fully filled section, the pressure flow will almost equal the forward flow, which will lead to considerable back mixing. Moreover, the leakage gap in our extruder is relatively large, which will also contribute to considerable mixing of the material. Both factors contribute to a much higher mixing degree than for a plug flow situation. This will give a lower conversion per centimeter extruder length. Consequently, the experimentally observed pressure drop per length unit is lower than anticipated from the model predictions. A correction factor of 1.3 is applicable in this case. To improve the residence distribution modeling, a residence time distribution model formulated by Pinto and Tadmor (15), based on the helical flow pattern in the channels of a single screw extruder may be of help. This model is also applicable for a self-wiping twin-screw extruder, even though the intermeshing zone will disrupt the flow pattern somewhat. The residence time distribution (RTD) analysis of Pinto and Tadmor (15) shows that the flow in the fully filled conveying elements falls within plug-flow and flow in a pipe. This observation of coincides with the correction factor noticed in the previous paragraph. In case the RTD approach of Pinto and Tadmor is used in the current extrusion model, it must be incorporated in the segmental structure. The plug flow assumption for transport elements used in the current model prevents difficulties that arise when a continuous laminar flow (as is the case for a screw channel) is subdivided in segments (as is he case for the current extrusion model). In that case, going from segment to segment, closed-closed boundary conditions are unsuitable. In a later stage, this problem will be tackled. To do so, the incoming flow can be divided in a group of small batch reactors that flow through the extruder. In a fully filled conveying zone, every batch reactor will follow a specific continuous path, and have a specific residence time, according to the RTD-function. Going from segment to segment, the flow-lines are not disturbed, so that a batch reactor will have an equal residence time in every segment. For every segment, the conversion of all
112
batch reactors can be calculated and averaged, to give the average conversion in a segment. 5.4.11 The effect of the residence time distribution on conversion For a second order reaction, a plug flow reactor is a far more efficient reactor (1.5 to 2 times). In fact, for all nth-order reactions with n > 1, a plug flow reactor is more efficient. If only the residence time distribution was important for reactive extrusion, the screw layout should be designed to approach as plug-flow as well as possible. Generally, transport elements are regarded as the screw elements that come closest to plug-flow. The reason that so many other types of elements are used lies in the fact that in an extruder different processes are combined, which require different type of elements. In process technology, a plug flow reactor is often approached with a cascade of continuous ideally stirred reactors. A similar analogy can be made for extrusion. A study performed by Todd (28) demonstrated that an extruder only filled with kneading elements showed mixing behavior that came closer to a plug flow reactor than an extruder filled with transport elements. However, the forward transport capacity and the energy consumption (and the related temperature rise) with a surplus of kneading paddles may be undesirable. Possibly new types of radial mixing elements (29) may benefit a narrow residence time distribution and improve the efficiency of a reactive extrusion process.
113
Chapter 5
5.5
Conclusions
A comparison of the experimental data with the model predictions demonstrates that the present model describes the polyurethane polymerization reaction in the extruder satisfactory, especially considering the engineering approach chosen for the model. The reverse reaction is also captured adequately in the model. The depolymerization reaction has a profound impact on the extruder performance by limiting the maximum conversion. At the same time, the depolymerization reaction may stabilize the extruder performance due to the considerable decrease of the reaction velocity near the die. Small disturbances at the inlet will be wiped out at the fully filled reaction zone near the die. From an extruder performance point of view, this stagnant zone is an inefficient use of expensive reactor volume. In addition, the consequence of the reverse reaction is that polyurethane that exits the extruder may continue to react in the bag. A prolonged presence of a polymerization-depolymerization equilibrium may be disadvantageous due to the possible occurrence of side reactions (e.g. allophanate formation). The depolymerization reaction is an extra complicating factor for understanding polyurethane extrusion. An extruder model is therefore a helpful tool for optimizing the polyurethane extruder. The current approach for the residence time distribution in the model is coarse. For example, at the relative high pressure to drag flow ratio that is present in the current extruder configuration, the residence time distribution in the filled transport elements comes closer to ideally stirred than to plug flow, while the latter approach is used in the model. However, a correction factor can be used in this case.
114
5.6
A0 A0,flow Awall C Cp [Cat] e EA EA,eq E(t) f ft fk h h i k k0 keq kf kp,t kp,k kr L m MW n n N [NCO] [NCO]0 [NCO]N [NCO]N-1 P/L
List of symbols
Reaction pre-exponential constant Flow pre-exponential constant Surface of the barrel wall Correction for entrance losses at the die Heat capacity Catalyst concentration Flight land width Reaction activation energy Equilibrium reaction activation energy Exit age function Functionality Filling degree of a not fully filled transport element Filling degree of a not fully filled kneading element Height of the screw channel Heat transfer coefficient Number of channels Power law consistency Forward reaction rate constant, catalyst independent Equilibrium reaction rate constant Forward reaction rate constant, catalyst dependent Shape factor for transport elements Shape factor for kneading elements Reverse reaction rate constant Length of the die Catalyst order Weight average molecular weight Reaction order Power law index Rotation speed Concentration isocyanate groups Initial concentration isocyanate groups Isocyanate concentration at the outlet of a reactor Isocyanate concentration at the inlet of a reactor Pressure gradient in the axial direction of the extruder mol/kgs Pasn m2 1/m3n J/kgK mg/g m J/mol J/mol m J/sm2K Pasn mol/kgs mol/kg kg/mols 1/s m g/mol 1/s mol/kg mol/kg mol/kg mol/kg Pa/m
115
Chapter 5
Q QS,max,T QS,max,K QL, drag QL, k R R RNCO t T u [U] UA v0 V w
Throughput Maximum conveying capacity, transport elements Maximum conveying capacity, kneading elements Drag term of leakage flow over the flight Leakage flow between the kneading paddles Gas constant Radius of the die Rate of isocyanate conversion Time Temperature Circumference of the eight-shaped barrel Concentration urethane bonds Flow activation energy Circumferential velocity of the screw Volume ATR reactor Width of the screw channel Viscous dissipation in the channel Viscous dissipation over the flight
kg/s kg/s kg/s kg/s kg/s J/mol K m mol/kgs s K m mol/kg J/mol m/s m3 m J/s J/s
& WC
& WF
Greek symbols R
Conversion (1 - [NCO] / [NCO]0) Clearance between barrel and flight tip Shear rate Viscosity in the channel Viscosity over the flight Intermeshing angle Pitch angle
m 1/s Pas Pas -
&
channel flight
116
5.7
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.
References
K.J. Ganzeveld, and L.P.B.M. Janssen, Polym. Eng. Sci., 32, 457 (1992). A. Bouilloux, C.W Macosko and T. Kotnour, Ind. Eng. Chem. Res.,30, 2431 (1991). P. Cassagnau, T. Nietch and A. Michel, Int. Polym. Process.,14, 144 (1999). M.E.Hyun and S.C. Kim, Polym. Eng. Sci., 28, 743 (1988). G. Lu , D.M. Kalyon, I. Yilgr and E. Yilgr, Polym. Eng. Sci., 43, 1863 (2003). V.W.A. Verhoeven, A.D. Padsalgikar, K.J. Ganzeveld and L.P.B.M. Janssen, J. Appl. Polym. Sci. , accepted for publication H.E. Meijer and P.H.M.Elemans, Polym. Eng. Sci., 28, 275 (1988). H. Potente, J. Ansahl and R. Wittemeier, Int. Polym. Process., 3, 208 (1990). H. Potente, J. Ansahl and B. Klarholz, Int. Polym. Process., 9, 11 (1994). W. Michaeli, A. Grefenstein and U. Berghaus, Polym. Eng. Sci., 35, 1485 (1995). H.Kye and J.L. White, Int. Polym. Process., 11, 129 (1996). W. Michaeli, and A. Grefenstein, Int. Polym. Process., 11, 121 (1996). B. Vergnes, G. Della Valle and L. Delamare, Polym. Eng. Sci., 38, 1781 (1998). K.R. Westerterp, W.P.M. Van Swaaij and A.A.C.M. Beenackers, Chemical Reactor Design and Operation, John Wiley & Sons, New York, Brisbane, Chichester, Toronto (1984).
15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29.
G. Pinto and Z. Tadmor, Polym. Eng. Sci., 10, 279 (1970). H. Potente, Untersuchung der Schweissbarkeit Thermoplastischer Kunststoffe mit Ultraschall, Aachen (1971). D.B. Todd, SPE ANTEC Tech. Papers, 34, 54 (1988). V.W.A. Verhoeven, M.P.Y. Van Vondel, K.J. Ganzeveld, L.P.B.M. Janssen, Polym. Eng. Sci., 44, 1648 (2004). D. W. Van Krevelen, Properties of Polymers, Elsevier, Amsterdam (1990). L.P.B.M. Janssen, Reactive Extrusion Systems, Marcel Dekker Inc., New York, Basel, (2004). D.B. Todd, Int. Polym. Process., 6, 143 (1991). T. Brouwer, D.B. Todd and L.P.B.M. Janssen, Intern. Polym. Process.,17, 26 (2002) Z. Tadmor, G. Gogos, Principles of Polymer Processing, John Wiley & Sons, New York, Brisbane, Chichester, Toronto (1979). R.A. De Graaf, D.J. Woldringh, and L.P.B.M. Janssen, Adv. Polym. Tech., 18, 295 (1999). J. No, Etude des coulements de polymres dans une extrudeuse bivis corotative., Phd-Thesis, Ecole des Mines Paris (1992). M.V. Karwe and S. Godavarti, J. Food Sci., 62, 367 (1997). O.S. Carneiro, J.A. Covas and B. Vergnes, J. Appl. Polym. Sci., 78, 1419 (2000). D.B. Todd, Chem. Eng. Prog., 69, p. 84 (1969). D.B. Todd, Plastic compounding, Hanser, Munich (1998).
117
The effect of premixing on the reactive extrusion of thermoplastic polyurethane
6.1
Introduction
For the reactive extrusion of polyurethane, the extruder performance may be improved by premixing of the monomers. The polyurethane monomers (diisocyanate and the di-alcohols) are immiscible. Therefore, mixing is required to attain a kinetically controlled reaction. Obviously, mixing takes place in the extruder. However, valuable extruder length may be saved by premixing, since in that case the reaction starts earlier. Moreover, the polymer formed may be more homogeneous in composition, since a better defined reaction mass enters the reactor. To validate these assumptions, the effect of premixing is investigated in the current chapter. The approach that is taken to measure the effect of premixing is straightforward. The conversion in the extruder is measured with and without premixing. Moreover, the effect of the degree of premixing is established with batch experiments. By measuring the reaction rate at different degrees of premixing, the effect of premixing is established. Adiabatic temperature rise (ATR) experiments are used to measure the reaction rate. For premixing, both static mixers and dynamic mixers (e.g. a high-speed rotating blade mixer) are used in practice. Due to its superior dispersive mixing action, at first glance a dynamic mixer seems to be a better choice for the current application. However, such a mixer is sensitive to clogging, especially at the inlet points of the monomers. A static mixer is less sensitive to congestion. Moreover, if a static mixer is clogged, it is much easier, faster, and cheaper to replace. Therefore, in the current investigation, a static mixer is used to premix the monomers. After premixing of the monomers, a dispersion of isocyanate drops in a continuous di-alcohol matrix is formed. To have a completely kinetically controlled reaction, the droplet diameter must be small enough so that the diffusion time is shorter than the reaction time. In several publications on reactive injection molding (RIM) of polyurethane, it was found that the reaction rate and the maximum conversion increase with the mixing intensity, up to a certain maximum (kinetically controlled) regime (1-3). Increasing the catalyst level increased the reaction velocity and the necessary mixing to reach a kinetically controlled regime. Most of these experiments were performed with cross-linking systems and impingement mixing. Cross-linking systems are more susceptible to diffusion limitations due to the lower
Chapter 6
mobility of the reactive groups near and beyond the gel point. Impingement mixing gives a much higher shear rate than present in our static mixer. Batch ATR experiments with (pre-) stirring were performed by Fields et al. and Lee et al. (4, 5). Similar results were obtained as with the impingement mixers.
6.2
Mixing
As stated in the introduction, the isocyanate droplets must be small enough to obtain a kinetically controlled reaction. The Fourier number, which compares the reaction time (treaction) with the diffusion time (d2 / ID), applies for this situation:
Fo =
ID t reaction d2
( 6.1 )
The Fourier number should be larger than one for a kinetically controlled reaction. An estimation for the current system shows that the droplets must be smaller than 20 m in that case (ID = 10-11 m2/s, treaction = 60 s, Fo > 2). The minimum droplet diameter that can be obtained in the static mixer can also be estimated through an analysis of a dimensionless number. The droplet diameter is governed by the Capillary number, which is the ratio between the force applied on a droplet ( ) and the interfacial forces ( / d) that keep the droplets together: &
Ca =
& d
( 6.2 )
As long as the capillary number is much larger than unity, the droplets size can still be reduced. At Ca 1 (6), the minimum droplet size is reached. For the static mixer used in this investigation, the minimum droplet diameter that can be reached is
& estimated at about 250 m (Cacr =1, = 0.01 N/m, = 160 s-1, = 0.25 Pas). In
practice, the droplet diameter may be somewhat smaller, since elongational forces are also present in Kenics-type static mixers. If we compare the two droplet diameters, it seems that a kinetically controlled reaction can never be reached in a static mixer alone. For a polymerization reaction, this is particularly troublesome, since high molecular weight barriers will appear at the interface, due to a rapid reaction on the isocyanate-polyol surface (7), preventing completion of the reaction. However, it was established by Machuga et al. (8) and Macosko (1) that the isocyanate polyol interface is unstable. The unstable
120
The effect of premixing on the reactive extrusion of polyurethane
surface results in an interfacial reaction zone that may stretch out to 100 m within a second, depending on the monomers used. In this zone, the di-isocyanates and the di-alcohols are mixed. The calculation of the minimum droplet diameter, based on the capillary number is therefore a worst-case approach. During the passage through the static mixer, the interfacial tension between di-isocyanate and dialcohol fraction is substantially reduced due to the interfacial reaction zone. This will give a much smaller droplet diameter. It is even conceivable that at the end of the mixer a single phase will appear. Nonetheless, the calculation of the final droplet diameter is not straightforward, since reaction, diffusion, and mixing for the polymerization reaction interact. A short analysis of the dynamics of the mixing process in a static mixer may help to estimate the efficiency of extra mixing elements. 6.2.1 Mixing dynamics
The mixing action of a static mixer is both dispersive and distributive. The isocyanate thread that enters the static mixer will be stretched and broken down into small droplets, and these droplets will be distributed evenly over the polyol phase with a narrow droplet size distribution. Far from the equilibrium diameter (Ca >> 1), the diameter of the thread that enters the mixer decreases exponentially, due to affine deformation and the bakers transformations that take place in the static mixer (9). Taking into account this exponential decay, the thread diameter will have reached the critical diameter within three elements (dintial = 2 mm, dcrit = 0.25 mm). Break-up is then not immediate; the relevant time scale for breaking up is (9):
t break up =
( 6.3 )
For the premixing in this investigation, the break-up time is about 0.01 second, which means that the thread is broken into droplets within one mixing element (the residence time for one mixing element = 0.05 seconds). According to the above dimensional analysis, the complete dispersive mixing process should be completed within about four mixing elements. Distributive mixing continues, but since the droplet diameter cannot be reduced further, the reaction velocity will only increase slightly by distributing the droplets evenly. However, the effect of the interfacial reaction zone must also be taken into account. In that case, it is conceivable that the interfacial reaction zone mixes continuously
121
Chapter 6
with the bulk through the static mixing action. In that case, also distributive mixing will improve the reaction considerably. A possible approach to estimate the effect of the interfacial zone is to look at the diffusion velocity of the interfacial zone. This diffusion velocity may be rate limiting. To estimate this effect, the penetration theory for non-stationary mass transport can be used. According to the penetration theory, the penetration depth is proportional to (IDtresidence)1/2. Since tresidence, the residence time, is proportional to the number of mixing elements, the penetration depth of the interfacial zone is proportional to N1/2. The striation thickness, which can be seen as a measure of the distributive mixing action, is inversely proportional to 2N in a Kenics-type static mixer. If we compare these two dependencies, it is clear that the interfacial diffusion velocity is rate limiting; the interfacial zone that is formed will be rapidly mixed in the bulk. Distributive mixing in the static mixer may therefore increase with N1/2.
6.3
Experimental setup
Polyol + diol + catalyst
MDI P1 P2 P3
Figure 6.1
The extruder layout for the premixing experiments.
Figure 6.1 shows a detailed picture of the static mixer and the extruder. The configuration is the same as for the extrusion experiments described in chapter 5. The description of the feeding equipment and the extruder can be found in paragraph 5.3.4. For the ATR experiments, a throughput of 15 kg/hour is maintained for all of the ATR experiments, for the extrusion experiments the throughput is 12.5 kg/hour. At the inlet of a static mixer, the isocyanate is inserted at the middle of the polyol stream through a small nozzle (d = 2 mm). The static mixer (Mixpac MC 06-32) consists of 32 mixing elements (Delement = 6.35 mm, L/D = 1). The number of mixing
122
elements can easily be reduced. The static mixer outlet is placed near the extruder inlet. If an ATR experiment is performed, the ATR reactor is placed under the outlet of the static mixer.
6.4
Materials
System 2, as described in chapter 4, was used for the premixing experiments. The system consists of a liquid form of diphenylmethane diisocyanate (an eutectic mixture of 2,4-MDI and 4,4-MDI), methyl-propane-diol and a polyester polyol (Mw = 2200, functionality =2). The treatment of the chemicals before usage is described in in chapter 2. The percentage of hard segments (isocyanate + methyl-propane-diol) was 24%. The reaction was catalyzed using bismuth octoate.
6.5
Adiabatic temperature rise analysis
The adiabatic reactor consisted of a disposable cup (diameter = 4cm) surrounded by a layer of urethane foam for insulation. The reactor could be closed with a lid. The reactor was equipped with a thin Copper Constantine thermocouple sticking in the middle of the reaction mass. At the start of an experiment, the reactor is filled with a continuous flow of reaction mass. The temperature of the entering material is measured. The time to fill the reactor at 15 kg/hour was 25 seconds, giving a final content of about 100 gram. For the lowest catalyst level, the filling time is not relevant. However, at the highest catalyst levels, the filling time of the reactor is relatively long compared to the reaction time (treaction = 80 seconds at 100 ppm; 120 seconds at 75 ppm). This makes the analysis of the reaction at the highest catalyst level not as straightforward as with the regular ATR experiments. On the other hand, the measurement is not intended for the determination of kinetic constants but for an evaluation of the effect of premixing. The first consequence of the relatively long filling time is that temperature gradients over the reaction mass may occur (Damkohler IV > 1): Material that enters the reactor initially will have reached a higher temperature than the last part of the material entering the reactor, since it has already reacted for 25 seconds. Combined with the fact that in the ATR reactor, the macromixing of the colder and warmer material is not perfect, no uniform reactor temperature may be reached. To estimate this effect, every experiment is performed in triplicate. In case large temperature gradients occur, the triplicate measurements should show large deviations. A second consequence of the large filling time is the presence of a distribution in reaction times in the reaction mass. The first part of the reaction mass has reacted 123
Chapter 6
for 25 seconds at the time the last part enters the reactor. The isocyanate concentration follows a similar distribution in the reactor, due to the poor micro mixing properties of the reacting polymer system. A micro segregated reaction mass with a variable isocyanate concentration is the result. This may bias the fitting procedure, especially at high catalyst concentrations. In case an isocyanate distribution is present, for a given ATR-curve, a standard ATR analysis will underestimate the activation energy and overestimate the reaction velocity (when a second order kinetic equation applies). A coarse approach was taken to investigate this effect (appendix 6.1). It was found that the activation energy and the reaction rate constant were slightly affected by micro segregation (< 3 % at the highest catalyst level). Considering the size of the effect, this systematic error was ignored. In order to derive kinetic data from the ATR experiments, a simplified heat balance (equation 4.1) and rate equation (equation 2.12) were solved simultaneously.
Cp
dT = R NCO HR h* (T Troom ) dt d[NCO] = A 0 e R T [NCO]n dt

E A
with
h* =
h A V
( 4.1 )
RNCO =
( 2.12 )
For the heat balance, quasi-adiabatic conditions were assumed, since the reactor was not completely adiabatic for the time under investigation. Depending on the reaction time, up to 4 % of the total reaction heat generated during the reaction was lost to the surroundings. The heat transfer coefficient h* was obtained by fitting the cooling curves of several experiments, using equation 4.1. We took the density and the specific heat to be constant over the whole measurement range. Although both the specific heat and the density are somewhat dependent on the temperature, the temperature effects of both constants counteract, so that the net effect is negligible (< 5%). HR was taken from the experiment that gave the largest temperature rise. A non-linear regression method (error controlled Runge-Kutta) was used to solve the differential equations. With a least square routine, the difference between the model and the measurement was minimized. The calculations were performed with the software program Scientist. For the model fit, only the first part of the ATR curve was utilized, up to a conversion of about 75 %. As described in chapter 4, at a higher conversion the reaction slows down due to phase separation of hard and soft segments. In order to determine the maximum temperature rise, the complete ATR curve was fitted.
124
6.6
Results
In figure 6.2, typical adiabatic temperature rise curves are shown. As the figure shows, a substantial part of the reaction occurs during filling of the reactor. After 25 seconds, the filling of the reactor has ended and the three temperature curves coincide, indicating a good reproducibility. The model fit is performed on the area between the two dotted lines. The low limit is of course related to the filling time of the reactor, while the upper limit is derived from an Arrhenius representation of the ATR curves (chapter 4). The upper limit represents the point at which the Arrhenius plot (not shown) deviates from a straight line. The reaction slows down at this point, due to the separation of the hard and soft segments (chapter 4), or more generally due to the crossing of the glass temperature of the material.
400 Temperature (K) 380 360 340 320 0 50 100 Time (s)
Figure 6.2 Typical adiabatic temperature rise curves, a triplicate experiment (16 mixing elements, [Cat] = 75 ppm).
150
200
6.6.1
The effect of premixing on the reaction velocity
The effect of catalyst level and number of mixers on the apparent reaction rate constant is shown in figure 6.3. The error bars in this figure represent the standard deviation based on three experiments. For the experiments most susceptible to diffusion limitations, the high catalyst experiments, the reaction rate increases with the mixing intensity, clearly indicating a diffusion controlled reaction at low mixing intensity. The reaction seems to approach the kinetically controlled regime at 32 elements. These results are contradictory to the analysis of the dimensionless numbers that was discussed in the theoretical section. In this analysis, no effect of mixing is expected with more than four mixing elements. Moreover, the effect of
125
Chapter 6
mixing should follow an exponential pattern, instead of the slow increase of the reaction velocity with the number of mixing elements as found in figure 6.3.
0.05 0.04 k (kg/mol s) 0.03 0.02 0.01 0 0 8 16 N (-) 24 32
Figure 6.3
The effect of the number of mixing elements (N) on the reaction rate constant for different catalyst levels ( = uncat, = 50 ppm, = 75 ppm, = 100 ppm).
Evidently, the analysis of dimensionless numbers in the theoretical section is based on best estimates of the physical constants, and the analysis may therefore give a deviation from the actual situation. However, the general conclusion of the analysis of dimensionless numbers that the minimum droplet diameter is in the range of 50 to 200 m diameter is supported by experimental results of Kolodziej et al. (3). Therefore, the difference in the analysis of dimensionless numbers and the experimental results must have a different origin than a faulty estimate. In the theoretical section, a possible scenario is sketched based on an effect that is typical for polyurethanes: the presence of an interfacial reaction zone. On the boundary layer of isocyanate and di-alcohol, rapid diffusion of the di-alcohols in the isocyanate droplets occurs, forming an interfacial reaction zone. This instable boundary layer will be continuously mixed in the bulk due to the distributive mixing effect of the static mixer. For this scenario, the reaction velocity increases with N1/2, in case the reaction is diffusion limited.
126
0.04 k (kg/mol s)
0.02
0 1 2 3 4 5 6 (N)^0.5 (-)
Figure 6.4 The effect of the square root of the number of mixing elements (N) on the reaction rate constant for different catalyst levels ( = uncat, = 50 ppm, 75 ppm, = 100 ppm). =
To test this scenario, figure 6.4 shows a plot of the reaction velocity versus N1/2. If the theoretical approach is valid, the reaction rate of the reaction with the highest catalyst level should rise linearly with N1/2. This seems not to be the case. The simplified theoretical approach therefore does not describe the measurements. The exact mixing mechanism must relate to a different mechanism.
32
16
0.04 k (kg/mol s)
0.02
0 0 0.1 1 / N (-) 0.2 0.3
Figure 6.5
The effect of the inverse of the number of mixing elements (N) on the reaction rate constant for different catalyst levels ( = uncat, = 50 ppm, = 75 ppm,
= 100 ppm).
127
Chapter 6
A least square fit of the reaction velocity of the highest catalyst level versus Nx shows that for x -1 a linear plot is obtained. The obvious effect of the decreasing efficiency of an extra mixing element is shown in figure 6.5. Looking at this figure, it seems that for the highest catalyst level, the reaction does not reach a kinetically controlled regime within 32 elements. For the 75 ppm experiments, a similar rise of the reaction rate with the number of mixing elements is visible, but for this catalyst level the kinetically controlled regime is reached at 16 mixing elements. The lowest catalyst level and uncatalyzed experiments do not seem to be bothered by any diffusion limitations. These observations agree with the idea that the faster the reaction, the more prone a reaction is to diffusion limitations. Moreover, a higher degree of mixing is necessary to reach a kinetically controlled regime for a faster reaction. 6.6.2 The effect of premixing on the adiabatic temperature rise
The adiabatic temperature rise is related to the initial mixing efficiency. In case the initial (micro-) mixing is insufficient, high molecular weight diffusion barriers may appear, which prevent the reaction to come to a full completion, giving a lower TAdiabatic. Both dispersive and distributive mixing are important in this case.
Adiabatic temperature rise (C)
75
70
65
60 0 10 20 N (-) 30 40
Figure 6.6
The effect of the number of mixing elements (N) on the adiabatic temperature rise for different catalyst levels ( = uncat, = 50 ppm, ppm). = 75 ppm, = 100
In figure 6.6, the effect of the number of mixing elements and the catalyst level on the adiabatic temperature rise is shown. The difference between the uncatalyzed and catalyzed series is striking. The adiabatic temperature rise for the uncatalyzed
128
reaction is unaffected by the number of mixing elements, while the catalyzed reactions show a steady increase of Tadiabatic with the number of mixing elements. The effect for the uncatalyzed experiments is probably related to the time after which an experiment was stopped. As stated before, the reaction slows down considerably at high conversions. As expected, the difference in Tad as a function of mixing degree is generated in this last part. Unfortunately, the temperature was monitored for too short a time to capture the complete tail of the uncatalyzed reactions. For the catalyzed reactions, it seems that the catalyst concentration has no significant effect on Tadiabatic, at least not within the experimental error of the current experiments. This observation is in agreement with a study on the interfacial activity, which does not see an effect of catalyst or the catalyst level on the initial formation velocity of an interfacial layer (8). As soon as catalyst is present, the micro mixing efficiency (expressed as Tadiabatic) increases, independent of the catalyst concentration. This catalyst independence implies that both the droplet diameter and the interfacial zone at the end of the static mixer are of equal size for all catalyst levels, or more general, the degree of mixing is similar for all catalyst levels and only dependent on the number of mixing elements.
Adiabatic temperature rise (C)
75
32
16
70
65
60 0 0.1 1 / N (-) 0.2 0.3
Figure 6.7
The effect of the inverse of the number of mixing elements (N) on the adiabatic temperature rise for different catalyst levels ( = uncat, = 50 ppm, ppm, = 100 ppm). = 75
In analogy with the reaction velocity, the adiabatic temperature rise is plotted versus the inverse of the number of mixing elements (figure 6.7). Although the significance level is not optimal, it seems that even with 32 elements no ideally
129
Chapter 6
micro mixed situation exists. This is in agreement with the high catalyst level experiments, which show that even with 32 elements the reaction is diffusion limited for the fastest reaction. However, for the low catalyst levels the reaction seems kinetically controlled for all mixing levels according to figure 6.5. This does not seem to agree with the adiabatic temperature rise data (figure 6.7), which show the opposite effect. Since the differences in adiabatic temperature rise (Tadiabatic) for the different mixing levels are made at the end of the reaction, it may be so that imperfect mixing only bothers the low catalyst experiments for the last part of the reaction. The kinetic data is obtained at lower conversions, before the reaction slows down. This difference may be the origin of the inconsistency. 6.6.3 The effect of premixing on the extruder performance
Rotation speed (RPM) 100 100 150 200

Table 6.1
Residence time (s) 114 110 117 193
Mw premixed 76 72.3 68.4 70.8
Mw nonpremixed 70.1 69.6 67.1 67.7
PDI premixed 2.1 2.2 2.1 2.0
PDI nonpremixed 2.2 2.3 2.1 2.1
The process parameters and results for the extrusion experiments (PDI = polydispersity index).
Extrusion experiments were carried with and without premixing (table 6.1). The mixing intensity and residence time in the extruder was varied to evaluate the effect of mixing in the extruder versus the effect of premixing. Model simulations (chapter 5) were performed to calculate the residence time. In table 6.1, the extrusion parameters are listed including the residence time model simulations.
130
100 75 Mw (kg/mol)
150
RPM
200
2.6 Polydispersity (-)
2.4 65
2.2
55 0.5 1.5 2.5 RPM x Time (-)

Figure 6.8
2 3.5 4.5
Weight average molecular weight with () and without () premixing and the polydispersity with () and without premixing ( ) as a function of the total strain in the extruder.
In figure 6.8, the molecular weight and the polydispersity of the formed polyurethane (with and without premixing) are plotted versus the product of the rotation speed and the residence time. The latter is a crude measure of the total strain the reaction mass has encountered during processing. If, due to more or less constant time and viscosity, the total strain coincides with the maximum shear stress, as is the case in figure 6.8, the x-axis can been seen as a combined mixing scale for both distributive as dispersive mixing. The catalyst level was varied in the above experiments, to obtain a more or less similar conversion for all experiments. As can been seen in figure 6.8, premixing has a slight effect on the final conversion. The weight average molecular weight is 5 tot 10 % higher for the premixed experiments. Since for condensation polymerization, the molecular weight is linearly dependent on the reaction time, 5 to 10% of reaction time (or extruder length) is saved by premixing the reaction mass. Figure 6.8 shows that the increase in conversion due to premixing is independent of the mixing degree in the extruder. The premixing benefit may be caused by a faster initial reaction velocity in the extruder due to premixing of the reaction mass. Then again, a higher degree of initial mixing may also lead to a higher final conversion, as for example the batch experiments show. In the latter case, high molecular weight diffusion barriers that occur due to the poor initial mixing are prevented by premixing. These barriers may be difficult to break, even in the flow field of the extruder, due to their high
131
Chapter 6
viscosity and small size. The reduction in reaction velocity at the end of the reaction is prevented and the reaction will carry on to a higher conversion. For all experiments the polydispersity decreases when the reaction mass is premixed. This observation reflects that a better defined reaction mass gives a polymer with a narrower molecular weight distribution. The mixing applied in the extruder does not seem to change that, even at a high mixing degree the difference in polydispersity remains between the premixed and non-premixed experiments. On the other hand, figure 6.8 shows also that the polydispersity decreases with the amount of mixing the extruder. This indicates that mixing in the extruder must have some beneficial effect on the homogeneity of the reaction mass. 6.6.4 A comparison of the results with the results of chapter 4
In chapter 4, different types of kinetic experiments were performed with the chemical system currently under investigation. The system was found to be very sensitive to mixing. Several findings led to this conclusion: For the ATR experiments, the activation energy decreased at higher catalyst levels and the reaction velocity seemed to reach a maximum value at higher catalyst levels. For the kneader and high temperature experiments, the experiments without mixing proceeded at a lower reaction velocity. If we look at the current experiments, the mixing sensitivity is also clearly present. However, it is interesting to compare the batch ATR results as presented in chapter 4 with the current continuous ATR results. For the batch experiments the reaction mass is premixed with a turbine stirrer, while a static mixer is used for the continuous experiments. In fact when comparing the two methods, superior dispersive mixing (turbine stirrer) is compared with superior distributive mixing. In table 6.2, the reaction rate constant of the uncatalyzed experiments and three catalyst levels are compared. For all catalyst levels, the experiments with the static premixing are faster than the dynamically premixed experiments. Moreover, with increasing catalyst level, the dynamically premixed experiments become relatively slower. Apparently, for this polyurethane, premixing with a static mixer is more efficient than with a turbine stirrer. Hence, the spatial distribution of the monomers must be imperfect for the dynamically premixed experiments, causing the observed diffusion limitations. The
132
better dispersive mixing action of a turbine stirrer does not seem to be decisive in this case. uncat static mixer turbine stirrer ratio
Table 6.2
50 ppm 0.009 0.007 1.3
75 ppm 0.024 0.012 2.0
100 ppm 0.041 0.016 2.6
0.003 0.002 1.4
The reaction rate constant (kg / mols) at 100C for the batch and continuous ATR experiments (using 32 mixing elements).
In light of the extrusion experiments performed with the current system, a fit of the dynamically premixed experiments (with 32 elements) is appropriate as an input for the extruder model. According to the analysis of the results of this chapter, only with 100 ppm catalyst, the reaction is not completely kinetically controlled. However, the effect of diffusion limitation is moderate, so that the kinetic constants will hardly be affected.
6.7
Conclusions
For the polyurethane under investigation, premixing has a small beneficial effect on the conversion at the end of the extruder. Moreover, premixing results in a narrower molecular weight distribution, giving improved material properties. Although the effect of premixing for this system is not substantial, it will increase with faster reacting monomers. For faster reactions, diffusion barriers at the start of the reaction are more dominant and premixing will help to level them. The investment costs for implementing (static) premixing are low, but whether to implement premixing depends on the clogging sensitivity of the static mixer, and the effect of a jam on the functioning of the feeding system. The benefit of premixing for the current polyurethane is not so great that the benefits always outweigh the added operational risks. Looking at the effect of the degree of premixing, it is clear that for the currently investigated polyurethane, the reaction velocity and the final conversion are affected by the degree of premixing. At low catalyst levels, the reaction velocity is independent on the degree of premixing, while at higher catalyst level premixing has a positive effect on the (apparent) reaction velocity.
133
Chapter 6
6.8
A A0 Cl Cp d ID EA h h* HR L n N [NCO] [NCO]0 MW t R RNCO T V
List of symbols
Surface area of ATR reactor Reaction pre-exponential constant Centerline distance Heat capacity Diameter Diffusion coefficient Reaction activation energy Heat transfer coefficient Overall heat transfer coefficient Heat of reaction Length Reaction order Number of mixing elements Concentration isocyanate groups Initial concentration isocyanate groups Weight average molecular weight Time Gas constant Rate of isocyanate conversion Temperature Volume ATR reactor m2 mol/kg s m J/kgK m m2/s J/mol J/m2sK J/kgsK J/mol m mol/kg mol/kg g/mol s J/mol K mol/kgs K m3
Greek symbols Clearance between barrel and flight tip Shear rate Viscosity Density Surface tension m 1/s Pas kg/m3 N/m
&

134
6.9
1. 2. 3. 4. 5. 6. 7. 8. 9. 10.
References
C. W. Macosko, RIM - Fundamentals of Reaction Injection Molding, Hanser, Munich, 1989. L.J. Lee, J.M. Ottino, W.E. Ranz, C.W. Macosko, Polym. Eng. Sci., 20, 868 (1980). Kolodziej, C.W. Macosko, and W.E. Ranz, Polym. Eng. Sci., 22, 388 (1982). S.D. Fields, and J.M. Ottino, AIChE J., 33, 157 (1987). Y.M. Lee, and L.J. Lee, Intern. Polym. Process., 1, 144 (1987). H.P. Grace, Chem. Eng. Commun., 14, 225 (1982). S.D. Fields, and J.M. Ottino, AIChE J., 33, 959 (1987). S.C. Machuga, H.L. Midje, J.S. Peanasky, C.W. Macosko, and W.E. Ranz, AIChE J., 34, 1057 (1988). J.M.H. Janssen, Ph. D. Thesis, Eindhoven University of technology (1993). V.W.A. Verhoeven, M. van Vondel, K.J. Ganzeveld, and L.P.B.M. Janssen, Polym. Eng. Sci., 44, 1648 (2004).
135
Chapter 6
6.10
Appendix 1
The average isocyanate concentration in the reactor just after filling is calculated by subtracting the measured inlet temperature from the measured temperature just after filling.
[NCO] =
(Tt=25 Tt=0 ) Cp
HR
( 6.4 )
Subsequently, a concentration distribution as should be present just after filling is drawn up, assuming a zero-order isothermal reaction (figure 6.9).
[NCO] t=0
[NCO]
0
Figure 6.9
Time
reactor.
25
Theoretical isocyanate concentration distribution during filling of the ATR
The ATR fit starts at this point. The ATR reactor is modeled as five different batch reactors, each with the same temperature but each with a different initial isocyanate concentration according to figure 6.9. The average concentration of the five batch reactors together is equal to the average isocyanate concentration in the complete reactor just after filling. For the model fit, the mass balance for every reactor is simultaneously solved together with the overall heat balance, as is done for the regular ATR fits. In table 6.2, the result for an ATR experiment at the highest catalyst level is shown.
136
EA (kJ/mol) Regular fit With isocyanate distribution 33.9 34.8 (3.2 %)
A0 (kg/mols) 1.89 2.48
k at 80C (kg/mols) 0.0206 0.0201 (2.5 %)
Table 6.2 Fitted reaction rate constants with and without an isocyanate concentration distribution ([Cat] = 100 ppm).
137
Conclusions
In case an engineer is asked to optimize or to develop a new reactive extrusion process, for example for polyurethane manufacturing, she or he can choose several ways to do the job. Most straightforward is to start doing small-scale extrusion experiments on a lab or pilot scale extruder. The experimental approach can be preceded by batch experiments. Through batch experiments, the feasibility of the process can be tested. The engineer can get an estimate if the process will work on the extruder, and if the desired product properties can be obtained. For the extruder experiments that follow, the targeted objective can be reached through a process in which iterative experiments are combined with previously gained knowledge. When phenomena are encountered that contradict expectations, or when the target is found to be difficult to reach, this engineer will look for background knowledge to understand what is happening or to get a clue where to go. For a reactive extrusion process, this will sometimes lead to inexplicable phenomena, because the response of the system is not always linear. Several processes interact: the reaction and the related change in viscosity, the flow in the extruder, and the heat transfer and generation. Moreover, when scaling up the extruder, surprises might appear. An approach that can be useful in addition to the experimental approach is to make use of an experimental design. With an experimental design, the number of experiments can be optimized and a statistical model can be built. Since the extruder is not a complete black box, the engineer can combine the obtained results with his or her knowledge to reach the targeted objective. Nonetheless, the interplay between experimental design and previous knowledge is a challenging undertaking. A different approach the engineer can take is to make use of a model that is based on the processes that take place in the extruder. The model can be used as a complete substitution of the experimental work or as an addition to experimental work. Complex phenomena may become better understandable with the use of such a model; moreover, extrapolation to other process conditions can be done with more confidence with the use of a model. Also before doing experimental work, model simulations can help to narrow the experimental window. The benefits of a model are clearly present; however, the applicability of such a model is hampered by several factors. First, for every polyurethane under investigation, rheo-kinetic data must be obtained. For every change in hard segment percentage, catalyst type,
Chapter 7
change in chain extender etcetera, new data must be generated, which is a cumbersome task. The benefit of a model concerning time and cost may not be that large due to this necessary experimental work. Second, no exact prediction can be obtained with the model. As discussed in paragraph 5.2.1, the status of extruder modeling is such that the results still are in between indicative and predictive. Last, the objective of the study must be expressible in a model parameter. The current thesis is aimed at reactive extrusion of thermoplastic polyurethane. To assist the above engineer, both background knowledge as well as better understanding of the process through a reactive model is presented. Three subjects were covered specifically: 1. 2. 3. The effect of the measurement method on the kinetic results, especially tailored for reactive extrusion. The modeling of the extruder and the effect of the depolymerization reaction on the extruder performance. The effect of premixing on the extruder performance.
In chapter three, a new method was presented for measuring the polyurethane kinetics, based on measurement kneader measurements. The method was validated; it was found that quantitative kinetic and rheological data could be obtained using this method. The method has advantages over other kinetic measurement methods since the reactants are mixed during an experiment, and the temperature is close to extrusion circumstances; mimicking real processing circumstances. Therefore, for applications where the reaction takes place under mixing conditions, as for reactive extrusion, the kinetic parameters obtained will be more accurate. Besides, the effect of mixing on the polymerization reaction can be investigated with this method. In chapter 4 the necessity of such a method was investigated for two different polyurethanes. It was found that for a typical thermoplastic polyurethane (system 1), the measurement method did not seem to matter. Also relatively low conversion adiabatic temperature rise experiments (no mixing) as high temperature experiments (with mixing) give the same result. Apparently, the reaction is uniform and kinetically controlled over a large range of temperatures and conversions. Adiabatic temperature rise experiments seem therefore the preferred choice for characterizing polyurethane polymerization. These types of experiments are easy to perform and to analyze. However, when analyzing the experiments, care must be
140
Conclusions
taken. It was found that diffusion limitations due to phase separation of soft and hard segments appear already at 70% conversion (depending on the temperature). If this is not noticed, the obtained kinetic constants will underestimate the real reaction velocity (4.4.1). For extrusion purposes, the kinetics of the reverse reaction must also be known. Unfortunately, a second kinetic measurement is necessary to establish the reverse kinetics. In paragraph 4.4.2, a method based on high temperature batch experiments is worked out. Reproducible results on the depolymerization reaction were obtained. Depending on the reaction velocity, the kinetic constants of the forward reaction can also be established, but the method is more cumbersome than performing adiabatic temperature rise experiments. The second polyurethane that was investigated showed a more complicated kinetic behavior. The ATR experiments showed a decrease of activation energy and a maximum in reaction velocity at higher catalyst levels, indicating a diffusion limited reaction. For this polyurethane, the formation of a broad interfacial reaction zone on the polyol-isocyanate surface may be hindered, due to the spatial conformation of the oligomers. This interfacial reaction zone is indispensable for a kinetically controlled reaction. As a confirmation on this diffusion limitation, it was found through high conversions that mixing did have an influence on the reaction velocity. For reactive extrusion modeling of this particular polyurethane, two kinetic measurement methods seem necessary, adiabatic temperature rise experiments for low conversions and kneader experiments for high conversions. The latter polyurethane was used for extrusion experiments. A reactive extrusion model was built to evaluate the results (chapter 5). The model was based on onedimensional equations. This type of model was chosen to have optimal flexibility and calculation speed. The model was validated with non-reactive data from own experiments and literature, and showed a satisfactory agreement. A comparison of the reactive experimental data with the model predictions demonstrates that the model describes the polyurethane polymerization reaction in the extruder satisfactory, especially considering the engineering approach chosen for the model. For model verification the rotation speed, the barrel wall temperature, the catalyst concentration, and the throughput were varied. The difference between the predicted molecular weight and the measured molecular weight was on average no more than 20 percent. More importantly, the response of the extrusion process towards changing process parameters was captured adequately. Therefore, the extrusion model seems a good tool for evaluating an extrusion process. However, an evaluation of the model with different polyurethanes and different extruder diameters would help to further validate the model.
141
Chapter 7
The reverse reaction was an important factor for the performed experiments. The presence of this reaction smoothens some effects, for example the effect of changing the catalyst concentration. Moreover, as explained in paragraph 5.4.9, the reverse reaction has several effects on the reactive extrusion process: The reverse reaction causes further reaction in the extruded material after extrusion The reverse reaction can give extra allophanate formation The reverse reaction stabilizes the extrusion process
For the extrusion model, a simple approach was chosen for the residence time distribution. An analysis showed (paragraph 5.4.10) that in some cases, such as for the transport elements, the approach oversimplifies the real situation. For the transport elements, the plug-flow assumption underestimates the spread in residence time. As is shown in paragraph 5.4.10, in case a section shows more ideally stirred behavior, the efficiency of that section drops (the efficiency expressed as conversion per meter extruder length). Therefore, for reactive polyurethane extrusion, screw elements that give a plug-flow behavior should prevail. In general, transport elements are considered closest to plug-flow. However, since a group of kneading paddles resembles a cascade of stirred reactors, the spread in residence time may be smaller than that of transport elements. Although many investigations have been conducted towards residence time distribution in extruders, the coupling with an extrusion model based on flow equations is still not completely developed. A start has been made by Poulesquen et al. (reference 20, chapter 2). Further development in this area will help to improve the choice for the optimal screw layout. The efficiency of the extruder may be improved by premixing of the monomers. Through premixing, valuable extruder length may be saved since the reaction gets a head-start. Moreover, the product may be more homogeneous in composition. The effect of premixing was investigated in chapter 6. It was concluded that premixing has a small positive contribution to the conversion at the end of the extruder, and that the final product has a narrower molecular weight distribution. Although in this investigation the effect of premixing may be dampened by the reverse reaction, the effect of premixing is rather small. For every situation, a careful consideration must be made if the risk of premixing (clogged equipment) is worth the improved extrusion performance and product quality.
142
8
8.1 Summary
Appendix
Thermoplastic polyurethane (TPU) is a widely used polymer that is used in automotive products, electronics, glazing, footwear and for industrial machinery. By changing the composition, the rubber-like material can be optimized for a specific application. The range of material properties is therefore large; some thermoplastic polyurethanes may be very elastic while other urethanes are very stiff. Thermoplastic polyurethane is often produced in an extruder. A typical process consists of feeding the monomers to the extruder, formation of the polymer in the extruder, and cutting the final polymer in small pellets at the exit of the extruder. The pellets can be processed to its final application in a later stage. Reactive extrusion of polyurethane is a relative expensive and a not very well understood process. The lack of understanding is caused by the complicated flow patterns in the extruder, and because the processes that occur in the extruder are difficult to measure. In industry, reactive extrusion is therefore often approached in a pragmatic way. However, improved knowledge could lead to cost benefits. Both the daily practice and the development or scale-up of new processes would benefit, leading to a more efficient extrusion process. An engineer working on the extrusion of polyurethane has several choices to reach his or hear goal. Firstly, available knowledge can be combined with batch or extrusion experiments. As an addition, the engineer can use experimental design techniques to optimize the data collection process. However, the knowledge buildup in this way is difficult to extrapolate, due to the non-linear character of the processes in an extruder. Moreover, local minima or maxima can be missed in this way. Another option for the engineer is to build an extrusion model. A reactive extrusion model is better able to describe the non-linear effects. Moreover, a model can be used to scan a whole range of parameters within a short time period, among which the effect of the screw profile (which is a cumbersome task when doing experiments). A disadvantage of a model is that for each polyurethane a new set of kinetic parameters must be obtained. In this thesis, a model was built and the predictive power was evaluated through a validation study. For such a model, the input parameters are equally important as the model itself. Especially, accurate kinetic parameters are indispensable. Therefore, different methods for obtaining the kinetic parameters were evaluated in chapter 3 and 4 of the thesis. The effect of mixing and reaction temperature on the
Chapter 8
kinetic data was established. Because of the immiscibility of the monomers, mixing was supposed to have an impact on the apparent reaction velocity and on the final conversion. A new measurement method was developed (chapter 3) to establish the effect of mixing on the kinetics at higher temperatures (150 200C); in a measurement kneader the torque and the molecular weight development were followed for a polyurethane polymerization reaction. In this way, extrusion conditions are mimicked and relevant kinetics data can be obtained. The kinetics were determined for a system consisting of a polyester polyol, methyl-propane-diol and a 50/50 mixture of 2,4- and 4,4-diphenylmethane diisocyanate (MDI). The reaction proceeded according to a second order reaction for which the activation energy was found to be equal to 61.3 kJ/mol, and the pre-exponential factor was equal to 2.18e6 mol/kg K. For the temperature range under investigation the flow activation energy was equal to 42.7 kJ/mol, which is comparable to that of a linear polymer. This indicates that the hard and soft segments are completely mixed at the temperatures investigated. The results obtained with the kneader were compared with other kinetic measurement methods in chapter 4. Two different polyurethane systems were investigated. Both polyurethane systems had the same large chain polyol (a polyester polyol, Mw = 2200) and the same percentage hard segments (24%). However, system 1 contained 4,4-diphenylmethane diisocyanate (4,4-MDI) and butane diol, while system 2 contained a 50/50 mixture of 2,4- and 4,4diphenylmethane diisocyanate and methyl-propane-diol. System 2 was used in the kneader experiments as described above. The monomers of system 2 are better compatible, which should lead to less diffusion limitations during the reaction. Three different kinetic methods were compared: adiabatic temperature rise, measurement kneader and isothermal high temperature measurements. For the less miscible polyurethane system (system 1), the reaction conditions did not seem to depend on the measurement temperature and the mixing conditions, implicating for all reaction conditions a kinetically controlled reaction. The reaction was second order in isocyanate concentration, 0.5-th order in catalyst concentration and had an activation energy of 52 kJ/mol. For the second (miscible) system (system 2), each of the three measurement methods showed a different behavior. Only at a low catalyst concentration, the adiabatic temperature rise experiments demonstrated a catalyst dependence, at a higher catalyst levels and for the other two measurement methods no catalyst dependency was present. Furthermore, the adiabatic temperature rise experiments
144
Appendix
showed a much higher reaction velocity in comparison to the other two methods. For this system, the rapid diffusion of the monomers and the oligomers through the interface between the species is probably hindered due to the presence of bulky oligomer molecules. The result is a diffusion-limited reaction at low conversions and an inhomogeneous distribution of species at higher conversions. The presumed better miscibility for system 2 was therefore not demonstrated. Contrary to system 1, the isocyanate and chain extender of system 2 are hardly used for commercial applications. For more regularly occurring systems based on 4,4-MDI, it seems probable (based on the results of system 1) that any kinetic measurement method is appropriate to establish the kinetics for reactive extrusion modeling. In chapter 5, the extrusion model is described. For the model, a one-dimensional approach was chosen. For this approach, the extruder was divided in zones with a length of 0.25D. For every zone the temperature, conversion, viscous dissipation, average shear rate, and the viscosity were calculated. The actual channel geometry and the effect of the intermeshing zone were incorporated in the flow equations. The average shear rate and the viscous dissipation were calculated on basis of a two-dimensional analysis. The leakage over the flight was taken into account for calculating the viscous dissipation, the pressure build up and the conveying capacity. In the model, a distinction was made between transport elements and kneading elements. For each element type, different flow equations were used. In addition, the residence time distribution was considered differently for each type of element. The reaction mass was considered to be micro-segregated for both type of elements, but the kneading zones were assumed to be ideally mixed, while the transport zones were considered as plug-flow reactors. The extrusion model was validated with literature data and through non-reactive extrusion experiments. Both validations showed a satisfactory agreement. Moreover, a reactive validation study was carried out with polyurethane system 2. The kinetics obtained in chapter 4 and 6 were used as input parameters. In the reactive validation study the catalyst level, throughput, rotation speed and the barrel wall temperature was varied. A comparison of the experiments with the model showed that the model predicted the polyurethane extrusion well. In chapter 4, it was established that the depolymerization reaction is of importance at temperatures higher than 150C. Since the extrusion conditions are generally above this temperature, the depolymerization reaction in polyurethanes extrusion is a relevant but often neglected phenomenon. In chapter 5 it was found that the extruder operation is strongly affected by the depolymerization reaction: the depolymerization reaction limits the maximal obtainable conversion, increases the
145
Chapter 8
amount of allophanate bindings that are formed, stabilizes the extruder operation, and causes undesired post-extrusion curing of the polyurethane. In the last regular chapter of this thesis (chapter 6), the effect of premixing the monomers (before extrusion) was investigated. To do so, a static mixer was placed at the feed port of the extruder. Extrusion trials were performed with and without premixing; the effect of premixing was evaluated through the final molecular weight and polydispersity of the product. For the polyurethane under investigation (system 2), it was found that premixing had a small beneficiary effect on the conversion at the end of the extruder. Moreover, premixing resulted in a narrower molecular weight distribution, giving improved material properties. However, the benefit of premixing is for the current polyurethane not of such extend that the benefit always outweighs the added operational risks (clogging of equipment). Still, although the effect of premixing for this system was not substantial, the implementation of pre-mixing is cheap and straightforward. Moreover, the beneficiary effect of premixing will increase with faster reacting monomers. The effect of the degree of premixing on the reaction is also described in chapter 6. The effect of the number of static mixer elements on adiabatic temperature rise experiments was established. A difference was found in the reaction velocity and adiabatic temperature rise as a function of the catalyst level and number of mixer elements. At low catalyst levels, the reaction velocity was independent of the degree of premixing, while at higher catalyst level an increased degree of premixing had a positive effect on the (apparent) reaction velocity. The adiabatic temperature rise showed a different behavior and was independent of the catalyst level but dependent on the number of mixing elements. This difference in behavior was attributed to the part of the reaction that is represented by the two parameters. The reaction velocity is related to the initial and middle part of the reaction, while the adiabatic temperature rise is related to the end of the reaction. The objective of this thesis was to increase the understanding of the reactive extrusion of thermoplastic polyurethane. Overall, several issues were identified: Using a relative simple extrusion model, the reactive extrusion process can be described. This model can be used to further investigate and optimize the reactive extrusion of thermoplastic polyurethane. Premixing has a small beneficiary effect on the efficiency of the extrusion process and the quality of the product formed.
146
Appendix
The depolymerization reaction has a large influence on the extrusion process For a regular polyurethane, the temperature and the mixing conditions do not affect the kinetic parameters over a wide temperature range.
147
Appendix
8.2
Samenvatting
Thermoplastisch polyurethaan is een veelgebruikt polymeer dat onder andere wordt toegepast in schoenzolen, sportuitrustingen, aandrijfriemen, oormerken voor vee en auto-onderdelen. Door de samenstelling van de grondstoffen te wijzigen kan het rubberachtige materiaal voor elke specifieke toepassing geoptimaliseerd worden; zo kan zijn dat het ene thermoplastische polyurethaan zeer elastisch is terwijl een ander thermoplastisch polyurethaan zeer stijf is. Thermoplastisch polyurethaan wordt vaak in een extruder geproduceerd, waarbij in de extruder de monomeren reageren tot het polymeer. Op het moment dat het polymeer uit de extruder komt wordt het versneden in korrels, die later verder verwerkt kunnen worden in de uiteindelijke toepassing. Reactieve extrusie van thermoplastisch polyurethaan is een relatief duur en onbegrepen proces. De onvolledige begripsvorming wordt veroorzaakt door de aanwezigheid van een ingewikkeld stromingsprofiel in de extruder, gecombineerd met het feit dat de processen die in de extruder plaatsvinden nauwelijks meetbaar zijn. In de industrie wordt reactieve extrusie om die reden vaak pragmatisch benaderd. Meer begrip van de processen die in de extruder plaatsvinden zou echter een kostenvoordeel kunnen opleveren. Zowel de dagelijkse praktijk als het ontwikkelen of opschalen van nieuwe processen kan dan met meer efficiency bedreven worden. Een ingenieur die aan extrusie van polyurethaan gaat werken heeft diverse keuzes om tot zijn of haar doel te komen. Aanwezige kennis kan gecombineerd worden met extra batch experimenten of extrusie experimenten. Hierbij kan eventueel gebruik gemaakt worden van experimental design technieken. De resultaten hiervan zijn echter lastig te extrapoleren, doordat niet-lineaire effecten onvermijdelijk aanwezig zijn in een extruder. Een ander optie is om gebruik te maken van een extrudermodel. Een reactief extrusiemodel kan de niet-lineaire effecten beter ondervangen. Bovendien kan met een model binnen een kort tijdsbestek een hele range aan parameters getest worden, waaronder het effect van het schroefprofiel (wat bij een experiment zeer tijdrovend is). Een nadeel van een model is dat voor elk verschillend type polyurethaan de kinetische parameters moeten worden vastgesteld. In dit proefschrift is een reactief extrusiemodel ontwikkeld en de betrouwbaarheid van het model is getest met een validatiestudie. Het model geeft meer inzicht in het proces en kan bovendien gebruikt worden om het proces te verbeteren. Voor een dergelijk reactief model zijn de ingebrachte parameters minstens zo belangrijk als
149
Chapter 8
het model zelf. In het bijzonder spelen de kinetische parameters een belangrijke rol. Daarom zijn in hoofdstuk 3 en 4 van het proefschrift verschillende kinetische methodes bekeken en met elkaar vergeleken. Het effect van de reactietemperatuur en menging is hierbij vastgesteld. Omdat de monomeren niet mengbaar zijn, was de verwachting dat menging een invloed op de reactie zou kunnen hebben. In hoofdstuk 3 is een nieuwe methode ontwikkeld om bij hogere temperaturen (150 200C) het effect van mengen op de kinetiek vast te stellen. Bij deze methode wordt het moment en het molecuulgewicht in een batch kneder in de tijd gevolgd. Op deze manier worden extrusie-omstandigheden nagebootst. Het bleek dat met deze methode relevante kinetische parameters gemeten konden worden. De kinetiek is bepaald voor een systeem dat bestond uit een polyester polyol, methylpropaan-diol en een 50/50 mengsel van 2,4- en 4,4-difenylmethaan diisocyanaat (MDI). De reactie verliep volgens een tweede orde reactievergelijking. De activeringsenergie was daarbij 61.3 kJ/mol en de pre-exponentiele factor was 2.18e6 mol/kg K. Binnen het onderzochte temperatuursgebied was de activeringsenergie 42.7 kJ/mol, wat overeen komt met een lineair polymeer. Dit betekent waarschijnlijk dat alle harde segmenten en zachte volledig gemengd zijn; de fasescheiding tussen deze twee onderdelen is opgeheven. In hoofdstuk 4 is een vergelijking gemaakt tussen de batch kneder metingen en andere kinetische meetmethodes. Twee verschillende polyurethanen zijn daarbij onderzocht. Het eerste systeem was hetzelfde polyurethaan waarmee de batch kneder experimenten zijn uitgevoerd. Omdat dit systeem een minder gangbaar polyurethaan is, is dit systeem aangeduid als systeem 2. Het andere systeem, systeem 1, had hetzelfde polyester polyol als systeem 2 en hetzelfde harde segmenten percentage (24%). Daarnaast bevatte systeem 1 butaan-diol en 4,4difenylmethaan diisocyanaat (MDI). De monomeren van systeem 1 zijn minder goed mengbaar. Dit zou tot meer diffusie effecten moeten leiden gedurende de reactie. Drie verschillende kinetische methodes zijn met elkaar vergeleken in hoofdstuk 4: adiabatische temperatuur stijging experimenten, batch kneder metingen en isothermische hoge temperatuur metingen. Het bleek dat voor systeem 1 de reactieomstandigheden (temperatuur en menging) geen effect hadden op de gevonden kinetische parameters. Dit betekent dat over een groot temperatuur- en mengbereik de reactie kinetisch gelimiteerd was. De reactie verliep volgens een tweede orde reactie met een activeringsenergie van 52 kJ/mol. De reactiesnelheid was recht evenredig met de wortel uit de katalysatorconcentratie. Voor systeem 2 gaven elk van de drie meetmethodes een ander resultaat. Alleen bij lage katalysatorconcentraties vertoonden de adiabatische temperatuurstijging
150
Appendix
experimenten een katalysatorafhankelijkheid. Bij hogere katalysatorconcentraties en voor de andere twee meetmethodes had de katalysatorconcentratie geen effect. Daarnaast was de reactiesnelheid bij de adiabatische temperatuurstijging experimenten veel hoger dan bij de andere experimenten. Deze waarnemingen duiden op diffusie limitaties. Meer specifiek was voor dit systeem de snelle oppervlaktediffusie een star die polyurethaanreacties waarschijnlijk kenmerkt door (het de polyol-isocynaat van oppervlak is instabiel waardoor beide fases veel sneller met elkaar mengen dan bij oppervlak) gehinderd aanwezigheid diffusiebeperkende oligomeermoleculen. Dit resulteerde in een diffusielimitatie bij lagere conversies en een inhomogeen verdeelde reactiemassa bij hoge conversies. De veronderstelde betere mengbaarheid van systeem 2 is daarom niet aangetoond, eerder het tegenovergestelde. In tegenstelling tot systeem 1 wordt het isocyanaat (en de ketenverlenger) van systeem 2 zelden gebruikt voor commercile toepassingen. Daarom lijkt het waarschijnlijk dat voor polyurethanen gebaseerd op 4,4-MDI elke meetmethode geschikt is om de kinetiek te bepalen voor reactieve extrusie. In hoofdstuk 5 staat het ontwikkelde extrusiemodel beschreven. Bij dit model is voor een eendimensionale benadering gekozen. De extruder werd daarbij verdeeld in zones met een lengte van 0.25D. Voor elke zone werd de temperatuur, conversie, viskeuze dissipatie, gemiddelde afschuifsnelheid en viscositeit berekend als functie van de vorige zone en de omstandigheden in de betreffende zone. De feitelijke kanaalgeometrie en het effect van de intermeshing zone zijn in de stromingvergelijkingen gebruikt. De gemiddelde afschuifsnelheid en de viskeuze dissipatie zijn berekend door middel van een tweedimensionale analyse. In de vergelijkingen voor de viskeuze dissipatie en voor de drukopbouw in de extruder is rekening gehouden met lekstroming over de flank van de schroef. In het model is onderscheid is gemaakt tussen transportelementen en kneedelementen. Voor elk elementtype zijn verschillende stromingsvergelijkingen gebruikt. Ook is voor elk elementtype een andere benadering gekozen voor de verblijftijdspreiding. Voor beide elementtypes is de reactiemassa als micro-segregated beschouwd. De kneedelementen zijn daarbij beschouwd als een serie van ideaal gemengde reactoren, terwijl de transportelementen als propstromingreactoren zijn gemodelleerd. Het extrusiemodel is gevalideerd met behulp van literatuurgegevens en nietreactieve extrusie-experimenten; het model en de gevonden meetwaarden kwamen daarbij goed overeen. Daarnaast is een reactieve validatie studie uitgevoerd met systeem 2. De kinetiek die in hoofdstuk 3, 4, en 6 gemeten was is daarbij gebruikt.
151
Chapter 8
In de reactieve validatiestudie is de katalysatorconcentratie, de doorzet, het toerental en de wandtemperatuur gevarieerd. Een vergelijking van de experimenten met het model toonde aan dat het model een goed voorspellend vermogen had; de meetresultaten en de modelsimulaties verschilden niet meer dan 20 %. In hoofdstuk 4 is vastgesteld dat de depolymerisatiereactie merkbaar aanwezig is boven een temperatuur van 150C. Doordat de reactieve extrusie van polyurethaan over het algemeen boven deze temperatuur plaatsvindt, is het effect van de depolymerisatie reactie op polyurethaan extrusie ook onderzocht in hoofdstuk 5. Daarbij is gevonden dat de depolymerisatiereactie een groot effect heeft op de polyurethaanextrusie: de depolymerisatie reactie limiteert de maximaal haalbare conversie, veroorzaakt mogelijk ongewenste allophanaat vorming, stabiliseert het extrusieproces en veroorzaakt ongewenst doorreageren na extrusie. In het laatste reguliere hoofdstuk van dit proefschrift, hoofdstuk 6, is het effect van voormenging van de monomeren onderzocht. Met voormenging wordt bedoeld dat de monomeren gemengd worden voordat ze aan de extruder worden gedoseerd. Voormenging zou kostbare extruder lengte kunnen besparen en een beter gedefinieerd polymeer kunnen geven. Om dit te kunnen onderzoeken is een statische menger bij de invoerpoort van de extruder geplaatst. Vervolgens zijn extrusie experimenten mt en zonder voormenging uitgevoerd (met systeem 2). Het effect van voormenging werd daarbij gevalueerd door het molecuulgewicht en polydispersiteit van het gereageerde product bij de spuitkop van de extruder te meten. Voor het onderzochte polyurethaan is gevonden dat voormenging een klein positief effect had op de conversie aan het einde van de extruder. Bovendien gaf voormenging een lagere polydispersiteit, wat een positief effect heeft op de materiaaleigenschappen. Daarbij moet gezegd worden dat effect klein was en mogelijk niet opweegt tegen het voornaamste risico dat met voormenging gepaard gaat: verstoppen van de apparatuur. De afweging om wel of niet voor te mengen zal daarom uit een kosten en risico analyse moeten volgen. Het effect van voormenging is waarschijnlijk groter bij sneller reagerende monomeren. Dit betekent dat het effect van voormenging voor andere polyurethanen groter kan zijn. Het effect van de mate van voormenging is ook onderzocht in hoofdstuk 6. Hierbij is het aantal statische mengelementen gevarieerd en is de reactie gevolgd met adiabatische temperatuurstijgingexperimenten. Daarbij is een verschil gevonden in de reactiesnelheid en de adiabatische temperatuurstijging als functie van de katalysator concentratie en het aantal mengelementen. Bij lage katalysatorconcentraties was de reactiesnelheid onafhankelijk van het aantal mengelementen, maar bij hogere katalysator concentraties had extra voormenging
152
Appendix
een
positief
effect
op
de
schijnbare
reactiesnelheid.
De
adiabatische
temperatuurstijging liet een ander beeld zien; deze stijging was onafhankelijk van de katalysator concentratie maar afhankelijk van het aantal mengelementen. Het verschil in gedrag is toegeschreven aan het deel van de reactie die vertegenwoordigd wordt door de twee parameters. De reactiesnelheid is gerelateerd aan het begin en het middelgedeelte van de reactie, terwijl de adiabatische temperatuurstijging gerelateerd is aan het einde van de reactie. Het doel van dit proefschrift was om meer begrip over reactieve extrusie van thermoplastisch polyurethaan te verkrijgen. Een aantal zaken is daarbij aan het licht gekomen: Voormenging heeft een gering positief effect op de eindconversie en polydispersiteit van het gevormde polyurethaan De depolymerisatie reactie heeft een groot effect op de extrusieprestatie Voor een gangbaar polyurethaan is de kinetische meetmethode van ondergeschikt belang voor de bepaling van relevante kinetische parameter Daarnaast is een extrusiemodel ontwikkeld en gevalideerd dat zowel in de dagelijkse praktijk als bij het ontwikkelen of opschalen van nieuwe processen gebruikt kan worden.
153
Appendix
8.3
List of publications
High temperature solution polymerization of butyl acrylate/methyl methacrylate: reactivity ratio estimation. M. Hakim, V. Verhoeven, N.T. Mcmanus, M.A. Dub, A. Penlidis, J. Appl. Polym. Sci., 77, 602 (2000). Rheo-kinetic Measurement of Thermoplastic Polyurethane Polymerization in a Measurement Kneader. V.W.A. Verhoeven, M. van Vondel, K.J. Ganzeveld and L.P.B.M. Janssen, Polym. Eng. Sci., 44, 1648 (2004). Method for producing a shaped cheese product, the cheese product obtained and an apparatus for continuously performing the method. V. Verhoeven, T. Jongsma and R. Fransen, EP 01520481A1 A Kinetic Investigation of Polyurethane Polymerization for Reactive Extrusion Purposes., V.W.A. Verhoeven, A.D. Padsalgikar, K.J. Ganzeveld and L.P.B.M. Janssen, J. Appl. Polym. Sci., accepted for publication The Reactive Extrusion of Thermoplastic Polyurethane and the Effect of the Depolymerization Reaction. V.W.A. Verhoeven, A.D. Padsalgikar, K.J. Ganzeveld and L.P.B.M. Janssen, Int. Polym. Process., accepted for publication The effect of premixing on the reactive extrusion of thermoplastic polyurethane. V.W.A. Verhoeven, A.D. Padsalgikar, K.J. Ganzeveld and L.P.B.M. Janssen, Polym. Eng. Sci., send in for publication
155
Appendix
8.4
Dankwoord
Dit proefschrift is met de bijdrage van vele mensen tot stand gekomen. Ik ben dankbaar dat ik de afgelopen jaren zoveel inhoudelijke, praktische en morele steun heb mogen ontvangen. Voor de inhoudelijke bijdrage wil ik ten eerste Lon bedanken, die mij de gelegenheid heeft geboden om dit onderzoek te starten en me de ruimte heeft gegeven om er zelf richting aan te geven. Daarbij heeft je positieve houding mij gesteund in het afronden van het proefschrift. Ineke, je bent iets later bij het project betrokken geraakt, maar je me daarna volle kracht geholpen met je kritische blik en je heldere visie op de grote lijn. Zelfs nadat je een nieuwe baan begonnen bent heb je nog heel wat versies doorgelezen. Ajay, thank you for all the support and guidance you gave during the project and for helping to find my way through the Huntsman organization. In addition, I received valuable help form Huntsman Polyurethanes, especially from Wim Vignero, Koen De Roovere, John Hobdell Valentina Gizzi, John Kendrick and Willem-Jan Leenslag. De leden van de leescommissie, Ton Broekhuis, Arend-Jan Schouten en Stephen Picken wil ik bedanken voor de tijd en aandacht die ze aan dit proefschrift hebben gegeven. Mijn afstudeerders die gekozen hebben om het TPU-avontuur in te stappen wil ik ook graag bedanken. Bas, Maarten Marille en Bernard, jullie werk is een waardevolle bijdrage gebleken voor dit proefschrift. Alle mensen die me vanuit de vakgroep hebben ondersteund wil ik graag ook noemen. Laurens, Marcel, Erwin en Anne, bedankt voor alle ondersteuning in de breedste zin om polyurethaan te kunnen extruderen en onderzoeken. Mijn dank gaat uit naar Jan-Henk voor de SEC en Gert voor de rheologische metingen. Marya, bedankt voor alle assistentie op afstand. Tijdens mijn periode aan de Nijenborgh 4 heb ik frustratie, euforie en vriendschap mogen delen met mijn mede-promovendi. Ten eerste met mijn kamergenoot en eeuwige collega Mario, met wie ik ondanks mijn introductie toch nog on speaking terms ben geraakt. Daarnaast met Jasper, Francesca, Cedric, Vincent, Marga en Mook, met wie ik gedurende 4 jaar veel vriendschap en broodjes van de maand heb gedeeld. Ook de overige vakgroepsgenoten, Erik Heeres, Sameer, Poppy, Iris, Jose, Gerald, Anette en Anant hebben mijn jaren aan de Nijenborgh 4 kleur gegeven. Mijn collegas bij Friesland Foods Cheeeezzze wil ik bedanken voor de zachte dwang en interesse in mijn proefschrift. Erik en Martijn, bedankt dat jullie mij op de dag zelf willen bijstaan als paranimfen.
157
Chapter 8
De afgelopen jaren ben ik in staat geweest mijn spaarzame vrije tijd gedoseerd in te zetten voor andere activiteiten. Alle Spicemen en fietsvrienden van Tandje Hoger, bedankt voor de vele uurtjes ontspanning, inspanning en gezelligheid die heel belangrijk zijn geweest om dit proefschrift te kunnen afronden. Daarnaast wil ik speciaal Martijn, Sander, Rogier, Eduard, Gerko en Hinko danken voor hun vriendschap. Het slotwoord van deze onderneming is voor mijn ouder en voor Ellen. Pap, mam, zonder jullie was ik nooit zover gekomen. Als laatste lieve Ellen, bedankt dat je er altijd voor me was. .
158

The Reactive Extrusion of Thermoplastic Polyurethane

Caricato da

Informazioni sul documento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

The Reactive Extrusion of Thermoplastic Polyurethane

Caricato da

Copyright:

Formati disponibili

The Reactive Extrusion of Thermoplastic Polyurethane

The Reactive Extrusion of Thermoplastic Polyurethane

Vincent Wilhelmus Andreas Verhoeven geboren op 24 mei 1973 te Waalre

ISBN 90-367-2520-8 ISBN 90-367-2521-6 (Electronic version)

LIST OF SYMBOLS LIST OF REFERENCES

133 135 134

136 139 143 143 155 149

Scope of the thesis

An Introduction to extrusion and polyurethanes

The closely intermeshing corotating twin-screw extruder

An introduction to extrusion and polyurethanes

Two closely intermeshing corotating screws.

Parallel channel representation of a corotating closely intermeshing twin-screw extruder (6).

The helical flow pattern in a single channel.

An introduction to extrusion and polyurethanes

heat of reaction conductive transport of heat

viscous dissipation conduction of heat

conductive heat transfer convective heat transfer

v 3 P w+e R sin Q L = u 0 sin cos R + 2 L tan 12 flight e u = 2 (D + 2 R ) ( )

An introduction to extrusion and polyurethanes

so that equation 2.3 becomes:

A kneading block for a corotating intermeshing twin-screw extruder.

An introduction to extrusion and polyurethanes

A typical screw profile for a corotating intermeshing twin-screw extruder (1).

An introduction to extrusion and polyurethanes

The basic unit in a urethane block-copolymer (24).

Flexible and rigid segments in a polyurethane elastomer.

Structure of 4,4-MDI (left) and 2,4-MDI (right).

An introduction to extrusion and polyurethanes

An introduction to extrusion and polyurethanes

H (1) Urethane Formation: N C O + OH

(2) Isocyanurate Formation:

H (3) Allophanate Formation: N C O O + N C O

(4) Uretidione Formation:

H (5) Urea Formation: H2 O + N C O N C O OH

The most commonly occurring isocyanate reactions.

Reaction 2: Isocyanurate formation

An introduction to extrusion and polyurethanes

Lewis base catalysis for urethane formation.

Lewis acid catalysis for urethane formation.

and Rest-. The resulting

An introduction to extrusion and polyurethanes

[Cat ] [OH] [NCO] 1+ K C2 [OH]

d[NCO] = RNCO, Uncat + RNCO, Cat = k f [NCO]n dt

[NCO] = [NCO]0 1+ k f (n 1 [NCO]0 )

MN = Mrep (1+ [NCO] 0 k f (T,[cat]) t ) M W = Mrep (1+ 2 [NCO]0 k( T,[cat]) t)

An introduction to extrusion and polyurethanes

d[NCO] = k f [NCO]2 k r [U] dt

A 0,eq e and [U] = [NCO]0 [NCO]

[U] eq MN MN Mrep k = K= f = = A 0,eq e 2 2 kr [NCO] eq Mrep [NCO] 0

An introduction to extrusion and polyurethanes

[U] v v0 w Greek symbols R

mol/kg m/s m/s m

m 1/s Pas W/mK m2/s kg/m3 -

Barrel wall Catalyzed Die Equilibrium Leakage Uncatalyzed

Rheo-kinetic measurements in a measurement kneader

Rheo-kinetic measurements in a measurement kneader

kneading paddles plug

The Brabender measurement kneader.

Rheo-kinetic measurements in a measurement kneader

Theory of measurement of the kinetics

A simplification of the flow geometry in the measurement kneader.

ND = app = app M & H

Torque = ( Area Shear Stress ) Lever Arm = (DW ) (D / 2)